diff options
Diffstat (limited to 'drivers/gpu/drm/i915')
277 files changed, 28288 insertions, 15659 deletions
diff --git a/drivers/gpu/drm/i915/Kconfig b/drivers/gpu/drm/i915/Kconfig index 0d21402945ab..ba9595960bbe 100644 --- a/drivers/gpu/drm/i915/Kconfig +++ b/drivers/gpu/drm/i915/Kconfig @@ -76,7 +76,7 @@ config DRM_I915_CAPTURE_ERROR This option enables capturing the GPU state when a hang is detected. This information is vital for triaging hangs and assists in debugging. Please report any hang to - https://bugs.freedesktop.org/enter_bug.cgi?product=DRI + https://bugs.freedesktop.org/enter_bug.cgi?product=DRI for triaging. If in doubt, say "Y". @@ -105,11 +105,11 @@ config DRM_I915_USERPTR If in doubt, say "Y". config DRM_I915_GVT - bool "Enable Intel GVT-g graphics virtualization host support" - depends on DRM_I915 - depends on 64BIT - default n - help + bool "Enable Intel GVT-g graphics virtualization host support" + depends on DRM_I915 + depends on 64BIT + default n + help Choose this option if you want to enable Intel GVT-g graphics virtualization technology host support with integrated graphics. With GVT-g, it's possible to have one integrated graphics @@ -148,3 +148,9 @@ menu "drm/i915 Profile Guided Optimisation" depends on DRM_I915 source "drivers/gpu/drm/i915/Kconfig.profile" endmenu + +menu "drm/i915 Unstable Evolution" + visible if EXPERT && STAGING && BROKEN + depends on DRM_I915 + source "drivers/gpu/drm/i915/Kconfig.unstable" +endmenu diff --git a/drivers/gpu/drm/i915/Kconfig.debug b/drivers/gpu/drm/i915/Kconfig.debug index 00786a142ff0..438040ff0179 100644 --- a/drivers/gpu/drm/i915/Kconfig.debug +++ b/drivers/gpu/drm/i915/Kconfig.debug @@ -1,34 +1,32 @@ # SPDX-License-Identifier: GPL-2.0-only config DRM_I915_WERROR - bool "Force GCC to throw an error instead of a warning when compiling" - # As this may inadvertently break the build, only allow the user - # to shoot oneself in the foot iff they aim really hard - depends on EXPERT - # We use the dependency on !COMPILE_TEST to not be enabled in - # allmodconfig or allyesconfig configurations - depends on !COMPILE_TEST - select HEADER_TEST - default n - help - Add -Werror to the build flags for (and only for) i915.ko. - Do not enable this unless you are writing code for the i915.ko module. - - Recommended for driver developers only. - - If in doubt, say "N". + bool "Force GCC to throw an error instead of a warning when compiling" + # As this may inadvertently break the build, only allow the user + # to shoot oneself in the foot iff they aim really hard + depends on EXPERT + # We use the dependency on !COMPILE_TEST to not be enabled in + # allmodconfig or allyesconfig configurations + depends on !COMPILE_TEST + default n + help + Add -Werror to the build flags for (and only for) i915.ko. + Do not enable this unless you are writing code for the i915.ko module. + + Recommended for driver developers only. + + If in doubt, say "N". config DRM_I915_DEBUG - bool "Enable additional driver debugging" - depends on DRM_I915 - select DEBUG_FS - select PREEMPT_COUNT - select REFCOUNT_FULL - select I2C_CHARDEV - select STACKDEPOT - select DRM_DP_AUX_CHARDEV - select X86_MSR # used by igt/pm_rpm - select DRM_VGEM # used by igt/prime_vgem (dmabuf interop checks) - select DRM_DEBUG_MM if DRM=y + bool "Enable additional driver debugging" + depends on DRM_I915 + select DEBUG_FS + select PREEMPT_COUNT + select I2C_CHARDEV + select STACKDEPOT + select DRM_DP_AUX_CHARDEV + select X86_MSR # used by igt/pm_rpm + select DRM_VGEM # used by igt/prime_vgem (dmabuf interop checks) + select DRM_DEBUG_MM if DRM=y select DRM_DEBUG_SELFTEST select DMABUF_SELFTESTS select SW_SYNC # signaling validation framework (igt/syncobj*) @@ -36,14 +34,14 @@ config DRM_I915_DEBUG select DRM_I915_SELFTEST select DRM_I915_DEBUG_RUNTIME_PM select DRM_I915_DEBUG_MMIO - default n - help - Choose this option to turn on extra driver debugging that may affect - performance but will catch some internal issues. + default n + help + Choose this option to turn on extra driver debugging that may affect + performance but will catch some internal issues. - Recommended for driver developers only. + Recommended for driver developers only. - If in doubt, say "N". + If in doubt, say "N". config DRM_I915_DEBUG_MMIO bool "Always insert extra checks around mmio access by default" @@ -59,16 +57,16 @@ config DRM_I915_DEBUG_MMIO If in doubt, say "N". config DRM_I915_DEBUG_GEM - bool "Insert extra checks into the GEM internals" - default n - depends on DRM_I915_WERROR - help - Enable extra sanity checks (including BUGs) along the GEM driver - paths that may slow the system down and if hit hang the machine. + bool "Insert extra checks into the GEM internals" + default n + depends on DRM_I915_WERROR + help + Enable extra sanity checks (including BUGs) along the GEM driver + paths that may slow the system down and if hit hang the machine. - Recommended for driver developers only. + Recommended for driver developers only. - If in doubt, say "N". + If in doubt, say "N". config DRM_I915_ERRLOG_GEM bool "Insert extra logging (very verbose) for common GEM errors" @@ -111,41 +109,41 @@ config DRM_I915_TRACE_GTT If in doubt, say "N". config DRM_I915_SW_FENCE_DEBUG_OBJECTS - bool "Enable additional driver debugging for fence objects" - depends on DRM_I915 - select DEBUG_OBJECTS - default n - help - Choose this option to turn on extra driver debugging that may affect - performance but will catch some internal issues. + bool "Enable additional driver debugging for fence objects" + depends on DRM_I915 + select DEBUG_OBJECTS + default n + help + Choose this option to turn on extra driver debugging that may affect + performance but will catch some internal issues. - Recommended for driver developers only. + Recommended for driver developers only. - If in doubt, say "N". + If in doubt, say "N". config DRM_I915_SW_FENCE_CHECK_DAG - bool "Enable additional driver debugging for detecting dependency cycles" - depends on DRM_I915 - default n - help - Choose this option to turn on extra driver debugging that may affect - performance but will catch some internal issues. + bool "Enable additional driver debugging for detecting dependency cycles" + depends on DRM_I915 + default n + help + Choose this option to turn on extra driver debugging that may affect + performance but will catch some internal issues. - Recommended for driver developers only. + Recommended for driver developers only. - If in doubt, say "N". + If in doubt, say "N". config DRM_I915_DEBUG_GUC - bool "Enable additional driver debugging for GuC" - depends on DRM_I915 - default n - help - Choose this option to turn on extra driver debugging that may affect - performance but will help resolve GuC related issues. + bool "Enable additional driver debugging for GuC" + depends on DRM_I915 + default n + help + Choose this option to turn on extra driver debugging that may affect + performance but will help resolve GuC related issues. - Recommended for driver developers only. + Recommended for driver developers only. - If in doubt, say "N". + If in doubt, say "N". config DRM_I915_SELFTEST bool "Enable selftests upon driver load" @@ -178,15 +176,15 @@ config DRM_I915_SELFTEST_BROKEN If in doubt, say "N". config DRM_I915_LOW_LEVEL_TRACEPOINTS - bool "Enable low level request tracing events" - depends on DRM_I915 - default n - help - Choose this option to turn on low level request tracing events. - This provides the ability to precisely monitor engine utilisation - and also analyze the request dependency resolving timeline. - - If in doubt, say "N". + bool "Enable low level request tracing events" + depends on DRM_I915 + default n + help + Choose this option to turn on low level request tracing events. + This provides the ability to precisely monitor engine utilisation + and also analyze the request dependency resolving timeline. + + If in doubt, say "N". config DRM_I915_DEBUG_VBLANK_EVADE bool "Enable extra debug warnings for vblank evasion" diff --git a/drivers/gpu/drm/i915/Kconfig.profile b/drivers/gpu/drm/i915/Kconfig.profile index 48df8889a88a..c280b6ae38eb 100644 --- a/drivers/gpu/drm/i915/Kconfig.profile +++ b/drivers/gpu/drm/i915/Kconfig.profile @@ -12,6 +12,29 @@ config DRM_I915_USERFAULT_AUTOSUSPEND May be 0 to disable the extra delay and solely use the device level runtime pm autosuspend delay tunable. +config DRM_I915_HEARTBEAT_INTERVAL + int "Interval between heartbeat pulses (ms)" + default 2500 # milliseconds + help + The driver sends a periodic heartbeat down all active engines to + check the health of the GPU and undertake regular house-keeping of + internal driver state. + + May be 0 to disable heartbeats and therefore disable automatic GPU + hang detection. + +config DRM_I915_PREEMPT_TIMEOUT + int "Preempt timeout (ms, jiffy granularity)" + default 640 # milliseconds + help + How long to wait (in milliseconds) for a preemption event to occur + when submitting a new context via execlists. If the current context + does not hit an arbitration point and yield to HW before the timer + expires, the HW will be reset to allow the more important context + to execute. + + May be 0 to disable the timeout. + config DRM_I915_SPIN_REQUEST int "Busywait for request completion (us)" default 5 # microseconds @@ -25,3 +48,29 @@ config DRM_I915_SPIN_REQUEST May be 0 to disable the initial spin. In practice, we estimate the cost of enabling the interrupt (if currently disabled) to be a few microseconds. + +config DRM_I915_STOP_TIMEOUT + int "How long to wait for an engine to quiesce gracefully before reset (ms)" + default 100 # milliseconds + help + By stopping submission and sleeping for a short time before resetting + the GPU, we allow the innocent contexts also on the system to quiesce. + It is then less likely for a hanging context to cause collateral + damage as the system is reset in order to recover. The corollary is + that the reset itself may take longer and so be more disruptive to + interactive or low latency workloads. + +config DRM_I915_TIMESLICE_DURATION + int "Scheduling quantum for userspace batches (ms, jiffy granularity)" + default 1 # milliseconds + help + When two user batches of equal priority are executing, we will + alternate execution of each batch to ensure forward progress of + all users. This is necessary in some cases where there may be + an implicit dependency between those batches that requires + concurrent execution in order for them to proceed, e.g. they + interact with each other via userspace semaphores. Each context + is scheduled for execution for the timeslice duration, before + switching to the next context. + + May be 0 to disable timeslicing. diff --git a/drivers/gpu/drm/i915/Kconfig.unstable b/drivers/gpu/drm/i915/Kconfig.unstable new file mode 100644 index 000000000000..0c2276155c2b --- /dev/null +++ b/drivers/gpu/drm/i915/Kconfig.unstable @@ -0,0 +1,29 @@ +# SPDX-License-Identifier: GPL-2.0-only +config DRM_I915_UNSTABLE + bool "Enable unstable API for early prototype development" + depends on EXPERT + depends on STAGING + depends on BROKEN # should never be enabled by distros! + # We use the dependency on !COMPILE_TEST to not be enabled in + # allmodconfig or allyesconfig configurations + depends on !COMPILE_TEST + default n + help + Enable prototype uAPI under general discussion before they are + finalized. Such prototypes may be withdrawn or substantially + changed before release. They are only enabled here so that a wide + number of interested parties (userspace driver developers) can + verify that the uAPI meet their expectations. These uAPI should + never be used in production. + + Recommended for driver developers _only_. + + If in the slightest bit of doubt, say "N". + +config DRM_I915_UNSTABLE_FAKE_LMEM + bool "Enable the experimental fake lmem" + depends on DRM_I915_UNSTABLE + default n + help + Convert some system memory into a fake local memory region for + testing. diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 2587ea834f06..90dcf09f52cc 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -46,10 +46,12 @@ i915-y += i915_drv.o \ i915_pci.o \ i915_scatterlist.o \ i915_suspend.o \ + i915_switcheroo.o \ i915_sysfs.o \ i915_utils.o \ intel_csr.o \ intel_device_info.o \ + intel_memory_region.o \ intel_pch.o \ intel_pm.o \ intel_runtime_pm.o \ @@ -76,19 +78,24 @@ gt-y += \ gt/intel_breadcrumbs.o \ gt/intel_context.o \ gt/intel_engine_cs.o \ - gt/intel_engine_pool.o \ + gt/intel_engine_heartbeat.o \ gt/intel_engine_pm.o \ + gt/intel_engine_pool.o \ gt/intel_engine_user.o \ gt/intel_gt.o \ gt/intel_gt_irq.o \ gt/intel_gt_pm.o \ gt/intel_gt_pm_irq.o \ - gt/intel_hangcheck.o \ + gt/intel_gt_requests.o \ + gt/intel_llc.o \ gt/intel_lrc.o \ + gt/intel_mocs.o \ + gt/intel_rc6.o \ gt/intel_renderstate.o \ gt/intel_reset.o \ - gt/intel_ringbuffer.o \ - gt/intel_mocs.o \ + gt/intel_ring.o \ + gt/intel_ring_submission.o \ + gt/intel_rps.o \ gt/intel_sseu.o \ gt/intel_timeline.o \ gt/intel_workarounds.o @@ -114,10 +121,12 @@ gem-y += \ gem/i915_gem_internal.o \ gem/i915_gem_object.o \ gem/i915_gem_object_blt.o \ + gem/i915_gem_lmem.o \ gem/i915_gem_mman.o \ gem/i915_gem_pages.o \ gem/i915_gem_phys.o \ gem/i915_gem_pm.o \ + gem/i915_gem_region.o \ gem/i915_gem_shmem.o \ gem/i915_gem_shrinker.o \ gem/i915_gem_stolen.o \ @@ -141,6 +150,7 @@ i915-y += \ i915_scheduler.o \ i915_trace_points.o \ i915_vma.o \ + intel_region_lmem.o \ intel_wopcm.o # general-purpose microcontroller (GuC) support @@ -172,6 +182,7 @@ i915-y += \ display/intel_display_power.o \ display/intel_dpio_phy.o \ display/intel_dpll_mgr.o \ + display/intel_dsb.o \ display/intel_fbc.o \ display/intel_fifo_underrun.o \ display/intel_frontbuffer.o \ @@ -182,7 +193,8 @@ i915-y += \ display/intel_psr.o \ display/intel_quirks.o \ display/intel_sprite.o \ - display/intel_tc.o + display/intel_tc.o \ + display/intel_vga.o i915-$(CONFIG_ACPI) += \ display/intel_acpi.o \ display/intel_opregion.o @@ -235,7 +247,8 @@ i915-y += \ oa/i915_oa_cflgt2.o \ oa/i915_oa_cflgt3.o \ oa/i915_oa_cnl.o \ - oa/i915_oa_icl.o + oa/i915_oa_icl.o \ + oa/i915_oa_tgl.o i915-y += i915_perf.o # Post-mortem debug and GPU hang state capture diff --git a/drivers/gpu/drm/i915/display/icl_dsi.c b/drivers/gpu/drm/i915/display/icl_dsi.c index 6e398c33a524..325df29b0447 100644 --- a/drivers/gpu/drm/i915/display/icl_dsi.c +++ b/drivers/gpu/drm/i915/display/icl_dsi.c @@ -1584,7 +1584,7 @@ void icl_dsi_init(struct drm_i915_private *dev_priv) encoder->get_hw_state = gen11_dsi_get_hw_state; encoder->type = INTEL_OUTPUT_DSI; encoder->cloneable = 0; - encoder->crtc_mask = BIT(PIPE_A) | BIT(PIPE_B) | BIT(PIPE_C); + encoder->pipe_mask = ~0; encoder->power_domain = POWER_DOMAIN_PORT_DSI; encoder->get_power_domains = gen11_dsi_get_power_domains; diff --git a/drivers/gpu/drm/i915/display/intel_atomic.c b/drivers/gpu/drm/i915/display/intel_atomic.c index d3fb75bb9eb1..c2875b10adf9 100644 --- a/drivers/gpu/drm/i915/display/intel_atomic.c +++ b/drivers/gpu/drm/i915/display/intel_atomic.c @@ -199,8 +199,8 @@ intel_crtc_duplicate_state(struct drm_crtc *crtc) crtc_state->disable_cxsr = false; crtc_state->update_wm_pre = false; crtc_state->update_wm_post = false; - crtc_state->fb_changed = false; crtc_state->fifo_changed = false; + crtc_state->preload_luts = false; crtc_state->wm.need_postvbl_update = false; crtc_state->fb_bits = 0; crtc_state->update_planes = 0; @@ -264,10 +264,13 @@ static void intel_atomic_setup_scaler(struct intel_crtc_scaler_state *scaler_sta */ mode = PS_SCALER_MODE_NORMAL; } else { + struct intel_plane *linked = + plane_state->planar_linked_plane; + mode = PS_SCALER_MODE_PLANAR; - if (plane_state->linked_plane) - mode |= PS_PLANE_Y_SEL(plane_state->linked_plane->id); + if (linked) + mode |= PS_PLANE_Y_SEL(linked->id); } } else if (INTEL_GEN(dev_priv) > 9 || IS_GEMINILAKE(dev_priv)) { mode = PS_SCALER_MODE_NORMAL; @@ -371,6 +374,15 @@ int intel_atomic_setup_scalers(struct drm_i915_private *dev_priv, */ if (!plane) { struct drm_plane_state *state; + + /* + * GLK+ scalers don't have a HQ mode so it + * isn't necessary to change between HQ and dyn mode + * on those platforms. + */ + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) + continue; + plane = drm_plane_from_index(&dev_priv->drm, i); state = drm_atomic_get_plane_state(drm_state, plane); if (IS_ERR(state)) { @@ -378,13 +390,6 @@ int intel_atomic_setup_scalers(struct drm_i915_private *dev_priv, plane->base.id); return PTR_ERR(state); } - - /* - * the plane is added after plane checks are run, - * but since this plane is unchanged just do the - * minimum required validation. - */ - crtc_state->base.planes_changed = true; } intel_plane = to_intel_plane(plane); @@ -425,6 +430,13 @@ void intel_atomic_state_clear(struct drm_atomic_state *s) struct intel_atomic_state *state = to_intel_atomic_state(s); drm_atomic_state_default_clear(&state->base); state->dpll_set = state->modeset = false; + state->global_state_changed = false; + state->active_pipes = 0; + memset(&state->min_cdclk, 0, sizeof(state->min_cdclk)); + memset(&state->min_voltage_level, 0, sizeof(state->min_voltage_level)); + memset(&state->cdclk.logical, 0, sizeof(state->cdclk.logical)); + memset(&state->cdclk.actual, 0, sizeof(state->cdclk.actual)); + state->cdclk.pipe = INVALID_PIPE; } struct intel_crtc_state * @@ -438,3 +450,40 @@ intel_atomic_get_crtc_state(struct drm_atomic_state *state, return to_intel_crtc_state(crtc_state); } + +int intel_atomic_lock_global_state(struct intel_atomic_state *state) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + struct intel_crtc *crtc; + + state->global_state_changed = true; + + for_each_intel_crtc(&dev_priv->drm, crtc) { + int ret; + + ret = drm_modeset_lock(&crtc->base.mutex, + state->base.acquire_ctx); + if (ret) + return ret; + } + + return 0; +} + +int intel_atomic_serialize_global_state(struct intel_atomic_state *state) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + struct intel_crtc *crtc; + + state->global_state_changed = true; + + for_each_intel_crtc(&dev_priv->drm, crtc) { + struct intel_crtc_state *crtc_state; + + crtc_state = intel_atomic_get_crtc_state(&state->base, crtc); + if (IS_ERR(crtc_state)) + return PTR_ERR(crtc_state); + } + + return 0; +} diff --git a/drivers/gpu/drm/i915/display/intel_atomic.h b/drivers/gpu/drm/i915/display/intel_atomic.h index 58065d3161a3..49d5cb1b9e0a 100644 --- a/drivers/gpu/drm/i915/display/intel_atomic.h +++ b/drivers/gpu/drm/i915/display/intel_atomic.h @@ -16,6 +16,7 @@ struct drm_crtc_state; struct drm_device; struct drm_i915_private; struct drm_property; +struct intel_atomic_state; struct intel_crtc; struct intel_crtc_state; @@ -46,4 +47,8 @@ int intel_atomic_setup_scalers(struct drm_i915_private *dev_priv, struct intel_crtc *intel_crtc, struct intel_crtc_state *crtc_state); +int intel_atomic_lock_global_state(struct intel_atomic_state *state); + +int intel_atomic_serialize_global_state(struct intel_atomic_state *state); + #endif /* __INTEL_ATOMIC_H__ */ diff --git a/drivers/gpu/drm/i915/display/intel_atomic_plane.c b/drivers/gpu/drm/i915/display/intel_atomic_plane.c index d1fcdf206da4..98f557a9f8ee 100644 --- a/drivers/gpu/drm/i915/display/intel_atomic_plane.c +++ b/drivers/gpu/drm/i915/display/intel_atomic_plane.c @@ -138,18 +138,58 @@ unsigned int intel_plane_data_rate(const struct intel_crtc_state *crtc_state, return cpp * crtc_state->pixel_rate; } +bool intel_plane_calc_min_cdclk(struct intel_atomic_state *state, + struct intel_plane *plane) +{ + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + const struct intel_plane_state *plane_state = + intel_atomic_get_new_plane_state(state, plane); + struct intel_crtc *crtc = to_intel_crtc(plane_state->base.crtc); + struct intel_crtc_state *crtc_state; + + if (!plane_state->base.visible || !plane->min_cdclk) + return false; + + crtc_state = intel_atomic_get_new_crtc_state(state, crtc); + + crtc_state->min_cdclk[plane->id] = + plane->min_cdclk(crtc_state, plane_state); + + /* + * Does the cdclk need to be bumbed up? + * + * Note: we obviously need to be called before the new + * cdclk frequency is calculated so state->cdclk.logical + * hasn't been populated yet. Hence we look at the old + * cdclk state under dev_priv->cdclk.logical. This is + * safe as long we hold at least one crtc mutex (which + * must be true since we have crtc_state). + */ + if (crtc_state->min_cdclk[plane->id] > dev_priv->cdclk.logical.cdclk) { + DRM_DEBUG_KMS("[PLANE:%d:%s] min_cdclk (%d kHz) > logical cdclk (%d kHz)\n", + plane->base.base.id, plane->base.name, + crtc_state->min_cdclk[plane->id], + dev_priv->cdclk.logical.cdclk); + return true; + } + + return false; +} + int intel_plane_atomic_check_with_state(const struct intel_crtc_state *old_crtc_state, struct intel_crtc_state *new_crtc_state, const struct intel_plane_state *old_plane_state, struct intel_plane_state *new_plane_state) { struct intel_plane *plane = to_intel_plane(new_plane_state->base.plane); + const struct drm_framebuffer *fb = new_plane_state->base.fb; int ret; new_crtc_state->active_planes &= ~BIT(plane->id); new_crtc_state->nv12_planes &= ~BIT(plane->id); new_crtc_state->c8_planes &= ~BIT(plane->id); new_crtc_state->data_rate[plane->id] = 0; + new_crtc_state->min_cdclk[plane->id] = 0; new_plane_state->base.visible = false; if (!new_plane_state->base.crtc && !old_plane_state->base.crtc) @@ -164,11 +204,11 @@ int intel_plane_atomic_check_with_state(const struct intel_crtc_state *old_crtc_ new_crtc_state->active_planes |= BIT(plane->id); if (new_plane_state->base.visible && - is_planar_yuv_format(new_plane_state->base.fb->format->format)) + drm_format_info_is_yuv_semiplanar(fb->format)) new_crtc_state->nv12_planes |= BIT(plane->id); if (new_plane_state->base.visible && - new_plane_state->base.fb->format->format == DRM_FORMAT_C8) + fb->format->format == DRM_FORMAT_C8) new_crtc_state->c8_planes |= BIT(plane->id); if (new_plane_state->base.visible || old_plane_state->base.visible) @@ -194,14 +234,11 @@ get_crtc_from_states(const struct intel_plane_state *old_plane_state, return NULL; } -static int intel_plane_atomic_check(struct drm_plane *_plane, - struct drm_plane_state *_new_plane_state) +int intel_plane_atomic_check(struct intel_atomic_state *state, + struct intel_plane *plane) { - struct intel_plane *plane = to_intel_plane(_plane); - struct intel_atomic_state *state = - to_intel_atomic_state(_new_plane_state->state); struct intel_plane_state *new_plane_state = - to_intel_plane_state(_new_plane_state); + intel_atomic_get_new_plane_state(state, plane); const struct intel_plane_state *old_plane_state = intel_atomic_get_old_plane_state(state, plane); struct intel_crtc *crtc = @@ -320,9 +357,9 @@ void skl_update_planes_on_crtc(struct intel_atomic_state *state, if (new_plane_state->base.visible) { intel_update_plane(plane, new_crtc_state, new_plane_state); - } else if (new_plane_state->slave) { + } else if (new_plane_state->planar_slave) { struct intel_plane *master = - new_plane_state->linked_plane; + new_plane_state->planar_linked_plane; /* * We update the slave plane from this function because @@ -368,5 +405,4 @@ void i9xx_update_planes_on_crtc(struct intel_atomic_state *state, const struct drm_plane_helper_funcs intel_plane_helper_funcs = { .prepare_fb = intel_prepare_plane_fb, .cleanup_fb = intel_cleanup_plane_fb, - .atomic_check = intel_plane_atomic_check, }; diff --git a/drivers/gpu/drm/i915/display/intel_atomic_plane.h b/drivers/gpu/drm/i915/display/intel_atomic_plane.h index cb7ef4f9eafd..e61e9a82aadf 100644 --- a/drivers/gpu/drm/i915/display/intel_atomic_plane.h +++ b/drivers/gpu/drm/i915/display/intel_atomic_plane.h @@ -41,9 +41,13 @@ int intel_plane_atomic_check_with_state(const struct intel_crtc_state *old_crtc_ struct intel_crtc_state *crtc_state, const struct intel_plane_state *old_plane_state, struct intel_plane_state *intel_state); +int intel_plane_atomic_check(struct intel_atomic_state *state, + struct intel_plane *plane); int intel_plane_atomic_calc_changes(const struct intel_crtc_state *old_crtc_state, struct intel_crtc_state *crtc_state, const struct intel_plane_state *old_plane_state, struct intel_plane_state *plane_state); +bool intel_plane_calc_min_cdclk(struct intel_atomic_state *state, + struct intel_plane *plane); #endif /* __INTEL_ATOMIC_PLANE_H__ */ diff --git a/drivers/gpu/drm/i915/display/intel_audio.c b/drivers/gpu/drm/i915/display/intel_audio.c index ddcccf4408c3..85e6b2bbb34f 100644 --- a/drivers/gpu/drm/i915/display/intel_audio.c +++ b/drivers/gpu/drm/i915/display/intel_audio.c @@ -28,6 +28,7 @@ #include <drm/i915_component.h> #include "i915_drv.h" +#include "intel_atomic.h" #include "intel_audio.h" #include "intel_display_types.h" #include "intel_lpe_audio.h" @@ -560,8 +561,9 @@ static void ilk_audio_codec_disable(struct intel_encoder *encoder, u32 tmp, eldv; i915_reg_t aud_config, aud_cntrl_st2; - DRM_DEBUG_KMS("Disable audio codec on port %c, pipe %c\n", - port_name(port), pipe_name(pipe)); + DRM_DEBUG_KMS("Disable audio codec on [ENCODER:%d:%s], pipe %c\n", + encoder->base.base.id, encoder->base.name, + pipe_name(pipe)); if (WARN_ON(port == PORT_A)) return; @@ -609,8 +611,9 @@ static void ilk_audio_codec_enable(struct intel_encoder *encoder, int len, i; i915_reg_t hdmiw_hdmiedid, aud_config, aud_cntl_st, aud_cntrl_st2; - DRM_DEBUG_KMS("Enable audio codec on port %c, pipe %c, %u bytes ELD\n", - port_name(port), pipe_name(pipe), drm_eld_size(eld)); + DRM_DEBUG_KMS("Enable audio codec on [ENCODER:%d:%s], pipe %c, %u bytes ELD\n", + encoder->base.base.id, encoder->base.name, + pipe_name(pipe), drm_eld_size(eld)); if (WARN_ON(port == PORT_A)) return; @@ -816,13 +819,8 @@ retry: to_intel_atomic_state(state)->cdclk.force_min_cdclk = enable ? 2 * 96000 : 0; - /* - * Protects dev_priv->cdclk.force_min_cdclk - * Need to lock this here in case we have no active pipes - * and thus wouldn't lock it during the commit otherwise. - */ - ret = drm_modeset_lock(&dev_priv->drm.mode_config.connection_mutex, - &ctx); + /* Protects dev_priv->cdclk.force_min_cdclk */ + ret = intel_atomic_lock_global_state(to_intel_atomic_state(state)); if (!ret) ret = drm_atomic_commit(state); @@ -850,11 +848,23 @@ static unsigned long i915_audio_component_get_power(struct device *kdev) ret = intel_display_power_get(dev_priv, POWER_DOMAIN_AUDIO); - /* Force CDCLK to 2*BCLK as long as we need audio to be powered. */ - if (dev_priv->audio_power_refcount++ == 0) - if (IS_CANNONLAKE(dev_priv) || IS_GEMINILAKE(dev_priv)) + if (dev_priv->audio_power_refcount++ == 0) { + if (IS_TIGERLAKE(dev_priv) || IS_ICELAKE(dev_priv)) { + I915_WRITE(AUD_FREQ_CNTRL, dev_priv->audio_freq_cntrl); + DRM_DEBUG_KMS("restored AUD_FREQ_CNTRL to 0x%x\n", + dev_priv->audio_freq_cntrl); + } + + /* Force CDCLK to 2*BCLK as long as we need audio powered. */ + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) glk_force_audio_cdclk(dev_priv, true); + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) + I915_WRITE(AUD_PIN_BUF_CTL, + (I915_READ(AUD_PIN_BUF_CTL) | + AUD_PIN_BUF_ENABLE)); + } + return ret; } @@ -865,7 +875,7 @@ static void i915_audio_component_put_power(struct device *kdev, /* Stop forcing CDCLK to 2*BCLK if no need for audio to be powered. */ if (--dev_priv->audio_power_refcount == 0) - if (IS_CANNONLAKE(dev_priv) || IS_GEMINILAKE(dev_priv)) + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) glk_force_audio_cdclk(dev_priv, false); intel_display_power_put(dev_priv, POWER_DOMAIN_AUDIO, cookie); @@ -1114,6 +1124,12 @@ static void i915_audio_component_init(struct drm_i915_private *dev_priv) return; } + if (IS_TIGERLAKE(dev_priv) || IS_ICELAKE(dev_priv)) { + dev_priv->audio_freq_cntrl = I915_READ(AUD_FREQ_CNTRL); + DRM_DEBUG_KMS("init value of AUD_FREQ_CNTRL of 0x%x\n", + dev_priv->audio_freq_cntrl); + } + dev_priv->audio_component_registered = true; } diff --git a/drivers/gpu/drm/i915/display/intel_bios.c b/drivers/gpu/drm/i915/display/intel_bios.c index efb39f350b19..63c1bd4c2954 100644 --- a/drivers/gpu/drm/i915/display/intel_bios.c +++ b/drivers/gpu/drm/i915/display/intel_bios.c @@ -1270,7 +1270,7 @@ static void sanitize_ddc_pin(struct drm_i915_private *dev_priv, DRM_DEBUG_KMS("port %c trying to use the same DDC pin (0x%x) as port %c, " "disabling port %c DVI/HDMI support\n", port_name(port), info->alternate_ddc_pin, - port_name(p), port_name(port)); + port_name(p), port_name(p)); /* * If we have multiple ports supposedly sharing the @@ -1278,9 +1278,14 @@ static void sanitize_ddc_pin(struct drm_i915_private *dev_priv, * port. Otherwise they share the same ddc bin and * system couldn't communicate with them separately. * - * Give child device order the priority, first come first - * served. + * Give inverse child device order the priority, + * last one wins. Yes, there are real machines + * (eg. Asrock B250M-HDV) where VBT has both + * port A and port E with the same AUX ch and + * we must pick port E :( */ + info = &dev_priv->vbt.ddi_port_info[p]; + info->supports_dvi = false; info->supports_hdmi = false; info->alternate_ddc_pin = 0; @@ -1316,7 +1321,7 @@ static void sanitize_aux_ch(struct drm_i915_private *dev_priv, DRM_DEBUG_KMS("port %c trying to use the same AUX CH (0x%x) as port %c, " "disabling port %c DP support\n", port_name(port), info->alternate_aux_channel, - port_name(p), port_name(port)); + port_name(p), port_name(p)); /* * If we have multiple ports supposedlt sharing the @@ -1324,9 +1329,14 @@ static void sanitize_aux_ch(struct drm_i915_private *dev_priv, * port. Otherwise they share the same aux channel * and system couldn't communicate with them separately. * - * Give child device order the priority, first come first - * served. + * Give inverse child device order the priority, + * last one wins. Yes, there are real machines + * (eg. Asrock B250M-HDV) where VBT has both + * port A and port E with the same AUX ch and + * we must pick port E :( */ + info = &dev_priv->vbt.ddi_port_info[p]; + info->supports_dp = false; info->alternate_aux_channel = 0; } @@ -1389,6 +1399,7 @@ static enum port dvo_port_to_port(u8 dvo_port) [PORT_D] = { DVO_PORT_HDMID, DVO_PORT_DPD, -1}, [PORT_E] = { DVO_PORT_CRT, DVO_PORT_HDMIE, DVO_PORT_DPE}, [PORT_F] = { DVO_PORT_HDMIF, DVO_PORT_DPF, -1}, + [PORT_G] = { DVO_PORT_HDMIG, DVO_PORT_DPG, -1}, }; enum port port; int i; @@ -1615,7 +1626,7 @@ parse_general_definitions(struct drm_i915_private *dev_priv, expected_size = 37; } else if (bdb->version <= 215) { expected_size = 38; - } else if (bdb->version <= 216) { + } else if (bdb->version <= 229) { expected_size = 39; } else { expected_size = sizeof(*child); @@ -1833,7 +1844,7 @@ void intel_bios_init(struct drm_i915_private *dev_priv) const struct bdb_header *bdb; u8 __iomem *bios = NULL; - if (!HAS_DISPLAY(dev_priv)) { + if (!HAS_DISPLAY(dev_priv) || !INTEL_DISPLAY_ENABLED(dev_priv)) { DRM_DEBUG_KMS("Skipping VBT init due to disabled display.\n"); return; } @@ -2248,6 +2259,9 @@ enum aux_ch intel_bios_port_aux_ch(struct drm_i915_private *dev_priv, case DP_AUX_F: aux_ch = AUX_CH_F; break; + case DP_AUX_G: + aux_ch = AUX_CH_G; + break; default: MISSING_CASE(info->alternate_aux_channel); aux_ch = AUX_CH_A; diff --git a/drivers/gpu/drm/i915/display/intel_bios.h b/drivers/gpu/drm/i915/display/intel_bios.h index 4969189e620f..98f064828a57 100644 --- a/drivers/gpu/drm/i915/display/intel_bios.h +++ b/drivers/gpu/drm/i915/display/intel_bios.h @@ -1,5 +1,5 @@ /* - * Copyright © 2016 Intel Corporation + * Copyright © 2016-2019 Intel Corporation * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -35,6 +35,7 @@ #include <drm/i915_drm.h> struct drm_i915_private; +enum port; enum intel_backlight_type { INTEL_BACKLIGHT_PMIC, diff --git a/drivers/gpu/drm/i915/display/intel_bw.c b/drivers/gpu/drm/i915/display/intel_bw.c index 688858ebe4d0..22e83f857de8 100644 --- a/drivers/gpu/drm/i915/display/intel_bw.c +++ b/drivers/gpu/drm/i915/display/intel_bw.c @@ -35,28 +35,54 @@ static int icl_pcode_read_mem_global_info(struct drm_i915_private *dev_priv, if (ret) return ret; - switch (val & 0xf) { - case 0: - qi->dram_type = INTEL_DRAM_DDR4; - break; - case 1: - qi->dram_type = INTEL_DRAM_DDR3; - break; - case 2: - qi->dram_type = INTEL_DRAM_LPDDR3; - break; - case 3: - qi->dram_type = INTEL_DRAM_LPDDR3; - break; - default: - MISSING_CASE(val & 0xf); - break; + if (IS_GEN(dev_priv, 12)) { + switch (val & 0xf) { + case 0: + qi->dram_type = INTEL_DRAM_DDR4; + break; + case 3: + qi->dram_type = INTEL_DRAM_LPDDR4; + break; + case 4: + qi->dram_type = INTEL_DRAM_DDR3; + break; + case 5: + qi->dram_type = INTEL_DRAM_LPDDR3; + break; + default: + MISSING_CASE(val & 0xf); + break; + } + } else if (IS_GEN(dev_priv, 11)) { + switch (val & 0xf) { + case 0: + qi->dram_type = INTEL_DRAM_DDR4; + break; + case 1: + qi->dram_type = INTEL_DRAM_DDR3; + break; + case 2: + qi->dram_type = INTEL_DRAM_LPDDR3; + break; + case 3: + qi->dram_type = INTEL_DRAM_LPDDR4; + break; + default: + MISSING_CASE(val & 0xf); + break; + } + } else { + MISSING_CASE(INTEL_GEN(dev_priv)); + qi->dram_type = INTEL_DRAM_LPDDR3; /* Conservative default */ } qi->num_channels = (val & 0xf0) >> 4; qi->num_points = (val & 0xf00) >> 8; - qi->t_bl = qi->dram_type == INTEL_DRAM_DDR4 ? 4 : 8; + if (IS_GEN(dev_priv, 12)) + qi->t_bl = qi->dram_type == INTEL_DRAM_DDR4 ? 4 : 16; + else if (IS_GEN(dev_priv, 11)) + qi->t_bl = qi->dram_type == INTEL_DRAM_DDR4 ? 4 : 8; return 0; } @@ -132,20 +158,25 @@ static int icl_sagv_max_dclk(const struct intel_qgv_info *qi) } struct intel_sa_info { - u8 deburst, mpagesize, deprogbwlimit, displayrtids; + u16 displayrtids; + u8 deburst, deprogbwlimit; }; static const struct intel_sa_info icl_sa_info = { .deburst = 8, - .mpagesize = 16, .deprogbwlimit = 25, /* GB/s */ .displayrtids = 128, }; -static int icl_get_bw_info(struct drm_i915_private *dev_priv) +static const struct intel_sa_info tgl_sa_info = { + .deburst = 16, + .deprogbwlimit = 34, /* GB/s */ + .displayrtids = 256, +}; + +static int icl_get_bw_info(struct drm_i915_private *dev_priv, const struct intel_sa_info *sa) { struct intel_qgv_info qi = {}; - const struct intel_sa_info *sa = &icl_sa_info; bool is_y_tile = true; /* assume y tile may be used */ int num_channels; int deinterleave; @@ -233,14 +264,16 @@ static unsigned int icl_max_bw(struct drm_i915_private *dev_priv, void intel_bw_init_hw(struct drm_i915_private *dev_priv) { - if (IS_GEN(dev_priv, 11)) - icl_get_bw_info(dev_priv); + if (IS_GEN(dev_priv, 12)) + icl_get_bw_info(dev_priv, &tgl_sa_info); + else if (IS_GEN(dev_priv, 11)) + icl_get_bw_info(dev_priv, &icl_sa_info); } static unsigned int intel_max_data_rate(struct drm_i915_private *dev_priv, int num_planes) { - if (IS_GEN(dev_priv, 11)) + if (INTEL_GEN(dev_priv) >= 11) /* * FIXME with SAGV disabled maybe we can assume * point 1 will always be used? Seems to match diff --git a/drivers/gpu/drm/i915/display/intel_cdclk.c b/drivers/gpu/drm/i915/display/intel_cdclk.c index d0bc42e5039c..ed8c7ce62119 100644 --- a/drivers/gpu/drm/i915/display/intel_cdclk.c +++ b/drivers/gpu/drm/i915/display/intel_cdclk.c @@ -21,6 +21,7 @@ * DEALINGS IN THE SOFTWARE. */ +#include "intel_atomic.h" #include "intel_cdclk.h" #include "intel_display_types.h" #include "intel_sideband.h" @@ -1161,28 +1162,88 @@ static void skl_uninit_cdclk(struct drm_i915_private *dev_priv) skl_set_cdclk(dev_priv, &cdclk_state, INVALID_PIPE); } -static int bxt_calc_cdclk(int min_cdclk) -{ - if (min_cdclk > 576000) - return 624000; - else if (min_cdclk > 384000) - return 576000; - else if (min_cdclk > 288000) - return 384000; - else if (min_cdclk > 144000) - return 288000; - else - return 144000; +static const struct intel_cdclk_vals bxt_cdclk_table[] = { + { .refclk = 19200, .cdclk = 144000, .divider = 8, .ratio = 60 }, + { .refclk = 19200, .cdclk = 288000, .divider = 4, .ratio = 60 }, + { .refclk = 19200, .cdclk = 384000, .divider = 3, .ratio = 60 }, + { .refclk = 19200, .cdclk = 576000, .divider = 2, .ratio = 60 }, + { .refclk = 19200, .cdclk = 624000, .divider = 2, .ratio = 65 }, + {} +}; + +static const struct intel_cdclk_vals glk_cdclk_table[] = { + { .refclk = 19200, .cdclk = 79200, .divider = 8, .ratio = 33 }, + { .refclk = 19200, .cdclk = 158400, .divider = 4, .ratio = 33 }, + { .refclk = 19200, .cdclk = 316800, .divider = 2, .ratio = 33 }, + {} +}; + +static const struct intel_cdclk_vals cnl_cdclk_table[] = { + { .refclk = 19200, .cdclk = 168000, .divider = 4, .ratio = 35 }, + { .refclk = 19200, .cdclk = 336000, .divider = 2, .ratio = 35 }, + { .refclk = 19200, .cdclk = 528000, .divider = 2, .ratio = 55 }, + + { .refclk = 24000, .cdclk = 168000, .divider = 4, .ratio = 28 }, + { .refclk = 24000, .cdclk = 336000, .divider = 2, .ratio = 28 }, + { .refclk = 24000, .cdclk = 528000, .divider = 2, .ratio = 44 }, + {} +}; + +static const struct intel_cdclk_vals icl_cdclk_table[] = { + { .refclk = 19200, .cdclk = 172800, .divider = 2, .ratio = 18 }, + { .refclk = 19200, .cdclk = 192000, .divider = 2, .ratio = 20 }, + { .refclk = 19200, .cdclk = 307200, .divider = 2, .ratio = 32 }, + { .refclk = 19200, .cdclk = 326400, .divider = 4, .ratio = 68 }, + { .refclk = 19200, .cdclk = 556800, .divider = 2, .ratio = 58 }, + { .refclk = 19200, .cdclk = 652800, .divider = 2, .ratio = 68 }, + + { .refclk = 24000, .cdclk = 180000, .divider = 2, .ratio = 15 }, + { .refclk = 24000, .cdclk = 192000, .divider = 2, .ratio = 16 }, + { .refclk = 24000, .cdclk = 312000, .divider = 2, .ratio = 26 }, + { .refclk = 24000, .cdclk = 324000, .divider = 4, .ratio = 54 }, + { .refclk = 24000, .cdclk = 552000, .divider = 2, .ratio = 46 }, + { .refclk = 24000, .cdclk = 648000, .divider = 2, .ratio = 54 }, + + { .refclk = 38400, .cdclk = 172800, .divider = 2, .ratio = 9 }, + { .refclk = 38400, .cdclk = 192000, .divider = 2, .ratio = 10 }, + { .refclk = 38400, .cdclk = 307200, .divider = 2, .ratio = 16 }, + { .refclk = 38400, .cdclk = 326400, .divider = 4, .ratio = 34 }, + { .refclk = 38400, .cdclk = 556800, .divider = 2, .ratio = 29 }, + { .refclk = 38400, .cdclk = 652800, .divider = 2, .ratio = 34 }, + {} +}; + +static int bxt_calc_cdclk(struct drm_i915_private *dev_priv, int min_cdclk) +{ + const struct intel_cdclk_vals *table = dev_priv->cdclk.table; + int i; + + for (i = 0; table[i].refclk; i++) + if (table[i].refclk == dev_priv->cdclk.hw.ref && + table[i].cdclk >= min_cdclk) + return table[i].cdclk; + + WARN(1, "Cannot satisfy minimum cdclk %d with refclk %u\n", + min_cdclk, dev_priv->cdclk.hw.ref); + return 0; } -static int glk_calc_cdclk(int min_cdclk) +static int bxt_calc_cdclk_pll_vco(struct drm_i915_private *dev_priv, int cdclk) { - if (min_cdclk > 158400) - return 316800; - else if (min_cdclk > 79200) - return 158400; - else - return 79200; + const struct intel_cdclk_vals *table = dev_priv->cdclk.table; + int i; + + if (cdclk == dev_priv->cdclk.hw.bypass) + return 0; + + for (i = 0; table[i].refclk; i++) + if (table[i].refclk == dev_priv->cdclk.hw.ref && + table[i].cdclk == cdclk) + return dev_priv->cdclk.hw.ref * table[i].ratio; + + WARN(1, "cdclk %d not valid for refclk %u\n", + cdclk, dev_priv->cdclk.hw.ref); + return 0; } static u8 bxt_calc_voltage_level(int cdclk) @@ -1190,69 +1251,101 @@ static u8 bxt_calc_voltage_level(int cdclk) return DIV_ROUND_UP(cdclk, 25000); } -static int bxt_de_pll_vco(struct drm_i915_private *dev_priv, int cdclk) +static u8 cnl_calc_voltage_level(int cdclk) { - int ratio; - - if (cdclk == dev_priv->cdclk.hw.bypass) + if (cdclk > 336000) + return 2; + else if (cdclk > 168000) + return 1; + else return 0; +} - switch (cdclk) { - default: - MISSING_CASE(cdclk); - /* fall through */ - case 144000: - case 288000: - case 384000: - case 576000: - ratio = 60; - break; - case 624000: - ratio = 65; - break; - } +static u8 icl_calc_voltage_level(int cdclk) +{ + if (cdclk > 556800) + return 2; + else if (cdclk > 312000) + return 1; + else + return 0; +} - return dev_priv->cdclk.hw.ref * ratio; +static u8 ehl_calc_voltage_level(int cdclk) +{ + if (cdclk > 326400) + return 3; + else if (cdclk > 312000) + return 2; + else if (cdclk > 180000) + return 1; + else + return 0; } -static int glk_de_pll_vco(struct drm_i915_private *dev_priv, int cdclk) +static void cnl_readout_refclk(struct drm_i915_private *dev_priv, + struct intel_cdclk_state *cdclk_state) { - int ratio; + if (I915_READ(SKL_DSSM) & CNL_DSSM_CDCLK_PLL_REFCLK_24MHz) + cdclk_state->ref = 24000; + else + cdclk_state->ref = 19200; +} - if (cdclk == dev_priv->cdclk.hw.bypass) - return 0; +static void icl_readout_refclk(struct drm_i915_private *dev_priv, + struct intel_cdclk_state *cdclk_state) +{ + u32 dssm = I915_READ(SKL_DSSM) & ICL_DSSM_CDCLK_PLL_REFCLK_MASK; - switch (cdclk) { + switch (dssm) { default: - MISSING_CASE(cdclk); + MISSING_CASE(dssm); /* fall through */ - case 79200: - case 158400: - case 316800: - ratio = 33; + case ICL_DSSM_CDCLK_PLL_REFCLK_24MHz: + cdclk_state->ref = 24000; + break; + case ICL_DSSM_CDCLK_PLL_REFCLK_19_2MHz: + cdclk_state->ref = 19200; + break; + case ICL_DSSM_CDCLK_PLL_REFCLK_38_4MHz: + cdclk_state->ref = 38400; break; } - - return dev_priv->cdclk.hw.ref * ratio; } -static void bxt_de_pll_update(struct drm_i915_private *dev_priv, - struct intel_cdclk_state *cdclk_state) +static void bxt_de_pll_readout(struct drm_i915_private *dev_priv, + struct intel_cdclk_state *cdclk_state) { - u32 val; + u32 val, ratio; - cdclk_state->ref = 19200; - cdclk_state->vco = 0; + if (INTEL_GEN(dev_priv) >= 11) + icl_readout_refclk(dev_priv, cdclk_state); + else if (IS_CANNONLAKE(dev_priv)) + cnl_readout_refclk(dev_priv, cdclk_state); + else + cdclk_state->ref = 19200; val = I915_READ(BXT_DE_PLL_ENABLE); - if ((val & BXT_DE_PLL_PLL_ENABLE) == 0) + if ((val & BXT_DE_PLL_PLL_ENABLE) == 0 || + (val & BXT_DE_PLL_LOCK) == 0) { + /* + * CDCLK PLL is disabled, the VCO/ratio doesn't matter, but + * setting it to zero is a way to signal that. + */ + cdclk_state->vco = 0; return; + } - if (WARN_ON((val & BXT_DE_PLL_LOCK) == 0)) - return; + /* + * CNL+ have the ratio directly in the PLL enable register, gen9lp had + * it in a separate PLL control register. + */ + if (INTEL_GEN(dev_priv) >= 10) + ratio = val & CNL_CDCLK_PLL_RATIO_MASK; + else + ratio = I915_READ(BXT_DE_PLL_CTL) & BXT_DE_PLL_RATIO_MASK; - val = I915_READ(BXT_DE_PLL_CTL); - cdclk_state->vco = (val & BXT_DE_PLL_RATIO_MASK) * cdclk_state->ref; + cdclk_state->vco = ratio * cdclk_state->ref; } static void bxt_get_cdclk(struct drm_i915_private *dev_priv, @@ -1261,12 +1354,19 @@ static void bxt_get_cdclk(struct drm_i915_private *dev_priv, u32 divider; int div; - bxt_de_pll_update(dev_priv, cdclk_state); + bxt_de_pll_readout(dev_priv, cdclk_state); - cdclk_state->cdclk = cdclk_state->bypass = cdclk_state->ref; + if (INTEL_GEN(dev_priv) >= 12) + cdclk_state->bypass = cdclk_state->ref / 2; + else if (INTEL_GEN(dev_priv) >= 11) + cdclk_state->bypass = 50000; + else + cdclk_state->bypass = cdclk_state->ref; - if (cdclk_state->vco == 0) + if (cdclk_state->vco == 0) { + cdclk_state->cdclk = cdclk_state->bypass; goto out; + } divider = I915_READ(CDCLK_CTL) & BXT_CDCLK_CD2X_DIV_SEL_MASK; @@ -1275,13 +1375,15 @@ static void bxt_get_cdclk(struct drm_i915_private *dev_priv, div = 2; break; case BXT_CDCLK_CD2X_DIV_SEL_1_5: - WARN(IS_GEMINILAKE(dev_priv), "Unsupported divider\n"); + WARN(IS_GEMINILAKE(dev_priv) || INTEL_GEN(dev_priv) >= 10, + "Unsupported divider\n"); div = 3; break; case BXT_CDCLK_CD2X_DIV_SEL_2: div = 4; break; case BXT_CDCLK_CD2X_DIV_SEL_4: + WARN(INTEL_GEN(dev_priv) >= 10, "Unsupported divider\n"); div = 8; break; default: @@ -1297,7 +1399,7 @@ static void bxt_get_cdclk(struct drm_i915_private *dev_priv, * at least what the CDCLK frequency requires. */ cdclk_state->voltage_level = - bxt_calc_voltage_level(cdclk_state->cdclk); + dev_priv->display.calc_voltage_level(cdclk_state->cdclk); } static void bxt_de_pll_disable(struct drm_i915_private *dev_priv) @@ -1332,259 +1434,6 @@ static void bxt_de_pll_enable(struct drm_i915_private *dev_priv, int vco) dev_priv->cdclk.hw.vco = vco; } -static void bxt_set_cdclk(struct drm_i915_private *dev_priv, - const struct intel_cdclk_state *cdclk_state, - enum pipe pipe) -{ - int cdclk = cdclk_state->cdclk; - int vco = cdclk_state->vco; - u32 val, divider; - int ret; - - /* cdclk = vco / 2 / div{1,1.5,2,4} */ - switch (DIV_ROUND_CLOSEST(vco, cdclk)) { - default: - WARN_ON(cdclk != dev_priv->cdclk.hw.bypass); - WARN_ON(vco != 0); - /* fall through */ - case 2: - divider = BXT_CDCLK_CD2X_DIV_SEL_1; - break; - case 3: - WARN(IS_GEMINILAKE(dev_priv), "Unsupported divider\n"); - divider = BXT_CDCLK_CD2X_DIV_SEL_1_5; - break; - case 4: - divider = BXT_CDCLK_CD2X_DIV_SEL_2; - break; - case 8: - divider = BXT_CDCLK_CD2X_DIV_SEL_4; - break; - } - - /* - * Inform power controller of upcoming frequency change. BSpec - * requires us to wait up to 150usec, but that leads to timeouts; - * the 2ms used here is based on experiment. - */ - ret = sandybridge_pcode_write_timeout(dev_priv, - HSW_PCODE_DE_WRITE_FREQ_REQ, - 0x80000000, 150, 2); - if (ret) { - DRM_ERROR("PCode CDCLK freq change notify failed (err %d, freq %d)\n", - ret, cdclk); - return; - } - - if (dev_priv->cdclk.hw.vco != 0 && - dev_priv->cdclk.hw.vco != vco) - bxt_de_pll_disable(dev_priv); - - if (dev_priv->cdclk.hw.vco != vco) - bxt_de_pll_enable(dev_priv, vco); - - val = divider | skl_cdclk_decimal(cdclk); - if (pipe == INVALID_PIPE) - val |= BXT_CDCLK_CD2X_PIPE_NONE; - else - val |= BXT_CDCLK_CD2X_PIPE(pipe); - /* - * Disable SSA Precharge when CD clock frequency < 500 MHz, - * enable otherwise. - */ - if (cdclk >= 500000) - val |= BXT_CDCLK_SSA_PRECHARGE_ENABLE; - I915_WRITE(CDCLK_CTL, val); - - if (pipe != INVALID_PIPE) - intel_wait_for_vblank(dev_priv, pipe); - - /* - * The timeout isn't specified, the 2ms used here is based on - * experiment. - * FIXME: Waiting for the request completion could be delayed until - * the next PCODE request based on BSpec. - */ - ret = sandybridge_pcode_write_timeout(dev_priv, - HSW_PCODE_DE_WRITE_FREQ_REQ, - cdclk_state->voltage_level, 150, 2); - if (ret) { - DRM_ERROR("PCode CDCLK freq set failed, (err %d, freq %d)\n", - ret, cdclk); - return; - } - - intel_update_cdclk(dev_priv); -} - -static void bxt_sanitize_cdclk(struct drm_i915_private *dev_priv) -{ - u32 cdctl, expected; - - intel_update_cdclk(dev_priv); - intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); - - if (dev_priv->cdclk.hw.vco == 0 || - dev_priv->cdclk.hw.cdclk == dev_priv->cdclk.hw.bypass) - goto sanitize; - - /* DPLL okay; verify the cdclock - * - * Some BIOS versions leave an incorrect decimal frequency value and - * set reserved MBZ bits in CDCLK_CTL at least during exiting from S4, - * so sanitize this register. - */ - cdctl = I915_READ(CDCLK_CTL); - /* - * Let's ignore the pipe field, since BIOS could have configured the - * dividers both synching to an active pipe, or asynchronously - * (PIPE_NONE). - */ - cdctl &= ~BXT_CDCLK_CD2X_PIPE_NONE; - - expected = (cdctl & BXT_CDCLK_CD2X_DIV_SEL_MASK) | - skl_cdclk_decimal(dev_priv->cdclk.hw.cdclk); - /* - * Disable SSA Precharge when CD clock frequency < 500 MHz, - * enable otherwise. - */ - if (dev_priv->cdclk.hw.cdclk >= 500000) - expected |= BXT_CDCLK_SSA_PRECHARGE_ENABLE; - - if (cdctl == expected) - /* All well; nothing to sanitize */ - return; - -sanitize: - DRM_DEBUG_KMS("Sanitizing cdclk programmed by pre-os\n"); - - /* force cdclk programming */ - dev_priv->cdclk.hw.cdclk = 0; - - /* force full PLL disable + enable */ - dev_priv->cdclk.hw.vco = -1; -} - -static void bxt_init_cdclk(struct drm_i915_private *dev_priv) -{ - struct intel_cdclk_state cdclk_state; - - bxt_sanitize_cdclk(dev_priv); - - if (dev_priv->cdclk.hw.cdclk != 0 && - dev_priv->cdclk.hw.vco != 0) - return; - - cdclk_state = dev_priv->cdclk.hw; - - /* - * FIXME: - * - The initial CDCLK needs to be read from VBT. - * Need to make this change after VBT has changes for BXT. - */ - if (IS_GEMINILAKE(dev_priv)) { - cdclk_state.cdclk = glk_calc_cdclk(0); - cdclk_state.vco = glk_de_pll_vco(dev_priv, cdclk_state.cdclk); - } else { - cdclk_state.cdclk = bxt_calc_cdclk(0); - cdclk_state.vco = bxt_de_pll_vco(dev_priv, cdclk_state.cdclk); - } - cdclk_state.voltage_level = bxt_calc_voltage_level(cdclk_state.cdclk); - - bxt_set_cdclk(dev_priv, &cdclk_state, INVALID_PIPE); -} - -static void bxt_uninit_cdclk(struct drm_i915_private *dev_priv) -{ - struct intel_cdclk_state cdclk_state = dev_priv->cdclk.hw; - - cdclk_state.cdclk = cdclk_state.bypass; - cdclk_state.vco = 0; - cdclk_state.voltage_level = bxt_calc_voltage_level(cdclk_state.cdclk); - - bxt_set_cdclk(dev_priv, &cdclk_state, INVALID_PIPE); -} - -static int cnl_calc_cdclk(int min_cdclk) -{ - if (min_cdclk > 336000) - return 528000; - else if (min_cdclk > 168000) - return 336000; - else - return 168000; -} - -static u8 cnl_calc_voltage_level(int cdclk) -{ - if (cdclk > 336000) - return 2; - else if (cdclk > 168000) - return 1; - else - return 0; -} - -static void cnl_cdclk_pll_update(struct drm_i915_private *dev_priv, - struct intel_cdclk_state *cdclk_state) -{ - u32 val; - - if (I915_READ(SKL_DSSM) & CNL_DSSM_CDCLK_PLL_REFCLK_24MHz) - cdclk_state->ref = 24000; - else - cdclk_state->ref = 19200; - - cdclk_state->vco = 0; - - val = I915_READ(BXT_DE_PLL_ENABLE); - if ((val & BXT_DE_PLL_PLL_ENABLE) == 0) - return; - - if (WARN_ON((val & BXT_DE_PLL_LOCK) == 0)) - return; - - cdclk_state->vco = (val & CNL_CDCLK_PLL_RATIO_MASK) * cdclk_state->ref; -} - -static void cnl_get_cdclk(struct drm_i915_private *dev_priv, - struct intel_cdclk_state *cdclk_state) -{ - u32 divider; - int div; - - cnl_cdclk_pll_update(dev_priv, cdclk_state); - - cdclk_state->cdclk = cdclk_state->bypass = cdclk_state->ref; - - if (cdclk_state->vco == 0) - goto out; - - divider = I915_READ(CDCLK_CTL) & BXT_CDCLK_CD2X_DIV_SEL_MASK; - - switch (divider) { - case BXT_CDCLK_CD2X_DIV_SEL_1: - div = 2; - break; - case BXT_CDCLK_CD2X_DIV_SEL_2: - div = 4; - break; - default: - MISSING_CASE(divider); - return; - } - - cdclk_state->cdclk = DIV_ROUND_CLOSEST(cdclk_state->vco, div); - - out: - /* - * Can't read this out :( Let's assume it's - * at least what the CDCLK frequency requires. - */ - cdclk_state->voltage_level = - cnl_calc_voltage_level(cdclk_state->cdclk); -} - static void cnl_cdclk_pll_disable(struct drm_i915_private *dev_priv) { u32 val; @@ -1618,7 +1467,27 @@ static void cnl_cdclk_pll_enable(struct drm_i915_private *dev_priv, int vco) dev_priv->cdclk.hw.vco = vco; } -static void cnl_set_cdclk(struct drm_i915_private *dev_priv, +static u32 bxt_cdclk_cd2x_pipe(struct drm_i915_private *dev_priv, enum pipe pipe) +{ + if (INTEL_GEN(dev_priv) >= 12) { + if (pipe == INVALID_PIPE) + return TGL_CDCLK_CD2X_PIPE_NONE; + else + return TGL_CDCLK_CD2X_PIPE(pipe); + } else if (INTEL_GEN(dev_priv) >= 11) { + if (pipe == INVALID_PIPE) + return ICL_CDCLK_CD2X_PIPE_NONE; + else + return ICL_CDCLK_CD2X_PIPE(pipe); + } else { + if (pipe == INVALID_PIPE) + return BXT_CDCLK_CD2X_PIPE_NONE; + else + return BXT_CDCLK_CD2X_PIPE(pipe); + } +} + +static void bxt_set_cdclk(struct drm_i915_private *dev_priv, const struct intel_cdclk_state *cdclk_state, enum pipe pipe) { @@ -1627,17 +1496,28 @@ static void cnl_set_cdclk(struct drm_i915_private *dev_priv, u32 val, divider; int ret; - ret = skl_pcode_request(dev_priv, SKL_PCODE_CDCLK_CONTROL, - SKL_CDCLK_PREPARE_FOR_CHANGE, - SKL_CDCLK_READY_FOR_CHANGE, - SKL_CDCLK_READY_FOR_CHANGE, 3); + /* Inform power controller of upcoming frequency change. */ + if (INTEL_GEN(dev_priv) >= 10) + ret = skl_pcode_request(dev_priv, SKL_PCODE_CDCLK_CONTROL, + SKL_CDCLK_PREPARE_FOR_CHANGE, + SKL_CDCLK_READY_FOR_CHANGE, + SKL_CDCLK_READY_FOR_CHANGE, 3); + else + /* + * BSpec requires us to wait up to 150usec, but that leads to + * timeouts; the 2ms used here is based on experiment. + */ + ret = sandybridge_pcode_write_timeout(dev_priv, + HSW_PCODE_DE_WRITE_FREQ_REQ, + 0x80000000, 150, 2); + if (ret) { - DRM_ERROR("Failed to inform PCU about cdclk change (%d)\n", - ret); + DRM_ERROR("Failed to inform PCU about cdclk change (err %d, freq %d)\n", + ret, cdclk); return; } - /* cdclk = vco / 2 / div{1,2} */ + /* cdclk = vco / 2 / div{1,1.5,2,4} */ switch (DIV_ROUND_CLOSEST(vco, cdclk)) { default: WARN_ON(cdclk != dev_priv->cdclk.hw.bypass); @@ -1646,67 +1526,87 @@ static void cnl_set_cdclk(struct drm_i915_private *dev_priv, case 2: divider = BXT_CDCLK_CD2X_DIV_SEL_1; break; + case 3: + WARN(IS_GEMINILAKE(dev_priv) || INTEL_GEN(dev_priv) >= 10, + "Unsupported divider\n"); + divider = BXT_CDCLK_CD2X_DIV_SEL_1_5; + break; case 4: divider = BXT_CDCLK_CD2X_DIV_SEL_2; break; + case 8: + WARN(INTEL_GEN(dev_priv) >= 10, "Unsupported divider\n"); + divider = BXT_CDCLK_CD2X_DIV_SEL_4; + break; } - if (dev_priv->cdclk.hw.vco != 0 && - dev_priv->cdclk.hw.vco != vco) - cnl_cdclk_pll_disable(dev_priv); - - if (dev_priv->cdclk.hw.vco != vco) - cnl_cdclk_pll_enable(dev_priv, vco); + if (INTEL_GEN(dev_priv) >= 10) { + if (dev_priv->cdclk.hw.vco != 0 && + dev_priv->cdclk.hw.vco != vco) + cnl_cdclk_pll_disable(dev_priv); - val = divider | skl_cdclk_decimal(cdclk); - if (pipe == INVALID_PIPE) - val |= BXT_CDCLK_CD2X_PIPE_NONE; - else - val |= BXT_CDCLK_CD2X_PIPE(pipe); - I915_WRITE(CDCLK_CTL, val); + if (dev_priv->cdclk.hw.vco != vco) + cnl_cdclk_pll_enable(dev_priv, vco); - if (pipe != INVALID_PIPE) - intel_wait_for_vblank(dev_priv, pipe); + } else { + if (dev_priv->cdclk.hw.vco != 0 && + dev_priv->cdclk.hw.vco != vco) + bxt_de_pll_disable(dev_priv); - /* inform PCU of the change */ - sandybridge_pcode_write(dev_priv, SKL_PCODE_CDCLK_CONTROL, - cdclk_state->voltage_level); + if (dev_priv->cdclk.hw.vco != vco) + bxt_de_pll_enable(dev_priv, vco); + } - intel_update_cdclk(dev_priv); + val = divider | skl_cdclk_decimal(cdclk) | + bxt_cdclk_cd2x_pipe(dev_priv, pipe); /* - * Can't read out the voltage level :( - * Let's just assume everything is as expected. + * Disable SSA Precharge when CD clock frequency < 500 MHz, + * enable otherwise. */ - dev_priv->cdclk.hw.voltage_level = cdclk_state->voltage_level; -} + if (IS_GEN9_LP(dev_priv) && cdclk >= 500000) + val |= BXT_CDCLK_SSA_PRECHARGE_ENABLE; + I915_WRITE(CDCLK_CTL, val); -static int cnl_cdclk_pll_vco(struct drm_i915_private *dev_priv, int cdclk) -{ - int ratio; + if (pipe != INVALID_PIPE) + intel_wait_for_vblank(dev_priv, pipe); - if (cdclk == dev_priv->cdclk.hw.bypass) - return 0; + if (INTEL_GEN(dev_priv) >= 10) { + ret = sandybridge_pcode_write(dev_priv, SKL_PCODE_CDCLK_CONTROL, + cdclk_state->voltage_level); + } else { + /* + * The timeout isn't specified, the 2ms used here is based on + * experiment. + * FIXME: Waiting for the request completion could be delayed + * until the next PCODE request based on BSpec. + */ + ret = sandybridge_pcode_write_timeout(dev_priv, + HSW_PCODE_DE_WRITE_FREQ_REQ, + cdclk_state->voltage_level, + 150, 2); + } - switch (cdclk) { - default: - MISSING_CASE(cdclk); - /* fall through */ - case 168000: - case 336000: - ratio = dev_priv->cdclk.hw.ref == 19200 ? 35 : 28; - break; - case 528000: - ratio = dev_priv->cdclk.hw.ref == 19200 ? 55 : 44; - break; + if (ret) { + DRM_ERROR("PCode CDCLK freq set failed, (err %d, freq %d)\n", + ret, cdclk); + return; } - return dev_priv->cdclk.hw.ref * ratio; + intel_update_cdclk(dev_priv); + + if (INTEL_GEN(dev_priv) >= 10) + /* + * Can't read out the voltage level :( + * Let's just assume everything is as expected. + */ + dev_priv->cdclk.hw.voltage_level = cdclk_state->voltage_level; } -static void cnl_sanitize_cdclk(struct drm_i915_private *dev_priv) +static void bxt_sanitize_cdclk(struct drm_i915_private *dev_priv) { u32 cdctl, expected; + int cdclk, vco; intel_update_cdclk(dev_priv); intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); @@ -1727,262 +1627,65 @@ static void cnl_sanitize_cdclk(struct drm_i915_private *dev_priv) * dividers both synching to an active pipe, or asynchronously * (PIPE_NONE). */ - cdctl &= ~BXT_CDCLK_CD2X_PIPE_NONE; - - expected = (cdctl & BXT_CDCLK_CD2X_DIV_SEL_MASK) | - skl_cdclk_decimal(dev_priv->cdclk.hw.cdclk); - - if (cdctl == expected) - /* All well; nothing to sanitize */ - return; - -sanitize: - DRM_DEBUG_KMS("Sanitizing cdclk programmed by pre-os\n"); + cdctl &= ~bxt_cdclk_cd2x_pipe(dev_priv, INVALID_PIPE); - /* force cdclk programming */ - dev_priv->cdclk.hw.cdclk = 0; + /* Make sure this is a legal cdclk value for the platform */ + cdclk = bxt_calc_cdclk(dev_priv, dev_priv->cdclk.hw.cdclk); + if (cdclk != dev_priv->cdclk.hw.cdclk) + goto sanitize; - /* force full PLL disable + enable */ - dev_priv->cdclk.hw.vco = -1; -} + /* Make sure the VCO is correct for the cdclk */ + vco = bxt_calc_cdclk_pll_vco(dev_priv, cdclk); + if (vco != dev_priv->cdclk.hw.vco) + goto sanitize; -static int icl_calc_cdclk(int min_cdclk, unsigned int ref) -{ - static const int ranges_24[] = { 180000, 192000, 312000, 552000, 648000 }; - static const int ranges_19_38[] = { 172800, 192000, 307200, 556800, 652800 }; - const int *ranges; - int len, i; + expected = skl_cdclk_decimal(cdclk); - switch (ref) { - default: - MISSING_CASE(ref); - /* fall through */ - case 24000: - ranges = ranges_24; - len = ARRAY_SIZE(ranges_24); - break; - case 19200: - case 38400: - ranges = ranges_19_38; - len = ARRAY_SIZE(ranges_19_38); + /* Figure out what CD2X divider we should be using for this cdclk */ + switch (DIV_ROUND_CLOSEST(dev_priv->cdclk.hw.vco, + dev_priv->cdclk.hw.cdclk)) { + case 2: + expected |= BXT_CDCLK_CD2X_DIV_SEL_1; break; - } - - for (i = 0; i < len; i++) { - if (min_cdclk <= ranges[i]) - return ranges[i]; - } - - WARN_ON(min_cdclk > ranges[len - 1]); - return ranges[len - 1]; -} - -static int icl_calc_cdclk_pll_vco(struct drm_i915_private *dev_priv, int cdclk) -{ - int ratio; - - if (cdclk == dev_priv->cdclk.hw.bypass) - return 0; - - switch (cdclk) { - default: - MISSING_CASE(cdclk); - /* fall through */ - case 172800: - case 307200: - case 556800: - case 652800: - WARN_ON(dev_priv->cdclk.hw.ref != 19200 && - dev_priv->cdclk.hw.ref != 38400); + case 3: + expected |= BXT_CDCLK_CD2X_DIV_SEL_1_5; break; - case 180000: - case 312000: - case 552000: - case 648000: - WARN_ON(dev_priv->cdclk.hw.ref != 24000); + case 4: + expected |= BXT_CDCLK_CD2X_DIV_SEL_2; break; - case 192000: - WARN_ON(dev_priv->cdclk.hw.ref != 19200 && - dev_priv->cdclk.hw.ref != 38400 && - dev_priv->cdclk.hw.ref != 24000); + case 8: + expected |= BXT_CDCLK_CD2X_DIV_SEL_4; break; - } - - ratio = cdclk / (dev_priv->cdclk.hw.ref / 2); - - return dev_priv->cdclk.hw.ref * ratio; -} - -static void icl_set_cdclk(struct drm_i915_private *dev_priv, - const struct intel_cdclk_state *cdclk_state, - enum pipe pipe) -{ - unsigned int cdclk = cdclk_state->cdclk; - unsigned int vco = cdclk_state->vco; - int ret; - - ret = skl_pcode_request(dev_priv, SKL_PCODE_CDCLK_CONTROL, - SKL_CDCLK_PREPARE_FOR_CHANGE, - SKL_CDCLK_READY_FOR_CHANGE, - SKL_CDCLK_READY_FOR_CHANGE, 3); - if (ret) { - DRM_ERROR("Failed to inform PCU about cdclk change (%d)\n", - ret); - return; - } - - if (dev_priv->cdclk.hw.vco != 0 && - dev_priv->cdclk.hw.vco != vco) - cnl_cdclk_pll_disable(dev_priv); - - if (dev_priv->cdclk.hw.vco != vco) - cnl_cdclk_pll_enable(dev_priv, vco); - - /* - * On ICL CD2X_DIV can only be 1, so we'll never end up changing the - * divider here synchronized to a pipe while CDCLK is on, nor will we - * need the corresponding vblank wait. - */ - I915_WRITE(CDCLK_CTL, ICL_CDCLK_CD2X_PIPE_NONE | - skl_cdclk_decimal(cdclk)); - - sandybridge_pcode_write(dev_priv, SKL_PCODE_CDCLK_CONTROL, - cdclk_state->voltage_level); - - intel_update_cdclk(dev_priv); - - /* - * Can't read out the voltage level :( - * Let's just assume everything is as expected. - */ - dev_priv->cdclk.hw.voltage_level = cdclk_state->voltage_level; -} - -static u8 icl_calc_voltage_level(struct drm_i915_private *dev_priv, int cdclk) -{ - if (IS_ELKHARTLAKE(dev_priv)) { - if (cdclk > 312000) - return 2; - else if (cdclk > 180000) - return 1; - else - return 0; - } else { - if (cdclk > 556800) - return 2; - else if (cdclk > 312000) - return 1; - else - return 0; - } -} - -static void icl_get_cdclk(struct drm_i915_private *dev_priv, - struct intel_cdclk_state *cdclk_state) -{ - u32 val; - - cdclk_state->bypass = 50000; - - val = I915_READ(SKL_DSSM); - switch (val & ICL_DSSM_CDCLK_PLL_REFCLK_MASK) { default: - MISSING_CASE(val); - /* fall through */ - case ICL_DSSM_CDCLK_PLL_REFCLK_24MHz: - cdclk_state->ref = 24000; - break; - case ICL_DSSM_CDCLK_PLL_REFCLK_19_2MHz: - cdclk_state->ref = 19200; - break; - case ICL_DSSM_CDCLK_PLL_REFCLK_38_4MHz: - cdclk_state->ref = 38400; - break; - } - - val = I915_READ(BXT_DE_PLL_ENABLE); - if ((val & BXT_DE_PLL_PLL_ENABLE) == 0 || - (val & BXT_DE_PLL_LOCK) == 0) { - /* - * CDCLK PLL is disabled, the VCO/ratio doesn't matter, but - * setting it to zero is a way to signal that. - */ - cdclk_state->vco = 0; - cdclk_state->cdclk = cdclk_state->bypass; - goto out; + goto sanitize; } - cdclk_state->vco = (val & BXT_DE_PLL_RATIO_MASK) * cdclk_state->ref; - - val = I915_READ(CDCLK_CTL); - WARN_ON((val & BXT_CDCLK_CD2X_DIV_SEL_MASK) != 0); - - cdclk_state->cdclk = cdclk_state->vco / 2; - -out: /* - * Can't read this out :( Let's assume it's - * at least what the CDCLK frequency requires. + * Disable SSA Precharge when CD clock frequency < 500 MHz, + * enable otherwise. */ - cdclk_state->voltage_level = - icl_calc_voltage_level(dev_priv, cdclk_state->cdclk); -} - -static void icl_init_cdclk(struct drm_i915_private *dev_priv) -{ - struct intel_cdclk_state sanitized_state; - u32 val; - - /* This sets dev_priv->cdclk.hw. */ - intel_update_cdclk(dev_priv); - intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); - - /* This means CDCLK disabled. */ - if (dev_priv->cdclk.hw.cdclk == dev_priv->cdclk.hw.bypass) - goto sanitize; - - val = I915_READ(CDCLK_CTL); - - if ((val & BXT_CDCLK_CD2X_DIV_SEL_MASK) != 0) - goto sanitize; - - if ((val & CDCLK_FREQ_DECIMAL_MASK) != - skl_cdclk_decimal(dev_priv->cdclk.hw.cdclk)) - goto sanitize; + if (IS_GEN9_LP(dev_priv) && dev_priv->cdclk.hw.cdclk >= 500000) + expected |= BXT_CDCLK_SSA_PRECHARGE_ENABLE; - return; + if (cdctl == expected) + /* All well; nothing to sanitize */ + return; sanitize: DRM_DEBUG_KMS("Sanitizing cdclk programmed by pre-os\n"); - sanitized_state.ref = dev_priv->cdclk.hw.ref; - sanitized_state.cdclk = icl_calc_cdclk(0, sanitized_state.ref); - sanitized_state.vco = icl_calc_cdclk_pll_vco(dev_priv, - sanitized_state.cdclk); - sanitized_state.voltage_level = - icl_calc_voltage_level(dev_priv, - sanitized_state.cdclk); - - icl_set_cdclk(dev_priv, &sanitized_state, INVALID_PIPE); -} - -static void icl_uninit_cdclk(struct drm_i915_private *dev_priv) -{ - struct intel_cdclk_state cdclk_state = dev_priv->cdclk.hw; - - cdclk_state.cdclk = cdclk_state.bypass; - cdclk_state.vco = 0; - cdclk_state.voltage_level = icl_calc_voltage_level(dev_priv, - cdclk_state.cdclk); + /* force cdclk programming */ + dev_priv->cdclk.hw.cdclk = 0; - icl_set_cdclk(dev_priv, &cdclk_state, INVALID_PIPE); + /* force full PLL disable + enable */ + dev_priv->cdclk.hw.vco = -1; } -static void cnl_init_cdclk(struct drm_i915_private *dev_priv) +static void bxt_init_cdclk(struct drm_i915_private *dev_priv) { struct intel_cdclk_state cdclk_state; - cnl_sanitize_cdclk(dev_priv); + bxt_sanitize_cdclk(dev_priv); if (dev_priv->cdclk.hw.cdclk != 0 && dev_priv->cdclk.hw.vco != 0) @@ -1990,22 +1693,29 @@ static void cnl_init_cdclk(struct drm_i915_private *dev_priv) cdclk_state = dev_priv->cdclk.hw; - cdclk_state.cdclk = cnl_calc_cdclk(0); - cdclk_state.vco = cnl_cdclk_pll_vco(dev_priv, cdclk_state.cdclk); - cdclk_state.voltage_level = cnl_calc_voltage_level(cdclk_state.cdclk); + /* + * FIXME: + * - The initial CDCLK needs to be read from VBT. + * Need to make this change after VBT has changes for BXT. + */ + cdclk_state.cdclk = bxt_calc_cdclk(dev_priv, 0); + cdclk_state.vco = bxt_calc_cdclk_pll_vco(dev_priv, cdclk_state.cdclk); + cdclk_state.voltage_level = + dev_priv->display.calc_voltage_level(cdclk_state.cdclk); - cnl_set_cdclk(dev_priv, &cdclk_state, INVALID_PIPE); + bxt_set_cdclk(dev_priv, &cdclk_state, INVALID_PIPE); } -static void cnl_uninit_cdclk(struct drm_i915_private *dev_priv) +static void bxt_uninit_cdclk(struct drm_i915_private *dev_priv) { struct intel_cdclk_state cdclk_state = dev_priv->cdclk.hw; cdclk_state.cdclk = cdclk_state.bypass; cdclk_state.vco = 0; - cdclk_state.voltage_level = cnl_calc_voltage_level(cdclk_state.cdclk); + cdclk_state.voltage_level = + dev_priv->display.calc_voltage_level(cdclk_state.cdclk); - cnl_set_cdclk(dev_priv, &cdclk_state, INVALID_PIPE); + bxt_set_cdclk(dev_priv, &cdclk_state, INVALID_PIPE); } /** @@ -2019,14 +1729,10 @@ static void cnl_uninit_cdclk(struct drm_i915_private *dev_priv) */ void intel_cdclk_init(struct drm_i915_private *i915) { - if (INTEL_GEN(i915) >= 11) - icl_init_cdclk(i915); - else if (IS_CANNONLAKE(i915)) - cnl_init_cdclk(i915); + if (IS_GEN9_LP(i915) || INTEL_GEN(i915) >= 10) + bxt_init_cdclk(i915); else if (IS_GEN9_BC(i915)) skl_init_cdclk(i915); - else if (IS_GEN9_LP(i915)) - bxt_init_cdclk(i915); } /** @@ -2038,14 +1744,10 @@ void intel_cdclk_init(struct drm_i915_private *i915) */ void intel_cdclk_uninit(struct drm_i915_private *i915) { - if (INTEL_GEN(i915) >= 11) - icl_uninit_cdclk(i915); - else if (IS_CANNONLAKE(i915)) - cnl_uninit_cdclk(i915); + if (INTEL_GEN(i915) >= 10 || IS_GEN9_LP(i915)) + bxt_uninit_cdclk(i915); else if (IS_GEN9_BC(i915)) skl_uninit_cdclk(i915); - else if (IS_GEN9_LP(i915)) - bxt_uninit_cdclk(i915); } /** @@ -2073,9 +1775,9 @@ bool intel_cdclk_needs_modeset(const struct intel_cdclk_state *a, * Returns: * True if the CDCLK states require just a cd2x divider update, false if not. */ -bool intel_cdclk_needs_cd2x_update(struct drm_i915_private *dev_priv, - const struct intel_cdclk_state *a, - const struct intel_cdclk_state *b) +static bool intel_cdclk_needs_cd2x_update(struct drm_i915_private *dev_priv, + const struct intel_cdclk_state *a, + const struct intel_cdclk_state *b) { /* Older hw doesn't have the capability */ if (INTEL_GEN(dev_priv) < 10 && !IS_GEN9_LP(dev_priv)) @@ -2094,8 +1796,8 @@ bool intel_cdclk_needs_cd2x_update(struct drm_i915_private *dev_priv, * Returns: * True if the CDCLK states don't match, false if they do. */ -bool intel_cdclk_changed(const struct intel_cdclk_state *a, - const struct intel_cdclk_state *b) +static bool intel_cdclk_changed(const struct intel_cdclk_state *a, + const struct intel_cdclk_state *b) { return intel_cdclk_needs_modeset(a, b) || a->voltage_level != b->voltage_level; @@ -2200,9 +1902,11 @@ intel_set_cdclk_post_plane_update(struct drm_i915_private *dev_priv, intel_set_cdclk(dev_priv, new_state, pipe); } -static int intel_pixel_rate_to_cdclk(struct drm_i915_private *dev_priv, - int pixel_rate) +static int intel_pixel_rate_to_cdclk(const struct intel_crtc_state *crtc_state) { + struct drm_i915_private *dev_priv = to_i915(crtc_state->base.crtc->dev); + int pixel_rate = crtc_state->pixel_rate; + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) return DIV_ROUND_UP(pixel_rate, 2); else if (IS_GEN(dev_priv, 9) || @@ -2210,10 +1914,25 @@ static int intel_pixel_rate_to_cdclk(struct drm_i915_private *dev_priv, return pixel_rate; else if (IS_CHERRYVIEW(dev_priv)) return DIV_ROUND_UP(pixel_rate * 100, 95); + else if (crtc_state->double_wide) + return DIV_ROUND_UP(pixel_rate * 100, 90 * 2); else return DIV_ROUND_UP(pixel_rate * 100, 90); } +static int intel_planes_min_cdclk(const struct intel_crtc_state *crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + struct intel_plane *plane; + int min_cdclk = 0; + + for_each_intel_plane_on_crtc(&dev_priv->drm, crtc, plane) + min_cdclk = max(crtc_state->min_cdclk[plane->id], min_cdclk); + + return min_cdclk; +} + int intel_crtc_compute_min_cdclk(const struct intel_crtc_state *crtc_state) { struct drm_i915_private *dev_priv = @@ -2223,7 +1942,7 @@ int intel_crtc_compute_min_cdclk(const struct intel_crtc_state *crtc_state) if (!crtc_state->base.enable) return 0; - min_cdclk = intel_pixel_rate_to_cdclk(dev_priv, crtc_state->pixel_rate); + min_cdclk = intel_pixel_rate_to_cdclk(crtc_state); /* pixel rate mustn't exceed 95% of cdclk with IPS on BDW */ if (IS_BROADWELL(dev_priv) && hsw_crtc_state_ips_capable(crtc_state)) @@ -2282,6 +2001,9 @@ int intel_crtc_compute_min_cdclk(const struct intel_crtc_state *crtc_state) IS_GEMINILAKE(dev_priv)) min_cdclk = max(158400, min_cdclk); + /* Account for additional needs from the planes */ + min_cdclk = max(intel_planes_min_cdclk(crtc_state), min_cdclk); + if (min_cdclk > dev_priv->max_cdclk_freq) { DRM_DEBUG_KMS("required cdclk (%d kHz) exceeds max (%d kHz)\n", min_cdclk, dev_priv->max_cdclk_freq); @@ -2303,11 +2025,20 @@ static int intel_compute_min_cdclk(struct intel_atomic_state *state) sizeof(state->min_cdclk)); for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) { + int ret; + min_cdclk = intel_crtc_compute_min_cdclk(crtc_state); if (min_cdclk < 0) return min_cdclk; + if (state->min_cdclk[i] == min_cdclk) + continue; + state->min_cdclk[i] = min_cdclk; + + ret = intel_atomic_lock_global_state(state); + if (ret) + return ret; } min_cdclk = state->cdclk.force_min_cdclk; @@ -2318,6 +2049,10 @@ static int intel_compute_min_cdclk(struct intel_atomic_state *state) } /* + * Account for port clock min voltage level requirements. + * This only really does something on CNL+ but can be + * called on earlier platforms as well. + * * Note that this functions assumes that 0 is * the lowest voltage value, and higher values * correspond to increasingly higher voltages. @@ -2326,7 +2061,7 @@ static int intel_compute_min_cdclk(struct intel_atomic_state *state) * future platforms this code will need to be * adjusted. */ -static u8 cnl_compute_min_voltage_level(struct intel_atomic_state *state) +static int bxt_compute_min_voltage_level(struct intel_atomic_state *state) { struct drm_i915_private *dev_priv = to_i915(state->base.dev); struct intel_crtc *crtc; @@ -2339,11 +2074,21 @@ static u8 cnl_compute_min_voltage_level(struct intel_atomic_state *state) sizeof(state->min_voltage_level)); for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) { + int ret; + if (crtc_state->base.enable) - state->min_voltage_level[i] = - crtc_state->min_voltage_level; + min_voltage_level = crtc_state->min_voltage_level; else - state->min_voltage_level[i] = 0; + min_voltage_level = 0; + + if (state->min_voltage_level[i] == min_voltage_level) + continue; + + state->min_voltage_level[i] = min_voltage_level; + + ret = intel_atomic_lock_global_state(state); + if (ret) + return ret; } min_voltage_level = 0; @@ -2369,7 +2114,7 @@ static int vlv_modeset_calc_cdclk(struct intel_atomic_state *state) state->cdclk.logical.voltage_level = vlv_calc_voltage_level(dev_priv, cdclk); - if (!state->active_crtcs) { + if (!state->active_pipes) { cdclk = vlv_calc_cdclk(dev_priv, state->cdclk.force_min_cdclk); state->cdclk.actual.cdclk = cdclk; @@ -2400,7 +2145,7 @@ static int bdw_modeset_calc_cdclk(struct intel_atomic_state *state) state->cdclk.logical.voltage_level = bdw_calc_voltage_level(cdclk); - if (!state->active_crtcs) { + if (!state->active_pipes) { cdclk = bdw_calc_cdclk(state->cdclk.force_min_cdclk); state->cdclk.actual.cdclk = cdclk; @@ -2470,7 +2215,7 @@ static int skl_modeset_calc_cdclk(struct intel_atomic_state *state) state->cdclk.logical.voltage_level = skl_calc_voltage_level(cdclk); - if (!state->active_crtcs) { + if (!state->active_pipes) { cdclk = skl_calc_cdclk(state->cdclk.force_min_cdclk, vco); state->cdclk.actual.vco = vco; @@ -2487,38 +2232,33 @@ static int skl_modeset_calc_cdclk(struct intel_atomic_state *state) static int bxt_modeset_calc_cdclk(struct intel_atomic_state *state) { struct drm_i915_private *dev_priv = to_i915(state->base.dev); - int min_cdclk, cdclk, vco; + int min_cdclk, min_voltage_level, cdclk, vco; min_cdclk = intel_compute_min_cdclk(state); if (min_cdclk < 0) return min_cdclk; - if (IS_GEMINILAKE(dev_priv)) { - cdclk = glk_calc_cdclk(min_cdclk); - vco = glk_de_pll_vco(dev_priv, cdclk); - } else { - cdclk = bxt_calc_cdclk(min_cdclk); - vco = bxt_de_pll_vco(dev_priv, cdclk); - } + min_voltage_level = bxt_compute_min_voltage_level(state); + if (min_voltage_level < 0) + return min_voltage_level; + + cdclk = bxt_calc_cdclk(dev_priv, min_cdclk); + vco = bxt_calc_cdclk_pll_vco(dev_priv, cdclk); state->cdclk.logical.vco = vco; state->cdclk.logical.cdclk = cdclk; state->cdclk.logical.voltage_level = - bxt_calc_voltage_level(cdclk); - - if (!state->active_crtcs) { - if (IS_GEMINILAKE(dev_priv)) { - cdclk = glk_calc_cdclk(state->cdclk.force_min_cdclk); - vco = glk_de_pll_vco(dev_priv, cdclk); - } else { - cdclk = bxt_calc_cdclk(state->cdclk.force_min_cdclk); - vco = bxt_de_pll_vco(dev_priv, cdclk); - } + max_t(int, min_voltage_level, + dev_priv->display.calc_voltage_level(cdclk)); + + if (!state->active_pipes) { + cdclk = bxt_calc_cdclk(dev_priv, state->cdclk.force_min_cdclk); + vco = bxt_calc_cdclk_pll_vco(dev_priv, cdclk); state->cdclk.actual.vco = vco; state->cdclk.actual.cdclk = cdclk; state->cdclk.actual.voltage_level = - bxt_calc_voltage_level(cdclk); + dev_priv->display.calc_voltage_level(cdclk); } else { state->cdclk.actual = state->cdclk.logical; } @@ -2526,70 +2266,138 @@ static int bxt_modeset_calc_cdclk(struct intel_atomic_state *state) return 0; } -static int cnl_modeset_calc_cdclk(struct intel_atomic_state *state) +static int intel_modeset_all_pipes(struct intel_atomic_state *state) { struct drm_i915_private *dev_priv = to_i915(state->base.dev); - int min_cdclk, cdclk, vco; + struct intel_crtc *crtc; - min_cdclk = intel_compute_min_cdclk(state); - if (min_cdclk < 0) - return min_cdclk; + /* + * Add all pipes to the state, and force + * a modeset on all the active ones. + */ + for_each_intel_crtc(&dev_priv->drm, crtc) { + struct intel_crtc_state *crtc_state; + int ret; - cdclk = cnl_calc_cdclk(min_cdclk); - vco = cnl_cdclk_pll_vco(dev_priv, cdclk); + crtc_state = intel_atomic_get_crtc_state(&state->base, crtc); + if (IS_ERR(crtc_state)) + return PTR_ERR(crtc_state); - state->cdclk.logical.vco = vco; - state->cdclk.logical.cdclk = cdclk; - state->cdclk.logical.voltage_level = - max(cnl_calc_voltage_level(cdclk), - cnl_compute_min_voltage_level(state)); + if (!crtc_state->base.active || + drm_atomic_crtc_needs_modeset(&crtc_state->base)) + continue; - if (!state->active_crtcs) { - cdclk = cnl_calc_cdclk(state->cdclk.force_min_cdclk); - vco = cnl_cdclk_pll_vco(dev_priv, cdclk); + crtc_state->base.mode_changed = true; - state->cdclk.actual.vco = vco; - state->cdclk.actual.cdclk = cdclk; - state->cdclk.actual.voltage_level = - cnl_calc_voltage_level(cdclk); - } else { - state->cdclk.actual = state->cdclk.logical; + ret = drm_atomic_add_affected_connectors(&state->base, + &crtc->base); + if (ret) + return ret; + + ret = drm_atomic_add_affected_planes(&state->base, + &crtc->base); + if (ret) + return ret; + + crtc_state->update_planes |= crtc_state->active_planes; } return 0; } -static int icl_modeset_calc_cdclk(struct intel_atomic_state *state) +static int fixed_modeset_calc_cdclk(struct intel_atomic_state *state) { - struct drm_i915_private *dev_priv = to_i915(state->base.dev); - unsigned int ref = state->cdclk.logical.ref; - int min_cdclk, cdclk, vco; + int min_cdclk; + /* + * We can't change the cdclk frequency, but we still want to + * check that the required minimum frequency doesn't exceed + * the actual cdclk frequency. + */ min_cdclk = intel_compute_min_cdclk(state); if (min_cdclk < 0) return min_cdclk; - cdclk = icl_calc_cdclk(min_cdclk, ref); - vco = icl_calc_cdclk_pll_vco(dev_priv, cdclk); + return 0; +} - state->cdclk.logical.vco = vco; - state->cdclk.logical.cdclk = cdclk; - state->cdclk.logical.voltage_level = - max(icl_calc_voltage_level(dev_priv, cdclk), - cnl_compute_min_voltage_level(state)); +int intel_modeset_calc_cdclk(struct intel_atomic_state *state) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + enum pipe pipe; + int ret; - if (!state->active_crtcs) { - cdclk = icl_calc_cdclk(state->cdclk.force_min_cdclk, ref); - vco = icl_calc_cdclk_pll_vco(dev_priv, cdclk); + ret = dev_priv->display.modeset_calc_cdclk(state); + if (ret) + return ret; - state->cdclk.actual.vco = vco; - state->cdclk.actual.cdclk = cdclk; - state->cdclk.actual.voltage_level = - icl_calc_voltage_level(dev_priv, cdclk); + /* + * Writes to dev_priv->cdclk.{actual,logical} must protected + * by holding all the crtc mutexes even if we don't end up + * touching the hardware + */ + if (intel_cdclk_changed(&dev_priv->cdclk.actual, + &state->cdclk.actual)) { + /* + * Also serialize commits across all crtcs + * if the actual hw needs to be poked. + */ + ret = intel_atomic_serialize_global_state(state); + if (ret) + return ret; + } else if (intel_cdclk_changed(&dev_priv->cdclk.logical, + &state->cdclk.logical)) { + ret = intel_atomic_lock_global_state(state); + if (ret) + return ret; } else { - state->cdclk.actual = state->cdclk.logical; + return 0; + } + + if (is_power_of_2(state->active_pipes) && + intel_cdclk_needs_cd2x_update(dev_priv, + &dev_priv->cdclk.actual, + &state->cdclk.actual)) { + struct intel_crtc *crtc; + struct intel_crtc_state *crtc_state; + + pipe = ilog2(state->active_pipes); + crtc = intel_get_crtc_for_pipe(dev_priv, pipe); + + crtc_state = intel_atomic_get_crtc_state(&state->base, crtc); + if (IS_ERR(crtc_state)) + return PTR_ERR(crtc_state); + + if (drm_atomic_crtc_needs_modeset(&crtc_state->base)) + pipe = INVALID_PIPE; + } else { + pipe = INVALID_PIPE; + } + + if (pipe != INVALID_PIPE) { + state->cdclk.pipe = pipe; + + DRM_DEBUG_KMS("Can change cdclk with pipe %c active\n", + pipe_name(pipe)); + } else if (intel_cdclk_needs_modeset(&dev_priv->cdclk.actual, + &state->cdclk.actual)) { + /* All pipes must be switched off while we change the cdclk. */ + ret = intel_modeset_all_pipes(state); + if (ret) + return ret; + + state->cdclk.pipe = INVALID_PIPE; + + DRM_DEBUG_KMS("Modeset required for cdclk change\n"); } + DRM_DEBUG_KMS("New cdclk calculated to be logical %u kHz, actual %u kHz\n", + state->cdclk.logical.cdclk, + state->cdclk.actual.cdclk); + DRM_DEBUG_KMS("New voltage level calculated to be logical %u, actual %u\n", + state->cdclk.logical.voltage_level, + state->cdclk.actual.voltage_level); + return 0; } @@ -2809,15 +2617,29 @@ void intel_update_rawclk(struct drm_i915_private *dev_priv) */ void intel_init_cdclk_hooks(struct drm_i915_private *dev_priv) { - if (INTEL_GEN(dev_priv) >= 11) { - dev_priv->display.set_cdclk = icl_set_cdclk; - dev_priv->display.modeset_calc_cdclk = icl_modeset_calc_cdclk; + if (IS_ELKHARTLAKE(dev_priv)) { + dev_priv->display.set_cdclk = bxt_set_cdclk; + dev_priv->display.modeset_calc_cdclk = bxt_modeset_calc_cdclk; + dev_priv->display.calc_voltage_level = ehl_calc_voltage_level; + dev_priv->cdclk.table = icl_cdclk_table; + } else if (INTEL_GEN(dev_priv) >= 11) { + dev_priv->display.set_cdclk = bxt_set_cdclk; + dev_priv->display.modeset_calc_cdclk = bxt_modeset_calc_cdclk; + dev_priv->display.calc_voltage_level = icl_calc_voltage_level; + dev_priv->cdclk.table = icl_cdclk_table; } else if (IS_CANNONLAKE(dev_priv)) { - dev_priv->display.set_cdclk = cnl_set_cdclk; - dev_priv->display.modeset_calc_cdclk = cnl_modeset_calc_cdclk; + dev_priv->display.set_cdclk = bxt_set_cdclk; + dev_priv->display.modeset_calc_cdclk = bxt_modeset_calc_cdclk; + dev_priv->display.calc_voltage_level = cnl_calc_voltage_level; + dev_priv->cdclk.table = cnl_cdclk_table; } else if (IS_GEN9_LP(dev_priv)) { dev_priv->display.set_cdclk = bxt_set_cdclk; dev_priv->display.modeset_calc_cdclk = bxt_modeset_calc_cdclk; + dev_priv->display.calc_voltage_level = bxt_calc_voltage_level; + if (IS_GEMINILAKE(dev_priv)) + dev_priv->cdclk.table = glk_cdclk_table; + else + dev_priv->cdclk.table = bxt_cdclk_table; } else if (IS_GEN9_BC(dev_priv)) { dev_priv->display.set_cdclk = skl_set_cdclk; dev_priv->display.modeset_calc_cdclk = skl_modeset_calc_cdclk; @@ -2830,13 +2652,11 @@ void intel_init_cdclk_hooks(struct drm_i915_private *dev_priv) } else if (IS_VALLEYVIEW(dev_priv)) { dev_priv->display.set_cdclk = vlv_set_cdclk; dev_priv->display.modeset_calc_cdclk = vlv_modeset_calc_cdclk; + } else { + dev_priv->display.modeset_calc_cdclk = fixed_modeset_calc_cdclk; } - if (INTEL_GEN(dev_priv) >= 11) - dev_priv->display.get_cdclk = icl_get_cdclk; - else if (IS_CANNONLAKE(dev_priv)) - dev_priv->display.get_cdclk = cnl_get_cdclk; - else if (IS_GEN9_LP(dev_priv)) + if (INTEL_GEN(dev_priv) >= 10 || IS_GEN9_LP(dev_priv)) dev_priv->display.get_cdclk = bxt_get_cdclk; else if (IS_GEN9_BC(dev_priv)) dev_priv->display.get_cdclk = skl_get_cdclk; diff --git a/drivers/gpu/drm/i915/display/intel_cdclk.h b/drivers/gpu/drm/i915/display/intel_cdclk.h index 4d6f7f5f8930..cf71394cc79c 100644 --- a/drivers/gpu/drm/i915/display/intel_cdclk.h +++ b/drivers/gpu/drm/i915/display/intel_cdclk.h @@ -15,6 +15,13 @@ struct intel_atomic_state; struct intel_cdclk_state; struct intel_crtc_state; +struct intel_cdclk_vals { + u16 refclk; + u32 cdclk; + u8 divider; /* CD2X divider * 2 */ + u8 ratio; +}; + int intel_crtc_compute_min_cdclk(const struct intel_crtc_state *crtc_state); void intel_cdclk_init(struct drm_i915_private *i915); void intel_cdclk_uninit(struct drm_i915_private *i915); @@ -22,13 +29,8 @@ void intel_init_cdclk_hooks(struct drm_i915_private *dev_priv); void intel_update_max_cdclk(struct drm_i915_private *dev_priv); void intel_update_cdclk(struct drm_i915_private *dev_priv); void intel_update_rawclk(struct drm_i915_private *dev_priv); -bool intel_cdclk_needs_cd2x_update(struct drm_i915_private *dev_priv, - const struct intel_cdclk_state *a, - const struct intel_cdclk_state *b); bool intel_cdclk_needs_modeset(const struct intel_cdclk_state *a, const struct intel_cdclk_state *b); -bool intel_cdclk_changed(const struct intel_cdclk_state *a, - const struct intel_cdclk_state *b); void intel_cdclk_swap_state(struct intel_atomic_state *state); void intel_set_cdclk_pre_plane_update(struct drm_i915_private *dev_priv, @@ -42,5 +44,6 @@ intel_set_cdclk_post_plane_update(struct drm_i915_private *dev_priv, enum pipe pipe); void intel_dump_cdclk_state(const struct intel_cdclk_state *cdclk_state, const char *context); +int intel_modeset_calc_cdclk(struct intel_atomic_state *state); #endif /* __INTEL_CDCLK_H__ */ diff --git a/drivers/gpu/drm/i915/display/intel_color.c b/drivers/gpu/drm/i915/display/intel_color.c index 71a0201437a9..aa3a063549c3 100644 --- a/drivers/gpu/drm/i915/display/intel_color.c +++ b/drivers/gpu/drm/i915/display/intel_color.c @@ -43,6 +43,21 @@ #define LEGACY_LUT_LENGTH 256 /* + * ILK+ csc matrix: + * + * |R/Cr| | c0 c1 c2 | ( |R/Cr| |preoff0| ) |postoff0| + * |G/Y | = | c3 c4 c5 | x ( |G/Y | + |preoff1| ) + |postoff1| + * |B/Cb| | c6 c7 c8 | ( |B/Cb| |preoff2| ) |postoff2| + * + * ILK/SNB don't have explicit post offsets, and instead + * CSC_MODE_YUV_TO_RGB and CSC_BLACK_SCREEN_OFFSET are used: + * CSC_MODE_YUV_TO_RGB=0 + CSC_BLACK_SCREEN_OFFSET=0 -> 1/2, 0, 1/2 + * CSC_MODE_YUV_TO_RGB=0 + CSC_BLACK_SCREEN_OFFSET=1 -> 1/2, 1/16, 1/2 + * CSC_MODE_YUV_TO_RGB=1 + CSC_BLACK_SCREEN_OFFSET=0 -> 0, 0, 0 + * CSC_MODE_YUV_TO_RGB=1 + CSC_BLACK_SCREEN_OFFSET=1 -> 1/16, 1/16, 1/16 + */ + +/* * Extract the CSC coefficient from a CTM coefficient (in U32.32 fixed point * format). This macro takes the coefficient we want transformed and the * number of fractional bits. @@ -59,37 +74,38 @@ #define ILK_CSC_POSTOFF_LIMITED_RANGE (16 * (1 << 12) / 255) +/* Nop pre/post offsets */ static const u16 ilk_csc_off_zero[3] = {}; +/* Identity matrix */ static const u16 ilk_csc_coeff_identity[9] = { ILK_CSC_COEFF_1_0, 0, 0, 0, ILK_CSC_COEFF_1_0, 0, 0, 0, ILK_CSC_COEFF_1_0, }; +/* Limited range RGB post offsets */ static const u16 ilk_csc_postoff_limited_range[3] = { ILK_CSC_POSTOFF_LIMITED_RANGE, ILK_CSC_POSTOFF_LIMITED_RANGE, ILK_CSC_POSTOFF_LIMITED_RANGE, }; +/* Full range RGB -> limited range RGB matrix */ static const u16 ilk_csc_coeff_limited_range[9] = { ILK_CSC_COEFF_LIMITED_RANGE, 0, 0, 0, ILK_CSC_COEFF_LIMITED_RANGE, 0, 0, 0, ILK_CSC_COEFF_LIMITED_RANGE, }; -/* - * These values are direct register values specified in the Bspec, - * for RGB->YUV conversion matrix (colorspace BT709) - */ +/* BT.709 full range RGB -> limited range YCbCr matrix */ static const u16 ilk_csc_coeff_rgb_to_ycbcr[9] = { 0x1e08, 0x9cc0, 0xb528, 0x2ba8, 0x09d8, 0x37e8, 0xbce8, 0x9ad8, 0x1e08, }; -/* Post offset values for RGB->YCBCR conversion */ +/* Limited range YCbCr post offsets */ static const u16 ilk_csc_postoff_rgb_to_ycbcr[3] = { 0x0800, 0x0100, 0x0800, }; @@ -611,12 +627,13 @@ static void bdw_load_lut_10(struct intel_crtc *crtc, static void ivb_load_lut_ext_max(struct intel_crtc *crtc) { struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + struct intel_dsb *dsb = intel_dsb_get(crtc); enum pipe pipe = crtc->pipe; /* Program the max register to clamp values > 1.0. */ - I915_WRITE(PREC_PAL_EXT_GC_MAX(pipe, 0), 1 << 16); - I915_WRITE(PREC_PAL_EXT_GC_MAX(pipe, 1), 1 << 16); - I915_WRITE(PREC_PAL_EXT_GC_MAX(pipe, 2), 1 << 16); + intel_dsb_reg_write(dsb, PREC_PAL_EXT_GC_MAX(pipe, 0), 1 << 16); + intel_dsb_reg_write(dsb, PREC_PAL_EXT_GC_MAX(pipe, 1), 1 << 16); + intel_dsb_reg_write(dsb, PREC_PAL_EXT_GC_MAX(pipe, 2), 1 << 16); /* * Program the gc max 2 register to clamp values > 1.0. @@ -624,10 +641,15 @@ static void ivb_load_lut_ext_max(struct intel_crtc *crtc) * from 3.0 to 7.0 */ if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) { - I915_WRITE(PREC_PAL_EXT2_GC_MAX(pipe, 0), 1 << 16); - I915_WRITE(PREC_PAL_EXT2_GC_MAX(pipe, 1), 1 << 16); - I915_WRITE(PREC_PAL_EXT2_GC_MAX(pipe, 2), 1 << 16); + intel_dsb_reg_write(dsb, PREC_PAL_EXT2_GC_MAX(pipe, 0), + 1 << 16); + intel_dsb_reg_write(dsb, PREC_PAL_EXT2_GC_MAX(pipe, 1), + 1 << 16); + intel_dsb_reg_write(dsb, PREC_PAL_EXT2_GC_MAX(pipe, 2), + 1 << 16); } + + intel_dsb_put(dsb); } static void ivb_load_luts(const struct intel_crtc_state *crtc_state) @@ -787,78 +809,83 @@ icl_load_gcmax(const struct intel_crtc_state *crtc_state, const struct drm_color_lut *color) { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); - struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + struct intel_dsb *dsb = intel_dsb_get(crtc); enum pipe pipe = crtc->pipe; /* Fixme: LUT entries are 16 bit only, so we can prog 0xFFFF max */ - I915_WRITE(PREC_PAL_GC_MAX(pipe, 0), color->red); - I915_WRITE(PREC_PAL_GC_MAX(pipe, 1), color->green); - I915_WRITE(PREC_PAL_GC_MAX(pipe, 2), color->blue); + intel_dsb_reg_write(dsb, PREC_PAL_GC_MAX(pipe, 0), color->red); + intel_dsb_reg_write(dsb, PREC_PAL_GC_MAX(pipe, 1), color->green); + intel_dsb_reg_write(dsb, PREC_PAL_GC_MAX(pipe, 2), color->blue); + intel_dsb_put(dsb); } static void icl_program_gamma_superfine_segment(const struct intel_crtc_state *crtc_state) { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); - struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); const struct drm_property_blob *blob = crtc_state->base.gamma_lut; const struct drm_color_lut *lut = blob->data; + struct intel_dsb *dsb = intel_dsb_get(crtc); enum pipe pipe = crtc->pipe; u32 i; /* - * Every entry in the multi-segment LUT is corresponding to a superfine - * segment step which is 1/(8 * 128 * 256). + * Program Super Fine segment (let's call it seg1)... * - * Superfine segment has 9 entries, corresponding to values - * 0, 1/(8 * 128 * 256), 2/(8 * 128 * 256) .... 8/(8 * 128 * 256). + * Super Fine segment's step is 1/(8 * 128 * 256) and it has + * 9 entries, corresponding to values 0, 1/(8 * 128 * 256), + * 2/(8 * 128 * 256) ... 8/(8 * 128 * 256). */ - I915_WRITE(PREC_PAL_MULTI_SEG_INDEX(pipe), PAL_PREC_AUTO_INCREMENT); + intel_dsb_reg_write(dsb, PREC_PAL_MULTI_SEG_INDEX(pipe), + PAL_PREC_AUTO_INCREMENT); for (i = 0; i < 9; i++) { const struct drm_color_lut *entry = &lut[i]; - I915_WRITE(PREC_PAL_MULTI_SEG_DATA(pipe), - ilk_lut_12p4_ldw(entry)); - I915_WRITE(PREC_PAL_MULTI_SEG_DATA(pipe), - ilk_lut_12p4_udw(entry)); + intel_dsb_indexed_reg_write(dsb, PREC_PAL_MULTI_SEG_DATA(pipe), + ilk_lut_12p4_ldw(entry)); + intel_dsb_indexed_reg_write(dsb, PREC_PAL_MULTI_SEG_DATA(pipe), + ilk_lut_12p4_udw(entry)); } + + intel_dsb_put(dsb); } static void icl_program_gamma_multi_segment(const struct intel_crtc_state *crtc_state) { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); - struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); const struct drm_property_blob *blob = crtc_state->base.gamma_lut; const struct drm_color_lut *lut = blob->data; const struct drm_color_lut *entry; + struct intel_dsb *dsb = intel_dsb_get(crtc); enum pipe pipe = crtc->pipe; u32 i; /* - * * Program Fine segment (let's call it seg2)... * - * Fine segment's step is 1/(128 * 256) ie 1/(128 * 256), 2/(128*256) - * ... 256/(128*256). So in order to program fine segment of LUT we - * need to pick every 8'th entry in LUT, and program 256 indexes. + * Fine segment's step is 1/(128 * 256) i.e. 1/(128 * 256), 2/(128 * 256) + * ... 256/(128 * 256). So in order to program fine segment of LUT we + * need to pick every 8th entry in the LUT, and program 256 indexes. * * PAL_PREC_INDEX[0] and PAL_PREC_INDEX[1] map to seg2[1], - * with seg2[0] being unused by the hardware. + * seg2[0] being unused by the hardware. */ - I915_WRITE(PREC_PAL_INDEX(pipe), PAL_PREC_AUTO_INCREMENT); + intel_dsb_reg_write(dsb, PREC_PAL_INDEX(pipe), PAL_PREC_AUTO_INCREMENT); for (i = 1; i < 257; i++) { entry = &lut[i * 8]; - I915_WRITE(PREC_PAL_DATA(pipe), ilk_lut_12p4_ldw(entry)); - I915_WRITE(PREC_PAL_DATA(pipe), ilk_lut_12p4_udw(entry)); + intel_dsb_indexed_reg_write(dsb, PREC_PAL_DATA(pipe), + ilk_lut_12p4_ldw(entry)); + intel_dsb_indexed_reg_write(dsb, PREC_PAL_DATA(pipe), + ilk_lut_12p4_udw(entry)); } /* * Program Coarse segment (let's call it seg3)... * - * Coarse segment's starts from index 0 and it's step is 1/256 ie 0, - * 1/256, 2/256 ...256/256. As per the description of each entry in LUT + * Coarse segment starts from index 0 and it's step is 1/256 ie 0, + * 1/256, 2/256 ... 256/256. As per the description of each entry in LUT * above, we need to pick every (8 * 128)th entry in LUT, and * program 256 of those. * @@ -868,20 +895,24 @@ icl_program_gamma_multi_segment(const struct intel_crtc_state *crtc_state) */ for (i = 0; i < 256; i++) { entry = &lut[i * 8 * 128]; - I915_WRITE(PREC_PAL_DATA(pipe), ilk_lut_12p4_ldw(entry)); - I915_WRITE(PREC_PAL_DATA(pipe), ilk_lut_12p4_udw(entry)); + intel_dsb_indexed_reg_write(dsb, PREC_PAL_DATA(pipe), + ilk_lut_12p4_ldw(entry)); + intel_dsb_indexed_reg_write(dsb, PREC_PAL_DATA(pipe), + ilk_lut_12p4_udw(entry)); } /* The last entry in the LUT is to be programmed in GCMAX */ entry = &lut[256 * 8 * 128]; icl_load_gcmax(crtc_state, entry); ivb_load_lut_ext_max(crtc); + intel_dsb_put(dsb); } static void icl_load_luts(const struct intel_crtc_state *crtc_state) { const struct drm_property_blob *gamma_lut = crtc_state->base.gamma_lut; struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct intel_dsb *dsb = intel_dsb_get(crtc); if (crtc_state->base.degamma_lut) glk_load_degamma_lut(crtc_state); @@ -890,16 +921,17 @@ static void icl_load_luts(const struct intel_crtc_state *crtc_state) case GAMMA_MODE_MODE_8BIT: i9xx_load_luts(crtc_state); break; - case GAMMA_MODE_MODE_12BIT_MULTI_SEGMENTED: icl_program_gamma_superfine_segment(crtc_state); icl_program_gamma_multi_segment(crtc_state); break; - default: bdw_load_lut_10(crtc, gamma_lut, PAL_PREC_INDEX_VALUE(0)); ivb_load_lut_ext_max(crtc); } + + intel_dsb_commit(dsb); + intel_dsb_put(dsb); } static u32 chv_cgm_degamma_ldw(const struct drm_color_lut *color) @@ -990,6 +1022,55 @@ void intel_color_commit(const struct intel_crtc_state *crtc_state) dev_priv->display.color_commit(crtc_state); } +static bool intel_can_preload_luts(const struct intel_crtc_state *new_crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(new_crtc_state->base.crtc); + struct intel_atomic_state *state = + to_intel_atomic_state(new_crtc_state->base.state); + const struct intel_crtc_state *old_crtc_state = + intel_atomic_get_old_crtc_state(state, crtc); + + return !old_crtc_state->base.gamma_lut && + !old_crtc_state->base.degamma_lut; +} + +static bool chv_can_preload_luts(const struct intel_crtc_state *new_crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(new_crtc_state->base.crtc); + struct intel_atomic_state *state = + to_intel_atomic_state(new_crtc_state->base.state); + const struct intel_crtc_state *old_crtc_state = + intel_atomic_get_old_crtc_state(state, crtc); + + /* + * CGM_PIPE_MODE is itself single buffered. We'd have to + * somehow split it out from chv_load_luts() if we wanted + * the ability to preload the CGM LUTs/CSC without tearing. + */ + if (old_crtc_state->cgm_mode || new_crtc_state->cgm_mode) + return false; + + return !old_crtc_state->base.gamma_lut; +} + +static bool glk_can_preload_luts(const struct intel_crtc_state *new_crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(new_crtc_state->base.crtc); + struct intel_atomic_state *state = + to_intel_atomic_state(new_crtc_state->base.state); + const struct intel_crtc_state *old_crtc_state = + intel_atomic_get_old_crtc_state(state, crtc); + + /* + * The hardware degamma is active whenever the pipe + * CSC is active. Thus even if the old state has no + * software degamma we need to avoid clobbering the + * linear hardware degamma mid scanout. + */ + return !old_crtc_state->csc_enable && + !old_crtc_state->base.gamma_lut; +} + int intel_color_check(struct intel_crtc_state *crtc_state) { struct drm_i915_private *dev_priv = to_i915(crtc_state->base.crtc->dev); @@ -1133,6 +1214,8 @@ static int i9xx_color_check(struct intel_crtc_state *crtc_state) if (ret) return ret; + crtc_state->preload_luts = intel_can_preload_luts(crtc_state); + return 0; } @@ -1185,6 +1268,8 @@ static int chv_color_check(struct intel_crtc_state *crtc_state) if (ret) return ret; + crtc_state->preload_luts = chv_can_preload_luts(crtc_state); + return 0; } @@ -1197,6 +1282,21 @@ static u32 ilk_gamma_mode(const struct intel_crtc_state *crtc_state) return GAMMA_MODE_MODE_10BIT; } +static u32 ilk_csc_mode(const struct intel_crtc_state *crtc_state) +{ + /* + * CSC comes after the LUT in RGB->YCbCr mode. + * RGB->YCbCr needs the limited range offsets added to + * the output. RGB limited range output is handled by + * the hw automagically elsewhere. + */ + if (crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB) + return CSC_BLACK_SCREEN_OFFSET; + + return CSC_MODE_YUV_TO_RGB | + CSC_POSITION_BEFORE_GAMMA; +} + static int ilk_color_check(struct intel_crtc_state *crtc_state) { int ret; @@ -1210,20 +1310,22 @@ static int ilk_color_check(struct intel_crtc_state *crtc_state) !crtc_state->c8_planes; /* - * We don't expose the ctm on ilk/snb currently, - * nor do we enable YCbCr output. Also RGB limited - * range output is handled by the hw automagically. + * We don't expose the ctm on ilk/snb currently, also RGB + * limited range output is handled by the hw automagically. */ - crtc_state->csc_enable = false; + crtc_state->csc_enable = + crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB; crtc_state->gamma_mode = ilk_gamma_mode(crtc_state); - crtc_state->csc_mode = 0; + crtc_state->csc_mode = ilk_csc_mode(crtc_state); ret = intel_color_add_affected_planes(crtc_state); if (ret) return ret; + crtc_state->preload_luts = intel_can_preload_luts(crtc_state); + return 0; } @@ -1281,6 +1383,8 @@ static int ivb_color_check(struct intel_crtc_state *crtc_state) if (ret) return ret; + crtc_state->preload_luts = intel_can_preload_luts(crtc_state); + return 0; } @@ -1319,6 +1423,8 @@ static int glk_color_check(struct intel_crtc_state *crtc_state) if (ret) return ret; + crtc_state->preload_luts = glk_can_preload_luts(crtc_state); + return 0; } @@ -1368,9 +1474,408 @@ static int icl_color_check(struct intel_crtc_state *crtc_state) crtc_state->csc_mode = icl_csc_mode(crtc_state); + crtc_state->preload_luts = intel_can_preload_luts(crtc_state); + return 0; } +static int i9xx_gamma_precision(const struct intel_crtc_state *crtc_state) +{ + if (!crtc_state->gamma_enable) + return 0; + + switch (crtc_state->gamma_mode) { + case GAMMA_MODE_MODE_8BIT: + return 8; + case GAMMA_MODE_MODE_10BIT: + return 16; + default: + MISSING_CASE(crtc_state->gamma_mode); + return 0; + } +} + +static int ilk_gamma_precision(const struct intel_crtc_state *crtc_state) +{ + if (!crtc_state->gamma_enable) + return 0; + + if ((crtc_state->csc_mode & CSC_POSITION_BEFORE_GAMMA) == 0) + return 0; + + switch (crtc_state->gamma_mode) { + case GAMMA_MODE_MODE_8BIT: + return 8; + case GAMMA_MODE_MODE_10BIT: + return 10; + default: + MISSING_CASE(crtc_state->gamma_mode); + return 0; + } +} + +static int chv_gamma_precision(const struct intel_crtc_state *crtc_state) +{ + if (crtc_state->cgm_mode & CGM_PIPE_MODE_GAMMA) + return 10; + else + return i9xx_gamma_precision(crtc_state); +} + +static int glk_gamma_precision(const struct intel_crtc_state *crtc_state) +{ + if (!crtc_state->gamma_enable) + return 0; + + switch (crtc_state->gamma_mode) { + case GAMMA_MODE_MODE_8BIT: + return 8; + case GAMMA_MODE_MODE_10BIT: + return 10; + default: + MISSING_CASE(crtc_state->gamma_mode); + return 0; + } +} + +int intel_color_get_gamma_bit_precision(const struct intel_crtc_state *crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + + if (HAS_GMCH(dev_priv)) { + if (IS_CHERRYVIEW(dev_priv)) + return chv_gamma_precision(crtc_state); + else + return i9xx_gamma_precision(crtc_state); + } else { + if (IS_CANNONLAKE(dev_priv) || IS_GEMINILAKE(dev_priv)) + return glk_gamma_precision(crtc_state); + else if (IS_IRONLAKE(dev_priv)) + return ilk_gamma_precision(crtc_state); + } + + return 0; +} + +static bool err_check(struct drm_color_lut *lut1, + struct drm_color_lut *lut2, u32 err) +{ + return ((abs((long)lut2->red - lut1->red)) <= err) && + ((abs((long)lut2->blue - lut1->blue)) <= err) && + ((abs((long)lut2->green - lut1->green)) <= err); +} + +static bool intel_color_lut_entry_equal(struct drm_color_lut *lut1, + struct drm_color_lut *lut2, + int lut_size, u32 err) +{ + int i; + + for (i = 0; i < lut_size; i++) { + if (!err_check(&lut1[i], &lut2[i], err)) + return false; + } + + return true; +} + +bool intel_color_lut_equal(struct drm_property_blob *blob1, + struct drm_property_blob *blob2, + u32 gamma_mode, u32 bit_precision) +{ + struct drm_color_lut *lut1, *lut2; + int lut_size1, lut_size2; + u32 err; + + if (!blob1 != !blob2) + return false; + + if (!blob1) + return true; + + lut_size1 = drm_color_lut_size(blob1); + lut_size2 = drm_color_lut_size(blob2); + + /* check sw and hw lut size */ + switch (gamma_mode) { + case GAMMA_MODE_MODE_8BIT: + case GAMMA_MODE_MODE_10BIT: + if (lut_size1 != lut_size2) + return false; + break; + default: + MISSING_CASE(gamma_mode); + return false; + } + + lut1 = blob1->data; + lut2 = blob2->data; + + err = 0xffff >> bit_precision; + + /* check sw and hw lut entry to be equal */ + switch (gamma_mode) { + case GAMMA_MODE_MODE_8BIT: + case GAMMA_MODE_MODE_10BIT: + if (!intel_color_lut_entry_equal(lut1, lut2, + lut_size2, err)) + return false; + break; + default: + MISSING_CASE(gamma_mode); + return false; + } + + return true; +} + +/* convert hw value with given bit_precision to lut property val */ +static u32 intel_color_lut_pack(u32 val, u32 bit_precision) +{ + u32 max = 0xffff >> (16 - bit_precision); + + val = clamp_val(val, 0, max); + + if (bit_precision < 16) + val <<= 16 - bit_precision; + + return val; +} + +static struct drm_property_blob * +i9xx_read_lut_8(const struct intel_crtc_state *crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + enum pipe pipe = crtc->pipe; + struct drm_property_blob *blob; + struct drm_color_lut *blob_data; + u32 i, val; + + blob = drm_property_create_blob(&dev_priv->drm, + sizeof(struct drm_color_lut) * LEGACY_LUT_LENGTH, + NULL); + if (IS_ERR(blob)) + return NULL; + + blob_data = blob->data; + + for (i = 0; i < LEGACY_LUT_LENGTH; i++) { + if (HAS_GMCH(dev_priv)) + val = I915_READ(PALETTE(pipe, i)); + else + val = I915_READ(LGC_PALETTE(pipe, i)); + + blob_data[i].red = intel_color_lut_pack(REG_FIELD_GET( + LGC_PALETTE_RED_MASK, val), 8); + blob_data[i].green = intel_color_lut_pack(REG_FIELD_GET( + LGC_PALETTE_GREEN_MASK, val), 8); + blob_data[i].blue = intel_color_lut_pack(REG_FIELD_GET( + LGC_PALETTE_BLUE_MASK, val), 8); + } + + return blob; +} + +static void i9xx_read_luts(struct intel_crtc_state *crtc_state) +{ + if (!crtc_state->gamma_enable) + return; + + crtc_state->base.gamma_lut = i9xx_read_lut_8(crtc_state); +} + +static struct drm_property_blob * +i965_read_lut_10p6(const struct intel_crtc_state *crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + u32 lut_size = INTEL_INFO(dev_priv)->color.gamma_lut_size; + enum pipe pipe = crtc->pipe; + struct drm_property_blob *blob; + struct drm_color_lut *blob_data; + u32 i, val1, val2; + + blob = drm_property_create_blob(&dev_priv->drm, + sizeof(struct drm_color_lut) * lut_size, + NULL); + if (IS_ERR(blob)) + return NULL; + + blob_data = blob->data; + + for (i = 0; i < lut_size - 1; i++) { + val1 = I915_READ(PALETTE(pipe, 2 * i + 0)); + val2 = I915_READ(PALETTE(pipe, 2 * i + 1)); + + blob_data[i].red = REG_FIELD_GET(PALETTE_RED_MASK, val2) << 8 | + REG_FIELD_GET(PALETTE_RED_MASK, val1); + blob_data[i].green = REG_FIELD_GET(PALETTE_GREEN_MASK, val2) << 8 | + REG_FIELD_GET(PALETTE_GREEN_MASK, val1); + blob_data[i].blue = REG_FIELD_GET(PALETTE_BLUE_MASK, val2) << 8 | + REG_FIELD_GET(PALETTE_BLUE_MASK, val1); + } + + blob_data[i].red = REG_FIELD_GET(PIPEGCMAX_RGB_MASK, + I915_READ(PIPEGCMAX(pipe, 0))); + blob_data[i].green = REG_FIELD_GET(PIPEGCMAX_RGB_MASK, + I915_READ(PIPEGCMAX(pipe, 1))); + blob_data[i].blue = REG_FIELD_GET(PIPEGCMAX_RGB_MASK, + I915_READ(PIPEGCMAX(pipe, 2))); + + return blob; +} + +static void i965_read_luts(struct intel_crtc_state *crtc_state) +{ + if (!crtc_state->gamma_enable) + return; + + if (crtc_state->gamma_mode == GAMMA_MODE_MODE_8BIT) + crtc_state->base.gamma_lut = i9xx_read_lut_8(crtc_state); + else + crtc_state->base.gamma_lut = i965_read_lut_10p6(crtc_state); +} + +static struct drm_property_blob * +chv_read_cgm_lut(const struct intel_crtc_state *crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + u32 lut_size = INTEL_INFO(dev_priv)->color.gamma_lut_size; + enum pipe pipe = crtc->pipe; + struct drm_property_blob *blob; + struct drm_color_lut *blob_data; + u32 i, val; + + blob = drm_property_create_blob(&dev_priv->drm, + sizeof(struct drm_color_lut) * lut_size, + NULL); + if (IS_ERR(blob)) + return NULL; + + blob_data = blob->data; + + for (i = 0; i < lut_size; i++) { + val = I915_READ(CGM_PIPE_GAMMA(pipe, i, 0)); + blob_data[i].green = intel_color_lut_pack(REG_FIELD_GET( + CGM_PIPE_GAMMA_GREEN_MASK, val), 10); + blob_data[i].blue = intel_color_lut_pack(REG_FIELD_GET( + CGM_PIPE_GAMMA_BLUE_MASK, val), 10); + + val = I915_READ(CGM_PIPE_GAMMA(pipe, i, 1)); + blob_data[i].red = intel_color_lut_pack(REG_FIELD_GET( + CGM_PIPE_GAMMA_RED_MASK, val), 10); + } + + return blob; +} + +static void chv_read_luts(struct intel_crtc_state *crtc_state) +{ + if (crtc_state->cgm_mode & CGM_PIPE_MODE_GAMMA) + crtc_state->base.gamma_lut = chv_read_cgm_lut(crtc_state); + else + i965_read_luts(crtc_state); +} + +static struct drm_property_blob * +ilk_read_lut_10(const struct intel_crtc_state *crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + u32 lut_size = INTEL_INFO(dev_priv)->color.gamma_lut_size; + enum pipe pipe = crtc->pipe; + struct drm_property_blob *blob; + struct drm_color_lut *blob_data; + u32 i, val; + + blob = drm_property_create_blob(&dev_priv->drm, + sizeof(struct drm_color_lut) * lut_size, + NULL); + if (IS_ERR(blob)) + return NULL; + + blob_data = blob->data; + + for (i = 0; i < lut_size; i++) { + val = I915_READ(PREC_PALETTE(pipe, i)); + + blob_data[i].red = intel_color_lut_pack(REG_FIELD_GET( + PREC_PALETTE_RED_MASK, val), 10); + blob_data[i].green = intel_color_lut_pack(REG_FIELD_GET( + PREC_PALETTE_GREEN_MASK, val), 10); + blob_data[i].blue = intel_color_lut_pack(REG_FIELD_GET( + PREC_PALETTE_BLUE_MASK, val), 10); + } + + return blob; +} + +static void ilk_read_luts(struct intel_crtc_state *crtc_state) +{ + if (!crtc_state->gamma_enable) + return; + + if ((crtc_state->csc_mode & CSC_POSITION_BEFORE_GAMMA) == 0) + return; + + if (crtc_state->gamma_mode == GAMMA_MODE_MODE_8BIT) + crtc_state->base.gamma_lut = i9xx_read_lut_8(crtc_state); + else + crtc_state->base.gamma_lut = ilk_read_lut_10(crtc_state); +} + +static struct drm_property_blob * +glk_read_lut_10(const struct intel_crtc_state *crtc_state, u32 prec_index) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + int hw_lut_size = ivb_lut_10_size(prec_index); + enum pipe pipe = crtc->pipe; + struct drm_property_blob *blob; + struct drm_color_lut *blob_data; + u32 i, val; + + blob = drm_property_create_blob(&dev_priv->drm, + sizeof(struct drm_color_lut) * hw_lut_size, + NULL); + if (IS_ERR(blob)) + return NULL; + + blob_data = blob->data; + + I915_WRITE(PREC_PAL_INDEX(pipe), prec_index | + PAL_PREC_AUTO_INCREMENT); + + for (i = 0; i < hw_lut_size; i++) { + val = I915_READ(PREC_PAL_DATA(pipe)); + + blob_data[i].red = intel_color_lut_pack(REG_FIELD_GET( + PREC_PAL_DATA_RED_MASK, val), 10); + blob_data[i].green = intel_color_lut_pack(REG_FIELD_GET( + PREC_PAL_DATA_GREEN_MASK, val), 10); + blob_data[i].blue = intel_color_lut_pack(REG_FIELD_GET( + PREC_PAL_DATA_BLUE_MASK, val), 10); + } + + I915_WRITE(PREC_PAL_INDEX(pipe), 0); + + return blob; +} + +static void glk_read_luts(struct intel_crtc_state *crtc_state) +{ + if (!crtc_state->gamma_enable) + return; + + if (crtc_state->gamma_mode == GAMMA_MODE_MODE_8BIT) + crtc_state->base.gamma_lut = i9xx_read_lut_8(crtc_state); + else + crtc_state->base.gamma_lut = glk_read_lut_10(crtc_state, PAL_PREC_INDEX_VALUE(0)); +} + void intel_color_init(struct intel_crtc *crtc) { struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); @@ -1383,14 +1888,17 @@ void intel_color_init(struct intel_crtc *crtc) dev_priv->display.color_check = chv_color_check; dev_priv->display.color_commit = i9xx_color_commit; dev_priv->display.load_luts = chv_load_luts; + dev_priv->display.read_luts = chv_read_luts; } else if (INTEL_GEN(dev_priv) >= 4) { dev_priv->display.color_check = i9xx_color_check; dev_priv->display.color_commit = i9xx_color_commit; dev_priv->display.load_luts = i965_load_luts; + dev_priv->display.read_luts = i965_read_luts; } else { dev_priv->display.color_check = i9xx_color_check; dev_priv->display.color_commit = i9xx_color_commit; dev_priv->display.load_luts = i9xx_load_luts; + dev_priv->display.read_luts = i9xx_read_luts; } } else { if (INTEL_GEN(dev_priv) >= 11) @@ -1409,16 +1917,19 @@ void intel_color_init(struct intel_crtc *crtc) else dev_priv->display.color_commit = ilk_color_commit; - if (INTEL_GEN(dev_priv) >= 11) + if (INTEL_GEN(dev_priv) >= 11) { dev_priv->display.load_luts = icl_load_luts; - else if (IS_CANNONLAKE(dev_priv) || IS_GEMINILAKE(dev_priv)) + } else if (IS_CANNONLAKE(dev_priv) || IS_GEMINILAKE(dev_priv)) { dev_priv->display.load_luts = glk_load_luts; - else if (INTEL_GEN(dev_priv) >= 8) + dev_priv->display.read_luts = glk_read_luts; + } else if (INTEL_GEN(dev_priv) >= 8) { dev_priv->display.load_luts = bdw_load_luts; - else if (INTEL_GEN(dev_priv) >= 7) + } else if (INTEL_GEN(dev_priv) >= 7) { dev_priv->display.load_luts = ivb_load_luts; - else + } else { dev_priv->display.load_luts = ilk_load_luts; + dev_priv->display.read_luts = ilk_read_luts; + } } drm_crtc_enable_color_mgmt(&crtc->base, diff --git a/drivers/gpu/drm/i915/display/intel_color.h b/drivers/gpu/drm/i915/display/intel_color.h index 057e8ac63555..173727aaa24d 100644 --- a/drivers/gpu/drm/i915/display/intel_color.h +++ b/drivers/gpu/drm/i915/display/intel_color.h @@ -6,13 +6,20 @@ #ifndef __INTEL_COLOR_H__ #define __INTEL_COLOR_H__ +#include <linux/types.h> + struct intel_crtc_state; struct intel_crtc; +struct drm_property_blob; void intel_color_init(struct intel_crtc *crtc); int intel_color_check(struct intel_crtc_state *crtc_state); void intel_color_commit(const struct intel_crtc_state *crtc_state); void intel_color_load_luts(const struct intel_crtc_state *crtc_state); void intel_color_get_config(struct intel_crtc_state *crtc_state); +int intel_color_get_gamma_bit_precision(const struct intel_crtc_state *crtc_state); +bool intel_color_lut_equal(struct drm_property_blob *blob1, + struct drm_property_blob *blob2, + u32 gamma_mode, u32 bit_precision); #endif /* __INTEL_COLOR_H__ */ diff --git a/drivers/gpu/drm/i915/display/intel_connector.c b/drivers/gpu/drm/i915/display/intel_connector.c index 308ec63207ee..1133c4e97bb4 100644 --- a/drivers/gpu/drm/i915/display/intel_connector.c +++ b/drivers/gpu/drm/i915/display/intel_connector.c @@ -277,7 +277,22 @@ intel_attach_aspect_ratio_property(struct drm_connector *connector) void intel_attach_colorspace_property(struct drm_connector *connector) { - if (!drm_mode_create_colorspace_property(connector)) - drm_object_attach_property(&connector->base, - connector->colorspace_property, 0); + switch (connector->connector_type) { + case DRM_MODE_CONNECTOR_HDMIA: + case DRM_MODE_CONNECTOR_HDMIB: + if (drm_mode_create_hdmi_colorspace_property(connector)) + return; + break; + case DRM_MODE_CONNECTOR_DisplayPort: + case DRM_MODE_CONNECTOR_eDP: + if (drm_mode_create_dp_colorspace_property(connector)) + return; + break; + default: + DRM_DEBUG_KMS("Colorspace property not supported\n"); + return; + } + + drm_object_attach_property(&connector->base, + connector->colorspace_property, 0); } diff --git a/drivers/gpu/drm/i915/display/intel_crt.c b/drivers/gpu/drm/i915/display/intel_crt.c index e6e8d4a82044..39cc6d79dc85 100644 --- a/drivers/gpu/drm/i915/display/intel_crt.c +++ b/drivers/gpu/drm/i915/display/intel_crt.c @@ -844,7 +844,7 @@ load_detect: } /* for pre-945g platforms use load detect */ - ret = intel_get_load_detect_pipe(connector, NULL, &tmp, ctx); + ret = intel_get_load_detect_pipe(connector, &tmp, ctx); if (ret > 0) { if (intel_crt_detect_ddc(connector)) status = connector_status_connected; @@ -864,6 +864,13 @@ load_detect: out: intel_display_power_put(dev_priv, intel_encoder->power_domain, wakeref); + + /* + * Make sure the refs for power wells enabled during detect are + * dropped to avoid a new detect cycle triggered by HPD polling. + */ + intel_display_power_flush_work(dev_priv); + return status; } @@ -994,9 +1001,9 @@ void intel_crt_init(struct drm_i915_private *dev_priv) crt->base.type = INTEL_OUTPUT_ANALOG; crt->base.cloneable = (1 << INTEL_OUTPUT_DVO) | (1 << INTEL_OUTPUT_HDMI); if (IS_I830(dev_priv)) - crt->base.crtc_mask = (1 << 0); + crt->base.pipe_mask = BIT(PIPE_A); else - crt->base.crtc_mask = (1 << 0) | (1 << 1) | (1 << 2); + crt->base.pipe_mask = ~0; if (IS_GEN(dev_priv, 2)) connector->interlace_allowed = 0; diff --git a/drivers/gpu/drm/i915/display/intel_ddi.c b/drivers/gpu/drm/i915/display/intel_ddi.c index 8eb2b3ec01ed..c7c2b349858d 100644 --- a/drivers/gpu/drm/i915/display/intel_ddi.c +++ b/drivers/gpu/drm/i915/display/intel_ddi.c @@ -45,6 +45,7 @@ #include "intel_lspcon.h" #include "intel_panel.h" #include "intel_psr.h" +#include "intel_sprite.h" #include "intel_tc.h" #include "intel_vdsc.h" @@ -586,6 +587,40 @@ static const struct icl_mg_phy_ddi_buf_trans icl_mg_phy_ddi_translations[] = { { 0x0, 0x00, 0x00 }, /* 3 0 */ }; +struct tgl_dkl_phy_ddi_buf_trans { + u32 dkl_vswing_control; + u32 dkl_preshoot_control; + u32 dkl_de_emphasis_control; +}; + +static const struct tgl_dkl_phy_ddi_buf_trans tgl_dkl_phy_dp_ddi_trans[] = { + /* VS pre-emp Non-trans mV Pre-emph dB */ + { 0x7, 0x0, 0x00 }, /* 0 0 400mV 0 dB */ + { 0x5, 0x0, 0x03 }, /* 0 1 400mV 3.5 dB */ + { 0x2, 0x0, 0x0b }, /* 0 2 400mV 6 dB */ + { 0x0, 0x0, 0x19 }, /* 0 3 400mV 9.5 dB */ + { 0x5, 0x0, 0x00 }, /* 1 0 600mV 0 dB */ + { 0x2, 0x0, 0x03 }, /* 1 1 600mV 3.5 dB */ + { 0x0, 0x0, 0x14 }, /* 1 2 600mV 6 dB */ + { 0x2, 0x0, 0x00 }, /* 2 0 800mV 0 dB */ + { 0x0, 0x0, 0x0B }, /* 2 1 800mV 3.5 dB */ + { 0x0, 0x0, 0x00 }, /* 3 0 1200mV 0 dB HDMI default */ +}; + +static const struct tgl_dkl_phy_ddi_buf_trans tgl_dkl_phy_hdmi_ddi_trans[] = { + /* HDMI Preset VS Pre-emph */ + { 0x7, 0x0, 0x0 }, /* 1 400mV 0dB */ + { 0x6, 0x0, 0x0 }, /* 2 500mV 0dB */ + { 0x4, 0x0, 0x0 }, /* 3 650mV 0dB */ + { 0x2, 0x0, 0x0 }, /* 4 800mV 0dB */ + { 0x0, 0x0, 0x0 }, /* 5 1000mV 0dB */ + { 0x0, 0x0, 0x5 }, /* 6 Full -1.5 dB */ + { 0x0, 0x0, 0x6 }, /* 7 Full -1.8 dB */ + { 0x0, 0x0, 0x7 }, /* 8 Full -2 dB */ + { 0x0, 0x0, 0x8 }, /* 9 Full -2.5 dB */ + { 0x0, 0x0, 0xA }, /* 10 Full -3 dB */ +}; + static const struct ddi_buf_trans * bdw_get_buf_trans_edp(struct drm_i915_private *dev_priv, int *n_entries) { @@ -872,7 +907,14 @@ static int intel_ddi_hdmi_level(struct drm_i915_private *dev_priv, enum port por level = dev_priv->vbt.ddi_port_info[port].hdmi_level_shift; - if (INTEL_GEN(dev_priv) >= 11) { + if (INTEL_GEN(dev_priv) >= 12) { + if (intel_phy_is_combo(dev_priv, phy)) + icl_get_combo_buf_trans(dev_priv, INTEL_OUTPUT_HDMI, + 0, &n_entries); + else + n_entries = ARRAY_SIZE(tgl_dkl_phy_hdmi_ddi_trans); + default_entry = n_entries - 1; + } else if (INTEL_GEN(dev_priv) == 11) { if (intel_phy_is_combo(dev_priv, phy)) icl_get_combo_buf_trans(dev_priv, INTEL_OUTPUT_HDMI, 0, &n_entries); @@ -1049,6 +1091,8 @@ static u32 icl_pll_to_ddi_clk_sel(struct intel_encoder *encoder, case DPLL_ID_ICL_MGPLL2: case DPLL_ID_ICL_MGPLL3: case DPLL_ID_ICL_MGPLL4: + case DPLL_ID_TGL_MGPLL5: + case DPLL_ID_TGL_MGPLL6: return DDI_CLK_SEL_MG; } } @@ -1413,11 +1457,30 @@ static int icl_calc_mg_pll_link(struct drm_i915_private *dev_priv, ref_clock = dev_priv->cdclk.hw.ref; - m1 = pll_state->mg_pll_div1 & MG_PLL_DIV1_FBPREDIV_MASK; - m2_int = pll_state->mg_pll_div0 & MG_PLL_DIV0_FBDIV_INT_MASK; - m2_frac = (pll_state->mg_pll_div0 & MG_PLL_DIV0_FRACNEN_H) ? - (pll_state->mg_pll_div0 & MG_PLL_DIV0_FBDIV_FRAC_MASK) >> - MG_PLL_DIV0_FBDIV_FRAC_SHIFT : 0; + if (INTEL_GEN(dev_priv) >= 12) { + m1 = pll_state->mg_pll_div0 & DKL_PLL_DIV0_FBPREDIV_MASK; + m1 = m1 >> DKL_PLL_DIV0_FBPREDIV_SHIFT; + m2_int = pll_state->mg_pll_div0 & DKL_PLL_DIV0_FBDIV_INT_MASK; + + if (pll_state->mg_pll_bias & DKL_PLL_BIAS_FRAC_EN_H) { + m2_frac = pll_state->mg_pll_bias & + DKL_PLL_BIAS_FBDIV_FRAC_MASK; + m2_frac = m2_frac >> DKL_PLL_BIAS_FBDIV_SHIFT; + } else { + m2_frac = 0; + } + } else { + m1 = pll_state->mg_pll_div1 & MG_PLL_DIV1_FBPREDIV_MASK; + m2_int = pll_state->mg_pll_div0 & MG_PLL_DIV0_FBDIV_INT_MASK; + + if (pll_state->mg_pll_div0 & MG_PLL_DIV0_FRACNEN_H) { + m2_frac = pll_state->mg_pll_div0 & + MG_PLL_DIV0_FBDIV_FRAC_MASK; + m2_frac = m2_frac >> MG_PLL_DIV0_FBDIV_FRAC_SHIFT; + } else { + m2_frac = 0; + } + } switch (pll_state->mg_clktop2_hsclkctl & MG_CLKTOP2_HSCLKCTL_HSDIV_RATIO_MASK) { @@ -1692,7 +1755,8 @@ static void intel_ddi_clock_get(struct intel_encoder *encoder, hsw_ddi_clock_get(encoder, pipe_config); } -void intel_ddi_set_pipe_settings(const struct intel_crtc_state *crtc_state) +void intel_ddi_set_dp_msa(const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); @@ -1704,44 +1768,50 @@ void intel_ddi_set_pipe_settings(const struct intel_crtc_state *crtc_state) WARN_ON(transcoder_is_dsi(cpu_transcoder)); - temp = TRANS_MSA_SYNC_CLK; - - if (crtc_state->limited_color_range) - temp |= TRANS_MSA_CEA_RANGE; + temp = DP_MSA_MISC_SYNC_CLOCK; switch (crtc_state->pipe_bpp) { case 18: - temp |= TRANS_MSA_6_BPC; + temp |= DP_MSA_MISC_6_BPC; break; case 24: - temp |= TRANS_MSA_8_BPC; + temp |= DP_MSA_MISC_8_BPC; break; case 30: - temp |= TRANS_MSA_10_BPC; + temp |= DP_MSA_MISC_10_BPC; break; case 36: - temp |= TRANS_MSA_12_BPC; + temp |= DP_MSA_MISC_12_BPC; break; default: MISSING_CASE(crtc_state->pipe_bpp); break; } + /* nonsense combination */ + WARN_ON(crtc_state->limited_color_range && + crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB); + + if (crtc_state->limited_color_range) + temp |= DP_MSA_MISC_COLOR_CEA_RGB; + /* * As per DP 1.2 spec section 2.3.4.3 while sending * YCBCR 444 signals we should program MSA MISC1/0 fields with - * colorspace information. The output colorspace encoding is BT601. + * colorspace information. */ if (crtc_state->output_format == INTEL_OUTPUT_FORMAT_YCBCR444) - temp |= TRANS_MSA_SAMPLING_444 | TRANS_MSA_CLRSP_YCBCR; + temp |= DP_MSA_MISC_COLOR_YCBCR_444_BT709; + /* * As per DP 1.4a spec section 2.2.4.3 [MSA Field for Indication * of Color Encoding Format and Content Color Gamut] while sending - * YCBCR 420 signals we should program MSA MISC1 fields which - * indicate VSC SDP for the Pixel Encoding/Colorimetry Format. + * YCBCR 420, HDR BT.2020 signals we should program MSA MISC1 fields + * which indicate VSC SDP for the Pixel Encoding/Colorimetry Format. */ - if (crtc_state->output_format == INTEL_OUTPUT_FORMAT_YCBCR420) - temp |= TRANS_MSA_USE_VSC_SDP; + if (intel_dp_needs_vsc_sdp(crtc_state, conn_state)) + temp |= DP_MSA_MISC_COLOR_VSC_SDP; + I915_WRITE(TRANS_MSA_MISC(cpu_transcoder), temp); } @@ -1761,7 +1831,14 @@ void intel_ddi_set_vc_payload_alloc(const struct intel_crtc_state *crtc_state, I915_WRITE(TRANS_DDI_FUNC_CTL(cpu_transcoder), temp); } -void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state) +/* + * Returns the TRANS_DDI_FUNC_CTL value based on CRTC state. + * + * Only intended to be used by intel_ddi_enable_transcoder_func() and + * intel_ddi_config_transcoder_func(). + */ +static u32 +intel_ddi_transcoder_func_reg_val_get(const struct intel_crtc_state *crtc_state) { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct intel_encoder *encoder = intel_ddi_get_crtc_encoder(crtc); @@ -1840,11 +1917,42 @@ void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state) } else if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP_MST)) { temp |= TRANS_DDI_MODE_SELECT_DP_MST; temp |= DDI_PORT_WIDTH(crtc_state->lane_count); + + if (INTEL_GEN(dev_priv) >= 12) + temp |= TRANS_DDI_MST_TRANSPORT_SELECT(crtc_state->cpu_transcoder); } else { temp |= TRANS_DDI_MODE_SELECT_DP_SST; temp |= DDI_PORT_WIDTH(crtc_state->lane_count); } + return temp; +} + +void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; + u32 temp; + + temp = intel_ddi_transcoder_func_reg_val_get(crtc_state); + I915_WRITE(TRANS_DDI_FUNC_CTL(cpu_transcoder), temp); +} + +/* + * Same as intel_ddi_enable_transcoder_func(), but it does not set the enable + * bit. + */ +static void +intel_ddi_config_transcoder_func(const struct intel_crtc_state *crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; + u32 temp; + + temp = intel_ddi_transcoder_func_reg_val_get(crtc_state); + temp &= ~TRANS_DDI_FUNC_ENABLE; I915_WRITE(TRANS_DDI_FUNC_CTL(cpu_transcoder), temp); } @@ -2045,18 +2153,20 @@ static void intel_ddi_get_encoder_pipes(struct intel_encoder *encoder, } if (!*pipe_mask) - DRM_DEBUG_KMS("No pipe for ddi port %c found\n", - port_name(port)); + DRM_DEBUG_KMS("No pipe for [ENCODER:%d:%s] found\n", + encoder->base.base.id, encoder->base.name); if (!mst_pipe_mask && hweight8(*pipe_mask) > 1) { - DRM_DEBUG_KMS("Multiple pipes for non DP-MST port %c (pipe_mask %02x)\n", - port_name(port), *pipe_mask); + DRM_DEBUG_KMS("Multiple pipes for [ENCODER:%d:%s] (pipe_mask %02x)\n", + encoder->base.base.id, encoder->base.name, + *pipe_mask); *pipe_mask = BIT(ffs(*pipe_mask) - 1); } if (mst_pipe_mask && mst_pipe_mask != *pipe_mask) - DRM_DEBUG_KMS("Conflicting MST and non-MST encoders for port %c (pipe_mask %02x mst_pipe_mask %02x)\n", - port_name(port), *pipe_mask, mst_pipe_mask); + DRM_DEBUG_KMS("Conflicting MST and non-MST state for [ENCODER:%d:%s] (pipe_mask %02x mst_pipe_mask %02x)\n", + encoder->base.base.id, encoder->base.name, + *pipe_mask, mst_pipe_mask); else *is_dp_mst = mst_pipe_mask; @@ -2066,8 +2176,9 @@ out: if ((tmp & (BXT_PHY_CMNLANE_POWERDOWN_ACK | BXT_PHY_LANE_POWERDOWN_ACK | BXT_PHY_LANE_ENABLED)) != BXT_PHY_LANE_ENABLED) - DRM_ERROR("Port %c enabled but PHY powered down? " - "(PHY_CTL %08x)\n", port_name(port), tmp); + DRM_ERROR("[ENCODER:%d:%s] enabled but PHY powered down? " + "(PHY_CTL %08x)\n", encoder->base.base.id, + encoder->base.name, tmp); } intel_display_power_put(dev_priv, encoder->power_domain, wakeref); @@ -2138,7 +2249,7 @@ static void intel_ddi_get_power_domains(struct intel_encoder *encoder, /* * VDSC power is needed when DSC is enabled */ - if (crtc_state->dsc_params.compression_enable) + if (crtc_state->dsc.compression_enable) intel_display_power_get(dev_priv, intel_dsc_power_domain(crtc_state)); } @@ -2269,7 +2380,13 @@ u8 intel_ddi_dp_voltage_max(struct intel_encoder *encoder) enum phy phy = intel_port_to_phy(dev_priv, port); int n_entries; - if (INTEL_GEN(dev_priv) >= 11) { + if (INTEL_GEN(dev_priv) >= 12) { + if (intel_phy_is_combo(dev_priv, phy)) + icl_get_combo_buf_trans(dev_priv, encoder->type, + intel_dp->link_rate, &n_entries); + else + n_entries = ARRAY_SIZE(tgl_dkl_phy_dp_ddi_trans); + } else if (INTEL_GEN(dev_priv) == 11) { if (intel_phy_is_combo(dev_priv, phy)) icl_get_combo_buf_trans(dev_priv, encoder->type, intel_dp->link_rate, &n_entries); @@ -2583,7 +2700,7 @@ static void icl_mg_phy_ddi_vswing_sequence(struct intel_encoder *encoder, u32 level) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = encoder->port; + enum tc_port tc_port = intel_port_to_tc(dev_priv, encoder->port); const struct icl_mg_phy_ddi_buf_trans *ddi_translations; u32 n_entries, val; int ln; @@ -2599,33 +2716,33 @@ static void icl_mg_phy_ddi_vswing_sequence(struct intel_encoder *encoder, /* Set MG_TX_LINK_PARAMS cri_use_fs32 to 0. */ for (ln = 0; ln < 2; ln++) { - val = I915_READ(MG_TX1_LINK_PARAMS(ln, port)); + val = I915_READ(MG_TX1_LINK_PARAMS(ln, tc_port)); val &= ~CRI_USE_FS32; - I915_WRITE(MG_TX1_LINK_PARAMS(ln, port), val); + I915_WRITE(MG_TX1_LINK_PARAMS(ln, tc_port), val); - val = I915_READ(MG_TX2_LINK_PARAMS(ln, port)); + val = I915_READ(MG_TX2_LINK_PARAMS(ln, tc_port)); val &= ~CRI_USE_FS32; - I915_WRITE(MG_TX2_LINK_PARAMS(ln, port), val); + I915_WRITE(MG_TX2_LINK_PARAMS(ln, tc_port), val); } /* Program MG_TX_SWINGCTRL with values from vswing table */ for (ln = 0; ln < 2; ln++) { - val = I915_READ(MG_TX1_SWINGCTRL(ln, port)); + val = I915_READ(MG_TX1_SWINGCTRL(ln, tc_port)); val &= ~CRI_TXDEEMPH_OVERRIDE_17_12_MASK; val |= CRI_TXDEEMPH_OVERRIDE_17_12( ddi_translations[level].cri_txdeemph_override_17_12); - I915_WRITE(MG_TX1_SWINGCTRL(ln, port), val); + I915_WRITE(MG_TX1_SWINGCTRL(ln, tc_port), val); - val = I915_READ(MG_TX2_SWINGCTRL(ln, port)); + val = I915_READ(MG_TX2_SWINGCTRL(ln, tc_port)); val &= ~CRI_TXDEEMPH_OVERRIDE_17_12_MASK; val |= CRI_TXDEEMPH_OVERRIDE_17_12( ddi_translations[level].cri_txdeemph_override_17_12); - I915_WRITE(MG_TX2_SWINGCTRL(ln, port), val); + I915_WRITE(MG_TX2_SWINGCTRL(ln, tc_port), val); } /* Program MG_TX_DRVCTRL with values from vswing table */ for (ln = 0; ln < 2; ln++) { - val = I915_READ(MG_TX1_DRVCTRL(ln, port)); + val = I915_READ(MG_TX1_DRVCTRL(ln, tc_port)); val &= ~(CRI_TXDEEMPH_OVERRIDE_11_6_MASK | CRI_TXDEEMPH_OVERRIDE_5_0_MASK); val |= CRI_TXDEEMPH_OVERRIDE_5_0( @@ -2633,9 +2750,9 @@ static void icl_mg_phy_ddi_vswing_sequence(struct intel_encoder *encoder, CRI_TXDEEMPH_OVERRIDE_11_6( ddi_translations[level].cri_txdeemph_override_11_6) | CRI_TXDEEMPH_OVERRIDE_EN; - I915_WRITE(MG_TX1_DRVCTRL(ln, port), val); + I915_WRITE(MG_TX1_DRVCTRL(ln, tc_port), val); - val = I915_READ(MG_TX2_DRVCTRL(ln, port)); + val = I915_READ(MG_TX2_DRVCTRL(ln, tc_port)); val &= ~(CRI_TXDEEMPH_OVERRIDE_11_6_MASK | CRI_TXDEEMPH_OVERRIDE_5_0_MASK); val |= CRI_TXDEEMPH_OVERRIDE_5_0( @@ -2643,7 +2760,7 @@ static void icl_mg_phy_ddi_vswing_sequence(struct intel_encoder *encoder, CRI_TXDEEMPH_OVERRIDE_11_6( ddi_translations[level].cri_txdeemph_override_11_6) | CRI_TXDEEMPH_OVERRIDE_EN; - I915_WRITE(MG_TX2_DRVCTRL(ln, port), val); + I915_WRITE(MG_TX2_DRVCTRL(ln, tc_port), val); /* FIXME: Program CRI_LOADGEN_SEL after the spec is updated */ } @@ -2654,17 +2771,17 @@ static void icl_mg_phy_ddi_vswing_sequence(struct intel_encoder *encoder, * values from table for which TX1 and TX2 enabled. */ for (ln = 0; ln < 2; ln++) { - val = I915_READ(MG_CLKHUB(ln, port)); + val = I915_READ(MG_CLKHUB(ln, tc_port)); if (link_clock < 300000) val |= CFG_LOW_RATE_LKREN_EN; else val &= ~CFG_LOW_RATE_LKREN_EN; - I915_WRITE(MG_CLKHUB(ln, port), val); + I915_WRITE(MG_CLKHUB(ln, tc_port), val); } /* Program the MG_TX_DCC<LN, port being used> based on the link frequency */ for (ln = 0; ln < 2; ln++) { - val = I915_READ(MG_TX1_DCC(ln, port)); + val = I915_READ(MG_TX1_DCC(ln, tc_port)); val &= ~CFG_AMI_CK_DIV_OVERRIDE_VAL_MASK; if (link_clock <= 500000) { val &= ~CFG_AMI_CK_DIV_OVERRIDE_EN; @@ -2672,9 +2789,9 @@ static void icl_mg_phy_ddi_vswing_sequence(struct intel_encoder *encoder, val |= CFG_AMI_CK_DIV_OVERRIDE_EN | CFG_AMI_CK_DIV_OVERRIDE_VAL(1); } - I915_WRITE(MG_TX1_DCC(ln, port), val); + I915_WRITE(MG_TX1_DCC(ln, tc_port), val); - val = I915_READ(MG_TX2_DCC(ln, port)); + val = I915_READ(MG_TX2_DCC(ln, tc_port)); val &= ~CFG_AMI_CK_DIV_OVERRIDE_VAL_MASK; if (link_clock <= 500000) { val &= ~CFG_AMI_CK_DIV_OVERRIDE_EN; @@ -2682,18 +2799,18 @@ static void icl_mg_phy_ddi_vswing_sequence(struct intel_encoder *encoder, val |= CFG_AMI_CK_DIV_OVERRIDE_EN | CFG_AMI_CK_DIV_OVERRIDE_VAL(1); } - I915_WRITE(MG_TX2_DCC(ln, port), val); + I915_WRITE(MG_TX2_DCC(ln, tc_port), val); } /* Program MG_TX_PISO_READLOAD with values from vswing table */ for (ln = 0; ln < 2; ln++) { - val = I915_READ(MG_TX1_PISO_READLOAD(ln, port)); + val = I915_READ(MG_TX1_PISO_READLOAD(ln, tc_port)); val |= CRI_CALCINIT; - I915_WRITE(MG_TX1_PISO_READLOAD(ln, port), val); + I915_WRITE(MG_TX1_PISO_READLOAD(ln, tc_port), val); - val = I915_READ(MG_TX2_PISO_READLOAD(ln, port)); + val = I915_READ(MG_TX2_PISO_READLOAD(ln, tc_port)); val |= CRI_CALCINIT; - I915_WRITE(MG_TX2_PISO_READLOAD(ln, port), val); + I915_WRITE(MG_TX2_PISO_READLOAD(ln, tc_port), val); } } @@ -2711,6 +2828,69 @@ static void icl_ddi_vswing_sequence(struct intel_encoder *encoder, icl_mg_phy_ddi_vswing_sequence(encoder, link_clock, level); } +static void +tgl_dkl_phy_ddi_vswing_sequence(struct intel_encoder *encoder, int link_clock, + u32 level) +{ + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + enum tc_port tc_port = intel_port_to_tc(dev_priv, encoder->port); + const struct tgl_dkl_phy_ddi_buf_trans *ddi_translations; + u32 n_entries, val, ln, dpcnt_mask, dpcnt_val; + + if (encoder->type == INTEL_OUTPUT_HDMI) { + n_entries = ARRAY_SIZE(tgl_dkl_phy_hdmi_ddi_trans); + ddi_translations = tgl_dkl_phy_hdmi_ddi_trans; + } else { + n_entries = ARRAY_SIZE(tgl_dkl_phy_dp_ddi_trans); + ddi_translations = tgl_dkl_phy_dp_ddi_trans; + } + + if (level >= n_entries) + level = n_entries - 1; + + dpcnt_mask = (DKL_TX_PRESHOOT_COEFF_MASK | + DKL_TX_DE_EMPAHSIS_COEFF_MASK | + DKL_TX_VSWING_CONTROL_MASK); + dpcnt_val = DKL_TX_VSWING_CONTROL(ddi_translations[level].dkl_vswing_control); + dpcnt_val |= DKL_TX_DE_EMPHASIS_COEFF(ddi_translations[level].dkl_de_emphasis_control); + dpcnt_val |= DKL_TX_PRESHOOT_COEFF(ddi_translations[level].dkl_preshoot_control); + + for (ln = 0; ln < 2; ln++) { + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, ln)); + + I915_WRITE(DKL_TX_PMD_LANE_SUS(tc_port), 0); + + /* All the registers are RMW */ + val = I915_READ(DKL_TX_DPCNTL0(tc_port)); + val &= ~dpcnt_mask; + val |= dpcnt_val; + I915_WRITE(DKL_TX_DPCNTL0(tc_port), val); + + val = I915_READ(DKL_TX_DPCNTL1(tc_port)); + val &= ~dpcnt_mask; + val |= dpcnt_val; + I915_WRITE(DKL_TX_DPCNTL1(tc_port), val); + + val = I915_READ(DKL_TX_DPCNTL2(tc_port)); + val &= ~DKL_TX_DP20BITMODE; + I915_WRITE(DKL_TX_DPCNTL2(tc_port), val); + } +} + +static void tgl_ddi_vswing_sequence(struct intel_encoder *encoder, + int link_clock, + u32 level, + enum intel_output_type type) +{ + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + enum phy phy = intel_port_to_phy(dev_priv, encoder->port); + + if (intel_phy_is_combo(dev_priv, phy)) + icl_combo_phy_ddi_vswing_sequence(encoder, level, type); + else + tgl_dkl_phy_ddi_vswing_sequence(encoder, link_clock, level); +} + static u32 translate_signal_level(int signal_levels) { int i; @@ -2742,7 +2922,10 @@ u32 bxt_signal_levels(struct intel_dp *intel_dp) struct intel_encoder *encoder = &dport->base; int level = intel_ddi_dp_level(intel_dp); - if (INTEL_GEN(dev_priv) >= 11) + if (INTEL_GEN(dev_priv) >= 12) + tgl_ddi_vswing_sequence(encoder, intel_dp->link_rate, + level, encoder->type); + else if (INTEL_GEN(dev_priv) >= 11) icl_ddi_vswing_sequence(encoder, intel_dp->link_rate, level, encoder->type); else if (IS_CANNONLAKE(dev_priv)) @@ -2989,130 +3172,141 @@ static void intel_ddi_clk_disable(struct intel_encoder *encoder) } } -static void icl_enable_phy_clock_gating(struct intel_digital_port *dig_port) +static void +icl_phy_set_clock_gating(struct intel_digital_port *dig_port, bool enable) { struct drm_i915_private *dev_priv = to_i915(dig_port->base.base.dev); - enum port port = dig_port->base.port; - enum tc_port tc_port = intel_port_to_tc(dev_priv, port); - u32 val; + enum tc_port tc_port = intel_port_to_tc(dev_priv, dig_port->base.port); + u32 val, bits; int ln; if (tc_port == PORT_TC_NONE) return; - for (ln = 0; ln < 2; ln++) { - val = I915_READ(MG_DP_MODE(ln, port)); - val |= MG_DP_MODE_CFG_TR2PWR_GATING | - MG_DP_MODE_CFG_TRPWR_GATING | - MG_DP_MODE_CFG_CLNPWR_GATING | - MG_DP_MODE_CFG_DIGPWR_GATING | - MG_DP_MODE_CFG_GAONPWR_GATING; - I915_WRITE(MG_DP_MODE(ln, port), val); - } - - val = I915_READ(MG_MISC_SUS0(tc_port)); - val |= MG_MISC_SUS0_SUSCLK_DYNCLKGATE_MODE(3) | - MG_MISC_SUS0_CFG_TR2PWR_GATING | - MG_MISC_SUS0_CFG_CL2PWR_GATING | - MG_MISC_SUS0_CFG_GAONPWR_GATING | - MG_MISC_SUS0_CFG_TRPWR_GATING | - MG_MISC_SUS0_CFG_CL1PWR_GATING | - MG_MISC_SUS0_CFG_DGPWR_GATING; - I915_WRITE(MG_MISC_SUS0(tc_port), val); -} + bits = MG_DP_MODE_CFG_TR2PWR_GATING | MG_DP_MODE_CFG_TRPWR_GATING | + MG_DP_MODE_CFG_CLNPWR_GATING | MG_DP_MODE_CFG_DIGPWR_GATING | + MG_DP_MODE_CFG_GAONPWR_GATING; -static void icl_disable_phy_clock_gating(struct intel_digital_port *dig_port) -{ - struct drm_i915_private *dev_priv = to_i915(dig_port->base.base.dev); - enum port port = dig_port->base.port; - enum tc_port tc_port = intel_port_to_tc(dev_priv, port); - u32 val; - int ln; + for (ln = 0; ln < 2; ln++) { + if (INTEL_GEN(dev_priv) >= 12) { + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, ln)); + val = I915_READ(DKL_DP_MODE(tc_port)); + } else { + val = I915_READ(MG_DP_MODE(ln, tc_port)); + } - if (tc_port == PORT_TC_NONE) - return; + if (enable) + val |= bits; + else + val &= ~bits; - for (ln = 0; ln < 2; ln++) { - val = I915_READ(MG_DP_MODE(ln, port)); - val &= ~(MG_DP_MODE_CFG_TR2PWR_GATING | - MG_DP_MODE_CFG_TRPWR_GATING | - MG_DP_MODE_CFG_CLNPWR_GATING | - MG_DP_MODE_CFG_DIGPWR_GATING | - MG_DP_MODE_CFG_GAONPWR_GATING); - I915_WRITE(MG_DP_MODE(ln, port), val); + if (INTEL_GEN(dev_priv) >= 12) + I915_WRITE(DKL_DP_MODE(tc_port), val); + else + I915_WRITE(MG_DP_MODE(ln, tc_port), val); } - val = I915_READ(MG_MISC_SUS0(tc_port)); - val &= ~(MG_MISC_SUS0_SUSCLK_DYNCLKGATE_MODE_MASK | - MG_MISC_SUS0_CFG_TR2PWR_GATING | - MG_MISC_SUS0_CFG_CL2PWR_GATING | - MG_MISC_SUS0_CFG_GAONPWR_GATING | - MG_MISC_SUS0_CFG_TRPWR_GATING | - MG_MISC_SUS0_CFG_CL1PWR_GATING | - MG_MISC_SUS0_CFG_DGPWR_GATING); - I915_WRITE(MG_MISC_SUS0(tc_port), val); + if (INTEL_GEN(dev_priv) == 11) { + bits = MG_MISC_SUS0_CFG_TR2PWR_GATING | + MG_MISC_SUS0_CFG_CL2PWR_GATING | + MG_MISC_SUS0_CFG_GAONPWR_GATING | + MG_MISC_SUS0_CFG_TRPWR_GATING | + MG_MISC_SUS0_CFG_CL1PWR_GATING | + MG_MISC_SUS0_CFG_DGPWR_GATING; + + val = I915_READ(MG_MISC_SUS0(tc_port)); + if (enable) + val |= (bits | MG_MISC_SUS0_SUSCLK_DYNCLKGATE_MODE(3)); + else + val &= ~(bits | MG_MISC_SUS0_SUSCLK_DYNCLKGATE_MODE_MASK); + I915_WRITE(MG_MISC_SUS0(tc_port), val); + } } -static void icl_program_mg_dp_mode(struct intel_digital_port *intel_dig_port) +static void +icl_program_mg_dp_mode(struct intel_digital_port *intel_dig_port, + const struct intel_crtc_state *crtc_state) { struct drm_i915_private *dev_priv = to_i915(intel_dig_port->base.base.dev); - enum port port = intel_dig_port->base.port; - u32 ln0, ln1, lane_mask; + enum tc_port tc_port = intel_port_to_tc(dev_priv, intel_dig_port->base.port); + u32 ln0, ln1, pin_assignment; + u8 width; if (intel_dig_port->tc_mode == TC_PORT_TBT_ALT) return; - ln0 = I915_READ(MG_DP_MODE(0, port)); - ln1 = I915_READ(MG_DP_MODE(1, port)); + if (INTEL_GEN(dev_priv) >= 12) { + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, 0x0)); + ln0 = I915_READ(DKL_DP_MODE(tc_port)); + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, 0x1)); + ln1 = I915_READ(DKL_DP_MODE(tc_port)); + } else { + ln0 = I915_READ(MG_DP_MODE(0, tc_port)); + ln1 = I915_READ(MG_DP_MODE(1, tc_port)); + } - switch (intel_dig_port->tc_mode) { - case TC_PORT_DP_ALT: - ln0 &= ~(MG_DP_MODE_CFG_DP_X1_MODE | MG_DP_MODE_CFG_DP_X2_MODE); - ln1 &= ~(MG_DP_MODE_CFG_DP_X1_MODE | MG_DP_MODE_CFG_DP_X2_MODE); + ln0 &= ~(MG_DP_MODE_CFG_DP_X1_MODE | MG_DP_MODE_CFG_DP_X1_MODE); + ln1 &= ~(MG_DP_MODE_CFG_DP_X1_MODE | MG_DP_MODE_CFG_DP_X2_MODE); - lane_mask = intel_tc_port_get_lane_mask(intel_dig_port); + /* DPPATC */ + pin_assignment = intel_tc_port_get_pin_assignment_mask(intel_dig_port); + width = crtc_state->lane_count; - switch (lane_mask) { - case 0x1: - case 0x4: - break; - case 0x2: + switch (pin_assignment) { + case 0x0: + WARN_ON(intel_dig_port->tc_mode != TC_PORT_LEGACY); + if (width == 1) { + ln1 |= MG_DP_MODE_CFG_DP_X1_MODE; + } else { + ln0 |= MG_DP_MODE_CFG_DP_X2_MODE; + ln1 |= MG_DP_MODE_CFG_DP_X2_MODE; + } + break; + case 0x1: + if (width == 4) { + ln0 |= MG_DP_MODE_CFG_DP_X2_MODE; + ln1 |= MG_DP_MODE_CFG_DP_X2_MODE; + } + break; + case 0x2: + if (width == 2) { + ln0 |= MG_DP_MODE_CFG_DP_X2_MODE; + ln1 |= MG_DP_MODE_CFG_DP_X2_MODE; + } + break; + case 0x3: + case 0x5: + if (width == 1) { ln0 |= MG_DP_MODE_CFG_DP_X1_MODE; - break; - case 0x3: - ln0 |= MG_DP_MODE_CFG_DP_X1_MODE | - MG_DP_MODE_CFG_DP_X2_MODE; - break; - case 0x8: ln1 |= MG_DP_MODE_CFG_DP_X1_MODE; - break; - case 0xC: - ln1 |= MG_DP_MODE_CFG_DP_X1_MODE | - MG_DP_MODE_CFG_DP_X2_MODE; - break; - case 0xF: - ln0 |= MG_DP_MODE_CFG_DP_X1_MODE | - MG_DP_MODE_CFG_DP_X2_MODE; - ln1 |= MG_DP_MODE_CFG_DP_X1_MODE | - MG_DP_MODE_CFG_DP_X2_MODE; - break; - default: - MISSING_CASE(lane_mask); + } else { + ln0 |= MG_DP_MODE_CFG_DP_X2_MODE; + ln1 |= MG_DP_MODE_CFG_DP_X2_MODE; } break; - - case TC_PORT_LEGACY: - ln0 |= MG_DP_MODE_CFG_DP_X1_MODE | MG_DP_MODE_CFG_DP_X2_MODE; - ln1 |= MG_DP_MODE_CFG_DP_X1_MODE | MG_DP_MODE_CFG_DP_X2_MODE; + case 0x4: + case 0x6: + if (width == 1) { + ln0 |= MG_DP_MODE_CFG_DP_X1_MODE; + ln1 |= MG_DP_MODE_CFG_DP_X1_MODE; + } else { + ln0 |= MG_DP_MODE_CFG_DP_X2_MODE; + ln1 |= MG_DP_MODE_CFG_DP_X2_MODE; + } break; - default: - MISSING_CASE(intel_dig_port->tc_mode); - return; + MISSING_CASE(pin_assignment); } - I915_WRITE(MG_DP_MODE(0, port), ln0); - I915_WRITE(MG_DP_MODE(1, port), ln1); + if (INTEL_GEN(dev_priv) >= 12) { + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, 0x0)); + I915_WRITE(DKL_DP_MODE(tc_port), ln0); + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, 0x1)); + I915_WRITE(DKL_DP_MODE(tc_port), ln1); + } else { + I915_WRITE(MG_DP_MODE(0, tc_port), ln0); + I915_WRITE(MG_DP_MODE(1, tc_port), ln1); + } } static void intel_dp_sink_set_fec_ready(struct intel_dp *intel_dp, @@ -3129,17 +3323,18 @@ static void intel_ddi_enable_fec(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = encoder->port; + struct intel_dp *intel_dp; u32 val; if (!crtc_state->fec_enable) return; - val = I915_READ(DP_TP_CTL(port)); + intel_dp = enc_to_intel_dp(&encoder->base); + val = I915_READ(intel_dp->regs.dp_tp_ctl); val |= DP_TP_CTL_FEC_ENABLE; - I915_WRITE(DP_TP_CTL(port), val); + I915_WRITE(intel_dp->regs.dp_tp_ctl, val); - if (intel_de_wait_for_set(dev_priv, DP_TP_STATUS(port), + if (intel_de_wait_for_set(dev_priv, intel_dp->regs.dp_tp_status, DP_TP_STATUS_FEC_ENABLE_LIVE, 1)) DRM_ERROR("Timed out waiting for FEC Enable Status\n"); } @@ -3148,21 +3343,205 @@ static void intel_ddi_disable_fec_state(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = encoder->port; + struct intel_dp *intel_dp; u32 val; if (!crtc_state->fec_enable) return; - val = I915_READ(DP_TP_CTL(port)); + intel_dp = enc_to_intel_dp(&encoder->base); + val = I915_READ(intel_dp->regs.dp_tp_ctl); val &= ~DP_TP_CTL_FEC_ENABLE; - I915_WRITE(DP_TP_CTL(port), val); - POSTING_READ(DP_TP_CTL(port)); + I915_WRITE(intel_dp->regs.dp_tp_ctl, val); + POSTING_READ(intel_dp->regs.dp_tp_ctl); } -static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder, - const struct intel_crtc_state *crtc_state, - const struct drm_connector_state *conn_state) +static void +tgl_clear_psr2_transcoder_exitline(const struct intel_crtc_state *cstate) +{ + struct drm_i915_private *dev_priv = to_i915(cstate->base.crtc->dev); + u32 val; + + if (!cstate->dc3co_exitline) + return; + + val = I915_READ(EXITLINE(cstate->cpu_transcoder)); + val &= ~(EXITLINE_MASK | EXITLINE_ENABLE); + I915_WRITE(EXITLINE(cstate->cpu_transcoder), val); +} + +static void +tgl_set_psr2_transcoder_exitline(const struct intel_crtc_state *cstate) +{ + u32 val, exit_scanlines; + struct drm_i915_private *dev_priv = to_i915(cstate->base.crtc->dev); + + if (!cstate->dc3co_exitline) + return; + + exit_scanlines = cstate->dc3co_exitline; + exit_scanlines <<= EXITLINE_SHIFT; + val = I915_READ(EXITLINE(cstate->cpu_transcoder)); + val &= ~(EXITLINE_MASK | EXITLINE_ENABLE); + val |= exit_scanlines; + val |= EXITLINE_ENABLE; + I915_WRITE(EXITLINE(cstate->cpu_transcoder), val); +} + +static void tgl_dc3co_exitline_compute_config(struct intel_encoder *encoder, + struct intel_crtc_state *cstate) +{ + u32 exit_scanlines; + struct drm_i915_private *dev_priv = to_i915(cstate->base.crtc->dev); + u32 crtc_vdisplay = cstate->base.adjusted_mode.crtc_vdisplay; + + cstate->dc3co_exitline = 0; + + if (!(dev_priv->csr.allowed_dc_mask & DC_STATE_EN_DC3CO)) + return; + + /* B.Specs:49196 DC3CO only works with pipeA and DDIA.*/ + if (to_intel_crtc(cstate->base.crtc)->pipe != PIPE_A || + encoder->port != PORT_A) + return; + + if (!cstate->has_psr2 || !cstate->base.active) + return; + + /* + * DC3CO Exit time 200us B.Spec 49196 + * PSR2 transcoder Early Exit scanlines = ROUNDUP(200 / line time) + 1 + */ + exit_scanlines = + intel_usecs_to_scanlines(&cstate->base.adjusted_mode, 200) + 1; + + if (WARN_ON(exit_scanlines > crtc_vdisplay)) + return; + + cstate->dc3co_exitline = crtc_vdisplay - exit_scanlines; + DRM_DEBUG_KMS("DC3CO exit scanlines %d\n", cstate->dc3co_exitline); +} + +static void tgl_dc3co_exitline_get_config(struct intel_crtc_state *crtc_state) +{ + u32 val; + struct drm_i915_private *dev_priv = to_i915(crtc_state->base.crtc->dev); + + if (INTEL_GEN(dev_priv) < 12) + return; + + val = I915_READ(EXITLINE(crtc_state->cpu_transcoder)); + + if (val & EXITLINE_ENABLE) + crtc_state->dc3co_exitline = val & EXITLINE_MASK; +} + +static void tgl_ddi_pre_enable_dp(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) +{ + struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + enum phy phy = intel_port_to_phy(dev_priv, encoder->port); + struct intel_digital_port *dig_port = enc_to_dig_port(&encoder->base); + bool is_mst = intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP_MST); + int level = intel_ddi_dp_level(intel_dp); + enum transcoder transcoder = crtc_state->cpu_transcoder; + + tgl_set_psr2_transcoder_exitline(crtc_state); + intel_dp_set_link_params(intel_dp, crtc_state->port_clock, + crtc_state->lane_count, is_mst); + + intel_dp->regs.dp_tp_ctl = TGL_DP_TP_CTL(transcoder); + intel_dp->regs.dp_tp_status = TGL_DP_TP_STATUS(transcoder); + + /* 1.a got on intel_atomic_commit_tail() */ + + /* 2. */ + intel_edp_panel_on(intel_dp); + + /* + * 1.b, 3. and 4.a is done before tgl_ddi_pre_enable_dp() by: + * haswell_crtc_enable()->intel_encoders_pre_pll_enable() and + * haswell_crtc_enable()->intel_enable_shared_dpll() + */ + + /* 4.b */ + intel_ddi_clk_select(encoder, crtc_state); + + /* 5. */ + if (!intel_phy_is_tc(dev_priv, phy) || + dig_port->tc_mode != TC_PORT_TBT_ALT) + intel_display_power_get(dev_priv, + dig_port->ddi_io_power_domain); + + /* 6. */ + icl_program_mg_dp_mode(dig_port, crtc_state); + + /* + * 7.a - Steps in this function should only be executed over MST + * master, what will be taken in care by MST hook + * intel_mst_pre_enable_dp() + */ + intel_ddi_enable_pipe_clock(crtc_state); + + /* 7.b */ + intel_ddi_config_transcoder_func(crtc_state); + + /* 7.d */ + icl_phy_set_clock_gating(dig_port, false); + + /* 7.e */ + tgl_ddi_vswing_sequence(encoder, crtc_state->port_clock, level, + encoder->type); + + /* 7.f */ + if (intel_phy_is_combo(dev_priv, phy)) { + bool lane_reversal = + dig_port->saved_port_bits & DDI_BUF_PORT_REVERSAL; + + intel_combo_phy_power_up_lanes(dev_priv, phy, false, + crtc_state->lane_count, + lane_reversal); + } + + /* 7.g */ + intel_ddi_init_dp_buf_reg(encoder); + + if (!is_mst) + intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_ON); + + intel_dp_sink_set_decompression_state(intel_dp, crtc_state, true); + /* + * DDI FEC: "anticipates enabling FEC encoding sets the FEC_READY bit + * in the FEC_CONFIGURATION register to 1 before initiating link + * training + */ + intel_dp_sink_set_fec_ready(intel_dp, crtc_state); + /* 7.c, 7.h, 7.i, 7.j */ + intel_dp_start_link_train(intel_dp); + + /* 7.k */ + if (!is_trans_port_sync_mode(crtc_state)) + intel_dp_stop_link_train(intel_dp); + + /* + * TODO: enable clock gating + * + * It is not written in DP enabling sequence but "PHY Clockgating + * programming" states that clock gating should be enabled after the + * link training but doing so causes all the following trainings to fail + * so not enabling it for now. + */ + + /* 7.l */ + intel_ddi_enable_fec(encoder, crtc_state); + intel_dsc_enable(encoder, crtc_state); +} + +static void hsw_ddi_pre_enable_dp(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); @@ -3177,6 +3556,9 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder, intel_dp_set_link_params(intel_dp, crtc_state->port_clock, crtc_state->lane_count, is_mst); + intel_dp->regs.dp_tp_ctl = DP_TP_CTL(port); + intel_dp->regs.dp_tp_status = DP_TP_STATUS(port); + intel_edp_panel_on(intel_dp); intel_ddi_clk_select(encoder, crtc_state); @@ -3186,8 +3568,8 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder, intel_display_power_get(dev_priv, dig_port->ddi_io_power_domain); - icl_program_mg_dp_mode(dig_port); - icl_disable_phy_clock_gating(dig_port); + icl_program_mg_dp_mode(dig_port, crtc_state); + icl_phy_set_clock_gating(dig_port, false); if (INTEL_GEN(dev_priv) >= 11) icl_ddi_vswing_sequence(encoder, crtc_state->port_clock, @@ -3215,12 +3597,13 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder, true); intel_dp_sink_set_fec_ready(intel_dp, crtc_state); intel_dp_start_link_train(intel_dp); - if (port != PORT_A || INTEL_GEN(dev_priv) >= 9) + if ((port != PORT_A || INTEL_GEN(dev_priv) >= 9) && + !is_trans_port_sync_mode(crtc_state)) intel_dp_stop_link_train(intel_dp); intel_ddi_enable_fec(encoder, crtc_state); - icl_enable_phy_clock_gating(dig_port); + icl_phy_set_clock_gating(dig_port, true); if (!is_mst) intel_ddi_enable_pipe_clock(crtc_state); @@ -3228,6 +3611,24 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder, intel_dsc_enable(encoder, crtc_state); } +static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) +{ + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + + if (INTEL_GEN(dev_priv) >= 12) + tgl_ddi_pre_enable_dp(encoder, crtc_state, conn_state); + else + hsw_ddi_pre_enable_dp(encoder, crtc_state, conn_state); + + /* MST will call a setting of MSA after an allocating of Virtual Channel + * from MST encoder pre_enable callback. + */ + if (!intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP_MST)) + intel_ddi_set_dp_msa(crtc_state, conn_state); +} + static void intel_ddi_pre_enable_hdmi(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state, const struct drm_connector_state *conn_state) @@ -3244,10 +3645,13 @@ static void intel_ddi_pre_enable_hdmi(struct intel_encoder *encoder, intel_display_power_get(dev_priv, dig_port->ddi_io_power_domain); - icl_program_mg_dp_mode(dig_port); - icl_disable_phy_clock_gating(dig_port); + icl_program_mg_dp_mode(dig_port, crtc_state); + icl_phy_set_clock_gating(dig_port, false); - if (INTEL_GEN(dev_priv) >= 11) + if (INTEL_GEN(dev_priv) >= 12) + tgl_ddi_vswing_sequence(encoder, crtc_state->port_clock, + level, INTEL_OUTPUT_HDMI); + else if (INTEL_GEN(dev_priv) == 11) icl_ddi_vswing_sequence(encoder, crtc_state->port_clock, level, INTEL_OUTPUT_HDMI); else if (IS_CANNONLAKE(dev_priv)) @@ -3257,7 +3661,7 @@ static void intel_ddi_pre_enable_hdmi(struct intel_encoder *encoder, else intel_prepare_hdmi_ddi_buffers(encoder, level); - icl_enable_phy_clock_gating(dig_port); + icl_phy_set_clock_gating(dig_port, true); if (IS_GEN9_BC(dev_priv)) skl_ddi_set_iboost(encoder, level, INTEL_OUTPUT_HDMI); @@ -3330,10 +3734,14 @@ static void intel_disable_ddi_buf(struct intel_encoder *encoder, wait = true; } - val = I915_READ(DP_TP_CTL(port)); - val &= ~(DP_TP_CTL_ENABLE | DP_TP_CTL_LINK_TRAIN_MASK); - val |= DP_TP_CTL_LINK_TRAIN_PAT1; - I915_WRITE(DP_TP_CTL(port), val); + if (intel_crtc_has_dp_encoder(crtc_state)) { + struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); + + val = I915_READ(intel_dp->regs.dp_tp_ctl); + val &= ~(DP_TP_CTL_ENABLE | DP_TP_CTL_LINK_TRAIN_MASK); + val |= DP_TP_CTL_LINK_TRAIN_PAT1; + I915_WRITE(intel_dp->regs.dp_tp_ctl, val); + } /* Disable FEC in DP Sink */ intel_ddi_disable_fec_state(encoder, crtc_state); @@ -3373,6 +3781,7 @@ static void intel_ddi_post_disable_dp(struct intel_encoder *encoder, dig_port->ddi_io_power_domain); intel_ddi_clk_disable(encoder); + tgl_clear_psr2_transcoder_exitline(old_crtc_state); } static void intel_ddi_post_disable_hdmi(struct intel_encoder *encoder, @@ -3475,7 +3884,8 @@ static void intel_enable_ddi_dp(struct intel_encoder *encoder, intel_edp_backlight_on(crtc_state, conn_state); intel_psr_enable(intel_dp, crtc_state); - intel_dp_ycbcr_420_enable(intel_dp, crtc_state); + intel_dp_vsc_enable(intel_dp, crtc_state, conn_state); + intel_dp_hdr_metadata_enable(intel_dp, crtc_state, conn_state); intel_edp_drrs_enable(intel_dp, crtc_state); if (crtc_state->has_audio) @@ -3486,12 +3896,12 @@ static i915_reg_t gen9_chicken_trans_reg_by_port(struct drm_i915_private *dev_priv, enum port port) { - static const i915_reg_t regs[] = { - [PORT_A] = CHICKEN_TRANS_EDP, - [PORT_B] = CHICKEN_TRANS_A, - [PORT_C] = CHICKEN_TRANS_B, - [PORT_D] = CHICKEN_TRANS_C, - [PORT_E] = CHICKEN_TRANS_A, + static const enum transcoder trans[] = { + [PORT_A] = TRANSCODER_EDP, + [PORT_B] = TRANSCODER_A, + [PORT_C] = TRANSCODER_B, + [PORT_D] = TRANSCODER_C, + [PORT_E] = TRANSCODER_A, }; WARN_ON(INTEL_GEN(dev_priv) < 9); @@ -3499,7 +3909,7 @@ gen9_chicken_trans_reg_by_port(struct drm_i915_private *dev_priv, if (WARN_ON(port < PORT_A || port > PORT_E)) port = PORT_A; - return regs[port]; + return CHICKEN_TRANS(trans[port]); } static void intel_enable_ddi_hdmi(struct intel_encoder *encoder, @@ -3633,7 +4043,7 @@ static void intel_ddi_update_pipe_dp(struct intel_encoder *encoder, { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - intel_ddi_set_pipe_settings(crtc_state); + intel_ddi_set_dp_msa(crtc_state, conn_state); intel_psr_update(intel_dp, crtc_state); intel_edp_drrs_enable(intel_dp, crtc_state); @@ -3761,7 +4171,7 @@ static void intel_ddi_prepare_link_retrain(struct intel_dp *intel_dp) u32 val; bool wait = false; - if (I915_READ(DP_TP_CTL(port)) & DP_TP_CTL_ENABLE) { + if (I915_READ(intel_dp->regs.dp_tp_ctl) & DP_TP_CTL_ENABLE) { val = I915_READ(DDI_BUF_CTL(port)); if (val & DDI_BUF_CTL_ENABLE) { val &= ~DDI_BUF_CTL_ENABLE; @@ -3769,11 +4179,11 @@ static void intel_ddi_prepare_link_retrain(struct intel_dp *intel_dp) wait = true; } - val = I915_READ(DP_TP_CTL(port)); + val = I915_READ(intel_dp->regs.dp_tp_ctl); val &= ~(DP_TP_CTL_ENABLE | DP_TP_CTL_LINK_TRAIN_MASK); val |= DP_TP_CTL_LINK_TRAIN_PAT1; - I915_WRITE(DP_TP_CTL(port), val); - POSTING_READ(DP_TP_CTL(port)); + I915_WRITE(intel_dp->regs.dp_tp_ctl, val); + POSTING_READ(intel_dp->regs.dp_tp_ctl); if (wait) intel_wait_ddi_buf_idle(dev_priv, port); @@ -3788,8 +4198,8 @@ static void intel_ddi_prepare_link_retrain(struct intel_dp *intel_dp) if (drm_dp_enhanced_frame_cap(intel_dp->dpcd)) val |= DP_TP_CTL_ENHANCED_FRAME_ENABLE; } - I915_WRITE(DP_TP_CTL(port), val); - POSTING_READ(DP_TP_CTL(port)); + I915_WRITE(intel_dp->regs.dp_tp_ctl, val); + POSTING_READ(intel_dp->regs.dp_tp_ctl); intel_dp->DP |= DDI_BUF_CTL_ENABLE; I915_WRITE(DDI_BUF_CTL(port), intel_dp->DP); @@ -3891,6 +4301,23 @@ void intel_ddi_get_config(struct intel_encoder *encoder, pipe_config->lane_count = ((temp & DDI_PORT_WIDTH_MASK) >> DDI_PORT_WIDTH_SHIFT) + 1; intel_dp_get_m_n(intel_crtc, pipe_config); + + if (INTEL_GEN(dev_priv) >= 11) { + i915_reg_t dp_tp_ctl; + + if (IS_GEN(dev_priv, 11)) + dp_tp_ctl = DP_TP_CTL(encoder->port); + else + dp_tp_ctl = TGL_DP_TP_CTL(pipe_config->cpu_transcoder); + + pipe_config->fec_enable = + I915_READ(dp_tp_ctl) & DP_TP_CTL_FEC_ENABLE; + + DRM_DEBUG_KMS("[ENCODER:%d:%s] Fec status: %u\n", + encoder->base.base.id, encoder->base.name, + pipe_config->fec_enable); + } + break; case TRANS_DDI_MODE_SELECT_DP_MST: pipe_config->output_types |= BIT(INTEL_OUTPUT_DP_MST); @@ -3902,6 +4329,9 @@ void intel_ddi_get_config(struct intel_encoder *encoder, break; } + if (encoder->type == INTEL_OUTPUT_EDP) + tgl_dc3co_exitline_get_config(pipe_config); + pipe_config->has_audio = intel_ddi_is_audio_enabled(dev_priv, cpu_transcoder); @@ -3979,10 +4409,13 @@ static int intel_ddi_compute_config(struct intel_encoder *encoder, if (HAS_TRANSCODER_EDP(dev_priv) && port == PORT_A) pipe_config->cpu_transcoder = TRANSCODER_EDP; - if (intel_crtc_has_type(pipe_config, INTEL_OUTPUT_HDMI)) + if (intel_crtc_has_type(pipe_config, INTEL_OUTPUT_HDMI)) { ret = intel_hdmi_compute_config(encoder, pipe_config, conn_state); - else + } else { ret = intel_dp_compute_config(encoder, pipe_config, conn_state); + tgl_dc3co_exitline_compute_config(encoder, pipe_config); + } + if (ret) return ret; @@ -4276,7 +4709,6 @@ void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port) struct intel_encoder *intel_encoder; struct drm_encoder *encoder; bool init_hdmi, init_dp, init_lspcon = false; - enum pipe pipe; enum phy phy = intel_port_to_phy(dev_priv, port); init_hdmi = port_info->supports_dvi || port_info->supports_hdmi; @@ -4328,8 +4760,7 @@ void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port) intel_encoder->power_domain = intel_port_to_power_domain(port); intel_encoder->port = port; intel_encoder->cloneable = 0; - for_each_pipe(dev_priv, pipe) - intel_encoder->crtc_mask |= BIT(pipe); + intel_encoder->pipe_mask = ~0; if (INTEL_GEN(dev_priv) >= 11) intel_dig_port->saved_port_bits = I915_READ(DDI_BUF_CTL(port)) & @@ -4351,46 +4782,9 @@ void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port) intel_encoder->update_complete = intel_ddi_update_complete; } - switch (port) { - case PORT_A: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_A_IO; - break; - case PORT_B: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_B_IO; - break; - case PORT_C: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_C_IO; - break; - case PORT_D: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_D_IO; - break; - case PORT_E: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_E_IO; - break; - case PORT_F: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_F_IO; - break; - case PORT_G: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_G_IO; - break; - case PORT_H: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_H_IO; - break; - case PORT_I: - intel_dig_port->ddi_io_power_domain = - POWER_DOMAIN_PORT_DDI_I_IO; - break; - default: - MISSING_CASE(port); - } + WARN_ON(port > PORT_I); + intel_dig_port->ddi_io_power_domain = POWER_DOMAIN_PORT_DDI_A_IO + + port - PORT_A; if (init_dp) { if (!intel_ddi_init_dp_connector(intel_dig_port)) diff --git a/drivers/gpu/drm/i915/display/intel_ddi.h b/drivers/gpu/drm/i915/display/intel_ddi.h index a08365da2643..19aeab1246ee 100644 --- a/drivers/gpu/drm/i915/display/intel_ddi.h +++ b/drivers/gpu/drm/i915/display/intel_ddi.h @@ -30,7 +30,8 @@ void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state) void intel_ddi_disable_transcoder_func(const struct intel_crtc_state *crtc_state); void intel_ddi_enable_pipe_clock(const struct intel_crtc_state *crtc_state); void intel_ddi_disable_pipe_clock(const struct intel_crtc_state *crtc_state); -void intel_ddi_set_pipe_settings(const struct intel_crtc_state *crtc_state); +void intel_ddi_set_dp_msa(const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state); bool intel_ddi_connector_get_hw_state(struct intel_connector *intel_connector); void intel_ddi_get_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config); diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c index b51d1ceb8739..6f5e3bd13ad1 100644 --- a/drivers/gpu/drm/i915/display/intel_display.c +++ b/drivers/gpu/drm/i915/display/intel_display.c @@ -31,7 +31,6 @@ #include <linux/module.h> #include <linux/dma-resv.h> #include <linux/slab.h> -#include <linux/vgaarb.h> #include <drm/drm_atomic.h> #include <drm/drm_atomic_helper.h> @@ -56,6 +55,8 @@ #include "display/intel_tv.h" #include "display/intel_vdsc.h" +#include "gt/intel_rps.h" + #include "i915_drv.h" #include "i915_trace.h" #include "intel_acpi.h" @@ -65,6 +66,7 @@ #include "intel_cdclk.h" #include "intel_color.h" #include "intel_display_types.h" +#include "intel_dp_link_training.h" #include "intel_fbc.h" #include "intel_fbdev.h" #include "intel_fifo_underrun.h" @@ -79,6 +81,7 @@ #include "intel_sideband.h" #include "intel_sprite.h" #include "intel_tc.h" +#include "intel_vga.h" /* Primary plane formats for gen <= 3 */ static const u32 i8xx_primary_formats[] = { @@ -88,7 +91,17 @@ static const u32 i8xx_primary_formats[] = { DRM_FORMAT_XRGB8888, }; -/* Primary plane formats for gen >= 4 */ +/* Primary plane formats for ivb (no fp16 due to hw issue) */ +static const u32 ivb_primary_formats[] = { + DRM_FORMAT_C8, + DRM_FORMAT_RGB565, + DRM_FORMAT_XRGB8888, + DRM_FORMAT_XBGR8888, + DRM_FORMAT_XRGB2101010, + DRM_FORMAT_XBGR2101010, +}; + +/* Primary plane formats for gen >= 4, except ivb */ static const u32 i965_primary_formats[] = { DRM_FORMAT_C8, DRM_FORMAT_RGB565, @@ -96,6 +109,7 @@ static const u32 i965_primary_formats[] = { DRM_FORMAT_XBGR8888, DRM_FORMAT_XRGB2101010, DRM_FORMAT_XBGR2101010, + DRM_FORMAT_XBGR16161616F, }; static const u64 i9xx_format_modifiers[] = { @@ -135,8 +149,6 @@ static void vlv_prepare_pll(struct intel_crtc *crtc, const struct intel_crtc_state *pipe_config); static void chv_prepare_pll(struct intel_crtc *crtc, const struct intel_crtc_state *pipe_config); -static void intel_begin_crtc_commit(struct intel_atomic_state *, struct intel_crtc *); -static void intel_finish_crtc_commit(struct intel_atomic_state *, struct intel_crtc *); static void intel_crtc_init_scalers(struct intel_crtc *crtc, struct intel_crtc_state *crtc_state); static void skylake_pfit_enable(const struct intel_crtc_state *crtc_state); @@ -490,7 +502,7 @@ static const struct intel_limit intel_limits_bxt = { /* WA Display #0827: Gen9:all */ static void -skl_wa_827(struct drm_i915_private *dev_priv, int pipe, bool enable) +skl_wa_827(struct drm_i915_private *dev_priv, enum pipe pipe, bool enable) { if (enable) I915_WRITE(CLKGATE_DIS_PSL(pipe), @@ -521,6 +533,20 @@ needs_modeset(const struct intel_crtc_state *state) return drm_atomic_crtc_needs_modeset(&state->base); } +bool +is_trans_port_sync_mode(const struct intel_crtc_state *crtc_state) +{ + return (crtc_state->master_transcoder != INVALID_TRANSCODER || + crtc_state->sync_mode_slaves_mask); +} + +static bool +is_trans_port_sync_master(const struct intel_crtc_state *crtc_state) +{ + return (crtc_state->master_transcoder == INVALID_TRANSCODER && + crtc_state->sync_mode_slaves_mask); +} + /* * Platform specific helpers to calculate the port PLL loopback- (clock.m), * and post-divider (clock.p) values, pre- (clock.vco) and post-divided fast @@ -1612,8 +1638,8 @@ void vlv_wait_port_ready(struct drm_i915_private *dev_priv, if (intel_de_wait_for_register(dev_priv, dpll_reg, port_mask, expected_mask, 1000)) - WARN(1, "timed out waiting for port %c ready: got 0x%x, expected 0x%x\n", - port_name(dport->base.port), + WARN(1, "timed out waiting for [ENCODER:%d:%s] port ready: got 0x%x, expected 0x%x\n", + dport->base.base.base.id, dport->base.base.name, I915_READ(dpll_reg) & port_mask, expected_mask); } @@ -2079,7 +2105,8 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb, unsigned int pinctl; u32 alignment; - WARN_ON(!mutex_is_locked(&dev->struct_mutex)); + if (WARN_ON(!i915_gem_object_is_framebuffer(obj))) + return ERR_PTR(-EINVAL); alignment = intel_surf_alignment(fb, 0); @@ -2161,8 +2188,6 @@ err: void intel_unpin_fb_vma(struct i915_vma *vma, unsigned long flags) { - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); - i915_gem_object_lock(vma->obj); if (flags & PLANE_HAS_FENCE) i915_vma_unpin_fence(vma); @@ -2504,6 +2529,9 @@ u32 intel_plane_fb_max_stride(struct drm_i915_private *dev_priv, * the highest stride limits of them all. */ crtc = intel_get_crtc_for_pipe(dev_priv, PIPE_A); + if (!crtc) + return 0; + plane = to_intel_plane(crtc->base.primary); return plane->max_stride(plane, pixel_format, modifier, @@ -2736,10 +2764,7 @@ intel_fill_fb_info(struct drm_i915_private *dev_priv, size++; /* rotate the x/y offsets to match the GTT view */ - r.x1 = x; - r.y1 = y; - r.x2 = x + width; - r.y2 = y + height; + drm_rect_init(&r, x, y, width, height); drm_rect_rotate(&r, rot_info->plane[i].width * tile_width, rot_info->plane[i].height * tile_height, @@ -2861,10 +2886,7 @@ intel_plane_remap_gtt(struct intel_plane_state *plane_state) struct drm_rect r; /* rotate the x/y offsets to match the GTT view */ - r.x1 = x; - r.y1 = y; - r.x2 = x + width; - r.y2 = y + height; + drm_rect_init(&r, x, y, width, height); drm_rect_rotate(&r, info->plane[i].width * tile_width, info->plane[i].height * tile_height, @@ -2966,6 +2988,8 @@ static int i9xx_format_to_fourcc(int format) return DRM_FORMAT_XRGB2101010; case DISPPLANE_RGBX101010: return DRM_FORMAT_XBGR2101010; + case DISPPLANE_RGBX161616: + return DRM_FORMAT_XBGR16161616F; } } @@ -3063,13 +3087,11 @@ intel_alloc_initial_plane_obj(struct intel_crtc *crtc, return false; } - mutex_lock(&dev->struct_mutex); obj = i915_gem_object_create_stolen_for_preallocated(dev_priv, base_aligned, base_aligned, size_aligned); - mutex_unlock(&dev->struct_mutex); - if (!obj) + if (IS_ERR(obj)) return false; switch (plane_config->tiling) { @@ -3151,6 +3173,7 @@ static void intel_plane_disable_noatomic(struct intel_crtc *crtc, intel_set_plane_visible(crtc_state, plane_state, false); fixup_active_planes(crtc_state); crtc_state->data_rate[plane->id] = 0; + crtc_state->min_cdclk[plane->id] = 0; if (plane->id == PLANE_PRIMARY) intel_pre_disable_primary_noatomic(&crtc->base); @@ -3230,13 +3253,11 @@ valid_fb: intel_state->color_plane[0].stride = intel_fb_pitch(fb, 0, intel_state->base.rotation); - mutex_lock(&dev->struct_mutex); intel_state->vma = intel_pin_and_fence_fb_obj(fb, &intel_state->view, intel_plane_uses_fence(intel_state), &intel_state->flags); - mutex_unlock(&dev->struct_mutex); if (IS_ERR(intel_state->vma)) { DRM_ERROR("failed to pin boot fb on pipe %d: %li\n", intel_crtc->pipe, PTR_ERR(intel_state->vma)); @@ -3280,7 +3301,20 @@ static int skl_max_plane_width(const struct drm_framebuffer *fb, switch (fb->modifier) { case DRM_FORMAT_MOD_LINEAR: case I915_FORMAT_MOD_X_TILED: - return 4096; + /* + * Validated limit is 4k, but has 5k should + * work apart from the following features: + * - Ytile (already limited to 4k) + * - FP16 (already limited to 4k) + * - render compression (already limited to 4k) + * - KVMR sprite and cursor (don't care) + * - horizontal panning (TODO verify this) + * - pipe and plane scaling (TODO verify this) + */ + if (cpp == 8) + return 4096; + else + return 5120; case I915_FORMAT_MOD_Y_TILED_CCS: case I915_FORMAT_MOD_Yf_TILED_CCS: /* FIXME AUX plane? */ @@ -3331,6 +3365,16 @@ static int icl_max_plane_width(const struct drm_framebuffer *fb, return 5120; } +static int skl_max_plane_height(void) +{ + return 4096; +} + +static int icl_max_plane_height(void) +{ + return 4320; +} + static bool skl_check_main_ccs_coordinates(struct intel_plane_state *plane_state, int main_x, int main_y, u32 main_offset) { @@ -3379,7 +3423,7 @@ static int skl_check_main_surface(struct intel_plane_state *plane_state) int w = drm_rect_width(&plane_state->base.src) >> 16; int h = drm_rect_height(&plane_state->base.src) >> 16; int max_width; - int max_height = 4096; + int max_height; u32 alignment, offset, aux_offset = plane_state->color_plane[1].offset; if (INTEL_GEN(dev_priv) >= 11) @@ -3389,6 +3433,11 @@ static int skl_check_main_surface(struct intel_plane_state *plane_state) else max_width = skl_max_plane_width(fb, 0, rotation); + if (INTEL_GEN(dev_priv) >= 11) + max_height = icl_max_plane_height(); + else + max_height = skl_max_plane_height(); + if (w > max_width || h > max_height) { DRM_DEBUG_KMS("requested Y/RGB source size %dx%d too big (limit %dx%d)\n", w, h, max_width, max_height); @@ -3455,9 +3504,8 @@ static int skl_check_main_surface(struct intel_plane_state *plane_state) * Put the final coordinates back so that the src * coordinate checks will see the right values. */ - drm_rect_translate(&plane_state->base.src, - (x << 16) - plane_state->base.src.x1, - (y << 16) - plane_state->base.src.y1); + drm_rect_translate_to(&plane_state->base.src, + x << 16, y << 16); return 0; } @@ -3528,7 +3576,7 @@ int skl_check_plane_surface(struct intel_plane_state *plane_state) * Handle the AUX surface first since * the main surface setup depends on it. */ - if (is_planar_yuv_format(fb->format->format)) { + if (drm_format_info_is_yuv_semiplanar(fb->format)) { ret = skl_check_nv12_aux_surface(plane_state); if (ret) return ret; @@ -3549,6 +3597,53 @@ int skl_check_plane_surface(struct intel_plane_state *plane_state) return 0; } +static void i9xx_plane_ratio(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state, + unsigned int *num, unsigned int *den) +{ + const struct drm_framebuffer *fb = plane_state->base.fb; + unsigned int cpp = fb->format->cpp[0]; + + /* + * g4x bspec says 64bpp pixel rate can't exceed 80% + * of cdclk when the sprite plane is enabled on the + * same pipe. ilk/snb bspec says 64bpp pixel rate is + * never allowed to exceed 80% of cdclk. Let's just go + * with the ilk/snb limit always. + */ + if (cpp == 8) { + *num = 10; + *den = 8; + } else { + *num = 1; + *den = 1; + } +} + +static int i9xx_plane_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) +{ + unsigned int pixel_rate; + unsigned int num, den; + + /* + * Note that crtc_state->pixel_rate accounts for both + * horizontal and vertical panel fitter downscaling factors. + * Pre-HSW bspec tells us to only consider the horizontal + * downscaling factor here. We ignore that and just consider + * both for simplicity. + */ + pixel_rate = crtc_state->pixel_rate; + + i9xx_plane_ratio(crtc_state, plane_state, &num, &den); + + /* two pixels per clock with double wide pipe */ + if (crtc_state->double_wide) + den *= 2; + + return DIV_ROUND_UP(pixel_rate * num, den); +} + unsigned int i9xx_plane_max_stride(struct intel_plane *plane, u32 pixel_format, u64 modifier, @@ -3631,6 +3726,9 @@ static u32 i9xx_plane_ctl(const struct intel_crtc_state *crtc_state, case DRM_FORMAT_XBGR2101010: dspcntr |= DISPPLANE_RGBX101010; break; + case DRM_FORMAT_XBGR16161616F: + dspcntr |= DISPPLANE_RGBX161616; + break; default: MISSING_CASE(fb->format->format); return 0; @@ -3653,7 +3751,8 @@ int i9xx_check_plane_surface(struct intel_plane_state *plane_state) { struct drm_i915_private *dev_priv = to_i915(plane_state->base.plane->dev); - int src_x, src_y; + const struct drm_framebuffer *fb = plane_state->base.fb; + int src_x, src_y, src_w; u32 offset; int ret; @@ -3664,9 +3763,14 @@ int i9xx_check_plane_surface(struct intel_plane_state *plane_state) if (!plane_state->base.visible) return 0; + src_w = drm_rect_width(&plane_state->base.src) >> 16; src_x = plane_state->base.src.x1 >> 16; src_y = plane_state->base.src.y1 >> 16; + /* Undocumented hardware limit on i965/g4x/vlv/chv */ + if (HAS_GMCH(dev_priv) && fb->format->cpp[0] == 8 && src_w > 2048) + return -EINVAL; + intel_add_fb_offsets(&src_x, &src_y, plane_state, 0); if (INTEL_GEN(dev_priv) >= 4) @@ -3679,9 +3783,8 @@ int i9xx_check_plane_surface(struct intel_plane_state *plane_state) * Put the final coordinates back so that the src * coordinate checks will see the right values. */ - drm_rect_translate(&plane_state->base.src, - (src_x << 16) - plane_state->base.src.x1, - (src_y << 16) - plane_state->base.src.y1); + drm_rect_translate_to(&plane_state->base.src, + src_x << 16, src_y << 16); /* HSW/BDW do this automagically in hardware */ if (!IS_HASWELL(dev_priv) && !IS_BROADWELL(dev_priv)) { @@ -4211,7 +4314,7 @@ __intel_display_resume(struct drm_device *dev, int i, ret; intel_modeset_setup_hw_state(dev, ctx); - i915_redisable_vga(to_i915(dev)); + intel_vga_redisable(to_i915(dev)); if (!state) return 0; @@ -4243,7 +4346,7 @@ __intel_display_resume(struct drm_device *dev, static bool gpu_reset_clobbers_display(struct drm_i915_private *dev_priv) { return (INTEL_INFO(dev_priv)->gpu_reset_clobbers_display && - intel_has_gpu_reset(dev_priv)); + intel_has_gpu_reset(&dev_priv->gt)); } void intel_prepare_reset(struct drm_i915_private *dev_priv) @@ -4330,7 +4433,7 @@ void intel_finish_reset(struct drm_i915_private *dev_priv) * so need a full re-initialization. */ intel_pps_unlock_regs_wa(dev_priv); - intel_modeset_init_hw(dev); + intel_modeset_init_hw(dev_priv); intel_init_clock_gating(dev_priv); spin_lock_irq(&dev_priv->irq_lock); @@ -4378,50 +4481,60 @@ static void icl_set_pipe_chicken(struct intel_crtc *crtc) I915_WRITE(PIPE_CHICKEN(pipe), tmp); } -static void intel_update_pipe_config(const struct intel_crtc_state *old_crtc_state, - const struct intel_crtc_state *new_crtc_state) +static void icl_enable_trans_port_sync(const struct intel_crtc_state *crtc_state) { - struct intel_crtc *crtc = to_intel_crtc(new_crtc_state->base.crtc); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - - /* drm_atomic_helper_update_legacy_modeset_state might not be called. */ - crtc->base.mode = new_crtc_state->base.mode; + u32 trans_ddi_func_ctl2_val; + u8 master_select; /* - * Update pipe size and adjust fitter if needed: the reason for this is - * that in compute_mode_changes we check the native mode (not the pfit - * mode) to see if we can flip rather than do a full mode set. In the - * fastboot case, we'll flip, but if we don't update the pipesrc and - * pfit state, we'll end up with a big fb scanned out into the wrong - * sized surface. + * Configure the master select and enable Transcoder Port Sync for + * Slave CRTCs transcoder. */ + if (crtc_state->master_transcoder == INVALID_TRANSCODER) + return; - I915_WRITE(PIPESRC(crtc->pipe), - ((new_crtc_state->pipe_src_w - 1) << 16) | - (new_crtc_state->pipe_src_h - 1)); + if (crtc_state->master_transcoder == TRANSCODER_EDP) + master_select = 0; + else + master_select = crtc_state->master_transcoder + 1; - /* on skylake this is done by detaching scalers */ - if (INTEL_GEN(dev_priv) >= 9) { - skl_detach_scalers(new_crtc_state); + /* Set the master select bits for Tranascoder Port Sync */ + trans_ddi_func_ctl2_val = (PORT_SYNC_MODE_MASTER_SELECT(master_select) & + PORT_SYNC_MODE_MASTER_SELECT_MASK) << + PORT_SYNC_MODE_MASTER_SELECT_SHIFT; + /* Enable Transcoder Port Sync */ + trans_ddi_func_ctl2_val |= PORT_SYNC_MODE_ENABLE; - if (new_crtc_state->pch_pfit.enabled) - skylake_pfit_enable(new_crtc_state); - } else if (HAS_PCH_SPLIT(dev_priv)) { - if (new_crtc_state->pch_pfit.enabled) - ironlake_pfit_enable(new_crtc_state); - else if (old_crtc_state->pch_pfit.enabled) - ironlake_pfit_disable(old_crtc_state); - } + I915_WRITE(TRANS_DDI_FUNC_CTL2(crtc_state->cpu_transcoder), + trans_ddi_func_ctl2_val); +} - if (INTEL_GEN(dev_priv) >= 11) - icl_set_pipe_chicken(crtc); +static void icl_disable_transcoder_port_sync(const struct intel_crtc_state *old_crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + i915_reg_t reg; + u32 trans_ddi_func_ctl2_val; + + if (old_crtc_state->master_transcoder == INVALID_TRANSCODER) + return; + + DRM_DEBUG_KMS("Disabling Transcoder Port Sync on Slave Transcoder %s\n", + transcoder_name(old_crtc_state->cpu_transcoder)); + + reg = TRANS_DDI_FUNC_CTL2(old_crtc_state->cpu_transcoder); + trans_ddi_func_ctl2_val = ~(PORT_SYNC_MODE_ENABLE | + PORT_SYNC_MODE_MASTER_SELECT_MASK); + I915_WRITE(reg, trans_ddi_func_ctl2_val); } static void intel_fdi_normal_train(struct intel_crtc *crtc) { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = crtc->pipe; + enum pipe pipe = crtc->pipe; i915_reg_t reg; u32 temp; @@ -4464,7 +4577,7 @@ static void ironlake_fdi_link_train(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = crtc->pipe; + enum pipe pipe = crtc->pipe; i915_reg_t reg; u32 temp, tries; @@ -4565,7 +4678,7 @@ static void gen6_fdi_link_train(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = crtc->pipe; + enum pipe pipe = crtc->pipe; i915_reg_t reg; u32 temp, i, retry; @@ -4698,7 +4811,7 @@ static void ivb_manual_fdi_link_train(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = crtc->pipe; + enum pipe pipe = crtc->pipe; i915_reg_t reg; u32 temp, i, j; @@ -4816,7 +4929,7 @@ static void ironlake_fdi_pll_enable(const struct intel_crtc_state *crtc_state) { struct intel_crtc *intel_crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(intel_crtc->base.dev); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; i915_reg_t reg; u32 temp; @@ -4853,7 +4966,7 @@ static void ironlake_fdi_pll_disable(struct intel_crtc *intel_crtc) { struct drm_device *dev = intel_crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; i915_reg_t reg; u32 temp; @@ -4884,7 +4997,7 @@ static void ironlake_fdi_disable(struct drm_crtc *crtc) struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; i915_reg_t reg; u32 temp; @@ -5199,7 +5312,7 @@ static void ironlake_pch_enable(const struct intel_atomic_state *state, struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = crtc->pipe; + enum pipe pipe = crtc->pipe; u32 temp; assert_pch_transcoder_disabled(dev_priv, pipe); @@ -5294,7 +5407,7 @@ static void lpt_pch_enable(const struct intel_atomic_state *state, lpt_enable_pch_transcoder(dev_priv, cpu_transcoder); } -static void cpt_verify_modeset(struct drm_device *dev, int pipe) +static void cpt_verify_modeset(struct drm_device *dev, enum pipe pipe) { struct drm_i915_private *dev_priv = to_i915(dev); i915_reg_t dslreg = PIPEDSL(pipe); @@ -5446,7 +5559,7 @@ skl_update_scaler(struct intel_crtc_state *crtc_state, bool force_detach, return 0; } - if (format && is_planar_yuv_format(format->format) && + if (format && drm_format_info_is_yuv_semiplanar(format) && (src_h < SKL_MIN_YUV_420_SRC_H || src_w < SKL_MIN_YUV_420_SRC_W)) { DRM_DEBUG_KMS("Planar YUV: src dimensions not met\n"); return -EINVAL; @@ -5523,7 +5636,7 @@ static int skl_update_scaler_plane(struct intel_crtc_state *crtc_state, /* Pre-gen11 and SDR planes always need a scaler for planar formats. */ if (!icl_is_hdr_plane(dev_priv, intel_plane->id) && - fb && is_planar_yuv_format(fb->format->format)) + fb && drm_format_info_is_yuv_semiplanar(fb->format)) need_scaler = true; ret = skl_update_scaler(crtc_state, force_detach, @@ -5555,10 +5668,6 @@ static int skl_update_scaler_plane(struct intel_crtc_state *crtc_state, case DRM_FORMAT_ARGB8888: case DRM_FORMAT_XRGB2101010: case DRM_FORMAT_XBGR2101010: - case DRM_FORMAT_XBGR16161616F: - case DRM_FORMAT_ABGR16161616F: - case DRM_FORMAT_XRGB16161616F: - case DRM_FORMAT_ARGB16161616F: case DRM_FORMAT_YUYV: case DRM_FORMAT_YVYU: case DRM_FORMAT_UYVY: @@ -5574,6 +5683,13 @@ static int skl_update_scaler_plane(struct intel_crtc_state *crtc_state, case DRM_FORMAT_XVYU12_16161616: case DRM_FORMAT_XVYU16161616: break; + case DRM_FORMAT_XBGR16161616F: + case DRM_FORMAT_ABGR16161616F: + case DRM_FORMAT_XRGB16161616F: + case DRM_FORMAT_ARGB16161616F: + if (INTEL_GEN(dev_priv) >= 11) + break; + /* fall through */ default: DRM_DEBUG_KMS("[PLANE:%d:%s] FB:%d unsupported scaling format 0x%x\n", intel_plane->base.base.id, intel_plane->base.name, @@ -5633,7 +5749,7 @@ static void ironlake_pfit_enable(const struct intel_crtc_state *crtc_state) { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - int pipe = crtc->pipe; + enum pipe pipe = crtc->pipe; if (crtc_state->pch_pfit.enabled) { /* Force use of hard-coded filter coefficients @@ -5715,13 +5831,8 @@ void hsw_disable_ips(const struct intel_crtc_state *crtc_state) static void intel_crtc_dpms_overlay_disable(struct intel_crtc *intel_crtc) { - if (intel_crtc->overlay) { - struct drm_device *dev = intel_crtc->base.dev; - - mutex_lock(&dev->struct_mutex); + if (intel_crtc->overlay) (void) intel_overlay_switch_off(intel_crtc->overlay); - mutex_unlock(&dev->struct_mutex); - } /* Let userspace switch the overlay on again. In most cases userspace * has to recompute where to put it anyway. @@ -5746,7 +5857,7 @@ intel_post_enable_primary(struct drm_crtc *crtc, struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; /* * Gen2 reports pipe underruns whenever all planes are disabled. @@ -5770,7 +5881,7 @@ intel_pre_disable_primary_noatomic(struct drm_crtc *crtc) struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; /* * Gen2 reports pipe underruns whenever all planes are disabled. @@ -6293,7 +6404,7 @@ static void ironlake_crtc_enable(struct intel_crtc_state *pipe_config, struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; if (WARN_ON(intel_crtc->active)) return; @@ -6426,7 +6537,7 @@ static void haswell_crtc_enable(struct intel_crtc_state *pipe_config, struct drm_crtc *crtc = pipe_config->base.crtc; struct drm_i915_private *dev_priv = to_i915(crtc->dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe, hsw_workaround_pipe; + enum pipe pipe = intel_crtc->pipe, hsw_workaround_pipe; enum transcoder cpu_transcoder = pipe_config->cpu_transcoder; bool psl_clkgate_wa; @@ -6446,6 +6557,9 @@ static void haswell_crtc_enable(struct intel_crtc_state *pipe_config, if (!transcoder_is_dsi(cpu_transcoder)) intel_set_pipe_timings(pipe_config); + if (INTEL_GEN(dev_priv) >= 11) + icl_enable_trans_port_sync(pipe_config); + intel_set_pipe_src_size(pipe_config); if (cpu_transcoder != TRANSCODER_EDP && @@ -6491,7 +6605,6 @@ static void haswell_crtc_enable(struct intel_crtc_state *pipe_config, if (INTEL_GEN(dev_priv) >= 11) icl_set_pipe_chicken(intel_crtc); - intel_ddi_set_pipe_settings(pipe_config); if (!transcoder_is_dsi(cpu_transcoder)) intel_ddi_enable_transcoder_func(pipe_config); @@ -6552,7 +6665,7 @@ static void ironlake_crtc_disable(struct intel_crtc_state *old_crtc_state, struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; /* * Sometimes spurious CPU pipe underruns happen when the @@ -6624,6 +6737,9 @@ static void haswell_crtc_disable(struct intel_crtc_state *old_crtc_state, if (intel_crtc_has_type(old_crtc_state, INTEL_OUTPUT_DP_MST)) intel_ddi_set_vc_payload_alloc(old_crtc_state, false); + if (INTEL_GEN(dev_priv) >= 11) + icl_disable_transcoder_port_sync(old_crtc_state); + if (!transcoder_is_dsi(cpu_transcoder)) intel_ddi_disable_transcoder_func(old_crtc_state); @@ -6721,6 +6837,8 @@ enum intel_display_power_domain intel_port_to_power_domain(enum port port) return POWER_DOMAIN_PORT_DDI_E_LANES; case PORT_F: return POWER_DOMAIN_PORT_DDI_F_LANES; + case PORT_G: + return POWER_DOMAIN_PORT_DDI_G_LANES; default: MISSING_CASE(port); return POWER_DOMAIN_PORT_OTHER; @@ -6737,16 +6855,18 @@ intel_aux_power_domain(struct intel_digital_port *dig_port) dig_port->tc_mode == TC_PORT_TBT_ALT) { switch (dig_port->aux_ch) { case AUX_CH_C: - return POWER_DOMAIN_AUX_TBT1; + return POWER_DOMAIN_AUX_C_TBT; case AUX_CH_D: - return POWER_DOMAIN_AUX_TBT2; + return POWER_DOMAIN_AUX_D_TBT; case AUX_CH_E: - return POWER_DOMAIN_AUX_TBT3; + return POWER_DOMAIN_AUX_E_TBT; case AUX_CH_F: - return POWER_DOMAIN_AUX_TBT4; + return POWER_DOMAIN_AUX_F_TBT; + case AUX_CH_G: + return POWER_DOMAIN_AUX_G_TBT; default: MISSING_CASE(dig_port->aux_ch); - return POWER_DOMAIN_AUX_TBT1; + return POWER_DOMAIN_AUX_C_TBT; } } @@ -6763,6 +6883,8 @@ intel_aux_power_domain(struct intel_digital_port *dig_port) return POWER_DOMAIN_AUX_E; case AUX_CH_F: return POWER_DOMAIN_AUX_F; + case AUX_CH_G: + return POWER_DOMAIN_AUX_G; default: MISSING_CASE(dig_port->aux_ch); return POWER_DOMAIN_AUX_A; @@ -6839,7 +6961,7 @@ static void valleyview_crtc_enable(struct intel_crtc_state *pipe_config, struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; if (WARN_ON(intel_crtc->active)) return; @@ -6971,7 +7093,7 @@ static void i9xx_crtc_disable(struct intel_crtc_state *old_crtc_state, struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; /* * On gen2 planes are double buffered but the pipe isn't, so we must @@ -7080,7 +7202,7 @@ static void intel_crtc_disable_noatomic(struct drm_crtc *crtc, intel_display_power_put_unchecked(dev_priv, domain); intel_crtc->enabled_power_domains = 0; - dev_priv->active_crtcs &= ~(1 << intel_crtc->pipe); + dev_priv->active_pipes &= ~BIT(intel_crtc->pipe); dev_priv->min_cdclk[intel_crtc->pipe] = 0; dev_priv->min_voltage_level[intel_crtc->pipe] = 0; @@ -7188,7 +7310,7 @@ static int ironlake_check_fdi_lanes(struct drm_device *dev, enum pipe pipe, } } - if (INTEL_INFO(dev_priv)->num_pipes == 2) + if (INTEL_NUM_PIPES(dev_priv) == 2) return 0; /* Ivybridge 3 pipe is really complicated */ @@ -7261,7 +7383,7 @@ retry: pipe_config->fdi_lanes = lane; intel_link_compute_m_n(pipe_config->pipe_bpp, lane, fdi_dotclock, - link_bw, &pipe_config->fdi_m_n, false); + link_bw, &pipe_config->fdi_m_n, false, false); ret = ironlake_check_fdi_lanes(dev, intel_crtc->pipe, pipe_config); if (ret == -EDEADLK) @@ -7508,11 +7630,15 @@ void intel_link_compute_m_n(u16 bits_per_pixel, int nlanes, int pixel_clock, int link_clock, struct intel_link_m_n *m_n, - bool constant_n) + bool constant_n, bool fec_enable) { - m_n->tu = 64; + u32 data_clock = bits_per_pixel * pixel_clock; + + if (fec_enable) + data_clock = intel_dp_mode_to_fec_clock(data_clock); - compute_m_n(bits_per_pixel * pixel_clock, + m_n->tu = 64; + compute_m_n(data_clock, link_clock * nlanes * 8, &m_n->gmch_m, &m_n->gmch_n, constant_n); @@ -7522,6 +7648,27 @@ intel_link_compute_m_n(u16 bits_per_pixel, int nlanes, constant_n); } +static void intel_panel_sanitize_ssc(struct drm_i915_private *dev_priv) +{ + /* + * There may be no VBT; and if the BIOS enabled SSC we can + * just keep using it to avoid unnecessary flicker. Whereas if the + * BIOS isn't using it, don't assume it will work even if the VBT + * indicates as much. + */ + if (HAS_PCH_IBX(dev_priv) || HAS_PCH_CPT(dev_priv)) { + bool bios_lvds_use_ssc = I915_READ(PCH_DREF_CONTROL) & + DREF_SSC1_ENABLE; + + if (dev_priv->vbt.lvds_use_ssc != bios_lvds_use_ssc) { + DRM_DEBUG_KMS("SSC %s by BIOS, overriding VBT which says %s\n", + enableddisabled(bios_lvds_use_ssc), + enableddisabled(dev_priv->vbt.lvds_use_ssc)); + dev_priv->vbt.lvds_use_ssc = bios_lvds_use_ssc; + } + } +} + static inline bool intel_panel_use_ssc(struct drm_i915_private *dev_priv) { if (i915_modparams.panel_use_ssc >= 0) @@ -8173,6 +8320,21 @@ static void intel_set_pipe_src_size(const struct intel_crtc_state *crtc_state) (crtc_state->pipe_src_h - 1)); } +static bool intel_pipe_is_interlaced(const struct intel_crtc_state *crtc_state) +{ + struct drm_i915_private *dev_priv = to_i915(crtc_state->base.crtc->dev); + enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; + + if (IS_GEN(dev_priv, 2)) + return false; + + if (INTEL_GEN(dev_priv) >= 9 || + IS_BROADWELL(dev_priv) || IS_HASWELL(dev_priv)) + return I915_READ(PIPECONF(cpu_transcoder)) & PIPECONF_INTERLACE_MASK_HSW; + else + return I915_READ(PIPECONF(cpu_transcoder)) & PIPECONF_INTERLACE_MASK; +} + static void intel_get_pipe_timings(struct intel_crtc *crtc, struct intel_crtc_state *pipe_config) { @@ -8211,7 +8373,7 @@ static void intel_get_pipe_timings(struct intel_crtc *crtc, pipe_config->base.adjusted_mode.crtc_vsync_start = (tmp & 0xffff) + 1; pipe_config->base.adjusted_mode.crtc_vsync_end = ((tmp >> 16) & 0xffff) + 1; - if (I915_READ(PIPECONF(cpu_transcoder)) & PIPECONF_INTERLACE_MASK) { + if (intel_pipe_is_interlaced(pipe_config)) { pipe_config->base.adjusted_mode.flags |= DRM_MODE_FLAG_INTERLACE; pipe_config->base.adjusted_mode.crtc_vtotal += 1; pipe_config->base.adjusted_mode.crtc_vblank_end += 1; @@ -8543,7 +8705,7 @@ static void vlv_crtc_clock_get(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = pipe_config->cpu_transcoder; + enum pipe pipe = crtc->pipe; struct dpll clock; u32 mdiv; int refclk = 100000; @@ -8653,7 +8815,7 @@ static void chv_crtc_clock_get(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = pipe_config->cpu_transcoder; + enum pipe pipe = crtc->pipe; enum dpio_channel port = vlv_pipe_to_channel(pipe); struct dpll clock; u32 cmn_dw13, pll_dw0, pll_dw1, pll_dw2, pll_dw3; @@ -8682,47 +8844,24 @@ static void chv_crtc_clock_get(struct intel_crtc *crtc, pipe_config->port_clock = chv_calc_dpll_params(refclk, &clock); } -static void intel_get_crtc_ycbcr_config(struct intel_crtc *crtc, - struct intel_crtc_state *pipe_config) +static enum intel_output_format +bdw_get_pipemisc_output_format(struct intel_crtc *crtc) { struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - enum intel_output_format output = INTEL_OUTPUT_FORMAT_RGB; - - pipe_config->lspcon_downsampling = false; + u32 tmp; - if (IS_BROADWELL(dev_priv) || INTEL_GEN(dev_priv) >= 9) { - u32 tmp = I915_READ(PIPEMISC(crtc->pipe)); + tmp = I915_READ(PIPEMISC(crtc->pipe)); - if (tmp & PIPEMISC_OUTPUT_COLORSPACE_YUV) { - bool ycbcr420_enabled = tmp & PIPEMISC_YUV420_ENABLE; - bool blend = tmp & PIPEMISC_YUV420_MODE_FULL_BLEND; + if (tmp & PIPEMISC_YUV420_ENABLE) { + /* We support 4:2:0 in full blend mode only */ + WARN_ON((tmp & PIPEMISC_YUV420_MODE_FULL_BLEND) == 0); - if (ycbcr420_enabled) { - /* We support 4:2:0 in full blend mode only */ - if (!blend) - output = INTEL_OUTPUT_FORMAT_INVALID; - else if (!(IS_GEMINILAKE(dev_priv) || - INTEL_GEN(dev_priv) >= 10)) - output = INTEL_OUTPUT_FORMAT_INVALID; - else - output = INTEL_OUTPUT_FORMAT_YCBCR420; - } else { - /* - * Currently there is no interface defined to - * check user preference between RGB/YCBCR444 - * or YCBCR420. So the only possible case for - * YCBCR444 usage is driving YCBCR420 output - * with LSPCON, when pipe is configured for - * YCBCR444 output and LSPCON takes care of - * downsampling it. - */ - pipe_config->lspcon_downsampling = true; - output = INTEL_OUTPUT_FORMAT_YCBCR444; - } - } + return INTEL_OUTPUT_FORMAT_YCBCR420; + } else if (tmp & PIPEMISC_OUTPUT_COLORSPACE_YUV) { + return INTEL_OUTPUT_FORMAT_YCBCR444; + } else { + return INTEL_OUTPUT_FORMAT_RGB; } - - pipe_config->output_format = output; } static void i9xx_get_pipe_color_config(struct intel_crtc_state *crtc_state) @@ -8760,6 +8899,7 @@ static bool i9xx_get_pipe_config(struct intel_crtc *crtc, pipe_config->output_format = INTEL_OUTPUT_FORMAT_RGB; pipe_config->cpu_transcoder = (enum transcoder) crtc->pipe; pipe_config->shared_dpll = NULL; + pipe_config->master_transcoder = INVALID_TRANSCODER; ret = false; @@ -9298,7 +9438,6 @@ static bool wrpll_uses_pch_ssc(struct drm_i915_private *dev_priv, static void lpt_init_pch_refclk(struct drm_i915_private *dev_priv) { struct intel_encoder *encoder; - bool pch_ssc_in_use = false; bool has_fdi = false; for_each_intel_encoder(&dev_priv->drm, encoder) { @@ -9326,22 +9465,24 @@ static void lpt_init_pch_refclk(struct drm_i915_private *dev_priv) * clock hierarchy. That would also allow us to do * clock bending finally. */ + dev_priv->pch_ssc_use = 0; + if (spll_uses_pch_ssc(dev_priv)) { DRM_DEBUG_KMS("SPLL using PCH SSC\n"); - pch_ssc_in_use = true; + dev_priv->pch_ssc_use |= BIT(DPLL_ID_SPLL); } if (wrpll_uses_pch_ssc(dev_priv, DPLL_ID_WRPLL1)) { DRM_DEBUG_KMS("WRPLL1 using PCH SSC\n"); - pch_ssc_in_use = true; + dev_priv->pch_ssc_use |= BIT(DPLL_ID_WRPLL1); } if (wrpll_uses_pch_ssc(dev_priv, DPLL_ID_WRPLL2)) { DRM_DEBUG_KMS("WRPLL2 using PCH SSC\n"); - pch_ssc_in_use = true; + dev_priv->pch_ssc_use |= BIT(DPLL_ID_WRPLL2); } - if (pch_ssc_in_use) + if (dev_priv->pch_ssc_use) return; if (has_fdi) { @@ -9398,9 +9539,19 @@ static void ironlake_set_pipeconf(const struct intel_crtc_state *crtc_state) else val |= PIPECONF_PROGRESSIVE; + /* + * This would end up with an odd purple hue over + * the entire display. Make sure we don't do it. + */ + WARN_ON(crtc_state->limited_color_range && + crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB); + if (crtc_state->limited_color_range) val |= PIPECONF_COLOR_RANGE_SELECT; + if (crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB) + val |= PIPECONF_OUTPUT_COLORSPACE_YUV709; + val |= PIPECONF_GAMMA_MODE(crtc_state->gamma_mode); I915_WRITE(PIPECONF(pipe), val); @@ -9422,6 +9573,10 @@ static void haswell_set_pipeconf(const struct intel_crtc_state *crtc_state) else val |= PIPECONF_PROGRESSIVE; + if (IS_HASWELL(dev_priv) && + crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB) + val |= PIPECONF_OUTPUT_COLORSPACE_YUV_HSW; + I915_WRITE(PIPECONF(cpu_transcoder), val); POSTING_READ(PIPECONF(cpu_transcoder)); } @@ -9572,7 +9727,7 @@ static void ironlake_compute_dpll(struct intel_crtc *crtc, * clear if it''s a win or loss power wise. No point in doing * this on ILK at all since it has a fixed DPLL<->pipe mapping. */ - if (INTEL_INFO(dev_priv)->num_pipes == 3 && + if (INTEL_NUM_PIPES(dev_priv) == 3 && intel_crtc_has_type(crtc_state, INTEL_OUTPUT_ANALOG)) dpll |= DPLL_SDVO_HIGH_SPEED; @@ -9871,8 +10026,8 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, offset = I915_READ(PLANE_OFFSET(pipe, plane_id)); val = I915_READ(PLANE_SIZE(pipe, plane_id)); - fb->height = ((val >> 16) & 0xfff) + 1; - fb->width = ((val >> 0) & 0x1fff) + 1; + fb->height = ((val >> 16) & 0xffff) + 1; + fb->width = ((val >> 0) & 0xffff) + 1; val = I915_READ(PLANE_STRIDE(pipe, plane_id)); stride_mult = skl_plane_stride_mult(fb, 0, DRM_MODE_ROTATE_0); @@ -9933,9 +10088,9 @@ static bool ironlake_get_pipe_config(struct intel_crtc *crtc, if (!wakeref) return false; - pipe_config->output_format = INTEL_OUTPUT_FORMAT_RGB; pipe_config->cpu_transcoder = (enum transcoder) crtc->pipe; pipe_config->shared_dpll = NULL; + pipe_config->master_transcoder = INVALID_TRANSCODER; ret = false; tmp = I915_READ(PIPECONF(crtc->pipe)); @@ -9962,6 +10117,16 @@ static bool ironlake_get_pipe_config(struct intel_crtc *crtc, if (tmp & PIPECONF_COLOR_RANGE_SELECT) pipe_config->limited_color_range = true; + switch (tmp & PIPECONF_OUTPUT_COLORSPACE_MASK) { + case PIPECONF_OUTPUT_COLORSPACE_YUV601: + case PIPECONF_OUTPUT_COLORSPACE_YUV709: + pipe_config->output_format = INTEL_OUTPUT_FORMAT_YCBCR444; + break; + default: + pipe_config->output_format = INTEL_OUTPUT_FORMAT_RGB; + break; + } + pipe_config->gamma_mode = (tmp & PIPECONF_GAMMA_MODE_MASK_ILK) >> PIPECONF_GAMMA_MODE_SHIFT; @@ -10376,6 +10541,59 @@ static void haswell_get_ddi_port_state(struct intel_crtc *crtc, } } +static enum transcoder transcoder_master_readout(struct drm_i915_private *dev_priv, + enum transcoder cpu_transcoder) +{ + u32 trans_port_sync, master_select; + + trans_port_sync = I915_READ(TRANS_DDI_FUNC_CTL2(cpu_transcoder)); + + if ((trans_port_sync & PORT_SYNC_MODE_ENABLE) == 0) + return INVALID_TRANSCODER; + + master_select = trans_port_sync & + PORT_SYNC_MODE_MASTER_SELECT_MASK; + if (master_select == 0) + return TRANSCODER_EDP; + else + return master_select - 1; +} + +static void icelake_get_trans_port_sync_config(struct intel_crtc_state *crtc_state) +{ + struct drm_i915_private *dev_priv = to_i915(crtc_state->base.crtc->dev); + u32 transcoders; + enum transcoder cpu_transcoder; + + crtc_state->master_transcoder = transcoder_master_readout(dev_priv, + crtc_state->cpu_transcoder); + + transcoders = BIT(TRANSCODER_A) | + BIT(TRANSCODER_B) | + BIT(TRANSCODER_C) | + BIT(TRANSCODER_D); + for_each_cpu_transcoder_masked(dev_priv, cpu_transcoder, transcoders) { + enum intel_display_power_domain power_domain; + intel_wakeref_t trans_wakeref; + + power_domain = POWER_DOMAIN_TRANSCODER(cpu_transcoder); + trans_wakeref = intel_display_power_get_if_enabled(dev_priv, + power_domain); + + if (!trans_wakeref) + continue; + + if (transcoder_master_readout(dev_priv, cpu_transcoder) == + crtc_state->cpu_transcoder) + crtc_state->sync_mode_slaves_mask |= BIT(cpu_transcoder); + + intel_display_power_put(dev_priv, power_domain, trans_wakeref); + } + + WARN_ON(crtc_state->master_transcoder != INVALID_TRANSCODER && + crtc_state->sync_mode_slaves_mask); +} + static bool haswell_get_pipe_config(struct intel_crtc *crtc, struct intel_crtc_state *pipe_config) { @@ -10387,6 +10605,8 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, intel_crtc_init_scalers(crtc, pipe_config); + pipe_config->master_transcoder = INVALID_TRANSCODER; + power_domain = POWER_DOMAIN_PIPE(crtc->pipe); wf = intel_display_power_get_if_enabled(dev_priv, power_domain); if (!wf) @@ -10417,7 +10637,30 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, } intel_get_pipe_src_size(crtc, pipe_config); - intel_get_crtc_ycbcr_config(crtc, pipe_config); + + if (IS_HASWELL(dev_priv)) { + u32 tmp = I915_READ(PIPECONF(pipe_config->cpu_transcoder)); + + if (tmp & PIPECONF_OUTPUT_COLORSPACE_YUV_HSW) + pipe_config->output_format = INTEL_OUTPUT_FORMAT_YCBCR444; + else + pipe_config->output_format = INTEL_OUTPUT_FORMAT_RGB; + } else { + pipe_config->output_format = + bdw_get_pipemisc_output_format(crtc); + + /* + * Currently there is no interface defined to + * check user preference between RGB/YCBCR444 + * or YCBCR420. So the only possible case for + * YCBCR444 usage is driving YCBCR420 output + * with LSPCON, when pipe is configured for + * YCBCR444 output and LSPCON takes care of + * downsampling it. + */ + pipe_config->lspcon_downsampling = + pipe_config->output_format == INTEL_OUTPUT_FORMAT_YCBCR444; + } pipe_config->gamma_mode = I915_READ(GAMMA_MODE(crtc->pipe)); @@ -10472,6 +10715,10 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, pipe_config->pixel_multiplier = 1; } + if (INTEL_GEN(dev_priv) >= 11 && + !transcoder_is_dsi(pipe_config->cpu_transcoder)) + icelake_get_trans_port_sync_config(pipe_config); + out: for_each_power_domain(power_domain, power_domain_mask) intel_display_power_put(dev_priv, @@ -10493,21 +10740,13 @@ static u32 intel_cursor_base(const struct intel_plane_state *plane_state) else base = intel_plane_ggtt_offset(plane_state); - base += plane_state->color_plane[0].offset; - - /* ILK+ do this automagically */ - if (HAS_GMCH(dev_priv) && - plane_state->base.rotation & DRM_MODE_ROTATE_180) - base += (plane_state->base.crtc_h * - plane_state->base.crtc_w - 1) * fb->format->cpp[0]; - - return base; + return base + plane_state->color_plane[0].offset; } static u32 intel_cursor_position(const struct intel_plane_state *plane_state) { - int x = plane_state->base.crtc_x; - int y = plane_state->base.crtc_y; + int x = plane_state->base.dst.x1; + int y = plane_state->base.dst.y1; u32 pos = 0; if (x < 0) { @@ -10529,8 +10768,8 @@ static bool intel_cursor_size_ok(const struct intel_plane_state *plane_state) { const struct drm_mode_config *config = &plane_state->base.plane->dev->mode_config; - int width = plane_state->base.crtc_w; - int height = plane_state->base.crtc_h; + int width = drm_rect_width(&plane_state->base.dst); + int height = drm_rect_height(&plane_state->base.dst); return width > 0 && width <= config->cursor_width && height > 0 && height <= config->cursor_height; @@ -10538,6 +10777,9 @@ static bool intel_cursor_size_ok(const struct intel_plane_state *plane_state) static int intel_cursor_check_surface(struct intel_plane_state *plane_state) { + struct drm_i915_private *dev_priv = + to_i915(plane_state->base.plane->dev); + unsigned int rotation = plane_state->base.rotation; int src_x, src_y; u32 offset; int ret; @@ -10549,8 +10791,8 @@ static int intel_cursor_check_surface(struct intel_plane_state *plane_state) if (!plane_state->base.visible) return 0; - src_x = plane_state->base.src_x >> 16; - src_y = plane_state->base.src_y >> 16; + src_x = plane_state->base.src.x1 >> 16; + src_y = plane_state->base.src.y1 >> 16; intel_add_fb_offsets(&src_x, &src_y, plane_state, 0); offset = intel_plane_compute_aligned_offset(&src_x, &src_y, @@ -10561,7 +10803,25 @@ static int intel_cursor_check_surface(struct intel_plane_state *plane_state) return -EINVAL; } + /* + * Put the final coordinates back so that the src + * coordinate checks will see the right values. + */ + drm_rect_translate_to(&plane_state->base.src, + src_x << 16, src_y << 16); + + /* ILK+ do this automagically in hardware */ + if (HAS_GMCH(dev_priv) && rotation & DRM_MODE_ROTATE_180) { + const struct drm_framebuffer *fb = plane_state->base.fb; + int src_w = drm_rect_width(&plane_state->base.src) >> 16; + int src_h = drm_rect_height(&plane_state->base.src) >> 16; + + offset += (src_h * src_w - 1) * fb->format->cpp[0]; + } + plane_state->color_plane[0].offset = offset; + plane_state->color_plane[0].x = src_x; + plane_state->color_plane[0].y = src_y; return 0; } @@ -10585,6 +10845,10 @@ static int intel_check_cursor(struct intel_crtc_state *crtc_state, if (ret) return ret; + /* Use the unclipped src/dst rectangles, which we program to hw */ + plane_state->base.src = drm_plane_state_src(&plane_state->base); + plane_state->base.dst = drm_plane_state_dest(&plane_state->base); + ret = intel_cursor_check_surface(plane_state); if (ret) return ret; @@ -10627,7 +10891,7 @@ static u32 i845_cursor_ctl(const struct intel_crtc_state *crtc_state, static bool i845_cursor_size_ok(const struct intel_plane_state *plane_state) { - int width = plane_state->base.crtc_w; + int width = drm_rect_width(&plane_state->base.dst); /* * 845g/865g are only limited by the width of their cursors, @@ -10653,8 +10917,8 @@ static int i845_check_cursor(struct intel_crtc_state *crtc_state, /* Check for which cursor types we support */ if (!i845_cursor_size_ok(plane_state)) { DRM_DEBUG("Cursor dimension %dx%d not supported\n", - plane_state->base.crtc_w, - plane_state->base.crtc_h); + drm_rect_width(&plane_state->base.dst), + drm_rect_height(&plane_state->base.dst)); return -EINVAL; } @@ -10687,8 +10951,8 @@ static void i845_update_cursor(struct intel_plane *plane, unsigned long irqflags; if (plane_state && plane_state->base.visible) { - unsigned int width = plane_state->base.crtc_w; - unsigned int height = plane_state->base.crtc_h; + unsigned int width = drm_rect_width(&plane_state->base.dst); + unsigned int height = drm_rect_height(&plane_state->base.dst); cntl = plane_state->ctl | i845_cursor_ctl_crtc(crtc_state); @@ -10790,7 +11054,7 @@ static u32 i9xx_cursor_ctl(const struct intel_crtc_state *crtc_state, if (IS_GEN(dev_priv, 6) || IS_IVYBRIDGE(dev_priv)) cntl |= MCURSOR_TRICKLE_FEED_DISABLE; - switch (plane_state->base.crtc_w) { + switch (drm_rect_width(&plane_state->base.dst)) { case 64: cntl |= MCURSOR_MODE_64_ARGB_AX; break; @@ -10801,7 +11065,7 @@ static u32 i9xx_cursor_ctl(const struct intel_crtc_state *crtc_state, cntl |= MCURSOR_MODE_256_ARGB_AX; break; default: - MISSING_CASE(plane_state->base.crtc_w); + MISSING_CASE(drm_rect_width(&plane_state->base.dst)); return 0; } @@ -10815,8 +11079,8 @@ static bool i9xx_cursor_size_ok(const struct intel_plane_state *plane_state) { struct drm_i915_private *dev_priv = to_i915(plane_state->base.plane->dev); - int width = plane_state->base.crtc_w; - int height = plane_state->base.crtc_h; + int width = drm_rect_width(&plane_state->base.dst); + int height = drm_rect_height(&plane_state->base.dst); if (!intel_cursor_size_ok(plane_state)) return false; @@ -10869,17 +11133,19 @@ static int i9xx_check_cursor(struct intel_crtc_state *crtc_state, /* Check for which cursor types we support */ if (!i9xx_cursor_size_ok(plane_state)) { DRM_DEBUG("Cursor dimension %dx%d not supported\n", - plane_state->base.crtc_w, - plane_state->base.crtc_h); + drm_rect_width(&plane_state->base.dst), + drm_rect_height(&plane_state->base.dst)); return -EINVAL; } WARN_ON(plane_state->base.visible && plane_state->color_plane[0].stride != fb->pitches[0]); - if (fb->pitches[0] != plane_state->base.crtc_w * fb->format->cpp[0]) { + if (fb->pitches[0] != + drm_rect_width(&plane_state->base.dst) * fb->format->cpp[0]) { DRM_DEBUG_KMS("Invalid cursor stride (%u) (cursor width %d)\n", - fb->pitches[0], plane_state->base.crtc_w); + fb->pitches[0], + drm_rect_width(&plane_state->base.dst)); return -EINVAL; } @@ -10894,7 +11160,7 @@ static int i9xx_check_cursor(struct intel_crtc_state *crtc_state, * Refuse the put the cursor into that compromised position. */ if (IS_CHERRYVIEW(dev_priv) && pipe == PIPE_C && - plane_state->base.visible && plane_state->base.crtc_x < 0) { + plane_state->base.visible && plane_state->base.dst.x1 < 0) { DRM_DEBUG_KMS("CHV cursor C not allowed to straddle the left screen edge\n"); return -EINVAL; } @@ -10914,11 +11180,14 @@ static void i9xx_update_cursor(struct intel_plane *plane, unsigned long irqflags; if (plane_state && plane_state->base.visible) { + unsigned width = drm_rect_width(&plane_state->base.dst); + unsigned height = drm_rect_height(&plane_state->base.dst); + cntl = plane_state->ctl | i9xx_cursor_ctl_crtc(crtc_state); - if (plane_state->base.crtc_h != plane_state->base.crtc_w) - fbc_ctl = CUR_FBC_CTL_EN | (plane_state->base.crtc_h - 1); + if (width != height) + fbc_ctl = CUR_FBC_CTL_EN | (height - 1); base = intel_cursor_base(plane_state); pos = intel_cursor_position(plane_state); @@ -11063,7 +11332,6 @@ static int intel_modeset_disable_planes(struct drm_atomic_state *state, } int intel_get_load_detect_pipe(struct drm_connector *connector, - const struct drm_display_mode *mode, struct intel_load_detect_pipe *old, struct drm_modeset_acquire_ctx *ctx) { @@ -11170,10 +11438,8 @@ found: crtc_state->base.active = crtc_state->base.enable = true; - if (!mode) - mode = &load_detect_mode; - - ret = drm_atomic_set_mode_for_crtc(&crtc_state->base, mode); + ret = drm_atomic_set_mode_for_crtc(&crtc_state->base, + &load_detect_mode); if (ret) goto fail; @@ -11265,7 +11531,7 @@ static void i9xx_crtc_clock_get(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - int pipe = pipe_config->cpu_transcoder; + enum pipe pipe = crtc->pipe; u32 dpll = pipe_config->dpll_hw_state.dpll; u32 fp; struct dpll clock; @@ -11489,7 +11755,6 @@ int intel_plane_atomic_calc_changes(const struct intel_crtc_state *old_crtc_stat bool was_crtc_enabled = old_crtc_state->base.active; bool is_crtc_enabled = crtc_state->base.active; bool turn_off, turn_on, visible, was_visible; - struct drm_framebuffer *fb = plane_state->base.fb; int ret; if (INTEL_GEN(dev_priv) >= 9 && plane->id != PLANE_CURSOR) { @@ -11518,24 +11783,18 @@ int intel_plane_atomic_calc_changes(const struct intel_crtc_state *old_crtc_stat plane_state->base.visible = visible = false; crtc_state->active_planes &= ~BIT(plane->id); crtc_state->data_rate[plane->id] = 0; + crtc_state->min_cdclk[plane->id] = 0; } if (!was_visible && !visible) return 0; - if (fb != old_plane_state->base.fb) - crtc_state->fb_changed = true; - turn_off = was_visible && (!visible || mode_changed); turn_on = visible && (!was_visible || mode_changed); - DRM_DEBUG_ATOMIC("[CRTC:%d:%s] has [PLANE:%d:%s] with fb %i\n", + DRM_DEBUG_ATOMIC("[CRTC:%d:%s] with [PLANE:%d:%s] visible %i -> %i, off %i, on %i, ms %i\n", crtc->base.base.id, crtc->base.name, plane->base.base.id, plane->base.name, - fb ? fb->base.id : -1); - - DRM_DEBUG_ATOMIC("[PLANE:%d:%s] visible %i -> %i, off %i, on %i, ms %i\n", - plane->base.base.id, plane->base.name, was_visible, visible, turn_off, turn_on, mode_changed); @@ -11644,7 +11903,7 @@ static int icl_add_linked_planes(struct intel_atomic_state *state) int i; for_each_new_intel_plane_in_state(state, plane, plane_state, i) { - linked = plane_state->linked_plane; + linked = plane_state->planar_linked_plane; if (!linked) continue; @@ -11653,8 +11912,8 @@ static int icl_add_linked_planes(struct intel_atomic_state *state) if (IS_ERR(linked_plane_state)) return PTR_ERR(linked_plane_state); - WARN_ON(linked_plane_state->linked_plane != plane); - WARN_ON(linked_plane_state->slave == plane_state->slave); + WARN_ON(linked_plane_state->planar_linked_plane != plane); + WARN_ON(linked_plane_state->planar_slave == plane_state->planar_slave); } return 0; @@ -11677,16 +11936,16 @@ static int icl_check_nv12_planes(struct intel_crtc_state *crtc_state) * in the crtc_state->active_planes mask. */ for_each_new_intel_plane_in_state(state, plane, plane_state, i) { - if (plane->pipe != crtc->pipe || !plane_state->linked_plane) + if (plane->pipe != crtc->pipe || !plane_state->planar_linked_plane) continue; - plane_state->linked_plane = NULL; - if (plane_state->slave && !plane_state->base.visible) { + plane_state->planar_linked_plane = NULL; + if (plane_state->planar_slave && !plane_state->base.visible) { crtc_state->active_planes &= ~BIT(plane->id); crtc_state->update_planes |= BIT(plane->id); } - plane_state->slave = false; + plane_state->planar_slave = false; } if (!crtc_state->nv12_planes) @@ -11720,10 +11979,10 @@ static int icl_check_nv12_planes(struct intel_crtc_state *crtc_state) return -EINVAL; } - plane_state->linked_plane = linked; + plane_state->planar_linked_plane = linked; - linked_state->slave = true; - linked_state->linked_plane = plane; + linked_state->planar_slave = true; + linked_state->planar_linked_plane = plane; crtc_state->active_planes |= BIT(linked->id); crtc_state->update_planes |= BIT(linked->id); DRM_DEBUG_KMS("Using %s as Y plane for %s\n", linked->base.name, plane->base.name); @@ -11743,25 +12002,108 @@ static bool c8_planes_changed(const struct intel_crtc_state *new_crtc_state) return !old_crtc_state->c8_planes != !new_crtc_state->c8_planes; } -static int intel_crtc_atomic_check(struct drm_crtc *crtc, - struct drm_crtc_state *crtc_state) +static int icl_add_sync_mode_crtcs(struct intel_crtc_state *crtc_state) { - struct drm_i915_private *dev_priv = to_i915(crtc->dev); - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - struct intel_crtc_state *pipe_config = - to_intel_crtc_state(crtc_state); + struct drm_crtc *crtc = crtc_state->base.crtc; + struct intel_atomic_state *state = to_intel_atomic_state(crtc_state->base.state); + struct drm_i915_private *dev_priv = to_i915(crtc_state->base.crtc->dev); + struct drm_connector *master_connector, *connector; + struct drm_connector_state *connector_state; + struct drm_connector_list_iter conn_iter; + struct drm_crtc *master_crtc = NULL; + struct drm_crtc_state *master_crtc_state; + struct intel_crtc_state *master_pipe_config; + int i, tile_group_id; + + if (INTEL_GEN(dev_priv) < 11) + return 0; + + /* + * In case of tiled displays there could be one or more slaves but there is + * only one master. Lets make the CRTC used by the connector corresponding + * to the last horizonal and last vertical tile a master/genlock CRTC. + * All the other CRTCs corresponding to other tiles of the same Tile group + * are the slave CRTCs and hold a pointer to their genlock CRTC. + */ + for_each_new_connector_in_state(&state->base, connector, connector_state, i) { + if (connector_state->crtc != crtc) + continue; + if (!connector->has_tile) + continue; + if (crtc_state->base.mode.hdisplay != connector->tile_h_size || + crtc_state->base.mode.vdisplay != connector->tile_v_size) + return 0; + if (connector->tile_h_loc == connector->num_h_tile - 1 && + connector->tile_v_loc == connector->num_v_tile - 1) + continue; + crtc_state->sync_mode_slaves_mask = 0; + tile_group_id = connector->tile_group->id; + drm_connector_list_iter_begin(&dev_priv->drm, &conn_iter); + drm_for_each_connector_iter(master_connector, &conn_iter) { + struct drm_connector_state *master_conn_state = NULL; + + if (!master_connector->has_tile) + continue; + if (master_connector->tile_h_loc != master_connector->num_h_tile - 1 || + master_connector->tile_v_loc != master_connector->num_v_tile - 1) + continue; + if (master_connector->tile_group->id != tile_group_id) + continue; + + master_conn_state = drm_atomic_get_connector_state(&state->base, + master_connector); + if (IS_ERR(master_conn_state)) { + drm_connector_list_iter_end(&conn_iter); + return PTR_ERR(master_conn_state); + } + if (master_conn_state->crtc) { + master_crtc = master_conn_state->crtc; + break; + } + } + drm_connector_list_iter_end(&conn_iter); + + if (!master_crtc) { + DRM_DEBUG_KMS("Could not find Master CRTC for Slave CRTC %d\n", + connector_state->crtc->base.id); + return -EINVAL; + } + + master_crtc_state = drm_atomic_get_crtc_state(&state->base, + master_crtc); + if (IS_ERR(master_crtc_state)) + return PTR_ERR(master_crtc_state); + + master_pipe_config = to_intel_crtc_state(master_crtc_state); + crtc_state->master_transcoder = master_pipe_config->cpu_transcoder; + master_pipe_config->sync_mode_slaves_mask |= + BIT(crtc_state->cpu_transcoder); + DRM_DEBUG_KMS("Master Transcoder = %s added for Slave CRTC = %d, slave transcoder bitmask = %d\n", + transcoder_name(crtc_state->master_transcoder), + crtc_state->base.crtc->base.id, + master_pipe_config->sync_mode_slaves_mask); + } + + return 0; +} + +static int intel_crtc_atomic_check(struct intel_atomic_state *state, + struct intel_crtc *crtc) +{ + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + struct intel_crtc_state *crtc_state = + intel_atomic_get_new_crtc_state(state, crtc); + bool mode_changed = needs_modeset(crtc_state); int ret; - bool mode_changed = needs_modeset(pipe_config); if (INTEL_GEN(dev_priv) < 5 && !IS_G4X(dev_priv) && - mode_changed && !crtc_state->active) - pipe_config->update_wm_post = true; + mode_changed && !crtc_state->base.active) + crtc_state->update_wm_post = true; - if (mode_changed && crtc_state->enable && + if (mode_changed && crtc_state->base.enable && dev_priv->display.crtc_compute_clock && - !WARN_ON(pipe_config->shared_dpll)) { - ret = dev_priv->display.crtc_compute_clock(intel_crtc, - pipe_config); + !WARN_ON(crtc_state->shared_dpll)) { + ret = dev_priv->display.crtc_compute_clock(crtc, crtc_state); if (ret) return ret; } @@ -11770,19 +12112,19 @@ static int intel_crtc_atomic_check(struct drm_crtc *crtc, * May need to update pipe gamma enable bits * when C8 planes are getting enabled/disabled. */ - if (c8_planes_changed(pipe_config)) - crtc_state->color_mgmt_changed = true; + if (c8_planes_changed(crtc_state)) + crtc_state->base.color_mgmt_changed = true; - if (mode_changed || pipe_config->update_pipe || - crtc_state->color_mgmt_changed) { - ret = intel_color_check(pipe_config); + if (mode_changed || crtc_state->update_pipe || + crtc_state->base.color_mgmt_changed) { + ret = intel_color_check(crtc_state); if (ret) return ret; } ret = 0; if (dev_priv->display.compute_pipe_wm) { - ret = dev_priv->display.compute_pipe_wm(pipe_config); + ret = dev_priv->display.compute_pipe_wm(crtc_state); if (ret) { DRM_DEBUG_KMS("Target pipe watermarks are invalid\n"); return ret; @@ -11798,7 +12140,7 @@ static int intel_crtc_atomic_check(struct drm_crtc *crtc, * old state and the new state. We can program these * immediately. */ - ret = dev_priv->display.compute_intermediate_wm(pipe_config); + ret = dev_priv->display.compute_intermediate_wm(crtc_state); if (ret) { DRM_DEBUG_KMS("No valid intermediate pipe watermarks are possible\n"); return ret; @@ -11806,29 +12148,19 @@ static int intel_crtc_atomic_check(struct drm_crtc *crtc, } if (INTEL_GEN(dev_priv) >= 9) { - if (mode_changed || pipe_config->update_pipe) - ret = skl_update_scaler_crtc(pipe_config); - - if (!ret) - ret = icl_check_nv12_planes(pipe_config); - if (!ret) - ret = skl_check_pipe_max_pixel_rate(intel_crtc, - pipe_config); + if (mode_changed || crtc_state->update_pipe) + ret = skl_update_scaler_crtc(crtc_state); if (!ret) - ret = intel_atomic_setup_scalers(dev_priv, intel_crtc, - pipe_config); + ret = intel_atomic_setup_scalers(dev_priv, crtc, + crtc_state); } if (HAS_IPS(dev_priv)) - pipe_config->ips_enabled = hsw_compute_ips_config(pipe_config); + crtc_state->ips_enabled = hsw_compute_ips_config(crtc_state); return ret; } -static const struct drm_crtc_helper_funcs intel_helper_funcs = { - .atomic_check = intel_crtc_atomic_check, -}; - static void intel_modeset_update_connector_atomic_state(struct drm_device *dev) { struct intel_connector *connector; @@ -12138,6 +12470,15 @@ static void intel_dump_pipe_config(const struct intel_crtc_state *pipe_config, intel_dpll_dump_hw_state(dev_priv, &pipe_config->dpll_hw_state); + if (IS_CHERRYVIEW(dev_priv)) + DRM_DEBUG_KMS("cgm_mode: 0x%x gamma_mode: 0x%x gamma_enable: %d csc_enable: %d\n", + pipe_config->cgm_mode, pipe_config->gamma_mode, + pipe_config->gamma_enable, pipe_config->csc_enable); + else + DRM_DEBUG_KMS("csc_mode: 0x%x gamma_mode: 0x%x gamma_enable: %d csc_enable: %d\n", + pipe_config->csc_mode, pipe_config->gamma_mode, + pipe_config->gamma_enable, pipe_config->csc_enable); + dump_planes: if (!state) return; @@ -12158,6 +12499,12 @@ static bool check_digital_port_conflicts(struct intel_atomic_state *state) bool ret = true; /* + * We're going to peek into connector->state, + * hence connection_mutex must be held. + */ + drm_modeset_lock_assert_held(&dev->mode_config.connection_mutex); + + /* * Walk the connector list instead of the encoder * list to detect the problem on ddi platforms * where there's just one encoder per digital port. @@ -12239,6 +12586,13 @@ clear_intel_crtc_state(struct intel_crtc_state *crtc_state) if (IS_G4X(dev_priv) || IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) saved_state->wm = crtc_state->wm; + /* + * Save the slave bitmask which gets filled for master crtc state during + * slave atomic check call. + */ + if (is_trans_port_sync_master(crtc_state)) + saved_state->sync_mode_slaves_mask = + crtc_state->sync_mode_slaves_mask; /* Keep base drm_crtc_state intact, only clear our extended struct */ BUILD_BUG_ON(offsetof(struct intel_crtc_state, base)); @@ -12332,6 +12686,15 @@ encoder_retry: drm_mode_set_crtcinfo(&pipe_config->base.adjusted_mode, CRTC_STEREO_DOUBLE); + /* Set the crtc_state defaults for trans_port_sync */ + pipe_config->master_transcoder = INVALID_TRANSCODER; + ret = icl_add_sync_mode_crtcs(pipe_config); + if (ret) { + DRM_DEBUG_KMS("Cannot assign Sync Mode CRTCs: %d\n", + ret); + return ret; + } + /* Pass our mode to the connectors and the CRTC to give them a chance to * adjust it according to limitations or connector properties, and also * a chance to reject the mode entirely. @@ -12464,22 +12827,23 @@ pipe_config_infoframe_mismatch(struct drm_i915_private *dev_priv, if ((drm_debug & DRM_UT_KMS) == 0) return; - drm_dbg(DRM_UT_KMS, "fastset mismatch in %s infoframe", name); - drm_dbg(DRM_UT_KMS, "expected:"); + DRM_DEBUG_KMS("fastset mismatch in %s infoframe\n", name); + DRM_DEBUG_KMS("expected:\n"); hdmi_infoframe_log(KERN_DEBUG, dev_priv->drm.dev, a); - drm_dbg(DRM_UT_KMS, "found"); + DRM_DEBUG_KMS("found:\n"); hdmi_infoframe_log(KERN_DEBUG, dev_priv->drm.dev, b); } else { - drm_err("mismatch in %s infoframe", name); - drm_err("expected:"); + DRM_ERROR("mismatch in %s infoframe\n", name); + DRM_ERROR("expected:\n"); hdmi_infoframe_log(KERN_ERR, dev_priv->drm.dev, a); - drm_err("found"); + DRM_ERROR("found:\n"); hdmi_infoframe_log(KERN_ERR, dev_priv->drm.dev, b); } } -static void __printf(3, 4) -pipe_config_mismatch(bool fastset, const char *name, const char *format, ...) +static void __printf(4, 5) +pipe_config_mismatch(bool fastset, const struct intel_crtc *crtc, + const char *name, const char *format, ...) { struct va_format vaf; va_list args; @@ -12489,9 +12853,11 @@ pipe_config_mismatch(bool fastset, const char *name, const char *format, ...) vaf.va = &args; if (fastset) - drm_dbg(DRM_UT_KMS, "fastset mismatch in %s %pV", name, &vaf); + DRM_DEBUG_KMS("[CRTC:%d:%s] fastset mismatch in %s %pV\n", + crtc->base.base.id, crtc->base.name, name, &vaf); else - drm_err("mismatch in %s %pV", name, &vaf); + DRM_ERROR("[CRTC:%d:%s] mismatch in %s %pV\n", + crtc->base.base.id, crtc->base.name, name, &vaf); va_end(args); } @@ -12519,7 +12885,9 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, bool fastset) { struct drm_i915_private *dev_priv = to_i915(current_config->base.crtc->dev); + struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); bool ret = true; + u32 bp_gamma = 0; bool fixup_inherited = fastset && (current_config->base.mode.private_flags & I915_MODE_FLAG_INHERITED) && !(pipe_config->base.mode.private_flags & I915_MODE_FLAG_INHERITED); @@ -12531,8 +12899,8 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, #define PIPE_CONF_CHECK_X(name) do { \ if (current_config->name != pipe_config->name) { \ - pipe_config_mismatch(fastset, __stringify(name), \ - "(expected 0x%08x, found 0x%08x)\n", \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ + "(expected 0x%08x, found 0x%08x)", \ current_config->name, \ pipe_config->name); \ ret = false; \ @@ -12541,8 +12909,8 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, #define PIPE_CONF_CHECK_I(name) do { \ if (current_config->name != pipe_config->name) { \ - pipe_config_mismatch(fastset, __stringify(name), \ - "(expected %i, found %i)\n", \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ + "(expected %i, found %i)", \ current_config->name, \ pipe_config->name); \ ret = false; \ @@ -12551,8 +12919,8 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, #define PIPE_CONF_CHECK_BOOL(name) do { \ if (current_config->name != pipe_config->name) { \ - pipe_config_mismatch(fastset, __stringify(name), \ - "(expected %s, found %s)\n", \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ + "(expected %s, found %s)", \ yesno(current_config->name), \ yesno(pipe_config->name)); \ ret = false; \ @@ -12568,8 +12936,8 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, if (!fixup_inherited || (!current_config->name && !pipe_config->name)) { \ PIPE_CONF_CHECK_BOOL(name); \ } else { \ - pipe_config_mismatch(fastset, __stringify(name), \ - "unable to verify whether state matches exactly, forcing modeset (expected %s, found %s)\n", \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ + "unable to verify whether state matches exactly, forcing modeset (expected %s, found %s)", \ yesno(current_config->name), \ yesno(pipe_config->name)); \ ret = false; \ @@ -12578,8 +12946,8 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, #define PIPE_CONF_CHECK_P(name) do { \ if (current_config->name != pipe_config->name) { \ - pipe_config_mismatch(fastset, __stringify(name), \ - "(expected %p, found %p)\n", \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ + "(expected %p, found %p)", \ current_config->name, \ pipe_config->name); \ ret = false; \ @@ -12590,9 +12958,9 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, if (!intel_compare_link_m_n(¤t_config->name, \ &pipe_config->name,\ !fastset)) { \ - pipe_config_mismatch(fastset, __stringify(name), \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ "(expected tu %i gmch %i/%i link %i/%i, " \ - "found tu %i, gmch %i/%i link %i/%i)\n", \ + "found tu %i, gmch %i/%i link %i/%i)", \ current_config->name.tu, \ current_config->name.gmch_m, \ current_config->name.gmch_n, \ @@ -12617,10 +12985,10 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, &pipe_config->name, !fastset) && \ !intel_compare_link_m_n(¤t_config->alt_name, \ &pipe_config->name, !fastset)) { \ - pipe_config_mismatch(fastset, __stringify(name), \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ "(expected tu %i gmch %i/%i link %i/%i, " \ "or tu %i gmch %i/%i link %i/%i, " \ - "found tu %i, gmch %i/%i link %i/%i)\n", \ + "found tu %i, gmch %i/%i link %i/%i)", \ current_config->name.tu, \ current_config->name.gmch_m, \ current_config->name.gmch_n, \ @@ -12642,8 +13010,8 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, #define PIPE_CONF_CHECK_FLAGS(name, mask) do { \ if ((current_config->name ^ pipe_config->name) & (mask)) { \ - pipe_config_mismatch(fastset, __stringify(name), \ - "(%x) (expected %i, found %i)\n", \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ + "(%x) (expected %i, found %i)", \ (mask), \ current_config->name & (mask), \ pipe_config->name & (mask)); \ @@ -12653,8 +13021,8 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, #define PIPE_CONF_CHECK_CLOCK_FUZZY(name) do { \ if (!intel_fuzzy_clock_check(current_config->name, pipe_config->name)) { \ - pipe_config_mismatch(fastset, __stringify(name), \ - "(expected %i, found %i)\n", \ + pipe_config_mismatch(fastset, crtc, __stringify(name), \ + "(expected %i, found %i)", \ current_config->name, \ pipe_config->name); \ ret = false; \ @@ -12671,6 +13039,24 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, } \ } while (0) +#define PIPE_CONF_CHECK_COLOR_LUT(name1, name2, bit_precision) do { \ + if (current_config->name1 != pipe_config->name1) { \ + pipe_config_mismatch(fastset, crtc, __stringify(name1), \ + "(expected %i, found %i, won't compare lut values)", \ + current_config->name1, \ + pipe_config->name1); \ + ret = false;\ + } else { \ + if (!intel_color_lut_equal(current_config->name2, \ + pipe_config->name2, pipe_config->name1, \ + bit_precision)) { \ + pipe_config_mismatch(fastset, crtc, __stringify(name2), \ + "hw_state doesn't match sw_state"); \ + ret = false; \ + } \ + } \ +} while (0) + #define PIPE_CONF_QUIRK(quirk) \ ((current_config->quirks | pipe_config->quirks) & (quirk)) @@ -12709,6 +13095,7 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, PIPE_CONF_CHECK_I(pixel_multiplier); PIPE_CONF_CHECK_I(output_format); + PIPE_CONF_CHECK_I(dc3co_exitline); PIPE_CONF_CHECK_BOOL(has_hdmi_sink); if ((INTEL_GEN(dev_priv) < 8 && !IS_HASWELL(dev_priv)) || IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) @@ -12717,6 +13104,7 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, PIPE_CONF_CHECK_BOOL(hdmi_scrambling); PIPE_CONF_CHECK_BOOL(hdmi_high_tmds_clock_ratio); PIPE_CONF_CHECK_BOOL(has_infoframe); + PIPE_CONF_CHECK_BOOL(fec_enable); PIPE_CONF_CHECK_BOOL_INCOMPLETE(has_audio); @@ -12766,6 +13154,11 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, PIPE_CONF_CHECK_X(csc_mode); PIPE_CONF_CHECK_BOOL(gamma_enable); PIPE_CONF_CHECK_BOOL(csc_enable); + + bp_gamma = intel_color_get_gamma_bit_precision(pipe_config); + if (bp_gamma) + PIPE_CONF_CHECK_COLOR_LUT(gamma_mode, base.gamma_lut, bp_gamma); + } PIPE_CONF_CHECK_BOOL(double_wide); @@ -12821,6 +13214,9 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, PIPE_CONF_CHECK_INFOFRAME(hdmi); PIPE_CONF_CHECK_INFOFRAME(drm); + PIPE_CONF_CHECK_I(sync_mode_slaves_mask); + PIPE_CONF_CHECK_I(master_transcoder); + #undef PIPE_CONF_CHECK_X #undef PIPE_CONF_CHECK_I #undef PIPE_CONF_CHECK_BOOL @@ -12828,6 +13224,7 @@ intel_pipe_config_compare(const struct intel_crtc_state *current_config, #undef PIPE_CONF_CHECK_P #undef PIPE_CONF_CHECK_FLAGS #undef PIPE_CONF_CHECK_CLOCK_FUZZY +#undef PIPE_CONF_CHECK_COLOR_LUT #undef PIPE_CONF_QUIRK return ret; @@ -13139,7 +13536,7 @@ intel_verify_planes(struct intel_atomic_state *state) for_each_new_intel_plane_in_state(state, plane, plane_state, i) - assert_plane(plane, plane_state->slave || + assert_plane(plane, plane_state->planar_slave || plane_state->base.visible); } @@ -13255,10 +13652,15 @@ intel_modeset_verify_disabled(struct drm_i915_private *dev_priv, verify_disabled_dpll_state(dev_priv); } -static void update_scanline_offset(const struct intel_crtc_state *crtc_state) +static void +intel_crtc_update_active_timings(const struct intel_crtc_state *crtc_state) { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + const struct drm_display_mode *adjusted_mode = + &crtc_state->base.adjusted_mode; + + drm_calc_timestamping_constants(&crtc->base, adjusted_mode); /* * The scanline counter increments at the leading edge of hsync. @@ -13288,7 +13690,6 @@ static void update_scanline_offset(const struct intel_crtc_state *crtc_state) * answer that's slightly in the future. */ if (IS_GEN(dev_priv, 2)) { - const struct drm_display_mode *adjusted_mode = &crtc_state->base.adjusted_mode; int vtotal; vtotal = adjusted_mode->crtc_vtotal; @@ -13299,8 +13700,9 @@ static void update_scanline_offset(const struct intel_crtc_state *crtc_state) } else if (HAS_DDI(dev_priv) && intel_crtc_has_type(crtc_state, INTEL_OUTPUT_HDMI)) { crtc->scanline_offset = 2; - } else + } else { crtc->scanline_offset = 1; + } } static void intel_modeset_clear_plls(struct intel_atomic_state *state) @@ -13382,158 +13784,43 @@ static int haswell_mode_set_planes_workaround(struct intel_atomic_state *state) return 0; } -static int intel_lock_all_pipes(struct intel_atomic_state *state) -{ - struct drm_i915_private *dev_priv = to_i915(state->base.dev); - struct intel_crtc *crtc; - - /* Add all pipes to the state */ - for_each_intel_crtc(&dev_priv->drm, crtc) { - struct intel_crtc_state *crtc_state; - - crtc_state = intel_atomic_get_crtc_state(&state->base, crtc); - if (IS_ERR(crtc_state)) - return PTR_ERR(crtc_state); - } - - return 0; -} - -static int intel_modeset_all_pipes(struct intel_atomic_state *state) -{ - struct drm_i915_private *dev_priv = to_i915(state->base.dev); - struct intel_crtc *crtc; - - /* - * Add all pipes to the state, and force - * a modeset on all the active ones. - */ - for_each_intel_crtc(&dev_priv->drm, crtc) { - struct intel_crtc_state *crtc_state; - int ret; - - crtc_state = intel_atomic_get_crtc_state(&state->base, crtc); - if (IS_ERR(crtc_state)) - return PTR_ERR(crtc_state); - - if (!crtc_state->base.active || needs_modeset(crtc_state)) - continue; - - crtc_state->base.mode_changed = true; - - ret = drm_atomic_add_affected_connectors(&state->base, - &crtc->base); - if (ret) - return ret; - - ret = drm_atomic_add_affected_planes(&state->base, - &crtc->base); - if (ret) - return ret; - } - - return 0; -} - static int intel_modeset_checks(struct intel_atomic_state *state) { struct drm_i915_private *dev_priv = to_i915(state->base.dev); struct intel_crtc_state *old_crtc_state, *new_crtc_state; struct intel_crtc *crtc; - int ret = 0, i; - - if (!check_digital_port_conflicts(state)) { - DRM_DEBUG_KMS("rejecting conflicting digital port configuration\n"); - return -EINVAL; - } + int ret, i; /* keep the current setting */ if (!state->cdclk.force_min_cdclk_changed) state->cdclk.force_min_cdclk = dev_priv->cdclk.force_min_cdclk; state->modeset = true; - state->active_crtcs = dev_priv->active_crtcs; + state->active_pipes = dev_priv->active_pipes; state->cdclk.logical = dev_priv->cdclk.logical; state->cdclk.actual = dev_priv->cdclk.actual; - state->cdclk.pipe = INVALID_PIPE; for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { if (new_crtc_state->base.active) - state->active_crtcs |= 1 << i; + state->active_pipes |= BIT(crtc->pipe); else - state->active_crtcs &= ~(1 << i); + state->active_pipes &= ~BIT(crtc->pipe); if (old_crtc_state->base.active != new_crtc_state->base.active) - state->active_pipe_changes |= drm_crtc_mask(&crtc->base); + state->active_pipe_changes |= BIT(crtc->pipe); } - /* - * See if the config requires any additional preparation, e.g. - * to adjust global state with pipes off. We need to do this - * here so we can get the modeset_pipe updated config for the new - * mode set on this crtc. For other crtcs we need to use the - * adjusted_mode bits in the crtc directly. - */ - if (dev_priv->display.modeset_calc_cdclk) { - enum pipe pipe; - - ret = dev_priv->display.modeset_calc_cdclk(state); - if (ret < 0) + if (state->active_pipe_changes) { + ret = intel_atomic_lock_global_state(state); + if (ret) return ret; - - /* - * Writes to dev_priv->cdclk.logical must protected by - * holding all the crtc locks, even if we don't end up - * touching the hardware - */ - if (intel_cdclk_changed(&dev_priv->cdclk.logical, - &state->cdclk.logical)) { - ret = intel_lock_all_pipes(state); - if (ret < 0) - return ret; - } - - if (is_power_of_2(state->active_crtcs)) { - struct intel_crtc *crtc; - struct intel_crtc_state *crtc_state; - - pipe = ilog2(state->active_crtcs); - crtc = intel_get_crtc_for_pipe(dev_priv, pipe); - crtc_state = intel_atomic_get_new_crtc_state(state, crtc); - if (crtc_state && needs_modeset(crtc_state)) - pipe = INVALID_PIPE; - } else { - pipe = INVALID_PIPE; - } - - /* All pipes must be switched off while we change the cdclk. */ - if (pipe != INVALID_PIPE && - intel_cdclk_needs_cd2x_update(dev_priv, - &dev_priv->cdclk.actual, - &state->cdclk.actual)) { - ret = intel_lock_all_pipes(state); - if (ret < 0) - return ret; - - state->cdclk.pipe = pipe; - } else if (intel_cdclk_needs_modeset(&dev_priv->cdclk.actual, - &state->cdclk.actual)) { - ret = intel_modeset_all_pipes(state); - if (ret < 0) - return ret; - - state->cdclk.pipe = INVALID_PIPE; - } - - DRM_DEBUG_KMS("New cdclk calculated to be logical %u kHz, actual %u kHz\n", - state->cdclk.logical.cdclk, - state->cdclk.actual.cdclk); - DRM_DEBUG_KMS("New voltage level calculated to be logical %u, actual %u\n", - state->cdclk.logical.voltage_level, - state->cdclk.actual.voltage_level); } + ret = intel_modeset_calc_cdclk(state); + if (ret) + return ret; + intel_modeset_clear_plls(state); if (IS_HASWELL(dev_priv)) @@ -13582,6 +13869,114 @@ static void intel_crtc_check_fastset(const struct intel_crtc_state *old_crtc_sta new_crtc_state->has_drrs = old_crtc_state->has_drrs; } +static int intel_crtc_add_planes_to_state(struct intel_atomic_state *state, + struct intel_crtc *crtc, + u8 plane_ids_mask) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + struct intel_plane *plane; + + for_each_intel_plane_on_crtc(&dev_priv->drm, crtc, plane) { + struct intel_plane_state *plane_state; + + if ((plane_ids_mask & BIT(plane->id)) == 0) + continue; + + plane_state = intel_atomic_get_plane_state(state, plane); + if (IS_ERR(plane_state)) + return PTR_ERR(plane_state); + } + + return 0; +} + +static bool active_planes_affects_min_cdclk(struct drm_i915_private *dev_priv) +{ + /* See {hsw,vlv,ivb}_plane_ratio() */ + return IS_BROADWELL(dev_priv) || IS_HASWELL(dev_priv) || + IS_CHERRYVIEW(dev_priv) || IS_VALLEYVIEW(dev_priv) || + IS_IVYBRIDGE(dev_priv); +} + +static int intel_atomic_check_planes(struct intel_atomic_state *state, + bool *need_modeset) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + struct intel_crtc_state *old_crtc_state, *new_crtc_state; + struct intel_plane_state *plane_state; + struct intel_plane *plane; + struct intel_crtc *crtc; + int i, ret; + + ret = icl_add_linked_planes(state); + if (ret) + return ret; + + for_each_new_intel_plane_in_state(state, plane, plane_state, i) { + ret = intel_plane_atomic_check(state, plane); + if (ret) { + DRM_DEBUG_ATOMIC("[PLANE:%d:%s] atomic driver check failed\n", + plane->base.base.id, plane->base.name); + return ret; + } + } + + for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state, + new_crtc_state, i) { + u8 old_active_planes, new_active_planes; + + ret = icl_check_nv12_planes(new_crtc_state); + if (ret) + return ret; + + /* + * On some platforms the number of active planes affects + * the planes' minimum cdclk calculation. Add such planes + * to the state before we compute the minimum cdclk. + */ + if (!active_planes_affects_min_cdclk(dev_priv)) + continue; + + old_active_planes = old_crtc_state->active_planes & ~BIT(PLANE_CURSOR); + new_active_planes = new_crtc_state->active_planes & ~BIT(PLANE_CURSOR); + + if (hweight8(old_active_planes) == hweight8(new_active_planes)) + continue; + + ret = intel_crtc_add_planes_to_state(state, crtc, new_active_planes); + if (ret) + return ret; + } + + /* + * active_planes bitmask has been updated, and potentially + * affected planes are part of the state. We can now + * compute the minimum cdclk for each plane. + */ + for_each_new_intel_plane_in_state(state, plane, plane_state, i) + *need_modeset |= intel_plane_calc_min_cdclk(state, plane); + + return 0; +} + +static int intel_atomic_check_crtcs(struct intel_atomic_state *state) +{ + struct intel_crtc_state *crtc_state; + struct intel_crtc *crtc; + int i; + + for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) { + int ret = intel_crtc_atomic_check(state, crtc); + if (ret) { + DRM_DEBUG_ATOMIC("[CRTC:%d:%s] atomic driver check failed\n", + crtc->base.base.id, crtc->base.name); + return ret; + } + } + + return 0; +} + /** * intel_atomic_check - validate state object * @dev: drm device @@ -13595,7 +13990,7 @@ static int intel_atomic_check(struct drm_device *dev, struct intel_crtc_state *old_crtc_state, *new_crtc_state; struct intel_crtc *crtc; int ret, i; - bool any_ms = state->cdclk.force_min_cdclk_changed; + bool any_ms = false; /* Catch I915_MODE_FLAG_INHERITED */ for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state, @@ -13629,10 +14024,22 @@ static int intel_atomic_check(struct drm_device *dev, any_ms = true; } + if (any_ms && !check_digital_port_conflicts(state)) { + DRM_DEBUG_KMS("rejecting conflicting digital port configuration\n"); + ret = EINVAL; + goto fail; + } + ret = drm_dp_mst_atomic_check(&state->base); if (ret) goto fail; + any_ms |= state->cdclk.force_min_cdclk_changed; + + ret = intel_atomic_check_planes(state, &any_ms); + if (ret) + goto fail; + if (any_ms) { ret = intel_modeset_checks(state); if (ret) @@ -13641,11 +14048,7 @@ static int intel_atomic_check(struct drm_device *dev, state->cdclk.logical = dev_priv->cdclk.logical; } - ret = icl_add_linked_planes(state); - if (ret) - goto fail; - - ret = drm_atomic_helper_check_planes(dev, &state->base); + ret = intel_atomic_check_crtcs(state); if (ret) goto fail; @@ -13703,25 +14106,110 @@ u32 intel_crtc_get_vblank_counter(struct intel_crtc *crtc) return crtc->base.funcs->get_vblank_counter(&crtc->base); } +void intel_crtc_arm_fifo_underrun(struct intel_crtc *crtc, + struct intel_crtc_state *crtc_state) +{ + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + + if (!IS_GEN(dev_priv, 2)) + intel_set_cpu_fifo_underrun_reporting(dev_priv, crtc->pipe, true); + + if (crtc_state->has_pch_encoder) { + enum pipe pch_transcoder = + intel_crtc_pch_transcoder(crtc); + + intel_set_pch_fifo_underrun_reporting(dev_priv, pch_transcoder, true); + } +} + +static void intel_pipe_fastset(const struct intel_crtc_state *old_crtc_state, + const struct intel_crtc_state *new_crtc_state) +{ + struct intel_crtc *crtc = to_intel_crtc(new_crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + + /* + * Update pipe size and adjust fitter if needed: the reason for this is + * that in compute_mode_changes we check the native mode (not the pfit + * mode) to see if we can flip rather than do a full mode set. In the + * fastboot case, we'll flip, but if we don't update the pipesrc and + * pfit state, we'll end up with a big fb scanned out into the wrong + * sized surface. + */ + intel_set_pipe_src_size(new_crtc_state); + + /* on skylake this is done by detaching scalers */ + if (INTEL_GEN(dev_priv) >= 9) { + skl_detach_scalers(new_crtc_state); + + if (new_crtc_state->pch_pfit.enabled) + skylake_pfit_enable(new_crtc_state); + } else if (HAS_PCH_SPLIT(dev_priv)) { + if (new_crtc_state->pch_pfit.enabled) + ironlake_pfit_enable(new_crtc_state); + else if (old_crtc_state->pch_pfit.enabled) + ironlake_pfit_disable(old_crtc_state); + } + + if (INTEL_GEN(dev_priv) >= 11) + icl_set_pipe_chicken(crtc); +} + +static void commit_pipe_config(struct intel_atomic_state *state, + struct intel_crtc_state *old_crtc_state, + struct intel_crtc_state *new_crtc_state) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + bool modeset = needs_modeset(new_crtc_state); + + /* + * During modesets pipe configuration was programmed as the + * CRTC was enabled. + */ + if (!modeset) { + if (new_crtc_state->base.color_mgmt_changed || + new_crtc_state->update_pipe) + intel_color_commit(new_crtc_state); + + if (INTEL_GEN(dev_priv) >= 9) + skl_detach_scalers(new_crtc_state); + + if (INTEL_GEN(dev_priv) >= 9 || IS_BROADWELL(dev_priv)) + bdw_set_pipemisc(new_crtc_state); + + if (new_crtc_state->update_pipe) + intel_pipe_fastset(old_crtc_state, new_crtc_state); + } + + if (dev_priv->display.atomic_update_watermarks) + dev_priv->display.atomic_update_watermarks(state, + new_crtc_state); +} + static void intel_update_crtc(struct intel_crtc *crtc, struct intel_atomic_state *state, struct intel_crtc_state *old_crtc_state, struct intel_crtc_state *new_crtc_state) { - struct drm_device *dev = state->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(state->base.dev); bool modeset = needs_modeset(new_crtc_state); struct intel_plane_state *new_plane_state = intel_atomic_get_new_plane_state(state, to_intel_plane(crtc->base.primary)); if (modeset) { - update_scanline_offset(new_crtc_state); + intel_crtc_update_active_timings(new_crtc_state); + dev_priv->display.crtc_enable(new_crtc_state, state); /* vblanks work again, re-enable pipe CRC. */ intel_crtc_enable_pipe_crc(crtc); } else { + if (new_crtc_state->preload_luts && + (new_crtc_state->base.color_mgmt_changed || + new_crtc_state->update_pipe)) + intel_color_load_luts(new_crtc_state); + intel_pre_plane_update(old_crtc_state, new_crtc_state); if (new_crtc_state->update_pipe) @@ -13733,17 +14221,151 @@ static void intel_update_crtc(struct intel_crtc *crtc, else if (new_plane_state) intel_fbc_enable(crtc, new_crtc_state, new_plane_state); - intel_begin_crtc_commit(state, crtc); + /* Perform vblank evasion around commit operation */ + intel_pipe_update_start(new_crtc_state); + + commit_pipe_config(state, old_crtc_state, new_crtc_state); if (INTEL_GEN(dev_priv) >= 9) skl_update_planes_on_crtc(state, crtc); else i9xx_update_planes_on_crtc(state, crtc); - intel_finish_crtc_commit(state, crtc); + intel_pipe_update_end(new_crtc_state); + + /* + * We usually enable FIFO underrun interrupts as part of the + * CRTC enable sequence during modesets. But when we inherit a + * valid pipe configuration from the BIOS we need to take care + * of enabling them on the CRTC's first fastset. + */ + if (new_crtc_state->update_pipe && !modeset && + old_crtc_state->base.mode.private_flags & I915_MODE_FLAG_INHERITED) + intel_crtc_arm_fifo_underrun(crtc, new_crtc_state); +} + +static struct intel_crtc *intel_get_slave_crtc(const struct intel_crtc_state *new_crtc_state) +{ + struct drm_i915_private *dev_priv = to_i915(new_crtc_state->base.crtc->dev); + enum transcoder slave_transcoder; + + WARN_ON(!is_power_of_2(new_crtc_state->sync_mode_slaves_mask)); + + slave_transcoder = ffs(new_crtc_state->sync_mode_slaves_mask) - 1; + return intel_get_crtc_for_pipe(dev_priv, + (enum pipe)slave_transcoder); +} + +static void intel_old_crtc_state_disables(struct intel_atomic_state *state, + struct intel_crtc_state *old_crtc_state, + struct intel_crtc_state *new_crtc_state, + struct intel_crtc *crtc) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + + intel_crtc_disable_planes(state, crtc); + + /* + * We need to disable pipe CRC before disabling the pipe, + * or we race against vblank off. + */ + intel_crtc_disable_pipe_crc(crtc); + + dev_priv->display.crtc_disable(old_crtc_state, state); + crtc->active = false; + intel_fbc_disable(crtc); + intel_disable_shared_dpll(old_crtc_state); + + /* + * Underruns don't always raise interrupts, + * so check manually. + */ + intel_check_cpu_fifo_underruns(dev_priv); + intel_check_pch_fifo_underruns(dev_priv); + + /* FIXME unify this for all platforms */ + if (!new_crtc_state->base.active && + !HAS_GMCH(dev_priv) && + dev_priv->display.initial_watermarks) + dev_priv->display.initial_watermarks(state, + new_crtc_state); +} + +static void intel_trans_port_sync_modeset_disables(struct intel_atomic_state *state, + struct intel_crtc *crtc, + struct intel_crtc_state *old_crtc_state, + struct intel_crtc_state *new_crtc_state) +{ + struct intel_crtc *slave_crtc = intel_get_slave_crtc(new_crtc_state); + struct intel_crtc_state *new_slave_crtc_state = + intel_atomic_get_new_crtc_state(state, slave_crtc); + struct intel_crtc_state *old_slave_crtc_state = + intel_atomic_get_old_crtc_state(state, slave_crtc); + + WARN_ON(!slave_crtc || !new_slave_crtc_state || + !old_slave_crtc_state); + + /* Disable Slave first */ + intel_pre_plane_update(old_slave_crtc_state, new_slave_crtc_state); + if (old_slave_crtc_state->base.active) + intel_old_crtc_state_disables(state, + old_slave_crtc_state, + new_slave_crtc_state, + slave_crtc); + + /* Disable Master */ + intel_pre_plane_update(old_crtc_state, new_crtc_state); + if (old_crtc_state->base.active) + intel_old_crtc_state_disables(state, + old_crtc_state, + new_crtc_state, + crtc); } -static void intel_update_crtcs(struct intel_atomic_state *state) +static void intel_commit_modeset_disables(struct intel_atomic_state *state) +{ + struct intel_crtc_state *new_crtc_state, *old_crtc_state; + struct intel_crtc *crtc; + int i; + + /* + * Disable CRTC/pipes in reverse order because some features(MST in + * TGL+) requires master and slave relationship between pipes, so it + * should always pick the lowest pipe as master as it will be enabled + * first and disable in the reverse order so the master will be the + * last one to be disabled. + */ + for_each_oldnew_intel_crtc_in_state_reverse(state, crtc, old_crtc_state, + new_crtc_state, i) { + if (!needs_modeset(new_crtc_state)) + continue; + + /* In case of Transcoder port Sync master slave CRTCs can be + * assigned in any order and we need to make sure that + * slave CRTCs are disabled first and then master CRTC since + * Slave vblanks are masked till Master Vblanks. + */ + if (is_trans_port_sync_mode(new_crtc_state)) { + if (is_trans_port_sync_master(new_crtc_state)) + intel_trans_port_sync_modeset_disables(state, + crtc, + old_crtc_state, + new_crtc_state); + else + continue; + } else { + intel_pre_plane_update(old_crtc_state, new_crtc_state); + + if (old_crtc_state->base.active) + intel_old_crtc_state_disables(state, + old_crtc_state, + new_crtc_state, + crtc); + } + } +} + +static void intel_commit_modeset_enables(struct intel_atomic_state *state) { struct intel_crtc *crtc; struct intel_crtc_state *old_crtc_state, *new_crtc_state; @@ -13758,14 +14380,120 @@ static void intel_update_crtcs(struct intel_atomic_state *state) } } -static void skl_update_crtcs(struct intel_atomic_state *state) +static void intel_crtc_enable_trans_port_sync(struct intel_crtc *crtc, + struct intel_atomic_state *state, + struct intel_crtc_state *new_crtc_state) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + + intel_crtc_update_active_timings(new_crtc_state); + dev_priv->display.crtc_enable(new_crtc_state, state); + intel_crtc_enable_pipe_crc(crtc); +} + +static void intel_set_dp_tp_ctl_normal(struct intel_crtc *crtc, + struct intel_atomic_state *state) +{ + struct drm_connector *uninitialized_var(conn); + struct drm_connector_state *conn_state; + struct intel_dp *intel_dp; + int i; + + for_each_new_connector_in_state(&state->base, conn, conn_state, i) { + if (conn_state->crtc == &crtc->base) + break; + } + intel_dp = enc_to_intel_dp(&intel_attached_encoder(conn)->base); + intel_dp_stop_link_train(intel_dp); +} + +static void intel_post_crtc_enable_updates(struct intel_crtc *crtc, + struct intel_atomic_state *state) +{ + struct intel_crtc_state *new_crtc_state = + intel_atomic_get_new_crtc_state(state, crtc); + struct intel_crtc_state *old_crtc_state = + intel_atomic_get_old_crtc_state(state, crtc); + struct intel_plane_state *new_plane_state = + intel_atomic_get_new_plane_state(state, + to_intel_plane(crtc->base.primary)); + bool modeset = needs_modeset(new_crtc_state); + + if (new_crtc_state->update_pipe && !new_crtc_state->enable_fbc) + intel_fbc_disable(crtc); + else if (new_plane_state) + intel_fbc_enable(crtc, new_crtc_state, new_plane_state); + + /* Perform vblank evasion around commit operation */ + intel_pipe_update_start(new_crtc_state); + commit_pipe_config(state, old_crtc_state, new_crtc_state); + skl_update_planes_on_crtc(state, crtc); + intel_pipe_update_end(new_crtc_state); + + /* + * We usually enable FIFO underrun interrupts as part of the + * CRTC enable sequence during modesets. But when we inherit a + * valid pipe configuration from the BIOS we need to take care + * of enabling them on the CRTC's first fastset. + */ + if (new_crtc_state->update_pipe && !modeset && + old_crtc_state->base.mode.private_flags & I915_MODE_FLAG_INHERITED) + intel_crtc_arm_fifo_underrun(crtc, new_crtc_state); +} + +static void intel_update_trans_port_sync_crtcs(struct intel_crtc *crtc, + struct intel_atomic_state *state, + struct intel_crtc_state *old_crtc_state, + struct intel_crtc_state *new_crtc_state) +{ + struct intel_crtc *slave_crtc = intel_get_slave_crtc(new_crtc_state); + struct intel_crtc_state *new_slave_crtc_state = + intel_atomic_get_new_crtc_state(state, slave_crtc); + struct intel_crtc_state *old_slave_crtc_state = + intel_atomic_get_old_crtc_state(state, slave_crtc); + + WARN_ON(!slave_crtc || !new_slave_crtc_state || + !old_slave_crtc_state); + + DRM_DEBUG_KMS("Updating Transcoder Port Sync Master CRTC = %d %s and Slave CRTC %d %s\n", + crtc->base.base.id, crtc->base.name, slave_crtc->base.base.id, + slave_crtc->base.name); + + /* Enable seq for slave with with DP_TP_CTL left Idle until the + * master is ready + */ + intel_crtc_enable_trans_port_sync(slave_crtc, + state, + new_slave_crtc_state); + + /* Enable seq for master with with DP_TP_CTL left Idle */ + intel_crtc_enable_trans_port_sync(crtc, + state, + new_crtc_state); + + /* Set Slave's DP_TP_CTL to Normal */ + intel_set_dp_tp_ctl_normal(slave_crtc, + state); + + /* Set Master's DP_TP_CTL To Normal */ + usleep_range(200, 400); + intel_set_dp_tp_ctl_normal(crtc, + state); + + /* Now do the post crtc enable for all master and slaves */ + intel_post_crtc_enable_updates(slave_crtc, + state); + intel_post_crtc_enable_updates(crtc, + state); +} + +static void skl_commit_modeset_enables(struct intel_atomic_state *state) { struct drm_i915_private *dev_priv = to_i915(state->base.dev); struct intel_crtc *crtc; struct intel_crtc_state *old_crtc_state, *new_crtc_state; unsigned int updated = 0; bool progress; - enum pipe pipe; int i; u8 hw_enabled_slices = dev_priv->wm.skl_hw.ddb.enabled_slices; u8 required_slices = state->wm_results.ddb.enabled_slices; @@ -13790,20 +14518,19 @@ static void skl_update_crtcs(struct intel_atomic_state *state) progress = false; for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { + enum pipe pipe = crtc->pipe; bool vbl_wait = false; - unsigned int cmask = drm_crtc_mask(&crtc->base); - - pipe = crtc->pipe; + bool modeset = needs_modeset(new_crtc_state); - if (updated & cmask || !new_crtc_state->base.active) + if (updated & BIT(crtc->pipe) || !new_crtc_state->base.active) continue; if (skl_ddb_allocation_overlaps(&new_crtc_state->wm.skl.ddb, entries, - INTEL_INFO(dev_priv)->num_pipes, i)) + INTEL_NUM_PIPES(dev_priv), i)) continue; - updated |= cmask; + updated |= BIT(pipe); entries[i] = new_crtc_state->wm.skl.ddb; /* @@ -13814,12 +14541,22 @@ static void skl_update_crtcs(struct intel_atomic_state *state) */ if (!skl_ddb_entry_equal(&new_crtc_state->wm.skl.ddb, &old_crtc_state->wm.skl.ddb) && - !new_crtc_state->base.active_changed && + !modeset && state->wm_results.dirty_pipes != updated) vbl_wait = true; - intel_update_crtc(crtc, state, old_crtc_state, - new_crtc_state); + if (modeset && is_trans_port_sync_mode(new_crtc_state)) { + if (is_trans_port_sync_master(new_crtc_state)) + intel_update_trans_port_sync_crtcs(crtc, + state, + old_crtc_state, + new_crtc_state); + else + continue; + } else { + intel_update_crtc(crtc, state, old_crtc_state, + new_crtc_state); + } if (vbl_wait) intel_wait_for_vblank(dev_priv, pipe); @@ -13908,49 +14645,18 @@ static void intel_atomic_commit_tail(struct intel_atomic_state *state) if (state->modeset) wakeref = intel_display_power_get(dev_priv, POWER_DOMAIN_MODESET); - for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { + for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state, + new_crtc_state, i) { if (needs_modeset(new_crtc_state) || new_crtc_state->update_pipe) { put_domains[crtc->pipe] = modeset_get_crtc_power_domains(new_crtc_state); } - - if (!needs_modeset(new_crtc_state)) - continue; - - intel_pre_plane_update(old_crtc_state, new_crtc_state); - - if (old_crtc_state->base.active) { - intel_crtc_disable_planes(state, crtc); - - /* - * We need to disable pipe CRC before disabling the pipe, - * or we race against vblank off. - */ - intel_crtc_disable_pipe_crc(crtc); - - dev_priv->display.crtc_disable(old_crtc_state, state); - crtc->active = false; - intel_fbc_disable(crtc); - intel_disable_shared_dpll(old_crtc_state); - - /* - * Underruns don't always raise - * interrupts, so check manually. - */ - intel_check_cpu_fifo_underruns(dev_priv); - intel_check_pch_fifo_underruns(dev_priv); - - /* FIXME unify this for all platforms */ - if (!new_crtc_state->base.active && - !HAS_GMCH(dev_priv) && - dev_priv->display.initial_watermarks) - dev_priv->display.initial_watermarks(state, - new_crtc_state); - } } + intel_commit_modeset_disables(state); + /* FIXME: Eventually get rid of our crtc->config pointer */ for_each_new_intel_crtc_in_state(state, crtc, new_crtc_state, i) crtc->config = new_crtc_state; @@ -13991,7 +14697,7 @@ static void intel_atomic_commit_tail(struct intel_atomic_state *state) intel_encoders_update_prepare(state); /* Now enable the clocks, plane, pipe, and connectors that we set up. */ - dev_priv->display.update_crtcs(state); + dev_priv->display.commit_modeset_enables(state); if (state->modeset) { intel_encoders_update_complete(state); @@ -14016,6 +14722,7 @@ static void intel_atomic_commit_tail(struct intel_atomic_state *state) for_each_new_intel_crtc_in_state(state, crtc, new_crtc_state, i) { if (new_crtc_state->base.active && !needs_modeset(new_crtc_state) && + !new_crtc_state->preload_luts && (new_crtc_state->base.color_mgmt_changed || new_crtc_state->update_pipe)) intel_color_load_luts(new_crtc_state); @@ -14121,6 +14828,14 @@ static void intel_atomic_track_fbs(struct intel_atomic_state *state) plane->frontbuffer_bit); } +static void assert_global_state_locked(struct drm_i915_private *dev_priv) +{ + struct intel_crtc *crtc; + + for_each_intel_crtc(&dev_priv->drm, crtc) + drm_modeset_lock_assert_held(&crtc->base.mutex); +} + static int intel_atomic_commit(struct drm_device *dev, struct drm_atomic_state *_state, bool nonblock) @@ -14186,12 +14901,14 @@ static int intel_atomic_commit(struct drm_device *dev, intel_shared_dpll_swap_state(state); intel_atomic_track_fbs(state); - if (state->modeset) { + if (state->global_state_changed) { + assert_global_state_locked(dev_priv); + memcpy(dev_priv->min_cdclk, state->min_cdclk, sizeof(state->min_cdclk)); memcpy(dev_priv->min_voltage_level, state->min_voltage_level, sizeof(state->min_voltage_level)); - dev_priv->active_crtcs = state->active_crtcs; + dev_priv->active_pipes = state->active_pipes; dev_priv->cdclk.force_min_cdclk = state->cdclk.force_min_cdclk; intel_cdclk_swap_state(state); @@ -14204,7 +14921,7 @@ static int intel_atomic_commit(struct drm_device *dev, if (nonblock && state->modeset) { queue_work(dev_priv->modeset_wq, &state->base.commit_work); } else if (nonblock) { - queue_work(system_unbound_wq, &state->base.commit_work); + queue_work(dev_priv->flip_wq, &state->base.commit_work); } else { if (state->modeset) flush_workqueue(dev_priv->modeset_wq); @@ -14233,7 +14950,7 @@ static int do_rps_boost(struct wait_queue_entry *_wait, * vblank without our intervention, so leave RPS alone. */ if (!i915_request_started(rq)) - gen6_rps_boost(rq); + intel_rps_boost(rq); i915_request_put(rq); drm_crtc_vblank_put(wait->crtc); @@ -14314,7 +15031,7 @@ static void intel_plane_unpin_fb(struct intel_plane_state *old_plane_state) static void fb_obj_bump_render_priority(struct drm_i915_gem_object *obj) { struct i915_sched_attr attr = { - .priority = I915_PRIORITY_DISPLAY, + .priority = I915_USER_PRIORITY(I915_PRIORITY_DISPLAY), }; i915_gem_object_wait_priority(obj, 0, &attr); @@ -14323,25 +15040,25 @@ static void fb_obj_bump_render_priority(struct drm_i915_gem_object *obj) /** * intel_prepare_plane_fb - Prepare fb for usage on plane * @plane: drm plane to prepare for - * @new_state: the plane state being prepared + * @_new_plane_state: the plane state being prepared * * Prepares a framebuffer for usage on a display plane. Generally this * involves pinning the underlying object and updating the frontbuffer tracking * bits. Some older platforms need special physical address handling for * cursor planes. * - * Must be called with struct_mutex held. - * * Returns 0 on success, negative error code on failure. */ int intel_prepare_plane_fb(struct drm_plane *plane, - struct drm_plane_state *new_state) + struct drm_plane_state *_new_plane_state) { + struct intel_plane_state *new_plane_state = + to_intel_plane_state(_new_plane_state); struct intel_atomic_state *intel_state = - to_intel_atomic_state(new_state->state); + to_intel_atomic_state(new_plane_state->base.state); struct drm_i915_private *dev_priv = to_i915(plane->dev); - struct drm_framebuffer *fb = new_state->fb; + struct drm_framebuffer *fb = new_plane_state->base.fb; struct drm_i915_gem_object *obj = intel_fb_obj(fb); struct drm_i915_gem_object *old_obj = intel_fb_obj(plane->state->fb); int ret; @@ -14372,9 +15089,9 @@ intel_prepare_plane_fb(struct drm_plane *plane, } } - if (new_state->fence) { /* explicit fencing */ + if (new_plane_state->base.fence) { /* explicit fencing */ ret = i915_sw_fence_await_dma_fence(&intel_state->commit_ready, - new_state->fence, + new_plane_state->base.fence, I915_FENCE_TIMEOUT, GFP_KERNEL); if (ret < 0) @@ -14388,15 +15105,8 @@ intel_prepare_plane_fb(struct drm_plane *plane, if (ret) return ret; - ret = mutex_lock_interruptible(&dev_priv->drm.struct_mutex); - if (ret) { - i915_gem_object_unpin_pages(obj); - return ret; - } - - ret = intel_plane_pin_fb(to_intel_plane_state(new_state)); + ret = intel_plane_pin_fb(new_plane_state); - mutex_unlock(&dev_priv->drm.struct_mutex); i915_gem_object_unpin_pages(obj); if (ret) return ret; @@ -14404,7 +15114,7 @@ intel_prepare_plane_fb(struct drm_plane *plane, fb_obj_bump_render_priority(obj); intel_frontbuffer_flush(obj->frontbuffer, ORIGIN_DIRTYFB); - if (!new_state->fence) { /* implicit fencing */ + if (!new_plane_state->base.fence) { /* implicit fencing */ struct dma_fence *fence; ret = i915_sw_fence_await_reservation(&intel_state->commit_ready, @@ -14416,11 +15126,13 @@ intel_prepare_plane_fb(struct drm_plane *plane, fence = dma_resv_get_excl_rcu(obj->base.resv); if (fence) { - add_rps_boost_after_vblank(new_state->crtc, fence); + add_rps_boost_after_vblank(new_plane_state->base.crtc, + fence); dma_fence_put(fence); } } else { - add_rps_boost_after_vblank(new_state->crtc, new_state->fence); + add_rps_boost_after_vblank(new_plane_state->base.crtc, + new_plane_state->base.fence); } /* @@ -14432,7 +15144,7 @@ intel_prepare_plane_fb(struct drm_plane *plane, * maximum clocks following a vblank miss (see do_rps_boost()). */ if (!intel_state->rps_interactive) { - intel_rps_mark_interactive(dev_priv, true); + intel_rps_mark_interactive(&dev_priv->gt.rps, true); intel_state->rps_interactive = true; } @@ -14442,130 +15154,27 @@ intel_prepare_plane_fb(struct drm_plane *plane, /** * intel_cleanup_plane_fb - Cleans up an fb after plane use * @plane: drm plane to clean up for - * @old_state: the state from the previous modeset + * @_old_plane_state: the state from the previous modeset * * Cleans up a framebuffer that has just been removed from a plane. - * - * Must be called with struct_mutex held. */ void intel_cleanup_plane_fb(struct drm_plane *plane, - struct drm_plane_state *old_state) + struct drm_plane_state *_old_plane_state) { + struct intel_plane_state *old_plane_state = + to_intel_plane_state(_old_plane_state); struct intel_atomic_state *intel_state = - to_intel_atomic_state(old_state->state); + to_intel_atomic_state(old_plane_state->base.state); struct drm_i915_private *dev_priv = to_i915(plane->dev); if (intel_state->rps_interactive) { - intel_rps_mark_interactive(dev_priv, false); + intel_rps_mark_interactive(&dev_priv->gt.rps, false); intel_state->rps_interactive = false; } /* Should only be called after a successful intel_prepare_plane_fb()! */ - mutex_lock(&dev_priv->drm.struct_mutex); - intel_plane_unpin_fb(to_intel_plane_state(old_state)); - mutex_unlock(&dev_priv->drm.struct_mutex); -} - -int -skl_max_scale(const struct intel_crtc_state *crtc_state, - u32 pixel_format) -{ - struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); - struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - int max_scale, mult; - int crtc_clock, max_dotclk, tmpclk1, tmpclk2; - - if (!crtc_state->base.enable) - return DRM_PLANE_HELPER_NO_SCALING; - - crtc_clock = crtc_state->base.adjusted_mode.crtc_clock; - max_dotclk = to_intel_atomic_state(crtc_state->base.state)->cdclk.logical.cdclk; - - if (IS_GEMINILAKE(dev_priv) || INTEL_GEN(dev_priv) >= 10) - max_dotclk *= 2; - - if (WARN_ON_ONCE(!crtc_clock || max_dotclk < crtc_clock)) - return DRM_PLANE_HELPER_NO_SCALING; - - /* - * skl max scale is lower of: - * close to 3 but not 3, -1 is for that purpose - * or - * cdclk/crtc_clock - */ - mult = is_planar_yuv_format(pixel_format) ? 2 : 3; - tmpclk1 = (1 << 16) * mult - 1; - tmpclk2 = (1 << 8) * ((max_dotclk << 8) / crtc_clock); - max_scale = min(tmpclk1, tmpclk2); - - return max_scale; -} - -static void intel_begin_crtc_commit(struct intel_atomic_state *state, - struct intel_crtc *crtc) -{ - struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - struct intel_crtc_state *old_crtc_state = - intel_atomic_get_old_crtc_state(state, crtc); - struct intel_crtc_state *new_crtc_state = - intel_atomic_get_new_crtc_state(state, crtc); - bool modeset = needs_modeset(new_crtc_state); - - /* Perform vblank evasion around commit operation */ - intel_pipe_update_start(new_crtc_state); - - if (modeset) - goto out; - - if (new_crtc_state->base.color_mgmt_changed || - new_crtc_state->update_pipe) - intel_color_commit(new_crtc_state); - - if (new_crtc_state->update_pipe) - intel_update_pipe_config(old_crtc_state, new_crtc_state); - else if (INTEL_GEN(dev_priv) >= 9) - skl_detach_scalers(new_crtc_state); - - if (INTEL_GEN(dev_priv) >= 9 || IS_BROADWELL(dev_priv)) - bdw_set_pipemisc(new_crtc_state); - -out: - if (dev_priv->display.atomic_update_watermarks) - dev_priv->display.atomic_update_watermarks(state, - new_crtc_state); -} - -void intel_crtc_arm_fifo_underrun(struct intel_crtc *crtc, - struct intel_crtc_state *crtc_state) -{ - struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - - if (!IS_GEN(dev_priv, 2)) - intel_set_cpu_fifo_underrun_reporting(dev_priv, crtc->pipe, true); - - if (crtc_state->has_pch_encoder) { - enum pipe pch_transcoder = - intel_crtc_pch_transcoder(crtc); - - intel_set_pch_fifo_underrun_reporting(dev_priv, pch_transcoder, true); - } -} - -static void intel_finish_crtc_commit(struct intel_atomic_state *state, - struct intel_crtc *crtc) -{ - struct intel_crtc_state *old_crtc_state = - intel_atomic_get_old_crtc_state(state, crtc); - struct intel_crtc_state *new_crtc_state = - intel_atomic_get_new_crtc_state(state, crtc); - - intel_pipe_update_end(new_crtc_state); - - if (new_crtc_state->update_pipe && - !needs_modeset(new_crtc_state) && - old_crtc_state->base.mode.private_flags & I915_MODE_FLAG_INHERITED) - intel_crtc_arm_fifo_underrun(crtc, new_crtc_state); + intel_plane_unpin_fb(old_plane_state); } /** @@ -14622,6 +15231,7 @@ static bool i965_plane_format_mod_supported(struct drm_plane *_plane, case DRM_FORMAT_XBGR8888: case DRM_FORMAT_XRGB2101010: case DRM_FORMAT_XBGR2101010: + case DRM_FORMAT_XBGR16161616F: return modifier == DRM_FORMAT_MOD_LINEAR || modifier == I915_FORMAT_MOD_X_TILED; default: @@ -14655,8 +15265,8 @@ static const struct drm_plane_funcs i8xx_plane_funcs = { }; static int -intel_legacy_cursor_update(struct drm_plane *plane, - struct drm_crtc *crtc, +intel_legacy_cursor_update(struct drm_plane *_plane, + struct drm_crtc *_crtc, struct drm_framebuffer *fb, int crtc_x, int crtc_y, unsigned int crtc_w, unsigned int crtc_h, @@ -14664,11 +15274,13 @@ intel_legacy_cursor_update(struct drm_plane *plane, u32 src_w, u32 src_h, struct drm_modeset_acquire_ctx *ctx) { - struct drm_i915_private *dev_priv = to_i915(crtc->dev); - struct drm_plane_state *old_plane_state, *new_plane_state; - struct intel_plane *intel_plane = to_intel_plane(plane); + struct intel_plane *plane = to_intel_plane(_plane); + struct intel_crtc *crtc = to_intel_crtc(_crtc); + struct intel_plane_state *old_plane_state = + to_intel_plane_state(plane->base.state); + struct intel_plane_state *new_plane_state; struct intel_crtc_state *crtc_state = - to_intel_crtc_state(crtc->state); + to_intel_crtc_state(crtc->base.state); struct intel_crtc_state *new_crtc_state; int ret; @@ -14680,14 +15292,13 @@ intel_legacy_cursor_update(struct drm_plane *plane, crtc_state->update_pipe) goto slow; - old_plane_state = plane->state; /* * Don't do an async update if there is an outstanding commit modifying * the plane. This prevents our async update's changes from getting * overridden by a previous synchronous update's state. */ - if (old_plane_state->commit && - !try_wait_for_completion(&old_plane_state->commit->hw_done)) + if (old_plane_state->base.commit && + !try_wait_for_completion(&old_plane_state->base.commit->hw_done)) goto slow; /* @@ -14695,56 +15306,51 @@ intel_legacy_cursor_update(struct drm_plane *plane, * take the slowpath. Only changing fb or position should be * in the fastpath. */ - if (old_plane_state->crtc != crtc || - old_plane_state->src_w != src_w || - old_plane_state->src_h != src_h || - old_plane_state->crtc_w != crtc_w || - old_plane_state->crtc_h != crtc_h || - !old_plane_state->fb != !fb) + if (old_plane_state->base.crtc != &crtc->base || + old_plane_state->base.src_w != src_w || + old_plane_state->base.src_h != src_h || + old_plane_state->base.crtc_w != crtc_w || + old_plane_state->base.crtc_h != crtc_h || + !old_plane_state->base.fb != !fb) goto slow; - new_plane_state = intel_plane_duplicate_state(plane); + new_plane_state = to_intel_plane_state(intel_plane_duplicate_state(&plane->base)); if (!new_plane_state) return -ENOMEM; - new_crtc_state = to_intel_crtc_state(intel_crtc_duplicate_state(crtc)); + new_crtc_state = to_intel_crtc_state(intel_crtc_duplicate_state(&crtc->base)); if (!new_crtc_state) { ret = -ENOMEM; goto out_free; } - drm_atomic_set_fb_for_plane(new_plane_state, fb); + drm_atomic_set_fb_for_plane(&new_plane_state->base, fb); - new_plane_state->src_x = src_x; - new_plane_state->src_y = src_y; - new_plane_state->src_w = src_w; - new_plane_state->src_h = src_h; - new_plane_state->crtc_x = crtc_x; - new_plane_state->crtc_y = crtc_y; - new_plane_state->crtc_w = crtc_w; - new_plane_state->crtc_h = crtc_h; + new_plane_state->base.src_x = src_x; + new_plane_state->base.src_y = src_y; + new_plane_state->base.src_w = src_w; + new_plane_state->base.src_h = src_h; + new_plane_state->base.crtc_x = crtc_x; + new_plane_state->base.crtc_y = crtc_y; + new_plane_state->base.crtc_w = crtc_w; + new_plane_state->base.crtc_h = crtc_h; ret = intel_plane_atomic_check_with_state(crtc_state, new_crtc_state, - to_intel_plane_state(old_plane_state), - to_intel_plane_state(new_plane_state)); + old_plane_state, new_plane_state); if (ret) goto out_free; - ret = mutex_lock_interruptible(&dev_priv->drm.struct_mutex); + ret = intel_plane_pin_fb(new_plane_state); if (ret) goto out_free; - ret = intel_plane_pin_fb(to_intel_plane_state(new_plane_state)); - if (ret) - goto out_unlock; - - intel_frontbuffer_flush(to_intel_frontbuffer(fb), ORIGIN_FLIP); - intel_frontbuffer_track(to_intel_frontbuffer(old_plane_state->fb), - to_intel_frontbuffer(fb), - intel_plane->frontbuffer_bit); + intel_frontbuffer_flush(to_intel_frontbuffer(new_plane_state->base.fb), ORIGIN_FLIP); + intel_frontbuffer_track(to_intel_frontbuffer(old_plane_state->base.fb), + to_intel_frontbuffer(new_plane_state->base.fb), + plane->frontbuffer_bit); /* Swap plane state */ - plane->state = new_plane_state; + plane->base.state = &new_plane_state->base; /* * We cannot swap crtc_state as it may be in use by an atomic commit or @@ -14758,27 +15364,24 @@ intel_legacy_cursor_update(struct drm_plane *plane, */ crtc_state->active_planes = new_crtc_state->active_planes; - if (plane->state->visible) - intel_update_plane(intel_plane, crtc_state, - to_intel_plane_state(plane->state)); + if (new_plane_state->base.visible) + intel_update_plane(plane, crtc_state, new_plane_state); else - intel_disable_plane(intel_plane, crtc_state); + intel_disable_plane(plane, crtc_state); - intel_plane_unpin_fb(to_intel_plane_state(old_plane_state)); + intel_plane_unpin_fb(old_plane_state); -out_unlock: - mutex_unlock(&dev_priv->drm.struct_mutex); out_free: if (new_crtc_state) - intel_crtc_destroy_state(crtc, &new_crtc_state->base); + intel_crtc_destroy_state(&crtc->base, &new_crtc_state->base); if (ret) - intel_plane_destroy_state(plane, new_plane_state); + intel_plane_destroy_state(&plane->base, &new_plane_state->base); else - intel_plane_destroy_state(plane, old_plane_state); + intel_plane_destroy_state(&plane->base, &old_plane_state->base); return ret; slow: - return drm_atomic_helper_update_plane(plane, crtc, fb, + return drm_atomic_helper_update_plane(&plane->base, &crtc->base, fb, crtc_x, crtc_y, crtc_w, crtc_h, src_x, src_y, src_w, src_h, ctx); } @@ -14819,7 +15422,7 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) const u64 *modifiers; const u32 *formats; int num_formats; - int ret; + int ret, zpos; if (INTEL_GEN(dev_priv) >= 9) return skl_universal_plane_create(dev_priv, pipe, @@ -14849,8 +15452,26 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) } if (INTEL_GEN(dev_priv) >= 4) { - formats = i965_primary_formats; - num_formats = ARRAY_SIZE(i965_primary_formats); + /* + * WaFP16GammaEnabling:ivb + * "Workaround : When using the 64-bit format, the plane + * output on each color channel has one quarter amplitude. + * It can be brought up to full amplitude by using pipe + * gamma correction or pipe color space conversion to + * multiply the plane output by four." + * + * There is no dedicated plane gamma for the primary plane, + * and using the pipe gamma/csc could conflict with other + * planes, so we choose not to expose fp16 on IVB primary + * planes. HSW primary planes no longer have this problem. + */ + if (IS_IVYBRIDGE(dev_priv)) { + formats = ivb_primary_formats; + num_formats = ARRAY_SIZE(ivb_primary_formats); + } else { + formats = i965_primary_formats; + num_formats = ARRAY_SIZE(i965_primary_formats); + } modifiers = i9xx_format_modifiers; plane->max_stride = i9xx_plane_max_stride; @@ -14859,6 +15480,15 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) plane->get_hw_state = i9xx_plane_get_hw_state; plane->check_plane = i9xx_plane_check; + if (IS_BROADWELL(dev_priv) || IS_HASWELL(dev_priv)) + plane->min_cdclk = hsw_plane_min_cdclk; + else if (IS_IVYBRIDGE(dev_priv)) + plane->min_cdclk = ivb_plane_min_cdclk; + else if (IS_CHERRYVIEW(dev_priv) || IS_VALLEYVIEW(dev_priv)) + plane->min_cdclk = vlv_plane_min_cdclk; + else + plane->min_cdclk = i9xx_plane_min_cdclk; + plane_funcs = &i965_plane_funcs; } else { formats = i8xx_primary_formats; @@ -14870,6 +15500,7 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) plane->disable_plane = i9xx_disable_plane; plane->get_hw_state = i9xx_plane_get_hw_state; plane->check_plane = i9xx_plane_check; + plane->min_cdclk = i9xx_plane_min_cdclk; plane_funcs = &i8xx_plane_funcs; } @@ -14908,6 +15539,9 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) DRM_MODE_ROTATE_0, supported_rotations); + zpos = 0; + drm_plane_create_zpos_immutable_property(&plane->base, zpos); + drm_plane_helper_add(&plane->base, &intel_plane_helper_funcs); return plane; @@ -14924,7 +15558,7 @@ intel_cursor_plane_create(struct drm_i915_private *dev_priv, { unsigned int possible_crtcs; struct intel_plane *cursor; - int ret; + int ret, zpos; cursor = intel_plane_alloc(); if (IS_ERR(cursor)) @@ -14973,6 +15607,9 @@ intel_cursor_plane_create(struct drm_i915_private *dev_priv, DRM_MODE_ROTATE_0 | DRM_MODE_ROTATE_180); + zpos = RUNTIME_INFO(dev_priv)->num_sprites[pipe] + 1; + drm_plane_create_zpos_immutable_property(&cursor->base, zpos); + drm_plane_helper_add(&cursor->base, &intel_plane_helper_funcs); return cursor; @@ -15048,12 +15685,12 @@ static const struct drm_crtc_funcs i965_crtc_funcs = { .disable_vblank = i965_disable_vblank, }; -static const struct drm_crtc_funcs i945gm_crtc_funcs = { +static const struct drm_crtc_funcs i915gm_crtc_funcs = { INTEL_CRTC_FUNCS, .get_vblank_counter = i915_get_vblank_counter, - .enable_vblank = i945gm_enable_vblank, - .disable_vblank = i945gm_disable_vblank, + .enable_vblank = i915gm_enable_vblank, + .disable_vblank = i915gm_disable_vblank, }; static const struct drm_crtc_funcs i915_crtc_funcs = { @@ -15124,8 +15761,8 @@ static int intel_crtc_init(struct drm_i915_private *dev_priv, enum pipe pipe) funcs = &g4x_crtc_funcs; else if (IS_GEN(dev_priv, 4)) funcs = &i965_crtc_funcs; - else if (IS_I945GM(dev_priv)) - funcs = &i945gm_crtc_funcs; + else if (IS_I945GM(dev_priv) || IS_I915GM(dev_priv)) + funcs = &i915gm_crtc_funcs; else if (IS_GEN(dev_priv, 3)) funcs = &i915_crtc_funcs; else @@ -15160,8 +15797,6 @@ static int intel_crtc_init(struct drm_i915_private *dev_priv, enum pipe pipe) dev_priv->plane_to_crtc_mapping[i9xx_plane] = intel_crtc; } - drm_crtc_helper_add(&intel_crtc->base, &intel_helper_funcs); - intel_color_init(intel_crtc); WARN_ON(drm_crtc_index(&intel_crtc->base) != intel_crtc->pipe); @@ -15196,21 +15831,32 @@ int intel_get_pipe_from_crtc_id_ioctl(struct drm_device *dev, void *data, return 0; } -static int intel_encoder_clones(struct intel_encoder *encoder) +static u32 intel_encoder_possible_clones(struct intel_encoder *encoder) { struct drm_device *dev = encoder->base.dev; struct intel_encoder *source_encoder; - int index_mask = 0; - int entry = 0; + u32 possible_clones = 0; for_each_intel_encoder(dev, source_encoder) { if (encoders_cloneable(encoder, source_encoder)) - index_mask |= (1 << entry); + possible_clones |= drm_encoder_mask(&source_encoder->base); + } + + return possible_clones; +} + +static u32 intel_encoder_possible_crtcs(struct intel_encoder *encoder) +{ + struct drm_device *dev = encoder->base.dev; + struct intel_crtc *crtc; + u32 possible_crtcs = 0; - entry++; + for_each_intel_crtc(dev, crtc) { + if (encoder->pipe_mask & BIT(crtc->pipe)) + possible_crtcs |= drm_crtc_mask(&crtc->base); } - return index_mask; + return possible_crtcs; } static bool ilk_has_edp_a(struct drm_i915_private *dev_priv) @@ -15292,13 +15938,18 @@ static void intel_setup_outputs(struct drm_i915_private *dev_priv) intel_pps_init(dev_priv); - if (!HAS_DISPLAY(dev_priv)) + if (!HAS_DISPLAY(dev_priv) || !INTEL_DISPLAY_ENABLED(dev_priv)) return; if (INTEL_GEN(dev_priv) >= 12) { - /* TODO: initialize TC ports as well */ intel_ddi_init(dev_priv, PORT_A); intel_ddi_init(dev_priv, PORT_B); + intel_ddi_init(dev_priv, PORT_D); + intel_ddi_init(dev_priv, PORT_E); + intel_ddi_init(dev_priv, PORT_F); + intel_ddi_init(dev_priv, PORT_G); + intel_ddi_init(dev_priv, PORT_H); + intel_ddi_init(dev_priv, PORT_I); icl_dsi_init(dev_priv); } else if (IS_ELKHARTLAKE(dev_priv)) { intel_ddi_init(dev_priv, PORT_A); @@ -15508,9 +16159,10 @@ static void intel_setup_outputs(struct drm_i915_private *dev_priv) intel_psr_init(dev_priv); for_each_intel_encoder(&dev_priv->drm, encoder) { - encoder->base.possible_crtcs = encoder->crtc_mask; + encoder->base.possible_crtcs = + intel_encoder_possible_crtcs(encoder); encoder->base.possible_clones = - intel_encoder_clones(encoder); + intel_encoder_possible_clones(encoder); } intel_init_pch_refclk(dev_priv); @@ -15765,8 +16417,14 @@ intel_mode_valid(struct drm_device *dev, DRM_MODE_FLAG_CLKDIV2)) return MODE_BAD; - if (INTEL_GEN(dev_priv) >= 9 || - IS_BROADWELL(dev_priv) || IS_HASWELL(dev_priv)) { + /* Transcoder timing limits */ + if (INTEL_GEN(dev_priv) >= 11) { + hdisplay_max = 16384; + vdisplay_max = 8192; + htotal_max = 16384; + vtotal_max = 8192; + } else if (INTEL_GEN(dev_priv) >= 9 || + IS_BROADWELL(dev_priv) || IS_HASWELL(dev_priv)) { hdisplay_max = 8192; /* FDI max 4096 handled elsewhere */ vdisplay_max = 4096; htotal_max = 8192; @@ -15795,6 +16453,56 @@ intel_mode_valid(struct drm_device *dev, mode->vtotal > vtotal_max) return MODE_V_ILLEGAL; + if (INTEL_GEN(dev_priv) >= 5) { + if (mode->hdisplay < 64 || + mode->htotal - mode->hdisplay < 32) + return MODE_H_ILLEGAL; + + if (mode->vtotal - mode->vdisplay < 5) + return MODE_V_ILLEGAL; + } else { + if (mode->htotal - mode->hdisplay < 32) + return MODE_H_ILLEGAL; + + if (mode->vtotal - mode->vdisplay < 3) + return MODE_V_ILLEGAL; + } + + return MODE_OK; +} + +enum drm_mode_status +intel_mode_valid_max_plane_size(struct drm_i915_private *dev_priv, + const struct drm_display_mode *mode) +{ + int plane_width_max, plane_height_max; + + /* + * intel_mode_valid() should be + * sufficient on older platforms. + */ + if (INTEL_GEN(dev_priv) < 9) + return MODE_OK; + + /* + * Most people will probably want a fullscreen + * plane so let's not advertize modes that are + * too big for that. + */ + if (INTEL_GEN(dev_priv) >= 11) { + plane_width_max = 5120; + plane_height_max = 4320; + } else { + plane_width_max = 5120; + plane_height_max = 4096; + } + + if (mode->hdisplay > plane_width_max) + return MODE_H_ILLEGAL; + + if (mode->vdisplay > plane_height_max) + return MODE_V_ILLEGAL; + return MODE_OK; } @@ -15898,47 +16606,17 @@ void intel_init_display_hooks(struct drm_i915_private *dev_priv) } if (INTEL_GEN(dev_priv) >= 9) - dev_priv->display.update_crtcs = skl_update_crtcs; + dev_priv->display.commit_modeset_enables = skl_commit_modeset_enables; else - dev_priv->display.update_crtcs = intel_update_crtcs; -} + dev_priv->display.commit_modeset_enables = intel_commit_modeset_enables; -static i915_reg_t i915_vgacntrl_reg(struct drm_i915_private *dev_priv) -{ - if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) - return VLV_VGACNTRL; - else if (INTEL_GEN(dev_priv) >= 5) - return CPU_VGACNTRL; - else - return VGACNTRL; } -/* Disable the VGA plane that we never use */ -static void i915_disable_vga(struct drm_i915_private *dev_priv) +void intel_modeset_init_hw(struct drm_i915_private *i915) { - struct pci_dev *pdev = dev_priv->drm.pdev; - u8 sr1; - i915_reg_t vga_reg = i915_vgacntrl_reg(dev_priv); - - /* WaEnableVGAAccessThroughIOPort:ctg,elk,ilk,snb,ivb,vlv,hsw */ - vga_get_uninterruptible(pdev, VGA_RSRC_LEGACY_IO); - outb(SR01, VGA_SR_INDEX); - sr1 = inb(VGA_SR_DATA); - outb(sr1 | 1<<5, VGA_SR_DATA); - vga_put(pdev, VGA_RSRC_LEGACY_IO); - udelay(300); - - I915_WRITE(vga_reg, VGA_DISP_DISABLE); - POSTING_READ(vga_reg); -} - -void intel_modeset_init_hw(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = to_i915(dev); - - intel_update_cdclk(dev_priv); - intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); - dev_priv->cdclk.logical = dev_priv->cdclk.actual = dev_priv->cdclk.hw; + intel_update_cdclk(i915); + intel_dump_cdclk_state(&i915->cdclk.hw, "Current CDCLK"); + i915->cdclk.logical = i915->cdclk.actual = i915->cdclk.hw; } /* @@ -16098,114 +16776,111 @@ out: return ret; } -int intel_modeset_init(struct drm_device *dev) +static void intel_mode_config_init(struct drm_i915_private *i915) { - struct drm_i915_private *dev_priv = to_i915(dev); - enum pipe pipe; - struct intel_crtc *crtc; - int ret; - - dev_priv->modeset_wq = alloc_ordered_workqueue("i915_modeset", 0); - - drm_mode_config_init(dev); - - ret = intel_bw_init(dev_priv); - if (ret) - return ret; - - dev->mode_config.min_width = 0; - dev->mode_config.min_height = 0; + struct drm_mode_config *mode_config = &i915->drm.mode_config; - dev->mode_config.preferred_depth = 24; - dev->mode_config.prefer_shadow = 1; + drm_mode_config_init(&i915->drm); - dev->mode_config.allow_fb_modifiers = true; + mode_config->min_width = 0; + mode_config->min_height = 0; - dev->mode_config.funcs = &intel_mode_funcs; + mode_config->preferred_depth = 24; + mode_config->prefer_shadow = 1; - init_llist_head(&dev_priv->atomic_helper.free_list); - INIT_WORK(&dev_priv->atomic_helper.free_work, - intel_atomic_helper_free_state_worker); - - intel_init_quirks(dev_priv); + mode_config->allow_fb_modifiers = true; - intel_fbc_init(dev_priv); - - intel_init_pm(dev_priv); - - /* - * There may be no VBT; and if the BIOS enabled SSC we can - * just keep using it to avoid unnecessary flicker. Whereas if the - * BIOS isn't using it, don't assume it will work even if the VBT - * indicates as much. - */ - if (HAS_PCH_IBX(dev_priv) || HAS_PCH_CPT(dev_priv)) { - bool bios_lvds_use_ssc = !!(I915_READ(PCH_DREF_CONTROL) & - DREF_SSC1_ENABLE); - - if (dev_priv->vbt.lvds_use_ssc != bios_lvds_use_ssc) { - DRM_DEBUG_KMS("SSC %sabled by BIOS, overriding VBT which says %sabled\n", - bios_lvds_use_ssc ? "en" : "dis", - dev_priv->vbt.lvds_use_ssc ? "en" : "dis"); - dev_priv->vbt.lvds_use_ssc = bios_lvds_use_ssc; - } - } + mode_config->funcs = &intel_mode_funcs; /* * Maximum framebuffer dimensions, chosen to match * the maximum render engine surface size on gen4+. */ - if (INTEL_GEN(dev_priv) >= 7) { - dev->mode_config.max_width = 16384; - dev->mode_config.max_height = 16384; - } else if (INTEL_GEN(dev_priv) >= 4) { - dev->mode_config.max_width = 8192; - dev->mode_config.max_height = 8192; - } else if (IS_GEN(dev_priv, 3)) { - dev->mode_config.max_width = 4096; - dev->mode_config.max_height = 4096; + if (INTEL_GEN(i915) >= 7) { + mode_config->max_width = 16384; + mode_config->max_height = 16384; + } else if (INTEL_GEN(i915) >= 4) { + mode_config->max_width = 8192; + mode_config->max_height = 8192; + } else if (IS_GEN(i915, 3)) { + mode_config->max_width = 4096; + mode_config->max_height = 4096; } else { - dev->mode_config.max_width = 2048; - dev->mode_config.max_height = 2048; + mode_config->max_width = 2048; + mode_config->max_height = 2048; } - if (IS_I845G(dev_priv) || IS_I865G(dev_priv)) { - dev->mode_config.cursor_width = IS_I845G(dev_priv) ? 64 : 512; - dev->mode_config.cursor_height = 1023; - } else if (IS_GEN(dev_priv, 2)) { - dev->mode_config.cursor_width = 64; - dev->mode_config.cursor_height = 64; + if (IS_I845G(i915) || IS_I865G(i915)) { + mode_config->cursor_width = IS_I845G(i915) ? 64 : 512; + mode_config->cursor_height = 1023; + } else if (IS_GEN(i915, 2)) { + mode_config->cursor_width = 64; + mode_config->cursor_height = 64; } else { - dev->mode_config.cursor_width = 256; - dev->mode_config.cursor_height = 256; + mode_config->cursor_width = 256; + mode_config->cursor_height = 256; } +} - DRM_DEBUG_KMS("%d display pipe%s available.\n", - INTEL_INFO(dev_priv)->num_pipes, - INTEL_INFO(dev_priv)->num_pipes > 1 ? "s" : ""); +int intel_modeset_init(struct drm_i915_private *i915) +{ + struct drm_device *dev = &i915->drm; + enum pipe pipe; + struct intel_crtc *crtc; + int ret; - for_each_pipe(dev_priv, pipe) { - ret = intel_crtc_init(dev_priv, pipe); - if (ret) { - drm_mode_config_cleanup(dev); - return ret; + i915->modeset_wq = alloc_ordered_workqueue("i915_modeset", 0); + i915->flip_wq = alloc_workqueue("i915_flip", WQ_HIGHPRI | + WQ_UNBOUND, WQ_UNBOUND_MAX_ACTIVE); + + intel_mode_config_init(i915); + + ret = intel_bw_init(i915); + if (ret) + return ret; + + init_llist_head(&i915->atomic_helper.free_list); + INIT_WORK(&i915->atomic_helper.free_work, + intel_atomic_helper_free_state_worker); + + intel_init_quirks(i915); + + intel_fbc_init(i915); + + intel_init_pm(i915); + + intel_panel_sanitize_ssc(i915); + + intel_gmbus_setup(i915); + + DRM_DEBUG_KMS("%d display pipe%s available.\n", + INTEL_NUM_PIPES(i915), + INTEL_NUM_PIPES(i915) > 1 ? "s" : ""); + + if (HAS_DISPLAY(i915) && INTEL_DISPLAY_ENABLED(i915)) { + for_each_pipe(i915, pipe) { + ret = intel_crtc_init(i915, pipe); + if (ret) { + drm_mode_config_cleanup(dev); + return ret; + } } } intel_shared_dpll_init(dev); - intel_update_fdi_pll_freq(dev_priv); + intel_update_fdi_pll_freq(i915); - intel_update_czclk(dev_priv); - intel_modeset_init_hw(dev); + intel_update_czclk(i915); + intel_modeset_init_hw(i915); - intel_hdcp_component_init(dev_priv); + intel_hdcp_component_init(i915); - if (dev_priv->max_cdclk_freq == 0) - intel_update_max_cdclk(dev_priv); + if (i915->max_cdclk_freq == 0) + intel_update_max_cdclk(i915); /* Just disable it once at startup */ - i915_disable_vga(dev_priv); - intel_setup_outputs(dev_priv); + intel_vga_disable(i915); + intel_setup_outputs(i915); drm_modeset_lock_all(dev); intel_modeset_setup_hw_state(dev, dev->mode_config.acquire_ctx); @@ -16224,8 +16899,7 @@ int intel_modeset_init(struct drm_device *dev) * can even allow for smooth boot transitions if the BIOS * fb is large enough for the active pipe configuration. */ - dev_priv->display.get_initial_plane_config(crtc, - &plane_config); + i915->display.get_initial_plane_config(crtc, &plane_config); /* * If the fb is shared between multiple heads, we'll @@ -16239,7 +16913,7 @@ int intel_modeset_init(struct drm_device *dev) * Note that we need to do this after reconstructing the BIOS fb's * since the watermark calculation done here will use pstate->fb. */ - if (!HAS_GMCH(dev_priv)) + if (!HAS_GMCH(i915)) sanitize_watermarks(dev); /* @@ -16564,39 +17238,6 @@ static void intel_sanitize_encoder(struct intel_encoder *encoder) icl_sanitize_encoder_pll_mapping(encoder); } -void i915_redisable_vga_power_on(struct drm_i915_private *dev_priv) -{ - i915_reg_t vga_reg = i915_vgacntrl_reg(dev_priv); - - if (!(I915_READ(vga_reg) & VGA_DISP_DISABLE)) { - DRM_DEBUG_KMS("Something enabled VGA plane, disabling it\n"); - i915_disable_vga(dev_priv); - } -} - -void i915_redisable_vga(struct drm_i915_private *dev_priv) -{ - intel_wakeref_t wakeref; - - /* - * This function can be called both from intel_modeset_setup_hw_state or - * at a very early point in our resume sequence, where the power well - * structures are not yet restored. Since this function is at a very - * paranoid "someone might have enabled VGA while we were not looking" - * level, just check if the power well is enabled instead of trying to - * follow the "don't touch the power well if we don't need it" policy - * the rest of the driver uses. - */ - wakeref = intel_display_power_get_if_enabled(dev_priv, - POWER_DOMAIN_VGA); - if (!wakeref) - return; - - i915_redisable_vga_power_on(dev_priv); - - intel_display_power_put(dev_priv, POWER_DOMAIN_VGA, wakeref); -} - /* FIXME read out full plane state for all planes */ static void readout_plane_state(struct drm_i915_private *dev_priv) { @@ -16640,7 +17281,7 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) struct drm_connector_list_iter conn_iter; int i; - dev_priv->active_crtcs = 0; + dev_priv->active_pipes = 0; for_each_intel_crtc(dev, crtc) { struct intel_crtc_state *crtc_state = @@ -16657,7 +17298,7 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) crtc->active = crtc_state->base.active; if (crtc_state->base.active) - dev_priv->active_crtcs |= 1 << crtc->pipe; + dev_priv->active_pipes |= BIT(crtc->pipe); DRM_DEBUG_KMS("[CRTC:%d:%s] hw state readout: %s\n", crtc->base.base.id, crtc->base.name, @@ -16717,24 +17358,28 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) drm_connector_list_iter_begin(dev, &conn_iter); for_each_intel_connector_iter(connector, &conn_iter) { if (connector->get_hw_state(connector)) { + struct intel_crtc_state *crtc_state; + struct intel_crtc *crtc; + connector->base.dpms = DRM_MODE_DPMS_ON; encoder = connector->encoder; connector->base.encoder = &encoder->base; - if (encoder->base.crtc && - encoder->base.crtc->state->active) { + crtc = to_intel_crtc(encoder->base.crtc); + crtc_state = crtc ? to_intel_crtc_state(crtc->base.state) : NULL; + + if (crtc_state && crtc_state->base.active) { /* * This has to be done during hardware readout * because anything calling .crtc_disable may * rely on the connector_mask being accurate. */ - encoder->base.crtc->state->connector_mask |= + crtc_state->base.connector_mask |= drm_connector_mask(&connector->base); - encoder->base.crtc->state->encoder_mask |= + crtc_state->base.encoder_mask |= drm_encoder_mask(&encoder->base); } - } else { connector->base.dpms = DRM_MODE_DPMS_OFF; connector->base.encoder = NULL; @@ -16753,13 +17398,16 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) struct intel_plane *plane; int min_cdclk = 0; - memset(&crtc->base.mode, 0, sizeof(crtc->base.mode)); if (crtc_state->base.active) { - intel_mode_from_pipe_config(&crtc->base.mode, crtc_state); - crtc->base.mode.hdisplay = crtc_state->pipe_src_w; - crtc->base.mode.vdisplay = crtc_state->pipe_src_h; - intel_mode_from_pipe_config(&crtc_state->base.adjusted_mode, crtc_state); - WARN_ON(drm_atomic_set_mode_for_crtc(crtc->base.state, &crtc->base.mode)); + struct drm_display_mode mode; + + intel_mode_from_pipe_config(&crtc_state->base.adjusted_mode, + crtc_state); + + mode = crtc_state->base.adjusted_mode; + mode.hdisplay = crtc_state->pipe_src_w; + mode.vdisplay = crtc_state->pipe_src_h; + WARN_ON(drm_atomic_set_mode_for_crtc(&crtc_state->base, &mode)); /* * The initial mode needs to be set in order to keep @@ -16774,21 +17422,9 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) intel_crtc_compute_pixel_rate(crtc_state); - if (dev_priv->display.modeset_calc_cdclk) { - min_cdclk = intel_crtc_compute_min_cdclk(crtc_state); - if (WARN_ON(min_cdclk < 0)) - min_cdclk = 0; - } - - drm_calc_timestamping_constants(&crtc->base, - &crtc_state->base.adjusted_mode); - update_scanline_offset(crtc_state); + intel_crtc_update_active_timings(crtc_state); } - dev_priv->min_cdclk[crtc->pipe] = min_cdclk; - dev_priv->min_voltage_level[crtc->pipe] = - crtc_state->min_voltage_level; - for_each_intel_plane_on_crtc(&dev_priv->drm, crtc, plane) { const struct intel_plane_state *plane_state = to_intel_plane_state(plane->base.state); @@ -16800,8 +17436,34 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) if (plane_state->base.visible) crtc_state->data_rate[plane->id] = 4 * crtc_state->pixel_rate; + /* + * FIXME don't have the fb yet, so can't + * use plane->min_cdclk() :( + */ + if (plane_state->base.visible && plane->min_cdclk) { + if (crtc_state->double_wide || + INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) + crtc_state->min_cdclk[plane->id] = + DIV_ROUND_UP(crtc_state->pixel_rate, 2); + else + crtc_state->min_cdclk[plane->id] = + crtc_state->pixel_rate; + } + DRM_DEBUG_KMS("[PLANE:%d:%s] min_cdclk %d kHz\n", + plane->base.base.id, plane->base.name, + crtc_state->min_cdclk[plane->id]); + } + + if (crtc_state->base.active) { + min_cdclk = intel_crtc_compute_min_cdclk(crtc_state); + if (WARN_ON(min_cdclk < 0)) + min_cdclk = 0; } + dev_priv->min_cdclk[crtc->pipe] = min_cdclk; + dev_priv->min_voltage_level[crtc->pipe] = + crtc_state->min_voltage_level; + intel_bw_crtc_update(bw_state, crtc_state); intel_pipe_config_sanity_check(dev_priv, crtc_state); @@ -17042,13 +17704,13 @@ void intel_display_resume(struct drm_device *dev) drm_atomic_state_put(state); } -static void intel_hpd_poll_fini(struct drm_device *dev) +static void intel_hpd_poll_fini(struct drm_i915_private *i915) { struct intel_connector *connector; struct drm_connector_list_iter conn_iter; /* Kill all the work that may have been queued by hpd. */ - drm_connector_list_iter_begin(dev, &conn_iter); + drm_connector_list_iter_begin(&i915->drm, &conn_iter); for_each_intel_connector_iter(connector, &conn_iter) { if (connector->modeset_retry_work.func) cancel_work_sync(&connector->modeset_retry_work); @@ -17060,78 +17722,49 @@ static void intel_hpd_poll_fini(struct drm_device *dev) drm_connector_list_iter_end(&conn_iter); } -void intel_modeset_driver_remove(struct drm_device *dev) +void intel_modeset_driver_remove(struct drm_i915_private *i915) { - struct drm_i915_private *dev_priv = to_i915(dev); + flush_workqueue(i915->flip_wq); + flush_workqueue(i915->modeset_wq); - flush_workqueue(dev_priv->modeset_wq); - - flush_work(&dev_priv->atomic_helper.free_work); - WARN_ON(!llist_empty(&dev_priv->atomic_helper.free_list)); + flush_work(&i915->atomic_helper.free_work); + WARN_ON(!llist_empty(&i915->atomic_helper.free_list)); /* * Interrupts and polling as the first thing to avoid creating havoc. * Too much stuff here (turning of connectors, ...) would * experience fancy races otherwise. */ - intel_irq_uninstall(dev_priv); + intel_irq_uninstall(i915); /* * Due to the hpd irq storm handling the hotplug work can re-arm the * poll handlers. Hence disable polling after hpd handling is shut down. */ - intel_hpd_poll_fini(dev); + intel_hpd_poll_fini(i915); /* poll work can call into fbdev, hence clean that up afterwards */ - intel_fbdev_fini(dev_priv); + intel_fbdev_fini(i915); intel_unregister_dsm_handler(); - intel_fbc_global_disable(dev_priv); + intel_fbc_global_disable(i915); /* flush any delayed tasks or pending work */ flush_scheduled_work(); - intel_hdcp_component_fini(dev_priv); - - drm_mode_config_cleanup(dev); + intel_hdcp_component_fini(i915); - intel_overlay_cleanup(dev_priv); + drm_mode_config_cleanup(&i915->drm); - intel_gmbus_teardown(dev_priv); + intel_overlay_cleanup(i915); - destroy_workqueue(dev_priv->modeset_wq); + intel_gmbus_teardown(i915); - intel_fbc_cleanup_cfb(dev_priv); -} + destroy_workqueue(i915->flip_wq); + destroy_workqueue(i915->modeset_wq); -/* - * set vga decode state - true == enable VGA decode - */ -int intel_modeset_vga_set_state(struct drm_i915_private *dev_priv, bool state) -{ - unsigned reg = INTEL_GEN(dev_priv) >= 6 ? SNB_GMCH_CTRL : INTEL_GMCH_CTRL; - u16 gmch_ctrl; - - if (pci_read_config_word(dev_priv->bridge_dev, reg, &gmch_ctrl)) { - DRM_ERROR("failed to read control word\n"); - return -EIO; - } - - if (!!(gmch_ctrl & INTEL_GMCH_VGA_DISABLE) == !state) - return 0; - - if (state) - gmch_ctrl &= ~INTEL_GMCH_VGA_DISABLE; - else - gmch_ctrl |= INTEL_GMCH_VGA_DISABLE; - - if (pci_write_config_word(dev_priv->bridge_dev, reg, gmch_ctrl)) { - DRM_ERROR("failed to write control word\n"); - return -EIO; - } - - return 0; + intel_fbc_cleanup_cfb(i915); } #if IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR) @@ -17194,7 +17827,7 @@ intel_display_capture_error_state(struct drm_i915_private *dev_priv) BUILD_BUG_ON(ARRAY_SIZE(transcoders) != ARRAY_SIZE(error->transcoder)); - if (!HAS_DISPLAY(dev_priv)) + if (!HAS_DISPLAY(dev_priv) || !INTEL_DISPLAY_ENABLED(dev_priv)) return NULL; error = kzalloc(sizeof(*error), GFP_ATOMIC); @@ -17273,7 +17906,7 @@ intel_display_print_error_state(struct drm_i915_error_state_buf *m, if (!error) return; - err_printf(m, "Num Pipes: %d\n", INTEL_INFO(dev_priv)->num_pipes); + err_printf(m, "Num Pipes: %d\n", INTEL_NUM_PIPES(dev_priv)); if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) err_printf(m, "PWR_WELL_CTL2: %08x\n", error->power_well_driver); diff --git a/drivers/gpu/drm/i915/display/intel_display.h b/drivers/gpu/drm/i915/display/intel_display.h index e57e6969051d..f417e0948001 100644 --- a/drivers/gpu/drm/i915/display/intel_display.h +++ b/drivers/gpu/drm/i915/display/intel_display.h @@ -1,5 +1,5 @@ /* - * Copyright © 2006-2017 Intel Corporation + * Copyright © 2006-2019 Intel Corporation * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -32,8 +32,10 @@ enum link_m_n_set; struct dpll; struct drm_connector; struct drm_device; +struct drm_display_mode; struct drm_encoder; struct drm_file; +struct drm_format_info; struct drm_framebuffer; struct drm_i915_error_state_buf; struct drm_i915_gem_object; @@ -52,6 +54,7 @@ struct intel_plane; struct intel_plane_state; struct intel_remapped_info; struct intel_rotation_info; +struct intel_crtc_state; enum i915_gpio { GPIOA, @@ -91,6 +94,7 @@ enum pipe { #define pipe_name(p) ((p) + 'A') enum transcoder { + INVALID_TRANSCODER = -1, /* * The following transcoders have a 1:1 transcoder -> pipe mapping, * keep their values fixed: the code assumes that TRANSCODER_A=0, the @@ -182,6 +186,24 @@ enum plane_id { for ((__p) = PLANE_PRIMARY; (__p) < I915_MAX_PLANES; (__p)++) \ for_each_if((__crtc)->plane_ids_mask & BIT(__p)) +enum port { + PORT_NONE = -1, + + PORT_A = 0, + PORT_B, + PORT_C, + PORT_D, + PORT_E, + PORT_F, + PORT_G, + PORT_H, + PORT_I, + + I915_MAX_PORTS +}; + +#define port_name(p) ((p) + 'A') + /* * Ports identifier referenced from other drivers. * Expected to remain stable over time @@ -251,6 +273,7 @@ enum aux_ch { AUX_CH_D, AUX_CH_E, /* ICL+ */ AUX_CH_F, + AUX_CH_G, }; #define aux_ch_name(a) ((a) + 'A') @@ -289,10 +312,10 @@ enum phy_fia { }; #define for_each_pipe(__dev_priv, __p) \ - for ((__p) = 0; (__p) < INTEL_INFO(__dev_priv)->num_pipes; (__p)++) + for ((__p) = 0; (__p) < INTEL_NUM_PIPES(__dev_priv); (__p)++) #define for_each_pipe_masked(__dev_priv, __p, __mask) \ - for ((__p) = 0; (__p) < INTEL_INFO(__dev_priv)->num_pipes; (__p)++) \ + for ((__p) = 0; (__p) < INTEL_NUM_PIPES(__dev_priv); (__p)++) \ for_each_if((__mask) & BIT(__p)) #define for_each_cpu_transcoder_masked(__dev_priv, __t, __mask) \ @@ -330,7 +353,7 @@ enum phy_fia { &(dev)->mode_config.plane_list, \ base.head) \ for_each_if((plane_mask) & \ - drm_plane_mask(&intel_plane->base))) + drm_plane_mask(&intel_plane->base)) #define for_each_intel_plane_on_crtc(dev, intel_crtc, intel_plane) \ list_for_each_entry(intel_plane, \ @@ -411,16 +434,37 @@ enum phy_fia { (__i)++) \ for_each_if(crtc) +#define for_each_oldnew_intel_crtc_in_state_reverse(__state, crtc, old_crtc_state, new_crtc_state, __i) \ + for ((__i) = (__state)->base.dev->mode_config.num_crtc - 1; \ + (__i) >= 0 && \ + ((crtc) = to_intel_crtc((__state)->base.crtcs[__i].ptr), \ + (old_crtc_state) = to_intel_crtc_state((__state)->base.crtcs[__i].old_state), \ + (new_crtc_state) = to_intel_crtc_state((__state)->base.crtcs[__i].new_state), 1); \ + (__i)--) \ + for_each_if(crtc) + +#define intel_atomic_crtc_state_for_each_plane_state( \ + plane, plane_state, \ + crtc_state) \ + for_each_intel_plane_mask(((crtc_state)->base.state->dev), (plane), \ + ((crtc_state)->base.plane_mask)) \ + for_each_if ((plane_state = \ + to_intel_plane_state(__drm_atomic_get_current_plane_state((crtc_state)->base.state, &plane->base)))) + void intel_link_compute_m_n(u16 bpp, int nlanes, int pixel_clock, int link_clock, struct intel_link_m_n *m_n, - bool constant_n); + bool constant_n, bool fec_enable); bool is_ccs_modifier(u64 modifier); void lpt_disable_clkout_dp(struct drm_i915_private *dev_priv); u32 intel_plane_fb_max_stride(struct drm_i915_private *dev_priv, u32 pixel_format, u64 modifier); bool intel_plane_can_remap(const struct intel_plane_state *plane_state); +enum drm_mode_status +intel_mode_valid_max_plane_size(struct drm_i915_private *dev_priv, + const struct drm_display_mode *mode); enum phy intel_port_to_phy(struct drm_i915_private *i915, enum port port); +bool is_trans_port_sync_mode(const struct intel_crtc_state *state); void intel_plane_destroy(struct drm_plane *plane); void i830_enable_pipe(struct drm_i915_private *dev_priv, enum pipe pipe); @@ -464,7 +508,6 @@ void vlv_wait_port_ready(struct drm_i915_private *dev_priv, struct intel_digital_port *dport, unsigned int expected_mask); int intel_get_load_detect_pipe(struct drm_connector *connector, - const struct drm_display_mode *mode, struct intel_load_detect_pipe *old, struct drm_modeset_acquire_ctx *ctx); void intel_release_load_detect_pipe(struct drm_connector *connector, @@ -499,8 +542,6 @@ void intel_dp_get_m_n(struct intel_crtc *crtc, struct intel_crtc_state *pipe_config); void intel_dp_set_m_n(const struct intel_crtc_state *crtc_state, enum link_m_n_set m_n); -void intel_dp_ycbcr_420_enable(struct intel_dp *intel_dp, - const struct intel_crtc_state *crtc_state); int intel_dotclock_calculate(int link_freq, const struct intel_link_m_n *m_n); bool bxt_find_best_dpll(struct intel_crtc_state *crtc_state, struct dpll *best_clock); @@ -520,8 +561,6 @@ void intel_crtc_arm_fifo_underrun(struct intel_crtc *crtc, u16 skl_scaler_calc_phase(int sub, int scale, bool chroma_center); int skl_update_scaler_crtc(struct intel_crtc_state *crtc_state); -int skl_max_scale(const struct intel_crtc_state *crtc_state, - u32 pixel_format); u32 glk_plane_color_ctl(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state); u32 glk_plane_color_ctl_crtc(const struct intel_crtc_state *crtc_state); @@ -544,13 +583,10 @@ void intel_display_print_error_state(struct drm_i915_error_state_buf *e, struct intel_display_error_state *error); /* modesetting */ -void intel_modeset_init_hw(struct drm_device *dev); -int intel_modeset_init(struct drm_device *dev); -void intel_modeset_driver_remove(struct drm_device *dev); -int intel_modeset_vga_set_state(struct drm_i915_private *dev_priv, bool state); +void intel_modeset_init_hw(struct drm_i915_private *i915); +int intel_modeset_init(struct drm_i915_private *i915); +void intel_modeset_driver_remove(struct drm_i915_private *i915); void intel_display_resume(struct drm_device *dev); -void i915_redisable_vga(struct drm_i915_private *dev_priv); -void i915_redisable_vga_power_on(struct drm_i915_private *dev_priv); void intel_init_pch_refclk(struct drm_i915_private *dev_priv); /* modesetting asserts */ diff --git a/drivers/gpu/drm/i915/display/intel_display_power.c b/drivers/gpu/drm/i915/display/intel_display_power.c index 12099760d99e..ce1b64f4dd44 100644 --- a/drivers/gpu/drm/i915/display/intel_display_power.c +++ b/drivers/gpu/drm/i915/display/intel_display_power.c @@ -3,8 +3,6 @@ * Copyright © 2019 Intel Corporation */ -#include <linux/vgaarb.h> - #include "display/intel_crt.h" #include "display/intel_dp.h" @@ -19,16 +17,14 @@ #include "intel_hotplug.h" #include "intel_sideband.h" #include "intel_tc.h" +#include "intel_vga.h" bool intel_display_power_well_is_enabled(struct drm_i915_private *dev_priv, enum i915_power_well_id power_well_id); const char * -intel_display_power_domain_str(struct drm_i915_private *i915, - enum intel_display_power_domain domain) +intel_display_power_domain_str(enum intel_display_power_domain domain) { - bool ddi_tc_ports = IS_GEN(i915, 12); - switch (domain) { case POWER_DOMAIN_DISPLAY_CORE: return "DISPLAY_CORE"; @@ -71,23 +67,17 @@ intel_display_power_domain_str(struct drm_i915_private *i915, case POWER_DOMAIN_PORT_DDI_C_LANES: return "PORT_DDI_C_LANES"; case POWER_DOMAIN_PORT_DDI_D_LANES: - BUILD_BUG_ON(POWER_DOMAIN_PORT_DDI_D_LANES != - POWER_DOMAIN_PORT_DDI_TC1_LANES); - return ddi_tc_ports ? "PORT_DDI_TC1_LANES" : "PORT_DDI_D_LANES"; + return "PORT_DDI_D_LANES"; case POWER_DOMAIN_PORT_DDI_E_LANES: - BUILD_BUG_ON(POWER_DOMAIN_PORT_DDI_E_LANES != - POWER_DOMAIN_PORT_DDI_TC2_LANES); - return ddi_tc_ports ? "PORT_DDI_TC2_LANES" : "PORT_DDI_E_LANES"; + return "PORT_DDI_E_LANES"; case POWER_DOMAIN_PORT_DDI_F_LANES: - BUILD_BUG_ON(POWER_DOMAIN_PORT_DDI_F_LANES != - POWER_DOMAIN_PORT_DDI_TC3_LANES); - return ddi_tc_ports ? "PORT_DDI_TC3_LANES" : "PORT_DDI_F_LANES"; - case POWER_DOMAIN_PORT_DDI_TC4_LANES: - return "PORT_DDI_TC4_LANES"; - case POWER_DOMAIN_PORT_DDI_TC5_LANES: - return "PORT_DDI_TC5_LANES"; - case POWER_DOMAIN_PORT_DDI_TC6_LANES: - return "PORT_DDI_TC6_LANES"; + return "PORT_DDI_F_LANES"; + case POWER_DOMAIN_PORT_DDI_G_LANES: + return "PORT_DDI_G_LANES"; + case POWER_DOMAIN_PORT_DDI_H_LANES: + return "PORT_DDI_H_LANES"; + case POWER_DOMAIN_PORT_DDI_I_LANES: + return "PORT_DDI_I_LANES"; case POWER_DOMAIN_PORT_DDI_A_IO: return "PORT_DDI_A_IO"; case POWER_DOMAIN_PORT_DDI_B_IO: @@ -95,23 +85,17 @@ intel_display_power_domain_str(struct drm_i915_private *i915, case POWER_DOMAIN_PORT_DDI_C_IO: return "PORT_DDI_C_IO"; case POWER_DOMAIN_PORT_DDI_D_IO: - BUILD_BUG_ON(POWER_DOMAIN_PORT_DDI_D_IO != - POWER_DOMAIN_PORT_DDI_TC1_IO); - return ddi_tc_ports ? "PORT_DDI_TC1_IO" : "PORT_DDI_D_IO"; + return "PORT_DDI_D_IO"; case POWER_DOMAIN_PORT_DDI_E_IO: - BUILD_BUG_ON(POWER_DOMAIN_PORT_DDI_E_IO != - POWER_DOMAIN_PORT_DDI_TC2_IO); - return ddi_tc_ports ? "PORT_DDI_TC2_IO" : "PORT_DDI_E_IO"; + return "PORT_DDI_E_IO"; case POWER_DOMAIN_PORT_DDI_F_IO: - BUILD_BUG_ON(POWER_DOMAIN_PORT_DDI_F_IO != - POWER_DOMAIN_PORT_DDI_TC3_IO); - return ddi_tc_ports ? "PORT_DDI_TC3_IO" : "PORT_DDI_F_IO"; - case POWER_DOMAIN_PORT_DDI_TC4_IO: - return "PORT_DDI_TC4_IO"; - case POWER_DOMAIN_PORT_DDI_TC5_IO: - return "PORT_DDI_TC5_IO"; - case POWER_DOMAIN_PORT_DDI_TC6_IO: - return "PORT_DDI_TC6_IO"; + return "PORT_DDI_F_IO"; + case POWER_DOMAIN_PORT_DDI_G_IO: + return "PORT_DDI_G_IO"; + case POWER_DOMAIN_PORT_DDI_H_IO: + return "PORT_DDI_H_IO"; + case POWER_DOMAIN_PORT_DDI_I_IO: + return "PORT_DDI_I_IO"; case POWER_DOMAIN_PORT_DSI: return "PORT_DSI"; case POWER_DOMAIN_PORT_CRT: @@ -129,34 +113,33 @@ intel_display_power_domain_str(struct drm_i915_private *i915, case POWER_DOMAIN_AUX_C: return "AUX_C"; case POWER_DOMAIN_AUX_D: - BUILD_BUG_ON(POWER_DOMAIN_AUX_D != POWER_DOMAIN_AUX_TC1); - return ddi_tc_ports ? "AUX_TC1" : "AUX_D"; + return "AUX_D"; case POWER_DOMAIN_AUX_E: - BUILD_BUG_ON(POWER_DOMAIN_AUX_E != POWER_DOMAIN_AUX_TC2); - return ddi_tc_ports ? "AUX_TC2" : "AUX_E"; + return "AUX_E"; case POWER_DOMAIN_AUX_F: - BUILD_BUG_ON(POWER_DOMAIN_AUX_F != POWER_DOMAIN_AUX_TC3); - return ddi_tc_ports ? "AUX_TC3" : "AUX_F"; - case POWER_DOMAIN_AUX_TC4: - return "AUX_TC4"; - case POWER_DOMAIN_AUX_TC5: - return "AUX_TC5"; - case POWER_DOMAIN_AUX_TC6: - return "AUX_TC6"; + return "AUX_F"; + case POWER_DOMAIN_AUX_G: + return "AUX_G"; + case POWER_DOMAIN_AUX_H: + return "AUX_H"; + case POWER_DOMAIN_AUX_I: + return "AUX_I"; case POWER_DOMAIN_AUX_IO_A: return "AUX_IO_A"; - case POWER_DOMAIN_AUX_TBT1: - return "AUX_TBT1"; - case POWER_DOMAIN_AUX_TBT2: - return "AUX_TBT2"; - case POWER_DOMAIN_AUX_TBT3: - return "AUX_TBT3"; - case POWER_DOMAIN_AUX_TBT4: - return "AUX_TBT4"; - case POWER_DOMAIN_AUX_TBT5: - return "AUX_TBT5"; - case POWER_DOMAIN_AUX_TBT6: - return "AUX_TBT6"; + case POWER_DOMAIN_AUX_C_TBT: + return "AUX_C_TBT"; + case POWER_DOMAIN_AUX_D_TBT: + return "AUX_D_TBT"; + case POWER_DOMAIN_AUX_E_TBT: + return "AUX_E_TBT"; + case POWER_DOMAIN_AUX_F_TBT: + return "AUX_F_TBT"; + case POWER_DOMAIN_AUX_G_TBT: + return "AUX_G_TBT"; + case POWER_DOMAIN_AUX_H_TBT: + return "AUX_H_TBT"; + case POWER_DOMAIN_AUX_I_TBT: + return "AUX_I_TBT"; case POWER_DOMAIN_GMBUS: return "GMBUS"; case POWER_DOMAIN_INIT: @@ -283,23 +266,8 @@ bool intel_display_power_is_enabled(struct drm_i915_private *dev_priv, static void hsw_power_well_post_enable(struct drm_i915_private *dev_priv, u8 irq_pipe_mask, bool has_vga) { - struct pci_dev *pdev = dev_priv->drm.pdev; - - /* - * After we re-enable the power well, if we touch VGA register 0x3d5 - * we'll get unclaimed register interrupts. This stops after we write - * anything to the VGA MSR register. The vgacon module uses this - * register all the time, so if we unbind our driver and, as a - * consequence, bind vgacon, we'll get stuck in an infinite loop at - * console_unlock(). So make here we touch the VGA MSR register, making - * sure vgacon can keep working normally without triggering interrupts - * and error messages. - */ - if (has_vga) { - vga_get_uninterruptible(pdev, VGA_RSRC_LEGACY_IO); - outb(inb(VGA_MSR_READ), VGA_MSR_WRITE); - vga_put(pdev, VGA_RSRC_LEGACY_IO); - } + if (has_vga) + intel_vga_reset_io_mem(dev_priv); if (irq_pipe_mask) gen8_irq_power_well_post_enable(dev_priv, irq_pipe_mask); @@ -578,6 +546,8 @@ static void icl_tc_port_assert_ref_held(struct drm_i915_private *dev_priv, #endif +#define TGL_AUX_PW_TO_TC_PORT(pw_idx) ((pw_idx) - TGL_PW_CTL_IDX_AUX_TC1) + static void icl_tc_phy_aux_power_well_enable(struct drm_i915_private *dev_priv, struct i915_power_well *power_well) @@ -594,6 +564,17 @@ icl_tc_phy_aux_power_well_enable(struct drm_i915_private *dev_priv, I915_WRITE(DP_AUX_CH_CTL(aux_ch), val); hsw_power_well_enable(dev_priv, power_well); + + if (INTEL_GEN(dev_priv) >= 12 && !power_well->desc->hsw.is_tc_tbt) { + enum tc_port tc_port; + + tc_port = TGL_AUX_PW_TO_TC_PORT(power_well->desc->hsw.idx); + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, 0x2)); + + if (intel_de_wait_for_set(dev_priv, DKL_CMN_UC_DW_27(tc_port), + DKL_CMN_UC_DW27_UC_HEALTH, 1)) + DRM_WARN("Timeout waiting TC uC health\n"); + } } static void @@ -714,7 +695,11 @@ static u32 gen9_dc_mask(struct drm_i915_private *dev_priv) u32 mask; mask = DC_STATE_EN_UPTO_DC5; - if (INTEL_GEN(dev_priv) >= 11) + + if (INTEL_GEN(dev_priv) >= 12) + mask |= DC_STATE_EN_DC3CO | DC_STATE_EN_UPTO_DC6 + | DC_STATE_EN_DC9; + else if (IS_GEN(dev_priv, 11)) mask |= DC_STATE_EN_UPTO_DC6 | DC_STATE_EN_DC9; else if (IS_GEN9_LP(dev_priv)) mask |= DC_STATE_EN_DC9; @@ -784,6 +769,52 @@ static void gen9_set_dc_state(struct drm_i915_private *dev_priv, u32 state) dev_priv->csr.dc_state = val & mask; } +static u32 +sanitize_target_dc_state(struct drm_i915_private *dev_priv, + u32 target_dc_state) +{ + u32 states[] = { + DC_STATE_EN_UPTO_DC6, + DC_STATE_EN_UPTO_DC5, + DC_STATE_EN_DC3CO, + DC_STATE_DISABLE, + }; + int i; + + for (i = 0; i < ARRAY_SIZE(states) - 1; i++) { + if (target_dc_state != states[i]) + continue; + + if (dev_priv->csr.allowed_dc_mask & target_dc_state) + break; + + target_dc_state = states[i + 1]; + } + + return target_dc_state; +} + +static void tgl_enable_dc3co(struct drm_i915_private *dev_priv) +{ + DRM_DEBUG_KMS("Enabling DC3CO\n"); + gen9_set_dc_state(dev_priv, DC_STATE_EN_DC3CO); +} + +static void tgl_disable_dc3co(struct drm_i915_private *dev_priv) +{ + u32 val; + + DRM_DEBUG_KMS("Disabling DC3CO\n"); + val = I915_READ(DC_STATE_EN); + val &= ~DC_STATE_DC3CO_STATUS; + I915_WRITE(DC_STATE_EN, val); + gen9_set_dc_state(dev_priv, DC_STATE_DISABLE); + /* + * Delay of 200us DC3CO Exit time B.Spec 49196 + */ + usleep_range(200, 210); +} + static void bxt_enable_dc9(struct drm_i915_private *dev_priv) { assert_can_enable_dc9(dev_priv); @@ -839,6 +870,51 @@ lookup_power_well(struct drm_i915_private *dev_priv, return &dev_priv->power_domains.power_wells[0]; } +/** + * intel_display_power_set_target_dc_state - Set target dc state. + * @dev_priv: i915 device + * @state: state which needs to be set as target_dc_state. + * + * This function set the "DC off" power well target_dc_state, + * based upon this target_dc_stste, "DC off" power well will + * enable desired DC state. + */ +void intel_display_power_set_target_dc_state(struct drm_i915_private *dev_priv, + u32 state) +{ + struct i915_power_well *power_well; + bool dc_off_enabled; + struct i915_power_domains *power_domains = &dev_priv->power_domains; + + mutex_lock(&power_domains->lock); + power_well = lookup_power_well(dev_priv, SKL_DISP_DC_OFF); + + if (WARN_ON(!power_well)) + goto unlock; + + state = sanitize_target_dc_state(dev_priv, state); + + if (state == dev_priv->csr.target_dc_state) + goto unlock; + + dc_off_enabled = power_well->desc->ops->is_enabled(dev_priv, + power_well); + /* + * If DC off power well is disabled, need to enable and disable the + * DC off power well to effect target DC state. + */ + if (!dc_off_enabled) + power_well->desc->ops->enable(dev_priv, power_well); + + dev_priv->csr.target_dc_state = state; + + if (!dc_off_enabled) + power_well->desc->ops->disable(dev_priv, power_well); + +unlock: + mutex_unlock(&power_domains->lock); +} + static void assert_can_enable_dc5(struct drm_i915_private *dev_priv) { bool pg2_enabled = intel_display_power_well_is_enabled(dev_priv, @@ -951,7 +1027,8 @@ static void bxt_verify_ddi_phy_power_wells(struct drm_i915_private *dev_priv) static bool gen9_dc_off_power_well_enabled(struct drm_i915_private *dev_priv, struct i915_power_well *power_well) { - return (I915_READ(DC_STATE_EN) & DC_STATE_EN_UPTO_DC5_DC6_MASK) == 0; + return ((I915_READ(DC_STATE_EN) & DC_STATE_EN_DC3CO) == 0 && + (I915_READ(DC_STATE_EN) & DC_STATE_EN_UPTO_DC5_DC6_MASK) == 0); } static void gen9_assert_dbuf_enabled(struct drm_i915_private *dev_priv) @@ -967,6 +1044,11 @@ static void gen9_disable_dc_states(struct drm_i915_private *dev_priv) { struct intel_cdclk_state cdclk_state = {}; + if (dev_priv->csr.target_dc_state == DC_STATE_EN_DC3CO) { + tgl_disable_dc3co(dev_priv); + return; + } + gen9_set_dc_state(dev_priv, DC_STATE_DISABLE); dev_priv->display.get_cdclk(dev_priv, &cdclk_state); @@ -999,10 +1081,17 @@ static void gen9_dc_off_power_well_disable(struct drm_i915_private *dev_priv, if (!dev_priv->csr.dmc_payload) return; - if (dev_priv->csr.allowed_dc_mask & DC_STATE_EN_UPTO_DC6) + switch (dev_priv->csr.target_dc_state) { + case DC_STATE_EN_DC3CO: + tgl_enable_dc3co(dev_priv); + break; + case DC_STATE_EN_UPTO_DC6: skl_enable_dc6(dev_priv); - else if (dev_priv->csr.allowed_dc_mask & DC_STATE_EN_UPTO_DC5) + break; + case DC_STATE_EN_UPTO_DC5: gen9_enable_dc5(dev_priv); + break; + } } static void i9xx_power_well_sync_hw_noop(struct drm_i915_private *dev_priv, @@ -1208,7 +1297,7 @@ static void vlv_display_power_well_init(struct drm_i915_private *dev_priv) intel_crt_reset(&encoder->base); } - i915_redisable_vga_power_on(dev_priv); + intel_vga_redisable_power_on(dev_priv); intel_pps_unlock_regs_wa(dev_priv); } @@ -1718,15 +1807,12 @@ __async_put_domains_state_ok(struct i915_power_domains *power_domains) static void print_power_domains(struct i915_power_domains *power_domains, const char *prefix, u64 mask) { - struct drm_i915_private *i915 = - container_of(power_domains, struct drm_i915_private, - power_domains); enum intel_display_power_domain domain; DRM_DEBUG_DRIVER("%s (%lu):\n", prefix, hweight64(mask)); for_each_power_domain(domain, mask) DRM_DEBUG_DRIVER("%s use_count %d\n", - intel_display_power_domain_str(i915, domain), + intel_display_power_domain_str(domain), power_domains->domain_use_count[domain]); } @@ -1896,7 +1982,7 @@ __intel_display_power_put_domain(struct drm_i915_private *dev_priv, { struct i915_power_domains *power_domains; struct i915_power_well *power_well; - const char *name = intel_display_power_domain_str(dev_priv, domain); + const char *name = intel_display_power_domain_str(domain); power_domains = &dev_priv->power_domains; @@ -2487,10 +2573,10 @@ void intel_display_power_put(struct drm_i915_private *dev_priv, BIT_ULL(POWER_DOMAIN_AUX_D) | \ BIT_ULL(POWER_DOMAIN_AUX_E) | \ BIT_ULL(POWER_DOMAIN_AUX_F) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT1) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT2) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT3) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT4) | \ + BIT_ULL(POWER_DOMAIN_AUX_C_TBT) | \ + BIT_ULL(POWER_DOMAIN_AUX_D_TBT) | \ + BIT_ULL(POWER_DOMAIN_AUX_E_TBT) | \ + BIT_ULL(POWER_DOMAIN_AUX_F_TBT) | \ BIT_ULL(POWER_DOMAIN_VGA) | \ BIT_ULL(POWER_DOMAIN_AUDIO) | \ BIT_ULL(POWER_DOMAIN_INIT)) @@ -2530,22 +2616,22 @@ void intel_display_power_put(struct drm_i915_private *dev_priv, BIT_ULL(POWER_DOMAIN_AUX_A)) #define ICL_AUX_B_IO_POWER_DOMAINS ( \ BIT_ULL(POWER_DOMAIN_AUX_B)) -#define ICL_AUX_C_IO_POWER_DOMAINS ( \ +#define ICL_AUX_C_TC1_IO_POWER_DOMAINS ( \ BIT_ULL(POWER_DOMAIN_AUX_C)) -#define ICL_AUX_D_IO_POWER_DOMAINS ( \ +#define ICL_AUX_D_TC2_IO_POWER_DOMAINS ( \ BIT_ULL(POWER_DOMAIN_AUX_D)) -#define ICL_AUX_E_IO_POWER_DOMAINS ( \ +#define ICL_AUX_E_TC3_IO_POWER_DOMAINS ( \ BIT_ULL(POWER_DOMAIN_AUX_E)) -#define ICL_AUX_F_IO_POWER_DOMAINS ( \ +#define ICL_AUX_F_TC4_IO_POWER_DOMAINS ( \ BIT_ULL(POWER_DOMAIN_AUX_F)) -#define ICL_AUX_TBT1_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TBT1)) -#define ICL_AUX_TBT2_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TBT2)) -#define ICL_AUX_TBT3_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TBT3)) -#define ICL_AUX_TBT4_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TBT4)) +#define ICL_AUX_C_TBT1_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_C_TBT)) +#define ICL_AUX_D_TBT2_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_D_TBT)) +#define ICL_AUX_E_TBT3_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_E_TBT)) +#define ICL_AUX_F_TBT4_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_F_TBT)) #define TGL_PW_5_POWER_DOMAINS ( \ BIT_ULL(POWER_DOMAIN_PIPE_D) | \ @@ -2565,24 +2651,24 @@ void intel_display_power_put(struct drm_i915_private *dev_priv, BIT_ULL(POWER_DOMAIN_PIPE_B) | \ BIT_ULL(POWER_DOMAIN_TRANSCODER_B) | \ BIT_ULL(POWER_DOMAIN_PIPE_B_PANEL_FITTER) | \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC1_LANES) | \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC2_LANES) | \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC3_LANES) | \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC4_LANES) | \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC5_LANES) | \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC6_LANES) | \ - BIT_ULL(POWER_DOMAIN_AUX_TC1) | \ - BIT_ULL(POWER_DOMAIN_AUX_TC2) | \ - BIT_ULL(POWER_DOMAIN_AUX_TC3) | \ - BIT_ULL(POWER_DOMAIN_AUX_TC4) | \ - BIT_ULL(POWER_DOMAIN_AUX_TC5) | \ - BIT_ULL(POWER_DOMAIN_AUX_TC6) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT1) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT2) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT3) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT4) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT5) | \ - BIT_ULL(POWER_DOMAIN_AUX_TBT6) | \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_D_LANES) | \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_E_LANES) | \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_F_LANES) | \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_G_LANES) | \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_H_LANES) | \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_I_LANES) | \ + BIT_ULL(POWER_DOMAIN_AUX_D) | \ + BIT_ULL(POWER_DOMAIN_AUX_E) | \ + BIT_ULL(POWER_DOMAIN_AUX_F) | \ + BIT_ULL(POWER_DOMAIN_AUX_G) | \ + BIT_ULL(POWER_DOMAIN_AUX_H) | \ + BIT_ULL(POWER_DOMAIN_AUX_I) | \ + BIT_ULL(POWER_DOMAIN_AUX_D_TBT) | \ + BIT_ULL(POWER_DOMAIN_AUX_E_TBT) | \ + BIT_ULL(POWER_DOMAIN_AUX_F_TBT) | \ + BIT_ULL(POWER_DOMAIN_AUX_G_TBT) | \ + BIT_ULL(POWER_DOMAIN_AUX_H_TBT) | \ + BIT_ULL(POWER_DOMAIN_AUX_I_TBT) | \ BIT_ULL(POWER_DOMAIN_VGA) | \ BIT_ULL(POWER_DOMAIN_AUDIO) | \ BIT_ULL(POWER_DOMAIN_INIT)) @@ -2596,37 +2682,54 @@ void intel_display_power_put(struct drm_i915_private *dev_priv, TGL_PW_2_POWER_DOMAINS | \ BIT_ULL(POWER_DOMAIN_MODESET) | \ BIT_ULL(POWER_DOMAIN_AUX_A) | \ + BIT_ULL(POWER_DOMAIN_AUX_B) | \ + BIT_ULL(POWER_DOMAIN_AUX_C) | \ BIT_ULL(POWER_DOMAIN_INIT)) -#define TGL_DDI_IO_TC1_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC1_IO)) -#define TGL_DDI_IO_TC2_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC2_IO)) -#define TGL_DDI_IO_TC3_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC3_IO)) -#define TGL_DDI_IO_TC4_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC4_IO)) -#define TGL_DDI_IO_TC5_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC5_IO)) -#define TGL_DDI_IO_TC6_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_PORT_DDI_TC6_IO)) - -#define TGL_AUX_TC1_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TC1)) -#define TGL_AUX_TC2_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TC2)) -#define TGL_AUX_TC3_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TC3)) -#define TGL_AUX_TC4_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TC4)) -#define TGL_AUX_TC5_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TC5)) -#define TGL_AUX_TC6_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TC6)) -#define TGL_AUX_TBT5_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TBT5)) -#define TGL_AUX_TBT6_IO_POWER_DOMAINS ( \ - BIT_ULL(POWER_DOMAIN_AUX_TBT6)) +#define TGL_DDI_IO_D_TC1_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_D_IO)) +#define TGL_DDI_IO_E_TC2_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_E_IO)) +#define TGL_DDI_IO_F_TC3_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_F_IO)) +#define TGL_DDI_IO_G_TC4_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_G_IO)) +#define TGL_DDI_IO_H_TC5_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_H_IO)) +#define TGL_DDI_IO_I_TC6_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_PORT_DDI_I_IO)) + +#define TGL_AUX_A_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_IO_A) | \ + BIT_ULL(POWER_DOMAIN_AUX_A)) +#define TGL_AUX_B_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_B)) +#define TGL_AUX_C_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_C)) +#define TGL_AUX_D_TC1_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_D)) +#define TGL_AUX_E_TC2_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_E)) +#define TGL_AUX_F_TC3_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_F)) +#define TGL_AUX_G_TC4_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_G)) +#define TGL_AUX_H_TC5_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_H)) +#define TGL_AUX_I_TC6_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_I)) +#define TGL_AUX_D_TBT1_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_D_TBT)) +#define TGL_AUX_E_TBT2_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_E_TBT)) +#define TGL_AUX_F_TBT3_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_F_TBT)) +#define TGL_AUX_G_TBT4_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_G_TBT)) +#define TGL_AUX_H_TBT5_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_H_TBT)) +#define TGL_AUX_I_TBT6_IO_POWER_DOMAINS ( \ + BIT_ULL(POWER_DOMAIN_AUX_I_TBT)) static const struct i915_power_well_ops i9xx_always_on_power_well_ops = { .sync_hw = i9xx_power_well_sync_hw_noop, @@ -2938,7 +3041,7 @@ static const struct i915_power_well_desc skl_power_wells[] = { .name = "DC off", .domains = SKL_DISPLAY_DC_OFF_POWER_DOMAINS, .ops = &gen9_dc_off_power_well_ops, - .id = DISP_PW_ID_NONE, + .id = SKL_DISP_DC_OFF, }, { .name = "power well 2", @@ -3020,7 +3123,7 @@ static const struct i915_power_well_desc bxt_power_wells[] = { .name = "DC off", .domains = BXT_DISPLAY_DC_OFF_POWER_DOMAINS, .ops = &gen9_dc_off_power_well_ops, - .id = DISP_PW_ID_NONE, + .id = SKL_DISP_DC_OFF, }, { .name = "power well 2", @@ -3080,7 +3183,7 @@ static const struct i915_power_well_desc glk_power_wells[] = { .name = "DC off", .domains = GLK_DISPLAY_DC_OFF_POWER_DOMAINS, .ops = &gen9_dc_off_power_well_ops, - .id = DISP_PW_ID_NONE, + .id = SKL_DISP_DC_OFF, }, { .name = "power well 2", @@ -3249,7 +3352,7 @@ static const struct i915_power_well_desc cnl_power_wells[] = { .name = "DC off", .domains = CNL_DISPLAY_DC_OFF_POWER_DOMAINS, .ops = &gen9_dc_off_power_well_ops, - .id = DISP_PW_ID_NONE, + .id = SKL_DISP_DC_OFF, }, { .name = "power well 2", @@ -3377,7 +3480,7 @@ static const struct i915_power_well_desc icl_power_wells[] = { .name = "DC off", .domains = ICL_DISPLAY_DC_OFF_POWER_DOMAINS, .ops = &gen9_dc_off_power_well_ops, - .id = DISP_PW_ID_NONE, + .id = SKL_DISP_DC_OFF, }, { .name = "power well 2", @@ -3484,8 +3587,8 @@ static const struct i915_power_well_desc icl_power_wells[] = { }, }, { - .name = "AUX C", - .domains = ICL_AUX_C_IO_POWER_DOMAINS, + .name = "AUX C TC1", + .domains = ICL_AUX_C_TC1_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3495,8 +3598,8 @@ static const struct i915_power_well_desc icl_power_wells[] = { }, }, { - .name = "AUX D", - .domains = ICL_AUX_D_IO_POWER_DOMAINS, + .name = "AUX D TC2", + .domains = ICL_AUX_D_TC2_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3506,8 +3609,8 @@ static const struct i915_power_well_desc icl_power_wells[] = { }, }, { - .name = "AUX E", - .domains = ICL_AUX_E_IO_POWER_DOMAINS, + .name = "AUX E TC3", + .domains = ICL_AUX_E_TC3_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3517,8 +3620,8 @@ static const struct i915_power_well_desc icl_power_wells[] = { }, }, { - .name = "AUX F", - .domains = ICL_AUX_F_IO_POWER_DOMAINS, + .name = "AUX F TC4", + .domains = ICL_AUX_F_TC4_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3528,8 +3631,8 @@ static const struct i915_power_well_desc icl_power_wells[] = { }, }, { - .name = "AUX TBT1", - .domains = ICL_AUX_TBT1_IO_POWER_DOMAINS, + .name = "AUX C TBT1", + .domains = ICL_AUX_C_TBT1_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3539,8 +3642,8 @@ static const struct i915_power_well_desc icl_power_wells[] = { }, }, { - .name = "AUX TBT2", - .domains = ICL_AUX_TBT2_IO_POWER_DOMAINS, + .name = "AUX D TBT2", + .domains = ICL_AUX_D_TBT2_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3550,8 +3653,8 @@ static const struct i915_power_well_desc icl_power_wells[] = { }, }, { - .name = "AUX TBT3", - .domains = ICL_AUX_TBT3_IO_POWER_DOMAINS, + .name = "AUX E TBT3", + .domains = ICL_AUX_E_TBT3_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3561,8 +3664,8 @@ static const struct i915_power_well_desc icl_power_wells[] = { }, }, { - .name = "AUX TBT4", - .domains = ICL_AUX_TBT4_IO_POWER_DOMAINS, + .name = "AUX F TBT4", + .domains = ICL_AUX_F_TBT4_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3610,7 +3713,7 @@ static const struct i915_power_well_desc tgl_power_wells[] = { .name = "DC off", .domains = TGL_DISPLAY_DC_OFF_POWER_DOMAINS, .ops = &gen9_dc_off_power_well_ops, - .id = DISP_PW_ID_NONE, + .id = SKL_DISP_DC_OFF, }, { .name = "power well 2", @@ -3667,8 +3770,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { } }, { - .name = "DDI TC1 IO", - .domains = TGL_DDI_IO_TC1_POWER_DOMAINS, + .name = "DDI D TC1 IO", + .domains = TGL_DDI_IO_D_TC1_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3677,8 +3780,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "DDI TC2 IO", - .domains = TGL_DDI_IO_TC2_POWER_DOMAINS, + .name = "DDI E TC2 IO", + .domains = TGL_DDI_IO_E_TC2_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3687,8 +3790,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "DDI TC3 IO", - .domains = TGL_DDI_IO_TC3_POWER_DOMAINS, + .name = "DDI F TC3 IO", + .domains = TGL_DDI_IO_F_TC3_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3697,8 +3800,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "DDI TC4 IO", - .domains = TGL_DDI_IO_TC4_POWER_DOMAINS, + .name = "DDI G TC4 IO", + .domains = TGL_DDI_IO_G_TC4_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3707,8 +3810,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "DDI TC5 IO", - .domains = TGL_DDI_IO_TC5_POWER_DOMAINS, + .name = "DDI H TC5 IO", + .domains = TGL_DDI_IO_H_TC5_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3717,8 +3820,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "DDI TC6 IO", - .domains = TGL_DDI_IO_TC6_POWER_DOMAINS, + .name = "DDI I TC6 IO", + .domains = TGL_DDI_IO_I_TC6_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3728,7 +3831,7 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, { .name = "AUX A", - .domains = ICL_AUX_A_IO_POWER_DOMAINS, + .domains = TGL_AUX_A_IO_POWER_DOMAINS, .ops = &icl_combo_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3738,7 +3841,7 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, { .name = "AUX B", - .domains = ICL_AUX_B_IO_POWER_DOMAINS, + .domains = TGL_AUX_B_IO_POWER_DOMAINS, .ops = &icl_combo_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3748,7 +3851,7 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, { .name = "AUX C", - .domains = ICL_AUX_C_IO_POWER_DOMAINS, + .domains = TGL_AUX_C_IO_POWER_DOMAINS, .ops = &icl_combo_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3757,8 +3860,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TC1", - .domains = TGL_AUX_TC1_IO_POWER_DOMAINS, + .name = "AUX D TC1", + .domains = TGL_AUX_D_TC1_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3768,8 +3871,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TC2", - .domains = TGL_AUX_TC2_IO_POWER_DOMAINS, + .name = "AUX E TC2", + .domains = TGL_AUX_E_TC2_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3779,8 +3882,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TC3", - .domains = TGL_AUX_TC3_IO_POWER_DOMAINS, + .name = "AUX F TC3", + .domains = TGL_AUX_F_TC3_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3790,8 +3893,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TC4", - .domains = TGL_AUX_TC4_IO_POWER_DOMAINS, + .name = "AUX G TC4", + .domains = TGL_AUX_G_TC4_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3801,8 +3904,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TC5", - .domains = TGL_AUX_TC5_IO_POWER_DOMAINS, + .name = "AUX H TC5", + .domains = TGL_AUX_H_TC5_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3812,8 +3915,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TC6", - .domains = TGL_AUX_TC6_IO_POWER_DOMAINS, + .name = "AUX I TC6", + .domains = TGL_AUX_I_TC6_IO_POWER_DOMAINS, .ops = &icl_tc_phy_aux_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3823,8 +3926,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TBT1", - .domains = ICL_AUX_TBT1_IO_POWER_DOMAINS, + .name = "AUX D TBT1", + .domains = TGL_AUX_D_TBT1_IO_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3834,8 +3937,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TBT2", - .domains = ICL_AUX_TBT2_IO_POWER_DOMAINS, + .name = "AUX E TBT2", + .domains = TGL_AUX_E_TBT2_IO_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3845,8 +3948,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TBT3", - .domains = ICL_AUX_TBT3_IO_POWER_DOMAINS, + .name = "AUX F TBT3", + .domains = TGL_AUX_F_TBT3_IO_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3856,8 +3959,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TBT4", - .domains = ICL_AUX_TBT4_IO_POWER_DOMAINS, + .name = "AUX G TBT4", + .domains = TGL_AUX_G_TBT4_IO_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3867,8 +3970,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TBT5", - .domains = TGL_AUX_TBT5_IO_POWER_DOMAINS, + .name = "AUX H TBT5", + .domains = TGL_AUX_H_TBT5_IO_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3878,8 +3981,8 @@ static const struct i915_power_well_desc tgl_power_wells[] = { }, }, { - .name = "AUX TBT6", - .domains = TGL_AUX_TBT6_IO_POWER_DOMAINS, + .name = "AUX I TBT6", + .domains = TGL_AUX_I_TBT6_IO_POWER_DOMAINS, .ops = &hsw_power_well_ops, .id = DISP_PW_ID_NONE, { @@ -3931,14 +4034,17 @@ static u32 get_allowed_dc_mask(const struct drm_i915_private *dev_priv, int requested_dc; int max_dc; - if (INTEL_GEN(dev_priv) >= 11) { - max_dc = 2; + if (INTEL_GEN(dev_priv) >= 12) { + max_dc = 4; /* * DC9 has a separate HW flow from the rest of the DC states, * not depending on the DMC firmware. It's needed by system * suspend/resume, so allow it unconditionally. */ mask = DC_STATE_EN_DC9; + } else if (IS_GEN(dev_priv, 11)) { + max_dc = 2; + mask = DC_STATE_EN_DC9; } else if (IS_GEN(dev_priv, 10) || IS_GEN9_BC(dev_priv)) { max_dc = 2; mask = 0; @@ -3957,7 +4063,7 @@ static u32 get_allowed_dc_mask(const struct drm_i915_private *dev_priv, requested_dc = enable_dc; } else if (enable_dc == -1) { requested_dc = max_dc; - } else if (enable_dc > max_dc && enable_dc <= 2) { + } else if (enable_dc > max_dc && enable_dc <= 4) { DRM_DEBUG_KMS("Adjusting requested max DC state (%d->%d)\n", enable_dc, max_dc); requested_dc = max_dc; @@ -3966,10 +4072,20 @@ static u32 get_allowed_dc_mask(const struct drm_i915_private *dev_priv, requested_dc = max_dc; } - if (requested_dc > 1) + switch (requested_dc) { + case 4: + mask |= DC_STATE_EN_DC3CO | DC_STATE_EN_UPTO_DC6; + break; + case 3: + mask |= DC_STATE_EN_DC3CO | DC_STATE_EN_UPTO_DC5; + break; + case 2: mask |= DC_STATE_EN_UPTO_DC6; - if (requested_dc > 0) + break; + case 1: mask |= DC_STATE_EN_UPTO_DC5; + break; + } DRM_DEBUG_KMS("Allowed DC state mask %02x\n", mask); @@ -4030,6 +4146,9 @@ int intel_power_domains_init(struct drm_i915_private *dev_priv) dev_priv->csr.allowed_dc_mask = get_allowed_dc_mask(dev_priv, i915_modparams.enable_dc); + dev_priv->csr.target_dc_state = + sanitize_target_dc_state(dev_priv, DC_STATE_EN_UPTO_DC6); + BUILD_BUG_ON(POWER_DOMAIN_NUM > 64); mutex_init(&power_domains->lock); @@ -4896,6 +5015,9 @@ void intel_power_domains_init_hw(struct drm_i915_private *i915, bool resume) power_domains->initializing = true; + /* Must happen before power domain init on VLV/CHV */ + intel_update_rawclk(i915); + if (INTEL_GEN(i915) >= 11) { icl_display_core_init(i915, resume); } else if (IS_CANNONLAKE(i915)) { @@ -5104,8 +5226,7 @@ static void intel_power_domains_dump_info(struct drm_i915_private *i915) for_each_power_domain(domain, power_well->desc->domains) DRM_DEBUG_DRIVER(" %-23s %d\n", - intel_display_power_domain_str(i915, - domain), + intel_display_power_domain_str(domain), power_domains->domain_use_count[domain]); } } diff --git a/drivers/gpu/drm/i915/display/intel_display_power.h b/drivers/gpu/drm/i915/display/intel_display_power.h index a50605b8b1ad..1da04f3e0fb3 100644 --- a/drivers/gpu/drm/i915/display/intel_display_power.h +++ b/drivers/gpu/drm/i915/display/intel_display_power.h @@ -36,29 +36,20 @@ enum intel_display_power_domain { POWER_DOMAIN_PORT_DDI_B_LANES, POWER_DOMAIN_PORT_DDI_C_LANES, POWER_DOMAIN_PORT_DDI_D_LANES, - POWER_DOMAIN_PORT_DDI_TC1_LANES = POWER_DOMAIN_PORT_DDI_D_LANES, POWER_DOMAIN_PORT_DDI_E_LANES, - POWER_DOMAIN_PORT_DDI_TC2_LANES = POWER_DOMAIN_PORT_DDI_E_LANES, POWER_DOMAIN_PORT_DDI_F_LANES, - POWER_DOMAIN_PORT_DDI_TC3_LANES = POWER_DOMAIN_PORT_DDI_F_LANES, - POWER_DOMAIN_PORT_DDI_TC4_LANES, - POWER_DOMAIN_PORT_DDI_TC5_LANES, - POWER_DOMAIN_PORT_DDI_TC6_LANES, + POWER_DOMAIN_PORT_DDI_G_LANES, + POWER_DOMAIN_PORT_DDI_H_LANES, + POWER_DOMAIN_PORT_DDI_I_LANES, POWER_DOMAIN_PORT_DDI_A_IO, POWER_DOMAIN_PORT_DDI_B_IO, POWER_DOMAIN_PORT_DDI_C_IO, POWER_DOMAIN_PORT_DDI_D_IO, - POWER_DOMAIN_PORT_DDI_TC1_IO = POWER_DOMAIN_PORT_DDI_D_IO, POWER_DOMAIN_PORT_DDI_E_IO, - POWER_DOMAIN_PORT_DDI_TC2_IO = POWER_DOMAIN_PORT_DDI_E_IO, POWER_DOMAIN_PORT_DDI_F_IO, - POWER_DOMAIN_PORT_DDI_TC3_IO = POWER_DOMAIN_PORT_DDI_F_IO, POWER_DOMAIN_PORT_DDI_G_IO, - POWER_DOMAIN_PORT_DDI_TC4_IO = POWER_DOMAIN_PORT_DDI_G_IO, POWER_DOMAIN_PORT_DDI_H_IO, - POWER_DOMAIN_PORT_DDI_TC5_IO = POWER_DOMAIN_PORT_DDI_H_IO, POWER_DOMAIN_PORT_DDI_I_IO, - POWER_DOMAIN_PORT_DDI_TC6_IO = POWER_DOMAIN_PORT_DDI_I_IO, POWER_DOMAIN_PORT_DSI, POWER_DOMAIN_PORT_CRT, POWER_DOMAIN_PORT_OTHER, @@ -68,21 +59,19 @@ enum intel_display_power_domain { POWER_DOMAIN_AUX_B, POWER_DOMAIN_AUX_C, POWER_DOMAIN_AUX_D, - POWER_DOMAIN_AUX_TC1 = POWER_DOMAIN_AUX_D, POWER_DOMAIN_AUX_E, - POWER_DOMAIN_AUX_TC2 = POWER_DOMAIN_AUX_E, POWER_DOMAIN_AUX_F, - POWER_DOMAIN_AUX_TC3 = POWER_DOMAIN_AUX_F, - POWER_DOMAIN_AUX_TC4, - POWER_DOMAIN_AUX_TC5, - POWER_DOMAIN_AUX_TC6, + POWER_DOMAIN_AUX_G, + POWER_DOMAIN_AUX_H, + POWER_DOMAIN_AUX_I, POWER_DOMAIN_AUX_IO_A, - POWER_DOMAIN_AUX_TBT1, - POWER_DOMAIN_AUX_TBT2, - POWER_DOMAIN_AUX_TBT3, - POWER_DOMAIN_AUX_TBT4, - POWER_DOMAIN_AUX_TBT5, - POWER_DOMAIN_AUX_TBT6, + POWER_DOMAIN_AUX_C_TBT, + POWER_DOMAIN_AUX_D_TBT, + POWER_DOMAIN_AUX_E_TBT, + POWER_DOMAIN_AUX_F_TBT, + POWER_DOMAIN_AUX_G_TBT, + POWER_DOMAIN_AUX_H_TBT, + POWER_DOMAIN_AUX_I_TBT, POWER_DOMAIN_GMBUS, POWER_DOMAIN_MODESET, POWER_DOMAIN_GT_IRQ, @@ -111,6 +100,7 @@ enum i915_power_well_id { SKL_DISP_PW_MISC_IO, SKL_DISP_PW_1, SKL_DISP_PW_2, + SKL_DISP_DC_OFF, }; #define POWER_DOMAIN_PIPE(pipe) ((pipe) + POWER_DOMAIN_PIPE_A) @@ -267,10 +257,11 @@ void intel_display_power_suspend_late(struct drm_i915_private *i915); void intel_display_power_resume_early(struct drm_i915_private *i915); void intel_display_power_suspend(struct drm_i915_private *i915); void intel_display_power_resume(struct drm_i915_private *i915); +void intel_display_power_set_target_dc_state(struct drm_i915_private *dev_priv, + u32 state); const char * -intel_display_power_domain_str(struct drm_i915_private *i915, - enum intel_display_power_domain domain); +intel_display_power_domain_str(enum intel_display_power_domain domain); bool intel_display_power_is_enabled(struct drm_i915_private *dev_priv, enum intel_display_power_domain domain); diff --git a/drivers/gpu/drm/i915/display/intel_display_types.h b/drivers/gpu/drm/i915/display/intel_display_types.h index 449abaea619f..1a7334dbe802 100644 --- a/drivers/gpu/drm/i915/display/intel_display_types.h +++ b/drivers/gpu/drm/i915/display/intel_display_types.h @@ -128,7 +128,8 @@ struct intel_encoder { enum intel_output_type type; enum port port; - unsigned int cloneable; + u16 cloneable; + u8 pipe_mask; enum intel_hotplug_state (*hotplug)(struct intel_encoder *encoder, struct intel_connector *connector, bool irq_received); @@ -187,7 +188,6 @@ struct intel_encoder { * device interrupts are disabled. */ void (*suspend)(struct intel_encoder *); - int crtc_mask; enum hpd_pin hpd_pin; enum intel_display_power_domain power_domain; /* for communication with audio component; protected by av_mutex */ @@ -388,6 +388,13 @@ struct intel_hdcp { wait_queue_head_t cp_irq_queue; atomic_t cp_irq_count; int cp_irq_count_cached; + + /* + * HDCP register access for gen12+ need the transcoder associated. + * Transcoder attached to the connector could be changed at modeset. + * Hence caching the transcoder here. + */ + enum transcoder cpu_transcoder; }; struct intel_connector { @@ -481,9 +488,9 @@ struct intel_atomic_state { * but the converse is not necessarily true; simply changing a mode may * not flip the final active status of any CRTC's */ - unsigned int active_pipe_changes; + u8 active_pipe_changes; - unsigned int active_crtcs; + u8 active_pipes; /* minimum acceptable cdclk for each pipe */ int min_cdclk[I915_MAX_PIPES]; /* minimum acceptable voltage level for each pipe */ @@ -499,6 +506,14 @@ struct intel_atomic_state { bool rps_interactive; + /* + * active_pipes + * min_cdclk[] + * min_voltage_level[] + * cdclk.* + */ + bool global_state_changed; + /* Gen9+ only */ struct skl_ddb_values wm_results; @@ -552,24 +567,24 @@ struct intel_plane_state { int scaler_id; /* - * linked_plane: + * planar_linked_plane: * * ICL planar formats require 2 planes that are updated as pairs. * This member is used to make sure the other plane is also updated * when required, and for update_slave() to find the correct * plane_state to pass as argument. */ - struct intel_plane *linked_plane; + struct intel_plane *planar_linked_plane; /* - * slave: + * planar_slave: * If set don't update use the linked plane's state for updating * this plane during atomic commit with the update_slave() callback. * * It's also used by the watermark code to ignore wm calculations on * this plane. They're calculated by the linked plane's wm code. */ - u32 slave; + u32 planar_slave; struct drm_intel_sprite_colorkey ckey; }; @@ -759,8 +774,8 @@ struct intel_crtc_state { bool update_pipe; /* can a fast modeset be performed? */ bool disable_cxsr; bool update_wm_pre, update_wm_post; /* watermarks are updated */ - bool fb_changed; /* fb on any of the planes is changed */ bool fifo_changed; /* FIFO split is changed */ + bool preload_luts; /* Pipe source size (ie. panel fitter input size) * All planes will be positioned inside this space, @@ -864,6 +879,7 @@ struct intel_crtc_state { bool has_psr; bool has_psr2; + u32 dc3co_exitline; /* * Frequence the dpll for the port should run at. Differs from the @@ -925,6 +941,8 @@ struct intel_crtc_state { struct intel_crtc_wm_state wm; + int min_cdclk[I915_MAX_PLANES]; + u32 data_rate[I915_MAX_PLANES]; /* Gamma mode programmed on the pipe */ @@ -979,11 +997,17 @@ struct intel_crtc_state { bool dsc_split; u16 compressed_bpp; u8 slice_count; - } dsc_params; - struct drm_dsc_config dp_dsc_cfg; + struct drm_dsc_config config; + } dsc; /* Forward Error correction State */ bool fec_enable; + + /* Pointer to master transcoder in case of tiled displays */ + enum transcoder master_transcoder; + + /* Bitmask to indicate slaves attached */ + u8 sync_mode_slaves_mask; }; struct intel_crtc { @@ -1026,6 +1050,9 @@ struct intel_crtc { /* scalers available on this crtc */ int num_scalers; + + /* per pipe DSB related info */ + struct intel_dsb dsb; }; struct intel_plane { @@ -1061,6 +1088,8 @@ struct intel_plane { bool (*get_hw_state)(struct intel_plane *plane, enum pipe *pipe); int (*check_plane)(struct intel_crtc_state *crtc_state, struct intel_plane_state *plane_state); + int (*min_cdclk)(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state); }; struct intel_watermark_params { @@ -1176,6 +1205,7 @@ struct intel_dp { /* sink or branch descriptor */ struct drm_dp_desc desc; struct drm_dp_aux aux; + u32 aux_busy_last_status; u8 train_set[4]; int panel_power_up_delay; int panel_power_down_delay; @@ -1211,6 +1241,15 @@ struct intel_dp { bool can_mst; /* this port supports mst */ bool is_mst; int active_mst_links; + + /* + * DP_TP_* registers may be either on port or transcoder register space. + */ + struct { + i915_reg_t dp_tp_ctl; + i915_reg_t dp_tp_status; + } regs; + /* connector directly attached - won't be use for modeset in mst world */ struct intel_connector *attached_connector; @@ -1269,6 +1308,7 @@ struct intel_digital_port { char tc_port_name[8]; enum tc_port_mode tc_mode; enum phy_fia tc_phy_fia; + u8 tc_phy_fia_idx; void (*write_infoframe)(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state, @@ -1509,7 +1549,7 @@ intel_wait_for_vblank(struct drm_i915_private *dev_priv, enum pipe pipe) drm_wait_one_vblank(&dev_priv->drm, pipe); } static inline void -intel_wait_for_vblank_if_active(struct drm_i915_private *dev_priv, int pipe) +intel_wait_for_vblank_if_active(struct drm_i915_private *dev_priv, enum pipe pipe) { const struct intel_crtc *crtc = intel_get_crtc_for_pipe(dev_priv, pipe); diff --git a/drivers/gpu/drm/i915/display/intel_dp.c b/drivers/gpu/drm/i915/display/intel_dp.c index 921ad0a2f7ba..050655a1a3d8 100644 --- a/drivers/gpu/drm/i915/display/intel_dp.c +++ b/drivers/gpu/drm/i915/display/intel_dp.c @@ -68,18 +68,13 @@ #define DP_DPRX_ESI_LEN 14 -/* DP DSC small joiner has 2 FIFOs each of 640 x 6 bytes */ -#define DP_DSC_MAX_SMALL_JOINER_RAM_BUFFER 61440 -#define DP_DSC_MIN_SUPPORTED_BPC 8 -#define DP_DSC_MAX_SUPPORTED_BPC 10 - /* DP DSC throughput values used for slice count calculations KPixels/s */ #define DP_DSC_PEAK_PIXEL_RATE 2720000 #define DP_DSC_MAX_ENC_THROUGHPUT_0 340000 #define DP_DSC_MAX_ENC_THROUGHPUT_1 400000 -/* DP DSC FEC Overhead factor = (100 - 2.4)/100 */ -#define DP_DSC_FEC_OVERHEAD_FACTOR 976 +/* DP DSC FEC Overhead factor = 1/(0.972261) */ +#define DP_DSC_FEC_OVERHEAD_FACTOR 972261 /* Compliance test status bits */ #define INTEL_DP_RESOLUTION_SHIFT_MASK 0 @@ -494,6 +489,127 @@ int intel_dp_get_link_train_fallback_values(struct intel_dp *intel_dp, return 0; } +u32 intel_dp_mode_to_fec_clock(u32 mode_clock) +{ + return div_u64(mul_u32_u32(mode_clock, 1000000U), + DP_DSC_FEC_OVERHEAD_FACTOR); +} + +static int +small_joiner_ram_size_bits(struct drm_i915_private *i915) +{ + if (INTEL_GEN(i915) >= 11) + return 7680 * 8; + else + return 6144 * 8; +} + +static u16 intel_dp_dsc_get_output_bpp(struct drm_i915_private *i915, + u32 link_clock, u32 lane_count, + u32 mode_clock, u32 mode_hdisplay) +{ + u32 bits_per_pixel, max_bpp_small_joiner_ram; + int i; + + /* + * Available Link Bandwidth(Kbits/sec) = (NumberOfLanes)* + * (LinkSymbolClock)* 8 * (TimeSlotsPerMTP) + * for SST -> TimeSlotsPerMTP is 1, + * for MST -> TimeSlotsPerMTP has to be calculated + */ + bits_per_pixel = (link_clock * lane_count * 8) / + intel_dp_mode_to_fec_clock(mode_clock); + DRM_DEBUG_KMS("Max link bpp: %u\n", bits_per_pixel); + + /* Small Joiner Check: output bpp <= joiner RAM (bits) / Horiz. width */ + max_bpp_small_joiner_ram = small_joiner_ram_size_bits(i915) / + mode_hdisplay; + DRM_DEBUG_KMS("Max small joiner bpp: %u\n", max_bpp_small_joiner_ram); + + /* + * Greatest allowed DSC BPP = MIN (output BPP from available Link BW + * check, output bpp from small joiner RAM check) + */ + bits_per_pixel = min(bits_per_pixel, max_bpp_small_joiner_ram); + + /* Error out if the max bpp is less than smallest allowed valid bpp */ + if (bits_per_pixel < valid_dsc_bpp[0]) { + DRM_DEBUG_KMS("Unsupported BPP %u, min %u\n", + bits_per_pixel, valid_dsc_bpp[0]); + return 0; + } + + /* Find the nearest match in the array of known BPPs from VESA */ + for (i = 0; i < ARRAY_SIZE(valid_dsc_bpp) - 1; i++) { + if (bits_per_pixel < valid_dsc_bpp[i + 1]) + break; + } + bits_per_pixel = valid_dsc_bpp[i]; + + /* + * Compressed BPP in U6.4 format so multiply by 16, for Gen 11, + * fractional part is 0 + */ + return bits_per_pixel << 4; +} + +static u8 intel_dp_dsc_get_slice_count(struct intel_dp *intel_dp, + int mode_clock, int mode_hdisplay) +{ + u8 min_slice_count, i; + int max_slice_width; + + if (mode_clock <= DP_DSC_PEAK_PIXEL_RATE) + min_slice_count = DIV_ROUND_UP(mode_clock, + DP_DSC_MAX_ENC_THROUGHPUT_0); + else + min_slice_count = DIV_ROUND_UP(mode_clock, + DP_DSC_MAX_ENC_THROUGHPUT_1); + + max_slice_width = drm_dp_dsc_sink_max_slice_width(intel_dp->dsc_dpcd); + if (max_slice_width < DP_DSC_MIN_SLICE_WIDTH_VALUE) { + DRM_DEBUG_KMS("Unsupported slice width %d by DP DSC Sink device\n", + max_slice_width); + return 0; + } + /* Also take into account max slice width */ + min_slice_count = min_t(u8, min_slice_count, + DIV_ROUND_UP(mode_hdisplay, + max_slice_width)); + + /* Find the closest match to the valid slice count values */ + for (i = 0; i < ARRAY_SIZE(valid_dsc_slicecount); i++) { + if (valid_dsc_slicecount[i] > + drm_dp_dsc_sink_max_slice_count(intel_dp->dsc_dpcd, + false)) + break; + if (min_slice_count <= valid_dsc_slicecount[i]) + return valid_dsc_slicecount[i]; + } + + DRM_DEBUG_KMS("Unsupported Slice Count %d\n", min_slice_count); + return 0; +} + +static bool intel_dp_hdisplay_bad(struct drm_i915_private *dev_priv, + int hdisplay) +{ + /* + * Older platforms don't like hdisplay==4096 with DP. + * + * On ILK/SNB/IVB the pipe seems to be somewhat running (scanline + * and frame counter increment), but we don't get vblank interrupts, + * and the pipe underruns immediately. The link also doesn't seem + * to get trained properly. + * + * On CHV the vblank interrupts don't seem to disappear but + * otherwise the symptoms are similar. + * + * TODO: confirm the behaviour on HSW+ + */ + return hdisplay == 4096 && !HAS_DDI(dev_priv); +} + static enum drm_mode_status intel_dp_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) @@ -529,6 +645,9 @@ intel_dp_mode_valid(struct drm_connector *connector, max_rate = intel_dp_max_data_rate(max_link_clock, max_lanes); mode_rate = intel_dp_link_required(target_clock, 18); + if (intel_dp_hdisplay_bad(dev_priv, mode->hdisplay)) + return MODE_H_ILLEGAL; + /* * Output bpp is stored in 6.4 format so right shift by 4 to get the * integer value since we support only integer values of bpp. @@ -543,7 +662,8 @@ intel_dp_mode_valid(struct drm_connector *connector, true); } else if (drm_dp_sink_supports_fec(intel_dp->fec_capable)) { dsc_max_output_bpp = - intel_dp_dsc_get_output_bpp(max_link_clock, + intel_dp_dsc_get_output_bpp(dev_priv, + max_link_clock, max_lanes, target_clock, mode->hdisplay) >> 4; @@ -564,7 +684,7 @@ intel_dp_mode_valid(struct drm_connector *connector, if (mode->flags & DRM_MODE_FLAG_DBLCLK) return MODE_H_ILLEGAL; - return MODE_OK; + return intel_mode_valid_max_plane_size(dev_priv, mode); } u32 intel_dp_pack_aux(const u8 *src, int src_bytes) @@ -641,12 +761,14 @@ vlv_power_sequencer_kick(struct intel_dp *intel_dp) u32 DP; if (WARN(I915_READ(intel_dp->output_reg) & DP_PORT_EN, - "skipping pipe %c power sequencer kick due to port %c being active\n", - pipe_name(pipe), port_name(intel_dig_port->base.port))) + "skipping pipe %c power sequencer kick due to [ENCODER:%d:%s] being active\n", + pipe_name(pipe), intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name)) return; - DRM_DEBUG_KMS("kicking pipe %c power sequencer for port %c\n", - pipe_name(pipe), port_name(intel_dig_port->base.port)); + DRM_DEBUG_KMS("kicking pipe %c power sequencer for [ENCODER:%d:%s]\n", + pipe_name(pipe), intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); /* Preserve the BIOS-computed detected bit. This is * supposed to be read-only. @@ -764,9 +886,10 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) vlv_steal_power_sequencer(dev_priv, pipe); intel_dp->pps_pipe = pipe; - DRM_DEBUG_KMS("picked pipe %c power sequencer for port %c\n", + DRM_DEBUG_KMS("picked pipe %c power sequencer for [ENCODER:%d:%s]\n", pipe_name(intel_dp->pps_pipe), - port_name(intel_dig_port->base.port)); + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); /* init power sequencer on this pipe and port */ intel_dp_init_panel_power_sequencer(intel_dp); @@ -874,13 +997,16 @@ vlv_initial_power_sequencer_setup(struct intel_dp *intel_dp) /* didn't find one? just let vlv_power_sequencer_pipe() pick one when needed */ if (intel_dp->pps_pipe == INVALID_PIPE) { - DRM_DEBUG_KMS("no initial power sequencer for port %c\n", - port_name(port)); + DRM_DEBUG_KMS("no initial power sequencer for [ENCODER:%d:%s]\n", + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); return; } - DRM_DEBUG_KMS("initial power sequencer for port %c: pipe %c\n", - port_name(port), pipe_name(intel_dp->pps_pipe)); + DRM_DEBUG_KMS("initial power sequencer for [ENCODER:%d:%s]: pipe %c\n", + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name, + pipe_name(intel_dp->pps_pipe)); intel_dp_init_panel_power_sequencer(intel_dp); intel_dp_init_panel_power_sequencer_registers(intel_dp, false); @@ -1053,18 +1179,20 @@ intel_dp_aux_wait_done(struct intel_dp *intel_dp) { struct drm_i915_private *i915 = dp_to_i915(intel_dp); i915_reg_t ch_ctl = intel_dp->aux_ch_ctl_reg(intel_dp); + const unsigned int timeout_ms = 10; u32 status; bool done; #define C (((status = intel_uncore_read_notrace(&i915->uncore, ch_ctl)) & DP_AUX_CH_CTL_SEND_BUSY) == 0) done = wait_event_timeout(i915->gmbus_wait_queue, C, - msecs_to_jiffies_timeout(10)); + msecs_to_jiffies_timeout(timeout_ms)); /* just trace the final value */ trace_i915_reg_rw(false, ch_ctl, status, sizeof(status), true); if (!done) - DRM_ERROR("dp aux hw did not signal timeout!\n"); + DRM_ERROR("%s did not complete or timeout within %ums (status 0x%08x)\n", + intel_dp->aux.name, timeout_ms, status); #undef C return status; @@ -1165,6 +1293,9 @@ static u32 skl_get_aux_send_ctl(struct intel_dp *intel_dp, u32 unused) { struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); + struct drm_i915_private *i915 = + to_i915(intel_dig_port->base.base.dev); + enum phy phy = intel_port_to_phy(i915, intel_dig_port->base.port); u32 ret; ret = DP_AUX_CH_CTL_SEND_BUSY | @@ -1177,7 +1308,8 @@ static u32 skl_get_aux_send_ctl(struct intel_dp *intel_dp, DP_AUX_CH_CTL_FW_SYNC_PULSE_SKL(32) | DP_AUX_CH_CTL_SYNC_PULSE_SKL(32); - if (intel_dig_port->tc_mode == TC_PORT_TBT_ALT) + if (intel_phy_is_tc(i915, phy) && + intel_dig_port->tc_mode == TC_PORT_TBT_ALT) ret |= DP_AUX_CH_CTL_TBT_IO; return ret; @@ -1243,13 +1375,12 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp, trace_i915_reg_rw(false, ch_ctl, status, sizeof(status), true); if (try == 3) { - static u32 last_status = -1; const u32 status = intel_uncore_read(uncore, ch_ctl); - if (status != last_status) { + if (status != intel_dp->aux_busy_last_status) { WARN(1, "dp_aux_ch not started status 0x%08x\n", status); - last_status = status; + intel_dp->aux_busy_last_status = status; } ret = -EBUSY; @@ -1541,6 +1672,7 @@ static i915_reg_t skl_aux_ctl_reg(struct intel_dp *intel_dp) case AUX_CH_D: case AUX_CH_E: case AUX_CH_F: + case AUX_CH_G: return DP_AUX_CH_CTL(aux_ch); default: MISSING_CASE(aux_ch); @@ -1561,6 +1693,7 @@ static i915_reg_t skl_aux_data_reg(struct intel_dp *intel_dp, int index) case AUX_CH_D: case AUX_CH_E: case AUX_CH_F: + case AUX_CH_G: return DP_AUX_CH_DATA(aux_ch, index); default: MISSING_CASE(aux_ch); @@ -1739,8 +1872,14 @@ static bool intel_dp_source_supports_fec(struct intel_dp *intel_dp, { struct drm_i915_private *dev_priv = dp_to_i915(intel_dp); - return INTEL_GEN(dev_priv) >= 11 && - pipe_config->cpu_transcoder != TRANSCODER_A; + /* On TGL, FEC is supported on all Pipes */ + if (INTEL_GEN(dev_priv) >= 12) + return true; + + if (IS_GEN(dev_priv, 11) && pipe_config->cpu_transcoder != TRANSCODER_A) + return true; + + return false; } static bool intel_dp_supports_fec(struct intel_dp *intel_dp, @@ -1755,8 +1894,18 @@ static bool intel_dp_source_supports_dsc(struct intel_dp *intel_dp, { struct drm_i915_private *dev_priv = dp_to_i915(intel_dp); - return INTEL_GEN(dev_priv) >= 10 && - pipe_config->cpu_transcoder != TRANSCODER_A; + if (!INTEL_INFO(dev_priv)->display.has_dsc) + return false; + + /* On TGL, DSC is supported on all Pipes */ + if (INTEL_GEN(dev_priv) >= 12) + return true; + + if (INTEL_GEN(dev_priv) >= 10 && + pipe_config->cpu_transcoder != TRANSCODER_A) + return true; + + return false; } static bool intel_dp_supports_dsc(struct intel_dp *intel_dp, @@ -1915,11 +2064,17 @@ static int intel_dp_dsc_compute_config(struct intel_dp *intel_dp, if (!intel_dp_supports_dsc(intel_dp, pipe_config)) return -EINVAL; - dsc_max_bpc = min_t(u8, DP_DSC_MAX_SUPPORTED_BPC, - conn_state->max_requested_bpc); + /* Max DSC Input BPC for ICL is 10 and for TGL+ is 12 */ + if (INTEL_GEN(dev_priv) >= 12) + dsc_max_bpc = min_t(u8, 12, conn_state->max_requested_bpc); + else + dsc_max_bpc = min_t(u8, 10, + conn_state->max_requested_bpc); pipe_bpp = intel_dp_dsc_compute_bpp(intel_dp, dsc_max_bpc); - if (pipe_bpp < DP_DSC_MIN_SUPPORTED_BPC * 3) { + + /* Min Input BPC for ICL+ is 8 */ + if (pipe_bpp < 8 * 3) { DRM_DEBUG_KMS("No DSC support for less than 8bpc\n"); return -EINVAL; } @@ -1934,10 +2089,10 @@ static int intel_dp_dsc_compute_config(struct intel_dp *intel_dp, pipe_config->lane_count = limits->max_lane_count; if (intel_dp_is_edp(intel_dp)) { - pipe_config->dsc_params.compressed_bpp = + pipe_config->dsc.compressed_bpp = min_t(u16, drm_edp_dsc_sink_output_bpp(intel_dp->dsc_dpcd) >> 4, pipe_config->pipe_bpp); - pipe_config->dsc_params.slice_count = + pipe_config->dsc.slice_count = drm_dp_dsc_sink_max_slice_count(intel_dp->dsc_dpcd, true); } else { @@ -1945,7 +2100,8 @@ static int intel_dp_dsc_compute_config(struct intel_dp *intel_dp, u8 dsc_dp_slice_count; dsc_max_output_bpp = - intel_dp_dsc_get_output_bpp(pipe_config->port_clock, + intel_dp_dsc_get_output_bpp(dev_priv, + pipe_config->port_clock, pipe_config->lane_count, adjusted_mode->crtc_clock, adjusted_mode->crtc_hdisplay); @@ -1957,10 +2113,10 @@ static int intel_dp_dsc_compute_config(struct intel_dp *intel_dp, DRM_DEBUG_KMS("Compressed BPP/Slice Count not supported\n"); return -EINVAL; } - pipe_config->dsc_params.compressed_bpp = min_t(u16, + pipe_config->dsc.compressed_bpp = min_t(u16, dsc_max_output_bpp >> 4, pipe_config->pipe_bpp); - pipe_config->dsc_params.slice_count = dsc_dp_slice_count; + pipe_config->dsc.slice_count = dsc_dp_slice_count; } /* * VDSC engine operates at 1 Pixel per clock, so if peak pixel rate @@ -1968,8 +2124,8 @@ static int intel_dp_dsc_compute_config(struct intel_dp *intel_dp, * then we need to use 2 VDSC instances. */ if (adjusted_mode->crtc_clock > dev_priv->max_cdclk_freq) { - if (pipe_config->dsc_params.slice_count > 1) { - pipe_config->dsc_params.dsc_split = true; + if (pipe_config->dsc.slice_count > 1) { + pipe_config->dsc.dsc_split = true; } else { DRM_DEBUG_KMS("Cannot split stream to use 2 VDSC instances\n"); return -EINVAL; @@ -1981,16 +2137,16 @@ static int intel_dp_dsc_compute_config(struct intel_dp *intel_dp, DRM_DEBUG_KMS("Cannot compute valid DSC parameters for Input Bpp = %d " "Compressed BPP = %d\n", pipe_config->pipe_bpp, - pipe_config->dsc_params.compressed_bpp); + pipe_config->dsc.compressed_bpp); return ret; } - pipe_config->dsc_params.compression_enable = true; + pipe_config->dsc.compression_enable = true; DRM_DEBUG_KMS("DP DSC computed with Input Bpp = %d " "Compressed Bpp = %d Slice Count = %d\n", pipe_config->pipe_bpp, - pipe_config->dsc_params.compressed_bpp, - pipe_config->dsc_params.slice_count); + pipe_config->dsc.compressed_bpp, + pipe_config->dsc.slice_count); return 0; } @@ -2064,15 +2220,15 @@ intel_dp_compute_link_config(struct intel_encoder *encoder, return ret; } - if (pipe_config->dsc_params.compression_enable) { + if (pipe_config->dsc.compression_enable) { DRM_DEBUG_KMS("DP lane count %d clock %d Input bpp %d Compressed bpp %d\n", pipe_config->lane_count, pipe_config->port_clock, pipe_config->pipe_bpp, - pipe_config->dsc_params.compressed_bpp); + pipe_config->dsc.compressed_bpp); DRM_DEBUG_KMS("DP link rate required %i available %i\n", intel_dp_link_required(adjusted_mode->crtc_clock, - pipe_config->dsc_params.compressed_bpp), + pipe_config->dsc.compressed_bpp), intel_dp_max_data_rate(pipe_config->port_clock, pipe_config->lane_count)); } else { @@ -2127,6 +2283,16 @@ bool intel_dp_limited_color_range(const struct intel_crtc_state *crtc_state, const struct drm_display_mode *adjusted_mode = &crtc_state->base.adjusted_mode; + /* + * Our YCbCr output is always limited range. + * crtc_state->limited_color_range only applies to RGB, + * and it must never be set for YCbCr or we risk setting + * some conflicting bits in PIPECONF which will mess up + * the colors on the monitor. + */ + if (crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB) + return false; + if (intel_conn_state->broadcast_rgb == INTEL_BROADCAST_RGB_AUTO) { /* * See: @@ -2164,6 +2330,7 @@ intel_dp_compute_config(struct intel_encoder *encoder, pipe_config->has_pch_encoder = true; pipe_config->output_format = INTEL_OUTPUT_FORMAT_RGB; + if (lspcon->active) lspcon_ycbcr420_config(&intel_connector->base, pipe_config); else @@ -2209,6 +2376,9 @@ intel_dp_compute_config(struct intel_encoder *encoder, if (adjusted_mode->flags & DRM_MODE_FLAG_DBLCLK) return -EINVAL; + if (intel_dp_hdisplay_bad(dev_priv, adjusted_mode->crtc_hdisplay)) + return -EINVAL; + ret = intel_dp_compute_link_config(encoder, pipe_config, conn_state); if (ret < 0) return ret; @@ -2216,8 +2386,8 @@ intel_dp_compute_config(struct intel_encoder *encoder, pipe_config->limited_color_range = intel_dp_limited_color_range(pipe_config, conn_state); - if (pipe_config->dsc_params.compression_enable) - output_bpp = pipe_config->dsc_params.compressed_bpp; + if (pipe_config->dsc.compression_enable) + output_bpp = pipe_config->dsc.compressed_bpp; else output_bpp = intel_dp_output_bpp(pipe_config, pipe_config->pipe_bpp); @@ -2226,7 +2396,7 @@ intel_dp_compute_config(struct intel_encoder *encoder, adjusted_mode->crtc_clock, pipe_config->port_clock, &pipe_config->dp_m_n, - constant_n); + constant_n, pipe_config->fec_enable); if (intel_connector->panel.downclock_mode != NULL && dev_priv->drrs.type == SEAMLESS_DRRS_SUPPORT) { @@ -2236,7 +2406,7 @@ intel_dp_compute_config(struct intel_encoder *encoder, intel_connector->panel.downclock_mode->clock, pipe_config->port_clock, &pipe_config->dp_m2_n2, - constant_n); + constant_n, pipe_config->fec_enable); } if (!HAS_DDI(dev_priv)) @@ -2244,6 +2414,9 @@ intel_dp_compute_config(struct intel_encoder *encoder, intel_psr_compute_config(intel_dp, pipe_config); + intel_hdcp_transcoder_config(intel_connector, + pipe_config->cpu_transcoder); + return 0; } @@ -2271,6 +2444,9 @@ static void intel_dp_prepare(struct intel_encoder *encoder, intel_crtc_has_type(pipe_config, INTEL_OUTPUT_DP_MST)); + intel_dp->regs.dp_tp_ctl = DP_TP_CTL(port); + intel_dp->regs.dp_tp_status = DP_TP_STATUS(port); + /* * There are four kinds of DP registers: * @@ -2472,8 +2648,9 @@ static bool edp_panel_vdd_on(struct intel_dp *intel_dp) intel_display_power_get(dev_priv, intel_aux_power_domain(intel_dig_port)); - DRM_DEBUG_KMS("Turning eDP port %c VDD on\n", - port_name(intel_dig_port->base.port)); + DRM_DEBUG_KMS("Turning [ENCODER:%d:%s] VDD on\n", + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); if (!edp_have_panel_power(intel_dp)) wait_panel_power_cycle(intel_dp); @@ -2492,8 +2669,9 @@ static bool edp_panel_vdd_on(struct intel_dp *intel_dp) * If the panel wasn't on, delay before accessing aux channel */ if (!edp_have_panel_power(intel_dp)) { - DRM_DEBUG_KMS("eDP port %c panel power wasn't enabled\n", - port_name(intel_dig_port->base.port)); + DRM_DEBUG_KMS("[ENCODER:%d:%s] panel power wasn't enabled\n", + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); msleep(intel_dp->panel_power_up_delay); } @@ -2518,8 +2696,9 @@ void intel_edp_panel_vdd_on(struct intel_dp *intel_dp) vdd = false; with_pps_lock(intel_dp, wakeref) vdd = edp_panel_vdd_on(intel_dp); - I915_STATE_WARN(!vdd, "eDP port %c VDD already requested on\n", - port_name(dp_to_dig_port(intel_dp)->base.port)); + I915_STATE_WARN(!vdd, "[ENCODER:%d:%s] VDD already requested on\n", + dp_to_dig_port(intel_dp)->base.base.base.id, + dp_to_dig_port(intel_dp)->base.base.name); } static void edp_panel_vdd_off_sync(struct intel_dp *intel_dp) @@ -2537,8 +2716,9 @@ static void edp_panel_vdd_off_sync(struct intel_dp *intel_dp) if (!edp_have_panel_vdd(intel_dp)) return; - DRM_DEBUG_KMS("Turning eDP port %c VDD off\n", - port_name(intel_dig_port->base.port)); + DRM_DEBUG_KMS("Turning [ENCODER:%d:%s] VDD off\n", + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); pp = ironlake_get_pp_control(intel_dp); pp &= ~EDP_FORCE_VDD; @@ -2600,8 +2780,9 @@ static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync) if (!intel_dp_is_edp(intel_dp)) return; - I915_STATE_WARN(!intel_dp->want_panel_vdd, "eDP port %c VDD not forced on", - port_name(dp_to_dig_port(intel_dp)->base.port)); + I915_STATE_WARN(!intel_dp->want_panel_vdd, "[ENCODER:%d:%s] VDD not forced on", + dp_to_dig_port(intel_dp)->base.base.base.id, + dp_to_dig_port(intel_dp)->base.base.name); intel_dp->want_panel_vdd = false; @@ -2622,12 +2803,14 @@ static void edp_panel_on(struct intel_dp *intel_dp) if (!intel_dp_is_edp(intel_dp)) return; - DRM_DEBUG_KMS("Turn eDP port %c panel power on\n", - port_name(dp_to_dig_port(intel_dp)->base.port)); + DRM_DEBUG_KMS("Turn [ENCODER:%d:%s] panel power on\n", + dp_to_dig_port(intel_dp)->base.base.base.id, + dp_to_dig_port(intel_dp)->base.base.name); if (WARN(edp_have_panel_power(intel_dp), - "eDP port %c panel power already on\n", - port_name(dp_to_dig_port(intel_dp)->base.port))) + "[ENCODER:%d:%s] panel power already on\n", + dp_to_dig_port(intel_dp)->base.base.base.id, + dp_to_dig_port(intel_dp)->base.base.name)) return; wait_panel_power_cycle(intel_dp); @@ -2682,11 +2865,11 @@ static void edp_panel_off(struct intel_dp *intel_dp) if (!intel_dp_is_edp(intel_dp)) return; - DRM_DEBUG_KMS("Turn eDP port %c panel power off\n", - port_name(dig_port->base.port)); + DRM_DEBUG_KMS("Turn [ENCODER:%d:%s] panel power off\n", + dig_port->base.base.base.id, dig_port->base.base.name); - WARN(!intel_dp->want_panel_vdd, "Need eDP port %c VDD to turn off panel\n", - port_name(dig_port->base.port)); + WARN(!intel_dp->want_panel_vdd, "Need [ENCODER:%d:%s] VDD to turn off panel\n", + dig_port->base.base.base.id, dig_port->base.base.name); pp = ironlake_get_pp_control(intel_dp); /* We need to switch off panel power _and_ force vdd, for otherwise some @@ -2831,8 +3014,8 @@ static void assert_dp_port(struct intel_dp *intel_dp, bool state) bool cur_state = I915_READ(intel_dp->output_reg) & DP_PORT_EN; I915_STATE_WARN(cur_state != state, - "DP port %c state assertion failure (expected %s, current %s)\n", - port_name(dig_port->base.port), + "[ENCODER:%d:%s] state assertion failure (expected %s, current %s)\n", + dig_port->base.base.base.id, dig_port->base.base.name, onoff(state), onoff(cur_state)); } #define assert_dp_port_disabled(d) assert_dp_port((d), false) @@ -2928,7 +3111,7 @@ void intel_dp_sink_set_decompression_state(struct intel_dp *intel_dp, { int ret; - if (!crtc_state->dsc_params.compression_enable) + if (!crtc_state->dsc.compression_enable) return; ret = drm_dp_dpcd_writeb(&intel_dp->aux, DP_DSC_ENABLE, @@ -3220,7 +3403,7 @@ _intel_dp_set_link_train(struct intel_dp *intel_dp, dp_train_pat & train_pat_mask); if (HAS_DDI(dev_priv)) { - u32 temp = I915_READ(DP_TP_CTL(port)); + u32 temp = I915_READ(intel_dp->regs.dp_tp_ctl); if (dp_train_pat & DP_LINK_SCRAMBLING_DISABLE) temp |= DP_TP_CTL_SCRAMBLE_DISABLE; @@ -3246,7 +3429,7 @@ _intel_dp_set_link_train(struct intel_dp *intel_dp, temp |= DP_TP_CTL_LINK_TRAIN_PAT4; break; } - I915_WRITE(DP_TP_CTL(port), temp); + I915_WRITE(intel_dp->regs.dp_tp_ctl, temp); } else if ((IS_IVYBRIDGE(dev_priv) && port == PORT_A) || (HAS_PCH_CPT(dev_priv) && port != PORT_A)) { @@ -3410,8 +3593,9 @@ static void vlv_detach_power_sequencer(struct intel_dp *intel_dp) * port select always when logically disconnecting a power sequencer * from a port. */ - DRM_DEBUG_KMS("detaching pipe %c power sequencer from port %c\n", - pipe_name(pipe), port_name(intel_dig_port->base.port)); + DRM_DEBUG_KMS("detaching pipe %c power sequencer from [ENCODER:%d:%s]\n", + pipe_name(pipe), intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); I915_WRITE(pp_on_reg, 0); POSTING_READ(pp_on_reg); @@ -3427,17 +3611,18 @@ static void vlv_steal_power_sequencer(struct drm_i915_private *dev_priv, for_each_intel_dp(&dev_priv->drm, encoder) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - enum port port = encoder->port; WARN(intel_dp->active_pipe == pipe, - "stealing pipe %c power sequencer from active (e)DP port %c\n", - pipe_name(pipe), port_name(port)); + "stealing pipe %c power sequencer from active [ENCODER:%d:%s]\n", + pipe_name(pipe), encoder->base.base.id, + encoder->base.name); if (intel_dp->pps_pipe != pipe) continue; - DRM_DEBUG_KMS("stealing pipe %c power sequencer from port %c\n", - pipe_name(pipe), port_name(port)); + DRM_DEBUG_KMS("stealing pipe %c power sequencer from [ENCODER:%d:%s]\n", + pipe_name(pipe), encoder->base.base.id, + encoder->base.name); /* make sure vdd is off before we steal it */ vlv_detach_power_sequencer(intel_dp); @@ -3479,8 +3664,9 @@ static void vlv_init_panel_power_sequencer(struct intel_encoder *encoder, /* now it's all ours */ intel_dp->pps_pipe = crtc->pipe; - DRM_DEBUG_KMS("initializing pipe %c power sequencer for port %c\n", - pipe_name(intel_dp->pps_pipe), port_name(encoder->port)); + DRM_DEBUG_KMS("initializing pipe %c power sequencer for [ENCODER:%d:%s]\n", + pipe_name(intel_dp->pps_pipe), encoder->base.base.id, + encoder->base.name); /* init power sequencer on this pipe and port */ intel_dp_init_panel_power_sequencer(intel_dp); @@ -3944,22 +4130,22 @@ void intel_dp_set_idle_link_train(struct intel_dp *intel_dp) if (!HAS_DDI(dev_priv)) return; - val = I915_READ(DP_TP_CTL(port)); + val = I915_READ(intel_dp->regs.dp_tp_ctl); val &= ~DP_TP_CTL_LINK_TRAIN_MASK; val |= DP_TP_CTL_LINK_TRAIN_IDLE; - I915_WRITE(DP_TP_CTL(port), val); + I915_WRITE(intel_dp->regs.dp_tp_ctl, val); /* - * On PORT_A we can have only eDP in SST mode. There the only reason - * we need to set idle transmission mode is to work around a HW issue - * where we enable the pipe while not in idle link-training mode. + * Until TGL on PORT_A we can have only eDP in SST mode. There the only + * reason we need to set idle transmission mode is to work around a HW + * issue where we enable the pipe while not in idle link-training mode. * In this case there is requirement to wait for a minimum number of * idle patterns to be sent. */ - if (port == PORT_A) + if (port == PORT_A && INTEL_GEN(dev_priv) < 12) return; - if (intel_de_wait_for_set(dev_priv, DP_TP_STATUS(port), + if (intel_de_wait_for_set(dev_priv, intel_dp->regs.dp_tp_status, DP_TP_STATUS_IDLE_DONE, 1)) DRM_ERROR("Timed out waiting for DP idle patterns\n"); } @@ -4301,9 +4487,10 @@ intel_dp_configure_mst(struct intel_dp *intel_dp) &dp_to_dig_port(intel_dp)->base; bool sink_can_mst = intel_dp_sink_can_mst(intel_dp); - DRM_DEBUG_KMS("MST support? port %c: %s, sink: %s, modparam: %s\n", - port_name(encoder->port), yesno(intel_dp->can_mst), - yesno(sink_can_mst), yesno(i915_modparams.enable_dp_mst)); + DRM_DEBUG_KMS("[ENCODER:%d:%s] MST support: port: %s, sink: %s, modparam: %s\n", + encoder->base.base.id, encoder->base.name, + yesno(intel_dp->can_mst), yesno(sink_can_mst), + yesno(i915_modparams.enable_dp_mst)); if (!intel_dp->can_mst) return; @@ -4323,94 +4510,36 @@ intel_dp_get_sink_irq_esi(struct intel_dp *intel_dp, u8 *sink_irq_vector) DP_DPRX_ESI_LEN; } -u16 intel_dp_dsc_get_output_bpp(int link_clock, u8 lane_count, - int mode_clock, int mode_hdisplay) +bool +intel_dp_needs_vsc_sdp(const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { - u16 bits_per_pixel, max_bpp_small_joiner_ram; - int i; - /* - * Available Link Bandwidth(Kbits/sec) = (NumberOfLanes)* - * (LinkSymbolClock)* 8 * ((100-FECOverhead)/100)*(TimeSlotsPerMTP) - * FECOverhead = 2.4%, for SST -> TimeSlotsPerMTP is 1, - * for MST -> TimeSlotsPerMTP has to be calculated + * As per DP 1.4a spec section 2.2.4.3 [MSA Field for Indication + * of Color Encoding Format and Content Color Gamut], in order to + * sending YCBCR 420 or HDR BT.2020 signals we should use DP VSC SDP. */ - bits_per_pixel = (link_clock * lane_count * 8 * - DP_DSC_FEC_OVERHEAD_FACTOR) / - mode_clock; - - /* Small Joiner Check: output bpp <= joiner RAM (bits) / Horiz. width */ - max_bpp_small_joiner_ram = DP_DSC_MAX_SMALL_JOINER_RAM_BUFFER / - mode_hdisplay; - - /* - * Greatest allowed DSC BPP = MIN (output BPP from avaialble Link BW - * check, output bpp from small joiner RAM check) - */ - bits_per_pixel = min(bits_per_pixel, max_bpp_small_joiner_ram); - - /* Error out if the max bpp is less than smallest allowed valid bpp */ - if (bits_per_pixel < valid_dsc_bpp[0]) { - DRM_DEBUG_KMS("Unsupported BPP %d\n", bits_per_pixel); - return 0; - } - - /* Find the nearest match in the array of known BPPs from VESA */ - for (i = 0; i < ARRAY_SIZE(valid_dsc_bpp) - 1; i++) { - if (bits_per_pixel < valid_dsc_bpp[i + 1]) - break; - } - bits_per_pixel = valid_dsc_bpp[i]; - - /* - * Compressed BPP in U6.4 format so multiply by 16, for Gen 11, - * fractional part is 0 - */ - return bits_per_pixel << 4; -} - -u8 intel_dp_dsc_get_slice_count(struct intel_dp *intel_dp, - int mode_clock, - int mode_hdisplay) -{ - u8 min_slice_count, i; - int max_slice_width; - - if (mode_clock <= DP_DSC_PEAK_PIXEL_RATE) - min_slice_count = DIV_ROUND_UP(mode_clock, - DP_DSC_MAX_ENC_THROUGHPUT_0); - else - min_slice_count = DIV_ROUND_UP(mode_clock, - DP_DSC_MAX_ENC_THROUGHPUT_1); - - max_slice_width = drm_dp_dsc_sink_max_slice_width(intel_dp->dsc_dpcd); - if (max_slice_width < DP_DSC_MIN_SLICE_WIDTH_VALUE) { - DRM_DEBUG_KMS("Unsupported slice width %d by DP DSC Sink device\n", - max_slice_width); - return 0; - } - /* Also take into account max slice width */ - min_slice_count = min_t(u8, min_slice_count, - DIV_ROUND_UP(mode_hdisplay, - max_slice_width)); + if (crtc_state->output_format == INTEL_OUTPUT_FORMAT_YCBCR420) + return true; - /* Find the closest match to the valid slice count values */ - for (i = 0; i < ARRAY_SIZE(valid_dsc_slicecount); i++) { - if (valid_dsc_slicecount[i] > - drm_dp_dsc_sink_max_slice_count(intel_dp->dsc_dpcd, - false)) - break; - if (min_slice_count <= valid_dsc_slicecount[i]) - return valid_dsc_slicecount[i]; + switch (conn_state->colorspace) { + case DRM_MODE_COLORIMETRY_SYCC_601: + case DRM_MODE_COLORIMETRY_OPYCC_601: + case DRM_MODE_COLORIMETRY_BT2020_YCC: + case DRM_MODE_COLORIMETRY_BT2020_RGB: + case DRM_MODE_COLORIMETRY_BT2020_CYCC: + return true; + default: + break; } - DRM_DEBUG_KMS("Unsupported Slice Count %d\n", min_slice_count); - return 0; + return false; } static void -intel_pixel_encoding_setup_vsc(struct intel_dp *intel_dp, - const struct intel_crtc_state *crtc_state) +intel_dp_setup_vsc_sdp(struct intel_dp *intel_dp, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); struct dp_sdp vsc_sdp = {}; @@ -4431,13 +4560,55 @@ intel_pixel_encoding_setup_vsc(struct intel_dp *intel_dp, */ vsc_sdp.sdp_header.HB3 = 0x13; - /* - * YCbCr 420 = 3h DB16[7:4] ITU-R BT.601 = 0h, ITU-R BT.709 = 1h - * DB16[3:0] DP 1.4a spec, Table 2-120 - */ - vsc_sdp.db[16] = 0x3 << 4; /* 0x3 << 4 , YCbCr 420*/ - /* RGB->YCBCR color conversion uses the BT.709 color space. */ - vsc_sdp.db[16] |= 0x1; /* 0x1, ITU-R BT.709 */ + /* DP 1.4a spec, Table 2-120 */ + switch (crtc_state->output_format) { + case INTEL_OUTPUT_FORMAT_YCBCR444: + vsc_sdp.db[16] = 0x1 << 4; /* YCbCr 444 : DB16[7:4] = 1h */ + break; + case INTEL_OUTPUT_FORMAT_YCBCR420: + vsc_sdp.db[16] = 0x3 << 4; /* YCbCr 420 : DB16[7:4] = 3h */ + break; + case INTEL_OUTPUT_FORMAT_RGB: + default: + /* RGB: DB16[7:4] = 0h */ + break; + } + + switch (conn_state->colorspace) { + case DRM_MODE_COLORIMETRY_BT709_YCC: + vsc_sdp.db[16] |= 0x1; + break; + case DRM_MODE_COLORIMETRY_XVYCC_601: + vsc_sdp.db[16] |= 0x2; + break; + case DRM_MODE_COLORIMETRY_XVYCC_709: + vsc_sdp.db[16] |= 0x3; + break; + case DRM_MODE_COLORIMETRY_SYCC_601: + vsc_sdp.db[16] |= 0x4; + break; + case DRM_MODE_COLORIMETRY_OPYCC_601: + vsc_sdp.db[16] |= 0x5; + break; + case DRM_MODE_COLORIMETRY_BT2020_CYCC: + case DRM_MODE_COLORIMETRY_BT2020_RGB: + vsc_sdp.db[16] |= 0x6; + break; + case DRM_MODE_COLORIMETRY_BT2020_YCC: + vsc_sdp.db[16] |= 0x7; + break; + case DRM_MODE_COLORIMETRY_DCI_P3_RGB_D65: + case DRM_MODE_COLORIMETRY_DCI_P3_RGB_THEATER: + vsc_sdp.db[16] |= 0x4; /* DCI-P3 (SMPTE RP 431-2) */ + break; + default: + /* sRGB (IEC 61966-2-1) / ITU-R BT.601: DB16[0:3] = 0h */ + + /* RGB->YCBCR color conversion uses the BT.709 color space. */ + if (crtc_state->output_format == INTEL_OUTPUT_FORMAT_YCBCR420) + vsc_sdp.db[16] |= 0x1; /* 0x1, ITU-R BT.709 */ + break; + } /* * For pixel encoding formats YCbCr444, YCbCr422, YCbCr420, and Y Only, @@ -4489,13 +4660,106 @@ intel_pixel_encoding_setup_vsc(struct intel_dp *intel_dp, crtc_state, DP_SDP_VSC, &vsc_sdp, sizeof(vsc_sdp)); } -void intel_dp_ycbcr_420_enable(struct intel_dp *intel_dp, - const struct intel_crtc_state *crtc_state) +static void +intel_dp_setup_hdr_metadata_infoframe_sdp(struct intel_dp *intel_dp, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) +{ + struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); + struct dp_sdp infoframe_sdp = {}; + struct hdmi_drm_infoframe drm_infoframe = {}; + const int infoframe_size = HDMI_INFOFRAME_HEADER_SIZE + HDMI_DRM_INFOFRAME_SIZE; + unsigned char buf[HDMI_INFOFRAME_HEADER_SIZE + HDMI_DRM_INFOFRAME_SIZE]; + ssize_t len; + int ret; + + ret = drm_hdmi_infoframe_set_hdr_metadata(&drm_infoframe, conn_state); + if (ret) { + DRM_DEBUG_KMS("couldn't set HDR metadata in infoframe\n"); + return; + } + + len = hdmi_drm_infoframe_pack_only(&drm_infoframe, buf, sizeof(buf)); + if (len < 0) { + DRM_DEBUG_KMS("buffer size is smaller than hdr metadata infoframe\n"); + return; + } + + if (len != infoframe_size) { + DRM_DEBUG_KMS("wrong static hdr metadata size\n"); + return; + } + + /* + * Set up the infoframe sdp packet for HDR static metadata. + * Prepare VSC Header for SU as per DP 1.4a spec, + * Table 2-100 and Table 2-101 + */ + + /* Packet ID, 00h for non-Audio INFOFRAME */ + infoframe_sdp.sdp_header.HB0 = 0; + /* + * Packet Type 80h + Non-audio INFOFRAME Type value + * HDMI_INFOFRAME_TYPE_DRM: 0x87, + */ + infoframe_sdp.sdp_header.HB1 = drm_infoframe.type; + /* + * Least Significant Eight Bits of (Data Byte Count – 1) + * infoframe_size - 1, + */ + infoframe_sdp.sdp_header.HB2 = 0x1D; + /* INFOFRAME SDP Version Number */ + infoframe_sdp.sdp_header.HB3 = (0x13 << 2); + /* CTA Header Byte 2 (INFOFRAME Version Number) */ + infoframe_sdp.db[0] = drm_infoframe.version; + /* CTA Header Byte 3 (Length of INFOFRAME): HDMI_DRM_INFOFRAME_SIZE */ + infoframe_sdp.db[1] = drm_infoframe.length; + /* + * Copy HDMI_DRM_INFOFRAME_SIZE size from a buffer after + * HDMI_INFOFRAME_HEADER_SIZE + */ + BUILD_BUG_ON(sizeof(infoframe_sdp.db) < HDMI_DRM_INFOFRAME_SIZE + 2); + memcpy(&infoframe_sdp.db[2], &buf[HDMI_INFOFRAME_HEADER_SIZE], + HDMI_DRM_INFOFRAME_SIZE); + + /* + * Size of DP infoframe sdp packet for HDR static metadata is consist of + * - DP SDP Header(struct dp_sdp_header): 4 bytes + * - Two Data Blocks: 2 bytes + * CTA Header Byte2 (INFOFRAME Version Number) + * CTA Header Byte3 (Length of INFOFRAME) + * - HDMI_DRM_INFOFRAME_SIZE: 26 bytes + * + * Prior to GEN11's GMP register size is identical to DP HDR static metadata + * infoframe size. But GEN11+ has larger than that size, write_infoframe + * will pad rest of the size. + */ + intel_dig_port->write_infoframe(&intel_dig_port->base, crtc_state, + HDMI_PACKET_TYPE_GAMUT_METADATA, + &infoframe_sdp, + sizeof(struct dp_sdp_header) + 2 + HDMI_DRM_INFOFRAME_SIZE); +} + +void intel_dp_vsc_enable(struct intel_dp *intel_dp, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { - if (crtc_state->output_format != INTEL_OUTPUT_FORMAT_YCBCR420) + if (!intel_dp_needs_vsc_sdp(crtc_state, conn_state)) return; - intel_pixel_encoding_setup_vsc(intel_dp, crtc_state); + intel_dp_setup_vsc_sdp(intel_dp, crtc_state, conn_state); +} + +void intel_dp_hdr_metadata_enable(struct intel_dp *intel_dp, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) +{ + if (!conn_state->hdr_output_metadata) + return; + + intel_dp_setup_hdr_metadata_infoframe_sdp(intel_dp, + crtc_state, + conn_state); } static u8 intel_dp_autotest_link_training(struct intel_dp *intel_dp) @@ -5212,12 +5476,13 @@ static bool bxt_digital_port_connected(struct intel_encoder *encoder) return I915_READ(GEN8_DE_PORT_ISR) & bit; } -static bool icl_combo_port_connected(struct drm_i915_private *dev_priv, - struct intel_digital_port *intel_dig_port) +static bool intel_combo_phy_connected(struct drm_i915_private *dev_priv, + enum phy phy) { - enum port port = intel_dig_port->base.port; + if (HAS_PCH_MCC(dev_priv) && phy == PHY_C) + return I915_READ(SDEISR) & SDE_TC_HOTPLUG_ICP(PORT_TC1); - return I915_READ(SDEISR) & SDE_DDI_HOTPLUG_ICP(port); + return I915_READ(SDEISR) & SDE_DDI_HOTPLUG_ICP(phy); } static bool icl_digital_port_connected(struct intel_encoder *encoder) @@ -5227,7 +5492,7 @@ static bool icl_digital_port_connected(struct intel_encoder *encoder) enum phy phy = intel_port_to_phy(dev_priv, encoder->port); if (intel_phy_is_combo(dev_priv, phy)) - return icl_combo_port_connected(dev_priv, dig_port); + return intel_combo_phy_connected(dev_priv, phy); else if (intel_phy_is_tc(dev_priv, phy)) return intel_tc_port_connected(dig_port); else @@ -5430,6 +5695,12 @@ out: if (status != connector_status_connected && !intel_dp->is_mst) intel_dp_unset_edid(intel_dp); + /* + * Make sure the refs for power wells enabled during detect are + * dropped to avoid a new detect cycle triggered by HPD polling. + */ + intel_display_power_flush_work(dev_priv); + return status; } @@ -5490,7 +5761,6 @@ static int intel_dp_connector_register(struct drm_connector *connector) { struct intel_dp *intel_dp = intel_attached_dp(connector); - struct drm_device *dev = connector->dev; int ret; ret = intel_connector_register(connector); @@ -5505,8 +5775,7 @@ intel_dp_connector_register(struct drm_connector *connector) intel_dp->aux.dev = connector->kdev; ret = drm_dp_aux_register(&intel_dp->aux); if (!ret) - drm_dp_cec_register_connector(&intel_dp->aux, - connector->name, dev->dev); + drm_dp_cec_register_connector(&intel_dp->aux, connector); return ret; } @@ -6264,13 +6533,15 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd) * would end up in an endless cycle of * "vdd off -> long hpd -> vdd on -> detect -> vdd off -> ..." */ - DRM_DEBUG_KMS("ignoring long hpd on eDP port %c\n", - port_name(intel_dig_port->base.port)); + DRM_DEBUG_KMS("ignoring long hpd on eDP [ENCODER:%d:%s]\n", + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); return IRQ_HANDLED; } - DRM_DEBUG_KMS("got hpd irq on port %c - %s\n", - port_name(intel_dig_port->base.port), + DRM_DEBUG_KMS("got hpd irq on [ENCODER:%d:%s] - %s\n", + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name, long_hpd ? "long" : "short"); if (long_hpd) { @@ -6337,6 +6608,13 @@ intel_dp_add_properties(struct intel_dp *intel_dp, struct drm_connector *connect else if (INTEL_GEN(dev_priv) >= 5) drm_connector_attach_max_bpc_property(connector, 6, 12); + intel_attach_colorspace_property(connector); + + if (IS_GEMINILAKE(dev_priv) || INTEL_GEN(dev_priv) >= 11) + drm_object_attach_property(&connector->base, + connector->dev->mode_config.hdr_output_metadata_property, + 0); + if (intel_dp_is_edp(intel_dp)) { u32 allowed_scalers; @@ -7134,8 +7412,9 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, intel_dp_modeset_retry_work_fn); if (WARN(intel_dig_port->max_lanes < 1, - "Not enough lanes (%d) for DP on port %c\n", - intel_dig_port->max_lanes, port_name(port))) + "Not enough lanes (%d) for DP on [ENCODER:%d:%s]\n", + intel_dig_port->max_lanes, intel_encoder->base.base.id, + intel_encoder->base.name)) return false; intel_dp_set_source_rates(intel_dp); @@ -7176,9 +7455,9 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, port != PORT_B && port != PORT_C)) return false; - DRM_DEBUG_KMS("Adding %s connector on port %c\n", - type == DRM_MODE_CONNECTOR_eDP ? "eDP" : "DP", - port_name(port)); + DRM_DEBUG_KMS("Adding %s connector on [ENCODER:%d:%s]\n", + type == DRM_MODE_CONNECTOR_eDP ? "eDP" : "DP", + intel_encoder->base.base.id, intel_encoder->base.name); drm_connector_init(dev, connector, &intel_dp_connector_funcs, type); drm_connector_helper_add(connector, &intel_dp_connector_helper_funcs); @@ -7202,11 +7481,8 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, intel_connector->get_hw_state = intel_connector_get_hw_state; /* init MST on ports that can support it */ - if (HAS_DP_MST(dev_priv) && !intel_dp_is_edp(intel_dp) && - (port == PORT_B || port == PORT_C || - port == PORT_D || port == PORT_F)) - intel_dp_mst_encoder_init(intel_dig_port, - intel_connector->base.base.id); + intel_dp_mst_encoder_init(intel_dig_port, + intel_connector->base.base.id); if (!intel_edp_init_connector(intel_dp, intel_connector)) { intel_dp_aux_fini(intel_dp); @@ -7297,11 +7573,11 @@ bool intel_dp_init(struct drm_i915_private *dev_priv, intel_encoder->power_domain = intel_port_to_power_domain(port); if (IS_CHERRYVIEW(dev_priv)) { if (port == PORT_D) - intel_encoder->crtc_mask = 1 << 2; + intel_encoder->pipe_mask = BIT(PIPE_C); else - intel_encoder->crtc_mask = (1 << 0) | (1 << 1); + intel_encoder->pipe_mask = BIT(PIPE_A) | BIT(PIPE_B); } else { - intel_encoder->crtc_mask = (1 << 0) | (1 << 1) | (1 << 2); + intel_encoder->pipe_mask = ~0; } intel_encoder->cloneable = 0; intel_encoder->port = port; @@ -7362,7 +7638,8 @@ void intel_dp_mst_resume(struct drm_i915_private *dev_priv) if (!intel_dp->can_mst) continue; - ret = drm_dp_mst_topology_mgr_resume(&intel_dp->mst_mgr); + ret = drm_dp_mst_topology_mgr_resume(&intel_dp->mst_mgr, + true); if (ret) { intel_dp->is_mst = false; drm_dp_mst_topology_mgr_set_mst(&intel_dp->mst_mgr, diff --git a/drivers/gpu/drm/i915/display/intel_dp.h b/drivers/gpu/drm/i915/display/intel_dp.h index 657bbb1f5ed0..3da166054788 100644 --- a/drivers/gpu/drm/i915/display/intel_dp.h +++ b/drivers/gpu/drm/i915/display/intel_dp.h @@ -13,6 +13,7 @@ #include "i915_reg.h" enum pipe; +enum port; struct drm_connector_state; struct drm_encoder; struct drm_i915_private; @@ -102,15 +103,19 @@ bool intel_dp_source_supports_hbr2(struct intel_dp *intel_dp); bool intel_dp_source_supports_hbr3(struct intel_dp *intel_dp); bool intel_dp_get_link_status(struct intel_dp *intel_dp, u8 *link_status); -u16 intel_dp_dsc_get_output_bpp(int link_clock, u8 lane_count, - int mode_clock, int mode_hdisplay); -u8 intel_dp_dsc_get_slice_count(struct intel_dp *intel_dp, int mode_clock, - int mode_hdisplay); bool intel_dp_read_dpcd(struct intel_dp *intel_dp); bool intel_dp_get_colorimetry_status(struct intel_dp *intel_dp); int intel_dp_link_required(int pixel_clock, int bpp); int intel_dp_max_data_rate(int max_link_clock, int max_lanes); +bool intel_dp_needs_vsc_sdp(const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state); +void intel_dp_vsc_enable(struct intel_dp *intel_dp, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state); +void intel_dp_hdr_metadata_enable(struct intel_dp *intel_dp, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state); bool intel_digital_port_connected(struct intel_encoder *encoder); static inline unsigned int intel_dp_unused_lane_mask(int lane_count) @@ -118,4 +123,6 @@ static inline unsigned int intel_dp_unused_lane_mask(int lane_count) return ~((1 << lane_count) - 1) & 0xf; } +u32 intel_dp_mode_to_fec_clock(u32 mode_clock); + #endif /* __INTEL_DP_H__ */ diff --git a/drivers/gpu/drm/i915/display/intel_dp_mst.c b/drivers/gpu/drm/i915/display/intel_dp_mst.c index 6df240a01b8c..03d1cba0b696 100644 --- a/drivers/gpu/drm/i915/display/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/display/intel_dp_mst.c @@ -81,7 +81,7 @@ static int intel_dp_mst_compute_link_config(struct intel_encoder *encoder, adjusted_mode->crtc_clock, crtc_state->port_clock, &crtc_state->dp_m_n, - constant_n); + constant_n, crtc_state->fec_enable); crtc_state->dp_m_n.tu = slots; return 0; @@ -215,7 +215,7 @@ static void intel_mst_disable_dp(struct intel_encoder *encoder, ret = drm_dp_update_payload_part1(&intel_dp->mst_mgr); if (ret) { - DRM_ERROR("failed to update payload %d\n", ret); + DRM_DEBUG_KMS("failed to update payload %d\n", ret); } if (old_crtc_state->has_audio) intel_audio_codec_disable(encoder, @@ -295,7 +295,6 @@ static void intel_mst_pre_enable_dp(struct intel_encoder *encoder, struct intel_digital_port *intel_dig_port = intel_mst->primary; struct intel_dp *intel_dp = &intel_dig_port->dp; struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_dig_port->base.port; struct intel_connector *connector = to_intel_connector(conn_state->connector); int ret; @@ -326,12 +325,14 @@ static void intel_mst_pre_enable_dp(struct intel_encoder *encoder, DRM_ERROR("failed to allocate vcpi\n"); intel_dp->active_mst_links++; - temp = I915_READ(DP_TP_STATUS(port)); - I915_WRITE(DP_TP_STATUS(port), temp); + temp = I915_READ(intel_dp->regs.dp_tp_status); + I915_WRITE(intel_dp->regs.dp_tp_status, temp); ret = drm_dp_update_payload_part1(&intel_dp->mst_mgr); intel_ddi_enable_pipe_clock(pipe_config); + + intel_ddi_set_dp_msa(pipe_config, conn_state); } static void intel_mst_enable_dp(struct intel_encoder *encoder, @@ -342,11 +343,10 @@ static void intel_mst_enable_dp(struct intel_encoder *encoder, struct intel_digital_port *intel_dig_port = intel_mst->primary; struct intel_dp *intel_dp = &intel_dig_port->dp; struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_dig_port->base.port; DRM_DEBUG_KMS("active links %d\n", intel_dp->active_mst_links); - if (intel_de_wait_for_set(dev_priv, DP_TP_STATUS(port), + if (intel_de_wait_for_set(dev_priv, intel_dp->regs.dp_tp_status, DP_TP_STATUS_ACT_SENT, 1)) DRM_ERROR("Timed out waiting for ACT sent\n"); @@ -393,20 +393,7 @@ static int intel_dp_mst_get_ddc_modes(struct drm_connector *connector) return ret; } -static enum drm_connector_status -intel_dp_mst_detect(struct drm_connector *connector, bool force) -{ - struct intel_connector *intel_connector = to_intel_connector(connector); - struct intel_dp *intel_dp = intel_connector->mst_port; - - if (drm_connector_is_unregistered(connector)) - return connector_status_disconnected; - return drm_dp_mst_detect_port(connector, &intel_dp->mst_mgr, - intel_connector->port); -} - static const struct drm_connector_funcs intel_dp_mst_connector_funcs = { - .detect = intel_dp_mst_detect, .fill_modes = drm_helper_probe_single_connector_modes, .atomic_get_property = intel_digital_connector_atomic_get_property, .atomic_set_property = intel_digital_connector_atomic_set_property, @@ -426,6 +413,7 @@ static enum drm_mode_status intel_dp_mst_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) { + struct drm_i915_private *dev_priv = to_i915(connector->dev); struct intel_connector *intel_connector = to_intel_connector(connector); struct intel_dp *intel_dp = intel_connector->mst_port; int max_dotclk = to_i915(connector->dev)->max_dotclk_freq; @@ -453,7 +441,7 @@ intel_dp_mst_mode_valid(struct drm_connector *connector, if (mode_rate > max_rate || mode->clock > max_dotclk) return MODE_CLOCK_HIGH; - return MODE_OK; + return intel_mode_valid_max_plane_size(dev_priv, mode); } static struct drm_encoder *intel_mst_atomic_best_encoder(struct drm_connector *connector, @@ -466,11 +454,26 @@ static struct drm_encoder *intel_mst_atomic_best_encoder(struct drm_connector *c return &intel_dp->mst_encoders[crtc->pipe]->base.base; } +static int +intel_dp_mst_detect(struct drm_connector *connector, + struct drm_modeset_acquire_ctx *ctx, bool force) +{ + struct intel_connector *intel_connector = to_intel_connector(connector); + struct intel_dp *intel_dp = intel_connector->mst_port; + + if (drm_connector_is_unregistered(connector)) + return connector_status_disconnected; + + return drm_dp_mst_detect_port(connector, ctx, &intel_dp->mst_mgr, + intel_connector->port); +} + static const struct drm_connector_helper_funcs intel_dp_mst_connector_helper_funcs = { .get_modes = intel_dp_mst_get_modes, .mode_valid = intel_dp_mst_mode_valid, .atomic_best_encoder = intel_mst_atomic_best_encoder, .atomic_check = intel_dp_mst_atomic_check, + .detect_ctx = intel_dp_mst_detect, }; static void intel_dp_mst_encoder_destroy(struct drm_encoder *encoder) @@ -615,8 +618,16 @@ intel_dp_create_fake_mst_encoder(struct intel_digital_port *intel_dig_port, enum intel_encoder->type = INTEL_OUTPUT_DP_MST; intel_encoder->power_domain = intel_dig_port->base.power_domain; intel_encoder->port = intel_dig_port->base.port; - intel_encoder->crtc_mask = BIT(pipe); intel_encoder->cloneable = 0; + /* + * This is wrong, but broken userspace uses the intersection + * of possible_crtcs of all the encoders of a given connector + * to figure out which crtcs can drive said connector. What + * should be used instead is the union of possible_crtcs. + * To keep such userspace functioning we must misconfigure + * this to make sure the intersection is not empty :( + */ + intel_encoder->pipe_mask = ~0; intel_encoder->compute_config = intel_dp_mst_compute_config; intel_encoder->disable = intel_mst_disable_dp; @@ -653,21 +664,31 @@ intel_dp_mst_encoder_active_links(struct intel_digital_port *intel_dig_port) int intel_dp_mst_encoder_init(struct intel_digital_port *intel_dig_port, int conn_base_id) { + struct drm_i915_private *i915 = to_i915(intel_dig_port->base.base.dev); struct intel_dp *intel_dp = &intel_dig_port->dp; - struct drm_device *dev = intel_dig_port->base.base.dev; + enum port port = intel_dig_port->base.port; int ret; - intel_dp->can_mst = true; + if (!HAS_DP_MST(i915) || intel_dp_is_edp(intel_dp)) + return 0; + + if (INTEL_GEN(i915) < 12 && port == PORT_A) + return 0; + + if (INTEL_GEN(i915) < 11 && port == PORT_E) + return 0; + intel_dp->mst_mgr.cbs = &mst_cbs; /* create encoders */ intel_dp_create_fake_mst_encoders(intel_dig_port); - ret = drm_dp_mst_topology_mgr_init(&intel_dp->mst_mgr, dev, + ret = drm_dp_mst_topology_mgr_init(&intel_dp->mst_mgr, &i915->drm, &intel_dp->aux, 16, 3, conn_base_id); - if (ret) { - intel_dp->can_mst = false; + if (ret) return ret; - } + + intel_dp->can_mst = true; + return 0; } diff --git a/drivers/gpu/drm/i915/display/intel_dpll_mgr.c b/drivers/gpu/drm/i915/display/intel_dpll_mgr.c index b8148f838354..3ce0a023eee0 100644 --- a/drivers/gpu/drm/i915/display/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/display/intel_dpll_mgr.c @@ -247,8 +247,7 @@ static struct intel_shared_dpll * intel_find_shared_dpll(struct intel_atomic_state *state, const struct intel_crtc *crtc, const struct intel_dpll_hw_state *pll_state, - enum intel_dpll_id range_min, - enum intel_dpll_id range_max) + unsigned long dpll_mask) { struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); struct intel_shared_dpll *pll, *unused_pll = NULL; @@ -257,7 +256,9 @@ intel_find_shared_dpll(struct intel_atomic_state *state, shared_dpll = intel_atomic_get_shared_dpll_state(&state->base); - for (i = range_min; i <= range_max; i++) { + WARN_ON(dpll_mask & ~(BIT(I915_NUM_PLLS) - 1)); + + for_each_set_bit(i, &dpll_mask, I915_NUM_PLLS) { pll = &dev_priv->shared_dplls[i]; /* Only want to check enabled timings first */ @@ -464,8 +465,8 @@ static bool ibx_get_dpll(struct intel_atomic_state *state, } else { pll = intel_find_shared_dpll(state, crtc, &crtc_state->dpll_hw_state, - DPLL_ID_PCH_PLL_A, - DPLL_ID_PCH_PLL_B); + BIT(DPLL_ID_PCH_PLL_B) | + BIT(DPLL_ID_PCH_PLL_A)); } if (!pll) @@ -525,16 +526,31 @@ static void hsw_ddi_wrpll_disable(struct drm_i915_private *dev_priv, val = I915_READ(WRPLL_CTL(id)); I915_WRITE(WRPLL_CTL(id), val & ~WRPLL_PLL_ENABLE); POSTING_READ(WRPLL_CTL(id)); + + /* + * Try to set up the PCH reference clock once all DPLLs + * that depend on it have been shut down. + */ + if (dev_priv->pch_ssc_use & BIT(id)) + intel_init_pch_refclk(dev_priv); } static void hsw_ddi_spll_disable(struct drm_i915_private *dev_priv, struct intel_shared_dpll *pll) { + enum intel_dpll_id id = pll->info->id; u32 val; val = I915_READ(SPLL_CTL); I915_WRITE(SPLL_CTL, val & ~SPLL_PLL_ENABLE); POSTING_READ(SPLL_CTL); + + /* + * Try to set up the PCH reference clock once all DPLLs + * that depend on it have been shut down. + */ + if (dev_priv->pch_ssc_use & BIT(id)) + intel_init_pch_refclk(dev_priv); } static bool hsw_ddi_wrpll_get_hw_state(struct drm_i915_private *dev_priv, @@ -814,7 +830,8 @@ hsw_ddi_hdmi_get_dpll(struct intel_atomic_state *state, pll = intel_find_shared_dpll(state, crtc, &crtc_state->dpll_hw_state, - DPLL_ID_WRPLL1, DPLL_ID_WRPLL2); + BIT(DPLL_ID_WRPLL2) | + BIT(DPLL_ID_WRPLL1)); if (!pll) return NULL; @@ -877,7 +894,7 @@ static bool hsw_get_dpll(struct intel_atomic_state *state, pll = intel_find_shared_dpll(state, crtc, &crtc_state->dpll_hw_state, - DPLL_ID_SPLL, DPLL_ID_SPLL); + BIT(DPLL_ID_SPLL)); } else { return false; } @@ -1447,13 +1464,13 @@ static bool skl_get_dpll(struct intel_atomic_state *state, if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_EDP)) pll = intel_find_shared_dpll(state, crtc, &crtc_state->dpll_hw_state, - DPLL_ID_SKL_DPLL0, - DPLL_ID_SKL_DPLL0); + BIT(DPLL_ID_SKL_DPLL0)); else pll = intel_find_shared_dpll(state, crtc, &crtc_state->dpll_hw_state, - DPLL_ID_SKL_DPLL1, - DPLL_ID_SKL_DPLL3); + BIT(DPLL_ID_SKL_DPLL3) | + BIT(DPLL_ID_SKL_DPLL2) | + BIT(DPLL_ID_SKL_DPLL1)); if (!pll) return false; @@ -2401,8 +2418,9 @@ static bool cnl_get_dpll(struct intel_atomic_state *state, pll = intel_find_shared_dpll(state, crtc, &crtc_state->dpll_hw_state, - DPLL_ID_SKL_DPLL0, - DPLL_ID_SKL_DPLL2); + BIT(DPLL_ID_SKL_DPLL2) | + BIT(DPLL_ID_SKL_DPLL1) | + BIT(DPLL_ID_SKL_DPLL0)); if (!pll) { DRM_DEBUG_KMS("No PLL selected\n"); return false; @@ -2520,6 +2538,18 @@ static const struct skl_wrpll_params icl_tbt_pll_19_2MHz_values = { .pdiv = 0x4 /* 5 */, .kdiv = 1, .qdiv_mode = 0, .qdiv_ratio = 0, }; +static const struct skl_wrpll_params tgl_tbt_pll_19_2MHz_values = { + .dco_integer = 0x54, .dco_fraction = 0x3000, + /* the following params are unused */ + .pdiv = 0, .kdiv = 0, .qdiv_mode = 0, .qdiv_ratio = 0, +}; + +static const struct skl_wrpll_params tgl_tbt_pll_24MHz_values = { + .dco_integer = 0x43, .dco_fraction = 0x4000, + /* the following params are unused */ + .pdiv = 0, .kdiv = 0, .qdiv_mode = 0, .qdiv_ratio = 0, +}; + static bool icl_calc_dp_combo_pll(struct intel_crtc_state *crtc_state, struct skl_wrpll_params *pll_params) { @@ -2547,8 +2577,34 @@ static bool icl_calc_tbt_pll(struct intel_crtc_state *crtc_state, { struct drm_i915_private *dev_priv = to_i915(crtc_state->base.crtc->dev); - *pll_params = dev_priv->cdclk.hw.ref == 24000 ? - icl_tbt_pll_24MHz_values : icl_tbt_pll_19_2MHz_values; + if (INTEL_GEN(dev_priv) >= 12) { + switch (dev_priv->cdclk.hw.ref) { + default: + MISSING_CASE(dev_priv->cdclk.hw.ref); + /* fall-through */ + case 19200: + case 38400: + *pll_params = tgl_tbt_pll_19_2MHz_values; + break; + case 24000: + *pll_params = tgl_tbt_pll_24MHz_values; + break; + } + } else { + switch (dev_priv->cdclk.hw.ref) { + default: + MISSING_CASE(dev_priv->cdclk.hw.ref); + /* fall-through */ + case 19200: + case 38400: + *pll_params = icl_tbt_pll_19_2MHz_values; + break; + case 24000: + *pll_params = icl_tbt_pll_24MHz_values; + break; + } + } + return true; } @@ -2607,7 +2663,8 @@ enum intel_dpll_id icl_tc_port_to_pll_id(enum tc_port tc_port) static bool icl_mg_pll_find_divisors(int clock_khz, bool is_dp, bool use_ssc, u32 *target_dco_khz, - struct intel_dpll_hw_state *state) + struct intel_dpll_hw_state *state, + bool is_dkl) { u32 dco_min_freq, dco_max_freq; int div1_vals[] = {7, 5, 3, 2}; @@ -2629,8 +2686,13 @@ static bool icl_mg_pll_find_divisors(int clock_khz, bool is_dp, bool use_ssc, continue; if (div2 >= 2) { + /* + * Note: a_divratio not matching TGL BSpec + * algorithm but matching hardcoded values and + * working on HW for DP alt-mode at least + */ a_divratio = is_dp ? 10 : 5; - tlinedrv = 2; + tlinedrv = is_dkl ? 1 : 2; } else { a_divratio = 5; tlinedrv = 0; @@ -2693,11 +2755,12 @@ static bool icl_calc_mg_pll_state(struct intel_crtc_state *crtc_state, u64 tmp; bool use_ssc = false; bool is_dp = !intel_crtc_has_type(crtc_state, INTEL_OUTPUT_HDMI); + bool is_dkl = INTEL_GEN(dev_priv) >= 12; memset(pll_state, 0, sizeof(*pll_state)); if (!icl_mg_pll_find_divisors(clock, is_dp, use_ssc, &dco_khz, - pll_state)) { + pll_state, is_dkl)) { DRM_DEBUG_KMS("Failed to find divisors for clock %d\n", clock); return false; } @@ -2705,8 +2768,11 @@ static bool icl_calc_mg_pll_state(struct intel_crtc_state *crtc_state, m1div = 2; m2div_int = dco_khz / (refclk_khz * m1div); if (m2div_int > 255) { - m1div = 4; - m2div_int = dco_khz / (refclk_khz * m1div); + if (!is_dkl) { + m1div = 4; + m2div_int = dco_khz / (refclk_khz * m1div); + } + if (m2div_int > 255) { DRM_DEBUG_KMS("Failed to find mdiv for clock %d\n", clock); @@ -2786,60 +2852,94 @@ static bool icl_calc_mg_pll_state(struct intel_crtc_state *crtc_state, } ssc_steplog = 4; - pll_state->mg_pll_div0 = (m2div_rem > 0 ? MG_PLL_DIV0_FRACNEN_H : 0) | - MG_PLL_DIV0_FBDIV_FRAC(m2div_frac) | - MG_PLL_DIV0_FBDIV_INT(m2div_int); - - pll_state->mg_pll_div1 = MG_PLL_DIV1_IREF_NDIVRATIO(iref_ndiv) | - MG_PLL_DIV1_DITHER_DIV_2 | - MG_PLL_DIV1_NDIVRATIO(1) | - MG_PLL_DIV1_FBPREDIV(m1div); - - pll_state->mg_pll_lf = MG_PLL_LF_TDCTARGETCNT(tdc_targetcnt) | - MG_PLL_LF_AFCCNTSEL_512 | - MG_PLL_LF_GAINCTRL(1) | - MG_PLL_LF_INT_COEFF(int_coeff) | - MG_PLL_LF_PROP_COEFF(prop_coeff); - - pll_state->mg_pll_frac_lock = MG_PLL_FRAC_LOCK_TRUELOCK_CRIT_32 | - MG_PLL_FRAC_LOCK_EARLYLOCK_CRIT_32 | - MG_PLL_FRAC_LOCK_LOCKTHRESH(10) | - MG_PLL_FRAC_LOCK_DCODITHEREN | - MG_PLL_FRAC_LOCK_FEEDFWRDGAIN(feedfwgain); - if (use_ssc || m2div_rem > 0) - pll_state->mg_pll_frac_lock |= MG_PLL_FRAC_LOCK_FEEDFWRDCAL_EN; - - pll_state->mg_pll_ssc = (use_ssc ? MG_PLL_SSC_EN : 0) | - MG_PLL_SSC_TYPE(2) | - MG_PLL_SSC_STEPLENGTH(ssc_steplen) | - MG_PLL_SSC_STEPNUM(ssc_steplog) | - MG_PLL_SSC_FLLEN | - MG_PLL_SSC_STEPSIZE(ssc_stepsize); - - pll_state->mg_pll_tdc_coldst_bias = MG_PLL_TDC_COLDST_COLDSTART | - MG_PLL_TDC_COLDST_IREFINT_EN | - MG_PLL_TDC_COLDST_REFBIAS_START_PULSE_W(iref_pulse_w) | - MG_PLL_TDC_TDCOVCCORR_EN | - MG_PLL_TDC_TDCSEL(3); - - pll_state->mg_pll_bias = MG_PLL_BIAS_BIAS_GB_SEL(3) | - MG_PLL_BIAS_INIT_DCOAMP(0x3F) | - MG_PLL_BIAS_BIAS_BONUS(10) | - MG_PLL_BIAS_BIASCAL_EN | - MG_PLL_BIAS_CTRIM(12) | - MG_PLL_BIAS_VREF_RDAC(4) | - MG_PLL_BIAS_IREFTRIM(iref_trim); - - if (refclk_khz == 38400) { - pll_state->mg_pll_tdc_coldst_bias_mask = MG_PLL_TDC_COLDST_COLDSTART; - pll_state->mg_pll_bias_mask = 0; + /* write pll_state calculations */ + if (is_dkl) { + pll_state->mg_pll_div0 = DKL_PLL_DIV0_INTEG_COEFF(int_coeff) | + DKL_PLL_DIV0_PROP_COEFF(prop_coeff) | + DKL_PLL_DIV0_FBPREDIV(m1div) | + DKL_PLL_DIV0_FBDIV_INT(m2div_int); + + pll_state->mg_pll_div1 = DKL_PLL_DIV1_IREF_TRIM(iref_trim) | + DKL_PLL_DIV1_TDC_TARGET_CNT(tdc_targetcnt); + + pll_state->mg_pll_ssc = DKL_PLL_SSC_IREF_NDIV_RATIO(iref_ndiv) | + DKL_PLL_SSC_STEP_LEN(ssc_steplen) | + DKL_PLL_SSC_STEP_NUM(ssc_steplog) | + (use_ssc ? DKL_PLL_SSC_EN : 0); + + pll_state->mg_pll_bias = (m2div_frac ? DKL_PLL_BIAS_FRAC_EN_H : 0) | + DKL_PLL_BIAS_FBDIV_FRAC(m2div_frac); + + pll_state->mg_pll_tdc_coldst_bias = + DKL_PLL_TDC_SSC_STEP_SIZE(ssc_stepsize) | + DKL_PLL_TDC_FEED_FWD_GAIN(feedfwgain); + } else { - pll_state->mg_pll_tdc_coldst_bias_mask = -1U; - pll_state->mg_pll_bias_mask = -1U; - } + pll_state->mg_pll_div0 = + (m2div_rem > 0 ? MG_PLL_DIV0_FRACNEN_H : 0) | + MG_PLL_DIV0_FBDIV_FRAC(m2div_frac) | + MG_PLL_DIV0_FBDIV_INT(m2div_int); + + pll_state->mg_pll_div1 = + MG_PLL_DIV1_IREF_NDIVRATIO(iref_ndiv) | + MG_PLL_DIV1_DITHER_DIV_2 | + MG_PLL_DIV1_NDIVRATIO(1) | + MG_PLL_DIV1_FBPREDIV(m1div); + + pll_state->mg_pll_lf = + MG_PLL_LF_TDCTARGETCNT(tdc_targetcnt) | + MG_PLL_LF_AFCCNTSEL_512 | + MG_PLL_LF_GAINCTRL(1) | + MG_PLL_LF_INT_COEFF(int_coeff) | + MG_PLL_LF_PROP_COEFF(prop_coeff); + + pll_state->mg_pll_frac_lock = + MG_PLL_FRAC_LOCK_TRUELOCK_CRIT_32 | + MG_PLL_FRAC_LOCK_EARLYLOCK_CRIT_32 | + MG_PLL_FRAC_LOCK_LOCKTHRESH(10) | + MG_PLL_FRAC_LOCK_DCODITHEREN | + MG_PLL_FRAC_LOCK_FEEDFWRDGAIN(feedfwgain); + if (use_ssc || m2div_rem > 0) + pll_state->mg_pll_frac_lock |= + MG_PLL_FRAC_LOCK_FEEDFWRDCAL_EN; + + pll_state->mg_pll_ssc = + (use_ssc ? MG_PLL_SSC_EN : 0) | + MG_PLL_SSC_TYPE(2) | + MG_PLL_SSC_STEPLENGTH(ssc_steplen) | + MG_PLL_SSC_STEPNUM(ssc_steplog) | + MG_PLL_SSC_FLLEN | + MG_PLL_SSC_STEPSIZE(ssc_stepsize); + + pll_state->mg_pll_tdc_coldst_bias = + MG_PLL_TDC_COLDST_COLDSTART | + MG_PLL_TDC_COLDST_IREFINT_EN | + MG_PLL_TDC_COLDST_REFBIAS_START_PULSE_W(iref_pulse_w) | + MG_PLL_TDC_TDCOVCCORR_EN | + MG_PLL_TDC_TDCSEL(3); + + pll_state->mg_pll_bias = + MG_PLL_BIAS_BIAS_GB_SEL(3) | + MG_PLL_BIAS_INIT_DCOAMP(0x3F) | + MG_PLL_BIAS_BIAS_BONUS(10) | + MG_PLL_BIAS_BIASCAL_EN | + MG_PLL_BIAS_CTRIM(12) | + MG_PLL_BIAS_VREF_RDAC(4) | + MG_PLL_BIAS_IREFTRIM(iref_trim); + + if (refclk_khz == 38400) { + pll_state->mg_pll_tdc_coldst_bias_mask = + MG_PLL_TDC_COLDST_COLDSTART; + pll_state->mg_pll_bias_mask = 0; + } else { + pll_state->mg_pll_tdc_coldst_bias_mask = -1U; + pll_state->mg_pll_bias_mask = -1U; + } - pll_state->mg_pll_tdc_coldst_bias &= pll_state->mg_pll_tdc_coldst_bias_mask; - pll_state->mg_pll_bias &= pll_state->mg_pll_bias_mask; + pll_state->mg_pll_tdc_coldst_bias &= + pll_state->mg_pll_tdc_coldst_bias_mask; + pll_state->mg_pll_bias &= pll_state->mg_pll_bias_mask; + } return true; } @@ -2893,7 +2993,7 @@ static bool icl_get_combo_phy_dpll(struct intel_atomic_state *state, &crtc_state->icl_port_dplls[ICL_PORT_DPLL_DEFAULT]; struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); enum port port = encoder->port; - bool has_dpll4 = false; + unsigned long dpll_mask; if (!icl_calc_dpll_state(crtc_state, encoder, &port_dpll->hw_state)) { DRM_DEBUG_KMS("Could not calculate combo PHY PLL state.\n"); @@ -2902,16 +3002,19 @@ static bool icl_get_combo_phy_dpll(struct intel_atomic_state *state, } if (IS_ELKHARTLAKE(dev_priv) && port != PORT_A) - has_dpll4 = true; + dpll_mask = + BIT(DPLL_ID_EHL_DPLL4) | + BIT(DPLL_ID_ICL_DPLL1) | + BIT(DPLL_ID_ICL_DPLL0); + else + dpll_mask = BIT(DPLL_ID_ICL_DPLL1) | BIT(DPLL_ID_ICL_DPLL0); port_dpll->pll = intel_find_shared_dpll(state, crtc, &port_dpll->hw_state, - DPLL_ID_ICL_DPLL0, - has_dpll4 ? DPLL_ID_EHL_DPLL4 - : DPLL_ID_ICL_DPLL1); + dpll_mask); if (!port_dpll->pll) { - DRM_DEBUG_KMS("No combo PHY PLL found for port %c\n", - port_name(encoder->port)); + DRM_DEBUG_KMS("No combo PHY PLL found for [ENCODER:%d:%s]\n", + encoder->base.base.id, encoder->base.name); return false; } @@ -2941,8 +3044,7 @@ static bool icl_get_tc_phy_dplls(struct intel_atomic_state *state, port_dpll->pll = intel_find_shared_dpll(state, crtc, &port_dpll->hw_state, - DPLL_ID_ICL_TBTPLL, - DPLL_ID_ICL_TBTPLL); + BIT(DPLL_ID_ICL_TBTPLL)); if (!port_dpll->pll) { DRM_DEBUG_KMS("No TBT-ALT PLL found\n"); return false; @@ -2961,8 +3063,7 @@ static bool icl_get_tc_phy_dplls(struct intel_atomic_state *state, encoder->port)); port_dpll->pll = intel_find_shared_dpll(state, crtc, &port_dpll->hw_state, - dpll_id, - dpll_id); + BIT(dpll_id)); if (!port_dpll->pll) { DRM_DEBUG_KMS("No MG PHY PLL found\n"); goto err_unreference_tbt_pll; @@ -3086,6 +3187,78 @@ out: return ret; } +static bool dkl_pll_get_hw_state(struct drm_i915_private *dev_priv, + struct intel_shared_dpll *pll, + struct intel_dpll_hw_state *hw_state) +{ + const enum intel_dpll_id id = pll->info->id; + enum tc_port tc_port = icl_pll_id_to_tc_port(id); + intel_wakeref_t wakeref; + bool ret = false; + u32 val; + + wakeref = intel_display_power_get_if_enabled(dev_priv, + POWER_DOMAIN_DISPLAY_CORE); + if (!wakeref) + return false; + + val = I915_READ(MG_PLL_ENABLE(tc_port)); + if (!(val & PLL_ENABLE)) + goto out; + + /* + * All registers read here have the same HIP_INDEX_REG even though + * they are on different building blocks + */ + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, 0x2)); + + hw_state->mg_refclkin_ctl = I915_READ(DKL_REFCLKIN_CTL(tc_port)); + hw_state->mg_refclkin_ctl &= MG_REFCLKIN_CTL_OD_2_MUX_MASK; + + hw_state->mg_clktop2_hsclkctl = + I915_READ(DKL_CLKTOP2_HSCLKCTL(tc_port)); + hw_state->mg_clktop2_hsclkctl &= + MG_CLKTOP2_HSCLKCTL_TLINEDRV_CLKSEL_MASK | + MG_CLKTOP2_HSCLKCTL_CORE_INPUTSEL_MASK | + MG_CLKTOP2_HSCLKCTL_HSDIV_RATIO_MASK | + MG_CLKTOP2_HSCLKCTL_DSDIV_RATIO_MASK; + + hw_state->mg_clktop2_coreclkctl1 = + I915_READ(DKL_CLKTOP2_CORECLKCTL1(tc_port)); + hw_state->mg_clktop2_coreclkctl1 &= + MG_CLKTOP2_CORECLKCTL1_A_DIVRATIO_MASK; + + hw_state->mg_pll_div0 = I915_READ(DKL_PLL_DIV0(tc_port)); + hw_state->mg_pll_div0 &= (DKL_PLL_DIV0_INTEG_COEFF_MASK | + DKL_PLL_DIV0_PROP_COEFF_MASK | + DKL_PLL_DIV0_FBPREDIV_MASK | + DKL_PLL_DIV0_FBDIV_INT_MASK); + + hw_state->mg_pll_div1 = I915_READ(DKL_PLL_DIV1(tc_port)); + hw_state->mg_pll_div1 &= (DKL_PLL_DIV1_IREF_TRIM_MASK | + DKL_PLL_DIV1_TDC_TARGET_CNT_MASK); + + hw_state->mg_pll_ssc = I915_READ(DKL_PLL_SSC(tc_port)); + hw_state->mg_pll_ssc &= (DKL_PLL_SSC_IREF_NDIV_RATIO_MASK | + DKL_PLL_SSC_STEP_LEN_MASK | + DKL_PLL_SSC_STEP_NUM_MASK | + DKL_PLL_SSC_EN); + + hw_state->mg_pll_bias = I915_READ(DKL_PLL_BIAS(tc_port)); + hw_state->mg_pll_bias &= (DKL_PLL_BIAS_FRAC_EN_H | + DKL_PLL_BIAS_FBDIV_FRAC_MASK); + + hw_state->mg_pll_tdc_coldst_bias = + I915_READ(DKL_PLL_TDC_COLDST_BIAS(tc_port)); + hw_state->mg_pll_tdc_coldst_bias &= (DKL_PLL_TDC_SSC_STEP_SIZE_MASK | + DKL_PLL_TDC_FEED_FWD_GAIN_MASK); + + ret = true; +out: + intel_display_power_put(dev_priv, POWER_DOMAIN_DISPLAY_CORE, wakeref); + return ret; +} + static bool icl_pll_get_hw_state(struct drm_i915_private *dev_priv, struct intel_shared_dpll *pll, struct intel_dpll_hw_state *hw_state, @@ -3220,6 +3393,75 @@ static void icl_mg_pll_write(struct drm_i915_private *dev_priv, POSTING_READ(MG_PLL_TDC_COLDST_BIAS(tc_port)); } +static void dkl_pll_write(struct drm_i915_private *dev_priv, + struct intel_shared_dpll *pll) +{ + struct intel_dpll_hw_state *hw_state = &pll->state.hw_state; + enum tc_port tc_port = icl_pll_id_to_tc_port(pll->info->id); + u32 val; + + /* + * All registers programmed here have the same HIP_INDEX_REG even + * though on different building block + */ + I915_WRITE(HIP_INDEX_REG(tc_port), HIP_INDEX_VAL(tc_port, 0x2)); + + /* All the registers are RMW */ + val = I915_READ(DKL_REFCLKIN_CTL(tc_port)); + val &= ~MG_REFCLKIN_CTL_OD_2_MUX_MASK; + val |= hw_state->mg_refclkin_ctl; + I915_WRITE(DKL_REFCLKIN_CTL(tc_port), val); + + val = I915_READ(DKL_CLKTOP2_CORECLKCTL1(tc_port)); + val &= ~MG_CLKTOP2_CORECLKCTL1_A_DIVRATIO_MASK; + val |= hw_state->mg_clktop2_coreclkctl1; + I915_WRITE(DKL_CLKTOP2_CORECLKCTL1(tc_port), val); + + val = I915_READ(DKL_CLKTOP2_HSCLKCTL(tc_port)); + val &= ~(MG_CLKTOP2_HSCLKCTL_TLINEDRV_CLKSEL_MASK | + MG_CLKTOP2_HSCLKCTL_CORE_INPUTSEL_MASK | + MG_CLKTOP2_HSCLKCTL_HSDIV_RATIO_MASK | + MG_CLKTOP2_HSCLKCTL_DSDIV_RATIO_MASK); + val |= hw_state->mg_clktop2_hsclkctl; + I915_WRITE(DKL_CLKTOP2_HSCLKCTL(tc_port), val); + + val = I915_READ(DKL_PLL_DIV0(tc_port)); + val &= ~(DKL_PLL_DIV0_INTEG_COEFF_MASK | + DKL_PLL_DIV0_PROP_COEFF_MASK | + DKL_PLL_DIV0_FBPREDIV_MASK | + DKL_PLL_DIV0_FBDIV_INT_MASK); + val |= hw_state->mg_pll_div0; + I915_WRITE(DKL_PLL_DIV0(tc_port), val); + + val = I915_READ(DKL_PLL_DIV1(tc_port)); + val &= ~(DKL_PLL_DIV1_IREF_TRIM_MASK | + DKL_PLL_DIV1_TDC_TARGET_CNT_MASK); + val |= hw_state->mg_pll_div1; + I915_WRITE(DKL_PLL_DIV1(tc_port), val); + + val = I915_READ(DKL_PLL_SSC(tc_port)); + val &= ~(DKL_PLL_SSC_IREF_NDIV_RATIO_MASK | + DKL_PLL_SSC_STEP_LEN_MASK | + DKL_PLL_SSC_STEP_NUM_MASK | + DKL_PLL_SSC_EN); + val |= hw_state->mg_pll_ssc; + I915_WRITE(DKL_PLL_SSC(tc_port), val); + + val = I915_READ(DKL_PLL_BIAS(tc_port)); + val &= ~(DKL_PLL_BIAS_FRAC_EN_H | + DKL_PLL_BIAS_FBDIV_FRAC_MASK); + val |= hw_state->mg_pll_bias; + I915_WRITE(DKL_PLL_BIAS(tc_port), val); + + val = I915_READ(DKL_PLL_TDC_COLDST_BIAS(tc_port)); + val &= ~(DKL_PLL_TDC_SSC_STEP_SIZE_MASK | + DKL_PLL_TDC_FEED_FWD_GAIN_MASK); + val |= hw_state->mg_pll_tdc_coldst_bias; + I915_WRITE(DKL_PLL_TDC_COLDST_BIAS(tc_port), val); + + POSTING_READ(DKL_PLL_TDC_COLDST_BIAS(tc_port)); +} + static void icl_pll_power_enable(struct drm_i915_private *dev_priv, struct intel_shared_dpll *pll, i915_reg_t enable_reg) @@ -3312,7 +3554,10 @@ static void mg_pll_enable(struct drm_i915_private *dev_priv, icl_pll_power_enable(dev_priv, pll, enable_reg); - icl_mg_pll_write(dev_priv, pll); + if (INTEL_GEN(dev_priv) >= 12) + dkl_pll_write(dev_priv, pll); + else + icl_mg_pll_write(dev_priv, pll); /* * DVFS pre sequence would be here, but in our driver the cdclk code @@ -3467,11 +3712,22 @@ static const struct intel_dpll_mgr ehl_pll_mgr = { .dump_hw_state = icl_dump_hw_state, }; +static const struct intel_shared_dpll_funcs dkl_pll_funcs = { + .enable = mg_pll_enable, + .disable = mg_pll_disable, + .get_hw_state = dkl_pll_get_hw_state, +}; + static const struct dpll_info tgl_plls[] = { { "DPLL 0", &combo_pll_funcs, DPLL_ID_ICL_DPLL0, 0 }, { "DPLL 1", &combo_pll_funcs, DPLL_ID_ICL_DPLL1, 0 }, { "TBT PLL", &tbt_pll_funcs, DPLL_ID_ICL_TBTPLL, 0 }, - /* TODO: Add typeC plls */ + { "TC PLL 1", &dkl_pll_funcs, DPLL_ID_ICL_MGPLL1, 0 }, + { "TC PLL 2", &dkl_pll_funcs, DPLL_ID_ICL_MGPLL2, 0 }, + { "TC PLL 3", &dkl_pll_funcs, DPLL_ID_ICL_MGPLL3, 0 }, + { "TC PLL 4", &dkl_pll_funcs, DPLL_ID_ICL_MGPLL4, 0 }, + { "TC PLL 5", &dkl_pll_funcs, DPLL_ID_TGL_MGPLL5, 0 }, + { "TC PLL 6", &dkl_pll_funcs, DPLL_ID_TGL_MGPLL6, 0 }, { }, }; @@ -3479,6 +3735,7 @@ static const struct intel_dpll_mgr tgl_pll_mgr = { .dpll_info = tgl_plls, .get_dplls = icl_get_dplls, .put_dplls = icl_put_dplls, + .update_active_dpll = icl_update_active_dpll, .dump_hw_state = icl_dump_hw_state, }; diff --git a/drivers/gpu/drm/i915/display/intel_dpll_mgr.h b/drivers/gpu/drm/i915/display/intel_dpll_mgr.h index e7588799fce5..2a104c64291d 100644 --- a/drivers/gpu/drm/i915/display/intel_dpll_mgr.h +++ b/drivers/gpu/drm/i915/display/intel_dpll_mgr.h @@ -147,11 +147,11 @@ enum intel_dpll_id { */ DPLL_ID_ICL_MGPLL4 = 6, /** - * @DPLL_ID_TGL_TCPLL5: TGL TC PLL port 5 (TC5) + * @DPLL_ID_TGL_MGPLL5: TGL TC PLL port 5 (TC5) */ DPLL_ID_TGL_MGPLL5 = 7, /** - * @DPLL_ID_TGL_TCPLL6: TGL TC PLL port 6 (TC6) + * @DPLL_ID_TGL_MGPLL6: TGL TC PLL port 6 (TC6) */ DPLL_ID_TGL_MGPLL6 = 8, }; @@ -337,6 +337,11 @@ struct intel_shared_dpll { * @info: platform specific info */ const struct dpll_info *info; + + /** + * @wakeref: In some platforms a device-level runtime pm reference may + * need to be grabbed to disable DC states while this DPLL is enabled + */ intel_wakeref_t wakeref; }; diff --git a/drivers/gpu/drm/i915/display/intel_dsb.c b/drivers/gpu/drm/i915/display/intel_dsb.c new file mode 100644 index 000000000000..bb5a0e91b370 --- /dev/null +++ b/drivers/gpu/drm/i915/display/intel_dsb.c @@ -0,0 +1,332 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + * + */ + +#include "i915_drv.h" +#include "intel_display_types.h" + +#define DSB_BUF_SIZE (2 * PAGE_SIZE) + +/** + * DOC: DSB + * + * A DSB (Display State Buffer) is a queue of MMIO instructions in the memory + * which can be offloaded to DSB HW in Display Controller. DSB HW is a DMA + * engine that can be programmed to download the DSB from memory. + * It allows driver to batch submit display HW programming. This helps to + * reduce loading time and CPU activity, thereby making the context switch + * faster. DSB Support added from Gen12 Intel graphics based platform. + * + * DSB's can access only the pipe, plane, and transcoder Data Island Packet + * registers. + * + * DSB HW can support only register writes (both indexed and direct MMIO + * writes). There are no registers reads possible with DSB HW engine. + */ + +/* DSB opcodes. */ +#define DSB_OPCODE_SHIFT 24 +#define DSB_OPCODE_MMIO_WRITE 0x1 +#define DSB_OPCODE_INDEXED_WRITE 0x9 +#define DSB_BYTE_EN 0xF +#define DSB_BYTE_EN_SHIFT 20 +#define DSB_REG_VALUE_MASK 0xfffff + +static inline bool is_dsb_busy(struct intel_dsb *dsb) +{ + struct intel_crtc *crtc = container_of(dsb, typeof(*crtc), dsb); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + enum pipe pipe = crtc->pipe; + + return DSB_STATUS & I915_READ(DSB_CTRL(pipe, dsb->id)); +} + +static inline bool intel_dsb_enable_engine(struct intel_dsb *dsb) +{ + struct intel_crtc *crtc = container_of(dsb, typeof(*crtc), dsb); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + enum pipe pipe = crtc->pipe; + u32 dsb_ctrl; + + dsb_ctrl = I915_READ(DSB_CTRL(pipe, dsb->id)); + if (DSB_STATUS & dsb_ctrl) { + DRM_DEBUG_KMS("DSB engine is busy.\n"); + return false; + } + + dsb_ctrl |= DSB_ENABLE; + I915_WRITE(DSB_CTRL(pipe, dsb->id), dsb_ctrl); + + POSTING_READ(DSB_CTRL(pipe, dsb->id)); + return true; +} + +static inline bool intel_dsb_disable_engine(struct intel_dsb *dsb) +{ + struct intel_crtc *crtc = container_of(dsb, typeof(*crtc), dsb); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + enum pipe pipe = crtc->pipe; + u32 dsb_ctrl; + + dsb_ctrl = I915_READ(DSB_CTRL(pipe, dsb->id)); + if (DSB_STATUS & dsb_ctrl) { + DRM_DEBUG_KMS("DSB engine is busy.\n"); + return false; + } + + dsb_ctrl &= ~DSB_ENABLE; + I915_WRITE(DSB_CTRL(pipe, dsb->id), dsb_ctrl); + + POSTING_READ(DSB_CTRL(pipe, dsb->id)); + return true; +} + +/** + * intel_dsb_get() - Allocate DSB context and return a DSB instance. + * @crtc: intel_crtc structure to get pipe info. + * + * This function provides handle of a DSB instance, for the further DSB + * operations. + * + * Returns: address of Intel_dsb instance requested for. + * Failure: Returns the same DSB instance, but without a command buffer. + */ + +struct intel_dsb * +intel_dsb_get(struct intel_crtc *crtc) +{ + struct drm_device *dev = crtc->base.dev; + struct drm_i915_private *i915 = to_i915(dev); + struct intel_dsb *dsb = &crtc->dsb; + struct drm_i915_gem_object *obj; + struct i915_vma *vma; + intel_wakeref_t wakeref; + + if (!HAS_DSB(i915)) + return dsb; + + if (atomic_add_return(1, &dsb->refcount) != 1) + return dsb; + + dsb->id = DSB1; + wakeref = intel_runtime_pm_get(&i915->runtime_pm); + + obj = i915_gem_object_create_internal(i915, DSB_BUF_SIZE); + if (IS_ERR(obj)) { + DRM_ERROR("Gem object creation failed\n"); + goto err; + } + + vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE); + if (IS_ERR(vma)) { + DRM_ERROR("Vma creation failed\n"); + i915_gem_object_put(obj); + atomic_dec(&dsb->refcount); + goto err; + } + + dsb->cmd_buf = i915_gem_object_pin_map(vma->obj, I915_MAP_WC); + if (IS_ERR(dsb->cmd_buf)) { + DRM_ERROR("Command buffer creation failed\n"); + i915_vma_unpin_and_release(&vma, 0); + dsb->cmd_buf = NULL; + atomic_dec(&dsb->refcount); + goto err; + } + dsb->vma = vma; + +err: + intel_runtime_pm_put(&i915->runtime_pm, wakeref); + return dsb; +} + +/** + * intel_dsb_put() - To destroy DSB context. + * @dsb: intel_dsb structure. + * + * This function destroys the DSB context allocated by a dsb_get(), by + * unpinning and releasing the VMA object associated with it. + */ + +void intel_dsb_put(struct intel_dsb *dsb) +{ + struct intel_crtc *crtc = container_of(dsb, typeof(*crtc), dsb); + struct drm_i915_private *i915 = to_i915(crtc->base.dev); + + if (!HAS_DSB(i915)) + return; + + if (WARN_ON(atomic_read(&dsb->refcount) == 0)) + return; + + if (atomic_dec_and_test(&dsb->refcount)) { + i915_vma_unpin_and_release(&dsb->vma, I915_VMA_RELEASE_MAP); + dsb->cmd_buf = NULL; + dsb->free_pos = 0; + dsb->ins_start_offset = 0; + } +} + +/** + * intel_dsb_indexed_reg_write() -Write to the DSB context for auto + * increment register. + * @dsb: intel_dsb structure. + * @reg: register address. + * @val: value. + * + * This function is used for writing register-value pair in command + * buffer of DSB for auto-increment register. During command buffer overflow, + * a warning is thrown and rest all erroneous condition register programming + * is done through mmio write. + */ + +void intel_dsb_indexed_reg_write(struct intel_dsb *dsb, i915_reg_t reg, + u32 val) +{ + struct intel_crtc *crtc = container_of(dsb, typeof(*crtc), dsb); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + u32 *buf = dsb->cmd_buf; + u32 reg_val; + + if (!buf) { + I915_WRITE(reg, val); + return; + } + + if (WARN_ON(dsb->free_pos >= DSB_BUF_SIZE)) { + DRM_DEBUG_KMS("DSB buffer overflow\n"); + return; + } + + /* + * For example the buffer will look like below for 3 dwords for auto + * increment register: + * +--------------------------------------------------------+ + * | size = 3 | offset &| value1 | value2 | value3 | zero | + * | | opcode | | | | | + * +--------------------------------------------------------+ + * + + + + + + + + * 0 4 8 12 16 20 24 + * Byte + * + * As every instruction is 8 byte aligned the index of dsb instruction + * will start always from even number while dealing with u32 array. If + * we are writing odd no of dwords, Zeros will be added in the end for + * padding. + */ + reg_val = buf[dsb->ins_start_offset + 1] & DSB_REG_VALUE_MASK; + if (reg_val != i915_mmio_reg_offset(reg)) { + /* Every instruction should be 8 byte aligned. */ + dsb->free_pos = ALIGN(dsb->free_pos, 2); + + dsb->ins_start_offset = dsb->free_pos; + + /* Update the size. */ + buf[dsb->free_pos++] = 1; + + /* Update the opcode and reg. */ + buf[dsb->free_pos++] = (DSB_OPCODE_INDEXED_WRITE << + DSB_OPCODE_SHIFT) | + i915_mmio_reg_offset(reg); + + /* Update the value. */ + buf[dsb->free_pos++] = val; + } else { + /* Update the new value. */ + buf[dsb->free_pos++] = val; + + /* Update the size. */ + buf[dsb->ins_start_offset]++; + } + + /* if number of data words is odd, then the last dword should be 0.*/ + if (dsb->free_pos & 0x1) + buf[dsb->free_pos] = 0; +} + +/** + * intel_dsb_reg_write() -Write to the DSB context for normal + * register. + * @dsb: intel_dsb structure. + * @reg: register address. + * @val: value. + * + * This function is used for writing register-value pair in command + * buffer of DSB. During command buffer overflow, a warning is thrown + * and rest all erroneous condition register programming is done + * through mmio write. + */ +void intel_dsb_reg_write(struct intel_dsb *dsb, i915_reg_t reg, u32 val) +{ + struct intel_crtc *crtc = container_of(dsb, typeof(*crtc), dsb); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + u32 *buf = dsb->cmd_buf; + + if (!buf) { + I915_WRITE(reg, val); + return; + } + + if (WARN_ON(dsb->free_pos >= DSB_BUF_SIZE)) { + DRM_DEBUG_KMS("DSB buffer overflow\n"); + return; + } + + dsb->ins_start_offset = dsb->free_pos; + buf[dsb->free_pos++] = val; + buf[dsb->free_pos++] = (DSB_OPCODE_MMIO_WRITE << DSB_OPCODE_SHIFT) | + (DSB_BYTE_EN << DSB_BYTE_EN_SHIFT) | + i915_mmio_reg_offset(reg); +} + +/** + * intel_dsb_commit() - Trigger workload execution of DSB. + * @dsb: intel_dsb structure. + * + * This function is used to do actual write to hardware using DSB. + * On errors, fall back to MMIO. Also this function help to reset the context. + */ +void intel_dsb_commit(struct intel_dsb *dsb) +{ + struct intel_crtc *crtc = container_of(dsb, typeof(*crtc), dsb); + struct drm_device *dev = crtc->base.dev; + struct drm_i915_private *dev_priv = to_i915(dev); + enum pipe pipe = crtc->pipe; + u32 tail; + + if (!dsb->free_pos) + return; + + if (!intel_dsb_enable_engine(dsb)) + goto reset; + + if (is_dsb_busy(dsb)) { + DRM_ERROR("HEAD_PTR write failed - dsb engine is busy.\n"); + goto reset; + } + I915_WRITE(DSB_HEAD(pipe, dsb->id), i915_ggtt_offset(dsb->vma)); + + tail = ALIGN(dsb->free_pos * 4, CACHELINE_BYTES); + if (tail > dsb->free_pos * 4) + memset(&dsb->cmd_buf[dsb->free_pos], 0, + (tail - dsb->free_pos * 4)); + + if (is_dsb_busy(dsb)) { + DRM_ERROR("TAIL_PTR write failed - dsb engine is busy.\n"); + goto reset; + } + DRM_DEBUG_KMS("DSB execution started - head 0x%x, tail 0x%x\n", + i915_ggtt_offset(dsb->vma), tail); + I915_WRITE(DSB_TAIL(pipe, dsb->id), i915_ggtt_offset(dsb->vma) + tail); + if (wait_for(!is_dsb_busy(dsb), 1)) { + DRM_ERROR("Timed out waiting for DSB workload completion.\n"); + goto reset; + } + +reset: + dsb->free_pos = 0; + dsb->ins_start_offset = 0; + intel_dsb_disable_engine(dsb); +} diff --git a/drivers/gpu/drm/i915/display/intel_dsb.h b/drivers/gpu/drm/i915/display/intel_dsb.h new file mode 100644 index 000000000000..6f95c8e909e6 --- /dev/null +++ b/drivers/gpu/drm/i915/display/intel_dsb.h @@ -0,0 +1,52 @@ +/* SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef _INTEL_DSB_H +#define _INTEL_DSB_H + +#include <linux/types.h> + +#include "i915_reg.h" + +struct intel_crtc; +struct i915_vma; + +enum dsb_id { + INVALID_DSB = -1, + DSB1, + DSB2, + DSB3, + MAX_DSB_PER_PIPE +}; + +struct intel_dsb { + atomic_t refcount; + enum dsb_id id; + u32 *cmd_buf; + struct i915_vma *vma; + + /* + * free_pos will point the first free entry position + * and help in calculating tail of command buffer. + */ + int free_pos; + + /* + * ins_start_offset will help to store start address of the dsb + * instuction and help in identifying the batch of auto-increment + * register. + */ + u32 ins_start_offset; +}; + +struct intel_dsb * +intel_dsb_get(struct intel_crtc *crtc); +void intel_dsb_put(struct intel_dsb *dsb); +void intel_dsb_reg_write(struct intel_dsb *dsb, i915_reg_t reg, u32 val); +void intel_dsb_indexed_reg_write(struct intel_dsb *dsb, i915_reg_t reg, + u32 val); +void intel_dsb_commit(struct intel_dsb *dsb); + +#endif diff --git a/drivers/gpu/drm/i915/display/intel_dsi.c b/drivers/gpu/drm/i915/display/intel_dsi.c index 5fec02aceaed..a2a937109a5a 100644 --- a/drivers/gpu/drm/i915/display/intel_dsi.c +++ b/drivers/gpu/drm/i915/display/intel_dsi.c @@ -55,6 +55,7 @@ int intel_dsi_get_modes(struct drm_connector *connector) enum drm_mode_status intel_dsi_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) { + struct drm_i915_private *dev_priv = to_i915(connector->dev); struct intel_connector *intel_connector = to_intel_connector(connector); const struct drm_display_mode *fixed_mode = intel_connector->panel.fixed_mode; int max_dotclk = to_i915(connector->dev)->max_dotclk_freq; @@ -73,7 +74,7 @@ enum drm_mode_status intel_dsi_mode_valid(struct drm_connector *connector, return MODE_CLOCK_HIGH; } - return MODE_OK; + return intel_mode_valid_max_plane_size(dev_priv, mode); } struct intel_dsi_host *intel_dsi_host_init(struct intel_dsi *intel_dsi, diff --git a/drivers/gpu/drm/i915/display/intel_dvo.c b/drivers/gpu/drm/i915/display/intel_dvo.c index 93baf366692e..bcfbcb743e7d 100644 --- a/drivers/gpu/drm/i915/display/intel_dvo.c +++ b/drivers/gpu/drm/i915/display/intel_dvo.c @@ -280,7 +280,7 @@ static void intel_dvo_pre_enable(struct intel_encoder *encoder, struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); const struct drm_display_mode *adjusted_mode = &pipe_config->base.adjusted_mode; struct intel_dvo *intel_dvo = enc_to_dvo(encoder); - int pipe = crtc->pipe; + enum pipe pipe = crtc->pipe; u32 dvo_val; i915_reg_t dvo_reg = intel_dvo->dev.dvo_reg; i915_reg_t dvo_srcdim_reg = intel_dvo->dev.dvo_srcdim_reg; @@ -505,7 +505,7 @@ void intel_dvo_init(struct drm_i915_private *dev_priv) intel_encoder->type = INTEL_OUTPUT_DVO; intel_encoder->power_domain = POWER_DOMAIN_PORT_OTHER; intel_encoder->port = port; - intel_encoder->crtc_mask = (1 << 0) | (1 << 1); + intel_encoder->pipe_mask = ~0; switch (dvo->type) { case INTEL_DVO_CHIP_TMDS: diff --git a/drivers/gpu/drm/i915/display/intel_fbc.c b/drivers/gpu/drm/i915/display/intel_fbc.c index 16ed44bfd734..3111ecaeabd0 100644 --- a/drivers/gpu/drm/i915/display/intel_fbc.c +++ b/drivers/gpu/drm/i915/display/intel_fbc.c @@ -343,8 +343,8 @@ static void gen7_fbc_activate(struct drm_i915_private *dev_priv) HSW_FBCQ_DIS); } - if (IS_GEN(dev_priv, 11)) - /* Wa_1409120013:icl,ehl */ + if (INTEL_GEN(dev_priv) >= 11) + /* Wa_1409120013:icl,ehl,tgl */ I915_WRITE(ILK_DPFC_CHICKEN, ILK_DPFC_CHICKEN_COMP_DUMMY_PIXEL); I915_WRITE(ILK_DPFC_CONTROL, dpfc_ctl | DPFC_CTL_EN); @@ -1320,6 +1320,9 @@ void intel_fbc_init(struct drm_i915_private *dev_priv) fbc->enabled = false; fbc->active = false; + if (!drm_mm_initialized(&dev_priv->mm.stolen)) + mkwrite_device_info(dev_priv)->display.has_fbc = false; + if (need_fbc_vtd_wa(dev_priv)) mkwrite_device_info(dev_priv)->display.has_fbc = false; diff --git a/drivers/gpu/drm/i915/display/intel_fbdev.c b/drivers/gpu/drm/i915/display/intel_fbdev.c index d59eee5c5d9c..48c960ca12fb 100644 --- a/drivers/gpu/drm/i915/display/intel_fbdev.c +++ b/drivers/gpu/drm/i915/display/intel_fbdev.c @@ -141,10 +141,10 @@ static int intelfb_alloc(struct drm_fb_helper *helper, /* If the FB is too big, just don't use it since fbdev is not very * important and we should probably use that space with FBC or other * features. */ - obj = NULL; + obj = ERR_PTR(-ENODEV); if (size * 2 < dev_priv->stolen_usable_size) obj = i915_gem_object_create_stolen(dev_priv, size); - if (obj == NULL) + if (IS_ERR(obj)) obj = i915_gem_object_create_shmem(dev_priv, size); if (IS_ERR(obj)) { DRM_ERROR("failed to allocate framebuffer\n"); @@ -204,7 +204,6 @@ static int intelfb_create(struct drm_fb_helper *helper, sizes->fb_height = intel_fb->base.height; } - mutex_lock(&dev->struct_mutex); wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); /* Pin the GGTT vma for our access via info->screen_base. @@ -235,6 +234,11 @@ static int intelfb_create(struct drm_fb_helper *helper, info->apertures->ranges[0].base = ggtt->gmadr.start; info->apertures->ranges[0].size = ggtt->mappable_end; + /* Our framebuffer is the entirety of fbdev's system memory */ + info->fix.smem_start = + (unsigned long)(ggtt->gmadr.start + vma->node.start); + info->fix.smem_len = vma->node.size; + vaddr = i915_vma_pin_iomap(vma); if (IS_ERR(vaddr)) { DRM_ERROR("Failed to remap framebuffer into virtual memory\n"); @@ -244,10 +248,6 @@ static int intelfb_create(struct drm_fb_helper *helper, info->screen_base = vaddr; info->screen_size = vma->node.size; - /* Our framebuffer is the entirety of fbdev's system memory */ - info->fix.smem_start = (unsigned long)info->screen_base; - info->fix.smem_len = info->screen_size; - drm_fb_helper_fill_info(info, &ifbdev->helper, sizes); /* If the object is shmemfs backed, it will have given us zeroed pages. @@ -266,7 +266,6 @@ static int intelfb_create(struct drm_fb_helper *helper, ifbdev->vma_flags = flags; intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); - mutex_unlock(&dev->struct_mutex); vga_switcheroo_client_fb_set(pdev, info); return 0; @@ -274,7 +273,6 @@ out_unpin: intel_unpin_fb_vma(vma, flags); out_unlock: intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); - mutex_unlock(&dev->struct_mutex); return ret; } @@ -291,11 +289,8 @@ static void intel_fbdev_destroy(struct intel_fbdev *ifbdev) drm_fb_helper_fini(&ifbdev->helper); - if (ifbdev->vma) { - mutex_lock(&ifbdev->helper.dev->struct_mutex); + if (ifbdev->vma) intel_unpin_fb_vma(ifbdev->vma, ifbdev->vma_flags); - mutex_unlock(&ifbdev->helper.dev->struct_mutex); - } if (ifbdev->fb) drm_framebuffer_remove(&ifbdev->fb->base); @@ -444,7 +439,7 @@ int intel_fbdev_init(struct drm_device *dev) struct intel_fbdev *ifbdev; int ret; - if (WARN_ON(!HAS_DISPLAY(dev_priv))) + if (WARN_ON(!HAS_DISPLAY(dev_priv) || !INTEL_DISPLAY_ENABLED(dev_priv))) return -ENODEV; ifbdev = kzalloc(sizeof(struct intel_fbdev), GFP_KERNEL); diff --git a/drivers/gpu/drm/i915/display/intel_frontbuffer.c b/drivers/gpu/drm/i915/display/intel_frontbuffer.c index 719379774fa5..84b164f31895 100644 --- a/drivers/gpu/drm/i915/display/intel_frontbuffer.c +++ b/drivers/gpu/drm/i915/display/intel_frontbuffer.c @@ -206,6 +206,7 @@ static int frontbuffer_active(struct i915_active *ref) return 0; } +__i915_active_call static void frontbuffer_retire(struct i915_active *ref) { struct intel_frontbuffer *front = @@ -220,11 +221,18 @@ static void frontbuffer_release(struct kref *ref) { struct intel_frontbuffer *front = container_of(ref, typeof(*front), ref); + struct drm_i915_gem_object *obj = front->obj; + struct i915_vma *vma; - front->obj->frontbuffer = NULL; - spin_unlock(&to_i915(front->obj->base.dev)->fb_tracking.lock); + spin_lock(&obj->vma.lock); + for_each_ggtt_vma(vma, obj) + vma->display_alignment = I915_GTT_MIN_ALIGNMENT; + spin_unlock(&obj->vma.lock); - i915_gem_object_put(front->obj); + obj->frontbuffer = NULL; + spin_unlock(&to_i915(obj->base.dev)->fb_tracking.lock); + + i915_gem_object_put(obj); kfree(front); } @@ -249,8 +257,9 @@ intel_frontbuffer_get(struct drm_i915_gem_object *obj) front->obj = obj; kref_init(&front->ref); atomic_set(&front->bits, 0); - i915_active_init(i915, &front->write, - frontbuffer_active, frontbuffer_retire); + i915_active_init(&front->write, + frontbuffer_active, + i915_active_may_sleep(frontbuffer_retire)); spin_lock(&i915->fb_tracking.lock); if (obj->frontbuffer) { diff --git a/drivers/gpu/drm/i915/display/intel_gmbus.c b/drivers/gpu/drm/i915/display/intel_gmbus.c index d6775a005726..3d4d19ac1d14 100644 --- a/drivers/gpu/drm/i915/display/intel_gmbus.c +++ b/drivers/gpu/drm/i915/display/intel_gmbus.c @@ -836,7 +836,7 @@ int intel_gmbus_setup(struct drm_i915_private *dev_priv) unsigned int pin; int ret; - if (!HAS_DISPLAY(dev_priv)) + if (!HAS_DISPLAY(dev_priv) || !INTEL_DISPLAY_ENABLED(dev_priv)) return 0; if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) diff --git a/drivers/gpu/drm/i915/display/intel_hdcp.c b/drivers/gpu/drm/i915/display/intel_hdcp.c index 6ec5ceeab601..f1f41ca8402b 100644 --- a/drivers/gpu/drm/i915/display/intel_hdcp.c +++ b/drivers/gpu/drm/i915/display/intel_hdcp.c @@ -1,9 +1,11 @@ /* SPDX-License-Identifier: MIT */ /* * Copyright (C) 2017 Google, Inc. + * Copyright _ 2017-2019, Intel Corporation. * * Authors: * Sean Paul <seanpaul@chromium.org> + * Ramalingam C <ramalingam.c@intel.com> */ #include <linux/component.h> @@ -18,6 +20,7 @@ #include "intel_display_types.h" #include "intel_hdcp.h" #include "intel_sideband.h" +#include "intel_connector.h" #define KEY_LOAD_TRIES 5 #define ENCRYPT_STATUS_CHANGE_TIMEOUT_MS 50 @@ -105,24 +108,20 @@ bool intel_hdcp2_capable(struct intel_connector *connector) return capable; } -static inline bool intel_hdcp_in_use(struct intel_connector *connector) +static inline +bool intel_hdcp_in_use(struct drm_i915_private *dev_priv, + enum transcoder cpu_transcoder, enum port port) { - struct drm_i915_private *dev_priv = to_i915(connector->base.dev); - enum port port = connector->encoder->port; - u32 reg; - - reg = I915_READ(PORT_HDCP_STATUS(port)); - return reg & HDCP_STATUS_ENC; + return I915_READ(HDCP_STATUS(dev_priv, cpu_transcoder, port)) & + HDCP_STATUS_ENC; } -static inline bool intel_hdcp2_in_use(struct intel_connector *connector) +static inline +bool intel_hdcp2_in_use(struct drm_i915_private *dev_priv, + enum transcoder cpu_transcoder, enum port port) { - struct drm_i915_private *dev_priv = to_i915(connector->base.dev); - enum port port = connector->encoder->port; - u32 reg; - - reg = I915_READ(HDCP2_STATUS_DDI(port)); - return reg & LINK_ENCRYPTION_STATUS; + return I915_READ(HDCP2_STATUS(dev_priv, cpu_transcoder, port)) & + LINK_ENCRYPTION_STATUS; } static int intel_hdcp_poll_ksv_fifo(struct intel_digital_port *intel_dig_port, @@ -253,9 +252,29 @@ static int intel_write_sha_text(struct drm_i915_private *dev_priv, u32 sha_text) } static -u32 intel_hdcp_get_repeater_ctl(struct intel_digital_port *intel_dig_port) +u32 intel_hdcp_get_repeater_ctl(struct drm_i915_private *dev_priv, + enum transcoder cpu_transcoder, enum port port) { - enum port port = intel_dig_port->base.port; + if (INTEL_GEN(dev_priv) >= 12) { + switch (cpu_transcoder) { + case TRANSCODER_A: + return HDCP_TRANSA_REP_PRESENT | + HDCP_TRANSA_SHA1_M0; + case TRANSCODER_B: + return HDCP_TRANSB_REP_PRESENT | + HDCP_TRANSB_SHA1_M0; + case TRANSCODER_C: + return HDCP_TRANSC_REP_PRESENT | + HDCP_TRANSC_SHA1_M0; + case TRANSCODER_D: + return HDCP_TRANSD_REP_PRESENT | + HDCP_TRANSD_SHA1_M0; + default: + DRM_ERROR("Unknown transcoder %d\n", cpu_transcoder); + return -EINVAL; + } + } + switch (port) { case PORT_A: return HDCP_DDIA_REP_PRESENT | HDCP_DDIA_SHA1_M0; @@ -268,18 +287,20 @@ u32 intel_hdcp_get_repeater_ctl(struct intel_digital_port *intel_dig_port) case PORT_E: return HDCP_DDIE_REP_PRESENT | HDCP_DDIE_SHA1_M0; default: - break; + DRM_ERROR("Unknown port %d\n", port); + return -EINVAL; } - DRM_ERROR("Unknown port %d\n", port); - return -EINVAL; } static -int intel_hdcp_validate_v_prime(struct intel_digital_port *intel_dig_port, +int intel_hdcp_validate_v_prime(struct intel_connector *connector, const struct intel_hdcp_shim *shim, u8 *ksv_fifo, u8 num_downstream, u8 *bstatus) { + struct intel_digital_port *intel_dig_port = conn_to_dig_port(connector); struct drm_i915_private *dev_priv; + enum transcoder cpu_transcoder = connector->hdcp.cpu_transcoder; + enum port port = intel_dig_port->base.port; u32 vprime, sha_text, sha_leftovers, rep_ctl; int ret, i, j, sha_idx; @@ -306,7 +327,7 @@ int intel_hdcp_validate_v_prime(struct intel_digital_port *intel_dig_port, sha_idx = 0; sha_text = 0; sha_leftovers = 0; - rep_ctl = intel_hdcp_get_repeater_ctl(intel_dig_port); + rep_ctl = intel_hdcp_get_repeater_ctl(dev_priv, cpu_transcoder, port); I915_WRITE(HDCP_REP_CTL, rep_ctl | HDCP_SHA1_TEXT_32); for (i = 0; i < num_downstream; i++) { unsigned int sha_empty; @@ -548,7 +569,7 @@ int intel_hdcp_auth_downstream(struct intel_connector *connector) * V prime atleast twice. */ for (i = 0; i < tries; i++) { - ret = intel_hdcp_validate_v_prime(intel_dig_port, shim, + ret = intel_hdcp_validate_v_prime(connector, shim, ksv_fifo, num_downstream, bstatus); if (!ret) @@ -576,6 +597,7 @@ static int intel_hdcp_auth(struct intel_connector *connector) struct drm_device *dev = connector->base.dev; const struct intel_hdcp_shim *shim = hdcp->shim; struct drm_i915_private *dev_priv; + enum transcoder cpu_transcoder = connector->hdcp.cpu_transcoder; enum port port; unsigned long r0_prime_gen_start; int ret, i, tries = 2; @@ -615,18 +637,21 @@ static int intel_hdcp_auth(struct intel_connector *connector) /* Initialize An with 2 random values and acquire it */ for (i = 0; i < 2; i++) - I915_WRITE(PORT_HDCP_ANINIT(port), get_random_u32()); - I915_WRITE(PORT_HDCP_CONF(port), HDCP_CONF_CAPTURE_AN); + I915_WRITE(HDCP_ANINIT(dev_priv, cpu_transcoder, port), + get_random_u32()); + I915_WRITE(HDCP_CONF(dev_priv, cpu_transcoder, port), + HDCP_CONF_CAPTURE_AN); /* Wait for An to be acquired */ - if (intel_de_wait_for_set(dev_priv, PORT_HDCP_STATUS(port), + if (intel_de_wait_for_set(dev_priv, + HDCP_STATUS(dev_priv, cpu_transcoder, port), HDCP_STATUS_AN_READY, 1)) { DRM_ERROR("Timed out waiting for An\n"); return -ETIMEDOUT; } - an.reg[0] = I915_READ(PORT_HDCP_ANLO(port)); - an.reg[1] = I915_READ(PORT_HDCP_ANHI(port)); + an.reg[0] = I915_READ(HDCP_ANLO(dev_priv, cpu_transcoder, port)); + an.reg[1] = I915_READ(HDCP_ANHI(dev_priv, cpu_transcoder, port)); ret = shim->write_an_aksv(intel_dig_port, an.shim); if (ret) return ret; @@ -644,24 +669,26 @@ static int intel_hdcp_auth(struct intel_connector *connector) return -EPERM; } - I915_WRITE(PORT_HDCP_BKSVLO(port), bksv.reg[0]); - I915_WRITE(PORT_HDCP_BKSVHI(port), bksv.reg[1]); + I915_WRITE(HDCP_BKSVLO(dev_priv, cpu_transcoder, port), bksv.reg[0]); + I915_WRITE(HDCP_BKSVHI(dev_priv, cpu_transcoder, port), bksv.reg[1]); ret = shim->repeater_present(intel_dig_port, &repeater_present); if (ret) return ret; if (repeater_present) I915_WRITE(HDCP_REP_CTL, - intel_hdcp_get_repeater_ctl(intel_dig_port)); + intel_hdcp_get_repeater_ctl(dev_priv, cpu_transcoder, + port)); ret = shim->toggle_signalling(intel_dig_port, true); if (ret) return ret; - I915_WRITE(PORT_HDCP_CONF(port), HDCP_CONF_AUTH_AND_ENC); + I915_WRITE(HDCP_CONF(dev_priv, cpu_transcoder, port), + HDCP_CONF_AUTH_AND_ENC); /* Wait for R0 ready */ - if (wait_for(I915_READ(PORT_HDCP_STATUS(port)) & + if (wait_for(I915_READ(HDCP_STATUS(dev_priv, cpu_transcoder, port)) & (HDCP_STATUS_R0_READY | HDCP_STATUS_ENC), 1)) { DRM_ERROR("Timed out waiting for R0 ready\n"); return -ETIMEDOUT; @@ -689,22 +716,25 @@ static int intel_hdcp_auth(struct intel_connector *connector) ret = shim->read_ri_prime(intel_dig_port, ri.shim); if (ret) return ret; - I915_WRITE(PORT_HDCP_RPRIME(port), ri.reg); + I915_WRITE(HDCP_RPRIME(dev_priv, cpu_transcoder, port), ri.reg); /* Wait for Ri prime match */ - if (!wait_for(I915_READ(PORT_HDCP_STATUS(port)) & + if (!wait_for(I915_READ(HDCP_STATUS(dev_priv, cpu_transcoder, + port)) & (HDCP_STATUS_RI_MATCH | HDCP_STATUS_ENC), 1)) break; } if (i == tries) { DRM_DEBUG_KMS("Timed out waiting for Ri prime match (%x)\n", - I915_READ(PORT_HDCP_STATUS(port))); + I915_READ(HDCP_STATUS(dev_priv, cpu_transcoder, + port))); return -ETIMEDOUT; } /* Wait for encryption confirmation */ - if (intel_de_wait_for_set(dev_priv, PORT_HDCP_STATUS(port), + if (intel_de_wait_for_set(dev_priv, + HDCP_STATUS(dev_priv, cpu_transcoder, port), HDCP_STATUS_ENC, ENCRYPT_STATUS_CHANGE_TIMEOUT_MS)) { DRM_ERROR("Timed out waiting for encryption\n"); @@ -729,15 +759,17 @@ static int _intel_hdcp_disable(struct intel_connector *connector) struct drm_i915_private *dev_priv = connector->base.dev->dev_private; struct intel_digital_port *intel_dig_port = conn_to_dig_port(connector); enum port port = intel_dig_port->base.port; + enum transcoder cpu_transcoder = hdcp->cpu_transcoder; int ret; DRM_DEBUG_KMS("[%s:%d] HDCP is being disabled...\n", connector->base.name, connector->base.base.id); hdcp->hdcp_encrypted = false; - I915_WRITE(PORT_HDCP_CONF(port), 0); - if (intel_de_wait_for_clear(dev_priv, PORT_HDCP_STATUS(port), ~0, - ENCRYPT_STATUS_CHANGE_TIMEOUT_MS)) { + I915_WRITE(HDCP_CONF(dev_priv, cpu_transcoder, port), 0); + if (intel_de_wait_for_clear(dev_priv, + HDCP_STATUS(dev_priv, cpu_transcoder, port), + ~0, ENCRYPT_STATUS_CHANGE_TIMEOUT_MS)) { DRM_ERROR("Failed to disable HDCP, timeout clearing status\n"); return -ETIMEDOUT; } @@ -808,9 +840,11 @@ static int intel_hdcp_check_link(struct intel_connector *connector) struct drm_i915_private *dev_priv = connector->base.dev->dev_private; struct intel_digital_port *intel_dig_port = conn_to_dig_port(connector); enum port port = intel_dig_port->base.port; + enum transcoder cpu_transcoder; int ret = 0; mutex_lock(&hdcp->mutex); + cpu_transcoder = hdcp->cpu_transcoder; /* Check_link valid only when HDCP1.4 is enabled */ if (hdcp->value != DRM_MODE_CONTENT_PROTECTION_ENABLED || @@ -819,10 +853,11 @@ static int intel_hdcp_check_link(struct intel_connector *connector) goto out; } - if (WARN_ON(!intel_hdcp_in_use(connector))) { + if (WARN_ON(!intel_hdcp_in_use(dev_priv, cpu_transcoder, port))) { DRM_ERROR("%s:%d HDCP link stopped encryption,%x\n", connector->base.name, connector->base.base.id, - I915_READ(PORT_HDCP_STATUS(port))); + I915_READ(HDCP_STATUS(dev_priv, cpu_transcoder, + port))); ret = -ENXIO; hdcp->value = DRM_MODE_CONTENT_PROTECTION_DESIRED; schedule_work(&hdcp->prop_work); @@ -887,7 +922,7 @@ static void intel_hdcp_prop_work(struct work_struct *work) bool is_hdcp_supported(struct drm_i915_private *dev_priv, enum port port) { /* PORT E doesn't have HDCP, and PORT F is disabled */ - return INTEL_GEN(dev_priv) >= 9 && port < PORT_E; + return INTEL_INFO(dev_priv)->display.has_hdcp && port < PORT_E; } static int @@ -1493,10 +1528,11 @@ static int hdcp2_enable_encryption(struct intel_connector *connector) struct drm_i915_private *dev_priv = to_i915(connector->base.dev); struct intel_hdcp *hdcp = &connector->hdcp; enum port port = connector->encoder->port; + enum transcoder cpu_transcoder = hdcp->cpu_transcoder; int ret; - WARN_ON(I915_READ(HDCP2_STATUS_DDI(port)) & LINK_ENCRYPTION_STATUS); - + WARN_ON(I915_READ(HDCP2_STATUS(dev_priv, cpu_transcoder, port)) & + LINK_ENCRYPTION_STATUS); if (hdcp->shim->toggle_signalling) { ret = hdcp->shim->toggle_signalling(intel_dig_port, true); if (ret) { @@ -1506,14 +1542,18 @@ static int hdcp2_enable_encryption(struct intel_connector *connector) } } - if (I915_READ(HDCP2_STATUS_DDI(port)) & LINK_AUTH_STATUS) { + if (I915_READ(HDCP2_STATUS(dev_priv, cpu_transcoder, port)) & + LINK_AUTH_STATUS) { /* Link is Authenticated. Now set for Encryption */ - I915_WRITE(HDCP2_CTL_DDI(port), - I915_READ(HDCP2_CTL_DDI(port)) | + I915_WRITE(HDCP2_CTL(dev_priv, cpu_transcoder, port), + I915_READ(HDCP2_CTL(dev_priv, cpu_transcoder, + port)) | CTL_LINK_ENCRYPTION_REQ); } - ret = intel_de_wait_for_set(dev_priv, HDCP2_STATUS_DDI(port), + ret = intel_de_wait_for_set(dev_priv, + HDCP2_STATUS(dev_priv, cpu_transcoder, + port), LINK_ENCRYPTION_STATUS, ENCRYPT_STATUS_CHANGE_TIMEOUT_MS); @@ -1526,14 +1566,19 @@ static int hdcp2_disable_encryption(struct intel_connector *connector) struct drm_i915_private *dev_priv = to_i915(connector->base.dev); struct intel_hdcp *hdcp = &connector->hdcp; enum port port = connector->encoder->port; + enum transcoder cpu_transcoder = hdcp->cpu_transcoder; int ret; - WARN_ON(!(I915_READ(HDCP2_STATUS_DDI(port)) & LINK_ENCRYPTION_STATUS)); + WARN_ON(!(I915_READ(HDCP2_STATUS(dev_priv, cpu_transcoder, port)) & + LINK_ENCRYPTION_STATUS)); - I915_WRITE(HDCP2_CTL_DDI(port), - I915_READ(HDCP2_CTL_DDI(port)) & ~CTL_LINK_ENCRYPTION_REQ); + I915_WRITE(HDCP2_CTL(dev_priv, cpu_transcoder, port), + I915_READ(HDCP2_CTL(dev_priv, cpu_transcoder, port)) & + ~CTL_LINK_ENCRYPTION_REQ); - ret = intel_de_wait_for_clear(dev_priv, HDCP2_STATUS_DDI(port), + ret = intel_de_wait_for_clear(dev_priv, + HDCP2_STATUS(dev_priv, cpu_transcoder, + port), LINK_ENCRYPTION_STATUS, ENCRYPT_STATUS_CHANGE_TIMEOUT_MS); if (ret == -ETIMEDOUT) @@ -1632,9 +1677,11 @@ static int intel_hdcp2_check_link(struct intel_connector *connector) struct drm_i915_private *dev_priv = to_i915(connector->base.dev); struct intel_hdcp *hdcp = &connector->hdcp; enum port port = connector->encoder->port; + enum transcoder cpu_transcoder; int ret = 0; mutex_lock(&hdcp->mutex); + cpu_transcoder = hdcp->cpu_transcoder; /* hdcp2_check_link is expected only when HDCP2.2 is Enabled */ if (hdcp->value != DRM_MODE_CONTENT_PROTECTION_ENABLED || @@ -1643,9 +1690,10 @@ static int intel_hdcp2_check_link(struct intel_connector *connector) goto out; } - if (WARN_ON(!intel_hdcp2_in_use(connector))) { + if (WARN_ON(!intel_hdcp2_in_use(dev_priv, cpu_transcoder, port))) { DRM_ERROR("HDCP2.2 link stopped the encryption, %x\n", - I915_READ(HDCP2_STATUS_DDI(port))); + I915_READ(HDCP2_STATUS(dev_priv, cpu_transcoder, + port))); ret = -ENXIO; hdcp->value = DRM_MODE_CONTENT_PROTECTION_DESIRED; schedule_work(&hdcp->prop_work); @@ -1749,13 +1797,71 @@ static const struct component_ops i915_hdcp_component_ops = { .unbind = i915_hdcp_component_unbind, }; +static inline +enum mei_fw_ddi intel_get_mei_fw_ddi_index(enum port port) +{ + switch (port) { + case PORT_A: + return MEI_DDI_A; + case PORT_B ... PORT_F: + return (enum mei_fw_ddi)port; + default: + return MEI_DDI_INVALID_PORT; + } +} + +static inline +enum mei_fw_tc intel_get_mei_fw_tc(enum transcoder cpu_transcoder) +{ + switch (cpu_transcoder) { + case TRANSCODER_A ... TRANSCODER_D: + return (enum mei_fw_tc)(cpu_transcoder | 0x10); + default: /* eDP, DSI TRANSCODERS are non HDCP capable */ + return MEI_INVALID_TRANSCODER; + } +} + +void intel_hdcp_transcoder_config(struct intel_connector *connector, + enum transcoder cpu_transcoder) +{ + struct drm_i915_private *dev_priv = to_i915(connector->base.dev); + struct intel_hdcp *hdcp = &connector->hdcp; + + if (!hdcp->shim) + return; + + if (INTEL_GEN(dev_priv) >= 12) { + mutex_lock(&hdcp->mutex); + hdcp->cpu_transcoder = cpu_transcoder; + hdcp->port_data.fw_tc = intel_get_mei_fw_tc(cpu_transcoder); + mutex_unlock(&hdcp->mutex); + } +} + static inline int initialize_hdcp_port_data(struct intel_connector *connector, const struct intel_hdcp_shim *shim) { + struct drm_i915_private *dev_priv = to_i915(connector->base.dev); struct intel_hdcp *hdcp = &connector->hdcp; struct hdcp_port_data *data = &hdcp->port_data; - data->port = connector->encoder->port; + if (INTEL_GEN(dev_priv) < 12) + data->fw_ddi = + intel_get_mei_fw_ddi_index(connector->encoder->port); + else + /* + * As per ME FW API expectation, for GEN 12+, fw_ddi is filled + * with zero(INVALID PORT index). + */ + data->fw_ddi = MEI_DDI_INVALID_PORT; + + /* + * As associated transcoder is set and modified at modeset, here fw_tc + * is initialized to zero (invalid transcoder index). This will be + * retained for <Gen12 forever. + */ + data->fw_tc = MEI_INVALID_TRANSCODER; + data->port_type = (u8)HDCP_PORT_TYPE_INTEGRATED; data->protocol = (u8)shim->protocol; diff --git a/drivers/gpu/drm/i915/display/intel_hdcp.h b/drivers/gpu/drm/i915/display/intel_hdcp.h index 13555b054930..41c1053d9e38 100644 --- a/drivers/gpu/drm/i915/display/intel_hdcp.h +++ b/drivers/gpu/drm/i915/display/intel_hdcp.h @@ -15,10 +15,14 @@ struct drm_connector_state; struct drm_i915_private; struct intel_connector; struct intel_hdcp_shim; +enum port; +enum transcoder; void intel_hdcp_atomic_check(struct drm_connector *connector, struct drm_connector_state *old_state, struct drm_connector_state *new_state); +void intel_hdcp_transcoder_config(struct intel_connector *connector, + enum transcoder cpu_transcoder); int intel_hdcp_init(struct intel_connector *connector, const struct intel_hdcp_shim *hdcp_shim); int intel_hdcp_enable(struct intel_connector *connector, u8 content_type); diff --git a/drivers/gpu/drm/i915/display/intel_hdmi.c b/drivers/gpu/drm/i915/display/intel_hdmi.c index e02f0faecf02..f6f5312205c4 100644 --- a/drivers/gpu/drm/i915/display/intel_hdmi.c +++ b/drivers/gpu/drm/i915/display/intel_hdmi.c @@ -189,13 +189,19 @@ hsw_dip_data_reg(struct drm_i915_private *dev_priv, } } -static int hsw_dip_data_size(unsigned int type) +static int hsw_dip_data_size(struct drm_i915_private *dev_priv, + unsigned int type) { switch (type) { case DP_SDP_VSC: return VIDEO_DIP_VSC_DATA_SIZE; case DP_SDP_PPS: return VIDEO_DIP_PPS_DATA_SIZE; + case HDMI_PACKET_TYPE_GAMUT_METADATA: + if (INTEL_GEN(dev_priv) >= 11) + return VIDEO_DIP_GMP_DATA_SIZE; + else + return VIDEO_DIP_DATA_SIZE; default: return VIDEO_DIP_DATA_SIZE; } @@ -514,7 +520,9 @@ static void hsw_write_infoframe(struct intel_encoder *encoder, int i; u32 val = I915_READ(ctl_reg); - data_size = hsw_dip_data_size(type); + data_size = hsw_dip_data_size(dev_priv, type); + + WARN_ON(len > data_size); val &= ~hsw_infoframe_enable(type); I915_WRITE(ctl_reg, val); @@ -724,11 +732,20 @@ intel_hdmi_compute_avi_infoframe(struct intel_encoder *encoder, drm_hdmi_avi_infoframe_colorspace(frame, conn_state); - drm_hdmi_avi_infoframe_quant_range(frame, connector, - adjusted_mode, - crtc_state->limited_color_range ? - HDMI_QUANTIZATION_RANGE_LIMITED : - HDMI_QUANTIZATION_RANGE_FULL); + /* nonsense combination */ + WARN_ON(crtc_state->limited_color_range && + crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB); + + if (crtc_state->output_format == INTEL_OUTPUT_FORMAT_RGB) { + drm_hdmi_avi_infoframe_quant_range(frame, connector, + adjusted_mode, + crtc_state->limited_color_range ? + HDMI_QUANTIZATION_RANGE_LIMITED : + HDMI_QUANTIZATION_RANGE_FULL); + } else { + frame->quantization_range = HDMI_QUANTIZATION_RANGE_DEFAULT; + frame->ycc_quantization_range = HDMI_YCC_QUANTIZATION_RANGE_LIMITED; + } drm_hdmi_avi_infoframe_content_type(frame, conn_state); @@ -1491,7 +1508,10 @@ bool intel_hdmi_hdcp_check_link(struct intel_digital_port *intel_dig_port) { struct drm_i915_private *dev_priv = intel_dig_port->base.base.dev->dev_private; + struct intel_connector *connector = + intel_dig_port->hdmi.attached_connector; enum port port = intel_dig_port->base.port; + enum transcoder cpu_transcoder = connector->hdcp.cpu_transcoder; int ret; union { u32 reg; @@ -1502,39 +1522,30 @@ bool intel_hdmi_hdcp_check_link(struct intel_digital_port *intel_dig_port) if (ret) return false; - I915_WRITE(PORT_HDCP_RPRIME(port), ri.reg); + I915_WRITE(HDCP_RPRIME(dev_priv, cpu_transcoder, port), ri.reg); /* Wait for Ri prime match */ - if (wait_for(I915_READ(PORT_HDCP_STATUS(port)) & + if (wait_for(I915_READ(HDCP_STATUS(dev_priv, cpu_transcoder, port)) & (HDCP_STATUS_RI_MATCH | HDCP_STATUS_ENC), 1)) { DRM_ERROR("Ri' mismatch detected, link check failed (%x)\n", - I915_READ(PORT_HDCP_STATUS(port))); + I915_READ(HDCP_STATUS(dev_priv, cpu_transcoder, + port))); return false; } return true; } -struct hdcp2_hdmi_msg_data { +struct hdcp2_hdmi_msg_timeout { u8 msg_id; - u32 timeout; - u32 timeout2; + u16 timeout; }; -static const struct hdcp2_hdmi_msg_data hdcp2_msg_data[] = { - { HDCP_2_2_AKE_INIT, 0, 0 }, - { HDCP_2_2_AKE_SEND_CERT, HDCP_2_2_CERT_TIMEOUT_MS, 0 }, - { HDCP_2_2_AKE_NO_STORED_KM, 0, 0 }, - { HDCP_2_2_AKE_STORED_KM, 0, 0 }, - { HDCP_2_2_AKE_SEND_HPRIME, HDCP_2_2_HPRIME_PAIRED_TIMEOUT_MS, - HDCP_2_2_HPRIME_NO_PAIRED_TIMEOUT_MS }, - { HDCP_2_2_AKE_SEND_PAIRING_INFO, HDCP_2_2_PAIRING_TIMEOUT_MS, 0 }, - { HDCP_2_2_LC_INIT, 0, 0 }, - { HDCP_2_2_LC_SEND_LPRIME, HDCP_2_2_HDMI_LPRIME_TIMEOUT_MS, 0 }, - { HDCP_2_2_SKE_SEND_EKS, 0, 0 }, - { HDCP_2_2_REP_SEND_RECVID_LIST, HDCP_2_2_RECVID_LIST_TIMEOUT_MS, 0 }, - { HDCP_2_2_REP_SEND_ACK, 0, 0 }, - { HDCP_2_2_REP_STREAM_MANAGE, 0, 0 }, - { HDCP_2_2_REP_STREAM_READY, HDCP_2_2_STREAM_READY_TIMEOUT_MS, 0 }, +static const struct hdcp2_hdmi_msg_timeout hdcp2_msg_timeout[] = { + { HDCP_2_2_AKE_SEND_CERT, HDCP_2_2_CERT_TIMEOUT_MS, }, + { HDCP_2_2_AKE_SEND_PAIRING_INFO, HDCP_2_2_PAIRING_TIMEOUT_MS, }, + { HDCP_2_2_LC_SEND_LPRIME, HDCP_2_2_HDMI_LPRIME_TIMEOUT_MS, }, + { HDCP_2_2_REP_SEND_RECVID_LIST, HDCP_2_2_RECVID_LIST_TIMEOUT_MS, }, + { HDCP_2_2_REP_STREAM_READY, HDCP_2_2_STREAM_READY_TIMEOUT_MS, }, }; static @@ -1551,12 +1562,17 @@ static int get_hdcp2_msg_timeout(u8 msg_id, bool is_paired) { int i; - for (i = 0; i < ARRAY_SIZE(hdcp2_msg_data); i++) - if (hdcp2_msg_data[i].msg_id == msg_id && - (msg_id != HDCP_2_2_AKE_SEND_HPRIME || is_paired)) - return hdcp2_msg_data[i].timeout; - else if (hdcp2_msg_data[i].msg_id == msg_id) - return hdcp2_msg_data[i].timeout2; + if (msg_id == HDCP_2_2_AKE_SEND_HPRIME) { + if (is_paired) + return HDCP_2_2_HPRIME_PAIRED_TIMEOUT_MS; + else + return HDCP_2_2_HPRIME_NO_PAIRED_TIMEOUT_MS; + } + + for (i = 0; i < ARRAY_SIZE(hdcp2_msg_timeout); i++) { + if (hdcp2_msg_timeout[i].msg_id == msg_id) + return hdcp2_msg_timeout[i].timeout; + } return -EINVAL; } @@ -2184,8 +2200,10 @@ intel_hdmi_mode_valid(struct drm_connector *connector, status = hdmi_port_clock_valid(hdmi, clock * 5 / 4, true, force_dvi); } + if (status != MODE_OK) + return status; - return status; + return intel_mode_valid_max_plane_size(dev_priv, mode); } static bool hdmi_deep_color_possible(const struct intel_crtc_state *crtc_state, @@ -2261,9 +2279,7 @@ static bool hdmi_deep_color_possible(const struct intel_crtc_state *crtc_state, static bool intel_hdmi_ycbcr420_config(struct drm_connector *connector, - struct intel_crtc_state *config, - int *clock_12bpc, int *clock_10bpc, - int *clock_8bpc) + struct intel_crtc_state *config) { struct intel_crtc *intel_crtc = to_intel_crtc(config->base.crtc); @@ -2272,11 +2288,6 @@ intel_hdmi_ycbcr420_config(struct drm_connector *connector, return false; } - /* YCBCR420 TMDS rate requirement is half the pixel clock */ - config->port_clock /= 2; - *clock_12bpc /= 2; - *clock_10bpc /= 2; - *clock_8bpc /= 2; config->output_format = INTEL_OUTPUT_FORMAT_YCBCR420; /* YCBCR 420 output conversion needs a scaler */ @@ -2291,6 +2302,104 @@ intel_hdmi_ycbcr420_config(struct drm_connector *connector, return true; } +static int intel_hdmi_port_clock(int clock, int bpc) +{ + /* + * Need to adjust the port link by: + * 1.5x for 12bpc + * 1.25x for 10bpc + */ + return clock * bpc / 8; +} + +static int intel_hdmi_compute_bpc(struct intel_encoder *encoder, + struct intel_crtc_state *crtc_state, + int clock, bool force_dvi) +{ + struct intel_hdmi *intel_hdmi = enc_to_intel_hdmi(&encoder->base); + int bpc; + + for (bpc = 12; bpc >= 10; bpc -= 2) { + if (hdmi_deep_color_possible(crtc_state, bpc) && + hdmi_port_clock_valid(intel_hdmi, + intel_hdmi_port_clock(clock, bpc), + true, force_dvi) == MODE_OK) + return bpc; + } + + return 8; +} + +static int intel_hdmi_compute_clock(struct intel_encoder *encoder, + struct intel_crtc_state *crtc_state, + bool force_dvi) +{ + struct intel_hdmi *intel_hdmi = enc_to_intel_hdmi(&encoder->base); + const struct drm_display_mode *adjusted_mode = + &crtc_state->base.adjusted_mode; + int bpc, clock = adjusted_mode->crtc_clock; + + if (adjusted_mode->flags & DRM_MODE_FLAG_DBLCLK) + clock *= 2; + + /* YCBCR420 TMDS rate requirement is half the pixel clock */ + if (crtc_state->output_format == INTEL_OUTPUT_FORMAT_YCBCR420) + clock /= 2; + + bpc = intel_hdmi_compute_bpc(encoder, crtc_state, + clock, force_dvi); + + crtc_state->port_clock = intel_hdmi_port_clock(clock, bpc); + + /* + * pipe_bpp could already be below 8bpc due to + * FDI bandwidth constraints. We shouldn't bump it + * back up to 8bpc in that case. + */ + if (crtc_state->pipe_bpp > bpc * 3) + crtc_state->pipe_bpp = bpc * 3; + + DRM_DEBUG_KMS("picking %d bpc for HDMI output (pipe bpp: %d)\n", + bpc, crtc_state->pipe_bpp); + + if (hdmi_port_clock_valid(intel_hdmi, crtc_state->port_clock, + false, force_dvi) != MODE_OK) { + DRM_DEBUG_KMS("unsupported HDMI clock (%d kHz), rejecting mode\n", + crtc_state->port_clock); + return -EINVAL; + } + + return 0; +} + +static bool intel_hdmi_limited_color_range(const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) +{ + const struct intel_digital_connector_state *intel_conn_state = + to_intel_digital_connector_state(conn_state); + const struct drm_display_mode *adjusted_mode = + &crtc_state->base.adjusted_mode; + + /* + * Our YCbCr output is always limited range. + * crtc_state->limited_color_range only applies to RGB, + * and it must never be set for YCbCr or we risk setting + * some conflicting bits in PIPECONF which will mess up + * the colors on the monitor. + */ + if (crtc_state->output_format != INTEL_OUTPUT_FORMAT_RGB) + return false; + + if (intel_conn_state->broadcast_rgb == INTEL_BROADCAST_RGB_AUTO) { + /* See CEA-861-E - 5.1 Default Encoding Parameters */ + return crtc_state->has_hdmi_sink && + drm_default_rgb_quant_range(adjusted_mode) == + HDMI_QUANTIZATION_RANGE_LIMITED; + } else { + return intel_conn_state->broadcast_rgb == INTEL_BROADCAST_RGB_LIMITED; + } +} + int intel_hdmi_compute_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config, struct drm_connector_state *conn_state) @@ -2302,11 +2411,8 @@ int intel_hdmi_compute_config(struct intel_encoder *encoder, struct drm_scdc *scdc = &connector->display_info.hdmi.scdc; struct intel_digital_connector_state *intel_conn_state = to_intel_digital_connector_state(conn_state); - int clock_8bpc = pipe_config->base.adjusted_mode.crtc_clock; - int clock_10bpc = clock_8bpc * 5 / 4; - int clock_12bpc = clock_8bpc * 3 / 2; - int desired_bpp; bool force_dvi = intel_conn_state->force_audio == HDMI_AUDIO_OFF_DVI; + int ret; if (adjusted_mode->flags & DRM_MODE_FLAG_DBLSCAN) return -EINVAL; @@ -2317,33 +2423,19 @@ int intel_hdmi_compute_config(struct intel_encoder *encoder, if (pipe_config->has_hdmi_sink) pipe_config->has_infoframe = true; - if (intel_conn_state->broadcast_rgb == INTEL_BROADCAST_RGB_AUTO) { - /* See CEA-861-E - 5.1 Default Encoding Parameters */ - pipe_config->limited_color_range = - pipe_config->has_hdmi_sink && - drm_default_rgb_quant_range(adjusted_mode) == - HDMI_QUANTIZATION_RANGE_LIMITED; - } else { - pipe_config->limited_color_range = - intel_conn_state->broadcast_rgb == INTEL_BROADCAST_RGB_LIMITED; - } - - if (adjusted_mode->flags & DRM_MODE_FLAG_DBLCLK) { + if (adjusted_mode->flags & DRM_MODE_FLAG_DBLCLK) pipe_config->pixel_multiplier = 2; - clock_8bpc *= 2; - clock_10bpc *= 2; - clock_12bpc *= 2; - } if (drm_mode_is_420_only(&connector->display_info, adjusted_mode)) { - if (!intel_hdmi_ycbcr420_config(connector, pipe_config, - &clock_12bpc, &clock_10bpc, - &clock_8bpc)) { + if (!intel_hdmi_ycbcr420_config(connector, pipe_config)) { DRM_ERROR("Can't support YCBCR420 output\n"); return -EINVAL; } } + pipe_config->limited_color_range = + intel_hdmi_limited_color_range(pipe_config, conn_state); + if (HAS_PCH_SPLIT(dev_priv) && !HAS_DDI(dev_priv)) pipe_config->has_pch_encoder = true; @@ -2355,43 +2447,9 @@ int intel_hdmi_compute_config(struct intel_encoder *encoder, intel_conn_state->force_audio == HDMI_AUDIO_ON; } - /* - * Note that g4x/vlv don't support 12bpc hdmi outputs. We also need - * to check that the higher clock still fits within limits. - */ - if (hdmi_deep_color_possible(pipe_config, 12) && - hdmi_port_clock_valid(intel_hdmi, clock_12bpc, - true, force_dvi) == MODE_OK) { - DRM_DEBUG_KMS("picking bpc to 12 for HDMI output\n"); - desired_bpp = 12*3; - - /* Need to adjust the port link by 1.5x for 12bpc. */ - pipe_config->port_clock = clock_12bpc; - } else if (hdmi_deep_color_possible(pipe_config, 10) && - hdmi_port_clock_valid(intel_hdmi, clock_10bpc, - true, force_dvi) == MODE_OK) { - DRM_DEBUG_KMS("picking bpc to 10 for HDMI output\n"); - desired_bpp = 10 * 3; - - /* Need to adjust the port link by 1.25x for 10bpc. */ - pipe_config->port_clock = clock_10bpc; - } else { - DRM_DEBUG_KMS("picking bpc to 8 for HDMI output\n"); - desired_bpp = 8*3; - - pipe_config->port_clock = clock_8bpc; - } - - if (!pipe_config->bw_constrained) { - DRM_DEBUG_KMS("forcing pipe bpp to %i for HDMI\n", desired_bpp); - pipe_config->pipe_bpp = desired_bpp; - } - - if (hdmi_port_clock_valid(intel_hdmi, pipe_config->port_clock, - false, force_dvi) != MODE_OK) { - DRM_DEBUG_KMS("unsupported HDMI clock, rejecting mode\n"); - return -EINVAL; - } + ret = intel_hdmi_compute_clock(encoder, pipe_config, force_dvi); + if (ret) + return ret; /* Set user selected PAR to incoming mode's member */ adjusted_mode->picture_aspect_ratio = conn_state->picture_aspect_ratio; @@ -2431,6 +2489,9 @@ int intel_hdmi_compute_config(struct intel_encoder *encoder, return -EINVAL; } + intel_hdcp_transcoder_config(intel_hdmi->attached_connector, + pipe_config->cpu_transcoder); + return 0; } @@ -2565,6 +2626,12 @@ out: if (status != connector_status_connected) cec_notifier_phys_addr_invalidate(intel_hdmi->cec_notifier); + /* + * Make sure the refs for power wells enabled during detect are + * dropped to avoid a new detect cycle triggered by HPD polling. + */ + intel_display_power_flush_work(dev_priv); + return status; } @@ -2751,8 +2818,9 @@ intel_hdmi_connector_register(struct drm_connector *connector) static void intel_hdmi_destroy(struct drm_connector *connector) { - if (intel_attached_hdmi(connector)->cec_notifier) - cec_notifier_put(intel_attached_hdmi(connector)->cec_notifier); + struct cec_notifier *n = intel_attached_hdmi(connector)->cec_notifier; + + cec_notifier_conn_unregister(n); intel_connector_destroy(connector); } @@ -3001,7 +3069,7 @@ static u8 intel_hdmi_ddc_pin(struct drm_i915_private *dev_priv, if (HAS_PCH_MCC(dev_priv)) ddc_pin = mcc_port_to_ddc_pin(dev_priv, port); - else if (HAS_PCH_TGP(dev_priv) || HAS_PCH_ICP(dev_priv)) + else if (INTEL_PCH_TYPE(dev_priv) >= PCH_ICP) ddc_pin = icl_port_to_ddc_pin(dev_priv, port); else if (HAS_PCH_CNP(dev_priv)) ddc_pin = cnp_port_to_ddc_pin(dev_priv, port); @@ -3067,13 +3135,15 @@ void intel_hdmi_init_connector(struct intel_digital_port *intel_dig_port, struct drm_device *dev = intel_encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); enum port port = intel_encoder->port; + struct cec_connector_info conn_info; - DRM_DEBUG_KMS("Adding HDMI connector on port %c\n", - port_name(port)); + DRM_DEBUG_KMS("Adding HDMI connector on [ENCODER:%d:%s]\n", + intel_encoder->base.base.id, intel_encoder->base.name); if (WARN(intel_dig_port->max_lanes < 4, - "Not enough lanes (%d) for HDMI on port %c\n", - intel_dig_port->max_lanes, port_name(port))) + "Not enough lanes (%d) for HDMI on [ENCODER:%d:%s]\n", + intel_dig_port->max_lanes, intel_encoder->base.base.id, + intel_encoder->base.name)) return; drm_connector_init(dev, connector, &intel_hdmi_connector_funcs, @@ -3119,8 +3189,11 @@ void intel_hdmi_init_connector(struct intel_digital_port *intel_dig_port, I915_WRITE(PEG_BAND_GAP_DATA, (temp & ~0xf) | 0xd); } - intel_hdmi->cec_notifier = cec_notifier_get_conn(dev->dev, - port_identifier(port)); + cec_fill_conn_info_from_drm(&conn_info, connector); + + intel_hdmi->cec_notifier = + cec_notifier_conn_register(dev->dev, port_identifier(port), + &conn_info); if (!intel_hdmi->cec_notifier) DRM_DEBUG_KMS("CEC notifier get failed\n"); } @@ -3210,11 +3283,11 @@ void intel_hdmi_init(struct drm_i915_private *dev_priv, intel_encoder->port = port; if (IS_CHERRYVIEW(dev_priv)) { if (port == PORT_D) - intel_encoder->crtc_mask = 1 << 2; + intel_encoder->pipe_mask = BIT(PIPE_C); else - intel_encoder->crtc_mask = (1 << 0) | (1 << 1); + intel_encoder->pipe_mask = BIT(PIPE_A) | BIT(PIPE_B); } else { - intel_encoder->crtc_mask = (1 << 0) | (1 << 1) | (1 << 2); + intel_encoder->pipe_mask = ~0; } intel_encoder->cloneable = 1 << INTEL_OUTPUT_ANALOG; /* diff --git a/drivers/gpu/drm/i915/display/intel_hdmi.h b/drivers/gpu/drm/i915/display/intel_hdmi.h index 106c2e0bc3c9..cf1ea5427639 100644 --- a/drivers/gpu/drm/i915/display/intel_hdmi.h +++ b/drivers/gpu/drm/i915/display/intel_hdmi.h @@ -23,6 +23,7 @@ struct intel_crtc_state; struct intel_hdmi; struct drm_connector_state; union hdmi_infoframe; +enum port; void intel_hdmi_init(struct drm_i915_private *dev_priv, i915_reg_t hdmi_reg, enum port port); diff --git a/drivers/gpu/drm/i915/display/intel_hotplug.c b/drivers/gpu/drm/i915/display/intel_hotplug.c index 56be20f6f47e..fc29046d48ea 100644 --- a/drivers/gpu/drm/i915/display/intel_hotplug.c +++ b/drivers/gpu/drm/i915/display/intel_hotplug.c @@ -481,7 +481,8 @@ void intel_hpd_irq_handler(struct drm_i915_private *dev_priv, long_hpd = long_mask & BIT(pin); - DRM_DEBUG_DRIVER("digital hpd port %c - %s\n", port_name(port), + DRM_DEBUG_DRIVER("digital hpd on [ENCODER:%d:%s] - %s\n", + encoder->base.base.id, encoder->base.name, long_hpd ? "long" : "short"); queue_dig = true; diff --git a/drivers/gpu/drm/i915/display/intel_hotplug.h b/drivers/gpu/drm/i915/display/intel_hotplug.h index b0cd447b7fbc..087b5f57b321 100644 --- a/drivers/gpu/drm/i915/display/intel_hotplug.h +++ b/drivers/gpu/drm/i915/display/intel_hotplug.h @@ -13,6 +13,7 @@ struct drm_i915_private; struct intel_connector; struct intel_encoder; +enum port; void intel_hpd_poll_init(struct drm_i915_private *dev_priv); enum intel_hotplug_state intel_encoder_hotplug(struct intel_encoder *encoder, diff --git a/drivers/gpu/drm/i915/display/intel_lpe_audio.c b/drivers/gpu/drm/i915/display/intel_lpe_audio.c index b19800b58442..0b67f7887cd0 100644 --- a/drivers/gpu/drm/i915/display/intel_lpe_audio.c +++ b/drivers/gpu/drm/i915/display/intel_lpe_audio.c @@ -114,7 +114,7 @@ lpe_audio_platdev_create(struct drm_i915_private *dev_priv) pinfo.size_data = sizeof(*pdata); pinfo.dma_mask = DMA_BIT_MASK(32); - pdata->num_pipes = INTEL_INFO(dev_priv)->num_pipes; + pdata->num_pipes = INTEL_NUM_PIPES(dev_priv); pdata->num_ports = IS_CHERRYVIEW(dev_priv) ? 3 : 2; /* B,C,D or B,C */ pdata->port[0].pipe = -1; pdata->port[1].pipe = -1; diff --git a/drivers/gpu/drm/i915/display/intel_lvds.c b/drivers/gpu/drm/i915/display/intel_lvds.c index b7c459a8931c..b1bc78623647 100644 --- a/drivers/gpu/drm/i915/display/intel_lvds.c +++ b/drivers/gpu/drm/i915/display/intel_lvds.c @@ -232,7 +232,7 @@ static void intel_pre_enable_lvds(struct intel_encoder *encoder, struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); const struct drm_display_mode *adjusted_mode = &pipe_config->base.adjusted_mode; - int pipe = crtc->pipe; + enum pipe pipe = crtc->pipe; u32 temp; if (HAS_PCH_SPLIT(dev_priv)) { @@ -899,12 +899,10 @@ void intel_lvds_init(struct drm_i915_private *dev_priv) intel_encoder->power_domain = POWER_DOMAIN_PORT_OTHER; intel_encoder->port = PORT_NONE; intel_encoder->cloneable = 0; - if (HAS_PCH_SPLIT(dev_priv)) - intel_encoder->crtc_mask = (1 << 0) | (1 << 1) | (1 << 2); - else if (IS_GEN(dev_priv, 4)) - intel_encoder->crtc_mask = (1 << 0) | (1 << 1); + if (INTEL_GEN(dev_priv) < 4) + intel_encoder->pipe_mask = BIT(PIPE_B); else - intel_encoder->crtc_mask = (1 << 1); + intel_encoder->pipe_mask = ~0; drm_connector_helper_add(connector, &intel_lvds_connector_helper_funcs); connector->display_info.subpixel_order = SubPixelHorizontalRGB; diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c index 29edfc343716..848ce07a8ec2 100644 --- a/drivers/gpu/drm/i915/display/intel_overlay.c +++ b/drivers/gpu/drm/i915/display/intel_overlay.c @@ -30,6 +30,7 @@ #include <drm/i915_drm.h> #include "gem/i915_gem_pm.h" +#include "gt/intel_ring.h" #include "i915_drv.h" #include "i915_reg.h" @@ -230,7 +231,7 @@ alloc_request(struct intel_overlay *overlay, void (*fn)(struct intel_overlay *)) if (IS_ERR(rq)) return rq; - err = i915_active_ref(&overlay->last_flip, rq->timeline, rq); + err = i915_active_add_request(&overlay->last_flip, rq); if (err) { i915_request_add(rq); return ERR_PTR(err); @@ -439,8 +440,6 @@ static int intel_overlay_release_old_vid(struct intel_overlay *overlay) struct i915_request *rq; u32 *cs; - lockdep_assert_held(&dev_priv->drm.struct_mutex); - /* * Only wait if there is actually an old frame to release to * guarantee forward progress. @@ -751,7 +750,6 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay, struct i915_vma *vma; int ret, tmp_width; - lockdep_assert_held(&dev_priv->drm.struct_mutex); WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex)); ret = intel_overlay_release_old_vid(overlay); @@ -852,7 +850,6 @@ int intel_overlay_switch_off(struct intel_overlay *overlay) struct drm_i915_private *dev_priv = overlay->i915; int ret; - lockdep_assert_held(&dev_priv->drm.struct_mutex); WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex)); ret = intel_overlay_recover_from_interrupt(overlay); @@ -1068,11 +1065,7 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data, if (!(params->flags & I915_OVERLAY_ENABLE)) { drm_modeset_lock_all(dev); - mutex_lock(&dev->struct_mutex); - ret = intel_overlay_switch_off(overlay); - - mutex_unlock(&dev->struct_mutex); drm_modeset_unlock_all(dev); return ret; @@ -1088,7 +1081,6 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data, return -ENOENT; drm_modeset_lock_all(dev); - mutex_lock(&dev->struct_mutex); if (i915_gem_object_is_tiled(new_bo)) { DRM_DEBUG_KMS("buffer used for overlay image can not be tiled\n"); @@ -1152,14 +1144,12 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data, if (ret != 0) goto out_unlock; - mutex_unlock(&dev->struct_mutex); drm_modeset_unlock_all(dev); i915_gem_object_put(new_bo); return 0; out_unlock: - mutex_unlock(&dev->struct_mutex); drm_modeset_unlock_all(dev); i915_gem_object_put(new_bo); @@ -1233,7 +1223,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data, } drm_modeset_lock_all(dev); - mutex_lock(&dev->struct_mutex); ret = -EINVAL; if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) { @@ -1290,7 +1279,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data, ret = 0; out_unlock: - mutex_unlock(&dev->struct_mutex); drm_modeset_unlock_all(dev); return ret; @@ -1303,15 +1291,11 @@ static int get_registers(struct intel_overlay *overlay, bool use_phys) struct i915_vma *vma; int err; - mutex_lock(&i915->drm.struct_mutex); - obj = i915_gem_object_create_stolen(i915, PAGE_SIZE); - if (obj == NULL) + if (IS_ERR(obj)) obj = i915_gem_object_create_internal(i915, PAGE_SIZE); - if (IS_ERR(obj)) { - err = PTR_ERR(obj); - goto err_unlock; - } + if (IS_ERR(obj)) + return PTR_ERR(obj); vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE); if (IS_ERR(vma)) { @@ -1332,13 +1316,10 @@ static int get_registers(struct intel_overlay *overlay, bool use_phys) } overlay->reg_bo = obj; - mutex_unlock(&i915->drm.struct_mutex); return 0; err_put_bo: i915_gem_object_put(obj); -err_unlock: - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -1367,8 +1348,7 @@ void intel_overlay_setup(struct drm_i915_private *dev_priv) overlay->contrast = 75; overlay->saturation = 146; - i915_active_init(dev_priv, - &overlay->last_flip, + i915_active_init(&overlay->last_flip, NULL, intel_overlay_last_flip_retire); ret = get_registers(overlay, OVERLAY_NEEDS_PHYSICAL(dev_priv)); diff --git a/drivers/gpu/drm/i915/display/intel_psr.c b/drivers/gpu/drm/i915/display/intel_psr.c index 3bfb720560c2..6a9f322d3fca 100644 --- a/drivers/gpu/drm/i915/display/intel_psr.c +++ b/drivers/gpu/drm/i915/display/intel_psr.c @@ -76,7 +76,7 @@ static bool intel_psr2_enabled(struct drm_i915_private *dev_priv, const struct intel_crtc_state *crtc_state) { /* Cannot enable DSC and PSR2 simultaneously */ - WARN_ON(crtc_state->dsc_params.compression_enable && + WARN_ON(crtc_state->dsc.compression_enable && crtc_state->has_psr2); switch (dev_priv->psr.debug & I915_PSR_DEBUG_MODE_MASK) { @@ -88,48 +88,35 @@ static bool intel_psr2_enabled(struct drm_i915_private *dev_priv, } } -static int edp_psr_shift(enum transcoder cpu_transcoder) +static void psr_irq_control(struct drm_i915_private *dev_priv) { - switch (cpu_transcoder) { - case TRANSCODER_A: - return EDP_PSR_TRANSCODER_A_SHIFT; - case TRANSCODER_B: - return EDP_PSR_TRANSCODER_B_SHIFT; - case TRANSCODER_C: - return EDP_PSR_TRANSCODER_C_SHIFT; - default: - MISSING_CASE(cpu_transcoder); - /* fallthrough */ - case TRANSCODER_EDP: - return EDP_PSR_TRANSCODER_EDP_SHIFT; - } -} - -void intel_psr_irq_control(struct drm_i915_private *dev_priv, u32 debug) -{ - u32 debug_mask, mask; - enum transcoder cpu_transcoder; - u32 transcoders = BIT(TRANSCODER_EDP); + enum transcoder trans_shift; + u32 mask, val; + i915_reg_t imr_reg; - if (INTEL_GEN(dev_priv) >= 8) - transcoders |= BIT(TRANSCODER_A) | - BIT(TRANSCODER_B) | - BIT(TRANSCODER_C); - - debug_mask = 0; - mask = 0; - for_each_cpu_transcoder_masked(dev_priv, cpu_transcoder, transcoders) { - int shift = edp_psr_shift(cpu_transcoder); - - mask |= EDP_PSR_ERROR(shift); - debug_mask |= EDP_PSR_POST_EXIT(shift) | - EDP_PSR_PRE_ENTRY(shift); + /* + * gen12+ has registers relative to transcoder and one per transcoder + * using the same bit definition: handle it as TRANSCODER_EDP to force + * 0 shift in bit definition + */ + if (INTEL_GEN(dev_priv) >= 12) { + trans_shift = 0; + imr_reg = TRANS_PSR_IMR(dev_priv->psr.transcoder); + } else { + trans_shift = dev_priv->psr.transcoder; + imr_reg = EDP_PSR_IMR; } - if (debug & I915_PSR_DEBUG_IRQ) - mask |= debug_mask; + mask = EDP_PSR_ERROR(trans_shift); + if (dev_priv->psr.debug & I915_PSR_DEBUG_IRQ) + mask |= EDP_PSR_POST_EXIT(trans_shift) | + EDP_PSR_PRE_ENTRY(trans_shift); - I915_WRITE(EDP_PSR_IMR, ~mask); + /* Warning: it is masking/setting reserved bits too */ + val = I915_READ(imr_reg); + val &= ~EDP_PSR_TRANS_MASK(trans_shift); + val |= ~mask; + I915_WRITE(imr_reg, val); } static void psr_event_print(u32 val, bool psr2_enabled) @@ -171,60 +158,58 @@ static void psr_event_print(u32 val, bool psr2_enabled) void intel_psr_irq_handler(struct drm_i915_private *dev_priv, u32 psr_iir) { - u32 transcoders = BIT(TRANSCODER_EDP); - enum transcoder cpu_transcoder; + enum transcoder cpu_transcoder = dev_priv->psr.transcoder; + enum transcoder trans_shift; + i915_reg_t imr_reg; ktime_t time_ns = ktime_get(); - u32 mask = 0; - if (INTEL_GEN(dev_priv) >= 8) - transcoders |= BIT(TRANSCODER_A) | - BIT(TRANSCODER_B) | - BIT(TRANSCODER_C); + if (INTEL_GEN(dev_priv) >= 12) { + trans_shift = 0; + imr_reg = TRANS_PSR_IMR(dev_priv->psr.transcoder); + } else { + trans_shift = dev_priv->psr.transcoder; + imr_reg = EDP_PSR_IMR; + } - for_each_cpu_transcoder_masked(dev_priv, cpu_transcoder, transcoders) { - int shift = edp_psr_shift(cpu_transcoder); + if (psr_iir & EDP_PSR_PRE_ENTRY(trans_shift)) { + dev_priv->psr.last_entry_attempt = time_ns; + DRM_DEBUG_KMS("[transcoder %s] PSR entry attempt in 2 vblanks\n", + transcoder_name(cpu_transcoder)); + } - if (psr_iir & EDP_PSR_ERROR(shift)) { - DRM_WARN("[transcoder %s] PSR aux error\n", - transcoder_name(cpu_transcoder)); + if (psr_iir & EDP_PSR_POST_EXIT(trans_shift)) { + dev_priv->psr.last_exit = time_ns; + DRM_DEBUG_KMS("[transcoder %s] PSR exit completed\n", + transcoder_name(cpu_transcoder)); - dev_priv->psr.irq_aux_error = true; + if (INTEL_GEN(dev_priv) >= 9) { + u32 val = I915_READ(PSR_EVENT(cpu_transcoder)); + bool psr2_enabled = dev_priv->psr.psr2_enabled; - /* - * If this interruption is not masked it will keep - * interrupting so fast that it prevents the scheduled - * work to run. - * Also after a PSR error, we don't want to arm PSR - * again so we don't care about unmask the interruption - * or unset irq_aux_error. - */ - mask |= EDP_PSR_ERROR(shift); - } - - if (psr_iir & EDP_PSR_PRE_ENTRY(shift)) { - dev_priv->psr.last_entry_attempt = time_ns; - DRM_DEBUG_KMS("[transcoder %s] PSR entry attempt in 2 vblanks\n", - transcoder_name(cpu_transcoder)); + I915_WRITE(PSR_EVENT(cpu_transcoder), val); + psr_event_print(val, psr2_enabled); } + } - if (psr_iir & EDP_PSR_POST_EXIT(shift)) { - dev_priv->psr.last_exit = time_ns; - DRM_DEBUG_KMS("[transcoder %s] PSR exit completed\n", - transcoder_name(cpu_transcoder)); + if (psr_iir & EDP_PSR_ERROR(trans_shift)) { + u32 val; - if (INTEL_GEN(dev_priv) >= 9) { - u32 val = I915_READ(PSR_EVENT(cpu_transcoder)); - bool psr2_enabled = dev_priv->psr.psr2_enabled; + DRM_WARN("[transcoder %s] PSR aux error\n", + transcoder_name(cpu_transcoder)); - I915_WRITE(PSR_EVENT(cpu_transcoder), val); - psr_event_print(val, psr2_enabled); - } - } - } + dev_priv->psr.irq_aux_error = true; - if (mask) { - mask |= I915_READ(EDP_PSR_IMR); - I915_WRITE(EDP_PSR_IMR, mask); + /* + * If this interruption is not masked it will keep + * interrupting so fast that it prevents the scheduled + * work to run. + * Also after a PSR error, we don't want to arm PSR + * again so we don't care about unmask the interruption + * or unset irq_aux_error. + */ + val = I915_READ(imr_reg); + val |= EDP_PSR_ERROR(trans_shift); + I915_WRITE(imr_reg, val); schedule_work(&dev_priv->psr.work); } @@ -283,6 +268,11 @@ void intel_psr_init_dpcd(struct intel_dp *intel_dp) struct drm_i915_private *dev_priv = to_i915(dp_to_dig_port(intel_dp)->base.base.dev); + if (dev_priv->psr.dp) { + DRM_WARN("More than one eDP panel found, PSR support should be extended\n"); + return; + } + drm_dp_dpcd_read(&intel_dp->aux, DP_PSR_SUPPORT, intel_dp->psr_dpcd, sizeof(intel_dp->psr_dpcd)); @@ -305,7 +295,6 @@ void intel_psr_init_dpcd(struct intel_dp *intel_dp) dev_priv->psr.sink_sync_latency = intel_dp_get_sink_sync_latency(intel_dp); - WARN_ON(dev_priv->psr.dp); dev_priv->psr.dp = intel_dp; if (INTEL_GEN(dev_priv) >= 9 && @@ -390,7 +379,7 @@ static void hsw_psr_setup_aux(struct intel_dp *intel_dp) BUILD_BUG_ON(sizeof(aux_msg) > 20); for (i = 0; i < sizeof(aux_msg); i += 4) - I915_WRITE(EDP_PSR_AUX_DATA(i >> 2), + I915_WRITE(EDP_PSR_AUX_DATA(dev_priv->psr.transcoder, i >> 2), intel_dp_pack_aux(&aux_msg[i], sizeof(aux_msg) - i)); aux_clock_divider = intel_dp->get_aux_clock_divider(intel_dp, 0); @@ -401,7 +390,7 @@ static void hsw_psr_setup_aux(struct intel_dp *intel_dp) /* Select only valid bits for SRD_AUX_CTL */ aux_ctl &= psr_aux_mask; - I915_WRITE(EDP_PSR_AUX_CTL, aux_ctl); + I915_WRITE(EDP_PSR_AUX_CTL(dev_priv->psr.transcoder), aux_ctl); } static void intel_psr_enable_sink(struct intel_dp *intel_dp) @@ -491,8 +480,9 @@ static void hsw_activate_psr1(struct intel_dp *intel_dp) if (INTEL_GEN(dev_priv) >= 8) val |= EDP_PSR_CRC_ENABLE; - val |= I915_READ(EDP_PSR_CTL) & EDP_PSR_RESTORE_PSR_ACTIVE_CTX_MASK; - I915_WRITE(EDP_PSR_CTL, val); + val |= (I915_READ(EDP_PSR_CTL(dev_priv->psr.transcoder)) & + EDP_PSR_RESTORE_PSR_ACTIVE_CTX_MASK); + I915_WRITE(EDP_PSR_CTL(dev_priv->psr.transcoder), val); } static void hsw_activate_psr2(struct intel_dp *intel_dp) @@ -528,9 +518,87 @@ static void hsw_activate_psr2(struct intel_dp *intel_dp) * PSR2 HW is incorrectly using EDP_PSR_TP1_TP3_SEL and BSpec is * recommending keep this bit unset while PSR2 is enabled. */ - I915_WRITE(EDP_PSR_CTL, 0); + I915_WRITE(EDP_PSR_CTL(dev_priv->psr.transcoder), 0); + + I915_WRITE(EDP_PSR2_CTL(dev_priv->psr.transcoder), val); +} - I915_WRITE(EDP_PSR2_CTL, val); +static bool +transcoder_has_psr2(struct drm_i915_private *dev_priv, enum transcoder trans) +{ + if (INTEL_GEN(dev_priv) < 9) + return false; + else if (INTEL_GEN(dev_priv) >= 12) + return trans == TRANSCODER_A; + else + return trans == TRANSCODER_EDP; +} + +static u32 intel_get_frame_time_us(const struct intel_crtc_state *cstate) +{ + if (!cstate || !cstate->base.active) + return 0; + + return DIV_ROUND_UP(1000 * 1000, + drm_mode_vrefresh(&cstate->base.adjusted_mode)); +} + +static void psr2_program_idle_frames(struct drm_i915_private *dev_priv, + u32 idle_frames) +{ + u32 val; + + idle_frames <<= EDP_PSR2_IDLE_FRAME_SHIFT; + val = I915_READ(EDP_PSR2_CTL(dev_priv->psr.transcoder)); + val &= ~EDP_PSR2_IDLE_FRAME_MASK; + val |= idle_frames; + I915_WRITE(EDP_PSR2_CTL(dev_priv->psr.transcoder), val); +} + +static void tgl_psr2_enable_dc3co(struct drm_i915_private *dev_priv) +{ + psr2_program_idle_frames(dev_priv, 0); + intel_display_power_set_target_dc_state(dev_priv, DC_STATE_EN_DC3CO); +} + +static void tgl_psr2_disable_dc3co(struct drm_i915_private *dev_priv) +{ + int idle_frames; + + intel_display_power_set_target_dc_state(dev_priv, DC_STATE_EN_UPTO_DC6); + /* + * Restore PSR2 idle frame let's use 6 as the minimum to cover all known + * cases including the off-by-one issue that HW has in some cases. + */ + idle_frames = max(6, dev_priv->vbt.psr.idle_frames); + idle_frames = max(idle_frames, dev_priv->psr.sink_sync_latency + 1); + psr2_program_idle_frames(dev_priv, idle_frames); +} + +static void tgl_dc5_idle_thread(struct work_struct *work) +{ + struct drm_i915_private *dev_priv = + container_of(work, typeof(*dev_priv), psr.idle_work.work); + + mutex_lock(&dev_priv->psr.lock); + /* If delayed work is pending, it is not idle */ + if (delayed_work_pending(&dev_priv->psr.idle_work)) + goto unlock; + + DRM_DEBUG_KMS("DC5/6 idle thread\n"); + tgl_psr2_disable_dc3co(dev_priv); +unlock: + mutex_unlock(&dev_priv->psr.lock); +} + +static void tgl_disallow_dc3co_on_psr2_exit(struct drm_i915_private *dev_priv) +{ + if (!dev_priv->psr.dc3co_enabled) + return; + + cancel_delayed_work(&dev_priv->psr.idle_work); + /* Before PSR2 exit disallow dc3co*/ + tgl_psr2_disable_dc3co(dev_priv); } static bool intel_psr2_config_valid(struct intel_dp *intel_dp, @@ -544,17 +612,26 @@ static bool intel_psr2_config_valid(struct intel_dp *intel_dp, if (!dev_priv->psr.sink_psr2_support) return false; + if (!transcoder_has_psr2(dev_priv, crtc_state->cpu_transcoder)) { + DRM_DEBUG_KMS("PSR2 not supported in transcoder %s\n", + transcoder_name(crtc_state->cpu_transcoder)); + return false; + } + /* * DSC and PSR2 cannot be enabled simultaneously. If a requested * resolution requires DSC to be enabled, priority is given to DSC * over PSR2. */ - if (crtc_state->dsc_params.compression_enable) { + if (crtc_state->dsc.compression_enable) { DRM_DEBUG_KMS("PSR2 cannot be enabled since DSC is enabled\n"); return false; } - if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) { + if (INTEL_GEN(dev_priv) >= 12) { + psr_max_h = 5120; + psr_max_v = 3200; + } else if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) { psr_max_h = 4096; psr_max_v = 2304; } else if (IS_GEN(dev_priv, 9)) { @@ -606,10 +683,9 @@ void intel_psr_compute_config(struct intel_dp *intel_dp, /* * HSW spec explicitly says PSR is tied to port A. - * BDW+ platforms with DDI implementation of PSR have different - * PSR registers per transcoder and we only implement transcoder EDP - * ones. Since by Display design transcoder EDP is tied to port A - * we can safely escape based on the port A. + * BDW+ platforms have a instance of PSR registers per transcoder but + * for now it only supports one instance of PSR, so lets keep it + * hardcoded to PORT_A */ if (dig_port->base.port != PORT_A) { DRM_DEBUG_KMS("PSR condition failed: Port not supported\n"); @@ -648,9 +724,10 @@ static void intel_psr_activate(struct intel_dp *intel_dp) { struct drm_i915_private *dev_priv = dp_to_i915(intel_dp); - if (INTEL_GEN(dev_priv) >= 9) - WARN_ON(I915_READ(EDP_PSR2_CTL) & EDP_PSR2_ENABLE); - WARN_ON(I915_READ(EDP_PSR_CTL) & EDP_PSR_ENABLE); + if (transcoder_has_psr2(dev_priv, dev_priv->psr.transcoder)) + WARN_ON(I915_READ(EDP_PSR2_CTL(dev_priv->psr.transcoder)) & EDP_PSR2_ENABLE); + + WARN_ON(I915_READ(EDP_PSR_CTL(dev_priv->psr.transcoder)) & EDP_PSR_ENABLE); WARN_ON(dev_priv->psr.active); lockdep_assert_held(&dev_priv->psr.lock); @@ -663,25 +740,6 @@ static void intel_psr_activate(struct intel_dp *intel_dp) dev_priv->psr.active = true; } -static i915_reg_t gen9_chicken_trans_reg(struct drm_i915_private *dev_priv, - enum transcoder cpu_transcoder) -{ - static const i915_reg_t regs[] = { - [TRANSCODER_A] = CHICKEN_TRANS_A, - [TRANSCODER_B] = CHICKEN_TRANS_B, - [TRANSCODER_C] = CHICKEN_TRANS_C, - [TRANSCODER_EDP] = CHICKEN_TRANS_EDP, - }; - - WARN_ON(INTEL_GEN(dev_priv) < 9); - - if (WARN_ON(cpu_transcoder >= ARRAY_SIZE(regs) || - !regs[cpu_transcoder].reg)) - cpu_transcoder = TRANSCODER_A; - - return regs[cpu_transcoder]; -} - static void intel_psr_enable_source(struct intel_dp *intel_dp, const struct intel_crtc_state *crtc_state) { @@ -697,8 +755,7 @@ static void intel_psr_enable_source(struct intel_dp *intel_dp, if (dev_priv->psr.psr2_enabled && (IS_GEN(dev_priv, 9) && !IS_GEMINILAKE(dev_priv))) { - i915_reg_t reg = gen9_chicken_trans_reg(dev_priv, - cpu_transcoder); + i915_reg_t reg = CHICKEN_TRANS(cpu_transcoder); u32 chicken = I915_READ(reg); chicken |= PSR2_VSC_ENABLE_PROG_HEADER | @@ -720,19 +777,46 @@ static void intel_psr_enable_source(struct intel_dp *intel_dp, if (INTEL_GEN(dev_priv) < 11) mask |= EDP_PSR_DEBUG_MASK_DISP_REG_WRITE; - I915_WRITE(EDP_PSR_DEBUG, mask); + I915_WRITE(EDP_PSR_DEBUG(dev_priv->psr.transcoder), mask); + + psr_irq_control(dev_priv); } static void intel_psr_enable_locked(struct drm_i915_private *dev_priv, const struct intel_crtc_state *crtc_state) { struct intel_dp *intel_dp = dev_priv->psr.dp; + u32 val; WARN_ON(dev_priv->psr.enabled); dev_priv->psr.psr2_enabled = intel_psr2_enabled(dev_priv, crtc_state); dev_priv->psr.busy_frontbuffer_bits = 0; dev_priv->psr.pipe = to_intel_crtc(crtc_state->base.crtc)->pipe; + dev_priv->psr.dc3co_enabled = !!crtc_state->dc3co_exitline; + dev_priv->psr.dc3co_exit_delay = intel_get_frame_time_us(crtc_state); + dev_priv->psr.transcoder = crtc_state->cpu_transcoder; + + /* + * If a PSR error happened and the driver is reloaded, the EDP_PSR_IIR + * will still keep the error set even after the reset done in the + * irq_preinstall and irq_uninstall hooks. + * And enabling in this situation cause the screen to freeze in the + * first time that PSR HW tries to activate so lets keep PSR disabled + * to avoid any rendering problems. + */ + if (INTEL_GEN(dev_priv) >= 12) { + val = I915_READ(TRANS_PSR_IIR(dev_priv->psr.transcoder)); + val &= EDP_PSR_ERROR(0); + } else { + val = I915_READ(EDP_PSR_IIR); + val &= EDP_PSR_ERROR(dev_priv->psr.transcoder); + } + if (val) { + dev_priv->psr.sink_not_reliable = true; + DRM_DEBUG_KMS("PSR interruption error set, not enabling PSR\n"); + return; + } DRM_DEBUG_KMS("Enabling PSR%s\n", dev_priv->psr.psr2_enabled ? "2" : "1"); @@ -782,20 +866,28 @@ static void intel_psr_exit(struct drm_i915_private *dev_priv) u32 val; if (!dev_priv->psr.active) { - if (INTEL_GEN(dev_priv) >= 9) - WARN_ON(I915_READ(EDP_PSR2_CTL) & EDP_PSR2_ENABLE); - WARN_ON(I915_READ(EDP_PSR_CTL) & EDP_PSR_ENABLE); + if (transcoder_has_psr2(dev_priv, dev_priv->psr.transcoder)) { + val = I915_READ(EDP_PSR2_CTL(dev_priv->psr.transcoder)); + WARN_ON(val & EDP_PSR2_ENABLE); + } + + val = I915_READ(EDP_PSR_CTL(dev_priv->psr.transcoder)); + WARN_ON(val & EDP_PSR_ENABLE); + return; } if (dev_priv->psr.psr2_enabled) { - val = I915_READ(EDP_PSR2_CTL); + tgl_disallow_dc3co_on_psr2_exit(dev_priv); + val = I915_READ(EDP_PSR2_CTL(dev_priv->psr.transcoder)); WARN_ON(!(val & EDP_PSR2_ENABLE)); - I915_WRITE(EDP_PSR2_CTL, val & ~EDP_PSR2_ENABLE); + val &= ~EDP_PSR2_ENABLE; + I915_WRITE(EDP_PSR2_CTL(dev_priv->psr.transcoder), val); } else { - val = I915_READ(EDP_PSR_CTL); + val = I915_READ(EDP_PSR_CTL(dev_priv->psr.transcoder)); WARN_ON(!(val & EDP_PSR_ENABLE)); - I915_WRITE(EDP_PSR_CTL, val & ~EDP_PSR_ENABLE); + val &= ~EDP_PSR_ENABLE; + I915_WRITE(EDP_PSR_CTL(dev_priv->psr.transcoder), val); } dev_priv->psr.active = false; } @@ -817,10 +909,10 @@ static void intel_psr_disable_locked(struct intel_dp *intel_dp) intel_psr_exit(dev_priv); if (dev_priv->psr.psr2_enabled) { - psr_status = EDP_PSR2_STATUS; + psr_status = EDP_PSR2_STATUS(dev_priv->psr.transcoder); psr_status_mask = EDP_PSR2_STATUS_STATE_MASK; } else { - psr_status = EDP_PSR_STATUS; + psr_status = EDP_PSR_STATUS(dev_priv->psr.transcoder); psr_status_mask = EDP_PSR_STATUS_STATE_MASK; } @@ -859,6 +951,7 @@ void intel_psr_disable(struct intel_dp *intel_dp, mutex_unlock(&dev_priv->psr.lock); cancel_work_sync(&dev_priv->psr.work); + cancel_delayed_work_sync(&dev_priv->psr.idle_work); } static void psr_force_hw_tracking_exit(struct drm_i915_private *dev_priv) @@ -963,7 +1056,8 @@ int intel_psr_wait_for_idle(const struct intel_crtc_state *new_crtc_state, * defensive enough to cover everything. */ - return __intel_wait_for_register(&dev_priv->uncore, EDP_PSR_STATUS, + return __intel_wait_for_register(&dev_priv->uncore, + EDP_PSR_STATUS(dev_priv->psr.transcoder), EDP_PSR_STATUS_STATE_MASK, EDP_PSR_STATUS_STATE_IDLE, 2, 50, out_value); @@ -979,10 +1073,10 @@ static bool __psr_wait_for_idle_locked(struct drm_i915_private *dev_priv) return false; if (dev_priv->psr.psr2_enabled) { - reg = EDP_PSR2_STATUS; + reg = EDP_PSR2_STATUS(dev_priv->psr.transcoder); mask = EDP_PSR2_STATUS_STATE_MASK; } else { - reg = EDP_PSR_STATUS; + reg = EDP_PSR_STATUS(dev_priv->psr.transcoder); mask = EDP_PSR_STATUS_STATE_MASK; } @@ -1067,7 +1161,13 @@ int intel_psr_debug_set(struct drm_i915_private *dev_priv, u64 val) old_mode = dev_priv->psr.debug & I915_PSR_DEBUG_MODE_MASK; dev_priv->psr.debug = val; - intel_psr_irq_control(dev_priv, dev_priv->psr.debug); + + /* + * Do it right away if it's already enabled, otherwise it will be done + * when enabling the source. + */ + if (dev_priv->psr.enabled) + psr_irq_control(dev_priv); mutex_unlock(&dev_priv->psr.lock); @@ -1159,6 +1259,44 @@ void intel_psr_invalidate(struct drm_i915_private *dev_priv, mutex_unlock(&dev_priv->psr.lock); } +/* + * When we will be completely rely on PSR2 S/W tracking in future, + * intel_psr_flush() will invalidate and flush the PSR for ORIGIN_FLIP + * event also therefore tgl_dc3co_flush() require to be changed + * accrodingly in future. + */ +static void +tgl_dc3co_flush(struct drm_i915_private *dev_priv, + unsigned int frontbuffer_bits, enum fb_op_origin origin) +{ + u32 delay; + + mutex_lock(&dev_priv->psr.lock); + + if (!dev_priv->psr.dc3co_enabled) + goto unlock; + + if (!dev_priv->psr.psr2_enabled || !dev_priv->psr.active) + goto unlock; + + /* + * At every frontbuffer flush flip event modified delay of delayed work, + * when delayed work schedules that means display has been idle. + */ + if (!(frontbuffer_bits & + INTEL_FRONTBUFFER_ALL_MASK(dev_priv->psr.pipe))) + goto unlock; + + tgl_psr2_enable_dc3co(dev_priv); + /* DC5/DC6 required idle frames = 6 */ + delay = 6 * dev_priv->psr.dc3co_exit_delay; + mod_delayed_work(system_wq, &dev_priv->psr.idle_work, + usecs_to_jiffies(delay)); + +unlock: + mutex_unlock(&dev_priv->psr.lock); +} + /** * intel_psr_flush - Flush PSR * @dev_priv: i915 device @@ -1178,8 +1316,10 @@ void intel_psr_flush(struct drm_i915_private *dev_priv, if (!CAN_PSR(dev_priv)) return; - if (origin == ORIGIN_FLIP) + if (origin == ORIGIN_FLIP) { + tgl_dc3co_flush(dev_priv, frontbuffer_bits, origin); return; + } mutex_lock(&dev_priv->psr.lock); if (!dev_priv->psr.enabled) { @@ -1208,45 +1348,34 @@ void intel_psr_flush(struct drm_i915_private *dev_priv, */ void intel_psr_init(struct drm_i915_private *dev_priv) { - u32 val; - if (!HAS_PSR(dev_priv)) return; - dev_priv->psr_mmio_base = IS_HASWELL(dev_priv) ? - HSW_EDP_PSR_BASE : BDW_EDP_PSR_BASE; - if (!dev_priv->psr.sink_support) return; + if (IS_HASWELL(dev_priv)) + /* + * HSW don't have PSR registers on the same space as transcoder + * so set this to a value that when subtract to the register + * in transcoder space results in the right offset for HSW + */ + dev_priv->hsw_psr_mmio_adjust = _SRD_CTL_EDP - _HSW_EDP_PSR_BASE; + if (i915_modparams.enable_psr == -1) if (INTEL_GEN(dev_priv) < 9 || !dev_priv->vbt.psr.enable) i915_modparams.enable_psr = 0; - /* - * If a PSR error happened and the driver is reloaded, the EDP_PSR_IIR - * will still keep the error set even after the reset done in the - * irq_preinstall and irq_uninstall hooks. - * And enabling in this situation cause the screen to freeze in the - * first time that PSR HW tries to activate so lets keep PSR disabled - * to avoid any rendering problems. - */ - val = I915_READ(EDP_PSR_IIR); - val &= EDP_PSR_ERROR(edp_psr_shift(TRANSCODER_EDP)); - if (val) { - DRM_DEBUG_KMS("PSR interruption error set\n"); - dev_priv->psr.sink_not_reliable = true; - } - /* Set link_standby x link_off defaults */ if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) /* HSW and BDW require workarounds that we don't implement. */ dev_priv->psr.link_standby = false; - else - /* For new platforms let's respect VBT back again */ + else if (INTEL_GEN(dev_priv) < 12) + /* For new platforms up to TGL let's respect VBT back again */ dev_priv->psr.link_standby = dev_priv->vbt.psr.full_link; INIT_WORK(&dev_priv->psr.work, intel_psr_work); + INIT_DELAYED_WORK(&dev_priv->psr.idle_work, tgl_dc5_idle_thread); mutex_init(&dev_priv->psr.lock); } @@ -1288,7 +1417,7 @@ void intel_psr_short_pulse(struct intel_dp *intel_dp) if (val & DP_PSR_VSC_SDP_UNCORRECTABLE_ERROR) DRM_DEBUG_KMS("PSR VSC SDP uncorrectable error, disabling PSR\n"); if (val & DP_PSR_LINK_CRC_ERROR) - DRM_ERROR("PSR Link CRC error, disabling PSR\n"); + DRM_DEBUG_KMS("PSR Link CRC error, disabling PSR\n"); if (val & ~errors) DRM_ERROR("PSR_ERROR_STATUS unhandled errors %x\n", diff --git a/drivers/gpu/drm/i915/display/intel_psr.h b/drivers/gpu/drm/i915/display/intel_psr.h index dc818826f36d..46e4de8b8cd5 100644 --- a/drivers/gpu/drm/i915/display/intel_psr.h +++ b/drivers/gpu/drm/i915/display/intel_psr.h @@ -30,7 +30,6 @@ void intel_psr_flush(struct drm_i915_private *dev_priv, void intel_psr_init(struct drm_i915_private *dev_priv); void intel_psr_compute_config(struct intel_dp *intel_dp, struct intel_crtc_state *crtc_state); -void intel_psr_irq_control(struct drm_i915_private *dev_priv, u32 debug); void intel_psr_irq_handler(struct drm_i915_private *dev_priv, u32 psr_iir); void intel_psr_short_pulse(struct intel_dp *intel_dp); int intel_psr_wait_for_idle(const struct intel_crtc_state *new_crtc_state, diff --git a/drivers/gpu/drm/i915/display/intel_sdvo.c b/drivers/gpu/drm/i915/display/intel_sdvo.c index adeb1c840976..5b7f4baf7348 100644 --- a/drivers/gpu/drm/i915/display/intel_sdvo.c +++ b/drivers/gpu/drm/i915/display/intel_sdvo.c @@ -2921,7 +2921,7 @@ intel_sdvo_output_setup(struct intel_sdvo *intel_sdvo, u16 flags) bytes[0], bytes[1]); return false; } - intel_sdvo->base.crtc_mask = (1 << 0) | (1 << 1) | (1 << 2); + intel_sdvo->base.pipe_mask = ~0; return true; } diff --git a/drivers/gpu/drm/i915/display/intel_sdvo.h b/drivers/gpu/drm/i915/display/intel_sdvo.h index c9e05bcdd141..a66f224aa17d 100644 --- a/drivers/gpu/drm/i915/display/intel_sdvo.h +++ b/drivers/gpu/drm/i915/display/intel_sdvo.h @@ -14,6 +14,7 @@ struct drm_i915_private; enum pipe; +enum port; bool intel_sdvo_port_enabled(struct drm_i915_private *dev_priv, i915_reg_t sdvo_reg, enum pipe *pipe); diff --git a/drivers/gpu/drm/i915/display/intel_sprite.c b/drivers/gpu/drm/i915/display/intel_sprite.c index dea63be1964f..72fda0430062 100644 --- a/drivers/gpu/drm/i915/display/intel_sprite.c +++ b/drivers/gpu/drm/i915/display/intel_sprite.c @@ -48,19 +48,6 @@ #include "intel_psr.h" #include "intel_sprite.h" -bool is_planar_yuv_format(u32 pixelformat) -{ - switch (pixelformat) { - case DRM_FORMAT_NV12: - case DRM_FORMAT_P010: - case DRM_FORMAT_P012: - case DRM_FORMAT_P016: - return true; - default: - return false; - } -} - int intel_usecs_to_scanlines(const struct drm_display_mode *adjusted_mode, int usecs) { @@ -300,10 +287,8 @@ int intel_plane_check_src_coordinates(struct intel_plane_state *plane_state) src_y = src->y1 >> 16; src_h = drm_rect_height(src) >> 16; - src->x1 = src_x << 16; - src->x2 = (src_x + src_w) << 16; - src->y1 = src_y << 16; - src->y2 = (src_y + src_h) << 16; + drm_rect_init(src, src_x << 16, src_y << 16, + src_w << 16, src_h << 16); if (!fb->format->is_yuv) return 0; @@ -337,6 +322,55 @@ bool icl_is_hdr_plane(struct drm_i915_private *dev_priv, enum plane_id plane_id) icl_hdr_plane_mask() & BIT(plane_id); } +static void +skl_plane_ratio(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state, + unsigned int *num, unsigned int *den) +{ + struct drm_i915_private *dev_priv = to_i915(plane_state->base.plane->dev); + const struct drm_framebuffer *fb = plane_state->base.fb; + + if (fb->format->cpp[0] == 8) { + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) { + *num = 10; + *den = 8; + } else { + *num = 9; + *den = 8; + } + } else { + *num = 1; + *den = 1; + } +} + +static int skl_plane_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) +{ + struct drm_i915_private *dev_priv = to_i915(plane_state->base.plane->dev); + unsigned int pixel_rate = crtc_state->pixel_rate; + unsigned int src_w, src_h, dst_w, dst_h; + unsigned int num, den; + + skl_plane_ratio(crtc_state, plane_state, &num, &den); + + /* two pixels per clock on glk+ */ + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) + den *= 2; + + src_w = drm_rect_width(&plane_state->base.src) >> 16; + src_h = drm_rect_height(&plane_state->base.src) >> 16; + dst_w = drm_rect_width(&plane_state->base.dst); + dst_h = drm_rect_height(&plane_state->base.dst); + + /* Downscaling limits the maximum pixel rate */ + dst_w = min(src_w, dst_w); + dst_h = min(src_h, dst_h); + + return DIV64_U64_ROUND_UP(mul_u32_u32(pixel_rate * num, src_w * src_h), + mul_u32_u32(den, dst_w * dst_h)); +} + static unsigned int skl_plane_max_stride(struct intel_plane *plane, u32 pixel_format, u64 modifier, @@ -361,6 +395,7 @@ skl_program_scaler(struct intel_plane *plane, const struct intel_plane_state *plane_state) { struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + const struct drm_framebuffer *fb = plane_state->base.fb; enum pipe pipe = plane->pipe; int scaler_id = plane_state->scaler_id; const struct intel_scaler *scaler = @@ -381,7 +416,7 @@ skl_program_scaler(struct intel_plane *plane, 0, INT_MAX); /* TODO: handle sub-pixel coordinates */ - if (is_planar_yuv_format(plane_state->base.fb->format->format) && + if (drm_format_info_is_yuv_semiplanar(fb->format) && !icl_is_hdr_plane(dev_priv, plane->id)) { y_hphase = skl_scaler_calc_phase(1, hscale, false); y_vphase = skl_scaler_calc_phase(1, vscale, false); @@ -554,7 +589,7 @@ skl_program_plane(struct intel_plane *plane, u32 y = plane_state->color_plane[color_plane].y; u32 src_w = drm_rect_width(&plane_state->base.src) >> 16; u32 src_h = drm_rect_height(&plane_state->base.src) >> 16; - struct intel_plane *linked = plane_state->linked_plane; + struct intel_plane *linked = plane_state->planar_linked_plane; const struct drm_framebuffer *fb = plane_state->base.fb; u8 alpha = plane_state->base.alpha >> 8; u32 plane_color_ctl = 0; @@ -653,7 +688,7 @@ skl_update_plane(struct intel_plane *plane, { int color_plane = 0; - if (plane_state->linked_plane) { + if (plane_state->planar_linked_plane) { /* Program the UV plane */ color_plane = 1; } @@ -825,6 +860,85 @@ vlv_update_clrc(const struct intel_plane_state *plane_state) SP_SH_SIN(sh_sin) | SP_SH_COS(sh_cos)); } +static void +vlv_plane_ratio(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state, + unsigned int *num, unsigned int *den) +{ + u8 active_planes = crtc_state->active_planes & ~BIT(PLANE_CURSOR); + const struct drm_framebuffer *fb = plane_state->base.fb; + unsigned int cpp = fb->format->cpp[0]; + + /* + * VLV bspec only considers cases where all three planes are + * enabled, and cases where the primary and one sprite is enabled. + * Let's assume the case with just two sprites enabled also + * maps to the latter case. + */ + if (hweight8(active_planes) == 3) { + switch (cpp) { + case 8: + *num = 11; + *den = 8; + break; + case 4: + *num = 18; + *den = 16; + break; + default: + *num = 1; + *den = 1; + break; + } + } else if (hweight8(active_planes) == 2) { + switch (cpp) { + case 8: + *num = 10; + *den = 8; + break; + case 4: + *num = 17; + *den = 16; + break; + default: + *num = 1; + *den = 1; + break; + } + } else { + switch (cpp) { + case 8: + *num = 10; + *den = 8; + break; + default: + *num = 1; + *den = 1; + break; + } + } +} + +int vlv_plane_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) +{ + unsigned int pixel_rate; + unsigned int num, den; + + /* + * Note that crtc_state->pixel_rate accounts for both + * horizontal and vertical panel fitter downscaling factors. + * Pre-HSW bspec tells us to only consider the horizontal + * downscaling factor here. We ignore that and just consider + * both for simplicity. + */ + pixel_rate = crtc_state->pixel_rate; + + vlv_plane_ratio(crtc_state, plane_state, &num, &den); + + return DIV_ROUND_UP(pixel_rate * num, den); +} + static u32 vlv_sprite_ctl_crtc(const struct intel_crtc_state *crtc_state) { u32 sprctl = 0; @@ -1031,6 +1145,164 @@ vlv_plane_get_hw_state(struct intel_plane *plane, return ret; } +static void ivb_plane_ratio(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state, + unsigned int *num, unsigned int *den) +{ + u8 active_planes = crtc_state->active_planes & ~BIT(PLANE_CURSOR); + const struct drm_framebuffer *fb = plane_state->base.fb; + unsigned int cpp = fb->format->cpp[0]; + + if (hweight8(active_planes) == 2) { + switch (cpp) { + case 8: + *num = 10; + *den = 8; + break; + case 4: + *num = 17; + *den = 16; + break; + default: + *num = 1; + *den = 1; + break; + } + } else { + switch (cpp) { + case 8: + *num = 9; + *den = 8; + break; + default: + *num = 1; + *den = 1; + break; + } + } +} + +static void ivb_plane_ratio_scaling(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state, + unsigned int *num, unsigned int *den) +{ + const struct drm_framebuffer *fb = plane_state->base.fb; + unsigned int cpp = fb->format->cpp[0]; + + switch (cpp) { + case 8: + *num = 12; + *den = 8; + break; + case 4: + *num = 19; + *den = 16; + break; + case 2: + *num = 33; + *den = 32; + break; + default: + *num = 1; + *den = 1; + break; + } +} + +int ivb_plane_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) +{ + unsigned int pixel_rate; + unsigned int num, den; + + /* + * Note that crtc_state->pixel_rate accounts for both + * horizontal and vertical panel fitter downscaling factors. + * Pre-HSW bspec tells us to only consider the horizontal + * downscaling factor here. We ignore that and just consider + * both for simplicity. + */ + pixel_rate = crtc_state->pixel_rate; + + ivb_plane_ratio(crtc_state, plane_state, &num, &den); + + return DIV_ROUND_UP(pixel_rate * num, den); +} + +static int ivb_sprite_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) +{ + unsigned int src_w, dst_w, pixel_rate; + unsigned int num, den; + + /* + * Note that crtc_state->pixel_rate accounts for both + * horizontal and vertical panel fitter downscaling factors. + * Pre-HSW bspec tells us to only consider the horizontal + * downscaling factor here. We ignore that and just consider + * both for simplicity. + */ + pixel_rate = crtc_state->pixel_rate; + + src_w = drm_rect_width(&plane_state->base.src) >> 16; + dst_w = drm_rect_width(&plane_state->base.dst); + + if (src_w != dst_w) + ivb_plane_ratio_scaling(crtc_state, plane_state, &num, &den); + else + ivb_plane_ratio(crtc_state, plane_state, &num, &den); + + /* Horizontal downscaling limits the maximum pixel rate */ + dst_w = min(src_w, dst_w); + + return DIV_ROUND_UP_ULL(mul_u32_u32(pixel_rate, num * src_w), + den * dst_w); +} + +static void hsw_plane_ratio(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state, + unsigned int *num, unsigned int *den) +{ + u8 active_planes = crtc_state->active_planes & ~BIT(PLANE_CURSOR); + const struct drm_framebuffer *fb = plane_state->base.fb; + unsigned int cpp = fb->format->cpp[0]; + + if (hweight8(active_planes) == 2) { + switch (cpp) { + case 8: + *num = 10; + *den = 8; + break; + default: + *num = 1; + *den = 1; + break; + } + } else { + switch (cpp) { + case 8: + *num = 9; + *den = 8; + break; + default: + *num = 1; + *den = 1; + break; + } + } +} + +int hsw_plane_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) +{ + unsigned int pixel_rate = crtc_state->pixel_rate; + unsigned int num, den; + + hsw_plane_ratio(crtc_state, plane_state, &num, &den); + + return DIV_ROUND_UP(pixel_rate * num, den); +} + static u32 ivb_sprite_ctl_crtc(const struct intel_crtc_state *crtc_state) { u32 sprctl = 0; @@ -1044,6 +1316,16 @@ static u32 ivb_sprite_ctl_crtc(const struct intel_crtc_state *crtc_state) return sprctl; } +static bool ivb_need_sprite_gamma(const struct intel_plane_state *plane_state) +{ + struct drm_i915_private *dev_priv = + to_i915(plane_state->base.plane->dev); + const struct drm_framebuffer *fb = plane_state->base.fb; + + return fb->format->cpp[0] == 8 && + (IS_IVYBRIDGE(dev_priv) || IS_HASWELL(dev_priv)); +} + static u32 ivb_sprite_ctl(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state) { @@ -1066,6 +1348,12 @@ static u32 ivb_sprite_ctl(const struct intel_crtc_state *crtc_state, case DRM_FORMAT_XRGB8888: sprctl |= SPRITE_FORMAT_RGBX888; break; + case DRM_FORMAT_XBGR16161616F: + sprctl |= SPRITE_FORMAT_RGBX161616 | SPRITE_RGB_ORDER_RGBX; + break; + case DRM_FORMAT_XRGB16161616F: + sprctl |= SPRITE_FORMAT_RGBX161616; + break; case DRM_FORMAT_YUYV: sprctl |= SPRITE_FORMAT_YUV422 | SPRITE_YUV_ORDER_YUYV; break; @@ -1083,7 +1371,8 @@ static u32 ivb_sprite_ctl(const struct intel_crtc_state *crtc_state, return 0; } - sprctl |= SPRITE_INT_GAMMA_DISABLE; + if (!ivb_need_sprite_gamma(plane_state)) + sprctl |= SPRITE_INT_GAMMA_DISABLE; if (plane_state->base.color_encoding == DRM_COLOR_YCBCR_BT709) sprctl |= SPRITE_YUV_TO_RGB_CSC_FORMAT_BT709; @@ -1105,12 +1394,26 @@ static u32 ivb_sprite_ctl(const struct intel_crtc_state *crtc_state, return sprctl; } -static void ivb_sprite_linear_gamma(u16 gamma[18]) +static void ivb_sprite_linear_gamma(const struct intel_plane_state *plane_state, + u16 gamma[18]) { - int i; + int scale, i; - for (i = 0; i < 17; i++) - gamma[i] = (i << 10) / 16; + /* + * WaFP16GammaEnabling:ivb,hsw + * "Workaround : When using the 64-bit format, the sprite output + * on each color channel has one quarter amplitude. It can be + * brought up to full amplitude by using sprite internal gamma + * correction, pipe gamma correction, or pipe color space + * conversion to multiply the sprite output by four." + */ + scale = 4; + + for (i = 0; i < 16; i++) + gamma[i] = min((scale * i << 10) / 16, (1 << 10) - 1); + + gamma[i] = min((scale * i << 10) / 16, 1 << 10); + i++; gamma[i] = 3 << 10; i++; @@ -1124,7 +1427,10 @@ static void ivb_update_gamma(const struct intel_plane_state *plane_state) u16 gamma[18]; int i; - ivb_sprite_linear_gamma(gamma); + if (!ivb_need_sprite_gamma(plane_state)) + return; + + ivb_sprite_linear_gamma(plane_state, gamma); /* FIXME these register are single buffered :( */ for (i = 0; i < 16; i++) @@ -1257,6 +1563,53 @@ ivb_plane_get_hw_state(struct intel_plane *plane, return ret; } +static int g4x_sprite_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) +{ + const struct drm_framebuffer *fb = plane_state->base.fb; + unsigned int hscale, pixel_rate; + unsigned int limit, decimate; + + /* + * Note that crtc_state->pixel_rate accounts for both + * horizontal and vertical panel fitter downscaling factors. + * Pre-HSW bspec tells us to only consider the horizontal + * downscaling factor here. We ignore that and just consider + * both for simplicity. + */ + pixel_rate = crtc_state->pixel_rate; + + /* Horizontal downscaling limits the maximum pixel rate */ + hscale = drm_rect_calc_hscale(&plane_state->base.src, + &plane_state->base.dst, + 0, INT_MAX); + if (hscale < 0x10000) + return pixel_rate; + + /* Decimation steps at 2x,4x,8x,16x */ + decimate = ilog2(hscale >> 16); + hscale >>= decimate; + + /* Starting limit is 90% of cdclk */ + limit = 9; + + /* -10% per decimation step */ + limit -= decimate; + + /* -10% for RGB */ + if (fb->format->cpp[0] >= 4) + limit--; /* -10% for RGB */ + + /* + * We should also do -10% if sprite scaling is enabled + * on the other pipe, but we can't really check for that, + * so we ignore it. + */ + + return DIV_ROUND_UP_ULL(mul_u32_u32(pixel_rate, 10 * hscale), + limit << 16); +} + static unsigned int g4x_sprite_max_stride(struct intel_plane *plane, u32 pixel_format, u64 modifier, @@ -1300,6 +1653,12 @@ static u32 g4x_sprite_ctl(const struct intel_crtc_state *crtc_state, case DRM_FORMAT_XRGB8888: dvscntr |= DVS_FORMAT_RGBX888; break; + case DRM_FORMAT_XBGR16161616F: + dvscntr |= DVS_FORMAT_RGBX161616 | DVS_RGB_ORDER_XBGR; + break; + case DRM_FORMAT_XRGB16161616F: + dvscntr |= DVS_FORMAT_RGBX161616; + break; case DRM_FORMAT_YUYV: dvscntr |= DVS_FORMAT_YUV422 | DVS_YUV_ORDER_YUYV; break; @@ -1513,6 +1872,11 @@ static bool intel_fb_scalable(const struct drm_framebuffer *fb) switch (fb->format->format) { case DRM_FORMAT_C8: return false; + case DRM_FORMAT_XRGB16161616F: + case DRM_FORMAT_ARGB16161616F: + case DRM_FORMAT_XBGR16161616F: + case DRM_FORMAT_ABGR16161616F: + return INTEL_GEN(to_i915(fb->dev)) >= 11; default: return true; } @@ -1528,6 +1892,7 @@ g4x_sprite_check_scaling(struct intel_crtc_state *crtc_state, int src_x, src_w, src_h, crtc_w, crtc_h; const struct drm_display_mode *adjusted_mode = &crtc_state->base.adjusted_mode; + unsigned int stride = plane_state->color_plane[0].stride; unsigned int cpp = fb->format->cpp[0]; unsigned int width_bytes; int min_width, min_height; @@ -1569,9 +1934,9 @@ g4x_sprite_check_scaling(struct intel_crtc_state *crtc_state, return -EINVAL; } - if (width_bytes > 4096 || fb->pitches[0] > 4096) { + if (stride > 4096) { DRM_DEBUG_KMS("Stride (%u) exceeds hardware max with scaling (%u)\n", - fb->pitches[0], 4096); + stride, 4096); return -EINVAL; } @@ -1790,7 +2155,7 @@ static int skl_plane_check_nv12_rotation(const struct intel_plane_state *plane_s int src_w = drm_rect_width(&plane_state->base.src) >> 16; /* Display WA #1106 */ - if (is_planar_yuv_format(fb->format->format) && src_w & 3 && + if (drm_format_info_is_yuv_semiplanar(fb->format) && src_w & 3 && (rotation == DRM_MODE_ROTATE_270 || rotation == (DRM_MODE_REFLECT_X | DRM_MODE_ROTATE_90))) { DRM_DEBUG_KMS("src width must be multiple of 4 for rotated planar YUV\n"); @@ -1800,6 +2165,22 @@ static int skl_plane_check_nv12_rotation(const struct intel_plane_state *plane_s return 0; } +static int skl_plane_max_scale(struct drm_i915_private *dev_priv, + const struct drm_framebuffer *fb) +{ + /* + * We don't yet know the final source width nor + * whether we can use the HQ scaler mode. Assume + * the best case. + * FIXME need to properly check this later. + */ + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv) || + !drm_format_info_is_yuv_semiplanar(fb->format)) + return 0x30000 - 1; + else + return 0x20000 - 1; +} + static int skl_plane_check(struct intel_crtc_state *crtc_state, struct intel_plane_state *plane_state) { @@ -1817,7 +2198,7 @@ static int skl_plane_check(struct intel_crtc_state *crtc_state, /* use scaler when colorkey is not required */ if (!plane_state->ckey.flags && intel_fb_scalable(fb)) { min_scale = 1; - max_scale = skl_max_scale(crtc_state, fb->format->format); + max_scale = skl_plane_max_scale(dev_priv, fb); } ret = drm_atomic_helper_check_plane_state(&plane_state->base, @@ -1992,8 +2373,10 @@ static const u64 i9xx_plane_format_modifiers[] = { }; static const u32 snb_plane_formats[] = { - DRM_FORMAT_XBGR8888, DRM_FORMAT_XRGB8888, + DRM_FORMAT_XBGR8888, + DRM_FORMAT_XRGB16161616F, + DRM_FORMAT_XBGR16161616F, DRM_FORMAT_YUYV, DRM_FORMAT_YVYU, DRM_FORMAT_UYVY, @@ -2023,6 +2406,8 @@ static const u32 skl_plane_formats[] = { DRM_FORMAT_ABGR8888, DRM_FORMAT_XRGB2101010, DRM_FORMAT_XBGR2101010, + DRM_FORMAT_XRGB16161616F, + DRM_FORMAT_XBGR16161616F, DRM_FORMAT_YUYV, DRM_FORMAT_YVYU, DRM_FORMAT_UYVY, @@ -2038,6 +2423,8 @@ static const u32 skl_planar_formats[] = { DRM_FORMAT_ABGR8888, DRM_FORMAT_XRGB2101010, DRM_FORMAT_XBGR2101010, + DRM_FORMAT_XRGB16161616F, + DRM_FORMAT_XBGR16161616F, DRM_FORMAT_YUYV, DRM_FORMAT_YVYU, DRM_FORMAT_UYVY, @@ -2054,6 +2441,8 @@ static const u32 glk_planar_formats[] = { DRM_FORMAT_ABGR8888, DRM_FORMAT_XRGB2101010, DRM_FORMAT_XBGR2101010, + DRM_FORMAT_XRGB16161616F, + DRM_FORMAT_XBGR16161616F, DRM_FORMAT_YUYV, DRM_FORMAT_YVYU, DRM_FORMAT_UYVY, @@ -2157,6 +2546,13 @@ static const u64 skl_plane_format_modifiers_ccs[] = { DRM_FORMAT_MOD_INVALID }; +static const u64 gen12_plane_format_modifiers_noccs[] = { + I915_FORMAT_MOD_Y_TILED, + I915_FORMAT_MOD_X_TILED, + DRM_FORMAT_MOD_LINEAR, + DRM_FORMAT_MOD_INVALID +}; + static bool g4x_sprite_format_mod_supported(struct drm_plane *_plane, u32 format, u64 modifier) { @@ -2197,6 +2593,8 @@ static bool snb_sprite_format_mod_supported(struct drm_plane *_plane, switch (format) { case DRM_FORMAT_XRGB8888: case DRM_FORMAT_XBGR8888: + case DRM_FORMAT_XRGB16161616F: + case DRM_FORMAT_XBGR16161616F: case DRM_FORMAT_YUYV: case DRM_FORMAT_YVYU: case DRM_FORMAT_UYVY: @@ -2305,6 +2703,55 @@ static bool skl_plane_format_mod_supported(struct drm_plane *_plane, } } +static bool gen12_plane_format_mod_supported(struct drm_plane *_plane, + u32 format, u64 modifier) +{ + switch (modifier) { + case DRM_FORMAT_MOD_LINEAR: + case I915_FORMAT_MOD_X_TILED: + case I915_FORMAT_MOD_Y_TILED: + break; + default: + return false; + } + + switch (format) { + case DRM_FORMAT_XRGB8888: + case DRM_FORMAT_XBGR8888: + case DRM_FORMAT_ARGB8888: + case DRM_FORMAT_ABGR8888: + case DRM_FORMAT_RGB565: + case DRM_FORMAT_XRGB2101010: + case DRM_FORMAT_XBGR2101010: + case DRM_FORMAT_YUYV: + case DRM_FORMAT_YVYU: + case DRM_FORMAT_UYVY: + case DRM_FORMAT_VYUY: + case DRM_FORMAT_NV12: + case DRM_FORMAT_P010: + case DRM_FORMAT_P012: + case DRM_FORMAT_P016: + case DRM_FORMAT_XVYU2101010: + case DRM_FORMAT_C8: + case DRM_FORMAT_XBGR16161616F: + case DRM_FORMAT_ABGR16161616F: + case DRM_FORMAT_XRGB16161616F: + case DRM_FORMAT_ARGB16161616F: + case DRM_FORMAT_Y210: + case DRM_FORMAT_Y212: + case DRM_FORMAT_Y216: + case DRM_FORMAT_XVYU12_16161616: + case DRM_FORMAT_XVYU16161616: + if (modifier == DRM_FORMAT_MOD_LINEAR || + modifier == I915_FORMAT_MOD_X_TILED || + modifier == I915_FORMAT_MOD_Y_TILED) + return true; + /* fall through */ + default: + return false; + } +} + static const struct drm_plane_funcs g4x_sprite_funcs = { .update_plane = drm_atomic_helper_update_plane, .disable_plane = drm_atomic_helper_disable_plane, @@ -2341,6 +2788,15 @@ static const struct drm_plane_funcs skl_plane_funcs = { .format_mod_supported = skl_plane_format_mod_supported, }; +static const struct drm_plane_funcs gen12_plane_funcs = { + .update_plane = drm_atomic_helper_update_plane, + .disable_plane = drm_atomic_helper_disable_plane, + .destroy = intel_plane_destroy, + .atomic_duplicate_state = intel_plane_duplicate_state, + .atomic_destroy_state = intel_plane_destroy_state, + .format_mod_supported = gen12_plane_format_mod_supported, +}; + static bool skl_plane_has_fbc(struct drm_i915_private *dev_priv, enum pipe pipe, enum plane_id plane_id) { @@ -2429,6 +2885,7 @@ struct intel_plane * skl_universal_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe, enum plane_id plane_id) { + const struct drm_plane_funcs *plane_funcs; struct intel_plane *plane; enum drm_plane_type plane_type; unsigned int supported_rotations; @@ -2458,6 +2915,7 @@ skl_universal_plane_create(struct drm_i915_private *dev_priv, plane->disable_plane = skl_disable_plane; plane->get_hw_state = skl_plane_get_hw_state; plane->check_plane = skl_plane_check; + plane->min_cdclk = skl_plane_min_cdclk; if (icl_is_nv12_y_plane(plane_id)) plane->update_slave = icl_update_slave; @@ -2471,11 +2929,19 @@ skl_universal_plane_create(struct drm_i915_private *dev_priv, formats = skl_get_plane_formats(dev_priv, pipe, plane_id, &num_formats); - plane->has_ccs = skl_plane_has_ccs(dev_priv, pipe, plane_id); - if (plane->has_ccs) - modifiers = skl_plane_format_modifiers_ccs; - else - modifiers = skl_plane_format_modifiers_noccs; + if (INTEL_GEN(dev_priv) >= 12) { + /* TODO: Implement support for gen-12 CCS modifiers */ + plane->has_ccs = false; + modifiers = gen12_plane_format_modifiers_noccs; + plane_funcs = &gen12_plane_funcs; + } else { + plane->has_ccs = skl_plane_has_ccs(dev_priv, pipe, plane_id); + if (plane->has_ccs) + modifiers = skl_plane_format_modifiers_ccs; + else + modifiers = skl_plane_format_modifiers_noccs; + plane_funcs = &skl_plane_funcs; + } if (plane_id == PLANE_PRIMARY) plane_type = DRM_PLANE_TYPE_PRIMARY; @@ -2485,7 +2951,7 @@ skl_universal_plane_create(struct drm_i915_private *dev_priv, possible_crtcs = BIT(pipe); ret = drm_universal_plane_init(&dev_priv->drm, &plane->base, - possible_crtcs, &skl_plane_funcs, + possible_crtcs, plane_funcs, formats, num_formats, modifiers, plane_type, "plane %d%c", plane_id + 1, @@ -2518,6 +2984,8 @@ skl_universal_plane_create(struct drm_i915_private *dev_priv, BIT(DRM_MODE_BLEND_PREMULTI) | BIT(DRM_MODE_BLEND_COVERAGE)); + drm_plane_create_zpos_immutable_property(&plane->base, plane_id); + drm_plane_helper_add(&plane->base, &intel_plane_helper_funcs); return plane; @@ -2539,7 +3007,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, const u64 *modifiers; const u32 *formats; int num_formats; - int ret; + int ret, zpos; if (INTEL_GEN(dev_priv) >= 9) return skl_universal_plane_create(dev_priv, pipe, @@ -2555,6 +3023,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, plane->disable_plane = vlv_disable_plane; plane->get_hw_state = vlv_plane_get_hw_state; plane->check_plane = vlv_sprite_check; + plane->min_cdclk = vlv_plane_min_cdclk; formats = vlv_plane_formats; num_formats = ARRAY_SIZE(vlv_plane_formats); @@ -2568,6 +3037,11 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, plane->get_hw_state = ivb_plane_get_hw_state; plane->check_plane = g4x_sprite_check; + if (IS_BROADWELL(dev_priv) || IS_HASWELL(dev_priv)) + plane->min_cdclk = hsw_plane_min_cdclk; + else + plane->min_cdclk = ivb_sprite_min_cdclk; + formats = snb_plane_formats; num_formats = ARRAY_SIZE(snb_plane_formats); modifiers = i9xx_plane_format_modifiers; @@ -2579,6 +3053,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, plane->disable_plane = g4x_disable_plane; plane->get_hw_state = g4x_plane_get_hw_state; plane->check_plane = g4x_sprite_check; + plane->min_cdclk = g4x_sprite_min_cdclk; modifiers = i9xx_plane_format_modifiers; if (IS_GEN(dev_priv, 6)) { @@ -2629,6 +3104,9 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, DRM_COLOR_YCBCR_BT709, DRM_COLOR_YCBCR_LIMITED_RANGE); + zpos = sprite + 1; + drm_plane_create_zpos_immutable_property(&plane->base, zpos); + drm_plane_helper_add(&plane->base, &intel_plane_helper_funcs); return plane; diff --git a/drivers/gpu/drm/i915/display/intel_sprite.h b/drivers/gpu/drm/i915/display/intel_sprite.h index 093a2d156f1e..5eeaa92420d1 100644 --- a/drivers/gpu/drm/i915/display/intel_sprite.h +++ b/drivers/gpu/drm/i915/display/intel_sprite.h @@ -17,7 +17,6 @@ struct drm_i915_private; struct intel_crtc_state; struct intel_plane_state; -bool is_planar_yuv_format(u32 pixelformat); int intel_usecs_to_scanlines(const struct drm_display_mode *adjusted_mode, int usecs); struct intel_plane *intel_sprite_plane_create(struct drm_i915_private *dev_priv, @@ -50,4 +49,11 @@ static inline u8 icl_hdr_plane_mask(void) bool icl_is_hdr_plane(struct drm_i915_private *dev_priv, enum plane_id plane_id); +int ivb_plane_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state); +int hsw_plane_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state); +int vlv_plane_min_cdclk(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state); + #endif /* __INTEL_SPRITE_H__ */ diff --git a/drivers/gpu/drm/i915/display/intel_tc.c b/drivers/gpu/drm/i915/display/intel_tc.c index 85743a43bee2..7773169b7331 100644 --- a/drivers/gpu/drm/i915/display/intel_tc.c +++ b/drivers/gpu/drm/i915/display/intel_tc.c @@ -23,32 +23,38 @@ static const char *tc_port_mode_name(enum tc_port_mode mode) return names[mode]; } -static bool has_modular_fia(struct drm_i915_private *i915) -{ - if (!INTEL_INFO(i915)->display.has_modular_fia) - return false; - - return intel_uncore_read(&i915->uncore, - PORT_TX_DFLEXDPSP(FIA1)) & MODULAR_FIA_MASK; -} - -static enum phy_fia tc_port_to_fia(struct drm_i915_private *i915, - enum tc_port tc_port) +static void +tc_port_load_fia_params(struct drm_i915_private *i915, + struct intel_digital_port *dig_port) { - if (!has_modular_fia(i915)) - return FIA1; + enum port port = dig_port->base.port; + enum tc_port tc_port = intel_port_to_tc(i915, port); + u32 modular_fia; + + if (INTEL_INFO(i915)->display.has_modular_fia) { + modular_fia = intel_uncore_read(&i915->uncore, + PORT_TX_DFLEXDPSP(FIA1)); + modular_fia &= MODULAR_FIA_MASK; + } else { + modular_fia = 0; + } /* * Each Modular FIA instance houses 2 TC ports. In SOC that has more * than two TC ports, there are multiple instances of Modular FIA. */ - return tc_port / 2; + if (modular_fia) { + dig_port->tc_phy_fia = tc_port / 2; + dig_port->tc_phy_fia_idx = tc_port % 2; + } else { + dig_port->tc_phy_fia = FIA1; + dig_port->tc_phy_fia_idx = tc_port; + } } u32 intel_tc_port_get_lane_mask(struct intel_digital_port *dig_port) { struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev); - enum tc_port tc_port = intel_port_to_tc(i915, dig_port->base.port); struct intel_uncore *uncore = &i915->uncore; u32 lane_mask; @@ -57,8 +63,23 @@ u32 intel_tc_port_get_lane_mask(struct intel_digital_port *dig_port) WARN_ON(lane_mask == 0xffffffff); - return (lane_mask & DP_LANE_ASSIGNMENT_MASK(tc_port)) >> - DP_LANE_ASSIGNMENT_SHIFT(tc_port); + lane_mask &= DP_LANE_ASSIGNMENT_MASK(dig_port->tc_phy_fia_idx); + return lane_mask >> DP_LANE_ASSIGNMENT_SHIFT(dig_port->tc_phy_fia_idx); +} + +u32 intel_tc_port_get_pin_assignment_mask(struct intel_digital_port *dig_port) +{ + struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev); + struct intel_uncore *uncore = &i915->uncore; + u32 pin_mask; + + pin_mask = intel_uncore_read(uncore, + PORT_TX_DFLEXPA1(dig_port->tc_phy_fia)); + + WARN_ON(pin_mask == 0xffffffff); + + return (pin_mask & DP_PIN_ASSIGNMENT_MASK(dig_port->tc_phy_fia_idx)) >> + DP_PIN_ASSIGNMENT_SHIFT(dig_port->tc_phy_fia_idx); } int intel_tc_port_fia_max_lane_count(struct intel_digital_port *dig_port) @@ -95,7 +116,6 @@ void intel_tc_port_set_fia_lane_count(struct intel_digital_port *dig_port, int required_lanes) { struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev); - enum tc_port tc_port = intel_port_to_tc(i915, dig_port->base.port); bool lane_reversal = dig_port->saved_port_bits & DDI_BUF_PORT_REVERSAL; struct intel_uncore *uncore = &i915->uncore; u32 val; @@ -104,19 +124,21 @@ void intel_tc_port_set_fia_lane_count(struct intel_digital_port *dig_port, val = intel_uncore_read(uncore, PORT_TX_DFLEXDPMLE1(dig_port->tc_phy_fia)); - val &= ~DFLEXDPMLE1_DPMLETC_MASK(tc_port); + val &= ~DFLEXDPMLE1_DPMLETC_MASK(dig_port->tc_phy_fia_idx); switch (required_lanes) { case 1: - val |= lane_reversal ? DFLEXDPMLE1_DPMLETC_ML3(tc_port) : - DFLEXDPMLE1_DPMLETC_ML0(tc_port); + val |= lane_reversal ? + DFLEXDPMLE1_DPMLETC_ML3(dig_port->tc_phy_fia_idx) : + DFLEXDPMLE1_DPMLETC_ML0(dig_port->tc_phy_fia_idx); break; case 2: - val |= lane_reversal ? DFLEXDPMLE1_DPMLETC_ML3_2(tc_port) : - DFLEXDPMLE1_DPMLETC_ML1_0(tc_port); + val |= lane_reversal ? + DFLEXDPMLE1_DPMLETC_ML3_2(dig_port->tc_phy_fia_idx) : + DFLEXDPMLE1_DPMLETC_ML1_0(dig_port->tc_phy_fia_idx); break; case 4: - val |= DFLEXDPMLE1_DPMLETC_ML3_0(tc_port); + val |= DFLEXDPMLE1_DPMLETC_ML3_0(dig_port->tc_phy_fia_idx); break; default: MISSING_CASE(required_lanes); @@ -164,9 +186,9 @@ static u32 tc_port_live_status_mask(struct intel_digital_port *dig_port) return mask; } - if (val & TC_LIVE_STATE_TBT(tc_port)) + if (val & TC_LIVE_STATE_TBT(dig_port->tc_phy_fia_idx)) mask |= BIT(TC_PORT_TBT_ALT); - if (val & TC_LIVE_STATE_TC(tc_port)) + if (val & TC_LIVE_STATE_TC(dig_port->tc_phy_fia_idx)) mask |= BIT(TC_PORT_DP_ALT); if (intel_uncore_read(uncore, SDEISR) & SDE_TC_HOTPLUG_ICP(tc_port)) @@ -182,7 +204,6 @@ static u32 tc_port_live_status_mask(struct intel_digital_port *dig_port) static bool icl_tc_phy_status_complete(struct intel_digital_port *dig_port) { struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev); - enum tc_port tc_port = intel_port_to_tc(i915, dig_port->base.port); struct intel_uncore *uncore = &i915->uncore; u32 val; @@ -194,14 +215,13 @@ static bool icl_tc_phy_status_complete(struct intel_digital_port *dig_port) return false; } - return val & DP_PHY_MODE_STATUS_COMPLETED(tc_port); + return val & DP_PHY_MODE_STATUS_COMPLETED(dig_port->tc_phy_fia_idx); } static bool icl_tc_phy_set_safe_mode(struct intel_digital_port *dig_port, bool enable) { struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev); - enum tc_port tc_port = intel_port_to_tc(i915, dig_port->base.port); struct intel_uncore *uncore = &i915->uncore; u32 val; @@ -215,9 +235,9 @@ static bool icl_tc_phy_set_safe_mode(struct intel_digital_port *dig_port, return false; } - val &= ~DP_PHY_MODE_STATUS_NOT_SAFE(tc_port); + val &= ~DP_PHY_MODE_STATUS_NOT_SAFE(dig_port->tc_phy_fia_idx); if (!enable) - val |= DP_PHY_MODE_STATUS_NOT_SAFE(tc_port); + val |= DP_PHY_MODE_STATUS_NOT_SAFE(dig_port->tc_phy_fia_idx); intel_uncore_write(uncore, PORT_TX_DFLEXDPCSSS(dig_port->tc_phy_fia), val); @@ -232,7 +252,6 @@ static bool icl_tc_phy_set_safe_mode(struct intel_digital_port *dig_port, static bool icl_tc_phy_is_in_safe_mode(struct intel_digital_port *dig_port) { struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev); - enum tc_port tc_port = intel_port_to_tc(i915, dig_port->base.port); struct intel_uncore *uncore = &i915->uncore; u32 val; @@ -244,7 +263,7 @@ static bool icl_tc_phy_is_in_safe_mode(struct intel_digital_port *dig_port) return true; } - return !(val & DP_PHY_MODE_STATUS_NOT_SAFE(tc_port)); + return !(val & DP_PHY_MODE_STATUS_NOT_SAFE(dig_port->tc_phy_fia_idx)); } /* @@ -540,5 +559,5 @@ void intel_tc_port_init(struct intel_digital_port *dig_port, bool is_legacy) mutex_init(&dig_port->tc_lock); dig_port->tc_legacy_port = is_legacy; dig_port->tc_link_refcount = 0; - dig_port->tc_phy_fia = tc_port_to_fia(i915, tc_port); + tc_port_load_fia_params(i915, dig_port); } diff --git a/drivers/gpu/drm/i915/display/intel_tc.h b/drivers/gpu/drm/i915/display/intel_tc.h index 783d75531435..463f1b3c836f 100644 --- a/drivers/gpu/drm/i915/display/intel_tc.h +++ b/drivers/gpu/drm/i915/display/intel_tc.h @@ -13,6 +13,7 @@ struct intel_digital_port; bool intel_tc_port_connected(struct intel_digital_port *dig_port); u32 intel_tc_port_get_lane_mask(struct intel_digital_port *dig_port); +u32 intel_tc_port_get_pin_assignment_mask(struct intel_digital_port *dig_port); int intel_tc_port_fia_max_lane_count(struct intel_digital_port *dig_port); void intel_tc_port_set_fia_lane_count(struct intel_digital_port *dig_port, int required_lanes); diff --git a/drivers/gpu/drm/i915/display/intel_tv.c b/drivers/gpu/drm/i915/display/intel_tv.c index b70221f5112a..9983fadf6c28 100644 --- a/drivers/gpu/drm/i915/display/intel_tv.c +++ b/drivers/gpu/drm/i915/display/intel_tv.c @@ -961,11 +961,10 @@ intel_tv_mode_valid(struct drm_connector *connector, return MODE_CLOCK_HIGH; /* Ensure TV refresh is close to desired refresh */ - if (tv_mode && abs(tv_mode->refresh - drm_mode_vrefresh(mode) * 1000) - < 1000) - return MODE_OK; + if (abs(tv_mode->refresh - drm_mode_vrefresh(mode) * 1000) >= 1000) + return MODE_CLOCK_RANGE; - return MODE_CLOCK_RANGE; + return MODE_OK; } static int @@ -1702,7 +1701,7 @@ intel_tv_detect(struct drm_connector *connector, struct intel_load_detect_pipe tmp; int ret; - ret = intel_get_load_detect_pipe(connector, NULL, &tmp, ctx); + ret = intel_get_load_detect_pipe(connector, &tmp, ctx); if (ret < 0) return ret; @@ -1948,9 +1947,8 @@ intel_tv_init(struct drm_i915_private *dev_priv) intel_encoder->type = INTEL_OUTPUT_TVOUT; intel_encoder->power_domain = POWER_DOMAIN_PORT_OTHER; intel_encoder->port = PORT_NONE; - intel_encoder->crtc_mask = (1 << 0) | (1 << 1); + intel_encoder->pipe_mask = ~0; intel_encoder->cloneable = 0; - intel_encoder->base.possible_crtcs = ((1 << 0) | (1 << 1)); intel_tv->type = DRM_MODE_CONNECTOR_Unknown; /* BIOS margin values */ diff --git a/drivers/gpu/drm/i915/display/intel_vbt_defs.h b/drivers/gpu/drm/i915/display/intel_vbt_defs.h index dfcd156b5094..69a7cb1fa121 100644 --- a/drivers/gpu/drm/i915/display/intel_vbt_defs.h +++ b/drivers/gpu/drm/i915/display/intel_vbt_defs.h @@ -114,6 +114,7 @@ enum bdb_block_id { BDB_LVDS_POWER = 44, BDB_MIPI_CONFIG = 52, BDB_MIPI_SEQUENCE = 53, + BDB_COMPRESSION_PARAMETERS = 56, BDB_SKIP = 254, /* VBIOS private block, ignore */ }; @@ -291,6 +292,8 @@ struct bdb_general_features { #define DVO_PORT_HDMIE 12 /* 193 */ #define DVO_PORT_DPF 13 /* N/A */ #define DVO_PORT_HDMIF 14 /* N/A */ +#define DVO_PORT_DPG 15 +#define DVO_PORT_HDMIG 16 #define DVO_PORT_MIPIA 21 /* 171 */ #define DVO_PORT_MIPIB 22 /* 171 */ #define DVO_PORT_MIPIC 23 /* 171 */ @@ -325,6 +328,7 @@ enum vbt_gmbus_ddi { #define DP_AUX_D 0x30 #define DP_AUX_E 0x50 #define DP_AUX_F 0x60 +#define DP_AUX_G 0x70 #define VBT_DP_MAX_LINK_RATE_HBR3 0 #define VBT_DP_MAX_LINK_RATE_HBR2 1 @@ -808,4 +812,55 @@ struct bdb_mipi_sequence { u8 data[0]; /* up to 6 variable length blocks */ } __packed; +/* + * Block 56 - Compression Parameters + */ + +#define VBT_RC_BUFFER_BLOCK_SIZE_1KB 0 +#define VBT_RC_BUFFER_BLOCK_SIZE_4KB 1 +#define VBT_RC_BUFFER_BLOCK_SIZE_16KB 2 +#define VBT_RC_BUFFER_BLOCK_SIZE_64KB 3 + +#define VBT_DSC_LINE_BUFFER_DEPTH(vbt_value) ((vbt_value) + 8) /* bits */ +#define VBT_DSC_MAX_BPP(vbt_value) (6 + (vbt_value) * 2) + +struct dsc_compression_parameters_entry { + u8 version_major:4; + u8 version_minor:4; + + u8 rc_buffer_block_size:2; + u8 reserved1:6; + + /* + * Buffer size in bytes: + * + * 4 ^ rc_buffer_block_size * 1024 * (rc_buffer_size + 1) bytes + */ + u8 rc_buffer_size; + u32 slices_per_line; + + u8 line_buffer_depth:4; + u8 reserved2:4; + + /* Flag Bits 1 */ + u8 block_prediction_enable:1; + u8 reserved3:7; + + u8 max_bpp; /* mapping */ + + /* Color depth capabilities */ + u8 reserved4:1; + u8 support_8bpc:1; + u8 support_10bpc:1; + u8 support_12bpc:1; + u8 reserved5:4; + + u16 slice_height; +} __packed; + +struct bdb_compression_parameters { + u16 entry_size; + struct dsc_compression_parameters_entry data[16]; +} __packed; + #endif /* _INTEL_VBT_DEFS_H_ */ diff --git a/drivers/gpu/drm/i915/display/intel_vdsc.c b/drivers/gpu/drm/i915/display/intel_vdsc.c index d4fb7f16f9f6..896b0c334f5e 100644 --- a/drivers/gpu/drm/i915/display/intel_vdsc.c +++ b/drivers/gpu/drm/i915/display/intel_vdsc.c @@ -322,8 +322,8 @@ static int get_column_index_for_rc_params(u8 bits_per_component) int intel_dp_compute_dsc_params(struct intel_dp *intel_dp, struct intel_crtc_state *pipe_config) { - struct drm_dsc_config *vdsc_cfg = &pipe_config->dp_dsc_cfg; - u16 compressed_bpp = pipe_config->dsc_params.compressed_bpp; + struct drm_dsc_config *vdsc_cfg = &pipe_config->dsc.config; + u16 compressed_bpp = pipe_config->dsc.compressed_bpp; u8 i = 0; int row_index = 0; int column_index = 0; @@ -332,7 +332,7 @@ int intel_dp_compute_dsc_params(struct intel_dp *intel_dp, vdsc_cfg->pic_width = pipe_config->base.adjusted_mode.crtc_hdisplay; vdsc_cfg->pic_height = pipe_config->base.adjusted_mode.crtc_vdisplay; vdsc_cfg->slice_width = DIV_ROUND_UP(vdsc_cfg->pic_width, - pipe_config->dsc_params.slice_count); + pipe_config->dsc.slice_count); /* * Slice Height of 8 works for all currently available panels. So start * with that if pic_height is an integral multiple of 8. @@ -485,13 +485,13 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - const struct drm_dsc_config *vdsc_cfg = &crtc_state->dp_dsc_cfg; + const struct drm_dsc_config *vdsc_cfg = &crtc_state->dsc.config; enum pipe pipe = crtc->pipe; enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; u32 pps_val = 0; u32 rc_buf_thresh_dword[4]; u32 rc_range_params_dword[8]; - u8 num_vdsc_instances = (crtc_state->dsc_params.dsc_split) ? 2 : 1; + u8 num_vdsc_instances = (crtc_state->dsc.dsc_split) ? 2 : 1; int i = 0; /* Populate PICTURE_PARAMETER_SET_0 registers */ @@ -514,11 +514,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_0, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_0(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_0(pipe), pps_val); } @@ -533,11 +533,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_1, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_1(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_1(pipe), pps_val); } @@ -553,11 +553,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_2, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_2(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_2(pipe), pps_val); } @@ -573,11 +573,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_3, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_3(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_3(pipe), pps_val); } @@ -593,11 +593,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_4, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_4(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_4(pipe), pps_val); } @@ -613,11 +613,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_5, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_5(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_5(pipe), pps_val); } @@ -635,11 +635,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_6, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_6(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_6(pipe), pps_val); } @@ -655,11 +655,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_7, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_7(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_7(pipe), pps_val); } @@ -675,11 +675,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_8, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_8(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_8(pipe), pps_val); } @@ -695,11 +695,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_9, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_9(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_9(pipe), pps_val); } @@ -717,11 +717,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_10, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_10(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_10(pipe), pps_val); } @@ -740,11 +740,11 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, * If 2 VDSC instances are needed, configure PPS for second * VDSC */ - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(DSCC_PICTURE_PARAMETER_SET_16, pps_val); } else { I915_WRITE(ICL_DSC0_PICTURE_PARAMETER_SET_16(pipe), pps_val); - if (crtc_state->dsc_params.dsc_split) + if (crtc_state->dsc.dsc_split) I915_WRITE(ICL_DSC1_PICTURE_PARAMETER_SET_16(pipe), pps_val); } @@ -763,7 +763,7 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, I915_WRITE(DSCA_RC_BUF_THRESH_0_UDW, rc_buf_thresh_dword[1]); I915_WRITE(DSCA_RC_BUF_THRESH_1, rc_buf_thresh_dword[2]); I915_WRITE(DSCA_RC_BUF_THRESH_1_UDW, rc_buf_thresh_dword[3]); - if (crtc_state->dsc_params.dsc_split) { + if (crtc_state->dsc.dsc_split) { I915_WRITE(DSCC_RC_BUF_THRESH_0, rc_buf_thresh_dword[0]); I915_WRITE(DSCC_RC_BUF_THRESH_0_UDW, @@ -782,7 +782,7 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, rc_buf_thresh_dword[2]); I915_WRITE(ICL_DSC0_RC_BUF_THRESH_1_UDW(pipe), rc_buf_thresh_dword[3]); - if (crtc_state->dsc_params.dsc_split) { + if (crtc_state->dsc.dsc_split) { I915_WRITE(ICL_DSC1_RC_BUF_THRESH_0(pipe), rc_buf_thresh_dword[0]); I915_WRITE(ICL_DSC1_RC_BUF_THRESH_0_UDW(pipe), @@ -824,7 +824,7 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, rc_range_params_dword[6]); I915_WRITE(DSCA_RC_RANGE_PARAMETERS_3_UDW, rc_range_params_dword[7]); - if (crtc_state->dsc_params.dsc_split) { + if (crtc_state->dsc.dsc_split) { I915_WRITE(DSCC_RC_RANGE_PARAMETERS_0, rc_range_params_dword[0]); I915_WRITE(DSCC_RC_RANGE_PARAMETERS_0_UDW, @@ -859,7 +859,7 @@ static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder, rc_range_params_dword[6]); I915_WRITE(ICL_DSC0_RC_RANGE_PARAMETERS_3_UDW(pipe), rc_range_params_dword[7]); - if (crtc_state->dsc_params.dsc_split) { + if (crtc_state->dsc.dsc_split) { I915_WRITE(ICL_DSC1_RC_RANGE_PARAMETERS_0(pipe), rc_range_params_dword[0]); I915_WRITE(ICL_DSC1_RC_RANGE_PARAMETERS_0_UDW(pipe), @@ -885,7 +885,7 @@ static void intel_dp_write_dsc_pps_sdp(struct intel_encoder *encoder, { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - const struct drm_dsc_config *vdsc_cfg = &crtc_state->dp_dsc_cfg; + const struct drm_dsc_config *vdsc_cfg = &crtc_state->dsc.config; struct drm_dsc_pps_infoframe dp_dsc_pps_sdp; /* Prepare DP SDP PPS header as per DP 1.4 spec, Table 2-123 */ @@ -909,7 +909,7 @@ void intel_dsc_enable(struct intel_encoder *encoder, u32 dss_ctl1_val = 0; u32 dss_ctl2_val = 0; - if (!crtc_state->dsc_params.compression_enable) + if (!crtc_state->dsc.compression_enable) return; /* Enable Power wells for VDSC/joining */ @@ -928,7 +928,7 @@ void intel_dsc_enable(struct intel_encoder *encoder, dss_ctl2_reg = ICL_PIPE_DSS_CTL2(pipe); } dss_ctl2_val |= LEFT_BRANCH_VDSC_ENABLE; - if (crtc_state->dsc_params.dsc_split) { + if (crtc_state->dsc.dsc_split) { dss_ctl2_val |= RIGHT_BRANCH_VDSC_ENABLE; dss_ctl1_val |= JOINER_ENABLE; } @@ -944,7 +944,7 @@ void intel_dsc_disable(const struct intel_crtc_state *old_crtc_state) i915_reg_t dss_ctl1_reg, dss_ctl2_reg; u32 dss_ctl1_val = 0, dss_ctl2_val = 0; - if (!old_crtc_state->dsc_params.compression_enable) + if (!old_crtc_state->dsc.compression_enable) return; if (old_crtc_state->cpu_transcoder == TRANSCODER_EDP) { diff --git a/drivers/gpu/drm/i915/display/intel_vga.c b/drivers/gpu/drm/i915/display/intel_vga.c new file mode 100644 index 000000000000..2ff7293986d4 --- /dev/null +++ b/drivers/gpu/drm/i915/display/intel_vga.c @@ -0,0 +1,160 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include <linux/pci.h> +#include <linux/vgaarb.h> + +#include <drm/i915_drm.h> + +#include "i915_drv.h" +#include "intel_vga.h" + +static i915_reg_t intel_vga_cntrl_reg(struct drm_i915_private *i915) +{ + if (IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915)) + return VLV_VGACNTRL; + else if (INTEL_GEN(i915) >= 5) + return CPU_VGACNTRL; + else + return VGACNTRL; +} + +/* Disable the VGA plane that we never use */ +void intel_vga_disable(struct drm_i915_private *dev_priv) +{ + struct pci_dev *pdev = dev_priv->drm.pdev; + i915_reg_t vga_reg = intel_vga_cntrl_reg(dev_priv); + u8 sr1; + + /* WaEnableVGAAccessThroughIOPort:ctg,elk,ilk,snb,ivb,vlv,hsw */ + vga_get_uninterruptible(pdev, VGA_RSRC_LEGACY_IO); + outb(SR01, VGA_SR_INDEX); + sr1 = inb(VGA_SR_DATA); + outb(sr1 | 1 << 5, VGA_SR_DATA); + vga_put(pdev, VGA_RSRC_LEGACY_IO); + udelay(300); + + I915_WRITE(vga_reg, VGA_DISP_DISABLE); + POSTING_READ(vga_reg); +} + +void intel_vga_redisable_power_on(struct drm_i915_private *dev_priv) +{ + i915_reg_t vga_reg = intel_vga_cntrl_reg(dev_priv); + + if (!(I915_READ(vga_reg) & VGA_DISP_DISABLE)) { + DRM_DEBUG_KMS("Something enabled VGA plane, disabling it\n"); + intel_vga_disable(dev_priv); + } +} + +void intel_vga_redisable(struct drm_i915_private *i915) +{ + intel_wakeref_t wakeref; + + /* + * This function can be called both from intel_modeset_setup_hw_state or + * at a very early point in our resume sequence, where the power well + * structures are not yet restored. Since this function is at a very + * paranoid "someone might have enabled VGA while we were not looking" + * level, just check if the power well is enabled instead of trying to + * follow the "don't touch the power well if we don't need it" policy + * the rest of the driver uses. + */ + wakeref = intel_display_power_get_if_enabled(i915, POWER_DOMAIN_VGA); + if (!wakeref) + return; + + intel_vga_redisable_power_on(i915); + + intel_display_power_put(i915, POWER_DOMAIN_VGA, wakeref); +} + +void intel_vga_reset_io_mem(struct drm_i915_private *i915) +{ + struct pci_dev *pdev = i915->drm.pdev; + + /* + * After we re-enable the power well, if we touch VGA register 0x3d5 + * we'll get unclaimed register interrupts. This stops after we write + * anything to the VGA MSR register. The vgacon module uses this + * register all the time, so if we unbind our driver and, as a + * consequence, bind vgacon, we'll get stuck in an infinite loop at + * console_unlock(). So make here we touch the VGA MSR register, making + * sure vgacon can keep working normally without triggering interrupts + * and error messages. + */ + vga_get_uninterruptible(pdev, VGA_RSRC_LEGACY_IO); + outb(inb(VGA_MSR_READ), VGA_MSR_WRITE); + vga_put(pdev, VGA_RSRC_LEGACY_IO); +} + +static int +intel_vga_set_state(struct drm_i915_private *i915, bool enable_decode) +{ + unsigned int reg = INTEL_GEN(i915) >= 6 ? SNB_GMCH_CTRL : INTEL_GMCH_CTRL; + u16 gmch_ctrl; + + if (pci_read_config_word(i915->bridge_dev, reg, &gmch_ctrl)) { + DRM_ERROR("failed to read control word\n"); + return -EIO; + } + + if (!!(gmch_ctrl & INTEL_GMCH_VGA_DISABLE) == !enable_decode) + return 0; + + if (enable_decode) + gmch_ctrl &= ~INTEL_GMCH_VGA_DISABLE; + else + gmch_ctrl |= INTEL_GMCH_VGA_DISABLE; + + if (pci_write_config_word(i915->bridge_dev, reg, gmch_ctrl)) { + DRM_ERROR("failed to write control word\n"); + return -EIO; + } + + return 0; +} + +static unsigned int +intel_vga_set_decode(void *cookie, bool enable_decode) +{ + struct drm_i915_private *i915 = cookie; + + intel_vga_set_state(i915, enable_decode); + + if (enable_decode) + return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM | + VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM; + else + return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM; +} + +int intel_vga_register(struct drm_i915_private *i915) +{ + struct pci_dev *pdev = i915->drm.pdev; + int ret; + + /* + * If we have > 1 VGA cards, then we need to arbitrate access to the + * common VGA resources. + * + * If we are a secondary display controller (!PCI_DISPLAY_CLASS_VGA), + * then we do not take part in VGA arbitration and the + * vga_client_register() fails with -ENODEV. + */ + ret = vga_client_register(pdev, i915, NULL, intel_vga_set_decode); + if (ret && ret != -ENODEV) + return ret; + + return 0; +} + +void intel_vga_unregister(struct drm_i915_private *i915) +{ + struct pci_dev *pdev = i915->drm.pdev; + + vga_client_register(pdev, NULL, NULL, NULL); +} diff --git a/drivers/gpu/drm/i915/display/intel_vga.h b/drivers/gpu/drm/i915/display/intel_vga.h new file mode 100644 index 000000000000..ba5b55b917f0 --- /dev/null +++ b/drivers/gpu/drm/i915/display/intel_vga.h @@ -0,0 +1,18 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __INTEL_VGA_H__ +#define __INTEL_VGA_H__ + +struct drm_i915_private; + +void intel_vga_reset_io_mem(struct drm_i915_private *i915); +void intel_vga_disable(struct drm_i915_private *i915); +void intel_vga_redisable(struct drm_i915_private *i915); +void intel_vga_redisable_power_on(struct drm_i915_private *i915); +int intel_vga_register(struct drm_i915_private *i915); +void intel_vga_unregister(struct drm_i915_private *i915); + +#endif /* __INTEL_VGA_H__ */ diff --git a/drivers/gpu/drm/i915/display/vlv_dsi.c b/drivers/gpu/drm/i915/display/vlv_dsi.c index a71b22bdd95b..0ca49b1604c6 100644 --- a/drivers/gpu/drm/i915/display/vlv_dsi.c +++ b/drivers/gpu/drm/i915/display/vlv_dsi.c @@ -749,7 +749,7 @@ static void intel_dsi_pre_enable(struct intel_encoder *encoder, struct drm_crtc *crtc = pipe_config->base.crtc; struct drm_i915_private *dev_priv = to_i915(crtc->dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; enum port port; u32 val; bool glk_cold_boot = false; @@ -1870,11 +1870,11 @@ void vlv_dsi_init(struct drm_i915_private *dev_priv) * port C. BXT isn't limited like this. */ if (IS_GEN9_LP(dev_priv)) - intel_encoder->crtc_mask = BIT(PIPE_A) | BIT(PIPE_B) | BIT(PIPE_C); + intel_encoder->pipe_mask = ~0; else if (port == PORT_A) - intel_encoder->crtc_mask = BIT(PIPE_A); + intel_encoder->pipe_mask = BIT(PIPE_A); else - intel_encoder->crtc_mask = BIT(PIPE_B); + intel_encoder->pipe_mask = BIT(PIPE_B); if (dev_priv->vbt.dsi.config->dual_link) intel_dsi->ports = BIT(PORT_A) | BIT(PORT_C); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c index f99920652751..81366aa4812b 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c @@ -155,7 +155,6 @@ static void clear_pages_dma_fence_cb(struct dma_fence *fence, static void clear_pages_worker(struct work_struct *work) { struct clear_pages_work *w = container_of(work, typeof(*w), work); - struct drm_i915_private *i915 = w->ce->engine->i915; struct drm_i915_gem_object *obj = w->sleeve->vma->obj; struct i915_vma *vma = w->sleeve->vma; struct i915_request *rq; @@ -173,11 +172,9 @@ static void clear_pages_worker(struct work_struct *work) obj->read_domains = I915_GEM_GPU_DOMAINS; obj->write_domain = 0; - /* XXX: we need to kill this */ - mutex_lock(&i915->drm.struct_mutex); err = i915_vma_pin(vma, 0, 0, PIN_USER); if (unlikely(err)) - goto out_unlock; + goto out_signal; batch = intel_emit_vma_fill_blt(w->ce, vma, w->value); if (IS_ERR(batch)) { @@ -211,7 +208,7 @@ static void clear_pages_worker(struct work_struct *work) * keep track of the GPU activity within this vma/request, and * propagate the signal from the request to w->dma. */ - err = i915_active_ref(&vma->active, rq->timeline, rq); + err = __i915_vma_move_to_active(vma, rq); if (err) goto out_request; @@ -229,8 +226,6 @@ out_batch: intel_emit_vma_release(w->ce, batch); out_unpin: i915_vma_unpin(vma); -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); out_signal: if (unlikely(err)) { dma_fence_set_error(&w->dma, err); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c index 1cdfe05514c3..337ba17b1e0e 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c @@ -69,8 +69,10 @@ #include <drm/i915_drm.h> -#include "gt/intel_lrc_reg.h" +#include "gt/intel_engine_heartbeat.h" #include "gt/intel_engine_user.h" +#include "gt/intel_lrc_reg.h" +#include "gt/intel_ring.h" #include "i915_gem_context.h" #include "i915_globals.h" @@ -167,97 +169,6 @@ lookup_user_engine(struct i915_gem_context *ctx, return i915_gem_context_get_engine(ctx, idx); } -static inline int new_hw_id(struct drm_i915_private *i915, gfp_t gfp) -{ - unsigned int max; - - lockdep_assert_held(&i915->contexts.mutex); - - if (INTEL_GEN(i915) >= 12) - max = GEN12_MAX_CONTEXT_HW_ID; - else if (INTEL_GEN(i915) >= 11) - max = GEN11_MAX_CONTEXT_HW_ID; - else if (USES_GUC_SUBMISSION(i915)) - /* - * When using GuC in proxy submission, GuC consumes the - * highest bit in the context id to indicate proxy submission. - */ - max = MAX_GUC_CONTEXT_HW_ID; - else - max = MAX_CONTEXT_HW_ID; - - return ida_simple_get(&i915->contexts.hw_ida, 0, max, gfp); -} - -static int steal_hw_id(struct drm_i915_private *i915) -{ - struct i915_gem_context *ctx, *cn; - LIST_HEAD(pinned); - int id = -ENOSPC; - - lockdep_assert_held(&i915->contexts.mutex); - - list_for_each_entry_safe(ctx, cn, - &i915->contexts.hw_id_list, hw_id_link) { - if (atomic_read(&ctx->hw_id_pin_count)) { - list_move_tail(&ctx->hw_id_link, &pinned); - continue; - } - - GEM_BUG_ON(!ctx->hw_id); /* perma-pinned kernel context */ - list_del_init(&ctx->hw_id_link); - id = ctx->hw_id; - break; - } - - /* - * Remember how far we got up on the last repossesion scan, so the - * list is kept in a "least recently scanned" order. - */ - list_splice_tail(&pinned, &i915->contexts.hw_id_list); - return id; -} - -static int assign_hw_id(struct drm_i915_private *i915, unsigned int *out) -{ - int ret; - - lockdep_assert_held(&i915->contexts.mutex); - - /* - * We prefer to steal/stall ourselves and our users over that of the - * entire system. That may be a little unfair to our users, and - * even hurt high priority clients. The choice is whether to oomkill - * something else, or steal a context id. - */ - ret = new_hw_id(i915, GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN); - if (unlikely(ret < 0)) { - ret = steal_hw_id(i915); - if (ret < 0) /* once again for the correct errno code */ - ret = new_hw_id(i915, GFP_KERNEL); - if (ret < 0) - return ret; - } - - *out = ret; - return 0; -} - -static void release_hw_id(struct i915_gem_context *ctx) -{ - struct drm_i915_private *i915 = ctx->i915; - - if (list_empty(&ctx->hw_id_link)) - return; - - mutex_lock(&i915->contexts.mutex); - if (!list_empty(&ctx->hw_id_link)) { - ida_simple_remove(&i915->contexts.hw_ida, ctx->hw_id); - list_del_init(&ctx->hw_id_link); - } - mutex_unlock(&i915->contexts.mutex); -} - static void __free_engines(struct i915_gem_engines *e, unsigned int count) { while (count--) { @@ -294,101 +205,241 @@ static struct i915_gem_engines *default_engines(struct i915_gem_context *ctx) for_each_engine(engine, gt, id) { struct intel_context *ce; + if (engine->legacy_idx == INVALID_ENGINE) + continue; + + GEM_BUG_ON(engine->legacy_idx >= I915_NUM_ENGINES); + GEM_BUG_ON(e->engines[engine->legacy_idx]); + ce = intel_context_create(ctx, engine); if (IS_ERR(ce)) { - __free_engines(e, id); + __free_engines(e, e->num_engines + 1); return ERR_CAST(ce); } - e->engines[id] = ce; - e->num_engines = id + 1; + e->engines[engine->legacy_idx] = ce; + e->num_engines = max(e->num_engines, engine->legacy_idx); } + e->num_engines++; return e; } static void i915_gem_context_free(struct i915_gem_context *ctx) { - lockdep_assert_held(&ctx->i915->drm.struct_mutex); GEM_BUG_ON(!i915_gem_context_is_closed(ctx)); - release_hw_id(ctx); - if (ctx->vm) - i915_vm_put(ctx->vm); + spin_lock(&ctx->i915->gem.contexts.lock); + list_del(&ctx->link); + spin_unlock(&ctx->i915->gem.contexts.lock); free_engines(rcu_access_pointer(ctx->engines)); mutex_destroy(&ctx->engines_mutex); + kfree(ctx->jump_whitelist); + if (ctx->timeline) intel_timeline_put(ctx->timeline); kfree(ctx->name); put_pid(ctx->pid); - list_del(&ctx->link); mutex_destroy(&ctx->mutex); kfree_rcu(ctx, rcu); } -static void contexts_free(struct drm_i915_private *i915) +static void contexts_free_all(struct llist_node *list) { - struct llist_node *freed = llist_del_all(&i915->contexts.free_list); struct i915_gem_context *ctx, *cn; - lockdep_assert_held(&i915->drm.struct_mutex); - - llist_for_each_entry_safe(ctx, cn, freed, free_link) + llist_for_each_entry_safe(ctx, cn, list, free_link) i915_gem_context_free(ctx); } -static void contexts_free_first(struct drm_i915_private *i915) +static void contexts_flush_free(struct i915_gem_contexts *gc) { - struct i915_gem_context *ctx; - struct llist_node *freed; - - lockdep_assert_held(&i915->drm.struct_mutex); - - freed = llist_del_first(&i915->contexts.free_list); - if (!freed) - return; - - ctx = container_of(freed, typeof(*ctx), free_link); - i915_gem_context_free(ctx); + contexts_free_all(llist_del_all(&gc->free_list)); } static void contexts_free_worker(struct work_struct *work) { - struct drm_i915_private *i915 = - container_of(work, typeof(*i915), contexts.free_work); + struct i915_gem_contexts *gc = + container_of(work, typeof(*gc), free_work); - mutex_lock(&i915->drm.struct_mutex); - contexts_free(i915); - mutex_unlock(&i915->drm.struct_mutex); + contexts_flush_free(gc); } void i915_gem_context_release(struct kref *ref) { struct i915_gem_context *ctx = container_of(ref, typeof(*ctx), ref); - struct drm_i915_private *i915 = ctx->i915; + struct i915_gem_contexts *gc = &ctx->i915->gem.contexts; trace_i915_context_free(ctx); - if (llist_add(&ctx->free_link, &i915->contexts.free_list)) - queue_work(i915->wq, &i915->contexts.free_work); + if (llist_add(&ctx->free_link, &gc->free_list)) + schedule_work(&gc->free_work); } -static void context_close(struct i915_gem_context *ctx) +static inline struct i915_gem_engines * +__context_engines_static(const struct i915_gem_context *ctx) { - mutex_lock(&ctx->mutex); + return rcu_dereference_protected(ctx->engines, true); +} - i915_gem_context_set_closed(ctx); - ctx->file_priv = ERR_PTR(-EBADF); +static bool __reset_engine(struct intel_engine_cs *engine) +{ + struct intel_gt *gt = engine->gt; + bool success = false; + + if (!intel_has_reset_engine(gt)) + return false; + + if (!test_and_set_bit(I915_RESET_ENGINE + engine->id, + >->reset.flags)) { + success = intel_engine_reset(engine, NULL) == 0; + clear_and_wake_up_bit(I915_RESET_ENGINE + engine->id, + >->reset.flags); + } + + return success; +} + +static void __reset_context(struct i915_gem_context *ctx, + struct intel_engine_cs *engine) +{ + intel_gt_handle_error(engine->gt, engine->mask, 0, + "context closure in %s", ctx->name); +} + +static bool __cancel_engine(struct intel_engine_cs *engine) +{ + /* + * Send a "high priority pulse" down the engine to cause the + * current request to be momentarily preempted. (If it fails to + * be preempted, it will be reset). As we have marked our context + * as banned, any incomplete request, including any running, will + * be skipped following the preemption. + * + * If there is no hangchecking (one of the reasons why we try to + * cancel the context) and no forced preemption, there may be no + * means by which we reset the GPU and evict the persistent hog. + * Ergo if we are unable to inject a preemptive pulse that can + * kill the banned context, we fallback to doing a local reset + * instead. + */ + if (IS_ACTIVE(CONFIG_DRM_I915_PREEMPT_TIMEOUT) && + !intel_engine_pulse(engine)) + return true; + + /* If we are unable to send a pulse, try resetting this engine. */ + return __reset_engine(engine); +} + +static struct intel_engine_cs *__active_engine(struct i915_request *rq) +{ + struct intel_engine_cs *engine, *locked; + + /* + * Serialise with __i915_request_submit() so that it sees + * is-banned?, or we know the request is already inflight. + */ + locked = READ_ONCE(rq->engine); + spin_lock_irq(&locked->active.lock); + while (unlikely(locked != (engine = READ_ONCE(rq->engine)))) { + spin_unlock(&locked->active.lock); + spin_lock(&engine->active.lock); + locked = engine; + } + + engine = NULL; + if (i915_request_is_active(rq) && !rq->fence.error) + engine = rq->engine; + + spin_unlock_irq(&locked->active.lock); + + return engine; +} + +static struct intel_engine_cs *active_engine(struct intel_context *ce) +{ + struct intel_engine_cs *engine = NULL; + struct i915_request *rq; + + if (!ce->timeline) + return NULL; + + mutex_lock(&ce->timeline->mutex); + list_for_each_entry_reverse(rq, &ce->timeline->requests, link) { + if (i915_request_completed(rq)) + break; + + /* Check with the backend if the request is inflight */ + engine = __active_engine(rq); + if (engine) + break; + } + mutex_unlock(&ce->timeline->mutex); + + return engine; +} + +static void kill_context(struct i915_gem_context *ctx) +{ + struct i915_gem_engines_iter it; + struct intel_context *ce; + + /* + * If we are already banned, it was due to a guilty request causing + * a reset and the entire context being evicted from the GPU. + */ + if (i915_gem_context_is_banned(ctx)) + return; + + i915_gem_context_set_banned(ctx); /* - * This context will never again be assinged to HW, so we can - * reuse its ID for the next context. + * Map the user's engine back to the actual engines; one virtual + * engine will be mapped to multiple engines, and using ctx->engine[] + * the same engine may be have multiple instances in the user's map. + * However, we only care about pending requests, so only include + * engines on which there are incomplete requests. */ - release_hw_id(ctx); + for_each_gem_engine(ce, __context_engines_static(ctx), it) { + struct intel_engine_cs *engine; + + /* + * Check the current active state of this context; if we + * are currently executing on the GPU we need to evict + * ourselves. On the other hand, if we haven't yet been + * submitted to the GPU or if everything is complete, + * we have nothing to do. + */ + engine = active_engine(ce); + + /* First attempt to gracefully cancel the context */ + if (engine && !__cancel_engine(engine)) + /* + * If we are unable to send a preemptive pulse to bump + * the context from the GPU, we have to resort to a full + * reset. We hope the collateral damage is worth it. + */ + __reset_context(ctx, engine); + } +} + +static void context_close(struct i915_gem_context *ctx) +{ + struct i915_address_space *vm; + + i915_gem_context_set_closed(ctx); + + mutex_lock(&ctx->mutex); + + vm = i915_gem_context_vm(ctx); + if (vm) + i915_vm_close(vm); + + ctx->file_priv = ERR_PTR(-EBADF); /* * The LUT uses the VMA as a backpointer to unref the object, @@ -398,9 +449,47 @@ static void context_close(struct i915_gem_context *ctx) lut_close(ctx); mutex_unlock(&ctx->mutex); + + /* + * If the user has disabled hangchecking, we can not be sure that + * the batches will ever complete after the context is closed, + * keeping the context and all resources pinned forever. So in this + * case we opt to forcibly kill off all remaining requests on + * context close. + */ + if (!i915_gem_context_is_persistent(ctx) || + !i915_modparams.enable_hangcheck) + kill_context(ctx); + i915_gem_context_put(ctx); } +static int __context_set_persistence(struct i915_gem_context *ctx, bool state) +{ + if (i915_gem_context_is_persistent(ctx) == state) + return 0; + + if (state) { + /* + * Only contexts that are short-lived [that will expire or be + * reset] are allowed to survive past termination. We require + * hangcheck to ensure that the persistent requests are healthy. + */ + if (!i915_modparams.enable_hangcheck) + return -EINVAL; + + i915_gem_context_set_persistence(ctx); + } else { + /* To cancel a context we use "preempt-to-idle" */ + if (!(ctx->i915->caps.scheduler & I915_SCHEDULER_CAP_PREEMPTION)) + return -ENODEV; + + i915_gem_context_clear_persistence(ctx); + } + + return 0; +} + static struct i915_gem_context * __create_context(struct drm_i915_private *i915) { @@ -414,7 +503,6 @@ __create_context(struct drm_i915_private *i915) return ERR_PTR(-ENOMEM); kref_init(&ctx->ref); - list_add_tail(&ctx->link, &i915->contexts.list); ctx->i915 = i915; ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_NORMAL); mutex_init(&ctx->mutex); @@ -428,7 +516,6 @@ __create_context(struct drm_i915_private *i915) RCU_INIT_POINTER(ctx->engines, e); INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL); - INIT_LIST_HEAD(&ctx->hw_id_link); /* NB: Mark all slices as needing a remap so that when the context first * loads it will restore whatever remap state already exists. If there @@ -437,10 +524,18 @@ __create_context(struct drm_i915_private *i915) i915_gem_context_set_bannable(ctx); i915_gem_context_set_recoverable(ctx); + __context_set_persistence(ctx, true /* cgroup hook? */); for (i = 0; i < ARRAY_SIZE(ctx->hang_timestamp); i++) ctx->hang_timestamp[i] = jiffies - CONTEXT_FAST_HANG_JIFFIES; + ctx->jump_whitelist = NULL; + ctx->jump_whitelist_cmds = 0; + + spin_lock(&i915->gem.contexts.lock); + list_add_tail(&ctx->link, &i915->gem.contexts.list); + spin_unlock(&i915->gem.contexts.lock); + return ctx; err_free: @@ -470,11 +565,11 @@ static void __apply_ppgtt(struct intel_context *ce, void *vm) static struct i915_address_space * __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm) { - struct i915_address_space *old = ctx->vm; + struct i915_address_space *old = i915_gem_context_vm(ctx); GEM_BUG_ON(old && i915_vm_is_4lvl(vm) != i915_vm_is_4lvl(old)); - ctx->vm = i915_vm_get(vm); + rcu_assign_pointer(ctx->vm, i915_vm_open(vm)); context_apply_all(ctx, __apply_ppgtt, vm); return old; @@ -483,12 +578,12 @@ __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm) static void __assign_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm) { - if (vm == ctx->vm) + if (vm == rcu_access_pointer(ctx->vm)) return; vm = __set_ppgtt(ctx, vm); if (vm) - i915_vm_put(vm); + i915_vm_close(vm); } static void __set_timeline(struct intel_timeline **dst, @@ -515,27 +610,25 @@ static void __assign_timeline(struct i915_gem_context *ctx, } static struct i915_gem_context * -i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags) +i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags) { struct i915_gem_context *ctx; - lockdep_assert_held(&dev_priv->drm.struct_mutex); - if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE && - !HAS_EXECLISTS(dev_priv)) + !HAS_EXECLISTS(i915)) return ERR_PTR(-EINVAL); - /* Reap the most stale context */ - contexts_free_first(dev_priv); + /* Reap the stale contexts */ + contexts_flush_free(&i915->gem.contexts); - ctx = __create_context(dev_priv); + ctx = __create_context(i915); if (IS_ERR(ctx)) return ctx; - if (HAS_FULL_PPGTT(dev_priv)) { + if (HAS_FULL_PPGTT(i915)) { struct i915_ppgtt *ppgtt; - ppgtt = i915_ppgtt_create(dev_priv); + ppgtt = i915_ppgtt_create(i915); if (IS_ERR(ppgtt)) { DRM_DEBUG_DRIVER("PPGTT setup failed (%ld)\n", PTR_ERR(ppgtt)); @@ -543,14 +636,17 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags) return ERR_CAST(ppgtt); } + mutex_lock(&ctx->mutex); __assign_ppgtt(ctx, &ppgtt->vm); + mutex_unlock(&ctx->mutex); + i915_vm_put(&ppgtt->vm); } if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE) { struct intel_timeline *timeline; - timeline = intel_timeline_create(&dev_priv->gt, NULL); + timeline = intel_timeline_create(&i915->gt, NULL); if (IS_ERR(timeline)) { context_close(ctx); return ERR_CAST(timeline); @@ -582,19 +678,13 @@ struct i915_gem_context * i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio) { struct i915_gem_context *ctx; - int err; ctx = i915_gem_create_context(i915, 0); if (IS_ERR(ctx)) return ctx; - err = i915_gem_context_pin_hw_id(ctx); - if (err) { - destroy_kernel_context(&ctx); - return ERR_PTR(err); - } - i915_gem_context_clear_bannable(ctx); + i915_gem_context_set_persistence(ctx); ctx->sched.priority = I915_USER_PRIORITY(prio); GEM_BUG_ON(!i915_gem_context_is_kernel(ctx)); @@ -602,62 +692,42 @@ i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio) return ctx; } -static void init_contexts(struct drm_i915_private *i915) +static void init_contexts(struct i915_gem_contexts *gc) { - mutex_init(&i915->contexts.mutex); - INIT_LIST_HEAD(&i915->contexts.list); - - /* Using the simple ida interface, the max is limited by sizeof(int) */ - BUILD_BUG_ON(MAX_CONTEXT_HW_ID > INT_MAX); - BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > INT_MAX); - ida_init(&i915->contexts.hw_ida); - INIT_LIST_HEAD(&i915->contexts.hw_id_list); + spin_lock_init(&gc->lock); + INIT_LIST_HEAD(&gc->list); - INIT_WORK(&i915->contexts.free_work, contexts_free_worker); - init_llist_head(&i915->contexts.free_list); + INIT_WORK(&gc->free_work, contexts_free_worker); + init_llist_head(&gc->free_list); } -int i915_gem_contexts_init(struct drm_i915_private *dev_priv) +int i915_gem_init_contexts(struct drm_i915_private *i915) { struct i915_gem_context *ctx; /* Reassure ourselves we are only called once */ - GEM_BUG_ON(dev_priv->kernel_context); + GEM_BUG_ON(i915->kernel_context); - init_contexts(dev_priv); + init_contexts(&i915->gem.contexts); /* lowest priority; idle task */ - ctx = i915_gem_context_create_kernel(dev_priv, I915_PRIORITY_MIN); + ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MIN); if (IS_ERR(ctx)) { DRM_ERROR("Failed to create default global context\n"); return PTR_ERR(ctx); } - /* - * For easy recognisablity, we want the kernel context to be 0 and then - * all user contexts will have non-zero hw_id. Kernel contexts are - * permanently pinned, so that we never suffer a stall and can - * use them from any allocation context (e.g. for evicting other - * contexts and from inside the shrinker). - */ - GEM_BUG_ON(ctx->hw_id); - GEM_BUG_ON(!atomic_read(&ctx->hw_id_pin_count)); - dev_priv->kernel_context = ctx; + i915->kernel_context = ctx; DRM_DEBUG_DRIVER("%s context support initialized\n", - DRIVER_CAPS(dev_priv)->has_logical_contexts ? + DRIVER_CAPS(i915)->has_logical_contexts ? "logical" : "fake"); return 0; } -void i915_gem_contexts_fini(struct drm_i915_private *i915) +void i915_gem_driver_release__contexts(struct drm_i915_private *i915) { - lockdep_assert_held(&i915->drm.struct_mutex); - destroy_kernel_context(&i915->kernel_context); - - /* Must free all deferred contexts (via flush_workqueue) first */ - GEM_BUG_ON(!list_empty(&i915->contexts.hw_id_list)); - ida_destroy(&i915->contexts.hw_ida); + flush_work(&i915->gem.contexts.free_work); } static int context_idr_cleanup(int id, void *p, void *data) @@ -675,11 +745,16 @@ static int vm_idr_cleanup(int id, void *p, void *data) static int gem_context_register(struct i915_gem_context *ctx, struct drm_i915_file_private *fpriv) { + struct i915_address_space *vm; int ret; ctx->file_priv = fpriv; - if (ctx->vm) - ctx->vm->file = fpriv; + + mutex_lock(&ctx->mutex); + vm = i915_gem_context_vm(ctx); + if (vm) + WRITE_ONCE(vm->file, fpriv); /* XXX */ + mutex_unlock(&ctx->mutex); ctx->pid = get_task_pid(current, PIDTYPE_PID); ctx->name = kasprintf(GFP_KERNEL, "%s[%d]", @@ -716,9 +791,7 @@ int i915_gem_context_open(struct drm_i915_private *i915, idr_init(&file_priv->context_idr); idr_init_base(&file_priv->vm_idr, 1); - mutex_lock(&i915->drm.struct_mutex); ctx = i915_gem_create_context(i915, 0); - mutex_unlock(&i915->drm.struct_mutex); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto err; @@ -746,6 +819,7 @@ err: void i915_gem_context_close(struct drm_file *file) { struct drm_i915_file_private *file_priv = file->driver_priv; + struct drm_i915_private *i915 = file_priv->dev_priv; idr_for_each(&file_priv->context_idr, context_idr_cleanup, NULL); idr_destroy(&file_priv->context_idr); @@ -754,6 +828,8 @@ void i915_gem_context_close(struct drm_file *file) idr_for_each(&file_priv->vm_idr, vm_idr_cleanup, NULL); idr_destroy(&file_priv->vm_idr); mutex_destroy(&file_priv->vm_idr_lock); + + contexts_flush_free(&i915->gem.contexts); } int i915_gem_vm_create_ioctl(struct drm_device *dev, void *data, @@ -846,6 +922,7 @@ struct context_barrier_task { void *data; }; +__i915_active_call static void cb_retire(struct i915_active *base) { struct context_barrier_task *cb = container_of(base, typeof(*cb), base); @@ -865,20 +942,18 @@ static int context_barrier_task(struct i915_gem_context *ctx, void (*task)(void *data), void *data) { - struct drm_i915_private *i915 = ctx->i915; struct context_barrier_task *cb; struct i915_gem_engines_iter it; struct intel_context *ce; int err = 0; - lockdep_assert_held(&i915->drm.struct_mutex); GEM_BUG_ON(!task); cb = kmalloc(sizeof(*cb), GFP_KERNEL); if (!cb) return -ENOMEM; - i915_active_init(i915, &cb->base, NULL, cb_retire); + i915_active_init(&cb->base, NULL, cb_retire); err = i915_active_acquire(&cb->base); if (err) { kfree(cb); @@ -910,7 +985,7 @@ static int context_barrier_task(struct i915_gem_context *ctx, if (emit) err = emit(rq, data); if (err == 0) - err = i915_active_ref(&cb->base, rq->timeline, rq); + err = i915_active_add_request(&cb->base, rq); i915_request_add(rq); if (err) @@ -933,16 +1008,12 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv, struct i915_address_space *vm; int ret; - if (!ctx->vm) + if (!rcu_access_pointer(ctx->vm)) return -ENODEV; - /* XXX rcu acquire? */ - ret = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex); - if (ret) - return ret; - + rcu_read_lock(); vm = i915_vm_get(ctx->vm); - mutex_unlock(&ctx->i915->drm.struct_mutex); + rcu_read_unlock(); ret = mutex_lock_interruptible(&file_priv->vm_idr_lock); if (ret) @@ -953,7 +1024,7 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv, if (ret < 0) goto err_unlock; - i915_vm_get(vm); + i915_vm_open(vm); args->size = 0; args->value = ret; @@ -973,7 +1044,7 @@ static void set_ppgtt_barrier(void *data) if (INTEL_GEN(old->i915) < 8) gen6_ppgtt_unpin_all(i915_vm_to_ppgtt(old)); - i915_vm_put(old); + i915_vm_close(old); } static int emit_ppgtt_update(struct i915_request *rq, void *data) @@ -1003,12 +1074,18 @@ static int emit_ppgtt_update(struct i915_request *rq, void *data) intel_ring_advance(rq, cs); } else if (HAS_LOGICAL_RING_CONTEXTS(engine->i915)) { struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); + int err; + + /* Magic required to prevent forcewake errors! */ + err = engine->emit_flush(rq, EMIT_INVALIDATE); + if (err) + return err; cs = intel_ring_begin(rq, 4 * GEN8_3LVL_PDPES + 2); if (IS_ERR(cs)) return PTR_ERR(cs); - *cs++ = MI_LOAD_REGISTER_IMM(2 * GEN8_3LVL_PDPES); + *cs++ = MI_LOAD_REGISTER_IMM(2 * GEN8_3LVL_PDPES) | MI_LRI_FORCE_POSTED; for (i = GEN8_3LVL_PDPES; i--; ) { const dma_addr_t pd_daddr = i915_page_dir_dma_addr(ppgtt, i); @@ -1045,34 +1122,34 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv, if (args->size) return -EINVAL; - if (!ctx->vm) + if (!rcu_access_pointer(ctx->vm)) return -ENODEV; if (upper_32_bits(args->value)) return -ENOENT; - err = mutex_lock_interruptible(&file_priv->vm_idr_lock); - if (err) - return err; - + rcu_read_lock(); vm = idr_find(&file_priv->vm_idr, args->value); - if (vm) - i915_vm_get(vm); - mutex_unlock(&file_priv->vm_idr_lock); + if (vm && !kref_get_unless_zero(&vm->ref)) + vm = NULL; + rcu_read_unlock(); if (!vm) return -ENOENT; - err = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex); + err = mutex_lock_interruptible(&ctx->mutex); if (err) goto out; - if (vm == ctx->vm) + if (i915_gem_context_is_closed(ctx)) { + err = -ENOENT; + goto unlock; + } + + if (vm == rcu_access_pointer(ctx->vm)) goto unlock; /* Teardown the existing obj:vma cache, it will have to be rebuilt. */ - mutex_lock(&ctx->mutex); lut_close(ctx); - mutex_unlock(&ctx->mutex); old = __set_ppgtt(ctx, vm); @@ -1087,13 +1164,12 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv, set_ppgtt_barrier, old); if (err) { - i915_vm_put(__set_ppgtt(ctx, old)); - i915_vm_put(old); + i915_vm_close(__set_ppgtt(ctx, old)); + i915_vm_close(old); } unlock: - mutex_unlock(&ctx->i915->drm.struct_mutex); - + mutex_unlock(&ctx->mutex); out: i915_vm_put(vm); return err; @@ -1112,7 +1188,7 @@ static int gen8_emit_rpcs_config(struct i915_request *rq, offset = i915_ggtt_offset(ce->state) + LRC_STATE_PN * PAGE_SIZE + - (CTX_R_PWR_CLK_STATE + 1) * 4; + CTX_R_PWR_CLK_STATE * 4; *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; *cs++ = lower_32_bits(offset); @@ -1155,8 +1231,7 @@ gen8_modify_rpcs(struct intel_context *ce, struct intel_sseu sseu) } static int -__intel_context_reconfigure_sseu(struct intel_context *ce, - struct intel_sseu sseu) +intel_context_reconfigure_sseu(struct intel_context *ce, struct intel_sseu sseu) { int ret; @@ -1180,23 +1255,6 @@ unlock: } static int -intel_context_reconfigure_sseu(struct intel_context *ce, struct intel_sseu sseu) -{ - struct drm_i915_private *i915 = ce->engine->i915; - int ret; - - ret = mutex_lock_interruptible(&i915->drm.struct_mutex); - if (ret) - return ret; - - ret = __intel_context_reconfigure_sseu(ce, sseu); - - mutex_unlock(&i915->drm.struct_mutex); - - return ret; -} - -static int user_to_context_sseu(struct drm_i915_private *i915, const struct drm_i915_gem_context_param_sseu *user, struct intel_sseu *context) @@ -1629,7 +1687,7 @@ replace: i915_gem_context_set_user_engines(ctx); else i915_gem_context_clear_user_engines(ctx); - rcu_swap_protected(ctx->engines, set.engines, 1); + set.engines = rcu_replace_pointer(ctx->engines, set.engines, 1); mutex_unlock(&ctx->engines_mutex); call_rcu(&set.engines->rcu, free_engines_rcu); @@ -1738,6 +1796,16 @@ err_free: return err; } +static int +set_persistence(struct i915_gem_context *ctx, + const struct drm_i915_gem_context_param *args) +{ + if (args->size) + return -EINVAL; + + return __context_set_persistence(ctx, args->value); +} + static int ctx_setparam(struct drm_i915_file_private *fpriv, struct i915_gem_context *ctx, struct drm_i915_gem_context_param *args) @@ -1815,6 +1883,10 @@ static int ctx_setparam(struct drm_i915_file_private *fpriv, ret = set_engines(ctx, args); break; + case I915_CONTEXT_PARAM_PERSISTENCE: + ret = set_persistence(ctx, args); + break; + case I915_CONTEXT_PARAM_BAN_PERIOD: default: ret = -EINVAL; @@ -1967,10 +2039,11 @@ static int clone_vm(struct i915_gem_context *dst, struct i915_gem_context *src) { struct i915_address_space *vm; + int err = 0; rcu_read_lock(); do { - vm = READ_ONCE(src->vm); + vm = rcu_dereference(src->vm); if (!vm) break; @@ -1992,7 +2065,7 @@ static int clone_vm(struct i915_gem_context *dst, * it cannot be reallocated elsewhere. */ - if (vm == READ_ONCE(src->vm)) + if (vm == rcu_access_pointer(src->vm)) break; i915_vm_put(vm); @@ -2000,11 +2073,16 @@ static int clone_vm(struct i915_gem_context *dst, rcu_read_unlock(); if (vm) { - __assign_ppgtt(dst, vm); + if (!mutex_lock_interruptible(&dst->mutex)) { + __assign_ppgtt(dst, vm); + mutex_unlock(&dst->mutex); + } else { + err = -EINTR; + } i915_vm_put(vm); } - return 0; + return err; } static int create_clone(struct i915_user_extension __user *ext, void *data) @@ -2094,12 +2172,7 @@ int i915_gem_context_create_ioctl(struct drm_device *dev, void *data, return -EIO; } - ret = i915_mutex_lock_interruptible(dev); - if (ret) - return ret; - ext_data.ctx = i915_gem_create_context(i915, args->flags); - mutex_unlock(&dev->struct_mutex); if (IS_ERR(ext_data.ctx)) return PTR_ERR(ext_data.ctx); @@ -2226,12 +2299,12 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data, case I915_CONTEXT_PARAM_GTT_SIZE: args->size = 0; - if (ctx->vm) - args->value = ctx->vm->total; - else if (to_i915(dev)->ggtt.alias) - args->value = to_i915(dev)->ggtt.alias->vm.total; + rcu_read_lock(); + if (rcu_access_pointer(ctx->vm)) + args->value = rcu_dereference(ctx->vm)->total; else args->value = to_i915(dev)->ggtt.vm.total; + rcu_read_unlock(); break; case I915_CONTEXT_PARAM_NO_ERROR_CAPTURE: @@ -2266,6 +2339,11 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data, ret = get_engines(ctx, args); break; + case I915_CONTEXT_PARAM_PERSISTENCE: + args->size = 0; + args->value = i915_gem_context_is_persistent(ctx); + break; + case I915_CONTEXT_PARAM_BAN_PERIOD: default: ret = -EINVAL; @@ -2297,7 +2375,7 @@ int i915_gem_context_setparam_ioctl(struct drm_device *dev, void *data, int i915_gem_context_reset_stats_ioctl(struct drm_device *dev, void *data, struct drm_file *file) { - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *i915 = to_i915(dev); struct drm_i915_reset_stats *args = data; struct i915_gem_context *ctx; int ret; @@ -2319,7 +2397,7 @@ int i915_gem_context_reset_stats_ioctl(struct drm_device *dev, */ if (capable(CAP_SYS_ADMIN)) - args->reset_count = i915_reset_count(&dev_priv->gpu_error); + args->reset_count = i915_reset_count(&i915->gpu_error); else args->reset_count = 0; @@ -2332,33 +2410,6 @@ out: return ret; } -int __i915_gem_context_pin_hw_id(struct i915_gem_context *ctx) -{ - struct drm_i915_private *i915 = ctx->i915; - int err = 0; - - mutex_lock(&i915->contexts.mutex); - - GEM_BUG_ON(i915_gem_context_is_closed(ctx)); - - if (list_empty(&ctx->hw_id_link)) { - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count)); - - err = assign_hw_id(i915, &ctx->hw_id); - if (err) - goto out_unlock; - - list_add_tail(&ctx->hw_id_link, &i915->contexts.hw_id_list); - } - - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == ~0u); - atomic_inc(&ctx->hw_id_pin_count); - -out_unlock: - mutex_unlock(&i915->contexts.mutex); - return err; -} - /* GEM context-engines iterator: for_each_gem_engine() */ struct intel_context * i915_gem_engines_iter_next(struct i915_gem_engines_iter *it) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h index 176978608b6f..18e50a769a6e 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h @@ -11,7 +11,9 @@ #include "gt/intel_context.h" +#include "i915_drv.h" #include "i915_gem.h" +#include "i915_gem_gtt.h" #include "i915_scheduler.h" #include "intel_device_info.h" @@ -74,6 +76,21 @@ static inline void i915_gem_context_clear_recoverable(struct i915_gem_context *c clear_bit(UCONTEXT_RECOVERABLE, &ctx->user_flags); } +static inline bool i915_gem_context_is_persistent(const struct i915_gem_context *ctx) +{ + return test_bit(UCONTEXT_PERSISTENCE, &ctx->user_flags); +} + +static inline void i915_gem_context_set_persistence(struct i915_gem_context *ctx) +{ + set_bit(UCONTEXT_PERSISTENCE, &ctx->user_flags); +} + +static inline void i915_gem_context_clear_persistence(struct i915_gem_context *ctx) +{ + clear_bit(UCONTEXT_PERSISTENCE, &ctx->user_flags); +} + static inline bool i915_gem_context_is_banned(const struct i915_gem_context *ctx) { return test_bit(CONTEXT_BANNED, &ctx->flags); @@ -112,19 +129,22 @@ i915_gem_context_clear_user_engines(struct i915_gem_context *ctx) clear_bit(CONTEXT_USER_ENGINES, &ctx->flags); } -int __i915_gem_context_pin_hw_id(struct i915_gem_context *ctx); -static inline int i915_gem_context_pin_hw_id(struct i915_gem_context *ctx) +static inline bool +i915_gem_context_nopreempt(const struct i915_gem_context *ctx) { - if (atomic_inc_not_zero(&ctx->hw_id_pin_count)) - return 0; + return test_bit(CONTEXT_NOPREEMPT, &ctx->flags); +} - return __i915_gem_context_pin_hw_id(ctx); +static inline void +i915_gem_context_set_nopreempt(struct i915_gem_context *ctx) +{ + set_bit(CONTEXT_NOPREEMPT, &ctx->flags); } -static inline void i915_gem_context_unpin_hw_id(struct i915_gem_context *ctx) +static inline void +i915_gem_context_clear_nopreempt(struct i915_gem_context *ctx) { - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == 0u); - atomic_dec(&ctx->hw_id_pin_count); + clear_bit(CONTEXT_NOPREEMPT, &ctx->flags); } static inline bool i915_gem_context_is_kernel(struct i915_gem_context *ctx) @@ -133,8 +153,8 @@ static inline bool i915_gem_context_is_kernel(struct i915_gem_context *ctx) } /* i915_gem_context.c */ -int __must_check i915_gem_contexts_init(struct drm_i915_private *dev_priv); -void i915_gem_contexts_fini(struct drm_i915_private *dev_priv); +int __must_check i915_gem_init_contexts(struct drm_i915_private *i915); +void i915_gem_driver_release__contexts(struct drm_i915_private *i915); int i915_gem_context_open(struct drm_i915_private *i915, struct drm_file *file); @@ -173,6 +193,27 @@ static inline void i915_gem_context_put(struct i915_gem_context *ctx) kref_put(&ctx->ref, i915_gem_context_release); } +static inline struct i915_address_space * +i915_gem_context_vm(struct i915_gem_context *ctx) +{ + return rcu_dereference_protected(ctx->vm, lockdep_is_held(&ctx->mutex)); +} + +static inline struct i915_address_space * +i915_gem_context_get_vm_rcu(struct i915_gem_context *ctx) +{ + struct i915_address_space *vm; + + rcu_read_lock(); + vm = rcu_dereference(ctx->vm); + if (!vm) + vm = &ctx->i915->ggtt.vm; + vm = i915_vm_get(vm); + rcu_read_unlock(); + + return vm; +} + static inline struct i915_gem_engines * i915_gem_context_engines(struct i915_gem_context *ctx) { diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h index 260d59cc3de8..3870dd5daaa0 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h @@ -88,7 +88,7 @@ struct i915_gem_context { * In other modes, this is a NULL pointer with the expectation that * the caller uses the shared global GTT. */ - struct i915_address_space *vm; + struct i915_address_space __rcu *vm; /** * @pid: process id of creator @@ -137,6 +137,7 @@ struct i915_gem_context { #define UCONTEXT_NO_ERROR_CAPTURE 1 #define UCONTEXT_BANNABLE 2 #define UCONTEXT_RECOVERABLE 3 +#define UCONTEXT_PERSISTENCE 4 /** * @flags: small set of booleans @@ -146,24 +147,7 @@ struct i915_gem_context { #define CONTEXT_CLOSED 1 #define CONTEXT_FORCE_SINGLE_SUBMISSION 2 #define CONTEXT_USER_ENGINES 3 - - /** - * @hw_id: - unique identifier for the context - * - * The hardware needs to uniquely identify the context for a few - * functions like fault reporting, PASID, scheduling. The - * &drm_i915_private.context_hw_ida is used to assign a unqiue - * id for the lifetime of the context. - * - * @hw_id_pin_count: - number of times this context had been pinned - * for use (should be, at most, once per engine). - * - * @hw_id_link: - all contexts with an assigned id are tracked - * for possible repossession. - */ - unsigned int hw_id; - atomic_t hw_id_pin_count; - struct list_head hw_id_link; +#define CONTEXT_NOPREEMPT 4 struct mutex mutex; @@ -192,6 +176,13 @@ struct i915_gem_context { * per vm, which may be one per context or shared with the global GTT) */ struct radix_tree_root handles_vma; + + /** jump_whitelist: Bit array for tracking cmds during cmdparsing + * Guarded by struct_mutex + */ + unsigned long *jump_whitelist; + /** jump_whitelist_cmds: No of cmd slots available */ + u32 jump_whitelist_cmds; }; #endif /* __I915_GEM_CONTEXT_TYPES_H__ */ diff --git a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c index 96ce95c8ac5a..eaea49d08eb5 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c @@ -256,6 +256,7 @@ static const struct drm_i915_gem_object_ops i915_gem_object_dmabuf_ops = { struct drm_gem_object *i915_gem_prime_import(struct drm_device *dev, struct dma_buf *dma_buf) { + static struct lock_class_key lock_class; struct dma_buf_attachment *attach; struct drm_i915_gem_object *obj; int ret; @@ -287,7 +288,7 @@ struct drm_gem_object *i915_gem_prime_import(struct drm_device *dev, } drm_gem_private_object_init(dev, &obj->base, dma_buf->size); - i915_gem_object_init(obj, &i915_gem_object_dmabuf_ops); + i915_gem_object_init(obj, &i915_gem_object_dmabuf_ops, &lock_class); obj->base.import_attach = attach; obj->base.resv = dma_buf->resv; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c index 9c58e8fac1d9..9937b4c341f1 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c @@ -27,7 +27,7 @@ static void __i915_gem_object_flush_for_display(struct drm_i915_gem_object *obj) void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj) { - if (!READ_ONCE(obj->pin_global)) + if (!i915_gem_object_is_framebuffer(obj)) return; i915_gem_object_lock(obj); @@ -288,14 +288,21 @@ restart: if (!drm_mm_node_allocated(&vma->node)) continue; - ret = i915_vma_bind(vma, cache_level, PIN_UPDATE); + /* Wait for an earlier async bind, need to rewrite it */ + ret = i915_vma_sync(vma); + if (ret) + return ret; + + ret = i915_vma_bind(vma, cache_level, PIN_UPDATE, NULL); if (ret) return ret; } } - list_for_each_entry(vma, &obj->vma.list, obj_link) - vma->node.color = cache_level; + list_for_each_entry(vma, &obj->vma.list, obj_link) { + if (i915_vm_has_cache_coloring(vma->vm)) + vma->node.color = cache_level; + } i915_gem_object_set_cache_coherency(obj, cache_level); obj->cache_dirty = true; /* Always invalidate stale cachelines */ @@ -389,16 +396,11 @@ int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data, if (ret) goto out; - ret = mutex_lock_interruptible(&i915->drm.struct_mutex); - if (ret) - goto out; - ret = i915_gem_object_lock_interruptible(obj); if (ret == 0) { ret = i915_gem_object_set_cache_level(obj, level); i915_gem_object_unlock(obj); } - mutex_unlock(&i915->drm.struct_mutex); out: i915_gem_object_put(obj); @@ -422,12 +424,8 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj, assert_object_held(obj); - /* Mark the global pin early so that we account for the - * display coherency whilst setting up the cache domains. - */ - obj->pin_global++; - - /* The display engine is not coherent with the LLC cache on gen6. As + /* + * The display engine is not coherent with the LLC cache on gen6. As * a result, we make sure that the pinning that is about to occur is * done with uncached PTEs. This is lowest common denominator for all * chipsets. @@ -439,12 +437,11 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj, ret = i915_gem_object_set_cache_level(obj, HAS_WT(to_i915(obj->base.dev)) ? I915_CACHE_WT : I915_CACHE_NONE); - if (ret) { - vma = ERR_PTR(ret); - goto err_unpin_global; - } + if (ret) + return ERR_PTR(ret); - /* As the user may map the buffer once pinned in the display plane + /* + * As the user may map the buffer once pinned in the display plane * (e.g. libkms for the bootup splash), we have to ensure that we * always use map_and_fenceable for all scanout buffers. However, * it may simply be too big to fit into mappable, in which case @@ -461,22 +458,19 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj, if (IS_ERR(vma)) vma = i915_gem_object_ggtt_pin(obj, view, 0, alignment, flags); if (IS_ERR(vma)) - goto err_unpin_global; + return vma; vma->display_alignment = max_t(u64, vma->display_alignment, alignment); __i915_gem_object_flush_for_display(obj); - /* It should now be out of any other write domains, and we can update + /* + * It should now be out of any other write domains, and we can update * the domain values for our changes. */ obj->read_domains |= I915_GEM_DOMAIN_GTT; return vma; - -err_unpin_global: - obj->pin_global--; - return vma; } static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj) @@ -491,6 +485,7 @@ static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj) if (!drm_mm_node_allocated(&vma->node)) continue; + GEM_BUG_ON(vma->vm != &i915->ggtt.vm); list_move_tail(&vma->vm_link, &vma->vm->bound_list); } mutex_unlock(&i915->ggtt.vm.mutex); @@ -500,7 +495,8 @@ static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj) spin_lock_irqsave(&i915->mm.obj_lock, flags); - if (obj->mm.madv == I915_MADV_WILLNEED) + if (obj->mm.madv == I915_MADV_WILLNEED && + !atomic_read(&obj->mm.shrink_pin)) list_move_tail(&obj->mm.link, &i915->mm.shrink_list); spin_unlock_irqrestore(&i915->mm.obj_lock, flags); @@ -514,12 +510,6 @@ i915_gem_object_unpin_from_display_plane(struct i915_vma *vma) assert_object_held(obj); - if (WARN_ON(obj->pin_global == 0)) - return; - - if (--obj->pin_global == 0) - vma->display_alignment = I915_GTT_MIN_ALIGNMENT; - /* Bump the LRU to try and avoid premature eviction whilst flipping */ i915_gem_object_bump_inactive_ggtt(obj); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index b5f6937369ea..f0998f1225af 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -19,6 +19,7 @@ #include "gt/intel_engine_pool.h" #include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" +#include "gt/intel_ring.h" #include "i915_drv.h" #include "i915_gem_clflush.h" @@ -252,6 +253,7 @@ struct i915_execbuffer { bool has_fence : 1; bool needs_unfenced : 1; + struct intel_context *ce; struct i915_request *rq; u32 *rq_cmd; unsigned int rq_size; @@ -296,7 +298,9 @@ static inline u64 gen8_noncanonical_addr(u64 address) static inline bool eb_use_cmdparser(const struct i915_execbuffer *eb) { - return intel_engine_needs_cmd_parser(eb->engine) && eb->batch_len; + return intel_engine_requires_cmd_parser(eb->engine) || + (intel_engine_using_cmd_parser(eb->engine) && + eb->args->batch_len); } static int eb_create(struct i915_execbuffer *eb) @@ -697,7 +701,9 @@ static int eb_reserve(struct i915_execbuffer *eb) case 1: /* Too fragmented, unbind everything and retry */ + mutex_lock(&eb->context->vm->mutex); err = i915_gem_evict_vm(eb->context->vm); + mutex_unlock(&eb->context->vm->mutex); if (err) return err; break; @@ -725,7 +731,7 @@ static int eb_select_context(struct i915_execbuffer *eb) return -ENOENT; eb->gem_context = ctx; - if (ctx->vm) + if (rcu_access_pointer(ctx->vm)) eb->invalid_flags |= EXEC_OBJECT_NEEDS_GTT; eb->context_flags = 0; @@ -880,6 +886,9 @@ static void eb_destroy(const struct i915_execbuffer *eb) { GEM_BUG_ON(eb->reloc_cache.rq); + if (eb->reloc_cache.ce) + intel_context_put(eb->reloc_cache.ce); + if (eb->lut_size > 0) kfree(eb->buckets); } @@ -902,7 +911,8 @@ static void reloc_cache_init(struct reloc_cache *cache, cache->use_64bit_reloc = HAS_64BIT_RELOC(i915); cache->has_fence = cache->gen < 4; cache->needs_unfenced = INTEL_INFO(i915)->unfenced_needs_alignment; - cache->node.allocated = false; + cache->node.flags = 0; + cache->ce = NULL; cache->rq = NULL; cache->rq_size = 0; } @@ -963,11 +973,13 @@ static void reloc_cache_reset(struct reloc_cache *cache) intel_gt_flush_ggtt_writes(ggtt->vm.gt); io_mapping_unmap_atomic((void __iomem *)vaddr); - if (cache->node.allocated) { + if (drm_mm_node_allocated(&cache->node)) { ggtt->vm.clear_range(&ggtt->vm, cache->node.start, cache->node.size); + mutex_lock(&ggtt->vm.mutex); drm_mm_remove_node(&cache->node); + mutex_unlock(&ggtt->vm.mutex); } else { i915_vma_unpin((struct i915_vma *)cache->node.mm); } @@ -1042,11 +1054,13 @@ static void *reloc_iomap(struct drm_i915_gem_object *obj, PIN_NOEVICT); if (IS_ERR(vma)) { memset(&cache->node, 0, sizeof(cache->node)); + mutex_lock(&ggtt->vm.mutex); err = drm_mm_insert_node_in_range (&ggtt->vm.mm, &cache->node, PAGE_SIZE, 0, I915_COLOR_UNEVICTABLE, 0, ggtt->mappable_end, DRM_MM_INSERT_LOW); + mutex_unlock(&ggtt->vm.mutex); if (err) /* no inactive aperture space, use cpu reloc */ return NULL; } else { @@ -1056,7 +1070,7 @@ static void *reloc_iomap(struct drm_i915_gem_object *obj, } offset = cache->node.start; - if (cache->node.allocated) { + if (drm_mm_node_allocated(&cache->node)) { ggtt->vm.insert_page(&ggtt->vm, i915_gem_object_get_dma_address(obj, page), offset, I915_CACHE_NONE, 0); @@ -1145,7 +1159,7 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb, u32 *cmd; int err; - pool = intel_engine_pool_get(&eb->engine->pool, PAGE_SIZE); + pool = intel_engine_get_pool(eb->engine, PAGE_SIZE); if (IS_ERR(pool)) return PTR_ERR(pool); @@ -1168,7 +1182,7 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb, if (err) goto err_unmap; - rq = i915_request_create(eb->context); + rq = intel_context_create_request(cache->ce); if (IS_ERR(rq)) { err = PTR_ERR(rq); goto err_unpin; @@ -1239,6 +1253,29 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb, if (!intel_engine_can_store_dword(eb->engine)) return ERR_PTR(-ENODEV); + if (!cache->ce) { + struct intel_context *ce; + + /* + * The CS pre-parser can pre-fetch commands across + * memory sync points and starting gen12 it is able to + * pre-fetch across BB_START and BB_END boundaries + * (within the same context). We therefore use a + * separate context gen12+ to guarantee that the reloc + * writes land before the parser gets to the target + * memory location. + */ + if (cache->gen >= 12) + ce = intel_context_create(eb->context->gem_context, + eb->engine); + else + ce = intel_context_get(eb->context); + if (IS_ERR(ce)) + return ERR_CAST(ce); + + cache->ce = ce; + } + err = __reloc_gpu_alloc(eb, vma, len); if (unlikely(err)) return ERR_PTR(err); @@ -1388,7 +1425,7 @@ eb_relocate_entry(struct i915_execbuffer *eb, if (reloc->write_domain == I915_GEM_DOMAIN_INSTRUCTION && IS_GEN(eb->i915, 6)) { err = i915_vma_bind(target, target->obj->cache_level, - PIN_GLOBAL); + PIN_GLOBAL, NULL); if (WARN_ONCE(err, "Unexpected failure to bind target VMA!")) return err; @@ -1955,40 +1992,94 @@ static int i915_reset_gen7_sol_offsets(struct i915_request *rq) return 0; } -static struct i915_vma *eb_parse(struct i915_execbuffer *eb, bool is_master) +static struct i915_vma * +shadow_batch_pin(struct i915_execbuffer *eb, struct drm_i915_gem_object *obj) +{ + struct drm_i915_private *dev_priv = eb->i915; + struct i915_vma * const vma = *eb->vma; + struct i915_address_space *vm; + u64 flags; + + /* + * PPGTT backed shadow buffers must be mapped RO, to prevent + * post-scan tampering + */ + if (CMDPARSER_USES_GGTT(dev_priv)) { + flags = PIN_GLOBAL; + vm = &dev_priv->ggtt.vm; + } else if (vma->vm->has_read_only) { + flags = PIN_USER; + vm = vma->vm; + i915_gem_object_set_readonly(obj); + } else { + DRM_DEBUG("Cannot prevent post-scan tampering without RO capable vm\n"); + return ERR_PTR(-EINVAL); + } + + return i915_gem_object_pin(obj, vm, NULL, 0, 0, flags); +} + +static struct i915_vma *eb_parse(struct i915_execbuffer *eb) { struct intel_engine_pool_node *pool; struct i915_vma *vma; + u64 batch_start; + u64 shadow_batch_start; int err; - pool = intel_engine_pool_get(&eb->engine->pool, eb->batch_len); + pool = intel_engine_get_pool(eb->engine, eb->batch_len); if (IS_ERR(pool)) return ERR_CAST(pool); - err = intel_engine_cmd_parser(eb->engine, + vma = shadow_batch_pin(eb, pool->obj); + if (IS_ERR(vma)) + goto err; + + batch_start = gen8_canonical_addr(eb->batch->node.start) + + eb->batch_start_offset; + + shadow_batch_start = gen8_canonical_addr(vma->node.start); + + err = intel_engine_cmd_parser(eb->gem_context, + eb->engine, eb->batch->obj, - pool->obj, + batch_start, eb->batch_start_offset, eb->batch_len, - is_master); + pool->obj, + shadow_batch_start); + if (err) { - if (err == -EACCES) /* unhandled chained batch */ + i915_vma_unpin(vma); + + /* + * Unsafe GGTT-backed buffers can still be submitted safely + * as non-secure. + * For PPGTT backing however, we have no choice but to forcibly + * reject unsafe buffers + */ + if (CMDPARSER_USES_GGTT(eb->i915) && (err == -EACCES)) + /* Execute original buffer non-secure */ vma = NULL; else vma = ERR_PTR(err); goto err; } - vma = i915_gem_object_ggtt_pin(pool->obj, NULL, 0, 0, 0); - if (IS_ERR(vma)) - goto err; - eb->vma[eb->buffer_count] = i915_vma_get(vma); eb->flags[eb->buffer_count] = __EXEC_OBJECT_HAS_PIN | __EXEC_OBJECT_HAS_REF; vma->exec_flags = &eb->flags[eb->buffer_count]; eb->buffer_count++; + eb->batch_start_offset = 0; + eb->batch = vma; + + if (CMDPARSER_USES_GGTT(eb->i915)) + eb->batch_flags |= I915_DISPATCH_SECURE; + + /* eb->batch_len unchanged */ + vma->private = pool; return vma; @@ -2043,6 +2134,9 @@ static int eb_submit(struct i915_execbuffer *eb) if (err) return err; + if (i915_gem_context_nopreempt(eb->gem_context)) + eb->request->flags |= I915_REQUEST_NOPREEMPT; + return 0; } @@ -2112,35 +2206,6 @@ static struct i915_request *eb_throttle(struct intel_context *ce) return i915_request_get(rq); } -static int -__eb_pin_context(struct i915_execbuffer *eb, struct intel_context *ce) -{ - int err; - - if (likely(atomic_inc_not_zero(&ce->pin_count))) - return 0; - - err = mutex_lock_interruptible(&eb->i915->drm.struct_mutex); - if (err) - return err; - - err = __intel_context_do_pin(ce); - mutex_unlock(&eb->i915->drm.struct_mutex); - - return err; -} - -static void -__eb_unpin_context(struct i915_execbuffer *eb, struct intel_context *ce) -{ - if (likely(atomic_add_unless(&ce->pin_count, -1, 1))) - return; - - mutex_lock(&eb->i915->drm.struct_mutex); - intel_context_unpin(ce); - mutex_unlock(&eb->i915->drm.struct_mutex); -} - static int __eb_pin_engine(struct i915_execbuffer *eb, struct intel_context *ce) { struct intel_timeline *tl; @@ -2160,7 +2225,7 @@ static int __eb_pin_engine(struct i915_execbuffer *eb, struct intel_context *ce) * GGTT space, so do this first before we reserve a seqno for * ourselves. */ - err = __eb_pin_context(eb, ce); + err = intel_context_pin(ce); if (err) return err; @@ -2204,7 +2269,7 @@ err_exit: intel_context_exit(ce); intel_context_timeline_unlock(tl); err_unpin: - __eb_unpin_context(eb, ce); + intel_context_unpin(ce); return err; } @@ -2217,7 +2282,7 @@ static void eb_unpin_engine(struct i915_execbuffer *eb) intel_context_exit(ce); mutex_unlock(&tl->mutex); - __eb_unpin_context(eb, ce); + intel_context_unpin(ce); } static unsigned int @@ -2421,6 +2486,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, struct drm_i915_gem_exec_object2 *exec, struct drm_syncobj **fences) { + struct drm_i915_private *i915 = to_i915(dev); struct i915_execbuffer eb; struct dma_fence *in_fence = NULL; struct dma_fence *exec_fence = NULL; @@ -2432,7 +2498,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, BUILD_BUG_ON(__EXEC_OBJECT_INTERNAL_FLAGS & ~__EXEC_OBJECT_UNKNOWN_FLAGS); - eb.i915 = to_i915(dev); + eb.i915 = i915; eb.file = file; eb.args = args; if (DBG_FORCE_RELOC || !(args->flags & I915_EXEC_NO_RELOC)) @@ -2452,8 +2518,15 @@ i915_gem_do_execbuffer(struct drm_device *dev, eb.batch_flags = 0; if (args->flags & I915_EXEC_SECURE) { + if (INTEL_GEN(i915) >= 11) + return -ENODEV; + + /* Return -EPERM to trigger fallback code on old binaries. */ + if (!HAS_SECURE_BATCHES(i915)) + return -EPERM; + if (!drm_is_current_master(file) || !capable(CAP_SYS_ADMIN)) - return -EPERM; + return -EPERM; eb.batch_flags |= I915_DISPATCH_SECURE; } @@ -2530,34 +2603,19 @@ i915_gem_do_execbuffer(struct drm_device *dev, goto err_vma; } + if (eb.batch_len == 0) + eb.batch_len = eb.batch->size - eb.batch_start_offset; + if (eb_use_cmdparser(&eb)) { struct i915_vma *vma; - vma = eb_parse(&eb, drm_is_current_master(file)); + vma = eb_parse(&eb); if (IS_ERR(vma)) { err = PTR_ERR(vma); goto err_vma; } - - if (vma) { - /* - * Batch parsed and accepted: - * - * Set the DISPATCH_SECURE bit to remove the NON_SECURE - * bit from MI_BATCH_BUFFER_START commands issued in - * the dispatch_execbuffer implementations. We - * specifically don't want that set on batches the - * command parser has accepted. - */ - eb.batch_flags |= I915_DISPATCH_SECURE; - eb.batch_start_offset = 0; - eb.batch = vma; - } } - if (eb.batch_len == 0) - eb.batch_len = eb.batch->size - eb.batch_start_offset; - /* * snb/ivb/vlv conflate the "batch in ppgtt" bit with the "non-secure * batch" bit. Hence we need to pin secure batches into the global gtt. diff --git a/drivers/gpu/drm/i915/gem/i915_gem_internal.c b/drivers/gpu/drm/i915/gem/i915_gem_internal.c index 0c41e04ab8fa..9cfb0e41ff06 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_internal.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_internal.c @@ -117,13 +117,6 @@ create_st: goto err; } - /* Mark the pages as dontneed whilst they are still pinned. As soon - * as they are unpinned they are allowed to be reaped by the shrinker, - * and the caller is expected to repopulate - the contents of this - * object are only valid whilst active and pinned. - */ - obj->mm.madv = I915_MADV_DONTNEED; - __i915_gem_object_set_pages(obj, st, sg_page_sizes); return 0; @@ -143,7 +136,6 @@ static void i915_gem_object_put_pages_internal(struct drm_i915_gem_object *obj, internal_free_pages(pages); obj->mm.dirty = false; - obj->mm.madv = I915_MADV_WILLNEED; } static const struct drm_i915_gem_object_ops i915_gem_object_internal_ops = { @@ -172,6 +164,7 @@ struct drm_i915_gem_object * i915_gem_object_create_internal(struct drm_i915_private *i915, phys_addr_t size) { + static struct lock_class_key lock_class; struct drm_i915_gem_object *obj; unsigned int cache_level; @@ -186,7 +179,16 @@ i915_gem_object_create_internal(struct drm_i915_private *i915, return ERR_PTR(-ENOMEM); drm_gem_private_object_init(&i915->drm, &obj->base, size); - i915_gem_object_init(obj, &i915_gem_object_internal_ops); + i915_gem_object_init(obj, &i915_gem_object_internal_ops, &lock_class); + + /* + * Mark the object as volatile, such that the pages are marked as + * dontneed whilst they are still pinned. As soon as they are unpinned + * they are allowed to be reaped by the shrinker, and the caller is + * expected to repopulate - the contents of this object are only valid + * whilst active and pinned. + */ + i915_gem_object_set_volatile(obj); obj->read_domains = I915_GEM_DOMAIN_CPU; obj->write_domain = I915_GEM_DOMAIN_CPU; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_lmem.c b/drivers/gpu/drm/i915/gem/i915_gem_lmem.c new file mode 100644 index 000000000000..0e2bf6b7e143 --- /dev/null +++ b/drivers/gpu/drm/i915/gem/i915_gem_lmem.c @@ -0,0 +1,99 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include "intel_memory_region.h" +#include "gem/i915_gem_region.h" +#include "gem/i915_gem_lmem.h" +#include "i915_drv.h" + +const struct drm_i915_gem_object_ops i915_gem_lmem_obj_ops = { + .flags = I915_GEM_OBJECT_HAS_IOMEM, + + .get_pages = i915_gem_object_get_pages_buddy, + .put_pages = i915_gem_object_put_pages_buddy, + .release = i915_gem_object_release_memory_region, +}; + +/* XXX: Time to vfunc your life up? */ +void __iomem * +i915_gem_object_lmem_io_map_page(struct drm_i915_gem_object *obj, + unsigned long n) +{ + resource_size_t offset; + + offset = i915_gem_object_get_dma_address(obj, n); + offset -= obj->mm.region->region.start; + + return io_mapping_map_wc(&obj->mm.region->iomap, offset, PAGE_SIZE); +} + +void __iomem * +i915_gem_object_lmem_io_map_page_atomic(struct drm_i915_gem_object *obj, + unsigned long n) +{ + resource_size_t offset; + + offset = i915_gem_object_get_dma_address(obj, n); + offset -= obj->mm.region->region.start; + + return io_mapping_map_atomic_wc(&obj->mm.region->iomap, offset); +} + +void __iomem * +i915_gem_object_lmem_io_map(struct drm_i915_gem_object *obj, + unsigned long n, + unsigned long size) +{ + resource_size_t offset; + + GEM_BUG_ON(!i915_gem_object_is_contiguous(obj)); + + offset = i915_gem_object_get_dma_address(obj, n); + offset -= obj->mm.region->region.start; + + return io_mapping_map_wc(&obj->mm.region->iomap, offset, size); +} + +bool i915_gem_object_is_lmem(struct drm_i915_gem_object *obj) +{ + return obj->ops == &i915_gem_lmem_obj_ops; +} + +struct drm_i915_gem_object * +i915_gem_object_create_lmem(struct drm_i915_private *i915, + resource_size_t size, + unsigned int flags) +{ + return i915_gem_object_create_region(i915->mm.regions[INTEL_REGION_LMEM], + size, flags); +} + +struct drm_i915_gem_object * +__i915_gem_lmem_object_create(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags) +{ + static struct lock_class_key lock_class; + struct drm_i915_private *i915 = mem->i915; + struct drm_i915_gem_object *obj; + + if (size > BIT(mem->mm.max_order) * mem->mm.chunk_size) + return ERR_PTR(-E2BIG); + + obj = i915_gem_object_alloc(); + if (!obj) + return ERR_PTR(-ENOMEM); + + drm_gem_private_object_init(&i915->drm, &obj->base, size); + i915_gem_object_init(obj, &i915_gem_lmem_obj_ops, &lock_class); + + obj->read_domains = I915_GEM_DOMAIN_WC | I915_GEM_DOMAIN_GTT; + + i915_gem_object_set_cache_coherency(obj, I915_CACHE_NONE); + + i915_gem_object_init_memory_region(obj, mem, flags); + + return obj; +} diff --git a/drivers/gpu/drm/i915/gem/i915_gem_lmem.h b/drivers/gpu/drm/i915/gem/i915_gem_lmem.h new file mode 100644 index 000000000000..7c176b8b7d2f --- /dev/null +++ b/drivers/gpu/drm/i915/gem/i915_gem_lmem.h @@ -0,0 +1,37 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __I915_GEM_LMEM_H +#define __I915_GEM_LMEM_H + +#include <linux/types.h> + +struct drm_i915_private; +struct drm_i915_gem_object; +struct intel_memory_region; + +extern const struct drm_i915_gem_object_ops i915_gem_lmem_obj_ops; + +void __iomem *i915_gem_object_lmem_io_map(struct drm_i915_gem_object *obj, + unsigned long n, unsigned long size); +void __iomem *i915_gem_object_lmem_io_map_page(struct drm_i915_gem_object *obj, + unsigned long n); +void __iomem * +i915_gem_object_lmem_io_map_page_atomic(struct drm_i915_gem_object *obj, + unsigned long n); + +bool i915_gem_object_is_lmem(struct drm_i915_gem_object *obj); + +struct drm_i915_gem_object * +i915_gem_object_create_lmem(struct drm_i915_private *i915, + resource_size_t size, + unsigned int flags); + +struct drm_i915_gem_object * +__i915_gem_lmem_object_create(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags); + +#endif /* !__I915_GEM_LMEM_H */ diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c index 261c9bd83f51..e3002849844b 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c @@ -8,6 +8,7 @@ #include <linux/sizes.h> #include "gt/intel_gt.h" +#include "gt/intel_gt_requests.h" #include "i915_drv.h" #include "i915_gem_gtt.h" @@ -245,21 +246,9 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf) wakeref = intel_runtime_pm_get(rpm); - srcu = intel_gt_reset_trylock(ggtt->vm.gt); - if (srcu < 0) { - ret = srcu; - goto err_rpm; - } - - ret = i915_mutex_lock_interruptible(dev); + ret = intel_gt_reset_trylock(ggtt->vm.gt, &srcu); if (ret) - goto err_reset; - - /* Access to snoopable pages through the GTT is incoherent. */ - if (obj->cache_level != I915_CACHE_NONE && !HAS_LLC(i915)) { - ret = -EFAULT; - goto err_unlock; - } + goto err_rpm; /* Now pin it into the GTT as needed */ vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, @@ -287,10 +276,19 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf) view.type = I915_GGTT_VIEW_PARTIAL; vma = i915_gem_object_ggtt_pin(obj, &view, 0, 0, flags); } + + /* The entire mappable GGTT is pinned? Unexpected! */ + GEM_BUG_ON(vma == ERR_PTR(-ENOSPC)); } if (IS_ERR(vma)) { ret = PTR_ERR(vma); - goto err_unlock; + goto err_reset; + } + + /* Access to snoopable pages through the GTT is incoherent. */ + if (obj->cache_level != I915_CACHE_NONE && !HAS_LLC(i915)) { + ret = -EFAULT; + goto err_unpin; } ret = i915_vma_pin_fence(vma); @@ -314,18 +312,20 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf) list_add(&obj->userfault_link, &i915->ggtt.userfault_list); mutex_unlock(&i915->ggtt.vm.mutex); - if (CONFIG_DRM_I915_USERFAULT_AUTOSUSPEND) + if (IS_ACTIVE(CONFIG_DRM_I915_USERFAULT_AUTOSUSPEND)) intel_wakeref_auto(&i915->ggtt.userfault_wakeref, msecs_to_jiffies_timeout(CONFIG_DRM_I915_USERFAULT_AUTOSUSPEND)); - i915_vma_set_ggtt_write(vma); + if (write) { + GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj)); + i915_vma_set_ggtt_write(vma); + obj->mm.dirty = true; + } err_fence: i915_vma_unpin_fence(vma); err_unpin: __i915_vma_unpin(vma); -err_unlock: - mutex_unlock(&dev->struct_mutex); err_reset: intel_gt_reset_unlock(ggtt->vm.gt, srcu); err_rpm: @@ -333,23 +333,20 @@ err_rpm: i915_gem_object_unpin_pages(obj); err: switch (ret) { - case -EIO: - /* - * We eat errors when the gpu is terminally wedged to avoid - * userspace unduly crashing (gl has no provisions for mmaps to - * fail). But any other -EIO isn't ours (e.g. swap in failure) - * and so needs to be reported. - */ - if (!intel_gt_is_wedged(ggtt->vm.gt)) - return VM_FAULT_SIGBUS; - /* else, fall through */ - case -EAGAIN: - /* - * EAGAIN means the gpu is hung and we'll wait for the error - * handler to reset everything when re-faulting in - * i915_mutex_lock_interruptible. - */ + default: + WARN_ONCE(ret, "unhandled error in %s: %i\n", __func__, ret); + /* fallthrough */ + case -EIO: /* shmemfs failure from swap device */ + case -EFAULT: /* purged object */ + case -ENODEV: /* bad object, how did you get here! */ + return VM_FAULT_SIGBUS; + + case -ENOSPC: /* shmemfs allocation failure */ + case -ENOMEM: /* our allocation failure */ + return VM_FAULT_OOM; + case 0: + case -EAGAIN: case -ERESTARTSYS: case -EINTR: case -EBUSY: @@ -358,14 +355,6 @@ err: * already did the job. */ return VM_FAULT_NOPAGE; - case -ENOMEM: - return VM_FAULT_OOM; - case -ENOSPC: - case -EFAULT: - return VM_FAULT_SIGBUS; - default: - WARN_ONCE(ret, "unhandled error in %s: %i\n", __func__, ret); - return VM_FAULT_SIGBUS; } } @@ -436,6 +425,7 @@ out: static int create_mmap_offset(struct drm_i915_gem_object *obj) { struct drm_i915_private *i915 = to_i915(obj->base.dev); + struct intel_gt *gt = &i915->gt; int err; err = drm_gem_create_mmap_offset(&obj->base); @@ -443,21 +433,12 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj) return 0; /* Attempt to reap some mmap space from dead objects */ - do { - err = i915_gem_wait_for_idle(i915, - I915_WAIT_INTERRUPTIBLE, - MAX_SCHEDULE_TIMEOUT); - if (err) - break; - - i915_gem_drain_freed_objects(i915); - err = drm_gem_create_mmap_offset(&obj->base); - if (!err) - break; + err = intel_gt_retire_requests_timeout(gt, MAX_SCHEDULE_TIMEOUT); + if (err) + return err; - } while (flush_delayed_work(&i915->gem.retire_work)); - - return err; + i915_gem_drain_freed_objects(i915); + return drm_gem_create_mmap_offset(&obj->base); } int @@ -473,10 +454,16 @@ i915_gem_mmap_gtt(struct drm_file *file, if (!obj) return -ENOENT; + if (i915_gem_object_never_bind_ggtt(obj)) { + ret = -ENODEV; + goto out; + } + ret = create_mmap_offset(obj); if (ret == 0) *offset = drm_vma_node_offset_addr(&obj->base.vma_node); +out: i915_gem_object_put(obj); return ret; } diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c index d7855dc5a5c5..a50296cce0d8 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c @@ -47,9 +47,10 @@ void i915_gem_object_free(struct drm_i915_gem_object *obj) } void i915_gem_object_init(struct drm_i915_gem_object *obj, - const struct drm_i915_gem_object_ops *ops) + const struct drm_i915_gem_object_ops *ops, + struct lock_class_key *key) { - mutex_init(&obj->mm.lock); + __mutex_init(&obj->mm.lock, "obj->mm.lock", key); spin_lock_init(&obj->vma.lock); INIT_LIST_HEAD(&obj->vma.list); @@ -155,21 +156,30 @@ static void __i915_gem_free_objects(struct drm_i915_private *i915, wakeref = intel_runtime_pm_get(&i915->runtime_pm); llist_for_each_entry_safe(obj, on, freed, freed) { - struct i915_vma *vma, *vn; - trace_i915_gem_object_destroy(obj); - mutex_lock(&i915->drm.struct_mutex); - - list_for_each_entry_safe(vma, vn, &obj->vma.list, obj_link) { - GEM_BUG_ON(i915_vma_is_active(vma)); - vma->flags &= ~I915_VMA_PIN_MASK; - i915_vma_destroy(vma); + if (!list_empty(&obj->vma.list)) { + struct i915_vma *vma; + + /* + * Note that the vma keeps an object reference while + * it is active, so it *should* not sleep while we + * destroy it. Our debug code errs insits it *might*. + * For the moment, play along. + */ + spin_lock(&obj->vma.lock); + while ((vma = list_first_entry_or_null(&obj->vma.list, + struct i915_vma, + obj_link))) { + GEM_BUG_ON(vma->obj != obj); + spin_unlock(&obj->vma.lock); + + i915_vma_destroy(vma); + + spin_lock(&obj->vma.lock); + } + spin_unlock(&obj->vma.lock); } - GEM_BUG_ON(!list_empty(&obj->vma.list)); - GEM_BUG_ON(!RB_EMPTY_ROOT(&obj->vma.tree)); - - mutex_unlock(&i915->drm.struct_mutex); GEM_BUG_ON(atomic_read(&obj->bind_count)); GEM_BUG_ON(obj->userfault_count); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h index 5efb9936e05b..458cd51331f1 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h @@ -23,12 +23,14 @@ struct drm_i915_gem_object *i915_gem_object_alloc(void); void i915_gem_object_free(struct drm_i915_gem_object *obj); void i915_gem_object_init(struct drm_i915_gem_object *obj, - const struct drm_i915_gem_object_ops *ops); + const struct drm_i915_gem_object_ops *ops, + struct lock_class_key *key); struct drm_i915_gem_object * -i915_gem_object_create_shmem(struct drm_i915_private *i915, u64 size); +i915_gem_object_create_shmem(struct drm_i915_private *i915, + resource_size_t size); struct drm_i915_gem_object * i915_gem_object_create_shmem_from_data(struct drm_i915_private *i915, - const void *data, size_t size); + const void *data, resource_size_t size); extern const struct drm_i915_gem_object_ops i915_gem_shmem_ops; void __i915_gem_object_release_shmem(struct drm_i915_gem_object *obj, @@ -106,6 +108,11 @@ static inline void i915_gem_object_lock(struct drm_i915_gem_object *obj) dma_resv_lock(obj->base.resv, NULL); } +static inline bool i915_gem_object_trylock(struct drm_i915_gem_object *obj) +{ + return dma_resv_trylock(obj->base.resv); +} + static inline int i915_gem_object_lock_interruptible(struct drm_i915_gem_object *obj) { @@ -135,27 +142,58 @@ i915_gem_object_is_readonly(const struct drm_i915_gem_object *obj) } static inline bool +i915_gem_object_is_contiguous(const struct drm_i915_gem_object *obj) +{ + return obj->flags & I915_BO_ALLOC_CONTIGUOUS; +} + +static inline bool +i915_gem_object_is_volatile(const struct drm_i915_gem_object *obj) +{ + return obj->flags & I915_BO_ALLOC_VOLATILE; +} + +static inline void +i915_gem_object_set_volatile(struct drm_i915_gem_object *obj) +{ + obj->flags |= I915_BO_ALLOC_VOLATILE; +} + +static inline bool +i915_gem_object_type_has(const struct drm_i915_gem_object *obj, + unsigned long flags) +{ + return obj->ops->flags & flags; +} + +static inline bool i915_gem_object_has_struct_page(const struct drm_i915_gem_object *obj) { - return obj->ops->flags & I915_GEM_OBJECT_HAS_STRUCT_PAGE; + return i915_gem_object_type_has(obj, I915_GEM_OBJECT_HAS_STRUCT_PAGE); } static inline bool i915_gem_object_is_shrinkable(const struct drm_i915_gem_object *obj) { - return obj->ops->flags & I915_GEM_OBJECT_IS_SHRINKABLE; + return i915_gem_object_type_has(obj, I915_GEM_OBJECT_IS_SHRINKABLE); } static inline bool i915_gem_object_is_proxy(const struct drm_i915_gem_object *obj) { - return obj->ops->flags & I915_GEM_OBJECT_IS_PROXY; + return i915_gem_object_type_has(obj, I915_GEM_OBJECT_IS_PROXY); +} + +static inline bool +i915_gem_object_never_bind_ggtt(const struct drm_i915_gem_object *obj) +{ + return i915_gem_object_type_has(obj, I915_GEM_OBJECT_NO_GGTT); } static inline bool i915_gem_object_needs_async_cancel(const struct drm_i915_gem_object *obj) { - return obj->ops->flags & I915_GEM_OBJECT_ASYNC_CANCEL; + return i915_gem_object_type_has(obj, I915_GEM_OBJECT_ASYNC_CANCEL); } static inline bool @@ -406,7 +444,8 @@ static inline bool cpu_write_needs_clflush(struct drm_i915_gem_object *obj) if (!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE)) return true; - return obj->pin_global; /* currently in use by HW, keep flushed */ + /* Currently in use by HW (display engine)? Keep flushed. */ + return i915_gem_object_is_framebuffer(obj); } static inline void __start_cpu_write(struct drm_i915_gem_object *obj) @@ -423,6 +462,5 @@ int i915_gem_object_wait(struct drm_i915_gem_object *obj, int i915_gem_object_wait_priority(struct drm_i915_gem_object *obj, unsigned int flags, const struct i915_sched_attr *attr); -#define I915_PRIORITY_DISPLAY I915_USER_PRIORITY(I915_PRIORITY_MAX) #endif diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_blt.c b/drivers/gpu/drm/i915/gem/i915_gem_object_blt.c index 6415f9a17e2d..70809d8897cd 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object_blt.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_blt.c @@ -8,6 +8,7 @@ #include "gt/intel_engine_pm.h" #include "gt/intel_engine_pool.h" #include "gt/intel_gt.h" +#include "gt/intel_ring.h" #include "i915_gem_clflush.h" #include "i915_gem_object_blt.h" @@ -16,7 +17,7 @@ struct i915_vma *intel_emit_vma_fill_blt(struct intel_context *ce, u32 value) { struct drm_i915_private *i915 = ce->vm->i915; - const u32 block_size = S16_MAX * PAGE_SIZE; + const u32 block_size = SZ_8M; /* ~1ms at 8GiB/s preemption delay */ struct intel_engine_pool_node *pool; struct i915_vma *batch; u64 offset; @@ -29,10 +30,10 @@ struct i915_vma *intel_emit_vma_fill_blt(struct intel_context *ce, GEM_BUG_ON(intel_engine_is_virtual(ce->engine)); intel_engine_pm_get(ce->engine); - count = div_u64(vma->size, block_size); + count = div_u64(round_up(vma->size, block_size), block_size); size = (1 + 8 * count) * sizeof(u32); size = round_up(size, PAGE_SIZE); - pool = intel_engine_pool_get(&ce->engine->pool, size); + pool = intel_engine_get_pool(ce->engine, size); if (IS_ERR(pool)) { err = PTR_ERR(pool); goto out_pm; @@ -200,7 +201,7 @@ struct i915_vma *intel_emit_vma_copy_blt(struct intel_context *ce, struct i915_vma *dst) { struct drm_i915_private *i915 = ce->vm->i915; - const u32 block_size = S16_MAX * PAGE_SIZE; + const u32 block_size = SZ_8M; /* ~1ms at 8GiB/s preemption delay */ struct intel_engine_pool_node *pool; struct i915_vma *batch; u64 src_offset, dst_offset; @@ -213,10 +214,10 @@ struct i915_vma *intel_emit_vma_copy_blt(struct intel_context *ce, GEM_BUG_ON(intel_engine_is_virtual(ce->engine)); intel_engine_pm_get(ce->engine); - count = div_u64(dst->size, block_size); + count = div_u64(round_up(dst->size, block_size), block_size); size = (1 + 11 * count) * sizeof(u32); size = round_up(size, PAGE_SIZE); - pool = intel_engine_pool_get(&ce->engine->pool, size); + pool = intel_engine_get_pool(ce->engine, size); if (IS_ERR(pool)) { err = PTR_ERR(pool); goto out_pm; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h index ede0eb4218a8..96008374a412 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h @@ -8,6 +8,7 @@ #define __I915_GEM_OBJECT_TYPES_H__ #include <drm/drm_gem.h> +#include <uapi/drm/i915_drm.h> #include "i915_active.h" #include "i915_selftest.h" @@ -30,9 +31,11 @@ struct i915_lut_handle { struct drm_i915_gem_object_ops { unsigned int flags; #define I915_GEM_OBJECT_HAS_STRUCT_PAGE BIT(0) -#define I915_GEM_OBJECT_IS_SHRINKABLE BIT(1) -#define I915_GEM_OBJECT_IS_PROXY BIT(2) -#define I915_GEM_OBJECT_ASYNC_CANCEL BIT(3) +#define I915_GEM_OBJECT_HAS_IOMEM BIT(1) +#define I915_GEM_OBJECT_IS_SHRINKABLE BIT(2) +#define I915_GEM_OBJECT_IS_PROXY BIT(3) +#define I915_GEM_OBJECT_NO_GGTT BIT(4) +#define I915_GEM_OBJECT_ASYNC_CANCEL BIT(5) /* Interface between the GEM object and its backing storage. * get_pages() is called once prior to the use of the associated set @@ -117,6 +120,11 @@ struct drm_i915_gem_object { I915_SELFTEST_DECLARE(struct list_head st_link); + unsigned long flags; +#define I915_BO_ALLOC_CONTIGUOUS BIT(0) +#define I915_BO_ALLOC_VOLATILE BIT(1) +#define I915_BO_ALLOC_FLAGS (I915_BO_ALLOC_CONTIGUOUS | I915_BO_ALLOC_VOLATILE) + /* * Is the object to be mapped as read-only to the GPU * Only honoured if hardware has relevant pte bit @@ -152,17 +160,30 @@ struct drm_i915_gem_object { /** Count of VMA actually bound by this object */ atomic_t bind_count; - /** Count of how many global VMA are currently pinned for use by HW */ - unsigned int pin_global; struct { struct mutex lock; /* protects the pages and their use */ atomic_t pages_pin_count; + atomic_t shrink_pin; + + /** + * Memory region for this object. + */ + struct intel_memory_region *region; + /** + * List of memory region blocks allocated for this object. + */ + struct list_head blocks; + /** + * Element within memory_region->objects or region->purgeable + * if the object is marked as DONTNEED. Access is protected by + * region->obj_lock. + */ + struct list_head region_link; struct sg_table *pages; void *mapping; - /* TODO: whack some of this into the error state */ struct i915_page_sizes { /** * The sg mask of the pages sg_table. i.e the mask of diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pages.c b/drivers/gpu/drm/i915/gem/i915_gem_pages.c index 18f0ce0135c1..29f4c2850745 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_pages.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_pages.c @@ -7,6 +7,7 @@ #include "i915_drv.h" #include "i915_gem_object.h" #include "i915_scatterlist.h" +#include "i915_gem_lmem.h" void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj, struct sg_table *pages, @@ -18,6 +19,9 @@ void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj, lockdep_assert_held(&obj->mm.lock); + if (i915_gem_object_is_volatile(obj)) + obj->mm.madv = I915_MADV_DONTNEED; + /* Make the pages coherent with the GPU (flushing any swapin). */ if (obj->cache_dirty) { obj->write_domain = 0; @@ -71,6 +75,7 @@ void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj, list = &i915->mm.shrink_list; list_add_tail(&obj->mm.link, list); + atomic_set(&obj->mm.shrink_pin, 0); spin_unlock_irqrestore(&i915->mm.obj_lock, flags); } } @@ -150,6 +155,16 @@ static void __i915_gem_object_reset_page_iter(struct drm_i915_gem_object *obj) rcu_read_unlock(); } +static void unmap_object(struct drm_i915_gem_object *obj, void *ptr) +{ + if (i915_gem_object_is_lmem(obj)) + io_mapping_unmap((void __force __iomem *)ptr); + else if (is_vmalloc_addr(ptr)) + vunmap(ptr); + else + kunmap(kmap_to_page(ptr)); +} + struct sg_table * __i915_gem_object_unset_pages(struct drm_i915_gem_object *obj) { @@ -159,17 +174,13 @@ __i915_gem_object_unset_pages(struct drm_i915_gem_object *obj) if (IS_ERR_OR_NULL(pages)) return pages; + if (i915_gem_object_is_volatile(obj)) + obj->mm.madv = I915_MADV_WILLNEED; + i915_gem_object_make_unshrinkable(obj); if (obj->mm.mapping) { - void *ptr; - - ptr = page_mask_bits(obj->mm.mapping); - if (is_vmalloc_addr(ptr)) - vunmap(ptr); - else - kunmap(kmap_to_page(ptr)); - + unmap_object(obj, page_mask_bits(obj->mm.mapping)); obj->mm.mapping = NULL; } @@ -224,7 +235,7 @@ unlock: } /* The 'mapping' part of i915_gem_object_pin_map() below */ -static void *i915_gem_object_map(const struct drm_i915_gem_object *obj, +static void *i915_gem_object_map(struct drm_i915_gem_object *obj, enum i915_map_type type) { unsigned long n_pages = obj->base.size >> PAGE_SHIFT; @@ -237,6 +248,16 @@ static void *i915_gem_object_map(const struct drm_i915_gem_object *obj, pgprot_t pgprot; void *addr; + if (i915_gem_object_is_lmem(obj)) { + void __iomem *io; + + if (type != I915_MAP_WC) + return NULL; + + io = i915_gem_object_lmem_io_map(obj, 0, obj->base.size); + return (void __force *)io; + } + /* A single page can always be kmapped */ if (n_pages == 1 && type == I915_MAP_WB) return kmap(sg_page(sgt->sgl)); @@ -278,11 +299,13 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj, enum i915_map_type type) { enum i915_map_type has_type; + unsigned int flags; bool pinned; void *ptr; int err; - if (unlikely(!i915_gem_object_has_struct_page(obj))) + flags = I915_GEM_OBJECT_HAS_STRUCT_PAGE | I915_GEM_OBJECT_HAS_IOMEM; + if (!i915_gem_object_type_has(obj, flags)) return ERR_PTR(-ENXIO); err = mutex_lock_interruptible(&obj->mm.lock); @@ -314,10 +337,7 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj, goto err_unpin; } - if (is_vmalloc_addr(ptr)) - vunmap(ptr); - else - kunmap(kmap_to_page(ptr)); + unmap_object(obj, ptr); ptr = obj->mm.mapping = NULL; } diff --git a/drivers/gpu/drm/i915/gem/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/i915_gem_phys.c index 768356908160..8043ff63d73f 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_phys.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_phys.c @@ -16,6 +16,7 @@ #include "gt/intel_gt.h" #include "i915_drv.h" #include "i915_gem_object.h" +#include "i915_gem_region.h" #include "i915_scatterlist.h" static int i915_gem_object_get_pages_phys(struct drm_i915_gem_object *obj) @@ -191,8 +192,10 @@ int i915_gem_object_attach_phys(struct drm_i915_gem_object *obj, int align) /* Perma-pin (until release) the physical set of pages */ __i915_gem_object_pin_pages(obj); - if (!IS_ERR_OR_NULL(pages)) + if (!IS_ERR_OR_NULL(pages)) { i915_gem_shmem_ops.put_pages(obj, pages); + i915_gem_object_release_memory_region(obj); + } mutex_unlock(&obj->mm.lock); return 0; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c index 92e53c25424c..f88ee1317bb4 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c @@ -7,138 +7,9 @@ #include "gem/i915_gem_pm.h" #include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" +#include "gt/intel_gt_requests.h" #include "i915_drv.h" -#include "i915_globals.h" - -static void call_idle_barriers(struct intel_engine_cs *engine) -{ - struct llist_node *node, *next; - - llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks)) { - struct i915_active_request *active = - container_of((struct list_head *)node, - typeof(*active), link); - - INIT_LIST_HEAD(&active->link); - RCU_INIT_POINTER(active->request, NULL); - - active->retire(active, NULL); - } -} - -static void i915_gem_park(struct drm_i915_private *i915) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - - lockdep_assert_held(&i915->drm.struct_mutex); - - for_each_engine(engine, i915, id) - call_idle_barriers(engine); /* cleanup after wedging */ - - i915_vma_parked(i915); - - i915_globals_park(); -} - -static void idle_work_handler(struct work_struct *work) -{ - struct drm_i915_private *i915 = - container_of(work, typeof(*i915), gem.idle_work); - bool park; - - cancel_delayed_work_sync(&i915->gem.retire_work); - mutex_lock(&i915->drm.struct_mutex); - - intel_wakeref_lock(&i915->gt.wakeref); - park = (!intel_wakeref_is_active(&i915->gt.wakeref) && - !work_pending(work)); - intel_wakeref_unlock(&i915->gt.wakeref); - if (park) - i915_gem_park(i915); - else - queue_delayed_work(i915->wq, - &i915->gem.retire_work, - round_jiffies_up_relative(HZ)); - - mutex_unlock(&i915->drm.struct_mutex); -} - -static void retire_work_handler(struct work_struct *work) -{ - struct drm_i915_private *i915 = - container_of(work, typeof(*i915), gem.retire_work.work); - - /* Come back later if the device is busy... */ - if (mutex_trylock(&i915->drm.struct_mutex)) { - i915_retire_requests(i915); - mutex_unlock(&i915->drm.struct_mutex); - } - - queue_delayed_work(i915->wq, - &i915->gem.retire_work, - round_jiffies_up_relative(HZ)); -} - -static int pm_notifier(struct notifier_block *nb, - unsigned long action, - void *data) -{ - struct drm_i915_private *i915 = - container_of(nb, typeof(*i915), gem.pm_notifier); - - switch (action) { - case INTEL_GT_UNPARK: - i915_globals_unpark(); - queue_delayed_work(i915->wq, - &i915->gem.retire_work, - round_jiffies_up_relative(HZ)); - break; - - case INTEL_GT_PARK: - queue_work(i915->wq, &i915->gem.idle_work); - break; - } - - return NOTIFY_OK; -} - -static bool switch_to_kernel_context_sync(struct intel_gt *gt) -{ - bool result = !intel_gt_is_wedged(gt); - - do { - if (i915_gem_wait_for_idle(gt->i915, - I915_WAIT_LOCKED | - I915_WAIT_FOR_IDLE_BOOST, - I915_GEM_IDLE_TIMEOUT) == -ETIME) { - /* XXX hide warning from gem_eio */ - if (i915_modparams.reset) { - dev_err(gt->i915->drm.dev, - "Failed to idle engines, declaring wedged!\n"); - GEM_TRACE_DUMP(); - } - - /* - * Forcibly cancel outstanding work and leave - * the gpu quiet. - */ - intel_gt_set_wedged(gt); - result = false; - } - } while (i915_retire_requests(gt->i915) && result); - - if (intel_gt_pm_wait_for_idle(gt)) - result = false; - - return result; -} - -bool i915_gem_load_power_context(struct drm_i915_private *i915) -{ - return switch_to_kernel_context_sync(&i915->gt); -} void i915_gem_suspend(struct drm_i915_private *i915) { @@ -147,8 +18,6 @@ void i915_gem_suspend(struct drm_i915_private *i915) intel_wakeref_auto(&i915->ggtt.userfault_wakeref, 0); flush_workqueue(i915->wq); - mutex_lock(&i915->drm.struct_mutex); - /* * We have to flush all the executing contexts to main memory so * that they can saved in the hibernation image. To ensure the last @@ -158,15 +27,9 @@ void i915_gem_suspend(struct drm_i915_private *i915) * state. Fortunately, the kernel_context is disposable and we do * not rely on its state. */ - switch_to_kernel_context_sync(&i915->gt); - - mutex_unlock(&i915->drm.struct_mutex); - - cancel_delayed_work_sync(&i915->gt.hangcheck.work); + intel_gt_suspend_prepare(&i915->gt); i915_gem_drain_freed_objects(i915); - - intel_uc_suspend(&i915->gt.uc); } static struct drm_i915_gem_object *first_mm_object(struct list_head *list) @@ -206,6 +69,8 @@ void i915_gem_suspend_late(struct drm_i915_private *i915) * machine in an unusable condition. */ + intel_gt_suspend_late(&i915->gt); + spin_lock_irqsave(&i915->mm.obj_lock, flags); for (phase = phases; *phase; phase++) { LIST_HEAD(keep); @@ -230,21 +95,15 @@ void i915_gem_suspend_late(struct drm_i915_private *i915) list_splice_tail(&keep, *phase); } spin_unlock_irqrestore(&i915->mm.obj_lock, flags); - - i915_gem_sanitize(i915); } void i915_gem_resume(struct drm_i915_private *i915) { GEM_TRACE("\n"); - mutex_lock(&i915->drm.struct_mutex); intel_uncore_forcewake_get(&i915->uncore, FORCEWAKE_ALL); - i915_gem_restore_gtt_mappings(i915); - i915_gem_restore_fences(i915); - - if (i915_gem_init_hw(i915)) + if (intel_gt_init_hw(&i915->gt)) goto err_wedged; /* @@ -255,15 +114,8 @@ void i915_gem_resume(struct drm_i915_private *i915) if (intel_gt_resume(&i915->gt)) goto err_wedged; - intel_uc_resume(&i915->gt.uc); - - /* Always reload a context for powersaving. */ - if (!i915_gem_load_power_context(i915)) - goto err_wedged; - out_unlock: intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL); - mutex_unlock(&i915->drm.struct_mutex); return; err_wedged: @@ -274,13 +126,3 @@ err_wedged: } goto out_unlock; } - -void i915_gem_init__pm(struct drm_i915_private *i915) -{ - INIT_WORK(&i915->gem.idle_work, idle_work_handler); - INIT_DELAYED_WORK(&i915->gem.retire_work, retire_work_handler); - - i915->gem.pm_notifier.notifier_call = pm_notifier; - blocking_notifier_chain_register(&i915->gt.pm_notifications, - &i915->gem.pm_notifier); -} diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.h b/drivers/gpu/drm/i915/gem/i915_gem_pm.h index 6f7d5d11ac3b..26b78dbdc225 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.h @@ -12,9 +12,6 @@ struct drm_i915_private; struct work_struct; -void i915_gem_init__pm(struct drm_i915_private *i915); - -bool i915_gem_load_power_context(struct drm_i915_private *i915); void i915_gem_resume(struct drm_i915_private *i915); void i915_gem_idle_work_handler(struct work_struct *work); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_region.c b/drivers/gpu/drm/i915/gem/i915_gem_region.c new file mode 100644 index 000000000000..2f7bcfb9c964 --- /dev/null +++ b/drivers/gpu/drm/i915/gem/i915_gem_region.c @@ -0,0 +1,174 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include "intel_memory_region.h" +#include "i915_gem_region.h" +#include "i915_drv.h" +#include "i915_trace.h" + +void +i915_gem_object_put_pages_buddy(struct drm_i915_gem_object *obj, + struct sg_table *pages) +{ + __intel_memory_region_put_pages_buddy(obj->mm.region, &obj->mm.blocks); + + obj->mm.dirty = false; + sg_free_table(pages); + kfree(pages); +} + +int +i915_gem_object_get_pages_buddy(struct drm_i915_gem_object *obj) +{ + struct intel_memory_region *mem = obj->mm.region; + struct list_head *blocks = &obj->mm.blocks; + resource_size_t size = obj->base.size; + resource_size_t prev_end; + struct i915_buddy_block *block; + unsigned int flags; + struct sg_table *st; + struct scatterlist *sg; + unsigned int sg_page_sizes; + int ret; + + st = kmalloc(sizeof(*st), GFP_KERNEL); + if (!st) + return -ENOMEM; + + if (sg_alloc_table(st, size >> ilog2(mem->mm.chunk_size), GFP_KERNEL)) { + kfree(st); + return -ENOMEM; + } + + flags = I915_ALLOC_MIN_PAGE_SIZE; + if (obj->flags & I915_BO_ALLOC_CONTIGUOUS) + flags |= I915_ALLOC_CONTIGUOUS; + + ret = __intel_memory_region_get_pages_buddy(mem, size, flags, blocks); + if (ret) + goto err_free_sg; + + GEM_BUG_ON(list_empty(blocks)); + + sg = st->sgl; + st->nents = 0; + sg_page_sizes = 0; + prev_end = (resource_size_t)-1; + + list_for_each_entry(block, blocks, link) { + u64 block_size, offset; + + block_size = min_t(u64, size, + i915_buddy_block_size(&mem->mm, block)); + offset = i915_buddy_block_offset(block); + + GEM_BUG_ON(overflows_type(block_size, sg->length)); + + if (offset != prev_end || + add_overflows_t(typeof(sg->length), sg->length, block_size)) { + if (st->nents) { + sg_page_sizes |= sg->length; + sg = __sg_next(sg); + } + + sg_dma_address(sg) = mem->region.start + offset; + sg_dma_len(sg) = block_size; + + sg->length = block_size; + + st->nents++; + } else { + sg->length += block_size; + sg_dma_len(sg) += block_size; + } + + prev_end = offset + block_size; + }; + + sg_page_sizes |= sg->length; + sg_mark_end(sg); + i915_sg_trim(st); + + __i915_gem_object_set_pages(obj, st, sg_page_sizes); + + return 0; + +err_free_sg: + sg_free_table(st); + kfree(st); + return ret; +} + +void i915_gem_object_init_memory_region(struct drm_i915_gem_object *obj, + struct intel_memory_region *mem, + unsigned long flags) +{ + INIT_LIST_HEAD(&obj->mm.blocks); + obj->mm.region = intel_memory_region_get(mem); + obj->flags |= flags; + + mutex_lock(&mem->objects.lock); + + if (obj->flags & I915_BO_ALLOC_VOLATILE) + list_add(&obj->mm.region_link, &mem->objects.purgeable); + else + list_add(&obj->mm.region_link, &mem->objects.list); + + mutex_unlock(&mem->objects.lock); +} + +void i915_gem_object_release_memory_region(struct drm_i915_gem_object *obj) +{ + struct intel_memory_region *mem = obj->mm.region; + + mutex_lock(&mem->objects.lock); + list_del(&obj->mm.region_link); + mutex_unlock(&mem->objects.lock); + + intel_memory_region_put(mem); +} + +struct drm_i915_gem_object * +i915_gem_object_create_region(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags) +{ + struct drm_i915_gem_object *obj; + + /* + * NB: Our use of resource_size_t for the size stems from using struct + * resource for the mem->region. We might need to revisit this in the + * future. + */ + + GEM_BUG_ON(flags & ~I915_BO_ALLOC_FLAGS); + + if (!mem) + return ERR_PTR(-ENODEV); + + size = round_up(size, mem->min_page_size); + + GEM_BUG_ON(!size); + GEM_BUG_ON(!IS_ALIGNED(size, I915_GTT_MIN_ALIGNMENT)); + + /* + * XXX: There is a prevalence of the assumption that we fit the + * object's page count inside a 32bit _signed_ variable. Let's document + * this and catch if we ever need to fix it. In the meantime, if you do + * spot such a local variable, please consider fixing! + */ + + if (size >> PAGE_SHIFT > INT_MAX) + return ERR_PTR(-E2BIG); + + if (overflows_type(size, obj->base.size)) + return ERR_PTR(-E2BIG); + + obj = mem->ops->create_object(mem, size, flags); + if (!IS_ERR(obj)) + trace_i915_gem_object_create(obj); + + return obj; +} diff --git a/drivers/gpu/drm/i915/gem/i915_gem_region.h b/drivers/gpu/drm/i915/gem/i915_gem_region.h new file mode 100644 index 000000000000..f2ff6f8bff74 --- /dev/null +++ b/drivers/gpu/drm/i915/gem/i915_gem_region.h @@ -0,0 +1,29 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __I915_GEM_REGION_H__ +#define __I915_GEM_REGION_H__ + +#include <linux/types.h> + +struct intel_memory_region; +struct drm_i915_gem_object; +struct sg_table; + +int i915_gem_object_get_pages_buddy(struct drm_i915_gem_object *obj); +void i915_gem_object_put_pages_buddy(struct drm_i915_gem_object *obj, + struct sg_table *pages); + +void i915_gem_object_init_memory_region(struct drm_i915_gem_object *obj, + struct intel_memory_region *mem, + unsigned long flags); +void i915_gem_object_release_memory_region(struct drm_i915_gem_object *obj); + +struct drm_i915_gem_object * +i915_gem_object_create_region(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags); + +#endif diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c index 4c4954e8ce0a..4d69c3fc3439 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c @@ -7,7 +7,9 @@ #include <linux/pagevec.h> #include <linux/swap.h> +#include "gem/i915_gem_region.h" #include "i915_drv.h" +#include "i915_gemfs.h" #include "i915_gem_object.h" #include "i915_scatterlist.h" #include "i915_trace.h" @@ -26,6 +28,7 @@ static void check_release_pagevec(struct pagevec *pvec) static int shmem_get_pages(struct drm_i915_gem_object *obj) { struct drm_i915_private *i915 = to_i915(obj->base.dev); + struct intel_memory_region *mem = obj->mm.region; const unsigned long page_count = obj->base.size / PAGE_SIZE; unsigned long i; struct address_space *mapping; @@ -52,7 +55,7 @@ static int shmem_get_pages(struct drm_i915_gem_object *obj) * If there's no chance of allocating enough pages for the whole * object, bail early. */ - if (page_count > totalram_pages()) + if (obj->base.size > resource_size(&mem->region)) return -ENOMEM; st = kmalloc(sizeof(*st), GFP_KERNEL); @@ -417,6 +420,8 @@ shmem_pwrite(struct drm_i915_gem_object *obj, static void shmem_release(struct drm_i915_gem_object *obj) { + i915_gem_object_release_memory_region(obj); + fput(obj->base.filp); } @@ -434,9 +439,9 @@ const struct drm_i915_gem_object_ops i915_gem_shmem_ops = { .release = shmem_release, }; -static int create_shmem(struct drm_i915_private *i915, - struct drm_gem_object *obj, - size_t size) +static int __create_shmem(struct drm_i915_private *i915, + struct drm_gem_object *obj, + resource_size_t size) { unsigned long flags = VM_NORESERVE; struct file *filp; @@ -455,31 +460,24 @@ static int create_shmem(struct drm_i915_private *i915, return 0; } -struct drm_i915_gem_object * -i915_gem_object_create_shmem(struct drm_i915_private *i915, u64 size) +static struct drm_i915_gem_object * +create_shmem(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags) { + static struct lock_class_key lock_class; + struct drm_i915_private *i915 = mem->i915; struct drm_i915_gem_object *obj; struct address_space *mapping; unsigned int cache_level; gfp_t mask; int ret; - /* There is a prevalence of the assumption that we fit the object's - * page count inside a 32bit _signed_ variable. Let's document this and - * catch if we ever need to fix it. In the meantime, if you do spot - * such a local variable, please consider fixing! - */ - if (size >> PAGE_SHIFT > INT_MAX) - return ERR_PTR(-E2BIG); - - if (overflows_type(size, obj->base.size)) - return ERR_PTR(-E2BIG); - obj = i915_gem_object_alloc(); if (!obj) return ERR_PTR(-ENOMEM); - ret = create_shmem(i915, &obj->base, size); + ret = __create_shmem(i915, &obj->base, size); if (ret) goto fail; @@ -494,7 +492,7 @@ i915_gem_object_create_shmem(struct drm_i915_private *i915, u64 size) mapping_set_gfp_mask(mapping, mask); GEM_BUG_ON(!(mapping_gfp_mask(mapping) & __GFP_RECLAIM)); - i915_gem_object_init(obj, &i915_gem_shmem_ops); + i915_gem_object_init(obj, &i915_gem_shmem_ops, &lock_class); obj->write_domain = I915_GEM_DOMAIN_CPU; obj->read_domains = I915_GEM_DOMAIN_CPU; @@ -518,7 +516,7 @@ i915_gem_object_create_shmem(struct drm_i915_private *i915, u64 size) i915_gem_object_set_cache_coherency(obj, cache_level); - trace_i915_gem_object_create(obj); + i915_gem_object_init_memory_region(obj, mem, 0); return obj; @@ -527,14 +525,22 @@ fail: return ERR_PTR(ret); } +struct drm_i915_gem_object * +i915_gem_object_create_shmem(struct drm_i915_private *i915, + resource_size_t size) +{ + return i915_gem_object_create_region(i915->mm.regions[INTEL_REGION_SMEM], + size, 0); +} + /* Allocate a new GEM object and fill it with the supplied data */ struct drm_i915_gem_object * i915_gem_object_create_shmem_from_data(struct drm_i915_private *dev_priv, - const void *data, size_t size) + const void *data, resource_size_t size) { struct drm_i915_gem_object *obj; struct file *file; - size_t offset; + resource_size_t offset; int err; obj = i915_gem_object_create_shmem(dev_priv, round_up(size, PAGE_SIZE)); @@ -577,3 +583,35 @@ fail: i915_gem_object_put(obj); return ERR_PTR(err); } + +static int init_shmem(struct intel_memory_region *mem) +{ + int err; + + err = i915_gemfs_init(mem->i915); + if (err) { + DRM_NOTE("Unable to create a private tmpfs mount, hugepage support will be disabled(%d).\n", + err); + } + + return 0; /* Don't error, we can simply fallback to the kernel mnt */ +} + +static void release_shmem(struct intel_memory_region *mem) +{ + i915_gemfs_fini(mem->i915); +} + +static const struct intel_memory_region_ops shmem_region_ops = { + .init = init_shmem, + .release = release_shmem, + .create_object = create_shmem, +}; + +struct intel_memory_region *i915_gem_shmem_setup(struct drm_i915_private *i915) +{ + return intel_memory_region_create(i915, 0, + totalram_pages() << PAGE_SHIFT, + PAGE_SIZE, 0, + &shmem_region_ops); +} diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c index edd21d14e64f..f2418a1cfe68 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c @@ -16,40 +16,6 @@ #include "i915_trace.h" -static bool shrinker_lock(struct drm_i915_private *i915, - unsigned int flags, - bool *unlock) -{ - struct mutex *m = &i915->drm.struct_mutex; - - switch (mutex_trylock_recursive(m)) { - case MUTEX_TRYLOCK_RECURSIVE: - *unlock = false; - return true; - - case MUTEX_TRYLOCK_FAILED: - *unlock = false; - if (flags & I915_SHRINK_ACTIVE && - mutex_lock_killable_nested(m, I915_MM_SHRINKER) == 0) - *unlock = true; - return *unlock; - - case MUTEX_TRYLOCK_SUCCESS: - *unlock = true; - return true; - } - - BUG(); -} - -static void shrinker_unlock(struct drm_i915_private *i915, bool unlock) -{ - if (!unlock) - return; - - mutex_unlock(&i915->drm.struct_mutex); -} - static bool swap_available(void) { return get_nr_swap_pages() > 0; @@ -61,7 +27,8 @@ static bool can_release_pages(struct drm_i915_gem_object *obj) if (!i915_gem_object_is_shrinkable(obj)) return false; - /* Only report true if by unbinding the object and putting its pages + /* + * Only report true if by unbinding the object and putting its pages * we can actually make forward progress towards freeing physical * pages. * @@ -72,16 +39,8 @@ static bool can_release_pages(struct drm_i915_gem_object *obj) if (atomic_read(&obj->mm.pages_pin_count) > atomic_read(&obj->bind_count)) return false; - /* If any vma are "permanently" pinned, it will prevent us from - * reclaiming the obj->mm.pages. We only allow scanout objects to claim - * a permanent pin, along with a few others like the context objects. - * To simplify the scan, and to avoid walking the list of vma under the - * object, we just check the count of its permanently pinned. - */ - if (READ_ONCE(obj->pin_global)) - return false; - - /* We can only return physical pages to the system if we can either + /* + * We can only return physical pages to the system if we can either * discard the contents (because the user has marked them as being * purgeable) or if we can move their contents out to swap. */ @@ -162,10 +121,6 @@ i915_gem_shrink(struct drm_i915_private *i915, intel_wakeref_t wakeref = 0; unsigned long count = 0; unsigned long scanned = 0; - bool unlock; - - if (!shrinker_lock(i915, shrink, &unlock)) - return 0; /* * When shrinking the active list, we should also consider active @@ -275,8 +230,6 @@ i915_gem_shrink(struct drm_i915_private *i915, if (shrink & I915_SHRINK_BOUND) intel_runtime_pm_put(&i915->runtime_pm, wakeref); - shrinker_unlock(i915, unlock); - if (nr_scanned) *nr_scanned += scanned; return count; @@ -346,19 +299,14 @@ i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc) struct drm_i915_private *i915 = container_of(shrinker, struct drm_i915_private, mm.shrinker); unsigned long freed; - bool unlock; sc->nr_scanned = 0; - if (!shrinker_lock(i915, 0, &unlock)) - return SHRINK_STOP; - freed = i915_gem_shrink(i915, sc->nr_to_scan, &sc->nr_scanned, I915_SHRINK_BOUND | - I915_SHRINK_UNBOUND | - I915_SHRINK_WRITEBACK); + I915_SHRINK_UNBOUND); if (sc->nr_scanned < sc->nr_to_scan && current_is_kswapd()) { intel_wakeref_t wakeref; @@ -373,8 +321,6 @@ i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc) } } - shrinker_unlock(i915, unlock); - return sc->nr_scanned ? freed : SHRINK_STOP; } @@ -391,6 +337,7 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr) freed_pages = 0; with_intel_runtime_pm(&i915->runtime_pm, wakeref) freed_pages += i915_gem_shrink(i915, -1UL, NULL, + I915_SHRINK_ACTIVE | I915_SHRINK_BOUND | I915_SHRINK_UNBOUND | I915_SHRINK_WRITEBACK); @@ -426,10 +373,6 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr struct i915_vma *vma, *next; unsigned long freed_pages = 0; intel_wakeref_t wakeref; - bool unlock; - - if (!shrinker_lock(i915, 0, &unlock)) - return NOTIFY_DONE; with_intel_runtime_pm(&i915->runtime_pm, wakeref) freed_pages += i915_gem_shrink(i915, -1UL, NULL, @@ -446,15 +389,11 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr if (!vma->iomap || i915_vma_is_active(vma)) continue; - mutex_unlock(&i915->ggtt.vm.mutex); - if (i915_vma_unbind(vma) == 0) + if (__i915_vma_unbind(vma) == 0) freed_pages += count; - mutex_lock(&i915->ggtt.vm.mutex); } mutex_unlock(&i915->ggtt.vm.mutex); - shrinker_unlock(i915, unlock); - *(unsigned long *)ptr += freed_pages; return NOTIFY_DONE; } @@ -497,72 +436,65 @@ void i915_gem_shrinker_taints_mutex(struct drm_i915_private *i915, fs_reclaim_acquire(GFP_KERNEL); - /* - * As we invariably rely on the struct_mutex within the shrinker, - * but have a complicated recursion dance, taint all the mutexes used - * within the shrinker with the struct_mutex. For completeness, we - * taint with all subclass of struct_mutex, even though we should - * only need tainting by I915_MM_NORMAL to catch possible ABBA - * deadlocks from using struct_mutex inside @mutex. - */ - mutex_acquire(&i915->drm.struct_mutex.dep_map, - I915_MM_SHRINKER, 0, _RET_IP_); - mutex_acquire(&mutex->dep_map, 0, 0, _RET_IP_); - mutex_release(&mutex->dep_map, 0, _RET_IP_); - - mutex_release(&i915->drm.struct_mutex.dep_map, 0, _RET_IP_); + mutex_release(&mutex->dep_map, _RET_IP_); fs_reclaim_release(GFP_KERNEL); if (unlock) - mutex_release(&i915->drm.struct_mutex.dep_map, 0, _RET_IP_); + mutex_release(&i915->drm.struct_mutex.dep_map, _RET_IP_); } #define obj_to_i915(obj__) to_i915((obj__)->base.dev) void i915_gem_object_make_unshrinkable(struct drm_i915_gem_object *obj) { + struct drm_i915_private *i915 = obj_to_i915(obj); + unsigned long flags; + /* * We can only be called while the pages are pinned or when * the pages are released. If pinned, we should only be called * from a single caller under controlled conditions; and on release * only one caller may release us. Neither the two may cross. */ - if (!list_empty(&obj->mm.link)) { /* pinned by caller */ - struct drm_i915_private *i915 = obj_to_i915(obj); - unsigned long flags; - - spin_lock_irqsave(&i915->mm.obj_lock, flags); - GEM_BUG_ON(list_empty(&obj->mm.link)); + if (atomic_add_unless(&obj->mm.shrink_pin, 1, 0)) + return; + spin_lock_irqsave(&i915->mm.obj_lock, flags); + if (!atomic_fetch_inc(&obj->mm.shrink_pin) && + !list_empty(&obj->mm.link)) { list_del_init(&obj->mm.link); i915->mm.shrink_count--; i915->mm.shrink_memory -= obj->base.size; - - spin_unlock_irqrestore(&i915->mm.obj_lock, flags); } + spin_unlock_irqrestore(&i915->mm.obj_lock, flags); } static void __i915_gem_object_make_shrinkable(struct drm_i915_gem_object *obj, struct list_head *head) { + struct drm_i915_private *i915 = obj_to_i915(obj); + unsigned long flags; + GEM_BUG_ON(!i915_gem_object_has_pages(obj)); - GEM_BUG_ON(!list_empty(&obj->mm.link)); + if (!i915_gem_object_is_shrinkable(obj)) + return; - if (i915_gem_object_is_shrinkable(obj)) { - struct drm_i915_private *i915 = obj_to_i915(obj); - unsigned long flags; + if (atomic_add_unless(&obj->mm.shrink_pin, -1, 1)) + return; - spin_lock_irqsave(&i915->mm.obj_lock, flags); - GEM_BUG_ON(!kref_read(&obj->base.refcount)); + spin_lock_irqsave(&i915->mm.obj_lock, flags); + GEM_BUG_ON(!kref_read(&obj->base.refcount)); + if (atomic_dec_and_test(&obj->mm.shrink_pin)) { + GEM_BUG_ON(!list_empty(&obj->mm.link)); list_add_tail(&obj->mm.link, head); i915->mm.shrink_count++; i915->mm.shrink_memory += obj->base.size; - spin_unlock_irqrestore(&i915->mm.obj_lock, flags); } + spin_unlock_irqrestore(&i915->mm.obj_lock, flags); } void i915_gem_object_make_shrinkable(struct drm_i915_gem_object *obj) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c index aa533b4ab5f5..a2d49c04e6a4 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c @@ -10,6 +10,7 @@ #include <drm/drm_mm.h> #include <drm/i915_drm.h> +#include "gem/i915_gem_region.h" #include "i915_drv.h" #include "i915_gem_stolen.h" @@ -150,7 +151,7 @@ static int i915_adjust_stolen(struct drm_i915_private *dev_priv, return 0; } -void i915_gem_cleanup_stolen(struct drm_i915_private *dev_priv) +static void i915_gem_cleanup_stolen(struct drm_i915_private *dev_priv) { if (!drm_mm_initialized(&dev_priv->mm.stolen)) return; @@ -355,7 +356,7 @@ static void icl_get_stolen_reserved(struct drm_i915_private *i915, } } -int i915_gem_init_stolen(struct drm_i915_private *dev_priv) +static int i915_gem_init_stolen(struct drm_i915_private *dev_priv) { resource_size_t reserved_base, stolen_top; resource_size_t reserved_total, reserved_size; @@ -425,8 +426,11 @@ int i915_gem_init_stolen(struct drm_i915_private *dev_priv) bdw_get_stolen_reserved(dev_priv, &reserved_base, &reserved_size); break; - case 11: default: + MISSING_CASE(INTEL_GEN(dev_priv)); + /* fall-through */ + case 11: + case 12: icl_get_stolen_reserved(dev_priv, &reserved_base, &reserved_size); break; @@ -536,6 +540,9 @@ i915_gem_object_release_stolen(struct drm_i915_gem_object *obj) i915_gem_stolen_remove_node(dev_priv, stolen); kfree(stolen); + + if (obj->mm.region) + i915_gem_object_release_memory_region(obj); } static const struct drm_i915_gem_object_ops i915_gem_object_stolen_ops = { @@ -545,65 +552,116 @@ static const struct drm_i915_gem_object_ops i915_gem_object_stolen_ops = { }; static struct drm_i915_gem_object * -_i915_gem_object_create_stolen(struct drm_i915_private *dev_priv, - struct drm_mm_node *stolen) +__i915_gem_object_create_stolen(struct drm_i915_private *dev_priv, + struct drm_mm_node *stolen, + struct intel_memory_region *mem) { + static struct lock_class_key lock_class; struct drm_i915_gem_object *obj; unsigned int cache_level; + int err = -ENOMEM; obj = i915_gem_object_alloc(); - if (obj == NULL) - return NULL; + if (!obj) + goto err; drm_gem_private_object_init(&dev_priv->drm, &obj->base, stolen->size); - i915_gem_object_init(obj, &i915_gem_object_stolen_ops); + i915_gem_object_init(obj, &i915_gem_object_stolen_ops, &lock_class); obj->stolen = stolen; obj->read_domains = I915_GEM_DOMAIN_CPU | I915_GEM_DOMAIN_GTT; cache_level = HAS_LLC(dev_priv) ? I915_CACHE_LLC : I915_CACHE_NONE; i915_gem_object_set_cache_coherency(obj, cache_level); - if (i915_gem_object_pin_pages(obj)) + err = i915_gem_object_pin_pages(obj); + if (err) goto cleanup; + if (mem) + i915_gem_object_init_memory_region(obj, mem, 0); + return obj; cleanup: i915_gem_object_free(obj); - return NULL; +err: + return ERR_PTR(err); } -struct drm_i915_gem_object * -i915_gem_object_create_stolen(struct drm_i915_private *dev_priv, - resource_size_t size) +static struct drm_i915_gem_object * +_i915_gem_object_create_stolen(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags) { + struct drm_i915_private *dev_priv = mem->i915; struct drm_i915_gem_object *obj; struct drm_mm_node *stolen; int ret; if (!drm_mm_initialized(&dev_priv->mm.stolen)) - return NULL; + return ERR_PTR(-ENODEV); if (size == 0) - return NULL; + return ERR_PTR(-EINVAL); stolen = kzalloc(sizeof(*stolen), GFP_KERNEL); if (!stolen) - return NULL; + return ERR_PTR(-ENOMEM); ret = i915_gem_stolen_insert_node(dev_priv, stolen, size, 4096); if (ret) { - kfree(stolen); - return NULL; + obj = ERR_PTR(ret); + goto err_free; } - obj = _i915_gem_object_create_stolen(dev_priv, stolen); - if (obj) - return obj; + obj = __i915_gem_object_create_stolen(dev_priv, stolen, mem); + if (IS_ERR(obj)) + goto err_remove; + return obj; + +err_remove: i915_gem_stolen_remove_node(dev_priv, stolen); +err_free: kfree(stolen); - return NULL; + return obj; +} + +struct drm_i915_gem_object * +i915_gem_object_create_stolen(struct drm_i915_private *dev_priv, + resource_size_t size) +{ + return i915_gem_object_create_region(dev_priv->mm.regions[INTEL_REGION_STOLEN], + size, I915_BO_ALLOC_CONTIGUOUS); +} + +static int init_stolen(struct intel_memory_region *mem) +{ + /* + * Initialise stolen early so that we may reserve preallocated + * objects for the BIOS to KMS transition. + */ + return i915_gem_init_stolen(mem->i915); +} + +static void release_stolen(struct intel_memory_region *mem) +{ + i915_gem_cleanup_stolen(mem->i915); +} + +static const struct intel_memory_region_ops i915_region_stolen_ops = { + .init = init_stolen, + .release = release_stolen, + .create_object = _i915_gem_object_create_stolen, +}; + +struct intel_memory_region *i915_gem_stolen_setup(struct drm_i915_private *i915) +{ + return intel_memory_region_create(i915, + intel_graphics_stolen_res.start, + resource_size(&intel_graphics_stolen_res), + PAGE_SIZE, 0, + &i915_region_stolen_ops); } struct drm_i915_gem_object * @@ -619,9 +677,7 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv int ret; if (!drm_mm_initialized(&dev_priv->mm.stolen)) - return NULL; - - lockdep_assert_held(&dev_priv->drm.struct_mutex); + return ERR_PTR(-ENODEV); DRM_DEBUG_DRIVER("creating preallocated stolen object: stolen_offset=%pa, gtt_offset=%pa, size=%pa\n", &stolen_offset, >t_offset, &size); @@ -630,11 +686,11 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv if (WARN_ON(size == 0) || WARN_ON(!IS_ALIGNED(size, I915_GTT_PAGE_SIZE)) || WARN_ON(!IS_ALIGNED(stolen_offset, I915_GTT_MIN_ALIGNMENT))) - return NULL; + return ERR_PTR(-EINVAL); stolen = kzalloc(sizeof(*stolen), GFP_KERNEL); if (!stolen) - return NULL; + return ERR_PTR(-ENOMEM); stolen->start = stolen_offset; stolen->size = size; @@ -644,15 +700,15 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv if (ret) { DRM_DEBUG_DRIVER("failed to allocate stolen space\n"); kfree(stolen); - return NULL; + return ERR_PTR(ret); } - obj = _i915_gem_object_create_stolen(dev_priv, stolen); - if (obj == NULL) { + obj = __i915_gem_object_create_stolen(dev_priv, stolen, NULL); + if (IS_ERR(obj)) { DRM_DEBUG_DRIVER("failed to allocate stolen object\n"); i915_gem_stolen_remove_node(dev_priv, stolen); kfree(stolen); - return NULL; + return obj; } /* Some objects just need physical mem from stolen space */ @@ -674,22 +730,26 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv * setting up the GTT space. The actual reservation will occur * later. */ + mutex_lock(&ggtt->vm.mutex); ret = i915_gem_gtt_reserve(&ggtt->vm, &vma->node, size, gtt_offset, obj->cache_level, 0); if (ret) { DRM_DEBUG_DRIVER("failed to allocate stolen GTT space\n"); + mutex_unlock(&ggtt->vm.mutex); goto err_pages; } GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); + GEM_BUG_ON(vma->pages); vma->pages = obj->mm.pages; - vma->flags |= I915_VMA_GLOBAL_BIND; + atomic_set(&vma->pages_count, I915_VMA_PAGES_ACTIVE); + + set_bit(I915_VMA_GLOBAL_BIND_BIT, __i915_vma_flags(vma)); __i915_vma_set_map_and_fenceable(vma); - mutex_lock(&ggtt->vm.mutex); - list_move_tail(&vma->vm_link, &ggtt->vm.bound_list); + list_add_tail(&vma->vm_link, &ggtt->vm.bound_list); mutex_unlock(&ggtt->vm.mutex); GEM_BUG_ON(i915_gem_object_is_shrinkable(obj)); @@ -701,5 +761,5 @@ err_pages: i915_gem_object_unpin_pages(obj); err: i915_gem_object_put(obj); - return NULL; + return ERR_PTR(ret); } diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.h b/drivers/gpu/drm/i915/gem/i915_gem_stolen.h index 2289644d8604..c1040627fbf3 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.h @@ -21,8 +21,7 @@ int i915_gem_stolen_insert_node_in_range(struct drm_i915_private *dev_priv, u64 end); void i915_gem_stolen_remove_node(struct drm_i915_private *dev_priv, struct drm_mm_node *node); -int i915_gem_init_stolen(struct drm_i915_private *dev_priv); -void i915_gem_cleanup_stolen(struct drm_i915_private *dev_priv); +struct intel_memory_region *i915_gem_stolen_setup(struct drm_i915_private *i915); struct drm_i915_gem_object * i915_gem_object_create_stolen(struct drm_i915_private *dev_priv, resource_size_t size); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_throttle.c b/drivers/gpu/drm/i915/gem/i915_gem_throttle.c index 1e372420771b..540ef0551789 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_throttle.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_throttle.c @@ -50,10 +50,8 @@ i915_gem_throttle_ioctl(struct drm_device *dev, void *data, if (time_after_eq(request->emitted_jiffies, recent_enough)) break; - if (target) { + if (target && xchg(&target->file_priv, NULL)) list_del(&target->client_link); - target->file_priv = NULL; - } target = request; } diff --git a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c index ca0c2f451742..1fa592d82af5 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c @@ -181,22 +181,25 @@ static int i915_gem_object_fence_prepare(struct drm_i915_gem_object *obj, int tiling_mode, unsigned int stride) { + struct i915_ggtt *ggtt = &to_i915(obj->base.dev)->ggtt; struct i915_vma *vma; - int ret; + int ret = 0; if (tiling_mode == I915_TILING_NONE) return 0; + mutex_lock(&ggtt->vm.mutex); for_each_ggtt_vma(vma, obj) { if (i915_vma_fence_prepare(vma, tiling_mode, stride)) continue; - ret = i915_vma_unbind(vma); + ret = __i915_vma_unbind(vma); if (ret) - return ret; + break; } + mutex_unlock(&ggtt->vm.mutex); - return 0; + return ret; } int @@ -212,7 +215,6 @@ i915_gem_object_set_tiling(struct drm_i915_gem_object *obj, GEM_BUG_ON(!i915_tiling_ok(obj, tiling, stride)); GEM_BUG_ON(!stride ^ (tiling == I915_TILING_NONE)); - lockdep_assert_held(&i915->drm.struct_mutex); if ((tiling | stride) == obj->tiling_and_stride) return 0; @@ -233,16 +235,18 @@ i915_gem_object_set_tiling(struct drm_i915_gem_object *obj, * whilst executing a fenced command for an untiled object. */ - err = i915_gem_object_fence_prepare(obj, tiling, stride); - if (err) - return err; - i915_gem_object_lock(obj); if (i915_gem_object_is_framebuffer(obj)) { i915_gem_object_unlock(obj); return -EBUSY; } + err = i915_gem_object_fence_prepare(obj, tiling, stride); + if (err) { + i915_gem_object_unlock(obj); + return err; + } + /* If the memory has unknown (i.e. varying) swizzling, we pin the * pages to prevent them being swapped out and causing corruption * due to the change in swizzling. @@ -313,10 +317,14 @@ int i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data, struct drm_file *file) { + struct drm_i915_private *dev_priv = to_i915(dev); struct drm_i915_gem_set_tiling *args = data; struct drm_i915_gem_object *obj; int err; + if (!dev_priv->ggtt.num_fences) + return -EOPNOTSUPP; + obj = i915_gem_object_lookup(file, args->handle); if (!obj) return -ENOENT; @@ -340,9 +348,9 @@ i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data, args->stride = 0; } else { if (args->tiling_mode == I915_TILING_X) - args->swizzle_mode = to_i915(dev)->mm.bit_6_swizzle_x; + args->swizzle_mode = to_i915(dev)->ggtt.bit_6_swizzle_x; else - args->swizzle_mode = to_i915(dev)->mm.bit_6_swizzle_y; + args->swizzle_mode = to_i915(dev)->ggtt.bit_6_swizzle_y; /* Hide bit 17 swizzling from the user. This prevents old Mesa * from aborting the application on sw fallbacks to bit 17, @@ -364,12 +372,7 @@ i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data, } } - err = mutex_lock_interruptible(&dev->struct_mutex); - if (err) - goto err; - err = i915_gem_object_set_tiling(obj, args->tiling_mode, args->stride); - mutex_unlock(&dev->struct_mutex); /* We have to maintain this existing ABI... */ args->stride = i915_gem_object_get_stride(obj); @@ -402,6 +405,9 @@ i915_gem_get_tiling_ioctl(struct drm_device *dev, void *data, struct drm_i915_gem_object *obj; int err = -ENOENT; + if (!dev_priv->ggtt.num_fences) + return -EOPNOTSUPP; + rcu_read_lock(); obj = i915_gem_object_lookup_rcu(file, args->handle); if (obj) { @@ -415,10 +421,10 @@ i915_gem_get_tiling_ioctl(struct drm_device *dev, void *data, switch (args->tiling_mode) { case I915_TILING_X: - args->swizzle_mode = dev_priv->mm.bit_6_swizzle_x; + args->swizzle_mode = dev_priv->ggtt.bit_6_swizzle_x; break; case I915_TILING_Y: - args->swizzle_mode = dev_priv->mm.bit_6_swizzle_y; + args->swizzle_mode = dev_priv->ggtt.bit_6_swizzle_y; break; default: case I915_TILING_NONE: diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c index 11b231c187c5..4c72d74d6576 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c @@ -92,7 +92,6 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn, struct i915_mmu_notifier *mn = container_of(_mn, struct i915_mmu_notifier, mn); struct interval_tree_node *it; - struct mutex *unlock = NULL; unsigned long end; int ret = 0; @@ -129,33 +128,13 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn, } spin_unlock(&mn->lock); - if (!unlock) { - unlock = &mn->mm->i915->drm.struct_mutex; - - switch (mutex_trylock_recursive(unlock)) { - default: - case MUTEX_TRYLOCK_FAILED: - if (mutex_lock_killable_nested(unlock, I915_MM_SHRINKER)) { - i915_gem_object_put(obj); - return -EINTR; - } - /* fall through */ - case MUTEX_TRYLOCK_SUCCESS: - break; - - case MUTEX_TRYLOCK_RECURSIVE: - unlock = ERR_PTR(-EEXIST); - break; - } - } - ret = i915_gem_object_unbind(obj, I915_GEM_OBJECT_UNBIND_ACTIVE); if (ret == 0) ret = __i915_gem_object_put_pages(obj, I915_MM_SHRINKER); i915_gem_object_put(obj); if (ret) - goto unlock; + return ret; spin_lock(&mn->lock); @@ -168,10 +147,6 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn, } spin_unlock(&mn->lock); -unlock: - if (!IS_ERR_OR_NULL(unlock)) - mutex_unlock(unlock); - return ret; } @@ -671,8 +646,28 @@ i915_gem_userptr_put_pages(struct drm_i915_gem_object *obj, obj->mm.dirty = false; for_each_sgt_page(page, sgt_iter, pages) { - if (obj->mm.dirty) + if (obj->mm.dirty && trylock_page(page)) { + /* + * As this may not be anonymous memory (e.g. shmem) + * but exist on a real mapping, we have to lock + * the page in order to dirty it -- holding + * the page reference is not sufficient to + * prevent the inode from being truncated. + * Play safe and take the lock. + * + * However...! + * + * The mmu-notifier can be invalidated for a + * migrate_page, that is alreadying holding the lock + * on the page. Such a try_to_unmap() will result + * in us calling put_pages() and so recursively try + * to lock the page. We avoid that deadlock with + * a trylock_page() and in exchange we risk missing + * some page dirtying. + */ set_page_dirty(page); + unlock_page(page); + } mark_page_accessed(page); put_page(page); @@ -702,6 +697,7 @@ i915_gem_userptr_dmabuf_export(struct drm_i915_gem_object *obj) static const struct drm_i915_gem_object_ops i915_gem_userptr_ops = { .flags = I915_GEM_OBJECT_HAS_STRUCT_PAGE | I915_GEM_OBJECT_IS_SHRINKABLE | + I915_GEM_OBJECT_NO_GGTT | I915_GEM_OBJECT_ASYNC_CANCEL, .get_pages = i915_gem_userptr_get_pages, .put_pages = i915_gem_userptr_put_pages, @@ -749,6 +745,7 @@ i915_gem_userptr_ioctl(struct drm_device *dev, void *data, struct drm_file *file) { + static struct lock_class_key lock_class; struct drm_i915_private *dev_priv = to_i915(dev); struct drm_i915_gem_userptr *args = data; struct drm_i915_gem_object *obj; @@ -782,7 +779,8 @@ i915_gem_userptr_ioctl(struct drm_device *dev, * On almost all of the older hw, we cannot tell the GPU that * a page is readonly. */ - vm = dev_priv->kernel_context->vm; + vm = rcu_dereference_protected(dev_priv->kernel_context->vm, + true); /* static vm */ if (!vm || !vm->has_read_only) return -ENODEV; } @@ -792,7 +790,7 @@ i915_gem_userptr_ioctl(struct drm_device *dev, return -ENOMEM; drm_gem_private_object_init(dev, &obj->base, args->user_size); - i915_gem_object_init(obj, &i915_gem_userptr_ops); + i915_gem_object_init(obj, &i915_gem_userptr_ops, &lock_class); obj->read_domains = I915_GEM_DOMAIN_CPU; obj->write_domain = I915_GEM_DOMAIN_CPU; i915_gem_object_set_cache_coherency(obj, I915_CACHE_LLC); diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c b/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c index 3c5d17b2b670..892d12db6c49 100644 --- a/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c +++ b/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c @@ -96,6 +96,7 @@ huge_gem_object(struct drm_i915_private *i915, phys_addr_t phys_size, dma_addr_t dma_size) { + static struct lock_class_key lock_class; struct drm_i915_gem_object *obj; unsigned int cache_level; @@ -111,7 +112,7 @@ huge_gem_object(struct drm_i915_private *i915, return ERR_PTR(-ENOMEM); drm_gem_private_object_init(&i915->drm, &obj->base, dma_size); - i915_gem_object_init(obj, &huge_ops); + i915_gem_object_init(obj, &huge_ops, &lock_class); obj->read_domains = I915_GEM_DOMAIN_CPU; obj->write_domain = I915_GEM_DOMAIN_CPU; diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c index 8de83c6d81f5..688c49a24f32 100644 --- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c @@ -8,6 +8,8 @@ #include "i915_selftest.h" +#include "gem/i915_gem_region.h" +#include "gem/i915_gem_lmem.h" #include "gem/i915_gem_pm.h" #include "gt/intel_gt.h" @@ -17,6 +19,7 @@ #include "selftests/mock_drm.h" #include "selftests/mock_gem_device.h" +#include "selftests/mock_region.h" #include "selftests/i915_random.h" static const unsigned int page_sizes[] = { @@ -113,8 +116,6 @@ static int get_huge_pages(struct drm_i915_gem_object *obj) if (i915_gem_gtt_prepare_pages(obj, st)) goto err; - obj->mm.madv = I915_MADV_DONTNEED; - GEM_BUG_ON(sg_page_sizes != obj->mm.page_mask); __i915_gem_object_set_pages(obj, st, sg_page_sizes); @@ -135,7 +136,6 @@ static void put_huge_pages(struct drm_i915_gem_object *obj, huge_pages_free_pages(pages); obj->mm.dirty = false; - obj->mm.madv = I915_MADV_WILLNEED; } static const struct drm_i915_gem_object_ops huge_page_ops = { @@ -150,6 +150,7 @@ huge_pages_object(struct drm_i915_private *i915, u64 size, unsigned int page_mask) { + static struct lock_class_key lock_class; struct drm_i915_gem_object *obj; GEM_BUG_ON(!size); @@ -166,7 +167,9 @@ huge_pages_object(struct drm_i915_private *i915, return ERR_PTR(-ENOMEM); drm_gem_private_object_init(&i915->drm, &obj->base, size); - i915_gem_object_init(obj, &huge_page_ops); + i915_gem_object_init(obj, &huge_page_ops, &lock_class); + + i915_gem_object_set_volatile(obj); obj->write_domain = I915_GEM_DOMAIN_CPU; obj->read_domains = I915_GEM_DOMAIN_CPU; @@ -227,8 +230,6 @@ static int fake_get_huge_pages(struct drm_i915_gem_object *obj) i915_sg_trim(st); - obj->mm.madv = I915_MADV_DONTNEED; - __i915_gem_object_set_pages(obj, st, sg_page_sizes); return 0; @@ -261,8 +262,6 @@ static int fake_get_huge_pages_single(struct drm_i915_gem_object *obj) sg_dma_len(sg) = obj->base.size; sg_dma_address(sg) = page_size; - obj->mm.madv = I915_MADV_DONTNEED; - __i915_gem_object_set_pages(obj, st, sg->length); return 0; @@ -281,7 +280,6 @@ static void fake_put_huge_pages(struct drm_i915_gem_object *obj, { fake_free_huge_pages(obj, pages); obj->mm.dirty = false; - obj->mm.madv = I915_MADV_WILLNEED; } static const struct drm_i915_gem_object_ops fake_ops = { @@ -299,6 +297,7 @@ static const struct drm_i915_gem_object_ops fake_ops_single = { static struct drm_i915_gem_object * fake_huge_pages_object(struct drm_i915_private *i915, u64 size, bool single) { + static struct lock_class_key lock_class; struct drm_i915_gem_object *obj; GEM_BUG_ON(!size); @@ -317,9 +316,11 @@ fake_huge_pages_object(struct drm_i915_private *i915, u64 size, bool single) drm_gem_private_object_init(&i915->drm, &obj->base, size); if (single) - i915_gem_object_init(obj, &fake_ops_single); + i915_gem_object_init(obj, &fake_ops_single, &lock_class); else - i915_gem_object_init(obj, &fake_ops); + i915_gem_object_init(obj, &fake_ops, &lock_class); + + i915_gem_object_set_volatile(obj); obj->write_domain = I915_GEM_DOMAIN_CPU; obj->read_domains = I915_GEM_DOMAIN_CPU; @@ -333,7 +334,12 @@ static int igt_check_page_sizes(struct i915_vma *vma) struct drm_i915_private *i915 = vma->vm->i915; unsigned int supported = INTEL_INFO(i915)->page_sizes; struct drm_i915_gem_object *obj = vma->obj; - int err = 0; + int err; + + /* We have to wait for the async bind to complete before our asserts */ + err = i915_vma_sync(vma); + if (err) + return err; if (!HAS_PAGE_SIZES(i915, vma->page_sizes.sg)) { pr_err("unsupported page_sizes.sg=%u, supported=%u\n", @@ -447,6 +453,88 @@ out_device: return err; } +static int igt_mock_memory_region_huge_pages(void *arg) +{ + const unsigned int flags[] = { 0, I915_BO_ALLOC_CONTIGUOUS }; + struct i915_ppgtt *ppgtt = arg; + struct drm_i915_private *i915 = ppgtt->vm.i915; + unsigned long supported = INTEL_INFO(i915)->page_sizes; + struct intel_memory_region *mem; + struct drm_i915_gem_object *obj; + struct i915_vma *vma; + int bit; + int err = 0; + + mem = mock_region_create(i915, 0, SZ_2G, I915_GTT_PAGE_SIZE_4K, 0); + if (IS_ERR(mem)) { + pr_err("%s failed to create memory region\n", __func__); + return PTR_ERR(mem); + } + + for_each_set_bit(bit, &supported, ilog2(I915_GTT_MAX_PAGE_SIZE) + 1) { + unsigned int page_size = BIT(bit); + resource_size_t phys; + int i; + + for (i = 0; i < ARRAY_SIZE(flags); ++i) { + obj = i915_gem_object_create_region(mem, page_size, + flags[i]); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + goto out_region; + } + + vma = i915_vma_instance(obj, &ppgtt->vm, NULL); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto out_put; + } + + err = i915_vma_pin(vma, 0, 0, PIN_USER); + if (err) + goto out_close; + + err = igt_check_page_sizes(vma); + if (err) + goto out_unpin; + + phys = i915_gem_object_get_dma_address(obj, 0); + if (!IS_ALIGNED(phys, page_size)) { + pr_err("%s addr misaligned(%pa) page_size=%u\n", + __func__, &phys, page_size); + err = -EINVAL; + goto out_unpin; + } + + if (vma->page_sizes.gtt != page_size) { + pr_err("%s page_sizes.gtt=%u, expected=%u\n", + __func__, vma->page_sizes.gtt, + page_size); + err = -EINVAL; + goto out_unpin; + } + + i915_vma_unpin(vma); + i915_vma_close(vma); + + __i915_gem_object_put_pages(obj, I915_MM_NORMAL); + i915_gem_object_put(obj); + } + } + + goto out_region; + +out_unpin: + i915_vma_unpin(vma); +out_close: + i915_vma_close(vma); +out_put: + i915_gem_object_put(obj); +out_region: + intel_memory_region_put(mem); + return err; +} + static int igt_mock_ppgtt_misaligned_dma(void *arg) { struct i915_ppgtt *ppgtt = arg; @@ -879,9 +967,8 @@ out_object_put: return err; } -static int gpu_write(struct i915_vma *vma, - struct i915_gem_context *ctx, - struct intel_engine_cs *engine, +static int gpu_write(struct intel_context *ce, + struct i915_vma *vma, u32 dw, u32 val) { @@ -893,11 +980,12 @@ static int gpu_write(struct i915_vma *vma, if (err) return err; - return igt_gpu_fill_dw(vma, ctx, engine, dw * sizeof(u32), + return igt_gpu_fill_dw(ce, vma, dw * sizeof(u32), vma->size >> PAGE_SHIFT, val); } -static int cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val) +static int +__cpu_check_shmem(struct drm_i915_gem_object *obj, u32 dword, u32 val) { unsigned int needs_flush; unsigned long n; @@ -929,18 +1017,61 @@ static int cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val) return err; } -static int __igt_write_huge(struct i915_gem_context *ctx, - struct intel_engine_cs *engine, +static int __cpu_check_lmem(struct drm_i915_gem_object *obj, u32 dword, u32 val) +{ + unsigned long n; + int err; + + i915_gem_object_lock(obj); + err = i915_gem_object_set_to_wc_domain(obj, false); + i915_gem_object_unlock(obj); + if (err) + return err; + + err = i915_gem_object_pin_pages(obj); + if (err) + return err; + + for (n = 0; n < obj->base.size >> PAGE_SHIFT; ++n) { + u32 __iomem *base; + u32 read_val; + + base = i915_gem_object_lmem_io_map_page_atomic(obj, n); + + read_val = ioread32(base + dword); + io_mapping_unmap_atomic(base); + if (read_val != val) { + pr_err("n=%lu base[%u]=%u, val=%u\n", + n, dword, read_val, val); + err = -EINVAL; + break; + } + } + + i915_gem_object_unpin_pages(obj); + return err; +} + +static int cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val) +{ + if (i915_gem_object_has_struct_page(obj)) + return __cpu_check_shmem(obj, dword, val); + else if (i915_gem_object_is_lmem(obj)) + return __cpu_check_lmem(obj, dword, val); + + return -ENODEV; +} + +static int __igt_write_huge(struct intel_context *ce, struct drm_i915_gem_object *obj, u64 size, u64 offset, u32 dword, u32 val) { - struct i915_address_space *vm = ctx->vm ?: &engine->gt->ggtt->vm; unsigned int flags = PIN_USER | PIN_OFFSET_FIXED; struct i915_vma *vma; int err; - vma = i915_vma_instance(obj, vm, NULL); + vma = i915_vma_instance(obj, ce->vm, NULL); if (IS_ERR(vma)) return PTR_ERR(vma); @@ -954,7 +1085,7 @@ static int __igt_write_huge(struct i915_gem_context *ctx, * The ggtt may have some pages reserved so * refrain from erroring out. */ - if (err == -ENOSPC && i915_is_ggtt(vm)) + if (err == -ENOSPC && i915_is_ggtt(ce->vm)) err = 0; goto out_vma_close; @@ -964,7 +1095,7 @@ static int __igt_write_huge(struct i915_gem_context *ctx, if (err) goto out_vma_unpin; - err = gpu_write(vma, ctx, engine, dword, val); + err = gpu_write(ce, vma, dword, val); if (err) { pr_err("gpu-write failed at offset=%llx\n", offset); goto out_vma_unpin; @@ -987,14 +1118,13 @@ out_vma_close: static int igt_write_huge(struct i915_gem_context *ctx, struct drm_i915_gem_object *obj) { - struct drm_i915_private *i915 = to_i915(obj->base.dev); - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm; - static struct intel_engine_cs *engines[I915_NUM_ENGINES]; - struct intel_engine_cs *engine; + struct i915_gem_engines *engines; + struct i915_gem_engines_iter it; + struct intel_context *ce; I915_RND_STATE(prng); IGT_TIMEOUT(end_time); unsigned int max_page_size; - unsigned int id; + unsigned int count; u64 max; u64 num; u64 size; @@ -1008,19 +1138,18 @@ static int igt_write_huge(struct i915_gem_context *ctx, if (obj->mm.page_sizes.sg & I915_GTT_PAGE_SIZE_64K) size = round_up(size, I915_GTT_PAGE_SIZE_2M); - max_page_size = rounddown_pow_of_two(obj->mm.page_sizes.sg); - max = div_u64((vm->total - size), max_page_size); - n = 0; - for_each_engine(engine, i915, id) { - if (!intel_engine_can_store_dword(engine)) { - pr_info("store-dword-imm not supported on engine=%u\n", - id); + count = 0; + max = U64_MAX; + for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { + count++; + if (!intel_engine_can_store_dword(ce->engine)) continue; - } - engines[n++] = engine; - } + max = min(max, ce->vm->total); + n++; + } + i915_gem_context_unlock_engines(ctx); if (!n) return 0; @@ -1029,23 +1158,30 @@ static int igt_write_huge(struct i915_gem_context *ctx, * randomized order, lets also make feeding to the same engine a few * times in succession a possibility by enlarging the permutation array. */ - order = i915_random_order(n * I915_NUM_ENGINES, &prng); + order = i915_random_order(count * count, &prng); if (!order) return -ENOMEM; + max_page_size = rounddown_pow_of_two(obj->mm.page_sizes.sg); + max = div_u64(max - size, max_page_size); + /* * Try various offsets in an ascending/descending fashion until we * timeout -- we want to avoid issues hidden by effectively always using * offset = 0. */ i = 0; + engines = i915_gem_context_lock_engines(ctx); for_each_prime_number_from(num, 0, max) { u64 offset_low = num * max_page_size; u64 offset_high = (max - num) * max_page_size; u32 dword = offset_in_page(num) / 4; + struct intel_context *ce; - engine = engines[order[i] % n]; - i = (i + 1) % (n * I915_NUM_ENGINES); + ce = engines->engines[order[i] % engines->num_engines]; + i = (i + 1) % (count * count); + if (!ce || !intel_engine_can_store_dword(ce->engine)) + continue; /* * In order to utilize 64K pages we need to both pad the vma @@ -1057,22 +1193,23 @@ static int igt_write_huge(struct i915_gem_context *ctx, offset_low = round_down(offset_low, I915_GTT_PAGE_SIZE_2M); - err = __igt_write_huge(ctx, engine, obj, size, offset_low, + err = __igt_write_huge(ce, obj, size, offset_low, dword, num + 1); if (err) break; - err = __igt_write_huge(ctx, engine, obj, size, offset_high, + err = __igt_write_huge(ce, obj, size, offset_high, dword, num + 1); if (err) break; if (igt_timeout(end_time, - "%s timed out on engine=%u, offset_low=%llx offset_high=%llx, max_page_size=%x\n", - __func__, engine->id, offset_low, offset_high, + "%s timed out on %s, offset_low=%llx offset_high=%llx, max_page_size=%x\n", + __func__, ce->engine->name, offset_low, offset_high, max_page_size)) break; } + i915_gem_context_unlock_engines(ctx); kfree(order); @@ -1180,131 +1317,235 @@ out_device: return err; } -static int igt_ppgtt_internal_huge(void *arg) +typedef struct drm_i915_gem_object * +(*igt_create_fn)(struct drm_i915_private *i915, u32 size, u32 flags); + +static inline bool igt_can_allocate_thp(struct drm_i915_private *i915) +{ + return i915->mm.gemfs && has_transparent_hugepage(); +} + +static struct drm_i915_gem_object * +igt_create_shmem(struct drm_i915_private *i915, u32 size, u32 flags) +{ + if (!igt_can_allocate_thp(i915)) { + pr_info("%s missing THP support, skipping\n", __func__); + return ERR_PTR(-ENODEV); + } + + return i915_gem_object_create_shmem(i915, size); +} + +static struct drm_i915_gem_object * +igt_create_internal(struct drm_i915_private *i915, u32 size, u32 flags) +{ + return i915_gem_object_create_internal(i915, size); +} + +static struct drm_i915_gem_object * +igt_create_system(struct drm_i915_private *i915, u32 size, u32 flags) +{ + return huge_pages_object(i915, size, size); +} + +static struct drm_i915_gem_object * +igt_create_local(struct drm_i915_private *i915, u32 size, u32 flags) +{ + return i915_gem_object_create_lmem(i915, size, flags); +} + +static u32 igt_random_size(struct rnd_state *prng, + u32 min_page_size, + u32 max_page_size) +{ + u64 mask; + u32 size; + + GEM_BUG_ON(!is_power_of_2(min_page_size)); + GEM_BUG_ON(!is_power_of_2(max_page_size)); + GEM_BUG_ON(min_page_size < PAGE_SIZE); + GEM_BUG_ON(min_page_size > max_page_size); + + mask = ((max_page_size << 1ULL) - 1) & PAGE_MASK; + size = prandom_u32_state(prng) & mask; + if (size < min_page_size) + size |= min_page_size; + + return size; +} + +static int igt_ppgtt_smoke_huge(void *arg) { struct i915_gem_context *ctx = arg; struct drm_i915_private *i915 = ctx->i915; struct drm_i915_gem_object *obj; - static const unsigned int sizes[] = { - SZ_64K, - SZ_128K, - SZ_256K, - SZ_512K, - SZ_1M, - SZ_2M, + I915_RND_STATE(prng); + struct { + igt_create_fn fn; + u32 min; + u32 max; + } backends[] = { + { igt_create_internal, SZ_64K, SZ_2M, }, + { igt_create_shmem, SZ_64K, SZ_32M, }, + { igt_create_local, SZ_64K, SZ_1G, }, }; - int i; int err; + int i; /* - * Sanity check that the HW uses huge pages correctly through internal - * -- ensure that our writes land in the right place. + * Sanity check that the HW uses huge pages correctly through our + * various backends -- ensure that our writes land in the right place. */ - for (i = 0; i < ARRAY_SIZE(sizes); ++i) { - unsigned int size = sizes[i]; + for (i = 0; i < ARRAY_SIZE(backends); ++i) { + u32 min = backends[i].min; + u32 max = backends[i].max; + u32 size = max; +try_again: + size = igt_random_size(&prng, min, rounddown_pow_of_two(size)); - obj = i915_gem_object_create_internal(i915, size); - if (IS_ERR(obj)) - return PTR_ERR(obj); + obj = backends[i].fn(i915, size, 0); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + if (err == -E2BIG) { + size >>= 1; + goto try_again; + } else if (err == -ENODEV) { + err = 0; + continue; + } + + return err; + } err = i915_gem_object_pin_pages(obj); - if (err) + if (err) { + if (err == -ENXIO) { + i915_gem_object_put(obj); + size >>= 1; + goto try_again; + } goto out_put; + } - if (obj->mm.page_sizes.phys < I915_GTT_PAGE_SIZE_64K) { - pr_info("internal unable to allocate huge-page(s) with size=%u\n", - size); + if (obj->mm.page_sizes.phys < min) { + pr_info("%s unable to allocate huge-page(s) with size=%u, i=%d\n", + __func__, size, i); + err = -ENOMEM; goto out_unpin; } err = igt_write_huge(ctx, obj); if (err) { - pr_err("internal write-huge failed with size=%u\n", - size); - goto out_unpin; + pr_err("%s write-huge failed with size=%u, i=%d\n", + __func__, size, i); } - +out_unpin: i915_gem_object_unpin_pages(obj); __i915_gem_object_put_pages(obj, I915_MM_NORMAL); +out_put: i915_gem_object_put(obj); - } - return 0; + if (err == -ENOMEM || err == -ENXIO) + err = 0; -out_unpin: - i915_gem_object_unpin_pages(obj); -out_put: - i915_gem_object_put(obj); + if (err) + break; - return err; -} + cond_resched(); + } -static inline bool igt_can_allocate_thp(struct drm_i915_private *i915) -{ - return i915->mm.gemfs && has_transparent_hugepage(); + return err; } -static int igt_ppgtt_gemfs_huge(void *arg) +static int igt_ppgtt_sanity_check(void *arg) { struct i915_gem_context *ctx = arg; struct drm_i915_private *i915 = ctx->i915; - struct drm_i915_gem_object *obj; - static const unsigned int sizes[] = { - SZ_2M, - SZ_4M, - SZ_8M, - SZ_16M, - SZ_32M, + unsigned int supported = INTEL_INFO(i915)->page_sizes; + struct { + igt_create_fn fn; + unsigned int flags; + } backends[] = { + { igt_create_system, 0, }, + { igt_create_local, I915_BO_ALLOC_CONTIGUOUS, }, }; - int i; + struct { + u32 size; + u32 pages; + } combos[] = { + { SZ_64K, SZ_64K }, + { SZ_2M, SZ_2M }, + { SZ_2M, SZ_64K }, + { SZ_2M - SZ_64K, SZ_64K }, + { SZ_2M - SZ_4K, SZ_64K | SZ_4K }, + { SZ_2M + SZ_4K, SZ_64K | SZ_4K }, + { SZ_2M + SZ_4K, SZ_2M | SZ_4K }, + { SZ_2M + SZ_64K, SZ_2M | SZ_64K }, + }; + int i, j; int err; + if (supported == I915_GTT_PAGE_SIZE_4K) + return 0; + /* - * Sanity check that the HW uses huge pages correctly through gemfs -- - * ensure that our writes land in the right place. + * Sanity check that the HW behaves with a limited set of combinations. + * We already have a bunch of randomised testing, which should give us + * a decent amount of variation between runs, however we should keep + * this to limit the chances of introducing a temporary regression, by + * testing the most obvious cases that might make something blow up. */ - if (!igt_can_allocate_thp(i915)) { - pr_info("missing THP support, skipping\n"); - return 0; - } + for (i = 0; i < ARRAY_SIZE(backends); ++i) { + for (j = 0; j < ARRAY_SIZE(combos); ++j) { + struct drm_i915_gem_object *obj; + u32 size = combos[j].size; + u32 pages = combos[j].pages; - for (i = 0; i < ARRAY_SIZE(sizes); ++i) { - unsigned int size = sizes[i]; + obj = backends[i].fn(i915, size, backends[i].flags); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + if (err == -ENODEV) { + pr_info("Device lacks local memory, skipping\n"); + err = 0; + break; + } - obj = i915_gem_object_create_shmem(i915, size); - if (IS_ERR(obj)) - return PTR_ERR(obj); + return err; + } - err = i915_gem_object_pin_pages(obj); - if (err) - goto out_put; + err = i915_gem_object_pin_pages(obj); + if (err) { + i915_gem_object_put(obj); + goto out; + } - if (obj->mm.page_sizes.phys < I915_GTT_PAGE_SIZE_2M) { - pr_info("finishing test early, gemfs unable to allocate huge-page(s) with size=%u\n", - size); - goto out_unpin; - } + GEM_BUG_ON(pages > obj->base.size); + pages = pages & supported; - err = igt_write_huge(ctx, obj); - if (err) { - pr_err("gemfs write-huge failed with size=%u\n", - size); - goto out_unpin; + if (pages) + obj->mm.page_sizes.sg = pages; + + err = igt_write_huge(ctx, obj); + + i915_gem_object_unpin_pages(obj); + __i915_gem_object_put_pages(obj, I915_MM_NORMAL); + i915_gem_object_put(obj); + + if (err) { + pr_err("%s write-huge failed with size=%u pages=%u i=%d, j=%d\n", + __func__, size, pages, i, j); + goto out; + } } - i915_gem_object_unpin_pages(obj); - __i915_gem_object_put_pages(obj, I915_MM_NORMAL); - i915_gem_object_put(obj); + cond_resched(); } - return 0; - -out_unpin: - i915_gem_object_unpin_pages(obj); -out_put: - i915_gem_object_put(obj); +out: + if (err == -ENOMEM) + err = 0; return err; } @@ -1314,15 +1555,15 @@ static int igt_ppgtt_pin_update(void *arg) struct i915_gem_context *ctx = arg; struct drm_i915_private *dev_priv = ctx->i915; unsigned long supported = INTEL_INFO(dev_priv)->page_sizes; - struct i915_address_space *vm = ctx->vm; struct drm_i915_gem_object *obj; + struct i915_gem_engines_iter it; + struct i915_address_space *vm; + struct intel_context *ce; struct i915_vma *vma; unsigned int flags = PIN_USER | PIN_OFFSET_FIXED; - struct intel_engine_cs *engine; - enum intel_engine_id id; unsigned int n; int first, last; - int err; + int err = 0; /* * Make sure there's no funny business when doing a PIN_UPDATE -- in the @@ -1332,9 +1573,10 @@ static int igt_ppgtt_pin_update(void *arg) * huge-gtt-pages. */ - if (!vm || !i915_vm_is_4lvl(vm)) { + vm = i915_gem_context_get_vm_rcu(ctx); + if (!i915_vm_is_4lvl(vm)) { pr_info("48b PPGTT not supported, skipping\n"); - return 0; + goto out_vm; } first = ilog2(I915_GTT_PAGE_SIZE_64K); @@ -1387,7 +1629,7 @@ static int igt_ppgtt_pin_update(void *arg) goto out_unpin; } - err = i915_vma_bind(vma, I915_CACHE_NONE, PIN_UPDATE); + err = i915_vma_bind(vma, I915_CACHE_NONE, PIN_UPDATE, NULL); if (err) goto out_unpin; @@ -1419,14 +1661,18 @@ static int igt_ppgtt_pin_update(void *arg) */ n = 0; - for_each_engine(engine, dev_priv, id) { - if (!intel_engine_can_store_dword(engine)) + for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { + if (!intel_engine_can_store_dword(ce->engine)) continue; - err = gpu_write(vma, ctx, engine, n++, 0xdeadbeaf); + err = gpu_write(ce, vma, n++, 0xdeadbeaf); if (err) - goto out_unpin; + break; } + i915_gem_context_unlock_engines(ctx); + if (err) + goto out_unpin; + while (n--) { err = cpu_check(obj, n, 0xdeadbeaf); if (err) @@ -1439,6 +1685,8 @@ out_close: i915_vma_close(vma); out_put: i915_gem_object_put(obj); +out_vm: + i915_vm_put(vm); return err; } @@ -1448,7 +1696,7 @@ static int igt_tmpfs_fallback(void *arg) struct i915_gem_context *ctx = arg; struct drm_i915_private *i915 = ctx->i915; struct vfsmount *gemfs = i915->mm.gemfs; - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm; + struct i915_address_space *vm = i915_gem_context_get_vm_rcu(ctx); struct drm_i915_gem_object *obj; struct i915_vma *vma; u32 *vaddr; @@ -1498,6 +1746,7 @@ out_put: out_restore: i915->mm.gemfs = gemfs; + i915_vm_put(vm); return err; } @@ -1505,14 +1754,14 @@ static int igt_shrink_thp(void *arg) { struct i915_gem_context *ctx = arg; struct drm_i915_private *i915 = ctx->i915; - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm; + struct i915_address_space *vm = i915_gem_context_get_vm_rcu(ctx); struct drm_i915_gem_object *obj; - struct intel_engine_cs *engine; - enum intel_engine_id id; + struct i915_gem_engines_iter it; + struct intel_context *ce; struct i915_vma *vma; unsigned int flags = PIN_USER; unsigned int n; - int err; + int err = 0; /* * Sanity check shrinking huge-paged object -- make sure nothing blows @@ -1521,12 +1770,14 @@ static int igt_shrink_thp(void *arg) if (!igt_can_allocate_thp(i915)) { pr_info("missing THP support, skipping\n"); - return 0; + goto out_vm; } obj = i915_gem_object_create_shmem(i915, SZ_2M); - if (IS_ERR(obj)) - return PTR_ERR(obj); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + goto out_vm; + } vma = i915_vma_instance(obj, vm, NULL); if (IS_ERR(vma)) { @@ -1548,16 +1799,19 @@ static int igt_shrink_thp(void *arg) goto out_unpin; n = 0; - for_each_engine(engine, i915, id) { - if (!intel_engine_can_store_dword(engine)) + + for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { + if (!intel_engine_can_store_dword(ce->engine)) continue; - err = gpu_write(vma, ctx, engine, n++, 0xdeadbeaf); + err = gpu_write(ce, vma, n++, 0xdeadbeaf); if (err) - goto out_unpin; + break; } - + i915_gem_context_unlock_engines(ctx); i915_vma_unpin(vma); + if (err) + goto out_close; /* * Now that the pages are *unpinned* shrink-all should invoke @@ -1583,16 +1837,17 @@ static int igt_shrink_thp(void *arg) while (n--) { err = cpu_check(obj, n, 0xdeadbeaf); if (err) - goto out_unpin; + break; } - out_unpin: i915_vma_unpin(vma); out_close: i915_vma_close(vma); out_put: i915_gem_object_put(obj); +out_vm: + i915_vm_put(vm); return err; } @@ -1601,6 +1856,7 @@ int i915_gem_huge_page_mock_selftests(void) { static const struct i915_subtest tests[] = { SUBTEST(igt_mock_exhaust_device_supported_pages), + SUBTEST(igt_mock_memory_region_huge_pages), SUBTEST(igt_mock_ppgtt_misaligned_dma), SUBTEST(igt_mock_ppgtt_huge_fill), SUBTEST(igt_mock_ppgtt_64K), @@ -1617,7 +1873,6 @@ int i915_gem_huge_page_mock_selftests(void) mkwrite_device_info(dev_priv)->ppgtt_type = INTEL_PPGTT_FULL; mkwrite_device_info(dev_priv)->ppgtt_size = 48; - mutex_lock(&dev_priv->drm.struct_mutex); ppgtt = i915_ppgtt_create(dev_priv); if (IS_ERR(ppgtt)) { err = PTR_ERR(ppgtt); @@ -1643,9 +1898,7 @@ out_close: i915_vm_put(&ppgtt->vm); out_unlock: - mutex_unlock(&dev_priv->drm.struct_mutex); drm_dev_put(&dev_priv->drm); - return err; } @@ -1656,12 +1909,12 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_ppgtt_pin_update), SUBTEST(igt_tmpfs_fallback), SUBTEST(igt_ppgtt_exhaust_huge), - SUBTEST(igt_ppgtt_gemfs_huge), - SUBTEST(igt_ppgtt_internal_huge), + SUBTEST(igt_ppgtt_smoke_huge), + SUBTEST(igt_ppgtt_sanity_check), }; struct drm_file *file; struct i915_gem_context *ctx; - intel_wakeref_t wakeref; + struct i915_address_space *vm; int err; if (!HAS_PPGTT(i915)) { @@ -1676,25 +1929,21 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - ctx = live_context(i915, file); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); - goto out_unlock; + goto out_file; } - if (ctx->vm) - ctx->vm->scrub_64K = true; + mutex_lock(&ctx->mutex); + vm = i915_gem_context_vm(ctx); + if (vm) + WRITE_ONCE(vm->scrub_64K, true); + mutex_unlock(&ctx->mutex); err = i915_subtests(tests, ctx); -out_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - +out_file: mock_file_free(i915, file); - return err; } diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c index d8804a847945..da8edee4fe0a 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c @@ -5,6 +5,7 @@ #include "i915_selftest.h" +#include "gt/intel_engine_user.h" #include "gt/intel_gt.h" #include "selftests/igt_flush_test.h" @@ -12,10 +13,9 @@ #include "huge_gem_object.h" #include "mock_context.h" -static int igt_client_fill(void *arg) +static int __igt_client_fill(struct intel_engine_cs *engine) { - struct drm_i915_private *i915 = arg; - struct intel_context *ce = i915->engine[BCS0]->kernel_context; + struct intel_context *ce = engine->kernel_context; struct drm_i915_gem_object *obj; struct rnd_state prng; IGT_TIMEOUT(end); @@ -37,7 +37,7 @@ static int igt_client_fill(void *arg) pr_debug("%s with phys_sz= %x, sz=%x, val=%x\n", __func__, phys_sz, sz, val); - obj = huge_gem_object(i915, phys_sz, sz); + obj = huge_gem_object(engine->i915, phys_sz, sz); if (IS_ERR(obj)) { err = PTR_ERR(obj); goto err_flush; @@ -103,6 +103,28 @@ err_flush: return err; } +static int igt_client_fill(void *arg) +{ + int inst = 0; + + do { + struct intel_engine_cs *engine; + int err; + + engine = intel_engine_lookup_user(arg, + I915_ENGINE_CLASS_COPY, + inst++); + if (!engine) + return 0; + + err = __igt_client_fill(engine); + if (err == -ENOMEM) + err = 0; + if (err) + return err; + } while (1); +} + int i915_gem_client_blt_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c index 0ff7a89aadca..2b29f6b4e1dd 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c @@ -7,13 +7,18 @@ #include <linux/prime_numbers.h> #include "gt/intel_gt.h" +#include "gt/intel_gt_pm.h" +#include "gt/intel_ring.h" #include "i915_selftest.h" #include "selftests/i915_random.h" -static int cpu_set(struct drm_i915_gem_object *obj, - unsigned long offset, - u32 v) +struct context { + struct drm_i915_gem_object *obj; + struct intel_engine_cs *engine; +}; + +static int cpu_set(struct context *ctx, unsigned long offset, u32 v) { unsigned int needs_clflush; struct page *page; @@ -21,11 +26,11 @@ static int cpu_set(struct drm_i915_gem_object *obj, u32 *cpu; int err; - err = i915_gem_object_prepare_write(obj, &needs_clflush); + err = i915_gem_object_prepare_write(ctx->obj, &needs_clflush); if (err) return err; - page = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT); + page = i915_gem_object_get_page(ctx->obj, offset >> PAGE_SHIFT); map = kmap_atomic(page); cpu = map + offset_in_page(offset); @@ -38,14 +43,12 @@ static int cpu_set(struct drm_i915_gem_object *obj, drm_clflush_virt_range(cpu, sizeof(*cpu)); kunmap_atomic(map); - i915_gem_object_finish_access(obj); + i915_gem_object_finish_access(ctx->obj); return 0; } -static int cpu_get(struct drm_i915_gem_object *obj, - unsigned long offset, - u32 *v) +static int cpu_get(struct context *ctx, unsigned long offset, u32 *v) { unsigned int needs_clflush; struct page *page; @@ -53,11 +56,11 @@ static int cpu_get(struct drm_i915_gem_object *obj, u32 *cpu; int err; - err = i915_gem_object_prepare_read(obj, &needs_clflush); + err = i915_gem_object_prepare_read(ctx->obj, &needs_clflush); if (err) return err; - page = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT); + page = i915_gem_object_get_page(ctx->obj, offset >> PAGE_SHIFT); map = kmap_atomic(page); cpu = map + offset_in_page(offset); @@ -67,136 +70,137 @@ static int cpu_get(struct drm_i915_gem_object *obj, *v = *cpu; kunmap_atomic(map); - i915_gem_object_finish_access(obj); + i915_gem_object_finish_access(ctx->obj); return 0; } -static int gtt_set(struct drm_i915_gem_object *obj, - unsigned long offset, - u32 v) +static int gtt_set(struct context *ctx, unsigned long offset, u32 v) { struct i915_vma *vma; u32 __iomem *map; - int err; + int err = 0; - i915_gem_object_lock(obj); - err = i915_gem_object_set_to_gtt_domain(obj, true); - i915_gem_object_unlock(obj); + i915_gem_object_lock(ctx->obj); + err = i915_gem_object_set_to_gtt_domain(ctx->obj, true); + i915_gem_object_unlock(ctx->obj); if (err) return err; - vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE); + vma = i915_gem_object_ggtt_pin(ctx->obj, NULL, 0, 0, PIN_MAPPABLE); if (IS_ERR(vma)) return PTR_ERR(vma); + intel_gt_pm_get(vma->vm->gt); + map = i915_vma_pin_iomap(vma); i915_vma_unpin(vma); - if (IS_ERR(map)) - return PTR_ERR(map); + if (IS_ERR(map)) { + err = PTR_ERR(map); + goto out_rpm; + } iowrite32(v, &map[offset / sizeof(*map)]); i915_vma_unpin_iomap(vma); - return 0; +out_rpm: + intel_gt_pm_put(vma->vm->gt); + return err; } -static int gtt_get(struct drm_i915_gem_object *obj, - unsigned long offset, - u32 *v) +static int gtt_get(struct context *ctx, unsigned long offset, u32 *v) { struct i915_vma *vma; u32 __iomem *map; - int err; + int err = 0; - i915_gem_object_lock(obj); - err = i915_gem_object_set_to_gtt_domain(obj, false); - i915_gem_object_unlock(obj); + i915_gem_object_lock(ctx->obj); + err = i915_gem_object_set_to_gtt_domain(ctx->obj, false); + i915_gem_object_unlock(ctx->obj); if (err) return err; - vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE); + vma = i915_gem_object_ggtt_pin(ctx->obj, NULL, 0, 0, PIN_MAPPABLE); if (IS_ERR(vma)) return PTR_ERR(vma); + intel_gt_pm_get(vma->vm->gt); + map = i915_vma_pin_iomap(vma); i915_vma_unpin(vma); - if (IS_ERR(map)) - return PTR_ERR(map); + if (IS_ERR(map)) { + err = PTR_ERR(map); + goto out_rpm; + } *v = ioread32(&map[offset / sizeof(*map)]); i915_vma_unpin_iomap(vma); - return 0; +out_rpm: + intel_gt_pm_put(vma->vm->gt); + return err; } -static int wc_set(struct drm_i915_gem_object *obj, - unsigned long offset, - u32 v) +static int wc_set(struct context *ctx, unsigned long offset, u32 v) { u32 *map; int err; - i915_gem_object_lock(obj); - err = i915_gem_object_set_to_wc_domain(obj, true); - i915_gem_object_unlock(obj); + i915_gem_object_lock(ctx->obj); + err = i915_gem_object_set_to_wc_domain(ctx->obj, true); + i915_gem_object_unlock(ctx->obj); if (err) return err; - map = i915_gem_object_pin_map(obj, I915_MAP_WC); + map = i915_gem_object_pin_map(ctx->obj, I915_MAP_WC); if (IS_ERR(map)) return PTR_ERR(map); map[offset / sizeof(*map)] = v; - i915_gem_object_unpin_map(obj); + i915_gem_object_unpin_map(ctx->obj); return 0; } -static int wc_get(struct drm_i915_gem_object *obj, - unsigned long offset, - u32 *v) +static int wc_get(struct context *ctx, unsigned long offset, u32 *v) { u32 *map; int err; - i915_gem_object_lock(obj); - err = i915_gem_object_set_to_wc_domain(obj, false); - i915_gem_object_unlock(obj); + i915_gem_object_lock(ctx->obj); + err = i915_gem_object_set_to_wc_domain(ctx->obj, false); + i915_gem_object_unlock(ctx->obj); if (err) return err; - map = i915_gem_object_pin_map(obj, I915_MAP_WC); + map = i915_gem_object_pin_map(ctx->obj, I915_MAP_WC); if (IS_ERR(map)) return PTR_ERR(map); *v = map[offset / sizeof(*map)]; - i915_gem_object_unpin_map(obj); + i915_gem_object_unpin_map(ctx->obj); return 0; } -static int gpu_set(struct drm_i915_gem_object *obj, - unsigned long offset, - u32 v) +static int gpu_set(struct context *ctx, unsigned long offset, u32 v) { - struct drm_i915_private *i915 = to_i915(obj->base.dev); struct i915_request *rq; struct i915_vma *vma; u32 *cs; int err; - i915_gem_object_lock(obj); - err = i915_gem_object_set_to_gtt_domain(obj, true); - i915_gem_object_unlock(obj); + i915_gem_object_lock(ctx->obj); + err = i915_gem_object_set_to_gtt_domain(ctx->obj, true); + i915_gem_object_unlock(ctx->obj); if (err) return err; - vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, 0); + vma = i915_gem_object_ggtt_pin(ctx->obj, NULL, 0, 0, 0); if (IS_ERR(vma)) return PTR_ERR(vma); - rq = i915_request_create(i915->engine[RCS0]->kernel_context); + rq = i915_request_create(ctx->engine->kernel_context); if (IS_ERR(rq)) { i915_vma_unpin(vma); return PTR_ERR(rq); @@ -209,12 +213,12 @@ static int gpu_set(struct drm_i915_gem_object *obj, return PTR_ERR(cs); } - if (INTEL_GEN(i915) >= 8) { + if (INTEL_GEN(ctx->engine->i915) >= 8) { *cs++ = MI_STORE_DWORD_IMM_GEN4 | 1 << 22; *cs++ = lower_32_bits(i915_ggtt_offset(vma) + offset); *cs++ = upper_32_bits(i915_ggtt_offset(vma) + offset); *cs++ = v; - } else if (INTEL_GEN(i915) >= 4) { + } else if (INTEL_GEN(ctx->engine->i915) >= 4) { *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; *cs++ = 0; *cs++ = i915_ggtt_offset(vma) + offset; @@ -239,32 +243,34 @@ static int gpu_set(struct drm_i915_gem_object *obj, return err; } -static bool always_valid(struct drm_i915_private *i915) +static bool always_valid(struct context *ctx) { return true; } -static bool needs_fence_registers(struct drm_i915_private *i915) +static bool needs_fence_registers(struct context *ctx) { - return !intel_gt_is_wedged(&i915->gt); -} + struct intel_gt *gt = ctx->engine->gt; -static bool needs_mi_store_dword(struct drm_i915_private *i915) -{ - if (intel_gt_is_wedged(&i915->gt)) + if (intel_gt_is_wedged(gt)) return false; - if (!HAS_ENGINE(i915, RCS0)) + return gt->ggtt->num_fences; +} + +static bool needs_mi_store_dword(struct context *ctx) +{ + if (intel_gt_is_wedged(ctx->engine->gt)) return false; - return intel_engine_can_store_dword(i915->engine[RCS0]); + return intel_engine_can_store_dword(ctx->engine); } static const struct igt_coherency_mode { const char *name; - int (*set)(struct drm_i915_gem_object *, unsigned long offset, u32 v); - int (*get)(struct drm_i915_gem_object *, unsigned long offset, u32 *v); - bool (*valid)(struct drm_i915_private *i915); + int (*set)(struct context *ctx, unsigned long offset, u32 v); + int (*get)(struct context *ctx, unsigned long offset, u32 *v); + bool (*valid)(struct context *ctx); } igt_coherency_mode[] = { { "cpu", cpu_set, cpu_get, always_valid }, { "gtt", gtt_set, gtt_get, needs_fence_registers }, @@ -273,19 +279,37 @@ static const struct igt_coherency_mode { { }, }; +static struct intel_engine_cs * +random_engine(struct drm_i915_private *i915, struct rnd_state *prng) +{ + struct intel_engine_cs *engine; + unsigned int count; + + count = 0; + for_each_uabi_engine(engine, i915) + count++; + + count = i915_prandom_u32_max_state(count, prng); + for_each_uabi_engine(engine, i915) + if (count-- == 0) + return engine; + + return NULL; +} + static int igt_gem_coherency(void *arg) { const unsigned int ncachelines = PAGE_SIZE/64; - I915_RND_STATE(prng); struct drm_i915_private *i915 = arg; const struct igt_coherency_mode *read, *write, *over; - struct drm_i915_gem_object *obj; - intel_wakeref_t wakeref; unsigned long count, n; u32 *offsets, *values; + I915_RND_STATE(prng); + struct context ctx; int err = 0; - /* We repeatedly write, overwrite and read from a sequence of + /* + * We repeatedly write, overwrite and read from a sequence of * cachelines in order to try and detect incoherency (unflushed writes * from either the CPU or GPU). Each setter/getter uses our cache * domain API which should prevent incoherency. @@ -299,34 +323,36 @@ static int igt_gem_coherency(void *arg) values = offsets + ncachelines; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + ctx.engine = random_engine(i915, &prng); + GEM_BUG_ON(!ctx.engine); + pr_info("%s: using %s\n", __func__, ctx.engine->name); + for (over = igt_coherency_mode; over->name; over++) { if (!over->set) continue; - if (!over->valid(i915)) + if (!over->valid(&ctx)) continue; for (write = igt_coherency_mode; write->name; write++) { if (!write->set) continue; - if (!write->valid(i915)) + if (!write->valid(&ctx)) continue; for (read = igt_coherency_mode; read->name; read++) { if (!read->get) continue; - if (!read->valid(i915)) + if (!read->valid(&ctx)) continue; for_each_prime_number_from(count, 1, ncachelines) { - obj = i915_gem_object_create_internal(i915, PAGE_SIZE); - if (IS_ERR(obj)) { - err = PTR_ERR(obj); - goto unlock; + ctx.obj = i915_gem_object_create_internal(i915, PAGE_SIZE); + if (IS_ERR(ctx.obj)) { + err = PTR_ERR(ctx.obj); + goto free; } i915_random_reorder(offsets, ncachelines, &prng); @@ -334,7 +360,7 @@ static int igt_gem_coherency(void *arg) values[n] = prandom_u32_state(&prng); for (n = 0; n < count; n++) { - err = over->set(obj, offsets[n], ~values[n]); + err = over->set(&ctx, offsets[n], ~values[n]); if (err) { pr_err("Failed to set stale value[%ld/%ld] in object using %s, err=%d\n", n, count, over->name, err); @@ -343,7 +369,7 @@ static int igt_gem_coherency(void *arg) } for (n = 0; n < count; n++) { - err = write->set(obj, offsets[n], values[n]); + err = write->set(&ctx, offsets[n], values[n]); if (err) { pr_err("Failed to set value[%ld/%ld] in object using %s, err=%d\n", n, count, write->name, err); @@ -354,7 +380,7 @@ static int igt_gem_coherency(void *arg) for (n = 0; n < count; n++) { u32 found; - err = read->get(obj, offsets[n], &found); + err = read->get(&ctx, offsets[n], &found); if (err) { pr_err("Failed to get value[%ld/%ld] in object using %s, err=%d\n", n, count, read->name, err); @@ -372,20 +398,18 @@ static int igt_gem_coherency(void *arg) } } - i915_gem_object_put(obj); + i915_gem_object_put(ctx.obj); } } } } -unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); +free: kfree(offsets); return err; put_object: - i915_gem_object_put(obj); - goto unlock; + i915_gem_object_put(ctx.obj); + goto free; } int i915_gem_coherency_live_selftests(struct drm_i915_private *i915) diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c index 3e6f4a65d356..62fabc023a83 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c @@ -8,6 +8,7 @@ #include "gem/i915_gem_pm.h" #include "gt/intel_gt.h" +#include "gt/intel_gt_requests.h" #include "gt/intel_reset.h" #include "i915_selftest.h" @@ -31,7 +32,6 @@ static int live_nop_switch(void *arg) struct drm_i915_private *i915 = arg; struct intel_engine_cs *engine; struct i915_gem_context **ctx; - enum intel_engine_id id; struct igt_live_test t; struct drm_file *file; unsigned long n; @@ -52,23 +52,21 @@ static int live_nop_switch(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - ctx = kcalloc(nctx, sizeof(*ctx), GFP_KERNEL); if (!ctx) { err = -ENOMEM; - goto out_unlock; + goto out_file; } for (n = 0; n < nctx; n++) { ctx[n] = live_context(i915, file); if (IS_ERR(ctx[n])) { err = PTR_ERR(ctx[n]); - goto out_unlock; + goto out_file; } } - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { struct i915_request *rq; unsigned long end_time, prime; ktime_t times[2] = {}; @@ -78,7 +76,7 @@ static int live_nop_switch(void *arg) rq = igt_request_alloc(ctx[n], engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); - goto out_unlock; + goto out_file; } i915_request_add(rq); } @@ -86,7 +84,7 @@ static int live_nop_switch(void *arg) pr_err("Failed to populated %d contexts\n", nctx); intel_gt_set_wedged(&i915->gt); err = -EIO; - goto out_unlock; + goto out_file; } times[1] = ktime_get_raw(); @@ -96,7 +94,7 @@ static int live_nop_switch(void *arg) err = igt_live_test_begin(&t, i915, __func__, engine->name); if (err) - goto out_unlock; + goto out_file; end_time = jiffies + i915_selftest.timeout_jiffies; for_each_prime_number_from(prime, 2, 8192) { @@ -106,7 +104,7 @@ static int live_nop_switch(void *arg) rq = igt_request_alloc(ctx[n % nctx], engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); - goto out_unlock; + goto out_file; } /* @@ -142,7 +140,7 @@ static int live_nop_switch(void *arg) err = igt_live_test_end(&t); if (err) - goto out_unlock; + goto out_file; pr_info("Switch latencies on %s: 1 = %lluns, %lu = %lluns\n", engine->name, @@ -150,8 +148,235 @@ static int live_nop_switch(void *arg) prime - 1, div64_u64(ktime_to_ns(times[1]), prime - 1)); } -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); +out_file: + mock_file_free(i915, file); + return err; +} + +struct parallel_switch { + struct task_struct *tsk; + struct intel_context *ce[2]; +}; + +static int __live_parallel_switch1(void *data) +{ + struct parallel_switch *arg = data; + IGT_TIMEOUT(end_time); + unsigned long count; + + count = 0; + do { + struct i915_request *rq = NULL; + int err, n; + + err = 0; + for (n = 0; !err && n < ARRAY_SIZE(arg->ce); n++) { + struct i915_request *prev = rq; + + rq = i915_request_create(arg->ce[n]); + if (IS_ERR(rq)) { + i915_request_put(prev); + return PTR_ERR(rq); + } + + i915_request_get(rq); + if (prev) { + err = i915_request_await_dma_fence(rq, &prev->fence); + i915_request_put(prev); + } + + i915_request_add(rq); + } + if (i915_request_wait(rq, 0, HZ / 5) < 0) + err = -ETIME; + i915_request_put(rq); + if (err) + return err; + + count++; + } while (!__igt_timeout(end_time, NULL)); + + pr_info("%s: %lu switches (sync)\n", arg->ce[0]->engine->name, count); + return 0; +} + +static int __live_parallel_switchN(void *data) +{ + struct parallel_switch *arg = data; + struct i915_request *rq = NULL; + IGT_TIMEOUT(end_time); + unsigned long count; + int n; + + count = 0; + do { + for (n = 0; n < ARRAY_SIZE(arg->ce); n++) { + struct i915_request *prev = rq; + int err = 0; + + rq = i915_request_create(arg->ce[n]); + if (IS_ERR(rq)) { + i915_request_put(prev); + return PTR_ERR(rq); + } + + i915_request_get(rq); + if (prev) { + err = i915_request_await_dma_fence(rq, &prev->fence); + i915_request_put(prev); + } + + i915_request_add(rq); + if (err) { + i915_request_put(rq); + return err; + } + } + + count++; + } while (!__igt_timeout(end_time, NULL)); + i915_request_put(rq); + + pr_info("%s: %lu switches (many)\n", arg->ce[0]->engine->name, count); + return 0; +} + +static int live_parallel_switch(void *arg) +{ + struct drm_i915_private *i915 = arg; + static int (* const func[])(void *arg) = { + __live_parallel_switch1, + __live_parallel_switchN, + NULL, + }; + struct parallel_switch *data = NULL; + struct i915_gem_engines *engines; + struct i915_gem_engines_iter it; + int (* const *fn)(void *arg); + struct i915_gem_context *ctx; + struct intel_context *ce; + struct drm_file *file; + int n, m, count; + int err = 0; + + /* + * Check we can process switches on all engines simultaneously. + */ + + if (!DRIVER_CAPS(i915)->has_logical_contexts) + return 0; + + file = mock_file(i915); + if (IS_ERR(file)) + return PTR_ERR(file); + + ctx = live_context(i915, file); + if (IS_ERR(ctx)) { + err = PTR_ERR(ctx); + goto out_file; + } + + engines = i915_gem_context_lock_engines(ctx); + count = engines->num_engines; + + data = kcalloc(count, sizeof(*data), GFP_KERNEL); + if (!data) { + i915_gem_context_unlock_engines(ctx); + err = -ENOMEM; + goto out_file; + } + + m = 0; /* Use the first context as our template for the engines */ + for_each_gem_engine(ce, engines, it) { + err = intel_context_pin(ce); + if (err) { + i915_gem_context_unlock_engines(ctx); + goto out; + } + data[m++].ce[0] = intel_context_get(ce); + } + i915_gem_context_unlock_engines(ctx); + + /* Clone the same set of engines into the other contexts */ + for (n = 1; n < ARRAY_SIZE(data->ce); n++) { + ctx = live_context(i915, file); + if (IS_ERR(ctx)) { + err = PTR_ERR(ctx); + goto out; + } + + for (m = 0; m < count; m++) { + if (!data[m].ce[0]) + continue; + + ce = intel_context_create(ctx, data[m].ce[0]->engine); + if (IS_ERR(ce)) + goto out; + + err = intel_context_pin(ce); + if (err) { + intel_context_put(ce); + goto out; + } + + data[m].ce[n] = ce; + } + } + + for (fn = func; !err && *fn; fn++) { + struct igt_live_test t; + int n; + + err = igt_live_test_begin(&t, i915, __func__, ""); + if (err) + break; + + for (n = 0; n < count; n++) { + if (!data[n].ce[0]) + continue; + + data[n].tsk = kthread_run(*fn, &data[n], + "igt/parallel:%s", + data[n].ce[0]->engine->name); + if (IS_ERR(data[n].tsk)) { + err = PTR_ERR(data[n].tsk); + break; + } + get_task_struct(data[n].tsk); + } + + yield(); /* start all threads before we kthread_stop() */ + + for (n = 0; n < count; n++) { + int status; + + if (IS_ERR_OR_NULL(data[n].tsk)) + continue; + + status = kthread_stop(data[n].tsk); + if (status && !err) + err = status; + + put_task_struct(data[n].tsk); + data[n].tsk = NULL; + } + + if (igt_live_test_end(&t)) + err = -EIO; + } + +out: + for (n = 0; n < count; n++) { + for (m = 0; m < ARRAY_SIZE(data->ce); m++) { + if (!data[n].ce[m]) + continue; + + intel_context_unpin(data[n].ce[m]); + intel_context_put(data[n].ce[m]); + } + } + kfree(data); +out_file: mock_file_free(i915, file); return err; } @@ -166,28 +391,20 @@ static unsigned long fake_page_count(struct drm_i915_gem_object *obj) return huge_gem_object_dma_size(obj) >> PAGE_SHIFT; } -static int gpu_fill(struct drm_i915_gem_object *obj, - struct i915_gem_context *ctx, - struct intel_engine_cs *engine, +static int gpu_fill(struct intel_context *ce, + struct drm_i915_gem_object *obj, unsigned int dw) { - struct i915_address_space *vm = ctx->vm ?: &engine->gt->ggtt->vm; struct i915_vma *vma; int err; - GEM_BUG_ON(obj->base.size > vm->total); - GEM_BUG_ON(!intel_engine_can_store_dword(engine)); + GEM_BUG_ON(obj->base.size > ce->vm->total); + GEM_BUG_ON(!intel_engine_can_store_dword(ce->engine)); - vma = i915_vma_instance(obj, vm, NULL); + vma = i915_vma_instance(obj, ce->vm, NULL); if (IS_ERR(vma)) return PTR_ERR(vma); - i915_gem_object_lock(obj); - err = i915_gem_object_set_to_gtt_domain(obj, true); - i915_gem_object_unlock(obj); - if (err) - return err; - err = i915_vma_pin(vma, 0, 0, PIN_HIGH | PIN_USER); if (err) return err; @@ -200,9 +417,7 @@ static int gpu_fill(struct drm_i915_gem_object *obj, * whilst checking that each context provides a unique view * into the object. */ - err = igt_gpu_fill_dw(vma, - ctx, - engine, + err = igt_gpu_fill_dw(ce, vma, (dw * real_page_count(obj)) << PAGE_SHIFT | (dw * sizeof(u32)), real_page_count(obj), @@ -305,22 +520,21 @@ static int file_add_object(struct drm_file *file, } static struct drm_i915_gem_object * -create_test_object(struct i915_gem_context *ctx, +create_test_object(struct i915_address_space *vm, struct drm_file *file, struct list_head *objects) { struct drm_i915_gem_object *obj; - struct i915_address_space *vm = ctx->vm ?: &ctx->i915->ggtt.vm; u64 size; int err; /* Keep in GEM's good graces */ - i915_retire_requests(ctx->i915); + intel_gt_retire_requests(vm->gt); size = min(vm->total / 2, 1024ull * DW_PER_PAGE * PAGE_SIZE); size = round_down(size, DW_PER_PAGE * PAGE_SIZE); - obj = huge_gem_object(ctx->i915, DW_PER_PAGE * PAGE_SIZE, size); + obj = huge_gem_object(vm->i915, DW_PER_PAGE * PAGE_SIZE, size); if (IS_ERR(obj)) return obj; @@ -348,11 +562,49 @@ static unsigned long max_dwords(struct drm_i915_gem_object *obj) return npages / DW_PER_PAGE; } +static void throttle_release(struct i915_request **q, int count) +{ + int i; + + for (i = 0; i < count; i++) { + if (IS_ERR_OR_NULL(q[i])) + continue; + + i915_request_put(fetch_and_zero(&q[i])); + } +} + +static int throttle(struct intel_context *ce, + struct i915_request **q, int count) +{ + int i; + + if (!IS_ERR_OR_NULL(q[0])) { + if (i915_request_wait(q[0], + I915_WAIT_INTERRUPTIBLE, + MAX_SCHEDULE_TIMEOUT) < 0) + return -EINTR; + + i915_request_put(q[0]); + } + + for (i = 0; i < count - 1; i++) + q[i] = q[i + 1]; + + q[i] = intel_context_create_request(ce); + if (IS_ERR(q[i])) + return PTR_ERR(q[i]); + + i915_request_get(q[i]); + i915_request_add(q[i]); + + return 0; +} + static int igt_ctx_exec(void *arg) { struct drm_i915_private *i915 = arg; struct intel_engine_cs *engine; - enum intel_engine_id id; int err = -ENODEV; /* @@ -364,9 +616,10 @@ static int igt_ctx_exec(void *arg) if (!DRIVER_CAPS(i915)->has_logical_contexts) return 0; - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { struct drm_i915_gem_object *obj = NULL; unsigned long ncontexts, ndwords, dw; + struct i915_request *tq[5] = {}; struct igt_live_test t; struct drm_file *file; IGT_TIMEOUT(end_time); @@ -382,39 +635,53 @@ static int igt_ctx_exec(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - err = igt_live_test_begin(&t, i915, __func__, engine->name); if (err) - goto out_unlock; + goto out_file; ncontexts = 0; ndwords = 0; dw = 0; while (!time_after(jiffies, end_time)) { struct i915_gem_context *ctx; + struct intel_context *ce; - ctx = live_context(i915, file); + ctx = kernel_context(i915); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); - goto out_unlock; + goto out_file; } + ce = i915_gem_context_get_engine(ctx, engine->legacy_idx); + GEM_BUG_ON(IS_ERR(ce)); + if (!obj) { - obj = create_test_object(ctx, file, &objects); + obj = create_test_object(ce->vm, file, &objects); if (IS_ERR(obj)) { err = PTR_ERR(obj); - goto out_unlock; + intel_context_put(ce); + kernel_context_close(ctx); + goto out_file; } } - err = gpu_fill(obj, ctx, engine, dw); + err = gpu_fill(ce, obj, dw); if (err) { - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n", + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n", ndwords, dw, max_dwords(obj), - engine->name, ctx->hw_id, - yesno(!!ctx->vm), err); - goto out_unlock; + engine->name, + yesno(!!rcu_access_pointer(ctx->vm)), + err); + intel_context_put(ce); + kernel_context_close(ctx); + goto out_file; + } + + err = throttle(ce, tq, ARRAY_SIZE(tq)); + if (err) { + intel_context_put(ce); + kernel_context_close(ctx); + goto out_file; } if (++dw == max_dwords(obj)) { @@ -424,6 +691,9 @@ static int igt_ctx_exec(void *arg) ndwords++; ncontexts++; + + intel_context_put(ce); + kernel_context_close(ctx); } pr_info("Submitted %lu contexts to %s, filling %lu dwords\n", @@ -441,10 +711,10 @@ static int igt_ctx_exec(void *arg) dw += rem; } -out_unlock: +out_file: + throttle_release(tq, ARRAY_SIZE(tq)); if (igt_live_test_end(&t)) err = -EIO; - mutex_unlock(&i915->drm.struct_mutex); mock_file_free(i915, file); if (err) @@ -459,9 +729,9 @@ out_unlock: static int igt_shared_ctx_exec(void *arg) { struct drm_i915_private *i915 = arg; + struct i915_request *tq[5] = {}; struct i915_gem_context *parent; struct intel_engine_cs *engine; - enum intel_engine_id id; struct igt_live_test t; struct drm_file *file; int err = 0; @@ -478,24 +748,22 @@ static int igt_shared_ctx_exec(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - parent = live_context(i915, file); if (IS_ERR(parent)) { err = PTR_ERR(parent); - goto out_unlock; + goto out_file; } if (!parent->vm) { /* not full-ppgtt; nothing to share */ err = 0; - goto out_unlock; + goto out_file; } err = igt_live_test_begin(&t, i915, __func__, ""); if (err) - goto out_unlock; + goto out_file; - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { unsigned long ncontexts, ndwords, dw; struct drm_i915_gem_object *obj = NULL; IGT_TIMEOUT(end_time); @@ -509,6 +777,7 @@ static int igt_shared_ctx_exec(void *arg) ncontexts = 0; while (!time_after(jiffies, end_time)) { struct i915_gem_context *ctx; + struct intel_context *ce; ctx = kernel_context(i915); if (IS_ERR(ctx)) { @@ -516,23 +785,38 @@ static int igt_shared_ctx_exec(void *arg) goto out_test; } + mutex_lock(&ctx->mutex); __assign_ppgtt(ctx, parent->vm); + mutex_unlock(&ctx->mutex); + + ce = i915_gem_context_get_engine(ctx, engine->legacy_idx); + GEM_BUG_ON(IS_ERR(ce)); if (!obj) { - obj = create_test_object(parent, file, &objects); + obj = create_test_object(parent->vm, file, &objects); if (IS_ERR(obj)) { err = PTR_ERR(obj); + intel_context_put(ce); kernel_context_close(ctx); goto out_test; } } - err = gpu_fill(obj, ctx, engine, dw); + err = gpu_fill(ce, obj, dw); if (err) { - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n", + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n", ndwords, dw, max_dwords(obj), - engine->name, ctx->hw_id, - yesno(!!ctx->vm), err); + engine->name, + yesno(!!rcu_access_pointer(ctx->vm)), + err); + intel_context_put(ce); + kernel_context_close(ctx); + goto out_test; + } + + err = throttle(ce, tq, ARRAY_SIZE(tq)); + if (err) { + intel_context_put(ce); kernel_context_close(ctx); goto out_test; } @@ -545,6 +829,7 @@ static int igt_shared_ctx_exec(void *arg) ndwords++; ncontexts++; + intel_context_put(ce); kernel_context_close(ctx); } pr_info("Submitted %lu contexts to %s, filling %lu dwords\n", @@ -562,16 +847,13 @@ static int igt_shared_ctx_exec(void *arg) dw += rem; } - mutex_unlock(&i915->drm.struct_mutex); i915_gem_drain_freed_objects(i915); - mutex_lock(&i915->drm.struct_mutex); } out_test: + throttle_release(tq, ARRAY_SIZE(tq)); if (igt_live_test_end(&t)) err = -EIO; -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - +out_file: mock_file_free(i915, file); return err; } @@ -604,6 +886,8 @@ static struct i915_vma *rpcs_query_batch(struct i915_vma *vma) __i915_gem_object_flush_map(obj, 0, 64); i915_gem_object_unpin_map(obj); + intel_gt_chipset_flush(vma->vm->gt); + vma = i915_vma_instance(obj, vma->vm, NULL); if (IS_ERR(vma)) { err = PTR_ERR(vma); @@ -681,10 +965,7 @@ emit_rpcs_query(struct drm_i915_gem_object *obj, if (err) goto skip_request; - i915_vma_unpin(batch); - i915_vma_close(batch); - i915_vma_put(batch); - + i915_vma_unpin_and_release(&batch, 0); i915_vma_unpin(vma); *rq_out = i915_request_get(rq); @@ -698,8 +979,7 @@ skip_request: err_request: i915_request_add(rq); err_batch: - i915_vma_unpin(batch); - i915_vma_put(batch); + i915_vma_unpin_and_release(&batch, 0); err_vma: i915_vma_unpin(vma); @@ -860,8 +1140,8 @@ out: igt_spinner_end(spin); if ((flags & TEST_IDLE) && ret == 0) { - ret = i915_gem_wait_for_idle(ce->engine->i915, - 0, MAX_SCHEDULE_TIMEOUT); + ret = intel_gt_wait_for_idle(ce->engine->gt, + MAX_SCHEDULE_TIMEOUT); if (ret) return ret; @@ -887,7 +1167,7 @@ __sseu_test(const char *name, if (ret) return ret; - ret = __intel_context_reconfigure_sseu(ce, sseu); + ret = intel_context_reconfigure_sseu(ce, sseu); if (ret) goto out_spin; @@ -908,106 +1188,97 @@ __igt_ctx_sseu(struct drm_i915_private *i915, const char *name, unsigned int flags) { - struct intel_engine_cs *engine = i915->engine[RCS0]; struct drm_i915_gem_object *obj; - struct i915_gem_context *ctx; - struct intel_context *ce; - struct intel_sseu pg_sseu; - struct drm_file *file; - int ret; - - if (INTEL_GEN(i915) < 9 || !engine) - return 0; - - if (!RUNTIME_INFO(i915)->sseu.has_slice_pg) - return 0; + int inst = 0; + int ret = 0; - if (hweight32(engine->sseu.slice_mask) < 2) + if (INTEL_GEN(i915) < 9 || !RUNTIME_INFO(i915)->sseu.has_slice_pg) return 0; - /* - * Gen11 VME friendly power-gated configuration with half enabled - * sub-slices. - */ - pg_sseu = engine->sseu; - pg_sseu.slice_mask = 1; - pg_sseu.subslice_mask = - ~(~0 << (hweight32(engine->sseu.subslice_mask) / 2)); - - pr_info("SSEU subtest '%s', flags=%x, def_slices=%u, pg_slices=%u\n", - name, flags, hweight32(engine->sseu.slice_mask), - hweight32(pg_sseu.slice_mask)); - - file = mock_file(i915); - if (IS_ERR(file)) - return PTR_ERR(file); - if (flags & TEST_RESET) igt_global_reset_lock(&i915->gt); - mutex_lock(&i915->drm.struct_mutex); - - ctx = live_context(i915, file); - if (IS_ERR(ctx)) { - ret = PTR_ERR(ctx); - goto out_unlock; - } - i915_gem_context_clear_bannable(ctx); /* to reset and beyond! */ - obj = i915_gem_object_create_internal(i915, PAGE_SIZE); if (IS_ERR(obj)) { ret = PTR_ERR(obj); goto out_unlock; } - ce = i915_gem_context_get_engine(ctx, RCS0); - if (IS_ERR(ce)) { - ret = PTR_ERR(ce); - goto out_put; - } + do { + struct intel_engine_cs *engine; + struct intel_context *ce; + struct intel_sseu pg_sseu; - ret = intel_context_pin(ce); - if (ret) - goto out_context; + engine = intel_engine_lookup_user(i915, + I915_ENGINE_CLASS_RENDER, + inst++); + if (!engine) + break; - /* First set the default mask. */ - ret = __sseu_test(name, flags, ce, obj, engine->sseu); - if (ret) - goto out_fail; + if (hweight32(engine->sseu.slice_mask) < 2) + continue; - /* Then set a power-gated configuration. */ - ret = __sseu_test(name, flags, ce, obj, pg_sseu); - if (ret) - goto out_fail; + /* + * Gen11 VME friendly power-gated configuration with + * half enabled sub-slices. + */ + pg_sseu = engine->sseu; + pg_sseu.slice_mask = 1; + pg_sseu.subslice_mask = + ~(~0 << (hweight32(engine->sseu.subslice_mask) / 2)); + + pr_info("%s: SSEU subtest '%s', flags=%x, def_slices=%u, pg_slices=%u\n", + engine->name, name, flags, + hweight32(engine->sseu.slice_mask), + hweight32(pg_sseu.slice_mask)); + + ce = intel_context_create(engine->kernel_context->gem_context, + engine); + if (IS_ERR(ce)) { + ret = PTR_ERR(ce); + goto out_put; + } - /* Back to defaults. */ - ret = __sseu_test(name, flags, ce, obj, engine->sseu); - if (ret) - goto out_fail; + ret = intel_context_pin(ce); + if (ret) + goto out_ce; - /* One last power-gated configuration for the road. */ - ret = __sseu_test(name, flags, ce, obj, pg_sseu); - if (ret) - goto out_fail; + /* First set the default mask. */ + ret = __sseu_test(name, flags, ce, obj, engine->sseu); + if (ret) + goto out_unpin; + + /* Then set a power-gated configuration. */ + ret = __sseu_test(name, flags, ce, obj, pg_sseu); + if (ret) + goto out_unpin; + + /* Back to defaults. */ + ret = __sseu_test(name, flags, ce, obj, engine->sseu); + if (ret) + goto out_unpin; -out_fail: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + /* One last power-gated configuration for the road. */ + ret = __sseu_test(name, flags, ce, obj, pg_sseu); + if (ret) + goto out_unpin; + +out_unpin: + intel_context_unpin(ce); +out_ce: + intel_context_put(ce); + } while (!ret); + + if (igt_flush_test(i915)) ret = -EIO; - intel_context_unpin(ce); -out_context: - intel_context_put(ce); out_put: i915_gem_object_put(obj); out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - if (flags & TEST_RESET) igt_global_reset_unlock(&i915->gt); - mock_file_free(i915, file); - if (ret) pr_err("%s: Failed with %d!\n", name, ret); @@ -1041,6 +1312,7 @@ static int igt_ctx_readonly(void *arg) { struct drm_i915_private *i915 = arg; struct drm_i915_gem_object *obj = NULL; + struct i915_request *tq[5] = {}; struct i915_address_space *vm; struct i915_gem_context *ctx; unsigned long idx, ndwords, dw; @@ -1061,52 +1333,63 @@ static int igt_ctx_readonly(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - err = igt_live_test_begin(&t, i915, __func__, ""); if (err) - goto out_unlock; + goto out_file; ctx = live_context(i915, file); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); - goto out_unlock; + goto out_file; } - vm = ctx->vm ?: &i915->ggtt.alias->vm; + rcu_read_lock(); + vm = rcu_dereference(ctx->vm) ?: &i915->ggtt.alias->vm; if (!vm || !vm->has_read_only) { + rcu_read_unlock(); err = 0; - goto out_unlock; + goto out_file; } + rcu_read_unlock(); ndwords = 0; dw = 0; while (!time_after(jiffies, end_time)) { - struct intel_engine_cs *engine; - unsigned int id; + struct i915_gem_engines_iter it; + struct intel_context *ce; - for_each_engine(engine, i915, id) { - if (!intel_engine_can_store_dword(engine)) + for_each_gem_engine(ce, + i915_gem_context_lock_engines(ctx), it) { + if (!intel_engine_can_store_dword(ce->engine)) continue; if (!obj) { - obj = create_test_object(ctx, file, &objects); + obj = create_test_object(ce->vm, file, &objects); if (IS_ERR(obj)) { err = PTR_ERR(obj); - goto out_unlock; + i915_gem_context_unlock_engines(ctx); + goto out_file; } if (prandom_u32_state(&prng) & 1) i915_gem_object_set_readonly(obj); } - err = gpu_fill(obj, ctx, engine, dw); + err = gpu_fill(ce, obj, dw); if (err) { - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n", + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n", ndwords, dw, max_dwords(obj), - engine->name, ctx->hw_id, - yesno(!!ctx->vm), err); - goto out_unlock; + ce->engine->name, + yesno(!!rcu_access_pointer(ctx->vm)), + err); + i915_gem_context_unlock_engines(ctx); + goto out_file; + } + + err = throttle(ce, tq, ARRAY_SIZE(tq)); + if (err) { + i915_gem_context_unlock_engines(ctx); + goto out_file; } if (++dw == max_dwords(obj)) { @@ -1115,6 +1398,7 @@ static int igt_ctx_readonly(void *arg) } ndwords++; } + i915_gem_context_unlock_engines(ctx); } pr_info("Submitted %lu dwords (across %u engines)\n", ndwords, RUNTIME_INFO(i915)->num_engines); @@ -1137,19 +1421,19 @@ static int igt_ctx_readonly(void *arg) dw += rem; } -out_unlock: +out_file: + throttle_release(tq, ARRAY_SIZE(tq)); if (igt_live_test_end(&t)) err = -EIO; - mutex_unlock(&i915->drm.struct_mutex); mock_file_free(i915, file); return err; } -static int check_scratch(struct i915_gem_context *ctx, u64 offset) +static int check_scratch(struct i915_address_space *vm, u64 offset) { struct drm_mm_node *node = - __drm_mm_interval_first(&ctx->vm->mm, + __drm_mm_interval_first(&vm->mm, offset, offset + sizeof(u32) - 1); if (!node || node->start > offset) return 0; @@ -1167,6 +1451,7 @@ static int write_to_scratch(struct i915_gem_context *ctx, { struct drm_i915_private *i915 = ctx->i915; struct drm_i915_gem_object *obj; + struct i915_address_space *vm; struct i915_request *rq; struct i915_vma *vma; u32 *cmd; @@ -1197,17 +1482,20 @@ static int write_to_scratch(struct i915_gem_context *ctx, __i915_gem_object_flush_map(obj, 0, 64); i915_gem_object_unpin_map(obj); - vma = i915_vma_instance(obj, ctx->vm, NULL); + intel_gt_chipset_flush(engine->gt); + + vm = i915_gem_context_get_vm_rcu(ctx); + vma = i915_vma_instance(obj, vm, NULL); if (IS_ERR(vma)) { err = PTR_ERR(vma); - goto err; + goto err_vm; } err = i915_vma_pin(vma, 0, 0, PIN_USER | PIN_OFFSET_FIXED); if (err) - goto err; + goto err_vm; - err = check_scratch(ctx, offset); + err = check_scratch(vm, offset); if (err) goto err_unpin; @@ -1229,12 +1517,11 @@ static int write_to_scratch(struct i915_gem_context *ctx, if (err) goto skip_request; - i915_vma_unpin(vma); - i915_vma_close(vma); - i915_vma_put(vma); + i915_vma_unpin_and_release(&vma, 0); i915_request_add(rq); + i915_vm_put(vm); return 0; skip_request: @@ -1243,6 +1530,8 @@ err_request: i915_request_add(rq); err_unpin: i915_vma_unpin(vma); +err_vm: + i915_vm_put(vm); err: i915_gem_object_put(obj); return err; @@ -1254,6 +1543,7 @@ static int read_from_scratch(struct i915_gem_context *ctx, { struct drm_i915_private *i915 = ctx->i915; struct drm_i915_gem_object *obj; + struct i915_address_space *vm; const u32 RCS_GPR0 = 0x2600; /* not all engines have their own GPR! */ const u32 result = 0x100; struct i915_request *rq; @@ -1296,17 +1586,20 @@ static int read_from_scratch(struct i915_gem_context *ctx, i915_gem_object_flush_map(obj); i915_gem_object_unpin_map(obj); - vma = i915_vma_instance(obj, ctx->vm, NULL); + intel_gt_chipset_flush(engine->gt); + + vm = i915_gem_context_get_vm_rcu(ctx); + vma = i915_vma_instance(obj, vm, NULL); if (IS_ERR(vma)) { err = PTR_ERR(vma); - goto err; + goto err_vm; } err = i915_vma_pin(vma, 0, 0, PIN_USER | PIN_OFFSET_FIXED); if (err) - goto err; + goto err_vm; - err = check_scratch(ctx, offset); + err = check_scratch(vm, offset); if (err) goto err_unpin; @@ -1337,12 +1630,12 @@ static int read_from_scratch(struct i915_gem_context *ctx, err = i915_gem_object_set_to_cpu_domain(obj, false); i915_gem_object_unlock(obj); if (err) - goto err; + goto err_vm; cmd = i915_gem_object_pin_map(obj, I915_MAP_WB); if (IS_ERR(cmd)) { err = PTR_ERR(cmd); - goto err; + goto err_vm; } *value = cmd[result / sizeof(*cmd)]; @@ -1357,6 +1650,8 @@ err_request: i915_request_add(rq); err_unpin: i915_vma_unpin(vma); +err_vm: + i915_vm_put(vm); err: i915_gem_object_put(obj); return err; @@ -1371,7 +1666,6 @@ static int igt_vm_isolation(void *arg) struct drm_file *file; I915_RND_STATE(prng); unsigned long count; - unsigned int id; u64 vm_total; int err; @@ -1387,34 +1681,32 @@ static int igt_vm_isolation(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - err = igt_live_test_begin(&t, i915, __func__, ""); if (err) - goto out_unlock; + goto out_file; ctx_a = live_context(i915, file); if (IS_ERR(ctx_a)) { err = PTR_ERR(ctx_a); - goto out_unlock; + goto out_file; } ctx_b = live_context(i915, file); if (IS_ERR(ctx_b)) { err = PTR_ERR(ctx_b); - goto out_unlock; + goto out_file; } /* We can only test vm isolation, if the vm are distinct */ if (ctx_a->vm == ctx_b->vm) - goto out_unlock; + goto out_file; vm_total = ctx_a->vm->total; GEM_BUG_ON(ctx_b->vm->total != vm_total); vm_total -= I915_GTT_PAGE_SIZE; count = 0; - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { IGT_TIMEOUT(end_time); unsigned long this = 0; @@ -1436,7 +1728,7 @@ static int igt_vm_isolation(void *arg) err = read_from_scratch(ctx_b, engine, offset, &value); if (err) - goto out_unlock; + goto out_file; if (value) { pr_err("%s: Read %08x from scratch (offset 0x%08x_%08x), after %lu reads!\n", @@ -1445,7 +1737,7 @@ static int igt_vm_isolation(void *arg) lower_32_bits(offset), this); err = -EINVAL; - goto out_unlock; + goto out_file; } this++; @@ -1455,30 +1747,13 @@ static int igt_vm_isolation(void *arg) pr_info("Checked %lu scratch offsets across %d engines\n", count, RUNTIME_INFO(i915)->num_engines); -out_unlock: +out_file: if (igt_live_test_end(&t)) err = -EIO; - mutex_unlock(&i915->drm.struct_mutex); - mock_file_free(i915, file); return err; } -static __maybe_unused const char * -__engine_name(struct drm_i915_private *i915, intel_engine_mask_t engines) -{ - struct intel_engine_cs *engine; - intel_engine_mask_t tmp; - - if (engines == ALL_ENGINES) - return "all"; - - for_each_engine_masked(engine, i915, engines, tmp) - return engine->name; - - return "none"; -} - static bool skip_unused_engines(struct intel_context *ce, void *data) { return !ce->state; @@ -1506,13 +1781,9 @@ static int mock_context_barrier(void *arg) * a request; useful for retiring old state after loading new. */ - mutex_lock(&i915->drm.struct_mutex); - ctx = mock_context(i915, "mock"); - if (!ctx) { - err = -ENOMEM; - goto unlock; - } + if (!ctx) + return -ENOMEM; counter = 0; err = context_barrier_task(ctx, 0, @@ -1585,8 +1856,6 @@ static int mock_context_barrier(void *arg) out: mock_context_close(ctx); -unlock: - mutex_unlock(&i915->drm.struct_mutex); return err; #undef pr_fmt #define pr_fmt(x) x @@ -1614,6 +1883,7 @@ int i915_gem_context_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(live_nop_switch), + SUBTEST(live_parallel_switch), SUBTEST(igt_ctx_exec), SUBTEST(igt_ctx_readonly), SUBTEST(igt_ctx_sseu), diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c index 1d27babff0ce..29b2077b73d2 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c @@ -10,6 +10,7 @@ #include "gt/intel_gt_pm.h" #include "huge_gem_object.h" #include "i915_selftest.h" +#include "selftests/i915_random.h" #include "selftests/igt_flush_test.h" struct tile { @@ -76,18 +77,103 @@ static u64 tiled_offset(const struct tile *tile, u64 v) static int check_partial_mapping(struct drm_i915_gem_object *obj, const struct tile *tile, - unsigned long end_time) + struct rnd_state *prng) { - const unsigned int nreal = obj->scratch / PAGE_SIZE; const unsigned long npages = obj->base.size / PAGE_SIZE; + struct i915_ggtt_view view; struct i915_vma *vma; unsigned long page; + u32 __iomem *io; + struct page *p; + unsigned int n; + u64 offset; + u32 *cpu; int err; - if (igt_timeout(end_time, - "%s: timed out before tiling=%d stride=%d\n", - __func__, tile->tiling, tile->stride)) - return -EINTR; + err = i915_gem_object_set_tiling(obj, tile->tiling, tile->stride); + if (err) { + pr_err("Failed to set tiling mode=%u, stride=%u, err=%d\n", + tile->tiling, tile->stride, err); + return err; + } + + GEM_BUG_ON(i915_gem_object_get_tiling(obj) != tile->tiling); + GEM_BUG_ON(i915_gem_object_get_stride(obj) != tile->stride); + + i915_gem_object_lock(obj); + err = i915_gem_object_set_to_gtt_domain(obj, true); + i915_gem_object_unlock(obj); + if (err) { + pr_err("Failed to flush to GTT write domain; err=%d\n", err); + return err; + } + + page = i915_prandom_u32_max_state(npages, prng); + view = compute_partial_view(obj, page, MIN_CHUNK_PAGES); + + vma = i915_gem_object_ggtt_pin(obj, &view, 0, 0, PIN_MAPPABLE); + if (IS_ERR(vma)) { + pr_err("Failed to pin partial view: offset=%lu; err=%d\n", + page, (int)PTR_ERR(vma)); + return PTR_ERR(vma); + } + + n = page - view.partial.offset; + GEM_BUG_ON(n >= view.partial.size); + + io = i915_vma_pin_iomap(vma); + i915_vma_unpin(vma); + if (IS_ERR(io)) { + pr_err("Failed to iomap partial view: offset=%lu; err=%d\n", + page, (int)PTR_ERR(io)); + err = PTR_ERR(io); + goto out; + } + + iowrite32(page, io + n * PAGE_SIZE / sizeof(*io)); + i915_vma_unpin_iomap(vma); + + offset = tiled_offset(tile, page << PAGE_SHIFT); + if (offset >= obj->base.size) + goto out; + + intel_gt_flush_ggtt_writes(&to_i915(obj->base.dev)->gt); + + p = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT); + cpu = kmap(p) + offset_in_page(offset); + drm_clflush_virt_range(cpu, sizeof(*cpu)); + if (*cpu != (u32)page) { + pr_err("Partial view for %lu [%u] (offset=%llu, size=%u [%llu, row size %u], fence=%d, tiling=%d, stride=%d) misalignment, expected write to page (%llu + %u [0x%llx]) of 0x%x, found 0x%x\n", + page, n, + view.partial.offset, + view.partial.size, + vma->size >> PAGE_SHIFT, + tile->tiling ? tile_row_pages(obj) : 0, + vma->fence ? vma->fence->id : -1, tile->tiling, tile->stride, + offset >> PAGE_SHIFT, + (unsigned int)offset_in_page(offset), + offset, + (u32)page, *cpu); + err = -EINVAL; + } + *cpu = 0; + drm_clflush_virt_range(cpu, sizeof(*cpu)); + kunmap(p); + +out: + i915_vma_destroy(vma); + return err; +} + +static int check_partial_mappings(struct drm_i915_gem_object *obj, + const struct tile *tile, + unsigned long end_time) +{ + const unsigned int nreal = obj->scratch / PAGE_SIZE; + const unsigned long npages = obj->base.size / PAGE_SIZE; + struct i915_vma *vma; + unsigned long page; + int err; err = i915_gem_object_set_tiling(obj, tile->tiling, tile->stride); if (err) { @@ -170,11 +256,42 @@ static int check_partial_mapping(struct drm_i915_gem_object *obj, return err; i915_vma_destroy(vma); + + if (igt_timeout(end_time, + "%s: timed out after tiling=%d stride=%d\n", + __func__, tile->tiling, tile->stride)) + return -EINTR; } return 0; } +static unsigned int +setup_tile_size(struct tile *tile, struct drm_i915_private *i915) +{ + if (INTEL_GEN(i915) <= 2) { + tile->height = 16; + tile->width = 128; + tile->size = 11; + } else if (tile->tiling == I915_TILING_Y && + HAS_128_BYTE_Y_TILING(i915)) { + tile->height = 32; + tile->width = 128; + tile->size = 12; + } else { + tile->height = 8; + tile->width = 512; + tile->size = 12; + } + + if (INTEL_GEN(i915) < 4) + return 8192 / tile->width; + else if (INTEL_GEN(i915) < 7) + return 128 * I965_FENCE_MAX_PITCH_VAL / tile->width; + else + return 128 * GEN7_FENCE_MAX_PITCH_VAL / tile->width; +} + static int igt_partial_tiling(void *arg) { const unsigned int nreal = 1 << 12; /* largest tile row x2 */ @@ -184,6 +301,9 @@ static int igt_partial_tiling(void *arg) int tiling; int err; + if (!i915_ggtt_has_aperture(&i915->ggtt)) + return 0; + /* We want to check the page mapping and fencing of a large object * mmapped through the GTT. The object we create is larger than can * possibly be mmaped as a whole, and so we must use partial GGTT vma. @@ -205,7 +325,6 @@ static int igt_partial_tiling(void *arg) goto out; } - mutex_lock(&i915->drm.struct_mutex); wakeref = intel_runtime_pm_get(&i915->runtime_pm); if (1) { @@ -219,7 +338,7 @@ static int igt_partial_tiling(void *arg) tile.swizzle = I915_BIT_6_SWIZZLE_NONE; tile.tiling = I915_TILING_NONE; - err = check_partial_mapping(obj, &tile, end); + err = check_partial_mappings(obj, &tile, end); if (err && err != -EINTR) goto out_unlock; } @@ -241,10 +360,10 @@ static int igt_partial_tiling(void *arg) tile.tiling = tiling; switch (tiling) { case I915_TILING_X: - tile.swizzle = i915->mm.bit_6_swizzle_x; + tile.swizzle = i915->ggtt.bit_6_swizzle_x; break; case I915_TILING_Y: - tile.swizzle = i915->mm.bit_6_swizzle_y; + tile.swizzle = i915->ggtt.bit_6_swizzle_y; break; } @@ -253,31 +372,11 @@ static int igt_partial_tiling(void *arg) tile.swizzle == I915_BIT_6_SWIZZLE_9_10_17) continue; - if (INTEL_GEN(i915) <= 2) { - tile.height = 16; - tile.width = 128; - tile.size = 11; - } else if (tile.tiling == I915_TILING_Y && - HAS_128_BYTE_Y_TILING(i915)) { - tile.height = 32; - tile.width = 128; - tile.size = 12; - } else { - tile.height = 8; - tile.width = 512; - tile.size = 12; - } - - if (INTEL_GEN(i915) < 4) - max_pitch = 8192 / tile.width; - else if (INTEL_GEN(i915) < 7) - max_pitch = 128 * I965_FENCE_MAX_PITCH_VAL / tile.width; - else - max_pitch = 128 * GEN7_FENCE_MAX_PITCH_VAL / tile.width; + max_pitch = setup_tile_size(&tile, i915); for (pitch = max_pitch; pitch; pitch >>= 1) { tile.stride = tile.width * pitch; - err = check_partial_mapping(obj, &tile, end); + err = check_partial_mappings(obj, &tile, end); if (err == -EINTR) goto next_tiling; if (err) @@ -285,7 +384,7 @@ static int igt_partial_tiling(void *arg) if (pitch > 2 && INTEL_GEN(i915) >= 4) { tile.stride = tile.width * (pitch - 1); - err = check_partial_mapping(obj, &tile, end); + err = check_partial_mappings(obj, &tile, end); if (err == -EINTR) goto next_tiling; if (err) @@ -294,7 +393,7 @@ static int igt_partial_tiling(void *arg) if (pitch < max_pitch && INTEL_GEN(i915) >= 4) { tile.stride = tile.width * (pitch + 1); - err = check_partial_mapping(obj, &tile, end); + err = check_partial_mappings(obj, &tile, end); if (err == -EINTR) goto next_tiling; if (err) @@ -305,7 +404,7 @@ static int igt_partial_tiling(void *arg) if (INTEL_GEN(i915) >= 4) { for_each_prime_number(pitch, max_pitch) { tile.stride = tile.width * pitch; - err = check_partial_mapping(obj, &tile, end); + err = check_partial_mappings(obj, &tile, end); if (err == -EINTR) goto next_tiling; if (err) @@ -318,7 +417,100 @@ next_tiling: ; out_unlock: intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); + i915_gem_object_unpin_pages(obj); +out: + i915_gem_object_put(obj); + return err; +} + +static int igt_smoke_tiling(void *arg) +{ + const unsigned int nreal = 1 << 12; /* largest tile row x2 */ + struct drm_i915_private *i915 = arg; + struct drm_i915_gem_object *obj; + intel_wakeref_t wakeref; + I915_RND_STATE(prng); + unsigned long count; + IGT_TIMEOUT(end); + int err; + + if (!i915_ggtt_has_aperture(&i915->ggtt)) + return 0; + + /* + * igt_partial_tiling() does an exhastive check of partial tiling + * chunking, but will undoubtably run out of time. Here, we do a + * randomised search and hope over many runs of 1s with different + * seeds we will do a thorough check. + * + * Remember to look at the st_seed if we see a flip-flop in BAT! + */ + + if (i915->quirks & QUIRK_PIN_SWIZZLED_PAGES) + return 0; + + obj = huge_gem_object(i915, + nreal << PAGE_SHIFT, + (1 + next_prime_number(i915->ggtt.vm.total >> PAGE_SHIFT)) << PAGE_SHIFT); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + err = i915_gem_object_pin_pages(obj); + if (err) { + pr_err("Failed to allocate %u pages (%lu total), err=%d\n", + nreal, obj->base.size / PAGE_SIZE, err); + goto out; + } + + wakeref = intel_runtime_pm_get(&i915->runtime_pm); + + count = 0; + do { + struct tile tile; + + tile.tiling = + i915_prandom_u32_max_state(I915_TILING_Y + 1, &prng); + switch (tile.tiling) { + case I915_TILING_NONE: + tile.height = 1; + tile.width = 1; + tile.size = 0; + tile.stride = 0; + tile.swizzle = I915_BIT_6_SWIZZLE_NONE; + break; + + case I915_TILING_X: + tile.swizzle = i915->ggtt.bit_6_swizzle_x; + break; + case I915_TILING_Y: + tile.swizzle = i915->ggtt.bit_6_swizzle_y; + break; + } + + if (tile.swizzle == I915_BIT_6_SWIZZLE_9_17 || + tile.swizzle == I915_BIT_6_SWIZZLE_9_10_17) + continue; + + if (tile.tiling != I915_TILING_NONE) { + unsigned int max_pitch = setup_tile_size(&tile, i915); + + tile.stride = + i915_prandom_u32_max_state(max_pitch, &prng); + tile.stride = (1 + tile.stride) * tile.width; + if (INTEL_GEN(i915) < 4) + tile.stride = rounddown_pow_of_two(tile.stride); + } + + err = check_partial_mapping(obj, &tile, &prng); + if (err) + break; + + count++; + } while (!__igt_timeout(end, NULL)); + + pr_info("%s: Completed %lu trials\n", __func__, count); + + intel_runtime_pm_put(&i915->runtime_pm, wakeref); i915_gem_object_unpin_pages(obj); out: i915_gem_object_put(obj); @@ -329,20 +521,19 @@ static int make_obj_busy(struct drm_i915_gem_object *obj) { struct drm_i915_private *i915 = to_i915(obj->base.dev); struct intel_engine_cs *engine; - enum intel_engine_id id; - struct i915_vma *vma; - int err; - vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL); - if (IS_ERR(vma)) - return PTR_ERR(vma); + for_each_uabi_engine(engine, i915) { + struct i915_request *rq; + struct i915_vma *vma; + int err; - err = i915_vma_pin(vma, 0, 0, PIN_USER); - if (err) - return err; + vma = i915_vma_instance(obj, &engine->gt->ggtt->vm, NULL); + if (IS_ERR(vma)) + return PTR_ERR(vma); - for_each_engine(engine, i915, id) { - struct i915_request *rq; + err = i915_vma_pin(vma, 0, 0, PIN_USER); + if (err) + return err; rq = i915_request_create(engine->kernel_context); if (IS_ERR(rq)) { @@ -358,12 +549,13 @@ static int make_obj_busy(struct drm_i915_gem_object *obj) i915_vma_unlock(vma); i915_request_add(rq); + i915_vma_unpin(vma); + if (err) + return err; } - i915_vma_unpin(vma); i915_gem_object_put(obj); /* leave it only alive via its active ref */ - - return err; + return 0; } static bool assert_mmap_offset(struct drm_i915_private *i915, @@ -386,21 +578,14 @@ static bool assert_mmap_offset(struct drm_i915_private *i915, static void disable_retire_worker(struct drm_i915_private *i915) { i915_gem_driver_unregister__shrinker(i915); - intel_gt_pm_get(&i915->gt); - - cancel_delayed_work_sync(&i915->gem.retire_work); - flush_work(&i915->gem.idle_work); + cancel_delayed_work_sync(&i915->gt.requests.retire_work); } static void restore_retire_worker(struct drm_i915_private *i915) { + igt_flush_test(i915); intel_gt_pm_put(&i915->gt); - - mutex_lock(&i915->drm.struct_mutex); - igt_flush_test(i915, I915_WAIT_LOCKED); - mutex_unlock(&i915->drm.struct_mutex); - i915_gem_driver_register__shrinker(i915); } @@ -490,9 +675,7 @@ static int igt_mmap_offset_exhaustion(void *arg) goto out; } - mutex_lock(&i915->drm.struct_mutex); err = make_obj_busy(obj); - mutex_unlock(&i915->drm.struct_mutex); if (err) { pr_err("[loop %d] Failed to busy the object\n", loop); goto err_obj; @@ -515,6 +698,7 @@ int i915_gem_mman_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(igt_partial_tiling), + SUBTEST(igt_smoke_tiling), SUBTEST(igt_mmap_offset_exhaustion), }; diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c index c21d747e7d05..e8132aca0bb6 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c @@ -3,40 +3,241 @@ * Copyright © 2019 Intel Corporation */ +#include <linux/sort.h> + #include "gt/intel_gt.h" +#include "gt/intel_engine_user.h" #include "i915_selftest.h" +#include "gem/i915_gem_context.h" #include "selftests/igt_flush_test.h" +#include "selftests/i915_random.h" #include "selftests/mock_drm.h" #include "huge_gem_object.h" #include "mock_context.h" -static int igt_fill_blt(void *arg) +static int wrap_ktime_compare(const void *A, const void *B) +{ + const ktime_t *a = A, *b = B; + + return ktime_compare(*a, *b); +} + +static int __perf_fill_blt(struct drm_i915_gem_object *obj) +{ + struct drm_i915_private *i915 = to_i915(obj->base.dev); + int inst = 0; + + do { + struct intel_engine_cs *engine; + ktime_t t[5]; + int pass; + int err; + + engine = intel_engine_lookup_user(i915, + I915_ENGINE_CLASS_COPY, + inst++); + if (!engine) + return 0; + + for (pass = 0; pass < ARRAY_SIZE(t); pass++) { + struct intel_context *ce = engine->kernel_context; + ktime_t t0, t1; + + t0 = ktime_get(); + + err = i915_gem_object_fill_blt(obj, ce, 0); + if (err) + return err; + + err = i915_gem_object_wait(obj, + I915_WAIT_ALL, + MAX_SCHEDULE_TIMEOUT); + if (err) + return err; + + t1 = ktime_get(); + t[pass] = ktime_sub(t1, t0); + } + + sort(t, ARRAY_SIZE(t), sizeof(*t), wrap_ktime_compare, NULL); + pr_info("%s: blt %zd KiB fill: %lld MiB/s\n", + engine->name, + obj->base.size >> 10, + div64_u64(mul_u32_u32(4 * obj->base.size, + 1000 * 1000 * 1000), + t[1] + 2 * t[2] + t[3]) >> 20); + } while (1); +} + +static int perf_fill_blt(void *arg) { struct drm_i915_private *i915 = arg; - struct intel_context *ce = i915->engine[BCS0]->kernel_context; - struct drm_i915_gem_object *obj; + static const unsigned long sizes[] = { + SZ_4K, + SZ_64K, + SZ_2M, + SZ_64M + }; + int i; + + for (i = 0; i < ARRAY_SIZE(sizes); i++) { + struct drm_i915_gem_object *obj; + int err; + + obj = i915_gem_object_create_internal(i915, sizes[i]); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + err = __perf_fill_blt(obj); + i915_gem_object_put(obj); + if (err) + return err; + } + + return 0; +} + +static int __perf_copy_blt(struct drm_i915_gem_object *src, + struct drm_i915_gem_object *dst) +{ + struct drm_i915_private *i915 = to_i915(src->base.dev); + int inst = 0; + + do { + struct intel_engine_cs *engine; + ktime_t t[5]; + int pass; + + engine = intel_engine_lookup_user(i915, + I915_ENGINE_CLASS_COPY, + inst++); + if (!engine) + return 0; + + for (pass = 0; pass < ARRAY_SIZE(t); pass++) { + struct intel_context *ce = engine->kernel_context; + ktime_t t0, t1; + int err; + + t0 = ktime_get(); + + err = i915_gem_object_copy_blt(src, dst, ce); + if (err) + return err; + + err = i915_gem_object_wait(dst, + I915_WAIT_ALL, + MAX_SCHEDULE_TIMEOUT); + if (err) + return err; + + t1 = ktime_get(); + t[pass] = ktime_sub(t1, t0); + } + + sort(t, ARRAY_SIZE(t), sizeof(*t), wrap_ktime_compare, NULL); + pr_info("%s: blt %zd KiB copy: %lld MiB/s\n", + engine->name, + src->base.size >> 10, + div64_u64(mul_u32_u32(4 * src->base.size, + 1000 * 1000 * 1000), + t[1] + 2 * t[2] + t[3]) >> 20); + } while (1); +} + +static int perf_copy_blt(void *arg) +{ + struct drm_i915_private *i915 = arg; + static const unsigned long sizes[] = { + SZ_4K, + SZ_64K, + SZ_2M, + SZ_64M + }; + int i; + + for (i = 0; i < ARRAY_SIZE(sizes); i++) { + struct drm_i915_gem_object *src, *dst; + int err; + + src = i915_gem_object_create_internal(i915, sizes[i]); + if (IS_ERR(src)) + return PTR_ERR(src); + + dst = i915_gem_object_create_internal(i915, sizes[i]); + if (IS_ERR(dst)) { + err = PTR_ERR(dst); + goto err_src; + } + + err = __perf_copy_blt(src, dst); + + i915_gem_object_put(dst); +err_src: + i915_gem_object_put(src); + if (err) + return err; + } + + return 0; +} + +struct igt_thread_arg { + struct drm_i915_private *i915; struct rnd_state prng; + unsigned int n_cpus; +}; + +static int igt_fill_blt_thread(void *arg) +{ + struct igt_thread_arg *thread = arg; + struct drm_i915_private *i915 = thread->i915; + struct rnd_state *prng = &thread->prng; + struct drm_i915_gem_object *obj; + struct i915_gem_context *ctx; + struct intel_context *ce; + struct drm_file *file; + unsigned int prio; IGT_TIMEOUT(end); - u32 *vaddr; - int err = 0; + int err; + + file = mock_file(i915); + if (IS_ERR(file)) + return PTR_ERR(file); + + ctx = live_context(i915, file); + if (IS_ERR(ctx)) { + err = PTR_ERR(ctx); + goto out_file; + } - prandom_seed_state(&prng, i915_selftest.random_seed); + prio = i915_prandom_u32_max_state(I915_PRIORITY_MAX, prng); + ctx->sched.priority = I915_USER_PRIORITY(prio); - /* - * XXX: needs some threads to scale all these tests, also maybe throw - * in submission from higher priority context to see if we are - * preempted for very large objects... - */ + ce = i915_gem_context_get_engine(ctx, BCS0); + GEM_BUG_ON(IS_ERR(ce)); do { const u32 max_block_size = S16_MAX * PAGE_SIZE; - u32 sz = min_t(u64, ce->vm->total >> 4, prandom_u32_state(&prng)); - u32 phys_sz = sz % (max_block_size + 1); - u32 val = prandom_u32_state(&prng); + u32 val = prandom_u32_state(prng); + u64 total = ce->vm->total; + u32 phys_sz; + u32 sz; + u32 *vaddr; u32 i; + /* + * If we have a tiny shared address space, like for the GGTT + * then we can't be too greedy. + */ + if (i915_is_ggtt(ce->vm)) + total = div64_u64(total, thread->n_cpus); + + sz = min_t(u64, total >> 4, prandom_u32_state(prng)); + phys_sz = sz % (max_block_size + 1); + sz = round_up(sz, PAGE_SIZE); phys_sz = round_up(phys_sz, PAGE_SIZE); @@ -65,9 +266,7 @@ static int igt_fill_blt(void *arg) if (!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE)) obj->cache_dirty = true; - mutex_lock(&i915->drm.struct_mutex); err = i915_gem_object_fill_blt(obj, ce, val); - mutex_unlock(&i915->drm.struct_mutex); if (err) goto err_unpin; @@ -100,28 +299,56 @@ err_flush: if (err == -ENOMEM) err = 0; + intel_context_put(ce); +out_file: + mock_file_free(i915, file); return err; } -static int igt_copy_blt(void *arg) +static int igt_copy_blt_thread(void *arg) { - struct drm_i915_private *i915 = arg; - struct intel_context *ce = i915->engine[BCS0]->kernel_context; + struct igt_thread_arg *thread = arg; + struct drm_i915_private *i915 = thread->i915; + struct rnd_state *prng = &thread->prng; struct drm_i915_gem_object *src, *dst; - struct rnd_state prng; + struct i915_gem_context *ctx; + struct intel_context *ce; + struct drm_file *file; + unsigned int prio; IGT_TIMEOUT(end); - u32 *vaddr; - int err = 0; + int err; + + file = mock_file(i915); + if (IS_ERR(file)) + return PTR_ERR(file); - prandom_seed_state(&prng, i915_selftest.random_seed); + ctx = live_context(i915, file); + if (IS_ERR(ctx)) { + err = PTR_ERR(ctx); + goto out_file; + } + + prio = i915_prandom_u32_max_state(I915_PRIORITY_MAX, prng); + ctx->sched.priority = I915_USER_PRIORITY(prio); + + ce = i915_gem_context_get_engine(ctx, BCS0); + GEM_BUG_ON(IS_ERR(ce)); do { const u32 max_block_size = S16_MAX * PAGE_SIZE; - u32 sz = min_t(u64, ce->vm->total >> 4, prandom_u32_state(&prng)); - u32 phys_sz = sz % (max_block_size + 1); - u32 val = prandom_u32_state(&prng); + u32 val = prandom_u32_state(prng); + u64 total = ce->vm->total; + u32 phys_sz; + u32 sz; + u32 *vaddr; u32 i; + if (i915_is_ggtt(ce->vm)) + total = div64_u64(total, thread->n_cpus); + + sz = min_t(u64, total >> 4, prandom_u32_state(prng)); + phys_sz = sz % (max_block_size + 1); + sz = round_up(sz, PAGE_SIZE); phys_sz = round_up(phys_sz, PAGE_SIZE); @@ -166,9 +393,7 @@ static int igt_copy_blt(void *arg) if (!(dst->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE)) dst->cache_dirty = true; - mutex_lock(&i915->drm.struct_mutex); err = i915_gem_object_copy_blt(src, dst, ce); - mutex_unlock(&i915->drm.struct_mutex); if (err) goto err_unpin; @@ -205,12 +430,85 @@ err_flush: if (err == -ENOMEM) err = 0; + intel_context_put(ce); +out_file: + mock_file_free(i915, file); + return err; +} + +static int igt_threaded_blt(struct drm_i915_private *i915, + int (*blt_fn)(void *arg)) +{ + struct igt_thread_arg *thread; + struct task_struct **tsk; + I915_RND_STATE(prng); + unsigned int n_cpus; + unsigned int i; + int err = 0; + + n_cpus = num_online_cpus() + 1; + + tsk = kcalloc(n_cpus, sizeof(struct task_struct *), GFP_KERNEL); + if (!tsk) + return 0; + + thread = kcalloc(n_cpus, sizeof(struct igt_thread_arg), GFP_KERNEL); + if (!thread) { + kfree(tsk); + return 0; + } + + for (i = 0; i < n_cpus; ++i) { + thread[i].i915 = i915; + thread[i].n_cpus = n_cpus; + thread[i].prng = + I915_RND_STATE_INITIALIZER(prandom_u32_state(&prng)); + + tsk[i] = kthread_run(blt_fn, &thread[i], "igt/blt-%d", i); + if (IS_ERR(tsk[i])) { + err = PTR_ERR(tsk[i]); + break; + } + + get_task_struct(tsk[i]); + } + + yield(); /* start all threads before we kthread_stop() */ + + for (i = 0; i < n_cpus; ++i) { + int status; + + if (IS_ERR_OR_NULL(tsk[i])) + continue; + + status = kthread_stop(tsk[i]); + if (status && !err) + err = status; + + put_task_struct(tsk[i]); + } + + kfree(tsk); + kfree(thread); + return err; } +static int igt_fill_blt(void *arg) +{ + return igt_threaded_blt(arg, igt_fill_blt_thread); +} + +static int igt_copy_blt(void *arg) +{ + return igt_threaded_blt(arg, igt_copy_blt_thread); +} + int i915_gem_object_blt_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { + SUBTEST(perf_fill_blt), + SUBTEST(perf_copy_blt), SUBTEST(igt_fill_blt), SUBTEST(igt_copy_blt), }; diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c index 94a15e3f6db8..34932871b3a5 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c @@ -25,9 +25,7 @@ static int mock_phys_object(void *arg) goto out; } - mutex_lock(&i915->drm.struct_mutex); err = i915_gem_object_attach_phys(obj, PAGE_SIZE); - mutex_unlock(&i915->drm.struct_mutex); if (err) { pr_err("i915_gem_object_attach_phys failed, err=%d\n", err); goto out_obj; diff --git a/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c index 57ece53c1075..6718da20f35d 100644 --- a/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c +++ b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c @@ -9,6 +9,7 @@ #include "gem/i915_gem_context.h" #include "gem/i915_gem_pm.h" #include "gt/intel_context.h" +#include "gt/intel_gt.h" #include "i915_vma.h" #include "i915_drv.h" @@ -84,6 +85,8 @@ igt_emit_store_dw(struct i915_vma *vma, *cmd = MI_BATCH_BUFFER_END; i915_gem_object_unpin_map(obj); + intel_gt_chipset_flush(vma->vm->gt); + vma = i915_vma_instance(obj, vma->vm, NULL); if (IS_ERR(vma)) { err = PTR_ERR(vma); @@ -101,40 +104,35 @@ err: return ERR_PTR(err); } -int igt_gpu_fill_dw(struct i915_vma *vma, - struct i915_gem_context *ctx, - struct intel_engine_cs *engine, - u64 offset, - unsigned long count, - u32 val) +int igt_gpu_fill_dw(struct intel_context *ce, + struct i915_vma *vma, u64 offset, + unsigned long count, u32 val) { - struct i915_address_space *vm = ctx->vm ?: &engine->gt->ggtt->vm; struct i915_request *rq; struct i915_vma *batch; unsigned int flags; int err; - GEM_BUG_ON(vma->size > vm->total); - GEM_BUG_ON(!intel_engine_can_store_dword(engine)); + GEM_BUG_ON(!intel_engine_can_store_dword(ce->engine)); GEM_BUG_ON(!i915_vma_is_pinned(vma)); batch = igt_emit_store_dw(vma, offset, count, val); if (IS_ERR(batch)) return PTR_ERR(batch); - rq = igt_request_alloc(ctx, engine); + rq = intel_context_create_request(ce); if (IS_ERR(rq)) { err = PTR_ERR(rq); goto err_batch; } flags = 0; - if (INTEL_GEN(vm->i915) <= 5) + if (INTEL_GEN(ce->vm->i915) <= 5) flags |= I915_DISPATCH_SECURE; - err = engine->emit_bb_start(rq, - batch->node.start, batch->node.size, - flags); + err = rq->engine->emit_bb_start(rq, + batch->node.start, batch->node.size, + flags); if (err) goto err_request; @@ -156,9 +154,7 @@ int igt_gpu_fill_dw(struct i915_vma *vma, i915_request_add(rq); - i915_vma_unpin(batch); - i915_vma_close(batch); - i915_vma_put(batch); + i915_vma_unpin_and_release(&batch, 0); return 0; @@ -167,7 +163,6 @@ skip_request: err_request: i915_request_add(rq); err_batch: - i915_vma_unpin(batch); - i915_vma_put(batch); + i915_vma_unpin_and_release(&batch, 0); return err; } diff --git a/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.h b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.h index 361a7ef866b0..4221cf84d175 100644 --- a/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.h +++ b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.h @@ -11,9 +11,11 @@ struct i915_request; struct i915_gem_context; -struct intel_engine_cs; struct i915_vma; +struct intel_context; +struct intel_engine_cs; + struct i915_request * igt_request_alloc(struct i915_gem_context *ctx, struct intel_engine_cs *engine); @@ -23,11 +25,8 @@ igt_emit_store_dw(struct i915_vma *vma, unsigned long count, u32 val); -int igt_gpu_fill_dw(struct i915_vma *vma, - struct i915_gem_context *ctx, - struct intel_engine_cs *engine, - u64 offset, - unsigned long count, - u32 val); +int igt_gpu_fill_dw(struct intel_context *ce, + struct i915_vma *vma, u64 offset, + unsigned long count, u32 val); #endif /* __IGT_GEM_UTILS_H__ */ diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c index be8974ccff24..29b8984f0e47 100644 --- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c +++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c @@ -13,7 +13,6 @@ mock_context(struct drm_i915_private *i915, { struct i915_gem_context *ctx; struct i915_gem_engines *e; - int ret; ctx = kzalloc(sizeof(*ctx), GFP_KERNEL); if (!ctx) @@ -23,6 +22,8 @@ mock_context(struct drm_i915_private *i915, INIT_LIST_HEAD(&ctx->link); ctx->i915 = i915; + i915_gem_context_set_persistence(ctx); + mutex_init(&ctx->engines_mutex); e = default_engines(ctx); if (IS_ERR(e)) @@ -30,13 +31,8 @@ mock_context(struct drm_i915_private *i915, RCU_INIT_POINTER(ctx->engines, e); INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL); - INIT_LIST_HEAD(&ctx->hw_id_link); mutex_init(&ctx->mutex); - ret = i915_gem_context_pin_hw_id(ctx); - if (ret < 0) - goto err_engines; - if (name) { struct i915_ppgtt *ppgtt; @@ -48,14 +44,15 @@ mock_context(struct drm_i915_private *i915, if (!ppgtt) goto err_put; + mutex_lock(&ctx->mutex); __set_ppgtt(ctx, &ppgtt->vm); + mutex_unlock(&ctx->mutex); + i915_vm_put(&ppgtt->vm); } return ctx; -err_engines: - free_engines(rcu_access_pointer(ctx->engines)); err_free: kfree(ctx); return NULL; @@ -73,7 +70,7 @@ void mock_context_close(struct i915_gem_context *ctx) void mock_init_contexts(struct drm_i915_private *i915) { - init_contexts(i915); + init_contexts(&i915->gem.contexts); } struct i915_gem_context * @@ -82,8 +79,6 @@ live_context(struct drm_i915_private *i915, struct drm_file *file) struct i915_gem_context *ctx; int err; - lockdep_assert_held(&i915->drm.struct_mutex); - ctx = i915_gem_create_context(i915, 0); if (IS_ERR(ctx)) return ctx; diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c index 09c68dda2098..55317081d48b 100644 --- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c +++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c @@ -120,7 +120,6 @@ __dma_fence_signal__notify(struct dma_fence *fence, struct dma_fence_cb *cur, *tmp; lockdep_assert_held(fence->lock); - lockdep_assert_irqs_disabled(); list_for_each_entry_safe(cur, tmp, list, node) { INIT_LIST_HEAD(&cur->node); @@ -134,9 +133,10 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine) const ktime_t timestamp = ktime_get(); struct intel_context *ce, *cn; struct list_head *pos, *next; + unsigned long flags; LIST_HEAD(signal); - spin_lock(&b->irq_lock); + spin_lock_irqsave(&b->irq_lock, flags); if (b->irq_armed && list_empty(&b->signalers)) __intel_breadcrumbs_disarm_irq(b); @@ -182,30 +182,23 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine) } } - spin_unlock(&b->irq_lock); + spin_unlock_irqrestore(&b->irq_lock, flags); list_for_each_safe(pos, next, &signal) { struct i915_request *rq = list_entry(pos, typeof(*rq), signal_link); struct list_head cb_list; - spin_lock(&rq->lock); + spin_lock_irqsave(&rq->lock, flags); list_replace(&rq->fence.cb_list, &cb_list); __dma_fence_signal__timestamp(&rq->fence, timestamp); __dma_fence_signal__notify(&rq->fence, &cb_list); - spin_unlock(&rq->lock); + spin_unlock_irqrestore(&rq->lock, flags); i915_request_put(rq); } } -void intel_engine_signal_breadcrumbs(struct intel_engine_cs *engine) -{ - local_irq_disable(); - intel_engine_breadcrumbs_irq(engine); - local_irq_enable(); -} - static void signal_irq_work(struct irq_work *work) { struct intel_engine_cs *engine = @@ -275,7 +268,6 @@ void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine) bool i915_request_enable_breadcrumb(struct i915_request *rq) { lockdep_assert_held(&rq->lock); - lockdep_assert_irqs_disabled(); if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags)) { struct intel_breadcrumbs *b = &rq->engine->breadcrumbs; @@ -325,7 +317,6 @@ void i915_request_cancel_breadcrumb(struct i915_request *rq) struct intel_breadcrumbs *b = &rq->engine->breadcrumbs; lockdep_assert_held(&rq->lock); - lockdep_assert_irqs_disabled(); /* * We must wait for b->irq_lock so that we know the interrupt handler diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c index f55691d151ae..ef7bc41ffffa 100644 --- a/drivers/gpu/drm/i915/gt/intel_context.c +++ b/drivers/gpu/drm/i915/gt/intel_context.c @@ -13,6 +13,7 @@ #include "intel_context.h" #include "intel_engine.h" #include "intel_engine_pm.h" +#include "intel_ring.h" static struct i915_global_context { struct i915_global base; @@ -62,7 +63,7 @@ int __intel_context_do_pin(struct intel_context *ce) } err = 0; - with_intel_runtime_pm(&ce->engine->i915->runtime_pm, wakeref) + with_intel_runtime_pm(ce->engine->uncore->rpm, wakeref) err = ce->ops->pin(ce); if (err) goto err; @@ -134,10 +135,11 @@ static int __context_pin_state(struct i915_vma *vma) static void __context_unpin_state(struct i915_vma *vma) { - __i915_vma_unpin(vma); i915_vma_make_shrinkable(vma); + __i915_vma_unpin(vma); } +__i915_active_call static void __intel_context_retire(struct i915_active *active) { struct intel_context *ce = container_of(active, typeof(*ce), active); @@ -150,6 +152,7 @@ static void __intel_context_retire(struct i915_active *active) intel_timeline_unpin(ce->timeline); intel_ring_unpin(ce->ring); + intel_context_put(ce); } @@ -219,12 +222,20 @@ intel_context_init(struct intel_context *ce, struct i915_gem_context *ctx, struct intel_engine_cs *engine) { + struct i915_address_space *vm; + GEM_BUG_ON(!engine->cops); kref_init(&ce->ref); ce->gem_context = ctx; - ce->vm = i915_vm_get(ctx->vm ?: &engine->gt->ggtt->vm); + rcu_read_lock(); + vm = rcu_dereference(ctx->vm); + if (vm) + ce->vm = i915_vm_get(vm); + else + ce->vm = i915_vm_get(&engine->gt->ggtt->vm); + rcu_read_unlock(); if (ctx->timeline) ce->timeline = intel_timeline_get(ctx->timeline); @@ -238,7 +249,7 @@ intel_context_init(struct intel_context *ce, mutex_init(&ce->pin_mutex); - i915_active_init(ctx->i915, &ce->active, + i915_active_init(&ce->active, __intel_context_active, __intel_context_retire); } @@ -298,14 +309,27 @@ int intel_context_prepare_remote_request(struct intel_context *ce, /* Only suitable for use in remotely modifying this context */ GEM_BUG_ON(rq->hw_context == ce); - if (rq->timeline != tl) { /* beware timeline sharing */ - err = mutex_lock_interruptible_nested(&tl->mutex, - SINGLE_DEPTH_NESTING); - if (err) - return err; + if (rcu_access_pointer(rq->timeline) != tl) { /* timeline sharing! */ + /* + * Ideally, we just want to insert our foreign fence as + * a barrier into the remove context, such that this operation + * occurs after all current operations in that context, and + * all future operations must occur after this. + * + * Currently, the timeline->last_request tracking is guarded + * by its mutex and so we must obtain that to atomically + * insert our barrier. However, since we already hold our + * timeline->mutex, we must be careful against potential + * inversion if we are the kernel_context as the remote context + * will itself poke at the kernel_context when it needs to + * unpin. Ergo, if already locked, we drop both locks and + * try again (through the magic of userspace repeating EAGAIN). + */ + if (!mutex_trylock(&tl->mutex)) + return -EAGAIN; /* Queue this switch after current activity by this context. */ - err = i915_active_request_set(&tl->last_request, rq); + err = i915_active_fence_set(&tl->last_request, rq); mutex_unlock(&tl->mutex); if (err) return err; @@ -319,7 +343,7 @@ int intel_context_prepare_remote_request(struct intel_context *ce, * words transfer the pinned ce object to tracked active request. */ GEM_BUG_ON(i915_active_is_idle(&ce->active)); - return i915_active_ref(&ce->active, rq->timeline, rq); + return i915_active_add_request(&ce->active, rq); } struct i915_request *intel_context_create_request(struct intel_context *ce) diff --git a/drivers/gpu/drm/i915/gt/intel_context.h b/drivers/gpu/drm/i915/gt/intel_context.h index dd742ac2fbdb..68b3d317d959 100644 --- a/drivers/gpu/drm/i915/gt/intel_context.h +++ b/drivers/gpu/drm/i915/gt/intel_context.h @@ -12,6 +12,7 @@ #include "i915_active.h" #include "intel_context_types.h" #include "intel_engine_types.h" +#include "intel_ring_types.h" #include "intel_timeline_types.h" void intel_context_init(struct intel_context *ce, diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h index bf9cedfccbf0..6959b05ae5f8 100644 --- a/drivers/gpu/drm/i915/gt/intel_context_types.h +++ b/drivers/gpu/drm/i915/gt/intel_context_types.h @@ -58,6 +58,7 @@ struct intel_context { u32 *lrc_reg_state; u64 lrc_desc; + u32 tag; /* cookie passed to HW to track this context on submission */ unsigned int active_count; /* protected by timeline->mutex */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index d3c6993f4f46..01765a7ec18f 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -19,6 +19,7 @@ #include "intel_workarounds.h" struct drm_printer; +struct intel_gt; /* Early gen2 devices have a cacheline of just 32 bytes, using 64 is overkill, * but keeps the logic simple. Indeed, the whole purpose of this macro is just @@ -89,38 +90,6 @@ struct drm_printer; /* seqno size is actually only a uint32, but since we plan to use MI_FLUSH_DW to * do the writes, and that must have qw aligned offsets, simply pretend it's 8b. */ -enum intel_engine_hangcheck_action { - ENGINE_IDLE = 0, - ENGINE_WAIT, - ENGINE_ACTIVE_SEQNO, - ENGINE_ACTIVE_HEAD, - ENGINE_ACTIVE_SUBUNITS, - ENGINE_WAIT_KICK, - ENGINE_DEAD, -}; - -static inline const char * -hangcheck_action_to_str(const enum intel_engine_hangcheck_action a) -{ - switch (a) { - case ENGINE_IDLE: - return "idle"; - case ENGINE_WAIT: - return "wait"; - case ENGINE_ACTIVE_SEQNO: - return "active seqno"; - case ENGINE_ACTIVE_HEAD: - return "active head"; - case ENGINE_ACTIVE_SUBUNITS: - return "active subunits"; - case ENGINE_WAIT_KICK: - return "wait kick"; - case ENGINE_DEAD: - return "dead"; - } - - return "unknown"; -} static inline unsigned int execlists_num_ports(const struct intel_engine_execlists * const execlists) @@ -131,9 +100,21 @@ execlists_num_ports(const struct intel_engine_execlists * const execlists) static inline struct i915_request * execlists_active(const struct intel_engine_execlists *execlists) { - GEM_BUG_ON(execlists->active - execlists->inflight > - execlists_num_ports(execlists)); - return READ_ONCE(*execlists->active); + return *READ_ONCE(execlists->active); +} + +static inline void +execlists_active_lock_bh(struct intel_engine_execlists *execlists) +{ + local_bh_disable(); /* prevent local softirq and lock recursion */ + tasklet_lock(&execlists->tasklet); +} + +static inline void +execlists_active_unlock_bh(struct intel_engine_execlists *execlists) +{ + tasklet_unlock(&execlists->tasklet); + local_bh_enable(); /* restore softirq, and kick ksoftirqd! */ } struct i915_request * @@ -192,126 +173,13 @@ intel_write_status_page(struct intel_engine_cs *engine, int reg, u32 value) #define I915_HWS_CSB_WRITE_INDEX 0x1f #define CNL_HWS_CSB_WRITE_INDEX 0x2f -struct intel_ring * -intel_engine_create_ring(struct intel_engine_cs *engine, int size); -int intel_ring_pin(struct intel_ring *ring); -void intel_ring_reset(struct intel_ring *ring, u32 tail); -unsigned int intel_ring_update_space(struct intel_ring *ring); -void intel_ring_unpin(struct intel_ring *ring); -void intel_ring_free(struct kref *ref); - -static inline struct intel_ring *intel_ring_get(struct intel_ring *ring) -{ - kref_get(&ring->ref); - return ring; -} - -static inline void intel_ring_put(struct intel_ring *ring) -{ - kref_put(&ring->ref, intel_ring_free); -} - void intel_engine_stop(struct intel_engine_cs *engine); void intel_engine_cleanup(struct intel_engine_cs *engine); -int __must_check intel_ring_cacheline_align(struct i915_request *rq); - -u32 __must_check *intel_ring_begin(struct i915_request *rq, unsigned int n); - -static inline void intel_ring_advance(struct i915_request *rq, u32 *cs) -{ - /* Dummy function. - * - * This serves as a placeholder in the code so that the reader - * can compare against the preceding intel_ring_begin() and - * check that the number of dwords emitted matches the space - * reserved for the command packet (i.e. the value passed to - * intel_ring_begin()). - */ - GEM_BUG_ON((rq->ring->vaddr + rq->ring->emit) != cs); -} - -static inline u32 intel_ring_wrap(const struct intel_ring *ring, u32 pos) -{ - return pos & (ring->size - 1); -} - -static inline bool -intel_ring_offset_valid(const struct intel_ring *ring, - unsigned int pos) -{ - if (pos & -ring->size) /* must be strictly within the ring */ - return false; - - if (!IS_ALIGNED(pos, 8)) /* must be qword aligned */ - return false; - - return true; -} - -static inline u32 intel_ring_offset(const struct i915_request *rq, void *addr) -{ - /* Don't write ring->size (equivalent to 0) as that hangs some GPUs. */ - u32 offset = addr - rq->ring->vaddr; - GEM_BUG_ON(offset > rq->ring->size); - return intel_ring_wrap(rq->ring, offset); -} - -static inline void -assert_ring_tail_valid(const struct intel_ring *ring, unsigned int tail) -{ - GEM_BUG_ON(!intel_ring_offset_valid(ring, tail)); - - /* - * "Ring Buffer Use" - * Gen2 BSpec "1. Programming Environment" / 1.4.4.6 - * Gen3 BSpec "1c Memory Interface Functions" / 2.3.4.5 - * Gen4+ BSpec "1c Memory Interface and Command Stream" / 5.3.4.5 - * "If the Ring Buffer Head Pointer and the Tail Pointer are on the - * same cacheline, the Head Pointer must not be greater than the Tail - * Pointer." - * - * We use ring->head as the last known location of the actual RING_HEAD, - * it may have advanced but in the worst case it is equally the same - * as ring->head and so we should never program RING_TAIL to advance - * into the same cacheline as ring->head. - */ -#define cacheline(a) round_down(a, CACHELINE_BYTES) - GEM_BUG_ON(cacheline(tail) == cacheline(ring->head) && - tail < ring->head); -#undef cacheline -} - -static inline unsigned int -intel_ring_set_tail(struct intel_ring *ring, unsigned int tail) -{ - /* Whilst writes to the tail are strictly order, there is no - * serialisation between readers and the writers. The tail may be - * read by i915_request_retire() just as it is being updated - * by execlists, as although the breadcrumb is complete, the context - * switch hasn't been seen. - */ - assert_ring_tail_valid(ring, tail); - ring->tail = tail; - return tail; -} - -static inline unsigned int -__intel_ring_space(unsigned int head, unsigned int tail, unsigned int size) -{ - /* - * "If the Ring Buffer Head Pointer and the Tail Pointer are on the - * same cacheline, the Head Pointer must not be greater than the Tail - * Pointer." - */ - GEM_BUG_ON(!is_power_of_2(size)); - return (head - tail - CACHELINE_BYTES) & (size - 1); -} - -int intel_engines_init_mmio(struct drm_i915_private *i915); -int intel_engines_setup(struct drm_i915_private *i915); -int intel_engines_init(struct drm_i915_private *i915); -void intel_engines_cleanup(struct drm_i915_private *i915); +int intel_engines_init_mmio(struct intel_gt *gt); +int intel_engines_setup(struct intel_gt *gt); +int intel_engines_init(struct intel_gt *gt); +void intel_engines_cleanup(struct intel_gt *gt); int intel_engine_init_common(struct intel_engine_cs *engine); void intel_engine_cleanup_common(struct intel_engine_cs *engine); @@ -335,7 +203,6 @@ void intel_engine_init_execlists(struct intel_engine_cs *engine); void intel_engine_init_breadcrumbs(struct intel_engine_cs *engine); void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine); -void intel_engine_signal_breadcrumbs(struct intel_engine_cs *engine); void intel_engine_disarm_breadcrumbs(struct intel_engine_cs *engine); static inline void @@ -408,8 +275,9 @@ static inline void __intel_engine_reset(struct intel_engine_cs *engine, engine->serial++; /* contexts lost */ } -bool intel_engine_is_idle(struct intel_engine_cs *engine); bool intel_engines_are_idle(struct intel_gt *gt); +bool intel_engine_is_idle(struct intel_engine_cs *engine); +void intel_engine_flush_submission(struct intel_engine_cs *engine); void intel_engines_reset_default_submission(struct intel_gt *gt); @@ -420,61 +288,6 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m, const char *header, ...); -static inline void intel_engine_context_in(struct intel_engine_cs *engine) -{ - unsigned long flags; - - if (READ_ONCE(engine->stats.enabled) == 0) - return; - - write_seqlock_irqsave(&engine->stats.lock, flags); - - if (engine->stats.enabled > 0) { - if (engine->stats.active++ == 0) - engine->stats.start = ktime_get(); - GEM_BUG_ON(engine->stats.active == 0); - } - - write_sequnlock_irqrestore(&engine->stats.lock, flags); -} - -static inline void intel_engine_context_out(struct intel_engine_cs *engine) -{ - unsigned long flags; - - if (READ_ONCE(engine->stats.enabled) == 0) - return; - - write_seqlock_irqsave(&engine->stats.lock, flags); - - if (engine->stats.enabled > 0) { - ktime_t last; - - if (engine->stats.active && --engine->stats.active == 0) { - /* - * Decrement the active context count and in case GPU - * is now idle add up to the running total. - */ - last = ktime_sub(ktime_get(), engine->stats.start); - - engine->stats.total = ktime_add(engine->stats.total, - last); - } else if (engine->stats.active == 0) { - /* - * After turning on engine stats, context out might be - * the first event in which case we account from the - * time stats gathering was turned on. - */ - last = ktime_sub(ktime_get(), engine->stats.enabled_at); - - engine->stats.total = ktime_add(engine->stats.total, - last); - } - } - - write_sequnlock_irqrestore(&engine->stats.lock, flags); -} - int intel_enable_engine_stats(struct intel_engine_cs *engine); void intel_disable_engine_stats(struct intel_engine_cs *engine); @@ -511,4 +324,22 @@ void intel_engine_init_active(struct intel_engine_cs *engine, #define ENGINE_MOCK 1 #define ENGINE_VIRTUAL 2 +static inline bool +intel_engine_has_preempt_reset(const struct intel_engine_cs *engine) +{ + if (!IS_ACTIVE(CONFIG_DRM_I915_PREEMPT_TIMEOUT)) + return false; + + return intel_engine_has_preemption(engine); +} + +static inline bool +intel_engine_has_timeslices(const struct intel_engine_cs *engine) +{ + if (!IS_ACTIVE(CONFIG_DRM_I915_TIMESLICE_DURATION)) + return false; + + return intel_engine_has_semaphores(engine); +} + #endif /* _INTEL_RINGBUFFER_H_ */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index 82630db0394b..813bd3a610d2 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -28,15 +28,16 @@ #include "i915_drv.h" -#include "gt/intel_gt.h" - +#include "intel_context.h" #include "intel_engine.h" #include "intel_engine_pm.h" #include "intel_engine_pool.h" #include "intel_engine_user.h" -#include "intel_context.h" +#include "intel_gt.h" +#include "intel_gt_requests.h" #include "intel_lrc.h" #include "intel_reset.h" +#include "intel_ring.h" /* Haswell does have the CXT_SIZE register however it does not appear to be * valid. Now, docs explain in dwords what is in the context object. The full @@ -277,6 +278,9 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id) BUILD_BUG_ON(MAX_ENGINE_CLASS >= BIT(GEN11_ENGINE_CLASS_WIDTH)); BUILD_BUG_ON(MAX_ENGINE_INSTANCE >= BIT(GEN11_ENGINE_INSTANCE_WIDTH)); + if (GEM_DEBUG_WARN_ON(id >= ARRAY_SIZE(gt->engine))) + return -EINVAL; + if (GEM_DEBUG_WARN_ON(info->class > MAX_ENGINE_CLASS)) return -EINVAL; @@ -293,6 +297,7 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id) BUILD_BUG_ON(BITS_PER_TYPE(engine->mask) < I915_NUM_ENGINES); engine->id = id; + engine->legacy_idx = INVALID_ENGINE; engine->mask = BIT(id); engine->i915 = gt->i915; engine->gt = gt; @@ -304,6 +309,15 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id) engine->instance = info->instance; __sprint_engine_name(engine); + engine->props.heartbeat_interval_ms = + CONFIG_DRM_I915_HEARTBEAT_INTERVAL; + engine->props.preempt_timeout_ms = + CONFIG_DRM_I915_PREEMPT_TIMEOUT; + engine->props.stop_timeout_ms = + CONFIG_DRM_I915_STOP_TIMEOUT; + engine->props.timeslice_duration_ms = + CONFIG_DRM_I915_TIMESLICE_DURATION; + /* * To be overridden by the backend on setup. However to facilitate * cleanup on error during setup, we always provide the destroy vfunc. @@ -328,6 +342,7 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id) intel_engine_sanitize_mmio(engine); gt->engine_class[info->class][info->instance] = engine; + gt->engine[id] = engine; intel_engine_add_user(engine); gt->i915->engine[id] = engine; @@ -365,38 +380,40 @@ static void __setup_engine_capabilities(struct intel_engine_cs *engine) } } -static void intel_setup_engine_capabilities(struct drm_i915_private *i915) +static void intel_setup_engine_capabilities(struct intel_gt *gt) { struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) + for_each_engine(engine, gt, id) __setup_engine_capabilities(engine); } /** * intel_engines_cleanup() - free the resources allocated for Command Streamers - * @i915: the i915 devic + * @gt: pointer to struct intel_gt */ -void intel_engines_cleanup(struct drm_i915_private *i915) +void intel_engines_cleanup(struct intel_gt *gt) { struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { engine->destroy(engine); - i915->engine[id] = NULL; + gt->engine[id] = NULL; + gt->i915->engine[id] = NULL; } } /** * intel_engines_init_mmio() - allocate and prepare the Engine Command Streamers - * @i915: the i915 device + * @gt: pointer to struct intel_gt * * Return: non-zero if the initialization failed. */ -int intel_engines_init_mmio(struct drm_i915_private *i915) +int intel_engines_init_mmio(struct intel_gt *gt) { + struct drm_i915_private *i915 = gt->i915; struct intel_device_info *device_info = mkwrite_device_info(i915); const unsigned int engine_mask = INTEL_INFO(i915)->engine_mask; unsigned int mask = 0; @@ -414,7 +431,7 @@ int intel_engines_init_mmio(struct drm_i915_private *i915) if (!HAS_ENGINE(i915, i)) continue; - err = intel_engine_setup(&i915->gt, i); + err = intel_engine_setup(gt, i); if (err) goto cleanup; @@ -431,36 +448,36 @@ int intel_engines_init_mmio(struct drm_i915_private *i915) RUNTIME_INFO(i915)->num_engines = hweight32(mask); - intel_gt_check_and_clear_faults(&i915->gt); + intel_gt_check_and_clear_faults(gt); - intel_setup_engine_capabilities(i915); + intel_setup_engine_capabilities(gt); return 0; cleanup: - intel_engines_cleanup(i915); + intel_engines_cleanup(gt); return err; } /** * intel_engines_init() - init the Engine Command Streamers - * @i915: i915 device private + * @gt: pointer to struct intel_gt * * Return: non-zero if the initialization failed. */ -int intel_engines_init(struct drm_i915_private *i915) +int intel_engines_init(struct intel_gt *gt) { int (*init)(struct intel_engine_cs *engine); struct intel_engine_cs *engine; enum intel_engine_id id; int err; - if (HAS_EXECLISTS(i915)) + if (HAS_EXECLISTS(gt->i915)) init = intel_execlists_submission_init; else init = intel_ring_submission_init; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { err = init(engine); if (err) goto cleanup; @@ -469,7 +486,7 @@ int intel_engines_init(struct drm_i915_private *i915) return 0; cleanup: - intel_engines_cleanup(i915); + intel_engines_cleanup(gt); return err; } @@ -513,7 +530,7 @@ static int pin_ggtt_status_page(struct intel_engine_cs *engine, unsigned int flags; flags = PIN_GLOBAL; - if (!HAS_LLC(engine->i915)) + if (!HAS_LLC(engine->i915) && i915_ggtt_has_aperture(engine->gt->ggtt)) /* * On g33, we cannot place HWS above 256MiB, so * restrict its pinning to the low mappable arena. @@ -597,9 +614,9 @@ static int intel_engine_setup_common(struct intel_engine_cs *engine) intel_engine_init_active(engine, ENGINE_PHYSICAL); intel_engine_init_breadcrumbs(engine); intel_engine_init_execlists(engine); - intel_engine_init_hangcheck(engine); intel_engine_init_cmd_parser(engine); intel_engine_init__pm(engine); + intel_engine_init_retire(engine); intel_engine_pool_init(&engine->pool); @@ -616,26 +633,26 @@ static int intel_engine_setup_common(struct intel_engine_cs *engine) /** * intel_engines_setup- setup engine state not requiring hw access - * @i915: Device to setup. + * @gt: pointer to struct intel_gt * * Initializes engine structure members shared between legacy and execlists * submission modes which do not require hardware access. * * Typically done early in the submission mode specific engine setup stage. */ -int intel_engines_setup(struct drm_i915_private *i915) +int intel_engines_setup(struct intel_gt *gt) { int (*setup)(struct intel_engine_cs *engine); struct intel_engine_cs *engine; enum intel_engine_id id; int err; - if (HAS_EXECLISTS(i915)) + if (HAS_EXECLISTS(gt->i915)) setup = intel_execlists_submission_setup; else setup = intel_ring_submission_setup; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { err = intel_engine_setup_common(engine); if (err) goto cleanup; @@ -653,7 +670,7 @@ int intel_engines_setup(struct drm_i915_private *i915) return 0; cleanup: - intel_engines_cleanup(i915); + intel_engines_cleanup(gt); return err; } @@ -680,6 +697,8 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine) engine->status_page.vma)) goto out_frame; + mutex_lock(&frame->timeline.mutex); + frame->ring.vaddr = frame->cs; frame->ring.size = sizeof(frame->cs); frame->ring.effective_size = frame->ring.size; @@ -688,18 +707,22 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine) frame->rq.i915 = engine->i915; frame->rq.engine = engine; frame->rq.ring = &frame->ring; - frame->rq.timeline = &frame->timeline; + rcu_assign_pointer(frame->rq.timeline, &frame->timeline); dw = intel_timeline_pin(&frame->timeline); if (dw < 0) goto out_timeline; + spin_lock_irq(&engine->active.lock); dw = engine->emit_fini_breadcrumb(&frame->rq, frame->cs) - frame->cs; + spin_unlock_irq(&engine->active.lock); + GEM_BUG_ON(dw & 1); /* RING_TAIL must be qword aligned */ intel_timeline_unpin(&frame->timeline); out_timeline: + mutex_unlock(&frame->timeline.mutex); intel_timeline_fini(&frame->timeline); out_frame: kfree(frame); @@ -730,6 +753,7 @@ intel_engine_init_active(struct intel_engine_cs *engine, unsigned int subclass) static struct intel_context * create_kernel_context(struct intel_engine_cs *engine) { + static struct lock_class_key kernel; struct intel_context *ce; int err; @@ -745,6 +769,14 @@ create_kernel_context(struct intel_engine_cs *engine) return ERR_PTR(err); } + /* + * Give our perma-pinned kernel timelines a separate lockdep class, + * so that we can use them from within the normal user timelines + * should we need to inject GPU operations during their request + * construction. + */ + lockdep_set_class(&ce->timeline->mutex, &kernel); + return ce; } @@ -807,6 +839,7 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) cleanup_status_page(engine); + intel_engine_fini_retire(engine); intel_engine_pool_fini(&engine->pool); intel_engine_fini_breadcrumbs(engine); intel_engine_cleanup_cmd_parser(engine); @@ -814,8 +847,10 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) if (engine->default_state) i915_gem_object_put(engine->default_state); - intel_context_unpin(engine->kernel_context); - intel_context_put(engine->kernel_context); + if (engine->kernel_context) { + intel_context_unpin(engine->kernel_context); + intel_context_put(engine->kernel_context); + } GEM_BUG_ON(!llist_empty(&engine->barrier_tasks)); intel_wa_list_free(&engine->ctx_wa_list); @@ -851,6 +886,21 @@ u64 intel_engine_get_last_batch_head(const struct intel_engine_cs *engine) return bbaddr; } +static unsigned long stop_timeout(const struct intel_engine_cs *engine) +{ + if (in_atomic() || irqs_disabled()) /* inside atomic preempt-reset? */ + return 0; + + /* + * If we are doing a normal GPU reset, we can take our time and allow + * the engine to quiesce. We've stopped submission to the engine, and + * if we wait long enough an innocent context should complete and + * leave the engine idle. So they should not be caught unaware by + * the forthcoming GPU reset (which usually follows the stop_cs)! + */ + return READ_ONCE(engine->props.stop_timeout_ms); +} + int intel_engine_stop_cs(struct intel_engine_cs *engine) { struct intel_uncore *uncore = engine->uncore; @@ -868,7 +918,7 @@ int intel_engine_stop_cs(struct intel_engine_cs *engine) err = 0; if (__intel_wait_for_register_fw(uncore, mode, MODE_IDLE, MODE_IDLE, - 1000, 0, + 1000, stop_timeout(engine), NULL)) { GEM_TRACE("%s: timed out on STOP_RING -> IDLE\n", engine->name); err = -ETIMEDOUT; @@ -948,6 +998,7 @@ void intel_engine_get_instdone(struct intel_engine_cs *engine, struct intel_instdone *instdone) { struct drm_i915_private *i915 = engine->i915; + const struct sseu_dev_info *sseu = &RUNTIME_INFO(i915)->sseu; struct intel_uncore *uncore = engine->uncore; u32 mmio_base = engine->mmio_base; int slice; @@ -965,7 +1016,7 @@ void intel_engine_get_instdone(struct intel_engine_cs *engine, instdone->slice_common = intel_uncore_read(uncore, GEN7_SC_INSTDONE); - for_each_instdone_slice_subslice(i915, slice, subslice) { + for_each_instdone_slice_subslice(i915, sseu, slice, subslice) { instdone->sampler[slice][subslice] = read_subslice_reg(engine, slice, subslice, GEN7_SAMPLER_INSTDONE); @@ -1031,6 +1082,25 @@ static bool ring_is_idle(struct intel_engine_cs *engine) return idle; } +void intel_engine_flush_submission(struct intel_engine_cs *engine) +{ + struct tasklet_struct *t = &engine->execlists.tasklet; + + if (__tasklet_is_scheduled(t)) { + local_bh_disable(); + if (tasklet_trylock(t)) { + /* Must wait for any GPU reset in progress. */ + if (__tasklet_is_enabled(t)) + t->func(t->data); + tasklet_unlock(t); + } + local_bh_enable(); + } + + /* Otherwise flush the tasklet if it was running on another cpu */ + tasklet_unlock_wait(t); +} + /** * intel_engine_is_idle() - Report if the engine has finished process all work * @engine: the intel_engine_cs @@ -1049,21 +1119,9 @@ bool intel_engine_is_idle(struct intel_engine_cs *engine) /* Waiting to drain ELSP? */ if (execlists_active(&engine->execlists)) { - struct tasklet_struct *t = &engine->execlists.tasklet; - synchronize_hardirq(engine->i915->drm.pdev->irq); - local_bh_disable(); - if (tasklet_trylock(t)) { - /* Must wait for any GPU reset in progress. */ - if (__tasklet_is_enabled(t)) - t->func(t->data); - tasklet_unlock(t); - } - local_bh_enable(); - - /* Otherwise flush the tasklet if it was on another cpu */ - tasklet_unlock_wait(t); + intel_engine_flush_submission(engine); if (execlists_active(&engine->execlists)) return false; @@ -1093,7 +1151,7 @@ bool intel_engines_are_idle(struct intel_gt *gt) if (!READ_ONCE(gt->awake)) return true; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { if (!intel_engine_is_idle(engine)) return false; } @@ -1106,7 +1164,7 @@ void intel_engines_reset_default_submission(struct intel_gt *gt) struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, gt->i915, id) + for_each_engine(engine, gt, id) engine->set_default_submission(engine); } @@ -1118,6 +1176,8 @@ bool intel_engine_can_store_dword(struct intel_engine_cs *engine) case 3: /* maybe only uses physical not virtual addresses */ return !(IS_I915G(engine->i915) || IS_I915GM(engine->i915)); + case 4: + return !IS_I965G(engine->i915); /* who knows! */ case 6: return engine->class != VIDEO_DECODE_CLASS; /* b0rked */ default: @@ -1193,13 +1253,43 @@ static void hexdump(struct drm_printer *m, const void *buf, size_t len) } } +static struct intel_timeline *get_timeline(struct i915_request *rq) +{ + struct intel_timeline *tl; + + /* + * Even though we are holding the engine->active.lock here, there + * is no control over the submission queue per-se and we are + * inspecting the active state at a random point in time, with an + * unknown queue. Play safe and make sure the timeline remains valid. + * (Only being used for pretty printing, one extra kref shouldn't + * cause a camel stampede!) + */ + rcu_read_lock(); + tl = rcu_dereference(rq->timeline); + if (!kref_get_unless_zero(&tl->kref)) + tl = NULL; + rcu_read_unlock(); + + return tl; +} + +static const char *repr_timer(const struct timer_list *t) +{ + if (!READ_ONCE(t->expires)) + return "inactive"; + + if (timer_pending(t)) + return "active"; + + return "expired"; +} + static void intel_engine_print_registers(struct intel_engine_cs *engine, struct drm_printer *m) { struct drm_i915_private *dev_priv = engine->i915; - const struct intel_engine_execlists * const execlists = - &engine->execlists; - unsigned long flags; + struct intel_engine_execlists * const execlists = &engine->execlists; u64 addr; if (engine->id == RENDER_CLASS && IS_GEN_RANGE(dev_priv, 4, 7)) @@ -1256,19 +1346,21 @@ static void intel_engine_print_registers(struct intel_engine_cs *engine, unsigned int idx; u8 read, write; - drm_printf(m, "\tExeclist status: 0x%08x %08x, entries %u\n", - ENGINE_READ(engine, RING_EXECLIST_STATUS_LO), - ENGINE_READ(engine, RING_EXECLIST_STATUS_HI), - num_entries); + drm_printf(m, "\tExeclist tasklet queued? %s (%s), preempt? %s, timeslice? %s\n", + yesno(test_bit(TASKLET_STATE_SCHED, + &engine->execlists.tasklet.state)), + enableddisabled(!atomic_read(&engine->execlists.tasklet.count)), + repr_timer(&engine->execlists.preempt), + repr_timer(&engine->execlists.timer)); read = execlists->csb_head; write = READ_ONCE(*execlists->csb_write); - drm_printf(m, "\tExeclist CSB read %d, write %d, tasklet queued? %s (%s)\n", - read, write, - yesno(test_bit(TASKLET_STATE_SCHED, - &engine->execlists.tasklet.state)), - enableddisabled(!atomic_read(&engine->execlists.tasklet.count))); + drm_printf(m, "\tExeclist status: 0x%08x %08x; CSB read:%d, write:%d, entries:%d\n", + ENGINE_READ(engine, RING_EXECLIST_STATUS_LO), + ENGINE_READ(engine, RING_EXECLIST_STATUS_HI), + read, write, num_entries); + if (read >= num_entries) read = 0; if (write >= num_entries) @@ -1281,35 +1373,47 @@ static void intel_engine_print_registers(struct intel_engine_cs *engine, idx, hws[idx * 2], hws[idx * 2 + 1]); } - spin_lock_irqsave(&engine->active.lock, flags); + execlists_active_lock_bh(execlists); + rcu_read_lock(); for (port = execlists->active; (rq = *port); port++) { char hdr[80]; int len; len = snprintf(hdr, sizeof(hdr), - "\t\tActive[%d: ", + "\t\tActive[%d]: ", (int)(port - execlists->active)); - if (!i915_request_signaled(rq)) + if (!i915_request_signaled(rq)) { + struct intel_timeline *tl = get_timeline(rq); + len += snprintf(hdr + len, sizeof(hdr) - len, "ring:{start:%08x, hwsp:%08x, seqno:%08x}, ", i915_ggtt_offset(rq->ring->vma), - rq->timeline->hwsp_offset, + tl ? tl->hwsp_offset : 0, hwsp_seqno(rq)); + + if (tl) + intel_timeline_put(tl); + } snprintf(hdr + len, sizeof(hdr) - len, "rq: "); print_request(m, rq, hdr); } for (port = execlists->pending; (rq = *port); port++) { + struct intel_timeline *tl = get_timeline(rq); char hdr[80]; snprintf(hdr, sizeof(hdr), "\t\tPending[%d] ring:{start:%08x, hwsp:%08x, seqno:%08x}, rq: ", (int)(port - execlists->pending), i915_ggtt_offset(rq->ring->vma), - rq->timeline->hwsp_offset, + tl ? tl->hwsp_offset : 0, hwsp_seqno(rq)); print_request(m, rq, hdr); + + if (tl) + intel_timeline_put(tl); } - spin_unlock_irqrestore(&engine->active.lock, flags); + rcu_read_unlock(); + execlists_active_unlock_bh(execlists); } else if (INTEL_GEN(dev_priv) > 6) { drm_printf(m, "\tPP_DIR_BASE: 0x%08x\n", ENGINE_READ(engine, RING_PP_DIR_BASE)); @@ -1374,8 +1478,13 @@ void intel_engine_dump(struct intel_engine_cs *engine, drm_printf(m, "*** WEDGED ***\n"); drm_printf(m, "\tAwake? %d\n", atomic_read(&engine->wakeref.count)); - drm_printf(m, "\tHangcheck: %d ms ago\n", - jiffies_to_msecs(jiffies - engine->hangcheck.action_timestamp)); + + rcu_read_lock(); + rq = READ_ONCE(engine->heartbeat.systole); + if (rq) + drm_printf(m, "\tHeartbeat: %d ms ago\n", + jiffies_to_msecs(jiffies - rq->emitted_jiffies)); + rcu_read_unlock(); drm_printf(m, "\tReset count: %d (global %d)\n", i915_reset_engine_count(error, engine), i915_reset_count(error)); @@ -1385,6 +1494,8 @@ void intel_engine_dump(struct intel_engine_cs *engine, spin_lock_irqsave(&engine->active.lock, flags); rq = intel_engine_find_active_request(engine); if (rq) { + struct intel_timeline *tl = get_timeline(rq); + print_request(m, rq, "\t\tactive "); drm_printf(m, "\t\tring->start: 0x%08x\n", @@ -1397,18 +1508,27 @@ void intel_engine_dump(struct intel_engine_cs *engine, rq->ring->emit); drm_printf(m, "\t\tring->space: 0x%08x\n", rq->ring->space); - drm_printf(m, "\t\tring->hwsp: 0x%08x\n", - rq->timeline->hwsp_offset); + + if (tl) { + drm_printf(m, "\t\tring->hwsp: 0x%08x\n", + tl->hwsp_offset); + intel_timeline_put(tl); + } print_request_ring(m, rq); + + if (rq->hw_context->lrc_reg_state) { + drm_printf(m, "Logical Ring Context:\n"); + hexdump(m, rq->hw_context->lrc_reg_state, PAGE_SIZE); + } } spin_unlock_irqrestore(&engine->active.lock, flags); drm_printf(m, "\tMMIO base: 0x%08x\n", engine->mmio_base); - wakeref = intel_runtime_pm_get_if_in_use(&engine->i915->runtime_pm); + wakeref = intel_runtime_pm_get_if_in_use(engine->uncore->rpm); if (wakeref) { intel_engine_print_registers(engine, m); - intel_runtime_pm_put(&engine->i915->runtime_pm, wakeref); + intel_runtime_pm_put(engine->uncore->rpm, wakeref); } else { drm_printf(m, "\tDevice is asleep; skipping register dump\n"); } @@ -1440,8 +1560,8 @@ int intel_enable_engine_stats(struct intel_engine_cs *engine) if (!intel_engine_supports_stats(engine)) return -ENODEV; - spin_lock_irqsave(&engine->active.lock, flags); - write_seqlock(&engine->stats.lock); + execlists_active_lock_bh(execlists); + write_seqlock_irqsave(&engine->stats.lock, flags); if (unlikely(engine->stats.enabled == ~0)) { err = -EBUSY; @@ -1469,8 +1589,8 @@ int intel_enable_engine_stats(struct intel_engine_cs *engine) } unlock: - write_sequnlock(&engine->stats.lock); - spin_unlock_irqrestore(&engine->active.lock, flags); + write_sequnlock_irqrestore(&engine->stats.lock, flags); + execlists_active_unlock_bh(execlists); return err; } diff --git a/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c new file mode 100644 index 000000000000..06aa14c7aa8c --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c @@ -0,0 +1,234 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include "i915_request.h" + +#include "intel_context.h" +#include "intel_engine_heartbeat.h" +#include "intel_engine_pm.h" +#include "intel_engine.h" +#include "intel_gt.h" +#include "intel_reset.h" + +/* + * While the engine is active, we send a periodic pulse along the engine + * to check on its health and to flush any idle-barriers. If that request + * is stuck, and we fail to preempt it, we declare the engine hung and + * issue a reset -- in the hope that restores progress. + */ + +static bool next_heartbeat(struct intel_engine_cs *engine) +{ + long delay; + + delay = READ_ONCE(engine->props.heartbeat_interval_ms); + if (!delay) + return false; + + delay = msecs_to_jiffies_timeout(delay); + if (delay >= HZ) + delay = round_jiffies_up_relative(delay); + schedule_delayed_work(&engine->heartbeat.work, delay); + + return true; +} + +static void idle_pulse(struct intel_engine_cs *engine, struct i915_request *rq) +{ + engine->wakeref_serial = READ_ONCE(engine->serial) + 1; + i915_request_add_active_barriers(rq); +} + +static void show_heartbeat(const struct i915_request *rq, + struct intel_engine_cs *engine) +{ + struct drm_printer p = drm_debug_printer("heartbeat"); + + intel_engine_dump(engine, &p, + "%s heartbeat {prio:%d} not ticking\n", + engine->name, + rq->sched.attr.priority); +} + +static void heartbeat(struct work_struct *wrk) +{ + struct i915_sched_attr attr = { + .priority = I915_USER_PRIORITY(I915_PRIORITY_MIN), + }; + struct intel_engine_cs *engine = + container_of(wrk, typeof(*engine), heartbeat.work.work); + struct intel_context *ce = engine->kernel_context; + struct i915_request *rq; + + if (!intel_engine_pm_get_if_awake(engine)) + return; + + rq = engine->heartbeat.systole; + if (rq && i915_request_completed(rq)) { + i915_request_put(rq); + engine->heartbeat.systole = NULL; + } + + if (intel_gt_is_wedged(engine->gt)) + goto out; + + if (engine->heartbeat.systole) { + if (engine->schedule && + rq->sched.attr.priority < I915_PRIORITY_BARRIER) { + /* + * Gradually raise the priority of the heartbeat to + * give high priority work [which presumably desires + * low latency and no jitter] the chance to naturally + * complete before being preempted. + */ + attr.priority = I915_PRIORITY_MASK; + if (rq->sched.attr.priority >= attr.priority) + attr.priority |= I915_USER_PRIORITY(I915_PRIORITY_HEARTBEAT); + if (rq->sched.attr.priority >= attr.priority) + attr.priority = I915_PRIORITY_BARRIER; + + local_bh_disable(); + engine->schedule(rq, &attr); + local_bh_enable(); + } else { + if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) + show_heartbeat(rq, engine); + + intel_gt_handle_error(engine->gt, engine->mask, + I915_ERROR_CAPTURE, + "stopped heartbeat on %s", + engine->name); + } + goto out; + } + + if (engine->wakeref_serial == engine->serial) + goto out; + + mutex_lock(&ce->timeline->mutex); + + intel_context_enter(ce); + rq = __i915_request_create(ce, GFP_NOWAIT | __GFP_NOWARN); + intel_context_exit(ce); + if (IS_ERR(rq)) + goto unlock; + + idle_pulse(engine, rq); + if (i915_modparams.enable_hangcheck) + engine->heartbeat.systole = i915_request_get(rq); + + __i915_request_commit(rq); + __i915_request_queue(rq, &attr); + +unlock: + mutex_unlock(&ce->timeline->mutex); +out: + if (!next_heartbeat(engine)) + i915_request_put(fetch_and_zero(&engine->heartbeat.systole)); + intel_engine_pm_put(engine); +} + +void intel_engine_unpark_heartbeat(struct intel_engine_cs *engine) +{ + if (!IS_ACTIVE(CONFIG_DRM_I915_HEARTBEAT_INTERVAL)) + return; + + next_heartbeat(engine); +} + +void intel_engine_park_heartbeat(struct intel_engine_cs *engine) +{ + if (cancel_delayed_work(&engine->heartbeat.work)) + i915_request_put(fetch_and_zero(&engine->heartbeat.systole)); +} + +void intel_engine_init_heartbeat(struct intel_engine_cs *engine) +{ + INIT_DELAYED_WORK(&engine->heartbeat.work, heartbeat); +} + +int intel_engine_set_heartbeat(struct intel_engine_cs *engine, + unsigned long delay) +{ + int err; + + /* Send one last pulse before to cleanup persistent hogs */ + if (!delay && IS_ACTIVE(CONFIG_DRM_I915_PREEMPT_TIMEOUT)) { + err = intel_engine_pulse(engine); + if (err) + return err; + } + + WRITE_ONCE(engine->props.heartbeat_interval_ms, delay); + + if (intel_engine_pm_get_if_awake(engine)) { + if (delay) + intel_engine_unpark_heartbeat(engine); + else + intel_engine_park_heartbeat(engine); + intel_engine_pm_put(engine); + } + + return 0; +} + +int intel_engine_pulse(struct intel_engine_cs *engine) +{ + struct i915_sched_attr attr = { .priority = I915_PRIORITY_BARRIER }; + struct intel_context *ce = engine->kernel_context; + struct i915_request *rq; + int err = 0; + + if (!intel_engine_has_preemption(engine)) + return -ENODEV; + + if (!intel_engine_pm_get_if_awake(engine)) + return 0; + + if (mutex_lock_interruptible(&ce->timeline->mutex)) + goto out_rpm; + + intel_context_enter(ce); + rq = __i915_request_create(ce, GFP_NOWAIT | __GFP_NOWARN); + intel_context_exit(ce); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto out_unlock; + } + + rq->flags |= I915_REQUEST_SENTINEL; + idle_pulse(engine, rq); + + __i915_request_commit(rq); + __i915_request_queue(rq, &attr); + +out_unlock: + mutex_unlock(&ce->timeline->mutex); +out_rpm: + intel_engine_pm_put(engine); + return err; +} + +int intel_engine_flush_barriers(struct intel_engine_cs *engine) +{ + struct i915_request *rq; + + if (llist_empty(&engine->barrier_tasks)) + return 0; + + rq = i915_request_create(engine->kernel_context); + if (IS_ERR(rq)) + return PTR_ERR(rq); + + idle_pulse(engine, rq); + i915_request_add(rq); + + return 0; +} + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "selftest_engine_heartbeat.c" +#endif diff --git a/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h new file mode 100644 index 000000000000..a7b8c0f9e005 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h @@ -0,0 +1,23 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_ENGINE_HEARTBEAT_H +#define INTEL_ENGINE_HEARTBEAT_H + +struct intel_engine_cs; + +void intel_engine_init_heartbeat(struct intel_engine_cs *engine); + +int intel_engine_set_heartbeat(struct intel_engine_cs *engine, + unsigned long delay); + +void intel_engine_park_heartbeat(struct intel_engine_cs *engine); +void intel_engine_unpark_heartbeat(struct intel_engine_cs *engine); + +int intel_engine_pulse(struct intel_engine_cs *engine); +int intel_engine_flush_barriers(struct intel_engine_cs *engine); + +#endif /* INTEL_ENGINE_HEARTBEAT_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c index 65b5ca74b394..c1dd0cd3efc7 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c @@ -7,10 +7,13 @@ #include "i915_drv.h" #include "intel_engine.h" +#include "intel_engine_heartbeat.h" #include "intel_engine_pm.h" #include "intel_engine_pool.h" #include "intel_gt.h" #include "intel_gt_pm.h" +#include "intel_rc6.h" +#include "intel_ring.h" static int __engine_unpark(struct intel_wakeref *wf) { @@ -33,7 +36,7 @@ static int __engine_unpark(struct intel_wakeref *wf) if (engine->unpark) engine->unpark(engine); - intel_engine_init_hangcheck(engine); + intel_engine_unpark_heartbeat(engine); return 0; } @@ -52,7 +55,7 @@ static inline unsigned long __timeline_mark_lock(struct intel_context *ce) static inline void __timeline_mark_unlock(struct intel_context *ce, unsigned long flags) { - mutex_release(&ce->timeline->mutex.dep_map, 0, _THIS_IP_); + mutex_release(&ce->timeline->mutex.dep_map, _THIS_IP_); local_irq_restore(flags); } @@ -70,8 +73,42 @@ static inline void __timeline_mark_unlock(struct intel_context *ce, #endif /* !IS_ENABLED(CONFIG_LOCKDEP) */ +static void +__queue_and_release_pm(struct i915_request *rq, + struct intel_timeline *tl, + struct intel_engine_cs *engine) +{ + struct intel_gt_timelines *timelines = &engine->gt->timelines; + + GEM_TRACE("%s\n", engine->name); + + /* + * We have to serialise all potential retirement paths with our + * submission, as we don't want to underflow either the + * engine->wakeref.counter or our timeline->active_count. + * + * Equally, we cannot allow a new submission to start until + * after we finish queueing, nor could we allow that submitter + * to retire us before we are ready! + */ + spin_lock(&timelines->lock); + + /* Let intel_gt_retire_requests() retire us (acquired under lock) */ + if (!atomic_fetch_inc(&tl->active_count)) + list_add_tail(&tl->link, &timelines->active_list); + + /* Hand the request over to HW and so engine_retire() */ + __i915_request_queue(rq, NULL); + + /* Let new submissions commence (and maybe retire this timeline) */ + __intel_wakeref_defer_park(&engine->wakeref); + + spin_unlock(&timelines->lock); +} + static bool switch_to_kernel_context(struct intel_engine_cs *engine) { + struct intel_context *ce = engine->kernel_context; struct i915_request *rq; unsigned long flags; bool result = true; @@ -95,34 +132,61 @@ static bool switch_to_kernel_context(struct intel_engine_cs *engine) * This should hold true as we can only park the engine after * retiring the last request, thus all rings should be empty and * all timelines idle. + * + * For unlocking, there are 2 other parties and the GPU who have a + * stake here. + * + * A new gpu user will be waiting on the engine-pm to start their + * engine_unpark. New waiters are predicated on engine->wakeref.count + * and so intel_wakeref_defer_park() acts like a mutex_unlock of the + * engine->wakeref. + * + * The other party is intel_gt_retire_requests(), which is walking the + * list of active timelines looking for completions. Meanwhile as soon + * as we call __i915_request_queue(), the GPU may complete our request. + * Ergo, if we put ourselves on the timelines.active_list + * (se intel_timeline_enter()) before we increment the + * engine->wakeref.count, we may see the request completion and retire + * it causing an undeflow of the engine->wakeref. */ - flags = __timeline_mark_lock(engine->kernel_context); + flags = __timeline_mark_lock(ce); + GEM_BUG_ON(atomic_read(&ce->timeline->active_count) < 0); - rq = __i915_request_create(engine->kernel_context, GFP_NOWAIT); + rq = __i915_request_create(ce, GFP_NOWAIT); if (IS_ERR(rq)) /* Context switch failed, hope for the best! Maybe reset? */ goto out_unlock; - intel_timeline_enter(rq->timeline); - /* Check again on the next retirement. */ engine->wakeref_serial = engine->serial + 1; i915_request_add_active_barriers(rq); /* Install ourselves as a preemption barrier */ - rq->sched.attr.priority = I915_PRIORITY_UNPREEMPTABLE; + rq->sched.attr.priority = I915_PRIORITY_BARRIER; __i915_request_commit(rq); - /* Release our exclusive hold on the engine */ - __intel_wakeref_defer_park(&engine->wakeref); - __i915_request_queue(rq, NULL); + /* Expose ourselves to the world */ + __queue_and_release_pm(rq, ce->timeline, engine); result = false; out_unlock: - __timeline_mark_unlock(engine->kernel_context, flags); + __timeline_mark_unlock(ce, flags); return result; } +static void call_idle_barriers(struct intel_engine_cs *engine) +{ + struct llist_node *node, *next; + + llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks)) { + struct dma_fence_cb *cb = + container_of((struct list_head *)node, + typeof(*cb), node); + + cb->func(NULL, cb); + } +} + static int __engine_park(struct intel_wakeref *wf) { struct intel_engine_cs *engine = @@ -142,6 +206,9 @@ static int __engine_park(struct intel_wakeref *wf) GEM_TRACE("%s\n", engine->name); + call_idle_barriers(engine); /* cleanup after wedging */ + + intel_engine_park_heartbeat(engine); intel_engine_disarm_breadcrumbs(engine); intel_engine_pool_park(&engine->pool); @@ -158,7 +225,8 @@ static int __engine_park(struct intel_wakeref *wf) engine->execlists.no_priolist = false; - intel_gt_pm_put(engine->gt); + /* While gt calls i915_vma_parked(), we have to break the lock cycle */ + intel_gt_pm_put_async(engine->gt); return 0; } @@ -169,9 +237,10 @@ static const struct intel_wakeref_ops wf_ops = { void intel_engine_init__pm(struct intel_engine_cs *engine) { - struct intel_runtime_pm *rpm = &engine->i915->runtime_pm; + struct intel_runtime_pm *rpm = engine->uncore->rpm; intel_wakeref_init(&engine->wakeref, rpm, &wf_ops); + intel_engine_init_heartbeat(engine); } #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.h b/drivers/gpu/drm/i915/gt/intel_engine_pm.h index 739c50fefcef..24e20344dc22 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pm.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.h @@ -31,6 +31,16 @@ static inline void intel_engine_pm_put(struct intel_engine_cs *engine) intel_wakeref_put(&engine->wakeref); } +static inline void intel_engine_pm_put_async(struct intel_engine_cs *engine) +{ + intel_wakeref_put_async(&engine->wakeref); +} + +static inline void intel_engine_pm_flush(struct intel_engine_cs *engine) +{ + intel_wakeref_unlock_wait(&engine->wakeref); +} + void intel_engine_init__pm(struct intel_engine_cs *engine); #endif /* INTEL_ENGINE_PM_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.c b/drivers/gpu/drm/i915/gt/intel_engine_pool.c index 4cd54c569911..397186818305 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pool.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.c @@ -61,6 +61,7 @@ static int pool_active(struct i915_active *ref) return 0; } +__i915_active_call static void pool_retire(struct i915_active *ref) { struct intel_engine_pool_node *node = @@ -94,7 +95,7 @@ node_create(struct intel_engine_pool *pool, size_t sz) return ERR_PTR(-ENOMEM); node->pool = pool; - i915_active_init(engine->i915, &node->active, pool_active, pool_retire); + i915_active_init(&node->active, pool_active, pool_retire); obj = i915_gem_object_create_internal(engine->i915, sz); if (IS_ERR(obj)) { @@ -103,13 +104,25 @@ node_create(struct intel_engine_pool *pool, size_t sz) return ERR_CAST(obj); } + i915_gem_object_set_readonly(obj); + node->obj = obj; return node; } +static struct intel_engine_pool *lookup_pool(struct intel_engine_cs *engine) +{ + if (intel_engine_is_virtual(engine)) + engine = intel_virtual_engine_get_sibling(engine, 0); + + GEM_BUG_ON(!engine); + return &engine->pool; +} + struct intel_engine_pool_node * -intel_engine_pool_get(struct intel_engine_pool *pool, size_t size) +intel_engine_get_pool(struct intel_engine_cs *engine, size_t size) { + struct intel_engine_pool *pool = lookup_pool(engine); struct intel_engine_pool_node *node; struct list_head *list; unsigned long flags; diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.h b/drivers/gpu/drm/i915/gt/intel_engine_pool.h index 8d069efd9457..1bd89cadc3b7 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pool.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.h @@ -12,13 +12,13 @@ #include "i915_request.h" struct intel_engine_pool_node * -intel_engine_pool_get(struct intel_engine_pool *pool, size_t size); +intel_engine_get_pool(struct intel_engine_cs *engine, size_t size); static inline int intel_engine_pool_mark_active(struct intel_engine_pool_node *node, struct i915_request *rq) { - return i915_active_ref(&node->active, rq->timeline, rq); + return i915_active_add_request(&node->active, rq); } static inline void diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h index a82cea95c2f2..17f1f1441efc 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h @@ -15,6 +15,7 @@ #include <linux/rbtree.h> #include <linux/timer.h> #include <linux/types.h> +#include <linux/workqueue.h> #include "i915_gem.h" #include "i915_pmu.h" @@ -58,6 +59,7 @@ struct i915_gem_context; struct i915_request; struct i915_sched_attr; struct intel_gt; +struct intel_ring; struct intel_uncore; typedef u8 intel_engine_mask_t; @@ -76,40 +78,6 @@ struct intel_instdone { u32 row[I915_MAX_SLICES][I915_MAX_SUBSLICES]; }; -struct intel_engine_hangcheck { - u64 acthd; - u32 last_ring; - u32 last_head; - unsigned long action_timestamp; - struct intel_instdone instdone; -}; - -struct intel_ring { - struct kref ref; - struct i915_vma *vma; - void *vaddr; - - /* - * As we have two types of rings, one global to the engine used - * by ringbuffer submission and those that are exclusive to a - * context used by execlists, we have to play safe and allow - * atomic updates to the pin_count. However, the actual pinning - * of the context is either done during initialisation for - * ringbuffer submission or serialised as part of the context - * pinning for execlists, and so we do not need a mutex ourselves - * to serialise intel_ring_pin/intel_ring_unpin. - */ - atomic_t pin_count; - - u32 head; - u32 tail; - u32 emit; - - u32 space; - u32 size; - u32 effective_size; -}; - /* * we use a single page to load ctx workarounds so all of these * values are referred in terms of dwords @@ -148,6 +116,7 @@ enum intel_engine_id { VECS1, #define _VECS(n) (VECS0 + (n)) I915_NUM_ENGINES +#define INVALID_ENGINE ((enum intel_engine_id)-1) }; struct st_preempt_hang { @@ -174,6 +143,11 @@ struct intel_engine_execlists { struct timer_list timer; /** + * @preempt: reset the current context if it fails to give way + */ + struct timer_list preempt; + + /** * @default_priolist: priority list for I915_PRIORITY_NORMAL */ struct i915_priolist default_priolist; @@ -303,10 +277,12 @@ struct intel_engine_cs { u8 uabi_class; u8 uabi_instance; + u32 uabi_capabilities; u32 context_size; u32 mmio_base; - u32 uabi_capabilities; + unsigned int context_tag; +#define NUM_CONTEXT_TAG roundup_pow_of_two(2 * EXECLIST_MAX_PORTS) struct rb_node uabi_node; @@ -323,6 +299,11 @@ struct intel_engine_cs { intel_engine_mask_t saturated; /* submitting semaphores too late? */ + struct { + struct delayed_work work; + struct i915_request *systole; + } heartbeat; + unsigned long serial; unsigned long wakeref_serial; @@ -470,17 +451,25 @@ struct intel_engine_cs { struct intel_engine_execlists execlists; + /* + * Keep track of completed timelines on this engine for early + * retirement with the goal of quickly enabling powersaving as + * soon as the engine is idle. + */ + struct intel_timeline *retire; + struct work_struct retire_work; + /* status_notifier: list of callbacks for context-switch changes */ struct atomic_notifier_head context_status_notifier; - struct intel_engine_hangcheck hangcheck; - -#define I915_ENGINE_NEEDS_CMD_PARSER BIT(0) +#define I915_ENGINE_USING_CMD_PARSER BIT(0) #define I915_ENGINE_SUPPORTS_STATS BIT(1) #define I915_ENGINE_HAS_PREEMPTION BIT(2) #define I915_ENGINE_HAS_SEMAPHORES BIT(3) #define I915_ENGINE_NEEDS_BREADCRUMB_TASKLET BIT(4) #define I915_ENGINE_IS_VIRTUAL BIT(5) +#define I915_ENGINE_HAS_RELATIVE_MMIO BIT(6) +#define I915_ENGINE_REQUIRES_CMD_PARSER BIT(7) unsigned int flags; /* @@ -538,12 +527,25 @@ struct intel_engine_cs { */ ktime_t total; } stats; + + struct { + unsigned long heartbeat_interval_ms; + unsigned long preempt_timeout_ms; + unsigned long stop_timeout_ms; + unsigned long timeslice_duration_ms; + } props; }; static inline bool -intel_engine_needs_cmd_parser(const struct intel_engine_cs *engine) +intel_engine_using_cmd_parser(const struct intel_engine_cs *engine) { - return engine->flags & I915_ENGINE_NEEDS_CMD_PARSER; + return engine->flags & I915_ENGINE_USING_CMD_PARSER; +} + +static inline bool +intel_engine_requires_cmd_parser(const struct intel_engine_cs *engine) +{ + return engine->flags & I915_ENGINE_REQUIRES_CMD_PARSER; } static inline bool @@ -576,20 +578,24 @@ intel_engine_is_virtual(const struct intel_engine_cs *engine) return engine->flags & I915_ENGINE_IS_VIRTUAL; } -#define instdone_slice_mask(dev_priv__) \ - (IS_GEN(dev_priv__, 7) ? \ - 1 : RUNTIME_INFO(dev_priv__)->sseu.slice_mask) +static inline bool +intel_engine_has_relative_mmio(const struct intel_engine_cs * const engine) +{ + return engine->flags & I915_ENGINE_HAS_RELATIVE_MMIO; +} -#define instdone_subslice_mask(dev_priv__) \ - (IS_GEN(dev_priv__, 7) ? \ - 1 : RUNTIME_INFO(dev_priv__)->sseu.subslice_mask[0]) +#define instdone_has_slice(dev_priv___, sseu___, slice___) \ + ((IS_GEN(dev_priv___, 7) ? 1 : ((sseu___)->slice_mask)) & BIT(slice___)) -#define for_each_instdone_slice_subslice(dev_priv__, slice__, subslice__) \ - for ((slice__) = 0, (subslice__) = 0; \ - (slice__) < I915_MAX_SLICES; \ - (subslice__) = ((subslice__) + 1) < I915_MAX_SUBSLICES ? (subslice__) + 1 : 0, \ - (slice__) += ((subslice__) == 0)) \ - for_each_if((BIT(slice__) & instdone_slice_mask(dev_priv__)) && \ - (BIT(subslice__) & instdone_subslice_mask(dev_priv__))) +#define instdone_has_subslice(dev_priv__, sseu__, slice__, subslice__) \ + (IS_GEN(dev_priv__, 7) ? (1 & BIT(subslice__)) : \ + intel_sseu_has_subslice(sseu__, 0, subslice__)) +#define for_each_instdone_slice_subslice(dev_priv_, sseu_, slice_, subslice_) \ + for ((slice_) = 0, (subslice_) = 0; (slice_) < I915_MAX_SLICES; \ + (subslice_) = ((subslice_) + 1) % I915_MAX_SUBSLICES, \ + (slice_) += ((subslice_) == 0)) \ + for_each_if((instdone_has_slice(dev_priv_, sseu_, slice_)) && \ + (instdone_has_subslice(dev_priv_, sseu_, slice_, \ + subslice_))) #endif /* __INTEL_ENGINE_TYPES_H__ */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c index 77cd5de83930..7f7150a733f4 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_user.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c @@ -160,10 +160,10 @@ static int legacy_ring_idx(const struct legacy_ring *ring) }; if (GEM_DEBUG_WARN_ON(ring->class >= ARRAY_SIZE(map))) - return -1; + return INVALID_ENGINE; if (GEM_DEBUG_WARN_ON(ring->instance >= map[ring->class].max)) - return -1; + return INVALID_ENGINE; return map[ring->class].base + ring->instance; } @@ -171,23 +171,15 @@ static int legacy_ring_idx(const struct legacy_ring *ring) static void add_legacy_ring(struct legacy_ring *ring, struct intel_engine_cs *engine) { - int idx; - if (engine->gt != ring->gt || engine->class != ring->class) { ring->gt = engine->gt; ring->class = engine->class; ring->instance = 0; } - idx = legacy_ring_idx(ring); - if (unlikely(idx == -1)) - return; - - GEM_BUG_ON(idx >= ARRAY_SIZE(ring->gt->engine)); - ring->gt->engine[idx] = engine; - ring->instance++; - - engine->legacy_idx = idx; + engine->legacy_idx = legacy_ring_idx(ring); + if (engine->legacy_idx != INVALID_ENGINE) + ring->instance++; } void intel_engines_driver_register(struct drm_i915_private *i915) diff --git a/drivers/gpu/drm/i915/gt/intel_gpu_commands.h b/drivers/gpu/drm/i915/gt/intel_gpu_commands.h index 86e00a2db8a4..4294f146f13c 100644 --- a/drivers/gpu/drm/i915/gt/intel_gpu_commands.h +++ b/drivers/gpu/drm/i915/gt/intel_gpu_commands.h @@ -112,6 +112,7 @@ #define MI_SEMAPHORE_SIGNAL MI_INSTR(0x1b, 0) /* GEN8+ */ #define MI_SEMAPHORE_TARGET(engine) ((engine)<<15) #define MI_SEMAPHORE_WAIT MI_INSTR(0x1c, 2) /* GEN8+ */ +#define MI_SEMAPHORE_WAIT_TOKEN MI_INSTR(0x1c, 3) /* GEN12+ */ #define MI_SEMAPHORE_POLL (1 << 15) #define MI_SEMAPHORE_SAD_GT_SDD (0 << 12) #define MI_SEMAPHORE_SAD_GTE_SDD (1 << 12) @@ -119,6 +120,8 @@ #define MI_SEMAPHORE_SAD_LTE_SDD (3 << 12) #define MI_SEMAPHORE_SAD_EQ_SDD (4 << 12) #define MI_SEMAPHORE_SAD_NEQ_SDD (5 << 12) +#define MI_SEMAPHORE_TOKEN_MASK REG_GENMASK(9, 5) +#define MI_SEMAPHORE_TOKEN_SHIFT 5 #define MI_STORE_DWORD_IMM MI_INSTR(0x20, 1) #define MI_STORE_DWORD_IMM_GEN4 MI_INSTR(0x20, 2) #define MI_MEM_VIRTUAL (1 << 22) /* 945,g33,965 */ @@ -132,7 +135,10 @@ * address/value pairs. Don't overdue it, though, x <= 2^4 must hold! */ #define MI_LOAD_REGISTER_IMM(x) MI_INSTR(0x22, 2*(x)-1) +/* Gen11+. addr = base + (ctx_restore ? offset & GENMASK(12,2) : offset) */ +#define MI_LRI_CS_MMIO (1<<19) #define MI_LRI_FORCE_POSTED (1<<12) +#define MI_LOAD_REGISTER_IMM_MAX_REGS (126) #define MI_STORE_REGISTER_MEM MI_INSTR(0x24, 1) #define MI_STORE_REGISTER_MEM_GEN8 MI_INSTR(0x24, 2) #define MI_SRM_LRM_GLOBAL_GTT (1<<22) @@ -147,6 +153,7 @@ #define MI_FLUSH_DW_USE_PPGTT (0<<2) #define MI_LOAD_REGISTER_MEM MI_INSTR(0x29, 1) #define MI_LOAD_REGISTER_MEM_GEN8 MI_INSTR(0x29, 2) +#define MI_LOAD_REGISTER_REG MI_INSTR(0x2A, 1) #define MI_BATCH_BUFFER MI_INSTR(0x30, 1) #define MI_BATCH_NON_SECURE (1) /* for snb/ivb/vlv this also means "batch in ppgtt" when ppgtt is enabled. */ @@ -156,7 +163,8 @@ #define MI_BATCH_BUFFER_START MI_INSTR(0x31, 0) #define MI_BATCH_GTT (2<<6) /* aliased with (1<<7) on gen4 */ #define MI_BATCH_BUFFER_START_GEN8 MI_INSTR(0x31, 1) -#define MI_BATCH_RESOURCE_STREAMER (1<<10) +#define MI_BATCH_RESOURCE_STREAMER REG_BIT(10) +#define MI_BATCH_PREDICATE REG_BIT(15) /* HSW+ on RCS only*/ /* * 3D instructions used by the kernel @@ -217,6 +225,7 @@ #define PIPE_CONTROL_CS_STALL (1<<20) #define PIPE_CONTROL_TLB_INVALIDATE (1<<18) #define PIPE_CONTROL_MEDIA_STATE_CLEAR (1<<16) +#define PIPE_CONTROL_WRITE_TIMESTAMP (3<<14) #define PIPE_CONTROL_QW_WRITE (1<<14) #define PIPE_CONTROL_POST_SYNC_OP_MASK (3<<14) #define PIPE_CONTROL_DEPTH_STALL (1<<13) @@ -224,7 +233,9 @@ #define PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH (1<<12) /* gen6+ */ #define PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE (1<<11) /* MBZ on ILK */ #define PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE (1<<10) /* GM45+ only */ +#define PIPE_CONTROL_L3_RO_CACHE_INVALIDATE REG_BIT(10) /* gen12 */ #define PIPE_CONTROL_INDIRECT_STATE_DISABLE (1<<9) +#define PIPE_CONTROL_HDC_PIPELINE_FLUSH REG_BIT(9) /* gen12 */ #define PIPE_CONTROL_NOTIFY (1<<8) #define PIPE_CONTROL_FLUSH_ENABLE (1<<7) /* gen7+ */ #define PIPE_CONTROL_DC_FLUSH_ENABLE (1<<5) @@ -235,6 +246,29 @@ #define PIPE_CONTROL_DEPTH_CACHE_FLUSH (1<<0) #define PIPE_CONTROL_GLOBAL_GTT (1<<2) /* in addr dword */ +#define MI_MATH(x) MI_INSTR(0x1a, (x) - 1) +#define MI_MATH_INSTR(opcode, op1, op2) ((opcode) << 20 | (op1) << 10 | (op2)) +/* Opcodes for MI_MATH_INSTR */ +#define MI_MATH_NOOP MI_MATH_INSTR(0x000, 0x0, 0x0) +#define MI_MATH_LOAD(op1, op2) MI_MATH_INSTR(0x080, op1, op2) +#define MI_MATH_LOADINV(op1, op2) MI_MATH_INSTR(0x480, op1, op2) +#define MI_MATH_LOAD0(op1) MI_MATH_INSTR(0x081, op1) +#define MI_MATH_LOAD1(op1) MI_MATH_INSTR(0x481, op1) +#define MI_MATH_ADD MI_MATH_INSTR(0x100, 0x0, 0x0) +#define MI_MATH_SUB MI_MATH_INSTR(0x101, 0x0, 0x0) +#define MI_MATH_AND MI_MATH_INSTR(0x102, 0x0, 0x0) +#define MI_MATH_OR MI_MATH_INSTR(0x103, 0x0, 0x0) +#define MI_MATH_XOR MI_MATH_INSTR(0x104, 0x0, 0x0) +#define MI_MATH_STORE(op1, op2) MI_MATH_INSTR(0x180, op1, op2) +#define MI_MATH_STOREINV(op1, op2) MI_MATH_INSTR(0x580, op1, op2) +/* Registers used as operands in MI_MATH_INSTR */ +#define MI_MATH_REG(x) (x) +#define MI_MATH_REG_SRCA 0x20 +#define MI_MATH_REG_SRCB 0x21 +#define MI_MATH_REG_ACCU 0x31 +#define MI_MATH_REG_ZF 0x32 +#define MI_MATH_REG_CF 0x33 + /* * Commands used only by the command parser */ @@ -251,7 +285,6 @@ #define MI_CLFLUSH MI_INSTR(0x27, 0) #define MI_REPORT_PERF_COUNT MI_INSTR(0x28, 0) #define MI_REPORT_PERF_COUNT_GGTT (1<<0) -#define MI_LOAD_REGISTER_REG MI_INSTR(0x2A, 0) #define MI_RS_STORE_DATA_IMM MI_INSTR(0x2B, 0) #define MI_LOAD_URB_MEM MI_INSTR(0x2C, 0) #define MI_STORE_URB_MEM MI_INSTR(0x2D, 0) diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c index d48ec9a76ed1..4c26daf7ee46 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.c +++ b/drivers/gpu/drm/i915/gt/intel_gt.c @@ -6,7 +6,12 @@ #include "i915_drv.h" #include "intel_gt.h" #include "intel_gt_pm.h" +#include "intel_gt_requests.h" +#include "intel_mocs.h" +#include "intel_rc6.h" +#include "intel_rps.h" #include "intel_uncore.h" +#include "intel_pm.h" void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915) { @@ -18,15 +23,108 @@ void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915) INIT_LIST_HEAD(>->closed_vma); spin_lock_init(>->closed_lock); - intel_gt_init_hangcheck(gt); intel_gt_init_reset(gt); + intel_gt_init_requests(gt); intel_gt_pm_init_early(gt); + + intel_rps_init_early(>->rps); intel_uc_init_early(>->uc); } -void intel_gt_init_hw(struct drm_i915_private *i915) +void intel_gt_init_hw_early(struct intel_gt *gt, struct i915_ggtt *ggtt) +{ + gt->ggtt = ggtt; + + intel_gt_sanitize(gt, false); +} + +static void init_unused_ring(struct intel_gt *gt, u32 base) { - i915->gt.ggtt = &i915->ggtt; + struct intel_uncore *uncore = gt->uncore; + + intel_uncore_write(uncore, RING_CTL(base), 0); + intel_uncore_write(uncore, RING_HEAD(base), 0); + intel_uncore_write(uncore, RING_TAIL(base), 0); + intel_uncore_write(uncore, RING_START(base), 0); +} + +static void init_unused_rings(struct intel_gt *gt) +{ + struct drm_i915_private *i915 = gt->i915; + + if (IS_I830(i915)) { + init_unused_ring(gt, PRB1_BASE); + init_unused_ring(gt, SRB0_BASE); + init_unused_ring(gt, SRB1_BASE); + init_unused_ring(gt, SRB2_BASE); + init_unused_ring(gt, SRB3_BASE); + } else if (IS_GEN(i915, 2)) { + init_unused_ring(gt, SRB0_BASE); + init_unused_ring(gt, SRB1_BASE); + } else if (IS_GEN(i915, 3)) { + init_unused_ring(gt, PRB1_BASE); + init_unused_ring(gt, PRB2_BASE); + } +} + +int intel_gt_init_hw(struct intel_gt *gt) +{ + struct drm_i915_private *i915 = gt->i915; + struct intel_uncore *uncore = gt->uncore; + int ret; + + BUG_ON(!i915->kernel_context); + ret = intel_gt_terminally_wedged(gt); + if (ret) + return ret; + + gt->last_init_time = ktime_get(); + + /* Double layer security blanket, see i915_gem_init() */ + intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); + + if (HAS_EDRAM(i915) && INTEL_GEN(i915) < 9) + intel_uncore_rmw(uncore, HSW_IDICR, 0, IDIHASHMSK(0xf)); + + if (IS_HASWELL(i915)) + intel_uncore_write(uncore, + MI_PREDICATE_RESULT_2, + IS_HSW_GT3(i915) ? + LOWER_SLICE_ENABLED : LOWER_SLICE_DISABLED); + + /* Apply the GT workarounds... */ + intel_gt_apply_workarounds(gt); + /* ...and determine whether they are sticking. */ + intel_gt_verify_workarounds(gt, "init"); + + intel_gt_init_swizzling(gt); + + /* + * At least 830 can leave some of the unused rings + * "active" (ie. head != tail) after resume which + * will prevent c3 entry. Makes sure all unused rings + * are totally idle. + */ + init_unused_rings(gt); + + ret = i915_ppgtt_init_hw(gt); + if (ret) { + DRM_ERROR("Enabling PPGTT failed (%d)\n", ret); + goto out; + } + + /* We can't enable contexts until all firmware is loaded */ + ret = intel_uc_init_hw(>->uc); + if (ret) { + i915_probe_error(i915, "Enabling uc failed (%d)\n", ret); + goto out; + } + + intel_mocs_init(gt); + +out: + intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); + return ret; } static void rmw_set(struct intel_uncore *uncore, i915_reg_t reg, u32 set) @@ -89,7 +187,7 @@ intel_gt_clear_error_registers(struct intel_gt *gt, struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine_masked(engine, i915, engine_mask, id) + for_each_engine_masked(engine, gt, engine_mask, id) gen8_clear_engine_error_register(engine); } } @@ -100,7 +198,7 @@ static void gen6_check_faults(struct intel_gt *gt) enum intel_engine_id id; u32 fault; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { fault = GEN6_RING_FAULT_REG_READ(engine); if (fault & RING_FAULT_VALID) { DRM_DEBUG_DRIVER("Unexpected fault\n" @@ -176,7 +274,7 @@ void intel_gt_check_and_clear_faults(struct intel_gt *gt) void intel_gt_flush_ggtt_writes(struct intel_gt *gt) { - struct drm_i915_private *i915 = gt->i915; + struct intel_uncore *uncore = gt->uncore; intel_wakeref_t wakeref; /* @@ -200,18 +298,18 @@ void intel_gt_flush_ggtt_writes(struct intel_gt *gt) wmb(); - if (INTEL_INFO(i915)->has_coherent_ggtt) + if (INTEL_INFO(gt->i915)->has_coherent_ggtt) return; intel_gt_chipset_flush(gt); - with_intel_runtime_pm(&i915->runtime_pm, wakeref) { - struct intel_uncore *uncore = gt->uncore; + with_intel_runtime_pm(uncore->rpm, wakeref) { + unsigned long flags; - spin_lock_irq(&uncore->lock); + spin_lock_irqsave(&uncore->lock, flags); intel_uncore_posting_read_fw(uncore, RING_HEAD(RENDER_RING_BASE)); - spin_unlock_irq(&uncore->lock); + spin_unlock_irqrestore(&uncore->lock, flags); } } @@ -222,7 +320,12 @@ void intel_gt_chipset_flush(struct intel_gt *gt) intel_gtt_chipset_flush(); } -int intel_gt_init_scratch(struct intel_gt *gt, unsigned int size) +void intel_gt_driver_register(struct intel_gt *gt) +{ + intel_rps_driver_register(>->rps); +} + +static int intel_gt_init_scratch(struct intel_gt *gt, unsigned int size) { struct drm_i915_private *i915 = gt->i915; struct drm_i915_gem_object *obj; @@ -230,7 +333,7 @@ int intel_gt_init_scratch(struct intel_gt *gt, unsigned int size) int ret; obj = i915_gem_object_create_stolen(i915, size); - if (!obj) + if (IS_ERR(obj)) obj = i915_gem_object_create_internal(i915, size); if (IS_ERR(obj)) { DRM_ERROR("Failed to allocate scratch page\n"); @@ -256,11 +359,40 @@ err_unref: return ret; } -void intel_gt_fini_scratch(struct intel_gt *gt) +static void intel_gt_fini_scratch(struct intel_gt *gt) { i915_vma_unpin_and_release(>->scratch, 0); } +int intel_gt_init(struct intel_gt *gt) +{ + int err; + + err = intel_gt_init_scratch(gt, IS_GEN(gt->i915, 2) ? SZ_256K : SZ_4K); + if (err) + return err; + + intel_gt_pm_init(gt); + + return 0; +} + +void intel_gt_driver_remove(struct intel_gt *gt) +{ + GEM_BUG_ON(gt->awake); +} + +void intel_gt_driver_unregister(struct intel_gt *gt) +{ + intel_rps_driver_unregister(>->rps); +} + +void intel_gt_driver_release(struct intel_gt *gt) +{ + intel_gt_pm_fini(gt); + intel_gt_fini_scratch(gt); +} + void intel_gt_driver_late_release(struct intel_gt *gt) { intel_uc_driver_late_release(>->uc); diff --git a/drivers/gpu/drm/i915/gt/intel_gt.h b/drivers/gpu/drm/i915/gt/intel_gt.h index 4920cb351f10..5436f8c30708 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.h +++ b/drivers/gpu/drm/i915/gt/intel_gt.h @@ -28,7 +28,14 @@ static inline struct intel_gt *huc_to_gt(struct intel_huc *huc) } void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915); -void intel_gt_init_hw(struct drm_i915_private *i915); +void intel_gt_init_hw_early(struct intel_gt *gt, struct i915_ggtt *ggtt); +int __must_check intel_gt_init_hw(struct intel_gt *gt); +int intel_gt_init(struct intel_gt *gt); +void intel_gt_driver_register(struct intel_gt *gt); + +void intel_gt_driver_unregister(struct intel_gt *gt); +void intel_gt_driver_remove(struct intel_gt *gt); +void intel_gt_driver_release(struct intel_gt *gt); void intel_gt_driver_late_release(struct intel_gt *gt); @@ -39,11 +46,6 @@ void intel_gt_clear_error_registers(struct intel_gt *gt, void intel_gt_flush_ggtt_writes(struct intel_gt *gt); void intel_gt_chipset_flush(struct intel_gt *gt); -void intel_gt_init_hangcheck(struct intel_gt *gt); - -int intel_gt_init_scratch(struct intel_gt *gt, unsigned int size); -void intel_gt_fini_scratch(struct intel_gt *gt); - static inline u32 intel_gt_scratch_offset(const struct intel_gt *gt, enum intel_gt_scratch_field field) { @@ -55,6 +57,4 @@ static inline bool intel_gt_is_wedged(struct intel_gt *gt) return __intel_reset_failed(>->reset); } -void intel_gt_queue_hangcheck(struct intel_gt *gt); - #endif /* __INTEL_GT_H__ */ diff --git a/drivers/gpu/drm/i915/gt/intel_gt_irq.c b/drivers/gpu/drm/i915/gt/intel_gt_irq.c index 34a4fb624bf7..973ee7eded64 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_irq.c +++ b/drivers/gpu/drm/i915/gt/intel_gt_irq.c @@ -11,6 +11,7 @@ #include "intel_gt.h" #include "intel_gt_irq.h" #include "intel_uncore.h" +#include "intel_rps.h" static void guc_irq_handler(struct intel_guc *guc, u16 iir) { @@ -77,7 +78,7 @@ gen11_other_irq_handler(struct intel_gt *gt, const u8 instance, return guc_irq_handler(>->uc.guc, iir); if (instance == OTHER_GTPM_INSTANCE) - return gen11_rps_irq_handler(gt, iir); + return gen11_rps_irq_handler(>->rps, iir); WARN_ONCE(1, "unhandled other interrupt instance=0x%x, iir=0x%x\n", instance, iir); @@ -336,7 +337,7 @@ void gen8_gt_irq_handler(struct intel_gt *gt, u32 master_ctl, u32 gt_iir[4]) } if (master_ctl & (GEN8_GT_PM_IRQ | GEN8_GT_GUC_IRQ)) { - gen6_rps_irq_handler(gt->i915, gt_iir[2]); + gen6_rps_irq_handler(>->rps, gt_iir[2]); guc_irq_handler(>->uc.guc, gt_iir[2] >> 16); } } diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c index 1363e069ec83..a459a42ad5c2 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c +++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c @@ -4,17 +4,38 @@ * Copyright © 2019 Intel Corporation */ +#include <linux/suspend.h> + #include "i915_drv.h" +#include "i915_globals.h" #include "i915_params.h" +#include "intel_context.h" #include "intel_engine_pm.h" #include "intel_gt.h" #include "intel_gt_pm.h" +#include "intel_gt_requests.h" +#include "intel_llc.h" #include "intel_pm.h" +#include "intel_rc6.h" +#include "intel_rps.h" #include "intel_wakeref.h" -static void pm_notify(struct drm_i915_private *i915, int state) +static void user_forcewake(struct intel_gt *gt, bool suspend) { - blocking_notifier_call_chain(&i915->gt.pm_notifications, state, i915); + int count = atomic_read(>->user_wakeref); + + /* Inside suspend/resume so single threaded, no races to worry about. */ + if (likely(!count)) + return; + + intel_gt_pm_get(gt); + if (suspend) { + GEM_BUG_ON(count > atomic_read(>->wakeref.count)); + atomic_sub(count, >->wakeref.count); + } else { + atomic_add(count, >->wakeref.count); + } + intel_gt_pm_put(gt); } static int __gt_unpark(struct intel_wakeref *wf) @@ -24,6 +45,8 @@ static int __gt_unpark(struct intel_wakeref *wf) GEM_TRACE("\n"); + i915_globals_unpark(); + /* * It seems that the DMC likes to transition between the DC states a lot * when there are no connected displays (no active power domains) during @@ -38,55 +61,66 @@ static int __gt_unpark(struct intel_wakeref *wf) gt->awake = intel_display_power_get(i915, POWER_DOMAIN_GT_IRQ); GEM_BUG_ON(!gt->awake); - intel_enable_gt_powersave(i915); - - i915_update_gfx_val(i915); - if (INTEL_GEN(i915) >= 6) - gen6_rps_busy(i915); + if (NEEDS_RC6_CTX_CORRUPTION_WA(i915)) + intel_uncore_forcewake_get(&i915->uncore, FORCEWAKE_ALL); + intel_rps_unpark(>->rps); i915_pmu_gt_unparked(i915); - intel_gt_queue_hangcheck(gt); - - pm_notify(i915, INTEL_GT_UNPARK); + intel_gt_unpark_requests(gt); return 0; } static int __gt_park(struct intel_wakeref *wf) { - struct drm_i915_private *i915 = - container_of(wf, typeof(*i915), gt.wakeref); - intel_wakeref_t wakeref = fetch_and_zero(&i915->gt.awake); + struct intel_gt *gt = container_of(wf, typeof(*gt), wakeref); + intel_wakeref_t wakeref = fetch_and_zero(>->awake); + struct drm_i915_private *i915 = gt->i915; GEM_TRACE("\n"); - pm_notify(i915, INTEL_GT_PARK); + intel_gt_park_requests(gt); + i915_vma_parked(gt); i915_pmu_gt_parked(i915); - if (INTEL_GEN(i915) >= 6) - gen6_rps_idle(i915); + intel_rps_park(>->rps); /* Everything switched off, flush any residual interrupt just in case */ intel_synchronize_irq(i915); + if (NEEDS_RC6_CTX_CORRUPTION_WA(i915)) { + intel_rc6_ctx_wa_check(&i915->gt.rc6); + intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL); + } + GEM_BUG_ON(!wakeref); intel_display_power_put(i915, POWER_DOMAIN_GT_IRQ, wakeref); + i915_globals_park(); + return 0; } static const struct intel_wakeref_ops wf_ops = { .get = __gt_unpark, .put = __gt_park, - .flags = INTEL_WAKEREF_PUT_ASYNC, }; void intel_gt_pm_init_early(struct intel_gt *gt) { - intel_wakeref_init(>->wakeref, >->i915->runtime_pm, &wf_ops); + intel_wakeref_init(>->wakeref, gt->uncore->rpm, &wf_ops); +} - BLOCKING_INIT_NOTIFIER_HEAD(>->pm_notifications); +void intel_gt_pm_init(struct intel_gt *gt) +{ + /* + * Enabling power-management should be "self-healing". If we cannot + * enable a feature, simply leave it disabled with a notice to the + * user. + */ + intel_rc6_init(>->rc6); + intel_rps_init(>->rps); } static bool reset_engines(struct intel_gt *gt) @@ -111,16 +145,47 @@ void intel_gt_sanitize(struct intel_gt *gt, bool force) { struct intel_engine_cs *engine; enum intel_engine_id id; + intel_wakeref_t wakeref; - GEM_TRACE("\n"); + GEM_TRACE("force:%s\n", yesno(force)); + + /* Use a raw wakeref to avoid calling intel_display_power_get early */ + wakeref = intel_runtime_pm_get(gt->uncore->rpm); + intel_uncore_forcewake_get(gt->uncore, FORCEWAKE_ALL); + + /* + * As we have just resumed the machine and woken the device up from + * deep PCI sleep (presumably D3_cold), assume the HW has been reset + * back to defaults, recovering from whatever wedged state we left it + * in and so worth trying to use the device once more. + */ + if (intel_gt_is_wedged(gt)) + intel_gt_unset_wedged(gt); intel_uc_sanitize(>->uc); - if (!reset_engines(gt) && !force) - return; + for_each_engine(engine, gt, id) + if (engine->reset.prepare) + engine->reset.prepare(engine); + + intel_uc_reset_prepare(>->uc); + + if (reset_engines(gt) || force) { + for_each_engine(engine, gt, id) + __intel_engine_reset(engine, false); + } + + for_each_engine(engine, gt, id) + if (engine->reset.finish) + engine->reset.finish(engine); - for_each_engine(engine, gt->i915, id) - __intel_engine_reset(engine, false); + intel_uncore_forcewake_put(gt->uncore, FORCEWAKE_ALL); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); +} + +void intel_gt_pm_fini(struct intel_gt *gt) +{ + intel_rc6_fini(>->rc6); } int intel_gt_resume(struct intel_gt *gt) @@ -129,6 +194,8 @@ int intel_gt_resume(struct intel_gt *gt) enum intel_engine_id id; int err = 0; + GEM_TRACE("\n"); + /* * After resume, we may need to poke into the pinned kernel * contexts to paper over any damage caused by the sudden suspend. @@ -136,14 +203,23 @@ int intel_gt_resume(struct intel_gt *gt) * allowing us to fixup the user contexts on their first pin. */ intel_gt_pm_get(gt); - for_each_engine(engine, gt->i915, id) { + + intel_uncore_forcewake_get(gt->uncore, FORCEWAKE_ALL); + intel_rc6_sanitize(>->rc6); + + intel_rps_enable(>->rps); + intel_llc_enable(>->llc); + + for_each_engine(engine, gt, id) { struct intel_context *ce; intel_engine_pm_get(engine); ce = engine->kernel_context; - if (ce) + if (ce) { + GEM_BUG_ON(!intel_context_is_pinned(ce)); ce->ops->reset(ce); + } engine->serial++; /* kernel context lost */ err = engine->resume(engine); @@ -156,19 +232,99 @@ int intel_gt_resume(struct intel_gt *gt) break; } } + + intel_rc6_enable(>->rc6); + + intel_uc_resume(>->uc); + + user_forcewake(gt, false); + + intel_uncore_forcewake_put(gt->uncore, FORCEWAKE_ALL); intel_gt_pm_put(gt); return err; } +static void wait_for_suspend(struct intel_gt *gt) +{ + if (!intel_gt_pm_is_awake(gt)) + return; + + if (intel_gt_wait_for_idle(gt, I915_GEM_IDLE_TIMEOUT) == -ETIME) { + /* + * Forcibly cancel outstanding work and leave + * the gpu quiet. + */ + intel_gt_set_wedged(gt); + } + + intel_gt_pm_wait_for_idle(gt); +} + +void intel_gt_suspend_prepare(struct intel_gt *gt) +{ + user_forcewake(gt, true); + wait_for_suspend(gt); + + intel_uc_suspend(>->uc); +} + +static suspend_state_t pm_suspend_target(void) +{ +#if IS_ENABLED(CONFIG_SUSPEND) && IS_ENABLED(CONFIG_PM_SLEEP) + return pm_suspend_target_state; +#else + return PM_SUSPEND_TO_IDLE; +#endif +} + +void intel_gt_suspend_late(struct intel_gt *gt) +{ + intel_wakeref_t wakeref; + + /* We expect to be idle already; but also want to be independent */ + wait_for_suspend(gt); + + /* + * On disabling the device, we want to turn off HW access to memory + * that we no longer own. + * + * However, not all suspend-states disable the device. S0 (s2idle) + * is effectively runtime-suspend, the device is left powered on + * but needs to be put into a low power state. We need to keep + * powermanagement enabled, but we also retain system state and so + * it remains safe to keep on using our allocated memory. + */ + if (pm_suspend_target() == PM_SUSPEND_TO_IDLE) + return; + + with_intel_runtime_pm(gt->uncore->rpm, wakeref) { + intel_rps_disable(>->rps); + intel_rc6_disable(>->rc6); + intel_llc_disable(>->llc); + } + + intel_gt_sanitize(gt, false); + + GEM_TRACE("\n"); +} + void intel_gt_runtime_suspend(struct intel_gt *gt) { intel_uc_runtime_suspend(>->uc); + + GEM_TRACE("\n"); } int intel_gt_runtime_resume(struct intel_gt *gt) { + GEM_TRACE("\n"); + intel_gt_init_swizzling(gt); return intel_uc_runtime_resume(>->uc); } + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "selftest_gt_pm.c" +#endif diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.h b/drivers/gpu/drm/i915/gt/intel_gt_pm.h index fb39d99cd6ee..990efc27a4e4 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_pm.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.h @@ -12,11 +12,6 @@ #include "intel_gt_types.h" #include "intel_wakeref.h" -enum { - INTEL_GT_UNPARK, - INTEL_GT_PARK, -}; - static inline bool intel_gt_pm_is_awake(const struct intel_gt *gt) { return intel_wakeref_is_active(>->wakeref); @@ -37,16 +32,32 @@ static inline void intel_gt_pm_put(struct intel_gt *gt) intel_wakeref_put(>->wakeref); } +static inline void intel_gt_pm_put_async(struct intel_gt *gt) +{ + intel_wakeref_put_async(>->wakeref); +} + static inline int intel_gt_pm_wait_for_idle(struct intel_gt *gt) { return intel_wakeref_wait_for_idle(>->wakeref); } void intel_gt_pm_init_early(struct intel_gt *gt); +void intel_gt_pm_init(struct intel_gt *gt); +void intel_gt_pm_fini(struct intel_gt *gt); void intel_gt_sanitize(struct intel_gt *gt, bool force); + +void intel_gt_suspend_prepare(struct intel_gt *gt); +void intel_gt_suspend_late(struct intel_gt *gt); int intel_gt_resume(struct intel_gt *gt); + void intel_gt_runtime_suspend(struct intel_gt *gt); int intel_gt_runtime_resume(struct intel_gt *gt); +static inline bool is_mock_gt(const struct intel_gt *gt) +{ + return I915_SELFTEST_ONLY(gt->awake == -ENODEV); +} + #endif /* INTEL_GT_PM_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_gt_requests.c b/drivers/gpu/drm/i915/gt/intel_gt_requests.c new file mode 100644 index 000000000000..3dc13ecf41bf --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_gt_requests.c @@ -0,0 +1,212 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include <linux/workqueue.h> + +#include "i915_drv.h" /* for_each_engine() */ +#include "i915_request.h" +#include "intel_gt.h" +#include "intel_gt_pm.h" +#include "intel_gt_requests.h" +#include "intel_timeline.h" + +static void retire_requests(struct intel_timeline *tl) +{ + struct i915_request *rq, *rn; + + list_for_each_entry_safe(rq, rn, &tl->requests, link) + if (!i915_request_retire(rq)) + break; +} + +static void flush_submission(struct intel_gt *gt) +{ + struct intel_engine_cs *engine; + enum intel_engine_id id; + + for_each_engine(engine, gt, id) + intel_engine_flush_submission(engine); +} + +static void engine_retire(struct work_struct *work) +{ + struct intel_engine_cs *engine = + container_of(work, typeof(*engine), retire_work); + struct intel_timeline *tl = xchg(&engine->retire, NULL); + + do { + struct intel_timeline *next = xchg(&tl->retire, NULL); + + /* + * Our goal here is to retire _idle_ timelines as soon as + * possible (as they are idle, we do not expect userspace + * to be cleaning up anytime soon). + * + * If the timeline is currently locked, either it is being + * retired elsewhere or about to be! + */ + if (mutex_trylock(&tl->mutex)) { + retire_requests(tl); + mutex_unlock(&tl->mutex); + } + intel_timeline_put(tl); + + GEM_BUG_ON(!next); + tl = ptr_mask_bits(next, 1); + } while (tl); +} + +static bool add_retire(struct intel_engine_cs *engine, + struct intel_timeline *tl) +{ + struct intel_timeline *first; + + /* + * We open-code a llist here to include the additional tag [BIT(0)] + * so that we know when the timeline is already on a + * retirement queue: either this engine or another. + * + * However, we rely on that a timeline can only be active on a single + * engine at any one time and that add_retire() is called before the + * engine releases the timeline and transferred to another to retire. + */ + + if (READ_ONCE(tl->retire)) /* already queued */ + return false; + + intel_timeline_get(tl); + first = READ_ONCE(engine->retire); + do + tl->retire = ptr_pack_bits(first, 1, 1); + while (!try_cmpxchg(&engine->retire, &first, tl)); + + return !first; +} + +void intel_engine_add_retire(struct intel_engine_cs *engine, + struct intel_timeline *tl) +{ + if (add_retire(engine, tl)) + schedule_work(&engine->retire_work); +} + +void intel_engine_init_retire(struct intel_engine_cs *engine) +{ + INIT_WORK(&engine->retire_work, engine_retire); +} + +void intel_engine_fini_retire(struct intel_engine_cs *engine) +{ + flush_work(&engine->retire_work); + GEM_BUG_ON(engine->retire); +} + +long intel_gt_retire_requests_timeout(struct intel_gt *gt, long timeout) +{ + struct intel_gt_timelines *timelines = >->timelines; + struct intel_timeline *tl, *tn; + unsigned long active_count = 0; + unsigned long flags; + bool interruptible; + LIST_HEAD(free); + + interruptible = true; + if (unlikely(timeout < 0)) + timeout = -timeout, interruptible = false; + + flush_submission(gt); /* kick the ksoftirqd tasklets */ + + spin_lock_irqsave(&timelines->lock, flags); + list_for_each_entry_safe(tl, tn, &timelines->active_list, link) { + if (!mutex_trylock(&tl->mutex)) { + active_count++; /* report busy to caller, try again? */ + continue; + } + + intel_timeline_get(tl); + GEM_BUG_ON(!atomic_read(&tl->active_count)); + atomic_inc(&tl->active_count); /* pin the list element */ + spin_unlock_irqrestore(&timelines->lock, flags); + + if (timeout > 0) { + struct dma_fence *fence; + + fence = i915_active_fence_get(&tl->last_request); + if (fence) { + timeout = dma_fence_wait_timeout(fence, + interruptible, + timeout); + dma_fence_put(fence); + } + } + + retire_requests(tl); + + spin_lock_irqsave(&timelines->lock, flags); + + /* Resume iteration after dropping lock */ + list_safe_reset_next(tl, tn, link); + if (atomic_dec_and_test(&tl->active_count)) + list_del(&tl->link); + else + active_count += !!rcu_access_pointer(tl->last_request.fence); + + mutex_unlock(&tl->mutex); + + /* Defer the final release to after the spinlock */ + if (refcount_dec_and_test(&tl->kref.refcount)) { + GEM_BUG_ON(atomic_read(&tl->active_count)); + list_add(&tl->link, &free); + } + } + spin_unlock_irqrestore(&timelines->lock, flags); + + list_for_each_entry_safe(tl, tn, &free, link) + __intel_timeline_free(&tl->kref); + + return active_count ? timeout : 0; +} + +int intel_gt_wait_for_idle(struct intel_gt *gt, long timeout) +{ + /* If the device is asleep, we have no requests outstanding */ + if (!intel_gt_pm_is_awake(gt)) + return 0; + + while ((timeout = intel_gt_retire_requests_timeout(gt, timeout)) > 0) { + cond_resched(); + if (signal_pending(current)) + return -EINTR; + } + + return timeout; +} + +static void retire_work_handler(struct work_struct *work) +{ + struct intel_gt *gt = + container_of(work, typeof(*gt), requests.retire_work.work); + + intel_gt_retire_requests(gt); + schedule_delayed_work(>->requests.retire_work, + round_jiffies_up_relative(HZ)); +} + +void intel_gt_init_requests(struct intel_gt *gt) +{ + INIT_DELAYED_WORK(>->requests.retire_work, retire_work_handler); +} + +void intel_gt_park_requests(struct intel_gt *gt) +{ + cancel_delayed_work(>->requests.retire_work); +} + +void intel_gt_unpark_requests(struct intel_gt *gt) +{ + schedule_delayed_work(>->requests.retire_work, + round_jiffies_up_relative(HZ)); +} diff --git a/drivers/gpu/drm/i915/gt/intel_gt_requests.h b/drivers/gpu/drm/i915/gt/intel_gt_requests.h new file mode 100644 index 000000000000..d626fb115386 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_gt_requests.h @@ -0,0 +1,31 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_GT_REQUESTS_H +#define INTEL_GT_REQUESTS_H + +struct intel_engine_cs; +struct intel_gt; +struct intel_timeline; + +long intel_gt_retire_requests_timeout(struct intel_gt *gt, long timeout); +static inline void intel_gt_retire_requests(struct intel_gt *gt) +{ + intel_gt_retire_requests_timeout(gt, 0); +} + +void intel_engine_init_retire(struct intel_engine_cs *engine); +void intel_engine_add_retire(struct intel_engine_cs *engine, + struct intel_timeline *tl); +void intel_engine_fini_retire(struct intel_engine_cs *engine); + +int intel_gt_wait_for_idle(struct intel_gt *gt, long timeout); + +void intel_gt_init_requests(struct intel_gt *gt); +void intel_gt_park_requests(struct intel_gt *gt); +void intel_gt_unpark_requests(struct intel_gt *gt); + +#endif /* INTEL_GT_REQUESTS_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_gt_types.h b/drivers/gpu/drm/i915/gt/intel_gt_types.h index dc295c196d11..d4e14dbd172e 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_types.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h @@ -17,7 +17,10 @@ #include "i915_vma.h" #include "intel_engine_types.h" +#include "intel_llc_types.h" #include "intel_reset_types.h" +#include "intel_rc6_types.h" +#include "intel_rps_types.h" #include "intel_wakeref.h" struct drm_i915_private; @@ -25,14 +28,6 @@ struct i915_ggtt; struct intel_engine_cs; struct intel_uncore; -struct intel_hangcheck { - /* For hangcheck timer */ -#define DRM_I915_HANGCHECK_PERIOD 1500 /* in ms */ -#define DRM_I915_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_I915_HANGCHECK_PERIOD) - - struct delayed_work work; -}; - struct intel_gt { struct drm_i915_private *i915; struct intel_uncore *uncore; @@ -49,12 +44,23 @@ struct intel_gt { struct list_head hwsp_free_list; } timelines; + struct intel_gt_requests { + /** + * We leave the user IRQ off as much as possible, + * but this means that requests will finish and never + * be retired once the system goes idle. Set a timer to + * fire periodically while the ring is running. When it + * fires, go retire requests. + */ + struct delayed_work retire_work; + } requests; + struct intel_wakeref wakeref; + atomic_t user_wakeref; struct list_head closed_vma; spinlock_t closed_lock; /* guards the list of closed_vma */ - struct intel_hangcheck hangcheck; struct intel_reset reset; /** @@ -66,7 +72,9 @@ struct intel_gt { */ intel_wakeref_t awake; - struct blocking_notifier_head pm_notifications; + struct intel_llc llc; + struct intel_rc6 rc6; + struct intel_rps rps; ktime_t last_init_time; @@ -89,14 +97,16 @@ enum intel_gt_scratch_field { INTEL_GT_SCRATCH_FIELD_DEFAULT = 0, /* 8 bytes */ - INTEL_GT_SCRATCH_FIELD_CLEAR_SLM_WA = 128, - - /* 8 bytes */ INTEL_GT_SCRATCH_FIELD_RENDER_FLUSH = 128, /* 8 bytes */ INTEL_GT_SCRATCH_FIELD_COHERENTL3_WA = 256, + /* 6 * 8 bytes */ + INTEL_GT_SCRATCH_FIELD_PERF_CS_GPR = 2048, + + /* 4 bytes */ + INTEL_GT_SCRATCH_FIELD_PERF_PREDICATE_RESULT_1 = 2096, }; #endif /* __INTEL_GT_TYPES_H__ */ diff --git a/drivers/gpu/drm/i915/gt/intel_hangcheck.c b/drivers/gpu/drm/i915/gt/intel_hangcheck.c deleted file mode 100644 index 05d042cdefe2..000000000000 --- a/drivers/gpu/drm/i915/gt/intel_hangcheck.c +++ /dev/null @@ -1,360 +0,0 @@ -/* - * Copyright © 2016 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS - * IN THE SOFTWARE. - * - */ - -#include "i915_drv.h" -#include "intel_engine.h" -#include "intel_gt.h" -#include "intel_reset.h" - -struct hangcheck { - u64 acthd; - u32 ring; - u32 head; - enum intel_engine_hangcheck_action action; - unsigned long action_timestamp; - int deadlock; - struct intel_instdone instdone; - bool wedged:1; - bool stalled:1; -}; - -static bool instdone_unchanged(u32 current_instdone, u32 *old_instdone) -{ - u32 tmp = current_instdone | *old_instdone; - bool unchanged; - - unchanged = tmp == *old_instdone; - *old_instdone |= tmp; - - return unchanged; -} - -static bool subunits_stuck(struct intel_engine_cs *engine) -{ - struct drm_i915_private *dev_priv = engine->i915; - struct intel_instdone instdone; - struct intel_instdone *accu_instdone = &engine->hangcheck.instdone; - bool stuck; - int slice; - int subslice; - - intel_engine_get_instdone(engine, &instdone); - - /* There might be unstable subunit states even when - * actual head is not moving. Filter out the unstable ones by - * accumulating the undone -> done transitions and only - * consider those as progress. - */ - stuck = instdone_unchanged(instdone.instdone, - &accu_instdone->instdone); - stuck &= instdone_unchanged(instdone.slice_common, - &accu_instdone->slice_common); - - for_each_instdone_slice_subslice(dev_priv, slice, subslice) { - stuck &= instdone_unchanged(instdone.sampler[slice][subslice], - &accu_instdone->sampler[slice][subslice]); - stuck &= instdone_unchanged(instdone.row[slice][subslice], - &accu_instdone->row[slice][subslice]); - } - - return stuck; -} - -static enum intel_engine_hangcheck_action -head_stuck(struct intel_engine_cs *engine, u64 acthd) -{ - if (acthd != engine->hangcheck.acthd) { - - /* Clear subunit states on head movement */ - memset(&engine->hangcheck.instdone, 0, - sizeof(engine->hangcheck.instdone)); - - return ENGINE_ACTIVE_HEAD; - } - - if (!subunits_stuck(engine)) - return ENGINE_ACTIVE_SUBUNITS; - - return ENGINE_DEAD; -} - -static enum intel_engine_hangcheck_action -engine_stuck(struct intel_engine_cs *engine, u64 acthd) -{ - enum intel_engine_hangcheck_action ha; - u32 tmp; - - ha = head_stuck(engine, acthd); - if (ha != ENGINE_DEAD) - return ha; - - if (IS_GEN(engine->i915, 2)) - return ENGINE_DEAD; - - /* Is the chip hanging on a WAIT_FOR_EVENT? - * If so we can simply poke the RB_WAIT bit - * and break the hang. This should work on - * all but the second generation chipsets. - */ - tmp = ENGINE_READ(engine, RING_CTL); - if (tmp & RING_WAIT) { - intel_gt_handle_error(engine->gt, engine->mask, 0, - "stuck wait on %s", engine->name); - ENGINE_WRITE(engine, RING_CTL, tmp); - return ENGINE_WAIT_KICK; - } - - return ENGINE_DEAD; -} - -static void hangcheck_load_sample(struct intel_engine_cs *engine, - struct hangcheck *hc) -{ - hc->acthd = intel_engine_get_active_head(engine); - hc->ring = ENGINE_READ(engine, RING_START); - hc->head = ENGINE_READ(engine, RING_HEAD); -} - -static void hangcheck_store_sample(struct intel_engine_cs *engine, - const struct hangcheck *hc) -{ - engine->hangcheck.acthd = hc->acthd; - engine->hangcheck.last_ring = hc->ring; - engine->hangcheck.last_head = hc->head; -} - -static enum intel_engine_hangcheck_action -hangcheck_get_action(struct intel_engine_cs *engine, - const struct hangcheck *hc) -{ - if (intel_engine_is_idle(engine)) - return ENGINE_IDLE; - - if (engine->hangcheck.last_ring != hc->ring) - return ENGINE_ACTIVE_SEQNO; - - if (engine->hangcheck.last_head != hc->head) - return ENGINE_ACTIVE_SEQNO; - - return engine_stuck(engine, hc->acthd); -} - -static void hangcheck_accumulate_sample(struct intel_engine_cs *engine, - struct hangcheck *hc) -{ - unsigned long timeout = I915_ENGINE_DEAD_TIMEOUT; - - hc->action = hangcheck_get_action(engine, hc); - - /* We always increment the progress - * if the engine is busy and still processing - * the same request, so that no single request - * can run indefinitely (such as a chain of - * batches). The only time we do not increment - * the hangcheck score on this ring, if this - * engine is in a legitimate wait for another - * engine. In that case the waiting engine is a - * victim and we want to be sure we catch the - * right culprit. Then every time we do kick - * the ring, make it as a progress as the seqno - * advancement might ensure and if not, it - * will catch the hanging engine. - */ - - switch (hc->action) { - case ENGINE_IDLE: - case ENGINE_ACTIVE_SEQNO: - /* Clear head and subunit states on seqno movement */ - hc->acthd = 0; - - memset(&engine->hangcheck.instdone, 0, - sizeof(engine->hangcheck.instdone)); - - /* Intentional fall through */ - case ENGINE_WAIT_KICK: - case ENGINE_WAIT: - engine->hangcheck.action_timestamp = jiffies; - break; - - case ENGINE_ACTIVE_HEAD: - case ENGINE_ACTIVE_SUBUNITS: - /* - * Seqno stuck with still active engine gets leeway, - * in hopes that it is just a long shader. - */ - timeout = I915_SEQNO_DEAD_TIMEOUT; - break; - - case ENGINE_DEAD: - break; - - default: - MISSING_CASE(hc->action); - } - - hc->stalled = time_after(jiffies, - engine->hangcheck.action_timestamp + timeout); - hc->wedged = time_after(jiffies, - engine->hangcheck.action_timestamp + - I915_ENGINE_WEDGED_TIMEOUT); -} - -static void hangcheck_declare_hang(struct intel_gt *gt, - intel_engine_mask_t hung, - intel_engine_mask_t stuck) -{ - struct intel_engine_cs *engine; - intel_engine_mask_t tmp; - char msg[80]; - int len; - - /* If some rings hung but others were still busy, only - * blame the hanging rings in the synopsis. - */ - if (stuck != hung) - hung &= ~stuck; - len = scnprintf(msg, sizeof(msg), - "%s on ", stuck == hung ? "no progress" : "hang"); - for_each_engine_masked(engine, gt->i915, hung, tmp) - len += scnprintf(msg + len, sizeof(msg) - len, - "%s, ", engine->name); - msg[len-2] = '\0'; - - return intel_gt_handle_error(gt, hung, I915_ERROR_CAPTURE, "%s", msg); -} - -/* - * This is called when the chip hasn't reported back with completed - * batchbuffers in a long time. We keep track per ring seqno progress and - * if there are no progress, hangcheck score for that ring is increased. - * Further, acthd is inspected to see if the ring is stuck. On stuck case - * we kick the ring. If we see no progress on three subsequent calls - * we assume chip is wedged and try to fix it by resetting the chip. - */ -static void hangcheck_elapsed(struct work_struct *work) -{ - struct intel_gt *gt = - container_of(work, typeof(*gt), hangcheck.work.work); - intel_engine_mask_t hung = 0, stuck = 0, wedged = 0; - struct intel_engine_cs *engine; - enum intel_engine_id id; - intel_wakeref_t wakeref; - - if (!i915_modparams.enable_hangcheck) - return; - - if (!READ_ONCE(gt->awake)) - return; - - if (intel_gt_is_wedged(gt)) - return; - - wakeref = intel_runtime_pm_get_if_in_use(>->i915->runtime_pm); - if (!wakeref) - return; - - /* As enabling the GPU requires fairly extensive mmio access, - * periodically arm the mmio checker to see if we are triggering - * any invalid access. - */ - intel_uncore_arm_unclaimed_mmio_detection(gt->uncore); - - for_each_engine(engine, gt->i915, id) { - struct hangcheck hc; - - intel_engine_signal_breadcrumbs(engine); - - hangcheck_load_sample(engine, &hc); - hangcheck_accumulate_sample(engine, &hc); - hangcheck_store_sample(engine, &hc); - - if (hc.stalled) { - hung |= engine->mask; - if (hc.action != ENGINE_DEAD) - stuck |= engine->mask; - } - - if (hc.wedged) - wedged |= engine->mask; - } - - if (GEM_SHOW_DEBUG() && (hung | stuck)) { - struct drm_printer p = drm_debug_printer("hangcheck"); - - for_each_engine(engine, gt->i915, id) { - if (intel_engine_is_idle(engine)) - continue; - - intel_engine_dump(engine, &p, "%s\n", engine->name); - } - } - - if (wedged) { - dev_err(gt->i915->drm.dev, - "GPU recovery timed out," - " cancelling all in-flight rendering.\n"); - GEM_TRACE_DUMP(); - intel_gt_set_wedged(gt); - } - - if (hung) - hangcheck_declare_hang(gt, hung, stuck); - - intel_runtime_pm_put(>->i915->runtime_pm, wakeref); - - /* Reset timer in case GPU hangs without another request being added */ - intel_gt_queue_hangcheck(gt); -} - -void intel_gt_queue_hangcheck(struct intel_gt *gt) -{ - unsigned long delay; - - if (unlikely(!i915_modparams.enable_hangcheck)) - return; - - /* - * Don't continually defer the hangcheck so that it is always run at - * least once after work has been scheduled on any ring. Otherwise, - * we will ignore a hung ring if a second ring is kept busy. - */ - - delay = round_jiffies_up_relative(DRM_I915_HANGCHECK_JIFFIES); - queue_delayed_work(system_long_wq, >->hangcheck.work, delay); -} - -void intel_engine_init_hangcheck(struct intel_engine_cs *engine) -{ - memset(&engine->hangcheck, 0, sizeof(engine->hangcheck)); - engine->hangcheck.action_timestamp = jiffies; -} - -void intel_gt_init_hangcheck(struct intel_gt *gt) -{ - INIT_DELAYED_WORK(>->hangcheck.work, hangcheck_elapsed); -} - -#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) -#include "selftest_hangcheck.c" -#endif diff --git a/drivers/gpu/drm/i915/gt/intel_llc.c b/drivers/gpu/drm/i915/gt/intel_llc.c new file mode 100644 index 000000000000..ceb785b75c25 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_llc.c @@ -0,0 +1,161 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include <linux/cpufreq.h> + +#include "i915_drv.h" +#include "intel_gt.h" +#include "intel_llc.h" +#include "intel_sideband.h" + +struct ia_constants { + unsigned int min_gpu_freq; + unsigned int max_gpu_freq; + + unsigned int min_ring_freq; + unsigned int max_ia_freq; +}; + +static struct intel_gt *llc_to_gt(struct intel_llc *llc) +{ + return container_of(llc, struct intel_gt, llc); +} + +static unsigned int cpu_max_MHz(void) +{ + struct cpufreq_policy *policy; + unsigned int max_khz; + + policy = cpufreq_cpu_get(0); + if (policy) { + max_khz = policy->cpuinfo.max_freq; + cpufreq_cpu_put(policy); + } else { + /* + * Default to measured freq if none found, PCU will ensure we + * don't go over + */ + max_khz = tsc_khz; + } + + return max_khz / 1000; +} + +static bool get_ia_constants(struct intel_llc *llc, + struct ia_constants *consts) +{ + struct drm_i915_private *i915 = llc_to_gt(llc)->i915; + struct intel_rps *rps = &llc_to_gt(llc)->rps; + + if (rps->max_freq <= rps->min_freq) + return false; + + consts->max_ia_freq = cpu_max_MHz(); + + consts->min_ring_freq = + intel_uncore_read(llc_to_gt(llc)->uncore, DCLK) & 0xf; + /* convert DDR frequency from units of 266.6MHz to bandwidth */ + consts->min_ring_freq = mult_frac(consts->min_ring_freq, 8, 3); + + consts->min_gpu_freq = rps->min_freq; + consts->max_gpu_freq = rps->max_freq; + if (INTEL_GEN(i915) >= 9) { + /* Convert GT frequency to 50 HZ units */ + consts->min_gpu_freq /= GEN9_FREQ_SCALER; + consts->max_gpu_freq /= GEN9_FREQ_SCALER; + } + + return true; +} + +static void calc_ia_freq(struct intel_llc *llc, + unsigned int gpu_freq, + const struct ia_constants *consts, + unsigned int *out_ia_freq, + unsigned int *out_ring_freq) +{ + struct drm_i915_private *i915 = llc_to_gt(llc)->i915; + const int diff = consts->max_gpu_freq - gpu_freq; + unsigned int ia_freq = 0, ring_freq = 0; + + if (INTEL_GEN(i915) >= 9) { + /* + * ring_freq = 2 * GT. ring_freq is in 100MHz units + * No floor required for ring frequency on SKL. + */ + ring_freq = gpu_freq; + } else if (INTEL_GEN(i915) >= 8) { + /* max(2 * GT, DDR). NB: GT is 50MHz units */ + ring_freq = max(consts->min_ring_freq, gpu_freq); + } else if (IS_HASWELL(i915)) { + ring_freq = mult_frac(gpu_freq, 5, 4); + ring_freq = max(consts->min_ring_freq, ring_freq); + /* leave ia_freq as the default, chosen by cpufreq */ + } else { + const int min_freq = 15; + const int scale = 180; + + /* + * On older processors, there is no separate ring + * clock domain, so in order to boost the bandwidth + * of the ring, we need to upclock the CPU (ia_freq). + * + * For GPU frequencies less than 750MHz, + * just use the lowest ring freq. + */ + if (gpu_freq < min_freq) + ia_freq = 800; + else + ia_freq = consts->max_ia_freq - diff * scale / 2; + ia_freq = DIV_ROUND_CLOSEST(ia_freq, 100); + } + + *out_ia_freq = ia_freq; + *out_ring_freq = ring_freq; +} + +static void gen6_update_ring_freq(struct intel_llc *llc) +{ + struct drm_i915_private *i915 = llc_to_gt(llc)->i915; + struct ia_constants consts; + unsigned int gpu_freq; + + if (!get_ia_constants(llc, &consts)) + return; + + /* + * For each potential GPU frequency, load a ring frequency we'd like + * to use for memory access. We do this by specifying the IA frequency + * the PCU should use as a reference to determine the ring frequency. + */ + for (gpu_freq = consts.max_gpu_freq; + gpu_freq >= consts.min_gpu_freq; + gpu_freq--) { + unsigned int ia_freq, ring_freq; + + calc_ia_freq(llc, gpu_freq, &consts, &ia_freq, &ring_freq); + sandybridge_pcode_write(i915, + GEN6_PCODE_WRITE_MIN_FREQ_TABLE, + ia_freq << GEN6_PCODE_FREQ_IA_RATIO_SHIFT | + ring_freq << GEN6_PCODE_FREQ_RING_RATIO_SHIFT | + gpu_freq); + } +} + +void intel_llc_enable(struct intel_llc *llc) +{ + if (HAS_LLC(llc_to_gt(llc)->i915)) + gen6_update_ring_freq(llc); +} + +void intel_llc_disable(struct intel_llc *llc) +{ + /* Currently there is no HW configuration to be done to disable. */ +} + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "selftest_llc.c" +#endif diff --git a/drivers/gpu/drm/i915/gt/intel_llc.h b/drivers/gpu/drm/i915/gt/intel_llc.h new file mode 100644 index 000000000000..ef09a890d2b7 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_llc.h @@ -0,0 +1,15 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_LLC_H +#define INTEL_LLC_H + +struct intel_llc; + +void intel_llc_enable(struct intel_llc *llc); +void intel_llc_disable(struct intel_llc *llc); + +#endif /* INTEL_LLC_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_llc_types.h b/drivers/gpu/drm/i915/gt/intel_llc_types.h new file mode 100644 index 000000000000..ecad4687b930 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_llc_types.h @@ -0,0 +1,13 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_LLC_TYPES_H +#define INTEL_LLC_TYPES_H + +struct intel_llc { +}; + +#endif /* INTEL_LLC_TYPES_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c index d42584439f51..9fdefbdc3546 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.c +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c @@ -142,9 +142,11 @@ #include "intel_engine_pm.h" #include "intel_gt.h" #include "intel_gt_pm.h" +#include "intel_gt_requests.h" #include "intel_lrc_reg.h" #include "intel_mocs.h" #include "intel_reset.h" +#include "intel_ring.h" #include "intel_workarounds.h" #define RING_EXECLIST_QFULL (1 << 0x2) @@ -230,9 +232,41 @@ static int __execlists_context_alloc(struct intel_context *ce, struct intel_engine_cs *engine); static void execlists_init_reg_state(u32 *reg_state, - struct intel_context *ce, - struct intel_engine_cs *engine, - struct intel_ring *ring); + const struct intel_context *ce, + const struct intel_engine_cs *engine, + const struct intel_ring *ring, + bool close); +static void +__execlists_update_reg_state(const struct intel_context *ce, + const struct intel_engine_cs *engine); + +static void mark_eio(struct i915_request *rq) +{ + if (i915_request_completed(rq)) + return; + + GEM_BUG_ON(i915_request_signaled(rq)); + + dma_fence_set_error(&rq->fence, -EIO); + i915_request_mark_complete(rq); +} + +static struct i915_request * +active_request(const struct intel_timeline * const tl, struct i915_request *rq) +{ + struct i915_request *active = rq; + + rcu_read_lock(); + list_for_each_entry_continue_reverse(rq, &tl->requests, link) { + if (i915_request_completed(rq)) + break; + + active = rq; + } + rcu_read_unlock(); + + return active; +} static inline u32 intel_hws_preempt_address(struct intel_engine_cs *engine) { @@ -330,10 +364,15 @@ static inline bool need_preempt(const struct intel_engine_cs *engine, * However, the priority hint is a mere hint that we may need to * preempt. If that hint is stale or we may be trying to preempt * ourselves, ignore the request. + * + * More naturally we would write + * prio >= max(0, last); + * except that we wish to prevent triggering preemption at the same + * priority level: the task that is running should remain running + * to preserve FIFO ordering of dependencies. */ - last_prio = effective_prio(rq); - if (!i915_scheduler_need_preempt(engine->execlists.queue_priority_hint, - last_prio)) + last_prio = max(effective_prio(rq), I915_PRIORITY_NORMAL - 1); + if (engine->execlists.queue_priority_hint <= last_prio) return false; /* @@ -422,12 +461,8 @@ assert_priority_queue(const struct i915_request *prev, static u64 lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine) { - struct i915_gem_context *ctx = ce->gem_context; u64 desc; - BUILD_BUG_ON(MAX_CONTEXT_HW_ID > (BIT(GEN8_CTX_ID_WIDTH))); - BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > (BIT(GEN11_SW_CTX_ID_WIDTH))); - desc = INTEL_LEGACY_32B_CONTEXT; if (i915_vm_is_4lvl(ce->vm)) desc = INTEL_LEGACY_64B_CONTEXT; @@ -437,33 +472,379 @@ lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine) if (IS_GEN(engine->i915, 8)) desc |= GEN8_CTX_L3LLC_COHERENT; - desc |= i915_ggtt_offset(ce->state) + LRC_HEADER_PAGES * PAGE_SIZE; - /* bits 12-31 */ + desc |= i915_ggtt_offset(ce->state); /* bits 12-31 */ /* * The following 32bits are copied into the OA reports (dword 2). * Consider updating oa_get_render_ctx_id in i915_perf.c when changing * anything below. */ if (INTEL_GEN(engine->i915) >= 11) { - GEM_BUG_ON(ctx->hw_id >= BIT(GEN11_SW_CTX_ID_WIDTH)); - desc |= (u64)ctx->hw_id << GEN11_SW_CTX_ID_SHIFT; - /* bits 37-47 */ - desc |= (u64)engine->instance << GEN11_ENGINE_INSTANCE_SHIFT; /* bits 48-53 */ - /* TODO: decide what to do with SW counter (bits 55-60) */ - desc |= (u64)engine->class << GEN11_ENGINE_CLASS_SHIFT; /* bits 61-63 */ - } else { - GEM_BUG_ON(ctx->hw_id >= BIT(GEN8_CTX_ID_WIDTH)); - desc |= (u64)ctx->hw_id << GEN8_CTX_ID_SHIFT; /* bits 32-52 */ } return desc; } +static u32 *set_offsets(u32 *regs, + const u8 *data, + const struct intel_engine_cs *engine) +#define NOP(x) (BIT(7) | (x)) +#define LRI(count, flags) ((flags) << 6 | (count)) +#define POSTED BIT(0) +#define REG(x) (((x) >> 2) | BUILD_BUG_ON_ZERO(x >= 0x200)) +#define REG16(x) \ + (((x) >> 9) | BIT(7) | BUILD_BUG_ON_ZERO(x >= 0x10000)), \ + (((x) >> 2) & 0x7f) +#define END() 0 +{ + const u32 base = engine->mmio_base; + + while (*data) { + u8 count, flags; + + if (*data & BIT(7)) { /* skip */ + regs += *data++ & ~BIT(7); + continue; + } + + count = *data & 0x3f; + flags = *data >> 6; + data++; + + *regs = MI_LOAD_REGISTER_IMM(count); + if (flags & POSTED) + *regs |= MI_LRI_FORCE_POSTED; + if (INTEL_GEN(engine->i915) >= 11) + *regs |= MI_LRI_CS_MMIO; + regs++; + + GEM_BUG_ON(!count); + do { + u32 offset = 0; + u8 v; + + do { + v = *data++; + offset <<= 7; + offset |= v & ~BIT(7); + } while (v & BIT(7)); + + *regs = base + (offset << 2); + regs += 2; + } while (--count); + } + + return regs; +} + +static const u8 gen8_xcs_offsets[] = { + NOP(1), + LRI(11, 0), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x11c), + REG(0x114), + REG(0x118), + + NOP(9), + LRI(9, 0), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + NOP(13), + LRI(2, 0), + REG16(0x200), + REG(0x028), + + END(), +}; + +static const u8 gen9_xcs_offsets[] = { + NOP(1), + LRI(14, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x11c), + REG(0x114), + REG(0x118), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + + NOP(3), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + NOP(13), + LRI(1, POSTED), + REG16(0x200), + + NOP(13), + LRI(44, POSTED), + REG(0x028), + REG(0x09c), + REG(0x0c0), + REG(0x178), + REG(0x17c), + REG16(0x358), + REG(0x170), + REG(0x150), + REG(0x154), + REG(0x158), + REG16(0x41c), + REG16(0x600), + REG16(0x604), + REG16(0x608), + REG16(0x60c), + REG16(0x610), + REG16(0x614), + REG16(0x618), + REG16(0x61c), + REG16(0x620), + REG16(0x624), + REG16(0x628), + REG16(0x62c), + REG16(0x630), + REG16(0x634), + REG16(0x638), + REG16(0x63c), + REG16(0x640), + REG16(0x644), + REG16(0x648), + REG16(0x64c), + REG16(0x650), + REG16(0x654), + REG16(0x658), + REG16(0x65c), + REG16(0x660), + REG16(0x664), + REG16(0x668), + REG16(0x66c), + REG16(0x670), + REG16(0x674), + REG16(0x678), + REG16(0x67c), + REG(0x068), + + END(), +}; + +static const u8 gen12_xcs_offsets[] = { + NOP(1), + LRI(13, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + REG(0x180), + REG16(0x2b4), + + NOP(5), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + END(), +}; + +static const u8 gen8_rcs_offsets[] = { + NOP(1), + LRI(14, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x11c), + REG(0x114), + REG(0x118), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + + NOP(3), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + NOP(13), + LRI(1, 0), + REG(0x0c8), + + END(), +}; + +static const u8 gen11_rcs_offsets[] = { + NOP(1), + LRI(15, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x11c), + REG(0x114), + REG(0x118), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + REG(0x180), + + NOP(1), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + LRI(1, POSTED), + REG(0x1b0), + + NOP(10), + LRI(1, 0), + REG(0x0c8), + + END(), +}; + +static const u8 gen12_rcs_offsets[] = { + NOP(1), + LRI(13, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + REG(0x180), + REG16(0x2b4), + + NOP(5), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + LRI(3, POSTED), + REG(0x1b0), + REG16(0x5a8), + REG16(0x5ac), + + NOP(6), + LRI(1, 0), + REG(0x0c8), + + END(), +}; + +#undef END +#undef REG16 +#undef REG +#undef LRI +#undef NOP + +static const u8 *reg_offsets(const struct intel_engine_cs *engine) +{ + /* + * The gen12+ lists only have the registers we program in the basic + * default state. We rely on the context image using relative + * addressing to automatic fixup the register state between the + * physical engines for virtual engine. + */ + GEM_BUG_ON(INTEL_GEN(engine->i915) >= 12 && + !intel_engine_has_relative_mmio(engine)); + + if (engine->class == RENDER_CLASS) { + if (INTEL_GEN(engine->i915) >= 12) + return gen12_rcs_offsets; + else if (INTEL_GEN(engine->i915) >= 11) + return gen11_rcs_offsets; + else + return gen8_rcs_offsets; + } else { + if (INTEL_GEN(engine->i915) >= 12) + return gen12_xcs_offsets; + else if (INTEL_GEN(engine->i915) >= 9) + return gen9_xcs_offsets; + else + return gen8_xcs_offsets; + } +} + static void unwind_wa_tail(struct i915_request *rq) { rq->tail = intel_ring_wrap(rq->ring, rq->wa_tail - WA_TAIL_BYTES); @@ -482,7 +863,6 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine) list_for_each_entry_safe_reverse(rq, rn, &engine->active.requests, sched.link) { - struct intel_engine_cs *owner; if (i915_request_completed(rq)) continue; /* XXX */ @@ -497,8 +877,7 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine) * engine so that it can be moved across onto another physical * engine as load dictates. */ - owner = rq->hw_context->engine; - if (likely(owner == engine)) { + if (likely(rq->execution_mask == engine->mask)) { GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID); if (rq_prio(rq) != prio) { prio = rq_prio(rq); @@ -509,6 +888,8 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine) list_move(&rq->sched.link, pl); active = rq; } else { + struct intel_engine_cs *owner = rq->hw_context->engine; + /* * Decouple the virtual breadcrumb before moving it * back to the virtual engine -- we don't want the @@ -518,7 +899,8 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine) */ if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &rq->fence.flags)) { - spin_lock(&rq->lock); + spin_lock_nested(&rq->lock, + SINGLE_DEPTH_NESTING); i915_request_cancel_breadcrumb(rq); spin_unlock(&rq->lock); } @@ -554,6 +936,114 @@ execlists_context_status_change(struct i915_request *rq, unsigned long status) status, rq); } +static void intel_engine_context_in(struct intel_engine_cs *engine) +{ + unsigned long flags; + + if (READ_ONCE(engine->stats.enabled) == 0) + return; + + write_seqlock_irqsave(&engine->stats.lock, flags); + + if (engine->stats.enabled > 0) { + if (engine->stats.active++ == 0) + engine->stats.start = ktime_get(); + GEM_BUG_ON(engine->stats.active == 0); + } + + write_sequnlock_irqrestore(&engine->stats.lock, flags); +} + +static void intel_engine_context_out(struct intel_engine_cs *engine) +{ + unsigned long flags; + + if (READ_ONCE(engine->stats.enabled) == 0) + return; + + write_seqlock_irqsave(&engine->stats.lock, flags); + + if (engine->stats.enabled > 0) { + ktime_t last; + + if (engine->stats.active && --engine->stats.active == 0) { + /* + * Decrement the active context count and in case GPU + * is now idle add up to the running total. + */ + last = ktime_sub(ktime_get(), engine->stats.start); + + engine->stats.total = ktime_add(engine->stats.total, + last); + } else if (engine->stats.active == 0) { + /* + * After turning on engine stats, context out might be + * the first event in which case we account from the + * time stats gathering was turned on. + */ + last = ktime_sub(ktime_get(), engine->stats.enabled_at); + + engine->stats.total = ktime_add(engine->stats.total, + last); + } + } + + write_sequnlock_irqrestore(&engine->stats.lock, flags); +} + +static void restore_default_state(struct intel_context *ce, + struct intel_engine_cs *engine) +{ + u32 *regs = ce->lrc_reg_state; + + if (engine->pinned_default_state) + memcpy(regs, /* skip restoring the vanilla PPHWSP */ + engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE, + engine->context_size - PAGE_SIZE); + + execlists_init_reg_state(regs, ce, engine, ce->ring, false); +} + +static void reset_active(struct i915_request *rq, + struct intel_engine_cs *engine) +{ + struct intel_context * const ce = rq->hw_context; + u32 head; + + /* + * The executing context has been cancelled. We want to prevent + * further execution along this context and propagate the error on + * to anything depending on its results. + * + * In __i915_request_submit(), we apply the -EIO and remove the + * requests' payloads for any banned requests. But first, we must + * rewind the context back to the start of the incomplete request so + * that we do not jump back into the middle of the batch. + * + * We preserve the breadcrumbs and semaphores of the incomplete + * requests so that inter-timeline dependencies (i.e other timelines) + * remain correctly ordered. And we defer to __i915_request_submit() + * so that all asynchronous waits are correctly handled. + */ + GEM_TRACE("%s(%s): { rq=%llx:%lld }\n", + __func__, engine->name, rq->fence.context, rq->fence.seqno); + + /* On resubmission of the active request, payload will be scrubbed */ + if (i915_request_completed(rq)) + head = rq->tail; + else + head = active_request(ce->timeline, rq)->head; + ce->ring->head = intel_ring_wrap(ce->ring, head); + intel_ring_update_space(ce->ring); + + /* Scrub the context image to prevent replaying the previous batch */ + restore_default_state(ce, engine); + __execlists_update_reg_state(ce, engine); + + /* We've switched away, so this should be a no-op, but intent matters */ + ce->lrc_desc |= CTX_DESC_FORCE_RESTORE; +} + static inline struct intel_engine_cs * __execlists_schedule_in(struct i915_request *rq) { @@ -562,6 +1052,21 @@ __execlists_schedule_in(struct i915_request *rq) intel_context_get(ce); + if (unlikely(i915_gem_context_is_banned(ce->gem_context))) + reset_active(rq, engine); + + if (ce->tag) { + /* Use a fixed tag for OA and friends */ + ce->lrc_desc |= (u64)ce->tag << 32; + } else { + /* We don't need a strict matching tag, just different values */ + ce->lrc_desc &= ~GENMASK_ULL(47, 37); + ce->lrc_desc |= + (u64)(engine->context_tag++ % NUM_CONTEXT_TAG) << + GEN11_SW_CTX_ID_SHIFT; + BUILD_BUG_ON(NUM_CONTEXT_TAG > GEN12_MAX_CONTEXT_HW_ID); + } + intel_gt_pm_get(engine->gt); execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN); intel_engine_context_in(engine); @@ -605,9 +1110,23 @@ __execlists_schedule_out(struct i915_request *rq, { struct intel_context * const ce = rq->hw_context; + /* + * NB process_csb() is not under the engine->active.lock and hence + * schedule_out can race with schedule_in meaning that we should + * refrain from doing non-trivial work here. + */ + + /* + * If we have just completed this context, the engine may now be + * idle and we want to re-enter powersaving. + */ + if (list_is_last(&rq->link, &ce->timeline->requests) && + i915_request_completed(rq)) + intel_engine_add_retire(engine, ce->timeline); + intel_engine_context_out(engine); execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT); - intel_gt_pm_put(engine->gt); + intel_gt_pm_put_async(engine->gt); /* * If this is part of a virtual engine, its next request may @@ -631,7 +1150,6 @@ execlists_schedule_out(struct i915_request *rq) struct intel_engine_cs *cur, *old; trace_i915_request_out(rq); - GEM_BUG_ON(intel_context_inflight(ce) != rq->engine); old = READ_ONCE(ce->inflight); do @@ -648,7 +1166,7 @@ static u64 execlists_update_context(const struct i915_request *rq) struct intel_context *ce = rq->hw_context; u64 desc; - ce->lrc_reg_state[CTX_RING_TAIL + 1] = + ce->lrc_reg_state[CTX_RING_TAIL] = intel_ring_set_tail(rq->ring, rq->tail); /* @@ -671,6 +1189,10 @@ static u64 execlists_update_context(const struct i915_request *rq) desc = ce->lrc_desc; ce->lrc_desc &= ~CTX_DESC_FORCE_RESTORE; + /* Wa_1607138340:tgl */ + if (IS_TGL_REVID(rq->i915, TGL_REVID_A0, TGL_REVID_A0)) + desc |= CTX_DESC_FORCE_RESTORE; + return desc; } @@ -693,6 +1215,9 @@ trace_ports(const struct intel_engine_execlists *execlists, const struct intel_engine_cs *engine = container_of(execlists, typeof(*engine), execlists); + if (!ports[0]) + return; + GEM_TRACE("%s: %s { %llx:%lld%s, %llx:%lld }\n", engine->name, msg, ports[0]->fence.context, @@ -713,25 +1238,45 @@ assert_pending_valid(const struct intel_engine_execlists *execlists, trace_ports(execlists, msg, execlists->pending); - if (!execlists->pending[0]) + if (!execlists->pending[0]) { + GEM_TRACE_ERR("Nothing pending for promotion!\n"); return false; + } - if (execlists->pending[execlists_num_ports(execlists)]) + if (execlists->pending[execlists_num_ports(execlists)]) { + GEM_TRACE_ERR("Excess pending[%d] for promotion!\n", + execlists_num_ports(execlists)); return false; + } for (port = execlists->pending; (rq = *port); port++) { - if (ce == rq->hw_context) + if (ce == rq->hw_context) { + GEM_TRACE_ERR("Duplicate context in pending[%zd]\n", + port - execlists->pending); return false; + } ce = rq->hw_context; if (i915_request_completed(rq)) continue; - if (i915_active_is_idle(&ce->active)) + if (i915_active_is_idle(&ce->active)) { + GEM_TRACE_ERR("Inactive context in pending[%zd]\n", + port - execlists->pending); return false; + } - if (!i915_vma_is_pinned(ce->state)) + if (!i915_vma_is_pinned(ce->state)) { + GEM_TRACE_ERR("Unpinned context in pending[%zd]\n", + port - execlists->pending); return false; + } + + if (!i915_vma_is_pinned(ce->ring->vma)) { + GEM_TRACE_ERR("Unpinned ringbuffer in pending[%zd]\n", + port - execlists->pending); + return false; + } } return ce; @@ -797,6 +1342,21 @@ static bool can_merge_rq(const struct i915_request *prev, GEM_BUG_ON(prev == next); GEM_BUG_ON(!assert_priority_queue(prev, next)); + /* + * We do not submit known completed requests. Therefore if the next + * request is already completed, we can pretend to merge it in + * with the previous context (and we will skip updating the ELSP + * and tracking). Thus hopefully keeping the ELSP full with active + * contexts, despite the best efforts of preempt-to-busy to confuse + * us. + */ + if (i915_request_completed(next)) + return true; + + if (unlikely((prev->flags ^ next->flags) & + (I915_REQUEST_NOPREEMPT | I915_REQUEST_SENTINEL))) + return false; + if (!can_merge_ctx(prev->hw_context, next->hw_context)) return false; @@ -806,47 +1366,7 @@ static bool can_merge_rq(const struct i915_request *prev, static void virtual_update_register_offsets(u32 *regs, struct intel_engine_cs *engine) { - u32 base = engine->mmio_base; - - /* Must match execlists_init_reg_state()! */ - - regs[CTX_CONTEXT_CONTROL] = - i915_mmio_reg_offset(RING_CONTEXT_CONTROL(base)); - regs[CTX_RING_HEAD] = i915_mmio_reg_offset(RING_HEAD(base)); - regs[CTX_RING_TAIL] = i915_mmio_reg_offset(RING_TAIL(base)); - regs[CTX_RING_BUFFER_START] = i915_mmio_reg_offset(RING_START(base)); - regs[CTX_RING_BUFFER_CONTROL] = i915_mmio_reg_offset(RING_CTL(base)); - - regs[CTX_BB_HEAD_U] = i915_mmio_reg_offset(RING_BBADDR_UDW(base)); - regs[CTX_BB_HEAD_L] = i915_mmio_reg_offset(RING_BBADDR(base)); - regs[CTX_BB_STATE] = i915_mmio_reg_offset(RING_BBSTATE(base)); - regs[CTX_SECOND_BB_HEAD_U] = - i915_mmio_reg_offset(RING_SBBADDR_UDW(base)); - regs[CTX_SECOND_BB_HEAD_L] = i915_mmio_reg_offset(RING_SBBADDR(base)); - regs[CTX_SECOND_BB_STATE] = i915_mmio_reg_offset(RING_SBBSTATE(base)); - - regs[CTX_CTX_TIMESTAMP] = - i915_mmio_reg_offset(RING_CTX_TIMESTAMP(base)); - regs[CTX_PDP3_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 3)); - regs[CTX_PDP3_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 3)); - regs[CTX_PDP2_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 2)); - regs[CTX_PDP2_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 2)); - regs[CTX_PDP1_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 1)); - regs[CTX_PDP1_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 1)); - regs[CTX_PDP0_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 0)); - regs[CTX_PDP0_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 0)); - - if (engine->class == RENDER_CLASS) { - regs[CTX_RCS_INDIRECT_CTX] = - i915_mmio_reg_offset(RING_INDIRECT_CTX(base)); - regs[CTX_RCS_INDIRECT_CTX_OFFSET] = - i915_mmio_reg_offset(RING_INDIRECT_CTX_OFFSET(base)); - regs[CTX_BB_PER_CTX_PTR] = - i915_mmio_reg_offset(RING_BB_PER_CTX_PTR(base)); - - regs[CTX_R_PWR_CLK_STATE] = - i915_mmio_reg_offset(GEN8_R_PWR_CLK_STATE); - } + set_offsets(regs, reg_offsets(engine), engine); } static bool virtual_matches(const struct virtual_engine *ve, @@ -893,7 +1413,7 @@ static void virtual_xfer_breadcrumbs(struct virtual_engine *ve, static struct i915_request * last_active(const struct intel_engine_execlists *execlists) { - struct i915_request * const *last = execlists->active; + struct i915_request * const *last = READ_ONCE(execlists->active); while (*last && i915_request_completed(*last)) last++; @@ -961,7 +1481,7 @@ need_timeslice(struct intel_engine_cs *engine, const struct i915_request *rq) { int hint; - if (!intel_engine_has_semaphores(engine)) + if (!intel_engine_has_timeslices(engine)) return false; if (list_is_last(&rq->sched.link, &engine->active.requests)) @@ -982,15 +1502,32 @@ switch_prio(struct intel_engine_cs *engine, const struct i915_request *rq) return rq_prio(list_next_entry(rq, sched.link)); } -static bool -enable_timeslice(const struct intel_engine_execlists *execlists) +static inline unsigned long +timeslice(const struct intel_engine_cs *engine) { - const struct i915_request *rq = *execlists->active; + return READ_ONCE(engine->props.timeslice_duration_ms); +} + +static unsigned long +active_timeslice(const struct intel_engine_cs *engine) +{ + const struct i915_request *rq = *engine->execlists.active; if (i915_request_completed(rq)) - return false; + return 0; + + if (engine->execlists.switch_priority_hint < effective_prio(rq)) + return 0; - return execlists->switch_priority_hint >= effective_prio(rq); + return timeslice(engine); +} + +static void set_timeslice(struct intel_engine_cs *engine) +{ + if (!intel_engine_has_timeslices(engine)) + return; + + set_timer_ms(&engine->execlists.timer, active_timeslice(engine)); } static void record_preemption(struct intel_engine_execlists *execlists) @@ -998,6 +1535,30 @@ static void record_preemption(struct intel_engine_execlists *execlists) (void)I915_SELFTEST_ONLY(execlists->preempt_hang.count++); } +static unsigned long active_preempt_timeout(struct intel_engine_cs *engine) +{ + struct i915_request *rq; + + rq = last_active(&engine->execlists); + if (!rq) + return 0; + + /* Force a fast reset for terminated contexts (ignoring sysfs!) */ + if (unlikely(i915_gem_context_is_banned(rq->gem_context))) + return 1; + + return READ_ONCE(engine->props.preempt_timeout_ms); +} + +static void set_preempt_timeout(struct intel_engine_cs *engine) +{ + if (!intel_engine_has_preempt_reset(engine)) + return; + + set_timer_ms(&engine->execlists.preempt, + active_preempt_timeout(engine)); +} + static void execlists_dequeue(struct intel_engine_cs *engine) { struct intel_engine_execlists * const execlists = &engine->execlists; @@ -1094,7 +1655,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine) last->hw_context->lrc_desc |= CTX_DESC_FORCE_RESTORE; last = NULL; } else if (need_timeslice(engine, last) && - !timer_pending(&engine->execlists.timer)) { + timer_expired(&engine->execlists.timer)) { GEM_TRACE("%s: expired last=%llx:%lld, prio=%d, hint=%d\n", engine->name, last->fence.context, @@ -1130,8 +1691,18 @@ static void execlists_dequeue(struct intel_engine_cs *engine) * submission. */ if (!list_is_last(&last->sched.link, - &engine->active.requests)) + &engine->active.requests)) { + /* + * Even if ELSP[1] is occupied and not worthy + * of timeslices, our queue might be. + */ + if (!execlists->timer.expires && + need_timeslice(engine, last)) + set_timer_ms(&execlists->timer, + timeslice(engine)); + return; + } /* * WaIdleLiteRestore:bdw,skl @@ -1172,21 +1743,6 @@ static void execlists_dequeue(struct intel_engine_cs *engine) continue; } - if (i915_request_completed(rq)) { - ve->request = NULL; - ve->base.execlists.queue_priority_hint = INT_MIN; - rb_erase_cached(rb, &execlists->virtual); - RB_CLEAR_NODE(rb); - - rq->engine = engine; - __i915_request_submit(rq); - - spin_unlock(&ve->base.active.lock); - - rb = rb_first_cached(&execlists->virtual); - continue; - } - if (last && !can_merge_rq(last, rq)) { spin_unlock(&ve->base.active.lock); return; /* leave this for another */ @@ -1214,7 +1770,10 @@ static void execlists_dequeue(struct intel_engine_cs *engine) unsigned int n; GEM_BUG_ON(READ_ONCE(ve->context.inflight)); - virtual_update_register_offsets(regs, engine); + + if (!intel_engine_has_relative_mmio(engine)) + virtual_update_register_offsets(regs, + engine); if (!list_empty(&ve->context.signals)) virtual_xfer_breadcrumbs(ve, engine); @@ -1237,11 +1796,24 @@ static void execlists_dequeue(struct intel_engine_cs *engine) GEM_BUG_ON(ve->siblings[0] != engine); } - __i915_request_submit(rq); - if (!i915_request_completed(rq)) { + if (__i915_request_submit(rq)) { submit = true; last = rq; } + i915_request_put(rq); + + /* + * Hmm, we have a bunch of virtual engine requests, + * but the first one was already completed (thanks + * preempt-to-busy!). Keep looking at the veng queue + * until we have no more relevant requests (i.e. + * the normal submit queue has higher priority). + */ + if (!submit) { + spin_unlock(&ve->base.active.lock); + rb = rb_first_cached(&execlists->virtual); + continue; + } } spin_unlock(&ve->base.active.lock); @@ -1254,8 +1826,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine) int i; priolist_for_each_request_consume(rq, rn, p, i) { - if (i915_request_completed(rq)) - goto skip; + bool merge = true; /* * Can we combine this request with the current port? @@ -1285,6 +1856,9 @@ static void execlists_dequeue(struct intel_engine_cs *engine) if (last->hw_context == rq->hw_context) goto done; + if (i915_request_has_sentinel(last)) + goto done; + /* * If GVT overrides us we only ever submit * port[0], leaving port[1] empty. Note that we @@ -1296,14 +1870,23 @@ static void execlists_dequeue(struct intel_engine_cs *engine) ctx_single_port_submission(rq->hw_context)) goto done; - *port = execlists_schedule_in(last, port - execlists->pending); - port++; + merge = false; } - last = rq; - submit = true; -skip: - __i915_request_submit(rq); + if (__i915_request_submit(rq)) { + if (!merge) { + *port = execlists_schedule_in(last, port - execlists->pending); + port++; + last = NULL; + } + + GEM_BUG_ON(last && + !can_merge_ctx(last->hw_context, + rq->hw_context)); + + submit = true; + last = rq; + } } rb_erase_cached(&p->node, &execlists->queue); @@ -1334,11 +1917,28 @@ done: if (submit) { *port = execlists_schedule_in(last, port - execlists->pending); - memset(port + 1, 0, (last_port - port) * sizeof(*port)); execlists->switch_priority_hint = switch_prio(engine, *execlists->pending); + + /* + * Skip if we ended up with exactly the same set of requests, + * e.g. trying to timeslice a pair of ordered contexts + */ + if (!memcmp(execlists->active, execlists->pending, + (port - execlists->pending + 1) * sizeof(*port))) { + do + execlists_schedule_out(fetch_and_zero(port)); + while (port-- != execlists->pending); + + goto skip_submit; + } + + memset(port + 1, 0, (last_port - port) * sizeof(*port)); execlists_submit_ports(engine); + + set_preempt_timeout(engine); } else { +skip_submit: ring_set_paused(engine, 0); } } @@ -1346,16 +1946,17 @@ done: static void cancel_port_requests(struct intel_engine_execlists * const execlists) { - struct i915_request * const *port, *rq; + struct i915_request * const *port; - for (port = execlists->pending; (rq = *port); port++) - execlists_schedule_out(rq); + for (port = execlists->pending; *port; port++) + execlists_schedule_out(*port); memset(execlists->pending, 0, sizeof(execlists->pending)); - for (port = execlists->active; (rq = *port); port++) - execlists_schedule_out(rq); - execlists->active = - memset(execlists->inflight, 0, sizeof(execlists->inflight)); + /* Mark the end of active before we overwrite *active */ + for (port = xchg(&execlists->active, execlists->pending); *port; port++) + execlists_schedule_out(*port); + WRITE_ONCE(execlists->active, + memset(execlists->inflight, 0, sizeof(execlists->inflight))); } static inline void @@ -1371,13 +1972,6 @@ reset_in_progress(const struct intel_engine_execlists *execlists) return unlikely(!__tasklet_is_enabled(&execlists->tasklet)); } -enum csb_step { - CSB_NOP, - CSB_PROMOTE, - CSB_PREEMPT, - CSB_COMPLETE, -}; - /* * Starting with Gen12, the status has a new format: * @@ -1404,7 +1998,7 @@ enum csb_step { * bits 47-57: sw context id of the lrc the GT switched away from * bits 58-63: sw counter of the lrc the GT switched away from */ -static inline enum csb_step +static inline bool gen12_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) { u32 lower_dw = csb[0]; @@ -1413,9 +2007,6 @@ gen12_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) bool ctx_away_valid = GEN12_CSB_CTX_VALID(upper_dw); bool new_queue = lower_dw & GEN12_CTX_STATUS_SWITCHED_TO_NEW_QUEUE; - if (!ctx_away_valid && ctx_to_valid) - return CSB_PROMOTE; - /* * The context switch detail is not guaranteed to be 5 when a preemption * occurs, so we can't just check for that. The check below works for @@ -1423,8 +2014,10 @@ gen12_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) * instructions and lite-restore. Preempt-to-idle via the CTRL register * would require some extra handling, but we don't support that. */ - if (new_queue && ctx_away_valid) - return CSB_PREEMPT; + if (!ctx_away_valid || new_queue) { + GEM_BUG_ON(!ctx_to_valid); + return true; + } /* * switch detail = 5 is covered by the case above and we do not expect a @@ -1432,30 +2025,13 @@ gen12_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) * use polling mode. */ GEM_BUG_ON(GEN12_CTX_SWITCH_DETAIL(upper_dw)); - - if (*execlists->active) { - GEM_BUG_ON(!ctx_away_valid); - return CSB_COMPLETE; - } - - return CSB_NOP; + return false; } -static inline enum csb_step +static inline bool gen8_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) { - unsigned int status = *csb; - - if (status & GEN8_CTX_STATUS_IDLE_ACTIVE) - return CSB_PROMOTE; - - if (status & GEN8_CTX_STATUS_PREEMPTED) - return CSB_PREEMPT; - - if (*execlists->active) - return CSB_COMPLETE; - - return CSB_NOP; + return *csb & (GEN8_CTX_STATUS_IDLE_ACTIVE | GEN8_CTX_STATUS_PREEMPTED); } static void process_csb(struct intel_engine_cs *engine) @@ -1465,7 +2041,14 @@ static void process_csb(struct intel_engine_cs *engine) const u8 num_entries = execlists->csb_size; u8 head, tail; - GEM_BUG_ON(USES_GUC_SUBMISSION(engine->i915)); + /* + * As we modify our execlists state tracking we require exclusive + * access. Either we are inside the tasklet, or the tasklet is disabled + * and we assume that is only inside the reset paths and so serialised. + */ + GEM_BUG_ON(!tasklet_is_locked(&execlists->tasklet) && + !reset_in_progress(execlists)); + GEM_BUG_ON(!intel_engine_in_execlists_submission_mode(engine)); /* * Note that csb_write, csb_status may be either in HWSP or mmio. @@ -1494,7 +2077,7 @@ static void process_csb(struct intel_engine_cs *engine) rmb(); do { - enum csb_step csb_step; + bool promote; if (++head == num_entries) head = 0; @@ -1522,37 +2105,37 @@ static void process_csb(struct intel_engine_cs *engine) buf[2 * head + 0], buf[2 * head + 1]); if (INTEL_GEN(engine->i915) >= 12) - csb_step = gen12_csb_parse(execlists, buf + 2 * head); + promote = gen12_csb_parse(execlists, buf + 2 * head); else - csb_step = gen8_csb_parse(execlists, buf + 2 * head); - - switch (csb_step) { - case CSB_PREEMPT: /* cancel old inflight, prepare for switch */ - trace_ports(execlists, "preempted", execlists->active); + promote = gen8_csb_parse(execlists, buf + 2 * head); + if (promote) { + struct i915_request * const *old = execlists->active; - while (*execlists->active) - execlists_schedule_out(*execlists->active++); - - /* fallthrough */ - case CSB_PROMOTE: /* switch pending to inflight */ - GEM_BUG_ON(*execlists->active); - GEM_BUG_ON(!assert_pending_valid(execlists, "promote")); - execlists->active = - memcpy(execlists->inflight, - execlists->pending, - execlists_num_ports(execlists) * - sizeof(*execlists->pending)); - - if (enable_timeslice(execlists)) - mod_timer(&execlists->timer, jiffies + 1); + /* Point active to the new ELSP; prevent overwriting */ + WRITE_ONCE(execlists->active, execlists->pending); + set_timeslice(engine); if (!inject_preempt_hang(execlists)) ring_set_paused(engine, 0); + /* cancel old inflight, prepare for switch */ + trace_ports(execlists, "preempted", old); + while (*old) + execlists_schedule_out(*old++); + + /* switch pending to inflight */ + GEM_BUG_ON(!assert_pending_valid(execlists, "promote")); + WRITE_ONCE(execlists->active, + memcpy(execlists->inflight, + execlists->pending, + execlists_num_ports(execlists) * + sizeof(*execlists->pending))); + WRITE_ONCE(execlists->pending[0], NULL); - break; + } else { + GEM_BUG_ON(!*execlists->active); - case CSB_COMPLETE: /* port0 completed, advanced to port1 */ + /* port0 completed, advanced to port1 */ trace_ports(execlists, "completed", execlists->active); /* @@ -1567,10 +2150,6 @@ static void process_csb(struct intel_engine_cs *engine) GEM_BUG_ON(execlists->active - execlists->inflight > execlists_num_ports(execlists)); - break; - - case CSB_NOP: - break; } } while (head != tail); @@ -1593,8 +2172,48 @@ static void process_csb(struct intel_engine_cs *engine) static void __execlists_submission_tasklet(struct intel_engine_cs *const engine) { lockdep_assert_held(&engine->active.lock); - if (!engine->execlists.pending[0]) + if (!engine->execlists.pending[0]) { + rcu_read_lock(); /* protect peeking at execlists->active */ execlists_dequeue(engine); + rcu_read_unlock(); + } +} + +static noinline void preempt_reset(struct intel_engine_cs *engine) +{ + const unsigned int bit = I915_RESET_ENGINE + engine->id; + unsigned long *lock = &engine->gt->reset.flags; + + if (i915_modparams.reset < 3) + return; + + if (test_and_set_bit(bit, lock)) + return; + + /* Mark this tasklet as disabled to avoid waiting for it to complete */ + tasklet_disable_nosync(&engine->execlists.tasklet); + + GEM_TRACE("%s: preempt timeout %lu+%ums\n", + engine->name, + READ_ONCE(engine->props.preempt_timeout_ms), + jiffies_to_msecs(jiffies - engine->execlists.preempt.expires)); + intel_engine_reset(engine, "preemption time out"); + + tasklet_enable(&engine->execlists.tasklet); + clear_and_wake_up_bit(bit, lock); +} + +static bool preempt_timeout(const struct intel_engine_cs *const engine) +{ + const struct timer_list *t = &engine->execlists.preempt; + + if (!CONFIG_DRM_I915_PREEMPT_TIMEOUT) + return false; + + if (!timer_expired(t)) + return false; + + return READ_ONCE(engine->execlists.pending[0]); } /* @@ -1604,23 +2223,39 @@ static void __execlists_submission_tasklet(struct intel_engine_cs *const engine) static void execlists_submission_tasklet(unsigned long data) { struct intel_engine_cs * const engine = (struct intel_engine_cs *)data; - unsigned long flags; + bool timeout = preempt_timeout(engine); process_csb(engine); - if (!READ_ONCE(engine->execlists.pending[0])) { + if (!READ_ONCE(engine->execlists.pending[0]) || timeout) { + unsigned long flags; + spin_lock_irqsave(&engine->active.lock, flags); __execlists_submission_tasklet(engine); spin_unlock_irqrestore(&engine->active.lock, flags); + + /* Recheck after serialising with direct-submission */ + if (timeout && preempt_timeout(engine)) + preempt_reset(engine); } } -static void execlists_submission_timer(struct timer_list *timer) +static void __execlists_kick(struct intel_engine_execlists *execlists) { - struct intel_engine_cs *engine = - from_timer(engine, timer, execlists.timer); - /* Kick the tasklet for some interrupt coalescing and reset handling */ - tasklet_hi_schedule(&engine->execlists.tasklet); + tasklet_hi_schedule(&execlists->tasklet); +} + +#define execlists_kick(t, member) \ + __execlists_kick(container_of(t, struct intel_engine_execlists, member)) + +static void execlists_timeslice(struct timer_list *timer) +{ + execlists_kick(timer, timer); +} + +static void execlists_preempt(struct timer_list *timer) +{ + execlists_kick(timer, preempt); } static void queue_request(struct intel_engine_cs *engine, @@ -1700,7 +2335,6 @@ set_redzone(void *vaddr, const struct intel_engine_cs *engine) if (!IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) return; - vaddr += LRC_HEADER_PAGES * PAGE_SIZE; vaddr += engine->context_size; memset(vaddr, POISON_INUSE, I915_GTT_PAGE_SIZE); @@ -1712,7 +2346,6 @@ check_redzone(const void *vaddr, const struct intel_engine_cs *engine) if (!IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) return; - vaddr += LRC_HEADER_PAGES * PAGE_SIZE; vaddr += engine->context_size; if (memchr_inv(vaddr, POISON_INUSE, I915_GTT_PAGE_SIZE)) @@ -1726,14 +2359,13 @@ static void execlists_context_unpin(struct intel_context *ce) check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE, ce->engine); - i915_gem_context_unpin_hw_id(ce->gem_context); i915_gem_object_unpin_map(ce->state->obj); intel_ring_reset(ce->ring, ce->ring->tail); } static void -__execlists_update_reg_state(struct intel_context *ce, - struct intel_engine_cs *engine) +__execlists_update_reg_state(const struct intel_context *ce, + const struct intel_engine_cs *engine) { struct intel_ring *ring = ce->ring; u32 *regs = ce->lrc_reg_state; @@ -1741,16 +2373,16 @@ __execlists_update_reg_state(struct intel_context *ce, GEM_BUG_ON(!intel_ring_offset_valid(ring, ring->head)); GEM_BUG_ON(!intel_ring_offset_valid(ring, ring->tail)); - regs[CTX_RING_BUFFER_START + 1] = i915_ggtt_offset(ring->vma); - regs[CTX_RING_HEAD + 1] = ring->head; - regs[CTX_RING_TAIL + 1] = ring->tail; + regs[CTX_RING_BUFFER_START] = i915_ggtt_offset(ring->vma); + regs[CTX_RING_HEAD] = ring->head; + regs[CTX_RING_TAIL] = ring->tail; /* RPCS */ if (engine->class == RENDER_CLASS) { - regs[CTX_R_PWR_CLK_STATE + 1] = + regs[CTX_R_PWR_CLK_STATE] = intel_sseu_make_rpcs(engine->i915, &ce->sseu); - i915_oa_init_reg_state(engine, ce, regs); + i915_oa_init_reg_state(ce, engine); } } @@ -1776,18 +2408,12 @@ __execlists_context_pin(struct intel_context *ce, goto unpin_active; } - ret = i915_gem_context_pin_hw_id(ce->gem_context); - if (ret) - goto unpin_map; - ce->lrc_desc = lrc_descriptor(ce, engine); ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE; __execlists_update_reg_state(ce, engine); return 0; -unpin_map: - i915_gem_object_unpin_map(ce->state->obj); unpin_active: intel_context_active_release(ce); err: @@ -1843,7 +2469,7 @@ static int gen8_emit_init_breadcrumb(struct i915_request *rq) { u32 *cs; - GEM_BUG_ON(!rq->timeline->has_initial_breadcrumb); + GEM_BUG_ON(!i915_request_timeline(rq)->has_initial_breadcrumb); cs = intel_ring_begin(rq, 6); if (IS_ERR(cs)) @@ -1859,7 +2485,7 @@ static int gen8_emit_init_breadcrumb(struct i915_request *rq) *cs++ = MI_NOOP; *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; - *cs++ = rq->timeline->hwsp_offset; + *cs++ = i915_request_timeline(rq)->hwsp_offset; *cs++ = 0; *cs++ = rq->fence.seqno - 1; @@ -1871,60 +2497,6 @@ static int gen8_emit_init_breadcrumb(struct i915_request *rq) return 0; } -static int emit_pdps(struct i915_request *rq) -{ - const struct intel_engine_cs * const engine = rq->engine; - struct i915_ppgtt * const ppgtt = i915_vm_to_ppgtt(rq->hw_context->vm); - int err, i; - u32 *cs; - - GEM_BUG_ON(intel_vgpu_active(rq->i915)); - - /* - * Beware ye of the dragons, this sequence is magic! - * - * Small changes to this sequence can cause anything from - * GPU hangs to forcewake errors and machine lockups! - */ - - /* Flush any residual operations from the context load */ - err = engine->emit_flush(rq, EMIT_FLUSH); - if (err) - return err; - - /* Magic required to prevent forcewake errors! */ - err = engine->emit_flush(rq, EMIT_INVALIDATE); - if (err) - return err; - - cs = intel_ring_begin(rq, 4 * GEN8_3LVL_PDPES + 2); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - /* Ensure the LRI have landed before we invalidate & continue */ - *cs++ = MI_LOAD_REGISTER_IMM(2 * GEN8_3LVL_PDPES) | MI_LRI_FORCE_POSTED; - for (i = GEN8_3LVL_PDPES; i--; ) { - const dma_addr_t pd_daddr = i915_page_dir_dma_addr(ppgtt, i); - u32 base = engine->mmio_base; - - *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, i)); - *cs++ = upper_32_bits(pd_daddr); - *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, i)); - *cs++ = lower_32_bits(pd_daddr); - } - *cs++ = MI_NOOP; - - intel_ring_advance(rq, cs); - - /* Be doubly sure the LRI have landed before proceeding */ - err = engine->emit_flush(rq, EMIT_FLUSH); - if (err) - return err; - - /* Re-invalidate the TLB for luck */ - return engine->emit_flush(rq, EMIT_INVALIDATE); -} - static int execlists_request_alloc(struct i915_request *request) { int ret; @@ -1947,10 +2519,7 @@ static int execlists_request_alloc(struct i915_request *request) */ /* Unconditionally invalidate GPU caches and TLBs. */ - if (i915_vm_is_4lvl(request->hw_context->vm)) - ret = request->engine->emit_flush(request, EMIT_INVALIDATE); - else - ret = emit_pdps(request); + ret = request->engine->emit_flush(request, EMIT_INVALIDATE); if (ret) return ret; @@ -2002,12 +2571,6 @@ gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch) return batch; } -static u32 slm_offset(struct intel_engine_cs *engine) -{ - return intel_gt_scratch_offset(engine->gt, - INTEL_GT_SCRATCH_FIELD_CLEAR_SLM_WA); -} - /* * Typically we only have one indirect_ctx and per_ctx batch buffer which are * initialized at the beginning and shared across all contexts but this field @@ -2036,10 +2599,10 @@ static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch) /* Actual scratch location is at 128 bytes offset */ batch = gen8_emit_pipe_control(batch, PIPE_CONTROL_FLUSH_L3 | - PIPE_CONTROL_GLOBAL_GTT_IVB | + PIPE_CONTROL_STORE_DATA_INDEX | PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE, - slm_offset(engine)); + LRC_PPHWSP_SCRATCH_ADDR); *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; @@ -2397,23 +2960,29 @@ static void reset_csb_pointers(struct intel_engine_cs *engine) &execlists->csb_status[reset_value]); } -static struct i915_request *active_request(struct i915_request *rq) +static int lrc_ring_mi_mode(const struct intel_engine_cs *engine) { - const struct list_head * const list = &rq->timeline->requests; - const struct intel_context * const ce = rq->hw_context; - struct i915_request *active = NULL; - - list_for_each_entry_from_reverse(rq, list, link) { - if (i915_request_completed(rq)) - break; + if (INTEL_GEN(engine->i915) >= 12) + return 0x60; + else if (INTEL_GEN(engine->i915) >= 9) + return 0x54; + else if (engine->class == RENDER_CLASS) + return 0x58; + else + return -1; +} - if (rq->hw_context != ce) - break; +static void __execlists_reset_reg_state(const struct intel_context *ce, + const struct intel_engine_cs *engine) +{ + u32 *regs = ce->lrc_reg_state; + int x; - active = rq; + x = lrc_ring_mi_mode(engine); + if (x != -1) { + regs[x + 1] &= ~STOP_RING; + regs[x + 1] |= STOP_RING << 16; } - - return active; } static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) @@ -2421,7 +2990,10 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) struct intel_engine_execlists * const execlists = &engine->execlists; struct intel_context *ce; struct i915_request *rq; - u32 *regs; + + mb(); /* paranoia: read the CSB pointers from after the reset */ + clflush(execlists->csb_write); + mb(); process_csb(engine); /* drain preemption events */ @@ -2437,16 +3009,23 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) if (!rq) goto unwind; + /* We still have requests in-flight; the engine should be active */ + GEM_BUG_ON(!intel_engine_pm_is_awake(engine)); + ce = rq->hw_context; - GEM_BUG_ON(i915_active_is_idle(&ce->active)); GEM_BUG_ON(!i915_vma_is_pinned(ce->state)); - rq = active_request(rq); - if (!rq) { - ce->ring->head = ce->ring->tail; + + if (i915_request_completed(rq)) { + /* Idle context; tidy up the ring so we can restart afresh */ + ce->ring->head = intel_ring_wrap(ce->ring, rq->tail); goto out_replay; } + /* Context has requests still in-flight; it should not be idle! */ + GEM_BUG_ON(i915_active_is_idle(&ce->active)); + rq = active_request(ce->timeline, rq); ce->ring->head = intel_ring_wrap(ce->ring, rq->head); + GEM_BUG_ON(ce->ring->head == ce->ring->tail); /* * If this request hasn't started yet, e.g. it is waiting on a @@ -2486,19 +3065,16 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) * future request will be after userspace has had the opportunity * to recreate its own state. */ - regs = ce->lrc_reg_state; - if (engine->pinned_default_state) { - memcpy(regs, /* skip restoring the vanilla PPHWSP */ - engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE, - engine->context_size - PAGE_SIZE); - } - execlists_init_reg_state(regs, ce, engine, ce->ring); + GEM_BUG_ON(!intel_context_is_pinned(ce)); + restore_default_state(ce, engine); out_replay: - GEM_TRACE("%s replay {head:%04x, tail:%04x\n", + GEM_TRACE("%s replay {head:%04x, tail:%04x}\n", engine->name, ce->ring->head, ce->ring->tail); intel_ring_update_space(ce->ring); + __execlists_reset_reg_state(ce, engine); __execlists_update_reg_state(ce, engine); + ce->lrc_desc |= CTX_DESC_FORCE_RESTORE; /* paranoid: GPU was reset! */ unwind: /* Push back any incomplete requests for replay after the reset. */ @@ -2552,12 +3128,8 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) __execlists_reset(engine, true); /* Mark all executing requests as skipped. */ - list_for_each_entry(rq, &engine->active.requests, sched.link) { - if (!i915_request_signaled(rq)) - dma_fence_set_error(&rq->fence, -EIO); - - i915_request_mark_complete(rq); - } + list_for_each_entry(rq, &engine->active.requests, sched.link) + mark_eio(rq); /* Flush the queued requests to the timeline list (for retiring). */ while ((rb = rb_first_cached(&execlists->queue))) { @@ -2565,10 +3137,8 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) int i; priolist_for_each_request_consume(rq, rn, p, i) { - list_del_init(&rq->sched.link); + mark_eio(rq); __i915_request_submit(rq); - dma_fence_set_error(&rq->fence, -EIO); - i915_request_mark_complete(rq); } rb_erase_cached(&p->node, &execlists->queue); @@ -2584,13 +3154,15 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) RB_CLEAR_NODE(rb); spin_lock(&ve->base.active.lock); - if (ve->request) { - ve->request->engine = engine; - __i915_request_submit(ve->request); - dma_fence_set_error(&ve->request->fence, -EIO); - i915_request_mark_complete(ve->request); + rq = fetch_and_zero(&ve->request); + if (rq) { + mark_eio(rq); + + rq->engine = engine; + __i915_request_submit(rq); + i915_request_put(rq); + ve->base.execlists.queue_priority_hint = INT_MIN; - ve->request = NULL; } spin_unlock(&ve->base.active.lock); } @@ -2723,7 +3295,7 @@ static int gen8_emit_flush(struct i915_request *request, u32 mode) } *cs++ = cmd; - *cs++ = I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT; + *cs++ = LRC_PPHWSP_SCRATCH_ADDR; *cs++ = 0; /* upper addr */ *cs++ = 0; /* value */ intel_ring_advance(request, cs); @@ -2734,10 +3306,6 @@ static int gen8_emit_flush(struct i915_request *request, u32 mode) static int gen8_emit_flush_render(struct i915_request *request, u32 mode) { - struct intel_engine_cs *engine = request->engine; - u32 scratch_addr = - intel_gt_scratch_offset(engine->gt, - INTEL_GT_SCRATCH_FIELD_RENDER_FLUSH); bool vf_flush_wa = false, dc_flush_wa = false; u32 *cs, flags = 0; int len; @@ -2759,7 +3327,7 @@ static int gen8_emit_flush_render(struct i915_request *request, flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; flags |= PIPE_CONTROL_QW_WRITE; - flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; + flags |= PIPE_CONTROL_STORE_DATA_INDEX; /* * On GEN9: before VF_CACHE_INVALIDATE we need to emit a NULL @@ -2792,7 +3360,7 @@ static int gen8_emit_flush_render(struct i915_request *request, cs = gen8_emit_pipe_control(cs, PIPE_CONTROL_DC_FLUSH_ENABLE, 0); - cs = gen8_emit_pipe_control(cs, flags, scratch_addr); + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); if (dc_flush_wa) cs = gen8_emit_pipe_control(cs, PIPE_CONTROL_CS_STALL, 0); @@ -2805,11 +3373,6 @@ static int gen8_emit_flush_render(struct i915_request *request, static int gen11_emit_flush_render(struct i915_request *request, u32 mode) { - struct intel_engine_cs *engine = request->engine; - const u32 scratch_addr = - intel_gt_scratch_offset(engine->gt, - INTEL_GT_SCRATCH_FIELD_RENDER_FLUSH); - if (mode & EMIT_FLUSH) { u32 *cs; u32 flags = 0; @@ -2822,13 +3385,13 @@ static int gen11_emit_flush_render(struct i915_request *request, flags |= PIPE_CONTROL_DC_FLUSH_ENABLE; flags |= PIPE_CONTROL_FLUSH_ENABLE; flags |= PIPE_CONTROL_QW_WRITE; - flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; + flags |= PIPE_CONTROL_STORE_DATA_INDEX; cs = intel_ring_begin(request, 6); if (IS_ERR(cs)) return PTR_ERR(cs); - cs = gen8_emit_pipe_control(cs, flags, scratch_addr); + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); intel_ring_advance(request, cs); } @@ -2846,16 +3409,108 @@ static int gen11_emit_flush_render(struct i915_request *request, flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; flags |= PIPE_CONTROL_QW_WRITE; - flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; + flags |= PIPE_CONTROL_STORE_DATA_INDEX; + + cs = intel_ring_begin(request, 6); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); + intel_ring_advance(request, cs); + } + + return 0; +} + +static u32 preparser_disable(bool state) +{ + return MI_ARB_CHECK | 1 << 8 | state; +} + +static int gen12_emit_flush_render(struct i915_request *request, + u32 mode) +{ + if (mode & EMIT_FLUSH) { + u32 flags = 0; + u32 *cs; + + flags |= PIPE_CONTROL_TILE_CACHE_FLUSH; + flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH; + flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH; + /* Wa_1409600907:tgl */ + flags |= PIPE_CONTROL_DEPTH_STALL; + flags |= PIPE_CONTROL_DC_FLUSH_ENABLE; + flags |= PIPE_CONTROL_FLUSH_ENABLE; + flags |= PIPE_CONTROL_HDC_PIPELINE_FLUSH; + + flags |= PIPE_CONTROL_STORE_DATA_INDEX; + flags |= PIPE_CONTROL_QW_WRITE; + + flags |= PIPE_CONTROL_CS_STALL; cs = intel_ring_begin(request, 6); if (IS_ERR(cs)) return PTR_ERR(cs); - cs = gen8_emit_pipe_control(cs, flags, scratch_addr); + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); intel_ring_advance(request, cs); } + if (mode & EMIT_INVALIDATE) { + u32 flags = 0; + u32 *cs; + + flags |= PIPE_CONTROL_COMMAND_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_TLB_INVALIDATE; + flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_L3_RO_CACHE_INVALIDATE; + + flags |= PIPE_CONTROL_STORE_DATA_INDEX; + flags |= PIPE_CONTROL_QW_WRITE; + + flags |= PIPE_CONTROL_CS_STALL; + + cs = intel_ring_begin(request, 8); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + /* + * Prevent the pre-parser from skipping past the TLB + * invalidate and loading a stale page for the batch + * buffer / request payload. + */ + *cs++ = preparser_disable(true); + + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); + + *cs++ = preparser_disable(false); + intel_ring_advance(request, cs); + + /* + * Wa_1604544889:tgl + */ + if (IS_TGL_REVID(request->i915, TGL_REVID_A0, TGL_REVID_A0)) { + flags = 0; + flags |= PIPE_CONTROL_CS_STALL; + flags |= PIPE_CONTROL_HDC_PIPELINE_FLUSH; + + flags |= PIPE_CONTROL_STORE_DATA_INDEX; + flags |= PIPE_CONTROL_QW_WRITE; + + cs = intel_ring_begin(request, 6); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + cs = gen8_emit_pipe_control(cs, flags, + LRC_PPHWSP_SCRATCH_ADDR); + intel_ring_advance(request, cs); + } + } + return 0; } @@ -2907,7 +3562,7 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs) { cs = gen8_emit_ggtt_write(cs, request->fence.seqno, - request->timeline->hwsp_offset, + i915_request_active_timeline(request)->hwsp_offset, 0); return gen8_emit_fini_breadcrumb_footer(request, cs); @@ -2915,28 +3570,28 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs) static u32 *gen8_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs) { - cs = gen8_emit_ggtt_write_rcs(cs, - request->fence.seqno, - request->timeline->hwsp_offset, - PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH | - PIPE_CONTROL_DEPTH_CACHE_FLUSH | - PIPE_CONTROL_DC_FLUSH_ENABLE); - - /* XXX flush+write+CS_STALL all in one upsets gem_concurrent_blt:kbl */ cs = gen8_emit_pipe_control(cs, - PIPE_CONTROL_FLUSH_ENABLE | - PIPE_CONTROL_CS_STALL, + PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH | + PIPE_CONTROL_DEPTH_CACHE_FLUSH | + PIPE_CONTROL_DC_FLUSH_ENABLE, 0); + /* XXX flush+write+CS_STALL all in one upsets gem_concurrent_blt:kbl */ + cs = gen8_emit_ggtt_write_rcs(cs, + request->fence.seqno, + i915_request_active_timeline(request)->hwsp_offset, + PIPE_CONTROL_FLUSH_ENABLE | + PIPE_CONTROL_CS_STALL); + return gen8_emit_fini_breadcrumb_footer(request, cs); } -static u32 *gen11_emit_fini_breadcrumb_rcs(struct i915_request *request, - u32 *cs) +static u32 * +gen11_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs) { cs = gen8_emit_ggtt_write_rcs(cs, request->fence.seqno, - request->timeline->hwsp_offset, + i915_request_active_timeline(request)->hwsp_offset, PIPE_CONTROL_CS_STALL | PIPE_CONTROL_TILE_CACHE_FLUSH | PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH | @@ -2947,9 +3602,88 @@ static u32 *gen11_emit_fini_breadcrumb_rcs(struct i915_request *request, return gen8_emit_fini_breadcrumb_footer(request, cs); } +/* + * Note that the CS instruction pre-parser will not stall on the breadcrumb + * flush and will continue pre-fetching the instructions after it before the + * memory sync is completed. On pre-gen12 HW, the pre-parser will stop at + * BB_START/END instructions, so, even though we might pre-fetch the pre-amble + * of the next request before the memory has been flushed, we're guaranteed that + * we won't access the batch itself too early. + * However, on gen12+ the parser can pre-fetch across the BB_START/END commands, + * so, if the current request is modifying an instruction in the next request on + * the same intel_context, we might pre-fetch and then execute the pre-update + * instruction. To avoid this, the users of self-modifying code should either + * disable the parser around the code emitting the memory writes, via a new flag + * added to MI_ARB_CHECK, or emit the writes from a different intel_context. For + * the in-kernel use-cases we've opted to use a separate context, see + * reloc_gpu() as an example. + * All the above applies only to the instructions themselves. Non-inline data + * used by the instructions is not pre-fetched. + */ + +static u32 *gen12_emit_preempt_busywait(struct i915_request *request, u32 *cs) +{ + *cs++ = MI_SEMAPHORE_WAIT_TOKEN | + MI_SEMAPHORE_GLOBAL_GTT | + MI_SEMAPHORE_POLL | + MI_SEMAPHORE_SAD_EQ_SDD; + *cs++ = 0; + *cs++ = intel_hws_preempt_address(request->engine); + *cs++ = 0; + *cs++ = 0; + *cs++ = MI_NOOP; + + return cs; +} + +static __always_inline u32* +gen12_emit_fini_breadcrumb_footer(struct i915_request *request, u32 *cs) +{ + *cs++ = MI_USER_INTERRUPT; + + *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; + if (intel_engine_has_semaphores(request->engine)) + cs = gen12_emit_preempt_busywait(request, cs); + + request->tail = intel_ring_offset(request, cs); + assert_ring_tail_valid(request->ring, request->tail); + + return gen8_emit_wa_tail(request, cs); +} + +static u32 *gen12_emit_fini_breadcrumb(struct i915_request *request, u32 *cs) +{ + cs = gen8_emit_ggtt_write(cs, + request->fence.seqno, + i915_request_active_timeline(request)->hwsp_offset, + 0); + + return gen12_emit_fini_breadcrumb_footer(request, cs); +} + +static u32 * +gen12_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs) +{ + cs = gen8_emit_ggtt_write_rcs(cs, + request->fence.seqno, + i915_request_active_timeline(request)->hwsp_offset, + PIPE_CONTROL_CS_STALL | + PIPE_CONTROL_TILE_CACHE_FLUSH | + PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH | + PIPE_CONTROL_DEPTH_CACHE_FLUSH | + /* Wa_1409600907:tgl */ + PIPE_CONTROL_DEPTH_STALL | + PIPE_CONTROL_DC_FLUSH_ENABLE | + PIPE_CONTROL_FLUSH_ENABLE | + PIPE_CONTROL_HDC_PIPELINE_FLUSH); + + return gen12_emit_fini_breadcrumb_footer(request, cs); +} + static void execlists_park(struct intel_engine_cs *engine) { - del_timer(&engine->execlists.timer); + cancel_timer(&engine->execlists.timer); + cancel_timer(&engine->execlists.preempt); } void intel_execlists_set_default_submission(struct intel_engine_cs *engine) @@ -2972,6 +3706,9 @@ void intel_execlists_set_default_submission(struct intel_engine_cs *engine) if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) engine->flags |= I915_ENGINE_HAS_PREEMPTION; } + + if (INTEL_GEN(engine->i915) >= 12) + engine->flags |= I915_ENGINE_HAS_RELATIVE_MMIO; } static void execlists_destroy(struct intel_engine_cs *engine) @@ -2999,6 +3736,8 @@ logical_ring_default_vfuncs(struct intel_engine_cs *engine) engine->emit_flush = gen8_emit_flush; engine->emit_init_breadcrumb = gen8_emit_init_breadcrumb; engine->emit_fini_breadcrumb = gen8_emit_fini_breadcrumb; + if (INTEL_GEN(engine->i915) >= 12) + engine->emit_fini_breadcrumb = gen12_emit_fini_breadcrumb; engine->set_default_submission = intel_execlists_set_default_submission; @@ -3044,6 +3783,9 @@ static void rcs_submission_override(struct intel_engine_cs *engine) { switch (INTEL_GEN(engine->i915)) { case 12: + engine->emit_flush = gen12_emit_flush_render; + engine->emit_fini_breadcrumb = gen12_emit_fini_breadcrumb_rcs; + break; case 11: engine->emit_flush = gen11_emit_flush_render; engine->emit_fini_breadcrumb = gen11_emit_fini_breadcrumb_rcs; @@ -3059,7 +3801,8 @@ int intel_execlists_submission_setup(struct intel_engine_cs *engine) { tasklet_init(&engine->execlists.tasklet, execlists_submission_tasklet, (unsigned long)engine); - timer_setup(&engine->execlists.timer, execlists_submission_timer, 0); + timer_setup(&engine->execlists.timer, execlists_timeslice, 0); + timer_setup(&engine->execlists.preempt, execlists_preempt, 0); logical_ring_default_vfuncs(engine); logical_ring_default_irqs(engine); @@ -3116,7 +3859,7 @@ int intel_execlists_submission_init(struct intel_engine_cs *engine) return 0; } -static u32 intel_lr_indirect_ctx_offset(struct intel_engine_cs *engine) +static u32 intel_lr_indirect_ctx_offset(const struct intel_engine_cs *engine) { u32 indirect_ctx_offset; @@ -3149,86 +3892,50 @@ static u32 intel_lr_indirect_ctx_offset(struct intel_engine_cs *engine) return indirect_ctx_offset; } -static void execlists_init_reg_state(u32 *regs, - struct intel_context *ce, - struct intel_engine_cs *engine, - struct intel_ring *ring) -{ - struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ce->vm); - bool rcs = engine->class == RENDER_CLASS; - u32 base = engine->mmio_base; - /* - * A context is actually a big batch buffer with several - * MI_LOAD_REGISTER_IMM commands followed by (reg, value) pairs. The - * values we are setting here are only for the first context restore: - * on a subsequent save, the GPU will recreate this batchbuffer with new - * values (including all the missing MI_LOAD_REGISTER_IMM commands that - * we are not initializing here). - * - * Must keep consistent with virtual_update_register_offsets(). - */ - regs[CTX_LRI_HEADER_0] = MI_LOAD_REGISTER_IMM(rcs ? 14 : 11) | - MI_LRI_FORCE_POSTED; - - CTX_REG(regs, CTX_CONTEXT_CONTROL, RING_CONTEXT_CONTROL(base), +static void init_common_reg_state(u32 * const regs, + const struct intel_engine_cs *engine, + const struct intel_ring *ring) +{ + regs[CTX_CONTEXT_CONTROL] = _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT) | - _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH)); - if (INTEL_GEN(engine->i915) < 11) { - regs[CTX_CONTEXT_CONTROL + 1] |= + _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH); + if (INTEL_GEN(engine->i915) < 11) + regs[CTX_CONTEXT_CONTROL] |= _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT | CTX_CTRL_RS_CTX_ENABLE); - } - CTX_REG(regs, CTX_RING_HEAD, RING_HEAD(base), 0); - CTX_REG(regs, CTX_RING_TAIL, RING_TAIL(base), 0); - CTX_REG(regs, CTX_RING_BUFFER_START, RING_START(base), 0); - CTX_REG(regs, CTX_RING_BUFFER_CONTROL, RING_CTL(base), - RING_CTL_SIZE(ring->size) | RING_VALID); - CTX_REG(regs, CTX_BB_HEAD_U, RING_BBADDR_UDW(base), 0); - CTX_REG(regs, CTX_BB_HEAD_L, RING_BBADDR(base), 0); - CTX_REG(regs, CTX_BB_STATE, RING_BBSTATE(base), RING_BB_PPGTT); - CTX_REG(regs, CTX_SECOND_BB_HEAD_U, RING_SBBADDR_UDW(base), 0); - CTX_REG(regs, CTX_SECOND_BB_HEAD_L, RING_SBBADDR(base), 0); - CTX_REG(regs, CTX_SECOND_BB_STATE, RING_SBBSTATE(base), 0); - if (rcs) { - struct i915_ctx_workarounds *wa_ctx = &engine->wa_ctx; - - CTX_REG(regs, CTX_RCS_INDIRECT_CTX, RING_INDIRECT_CTX(base), 0); - CTX_REG(regs, CTX_RCS_INDIRECT_CTX_OFFSET, - RING_INDIRECT_CTX_OFFSET(base), 0); - if (wa_ctx->indirect_ctx.size) { - u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma); - - regs[CTX_RCS_INDIRECT_CTX + 1] = - (ggtt_offset + wa_ctx->indirect_ctx.offset) | - (wa_ctx->indirect_ctx.size / CACHELINE_BYTES); - - regs[CTX_RCS_INDIRECT_CTX_OFFSET + 1] = - intel_lr_indirect_ctx_offset(engine) << 6; - } - CTX_REG(regs, CTX_BB_PER_CTX_PTR, RING_BB_PER_CTX_PTR(base), 0); - if (wa_ctx->per_ctx.size) { - u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma); + regs[CTX_RING_BUFFER_CONTROL] = RING_CTL_SIZE(ring->size) | RING_VALID; + regs[CTX_BB_STATE] = RING_BB_PPGTT; +} - regs[CTX_BB_PER_CTX_PTR + 1] = - (ggtt_offset + wa_ctx->per_ctx.offset) | 0x01; - } +static void init_wa_bb_reg_state(u32 * const regs, + const struct intel_engine_cs *engine, + u32 pos_bb_per_ctx) +{ + const struct i915_ctx_workarounds * const wa_ctx = &engine->wa_ctx; + + if (wa_ctx->per_ctx.size) { + const u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma); + + regs[pos_bb_per_ctx] = + (ggtt_offset + wa_ctx->per_ctx.offset) | 0x01; } - regs[CTX_LRI_HEADER_1] = MI_LOAD_REGISTER_IMM(9) | MI_LRI_FORCE_POSTED; + if (wa_ctx->indirect_ctx.size) { + const u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma); - CTX_REG(regs, CTX_CTX_TIMESTAMP, RING_CTX_TIMESTAMP(base), 0); - /* PDP values well be assigned later if needed */ - CTX_REG(regs, CTX_PDP3_UDW, GEN8_RING_PDP_UDW(base, 3), 0); - CTX_REG(regs, CTX_PDP3_LDW, GEN8_RING_PDP_LDW(base, 3), 0); - CTX_REG(regs, CTX_PDP2_UDW, GEN8_RING_PDP_UDW(base, 2), 0); - CTX_REG(regs, CTX_PDP2_LDW, GEN8_RING_PDP_LDW(base, 2), 0); - CTX_REG(regs, CTX_PDP1_UDW, GEN8_RING_PDP_UDW(base, 1), 0); - CTX_REG(regs, CTX_PDP1_LDW, GEN8_RING_PDP_LDW(base, 1), 0); - CTX_REG(regs, CTX_PDP0_UDW, GEN8_RING_PDP_UDW(base, 0), 0); - CTX_REG(regs, CTX_PDP0_LDW, GEN8_RING_PDP_LDW(base, 0), 0); + regs[pos_bb_per_ctx + 2] = + (ggtt_offset + wa_ctx->indirect_ctx.offset) | + (wa_ctx->indirect_ctx.size / CACHELINE_BYTES); + regs[pos_bb_per_ctx + 4] = + intel_lr_indirect_ctx_offset(engine) << 6; + } +} + +static void init_ppgtt_reg_state(u32 *regs, const struct i915_ppgtt *ppgtt) +{ if (i915_vm_is_4lvl(&ppgtt->vm)) { /* 64b PPGTT (48bit canonical) * PDP0_DESCRIPTOR contains the base address to PML4 and @@ -3241,15 +3948,47 @@ static void execlists_init_reg_state(u32 *regs, ASSIGN_CTX_PDP(ppgtt, regs, 1); ASSIGN_CTX_PDP(ppgtt, regs, 0); } +} + +static struct i915_ppgtt *vm_alias(struct i915_address_space *vm) +{ + if (i915_is_ggtt(vm)) + return i915_vm_to_ggtt(vm)->alias; + else + return i915_vm_to_ppgtt(vm); +} - if (rcs) { - regs[CTX_LRI_HEADER_2] = MI_LOAD_REGISTER_IMM(1); - CTX_REG(regs, CTX_R_PWR_CLK_STATE, GEN8_R_PWR_CLK_STATE, 0); +static void execlists_init_reg_state(u32 *regs, + const struct intel_context *ce, + const struct intel_engine_cs *engine, + const struct intel_ring *ring, + bool close) +{ + /* + * A context is actually a big batch buffer with several + * MI_LOAD_REGISTER_IMM commands followed by (reg, value) pairs. The + * values we are setting here are only for the first context restore: + * on a subsequent save, the GPU will recreate this batchbuffer with new + * values (including all the missing MI_LOAD_REGISTER_IMM commands that + * we are not initializing here). + * + * Must keep consistent with virtual_update_register_offsets(). + */ + u32 *bbe = set_offsets(regs, reg_offsets(engine), engine); + + if (close) { /* Close the batch; used mainly by live_lrc_layout() */ + *bbe = MI_BATCH_BUFFER_END; + if (INTEL_GEN(engine->i915) >= 10) + *bbe |= BIT(0); } - regs[CTX_END] = MI_BATCH_BUFFER_END; - if (INTEL_GEN(engine->i915) >= 10) - regs[CTX_END] |= BIT(0); + init_common_reg_state(regs, engine, ring); + init_ppgtt_reg_state(regs, vm_alias(ce->vm)); + + init_wa_bb_reg_state(regs, engine, + INTEL_GEN(engine->i915) >= 12 ? + GEN12_CTX_BB_PER_CTX_PTR : + CTX_BB_PER_CTX_PTR); } static int @@ -3258,6 +3997,7 @@ populate_lr_context(struct intel_context *ce, struct intel_engine_cs *engine, struct intel_ring *ring) { + bool inhibit = true; void *vaddr; u32 *regs; int ret; @@ -3272,12 +4012,6 @@ populate_lr_context(struct intel_context *ce, set_redzone(vaddr, engine); if (engine->default_state) { - /* - * We only want to copy over the template context state; - * skipping over the headers reserved for GuC communication, - * leaving those as zero. - */ - const unsigned long start = LRC_HEADER_PAGES * PAGE_SIZE; void *defaults; defaults = i915_gem_object_pin_map(engine->default_state, @@ -3287,23 +4021,22 @@ populate_lr_context(struct intel_context *ce, goto err_unpin_ctx; } - memcpy(vaddr + start, defaults + start, engine->context_size); + memcpy(vaddr, defaults, engine->context_size); i915_gem_object_unpin_map(engine->default_state); + inhibit = false; } /* The second page of the context object contains some fields which must * be set up prior to the first execution. */ regs = vaddr + LRC_STATE_PN * PAGE_SIZE; - execlists_init_reg_state(regs, ce, engine, ring); - if (!engine->default_state) - regs[CTX_CONTEXT_CONTROL + 1] |= + execlists_init_reg_state(regs, ce, engine, ring, inhibit); + if (inhibit) + regs[CTX_CONTEXT_CONTROL] |= _MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT); ret = 0; err_unpin_ctx: - __i915_gem_object_flush_map(ctx_obj, - LRC_HEADER_PAGES * PAGE_SIZE, - engine->context_size); + __i915_gem_object_flush_map(ctx_obj, 0, engine->context_size); i915_gem_object_unpin_map(ctx_obj); return ret; } @@ -3320,11 +4053,6 @@ static int __execlists_context_alloc(struct intel_context *ce, GEM_BUG_ON(ce->state); context_size = round_up(engine->context_size, I915_GTT_PAGE_SIZE); - /* - * Before the actual start of the context image, we insert a few pages - * for our own use and for sharing with the GuC. - */ - context_size += LRC_HEADER_PAGES * PAGE_SIZE; if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) context_size += I915_GTT_PAGE_SIZE; /* for redzone */ @@ -3436,8 +4164,9 @@ static void virtual_engine_initial_hint(struct virtual_engine *ve) return; swap(ve->siblings[swp], ve->siblings[0]); - virtual_update_register_offsets(ve->context.lrc_reg_state, - ve->siblings[0]); + if (!intel_engine_has_relative_mmio(ve->siblings[0])) + virtual_update_register_offsets(ve->context.lrc_reg_state, + ve->siblings[0]); } static int virtual_context_pin(struct intel_context *ce) @@ -3594,6 +4323,8 @@ submit_engine: static void virtual_submit_request(struct i915_request *rq) { struct virtual_engine *ve = to_virtual_engine(rq->engine); + struct i915_request *old; + unsigned long flags; GEM_TRACE("%s: rq=%llx:%lld\n", ve->base.name, @@ -3602,15 +4333,31 @@ static void virtual_submit_request(struct i915_request *rq) GEM_BUG_ON(ve->base.submit_request != virtual_submit_request); - GEM_BUG_ON(ve->request); - GEM_BUG_ON(!list_empty(virtual_queue(ve))); + spin_lock_irqsave(&ve->base.active.lock, flags); - ve->base.execlists.queue_priority_hint = rq_prio(rq); - WRITE_ONCE(ve->request, rq); + old = ve->request; + if (old) { /* background completion event from preempt-to-busy */ + GEM_BUG_ON(!i915_request_completed(old)); + __i915_request_submit(old); + i915_request_put(old); + } + + if (i915_request_completed(rq)) { + __i915_request_submit(rq); + + ve->base.execlists.queue_priority_hint = INT_MIN; + ve->request = NULL; + } else { + ve->base.execlists.queue_priority_hint = rq_prio(rq); + ve->request = i915_request_get(rq); - list_move_tail(&rq->sched.link, virtual_queue(ve)); + GEM_BUG_ON(!list_empty(virtual_queue(ve))); + list_move_tail(&rq->sched.link, virtual_queue(ve)); - tasklet_schedule(&ve->base.execlists.tasklet); + tasklet_schedule(&ve->base.execlists.tasklet); + } + + spin_unlock_irqrestore(&ve->base.active.lock, flags); } static struct ve_bond * @@ -3631,18 +4378,22 @@ static void virtual_bond_execute(struct i915_request *rq, struct dma_fence *signal) { struct virtual_engine *ve = to_virtual_engine(rq->engine); + intel_engine_mask_t allowed, exec; struct ve_bond *bond; + allowed = ~to_request(signal)->engine->mask; + bond = virtual_find_bond(ve, to_request(signal)->engine); - if (bond) { - intel_engine_mask_t old, new, cmp; + if (bond) + allowed &= bond->sibling_mask; - cmp = READ_ONCE(rq->execution_mask); - do { - old = cmp; - new = cmp & bond->sibling_mask; - } while ((cmp = cmpxchg(&rq->execution_mask, old, new)) != old); - } + /* Restrict the bonded request to run on only the available engines */ + exec = READ_ONCE(rq->execution_mask); + while (!try_cmpxchg(&rq->execution_mask, &exec, exec & allowed)) + ; + + /* Prevent the master from being re-run on the bonded engines */ + to_request(signal)->execution_mask &= ~allowed; } struct intel_context * @@ -3666,6 +4417,7 @@ intel_execlists_create_virtual(struct i915_gem_context *ctx, ve->base.i915 = ctx->i915; ve->base.gt = siblings[0]->gt; + ve->base.uncore = siblings[0]->uncore; ve->base.id = -1; ve->base.class = OTHER_CLASS; ve->base.uabi_class = I915_ENGINE_CLASS_INVALID; @@ -3689,6 +4441,7 @@ intel_execlists_create_virtual(struct i915_gem_context *ctx, snprintf(ve->base.name, sizeof(ve->base.name), "virtual"); intel_engine_init_active(&ve->base, ENGINE_VIRTUAL); + intel_engine_init_breadcrumbs(&ve->base); intel_engine_init_execlists(&ve->base); @@ -3851,6 +4604,18 @@ int intel_virtual_engine_attach_bond(struct intel_engine_cs *engine, return 0; } +struct intel_engine_cs * +intel_virtual_engine_get_sibling(struct intel_engine_cs *engine, + unsigned int sibling) +{ + struct virtual_engine *ve = to_virtual_engine(engine); + + if (sibling >= ve->num_siblings) + return NULL; + + return ve->siblings[sibling]; +} + void intel_execlists_show_requests(struct intel_engine_cs *engine, struct drm_printer *m, void (*show_request)(struct drm_printer *m, @@ -3939,6 +4704,8 @@ void intel_lr_context_reset(struct intel_engine_cs *engine, u32 head, bool scrub) { + GEM_BUG_ON(!intel_context_is_pinned(ce)); + /* * We want a simple context + ring to execute the breadcrumb update. * We cannot rely on the context being intact across the GPU hang, @@ -3947,16 +4714,8 @@ void intel_lr_context_reset(struct intel_engine_cs *engine, * future request will be after userspace has had the opportunity * to recreate its own state. */ - if (scrub) { - u32 *regs = ce->lrc_reg_state; - - if (engine->pinned_default_state) { - memcpy(regs, /* skip restoring the vanilla PPHWSP */ - engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE, - engine->context_size - PAGE_SIZE); - } - execlists_init_reg_state(regs, ce, engine, ce->ring); - } + if (scrub) + restore_default_state(ce, engine); /* Rerun the request; its payload has been neutered (if guilty). */ ce->ring->head = head; @@ -3965,6 +4724,13 @@ void intel_lr_context_reset(struct intel_engine_cs *engine, __execlists_update_reg_state(ce, engine); } +bool +intel_engine_in_execlists_submission_mode(const struct intel_engine_cs *engine) +{ + return engine->set_default_submission == + intel_execlists_set_default_submission; +} + #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) #include "selftest_lrc.c" #endif diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h index c2bba82bcc16..04511d8ebdc1 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.h +++ b/drivers/gpu/drm/i915/gt/intel_lrc.h @@ -43,6 +43,7 @@ struct intel_engine_cs; #define CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT (1 << 0) #define CTX_CTRL_RS_CTX_ENABLE (1 << 1) #define CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT (1 << 2) +#define GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE (1 << 8) #define RING_CONTEXT_STATUS_PTR(base) _MMIO((base) + 0x3a0) #define RING_EXECLIST_SQ_CONTENTS(base) _MMIO((base) + 0x510) #define RING_EXECLIST_CONTROL(base) _MMIO((base) + 0x550) @@ -66,6 +67,12 @@ struct intel_engine_cs; #define GEN11_CSB_READ_PTR_MASK (GEN11_CSB_PTR_MASK << 8) #define GEN11_CSB_WRITE_PTR_MASK (GEN11_CSB_PTR_MASK << 0) +#define MAX_CONTEXT_HW_ID (1<<21) /* exclusive */ +#define MAX_GUC_CONTEXT_HW_ID (1 << 20) /* exclusive */ +#define GEN11_MAX_CONTEXT_HW_ID (1<<11) /* exclusive */ +/* in Gen12 ID 0x7FF is reserved to indicate idle */ +#define GEN12_MAX_CONTEXT_HW_ID (GEN11_MAX_CONTEXT_HW_ID - 1) + enum { INTEL_CONTEXT_SCHEDULE_IN = 0, INTEL_CONTEXT_SCHEDULE_OUT, @@ -79,30 +86,15 @@ int intel_execlists_submission_setup(struct intel_engine_cs *engine); int intel_execlists_submission_init(struct intel_engine_cs *engine); /* Logical Ring Contexts */ - -/* - * We allocate a header at the start of the context image for our own - * use, therefore the actual location of the logical state is offset - * from the start of the VMA. The layout is - * - * | [guc] | [hwsp] [logical state] | - * |<- our header ->|<- context image ->| - * - */ -/* The first page is used for sharing data with the GuC */ -#define LRC_GUCSHR_PN (0) -#define LRC_GUCSHR_SZ (1) /* At the start of the context image is its per-process HWS page */ -#define LRC_PPHWSP_PN (LRC_GUCSHR_PN + LRC_GUCSHR_SZ) +#define LRC_PPHWSP_PN (0) #define LRC_PPHWSP_SZ (1) -/* Finally we have the logical state for the context */ +/* After the PPHWSP we have the logical state for the context */ #define LRC_STATE_PN (LRC_PPHWSP_PN + LRC_PPHWSP_SZ) -/* - * Currently we include the PPHWSP in __intel_engine_context_size() so - * the size of the header is synonymous with the start of the PPHWSP. - */ -#define LRC_HEADER_PAGES LRC_PPHWSP_PN +/* Space within PPHWSP reserved to be used as scratch */ +#define LRC_PPHWSP_SCRATCH 0x34 +#define LRC_PPHWSP_SCRATCH_ADDR (LRC_PPHWSP_SCRATCH * sizeof(u32)) void intel_execlists_set_default_submission(struct intel_engine_cs *engine); @@ -131,4 +123,11 @@ int intel_virtual_engine_attach_bond(struct intel_engine_cs *engine, const struct intel_engine_cs *master, const struct intel_engine_cs *sibling); +struct intel_engine_cs * +intel_virtual_engine_get_sibling(struct intel_engine_cs *engine, + unsigned int sibling); + +bool +intel_engine_in_execlists_submission_mode(const struct intel_engine_cs *engine); + #endif /* _INTEL_LRC_H_ */ diff --git a/drivers/gpu/drm/i915/gt/intel_lrc_reg.h b/drivers/gpu/drm/i915/gt/intel_lrc_reg.h index b8f20ad71169..06ab0276e10e 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc_reg.h +++ b/drivers/gpu/drm/i915/gt/intel_lrc_reg.h @@ -9,55 +9,41 @@ #include <linux/types.h> -/* GEN8+ Reg State Context */ -#define CTX_LRI_HEADER_0 0x01 -#define CTX_CONTEXT_CONTROL 0x02 -#define CTX_RING_HEAD 0x04 -#define CTX_RING_TAIL 0x06 -#define CTX_RING_BUFFER_START 0x08 -#define CTX_RING_BUFFER_CONTROL 0x0a -#define CTX_BB_HEAD_U 0x0c -#define CTX_BB_HEAD_L 0x0e -#define CTX_BB_STATE 0x10 -#define CTX_SECOND_BB_HEAD_U 0x12 -#define CTX_SECOND_BB_HEAD_L 0x14 -#define CTX_SECOND_BB_STATE 0x16 -#define CTX_BB_PER_CTX_PTR 0x18 -#define CTX_RCS_INDIRECT_CTX 0x1a -#define CTX_RCS_INDIRECT_CTX_OFFSET 0x1c -#define CTX_LRI_HEADER_1 0x21 -#define CTX_CTX_TIMESTAMP 0x22 -#define CTX_PDP3_UDW 0x24 -#define CTX_PDP3_LDW 0x26 -#define CTX_PDP2_UDW 0x28 -#define CTX_PDP2_LDW 0x2a -#define CTX_PDP1_UDW 0x2c -#define CTX_PDP1_LDW 0x2e -#define CTX_PDP0_UDW 0x30 -#define CTX_PDP0_LDW 0x32 -#define CTX_LRI_HEADER_2 0x41 -#define CTX_R_PWR_CLK_STATE 0x42 -#define CTX_END 0x44 - -#define CTX_REG(reg_state, pos, reg, val) do { \ - u32 *reg_state__ = (reg_state); \ - const u32 pos__ = (pos); \ - (reg_state__)[(pos__) + 0] = i915_mmio_reg_offset(reg); \ - (reg_state__)[(pos__) + 1] = (val); \ -} while (0) +/* GEN8 to GEN11 Reg State Context */ +#define CTX_CONTEXT_CONTROL (0x02 + 1) +#define CTX_RING_HEAD (0x04 + 1) +#define CTX_RING_TAIL (0x06 + 1) +#define CTX_RING_BUFFER_START (0x08 + 1) +#define CTX_RING_BUFFER_CONTROL (0x0a + 1) +#define CTX_BB_STATE (0x10 + 1) +#define CTX_BB_PER_CTX_PTR (0x18 + 1) +#define CTX_PDP3_UDW (0x24 + 1) +#define CTX_PDP3_LDW (0x26 + 1) +#define CTX_PDP2_UDW (0x28 + 1) +#define CTX_PDP2_LDW (0x2a + 1) +#define CTX_PDP1_UDW (0x2c + 1) +#define CTX_PDP1_LDW (0x2e + 1) +#define CTX_PDP0_UDW (0x30 + 1) +#define CTX_PDP0_LDW (0x32 + 1) +#define CTX_R_PWR_CLK_STATE (0x42 + 1) + +#define GEN9_CTX_RING_MI_MODE 0x54 + +/* GEN12+ Reg State Context */ +#define GEN12_CTX_BB_PER_CTX_PTR (0x12 + 1) #define ASSIGN_CTX_PDP(ppgtt, reg_state, n) do { \ u32 *reg_state__ = (reg_state); \ const u64 addr__ = i915_page_dir_dma_addr((ppgtt), (n)); \ - (reg_state__)[CTX_PDP ## n ## _UDW + 1] = upper_32_bits(addr__); \ - (reg_state__)[CTX_PDP ## n ## _LDW + 1] = lower_32_bits(addr__); \ + (reg_state__)[CTX_PDP ## n ## _UDW] = upper_32_bits(addr__); \ + (reg_state__)[CTX_PDP ## n ## _LDW] = lower_32_bits(addr__); \ } while (0) #define ASSIGN_CTX_PML4(ppgtt, reg_state) do { \ u32 *reg_state__ = (reg_state); \ const u64 addr__ = px_dma(ppgtt->pd); \ - (reg_state__)[CTX_PDP0_UDW + 1] = upper_32_bits(addr__); \ - (reg_state__)[CTX_PDP0_LDW + 1] = lower_32_bits(addr__); \ + (reg_state__)[CTX_PDP0_UDW] = upper_32_bits(addr__); \ + (reg_state__)[CTX_PDP0_LDW] = lower_32_bits(addr__); \ } while (0) #define GEN8_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT 0x17 diff --git a/drivers/gpu/drm/i915/gt/intel_mocs.c b/drivers/gpu/drm/i915/gt/intel_mocs.c index 728704bbbe18..2b977991b785 100644 --- a/drivers/gpu/drm/i915/gt/intel_mocs.c +++ b/drivers/gpu/drm/i915/gt/intel_mocs.c @@ -26,6 +26,7 @@ #include "intel_gt.h" #include "intel_mocs.h" #include "intel_lrc.h" +#include "intel_ring.h" /* structures required */ struct drm_i915_mocs_entry { @@ -199,14 +200,6 @@ static const struct drm_i915_mocs_entry broxton_mocs_table[] = { MOCS_ENTRY(15, \ LE_3_WB | LE_TC_1_LLC | LE_LRUM(2) | LE_AOM(1), \ L3_3_WB), \ - /* Bypass LLC - Uncached (EHL+) */ \ - MOCS_ENTRY(16, \ - LE_1_UC | LE_TC_1_LLC | LE_SCF(1), \ - L3_1_UC), \ - /* Bypass LLC - L3 (Read-Only) (EHL+) */ \ - MOCS_ENTRY(17, \ - LE_1_UC | LE_TC_1_LLC | LE_SCF(1), \ - L3_3_WB), \ /* Self-Snoop - L3 + LLC */ \ MOCS_ENTRY(18, \ LE_3_WB | LE_TC_1_LLC | LE_LRUM(3) | LE_SSE(3), \ @@ -270,7 +263,7 @@ static const struct drm_i915_mocs_entry tigerlake_mocs_table[] = { L3_1_UC), /* HW Special Case (Displayable) */ MOCS_ENTRY(61, - LE_1_UC | LE_TC_1_LLC | LE_SCF(1), + LE_1_UC | LE_TC_1_LLC, L3_3_WB), }; @@ -287,10 +280,9 @@ static const struct drm_i915_mocs_entry icelake_mocs_table[] = { GEN11_MOCS_ENTRIES }; -static bool get_mocs_settings(struct intel_gt *gt, +static bool get_mocs_settings(const struct drm_i915_private *i915, struct drm_i915_mocs_table *table) { - struct drm_i915_private *i915 = gt->i915; bool result = false; if (INTEL_GEN(i915) >= 12) { @@ -331,9 +323,9 @@ static bool get_mocs_settings(struct intel_gt *gt, return result; } -static i915_reg_t mocs_register(enum intel_engine_id engine_id, int index) +static i915_reg_t mocs_register(const struct intel_engine_cs *engine, int index) { - switch (engine_id) { + switch (engine->id) { case RCS0: return GEN9_GFX_MOCS(index); case VCS0: @@ -347,7 +339,7 @@ static i915_reg_t mocs_register(enum intel_engine_id engine_id, int index) case VCS2: return GEN11_MFX2_MOCS(index); default: - MISSING_CASE(engine_id); + MISSING_CASE(engine->id); return INVALID_MMIO_REG; } } @@ -365,118 +357,25 @@ static u32 get_entry_control(const struct drm_i915_mocs_table *table, return table->table[I915_MOCS_PTE].control_value; } -/** - * intel_mocs_init_engine() - emit the mocs control table - * @engine: The engine for whom to emit the registers. - * - * This function simply emits a MI_LOAD_REGISTER_IMM command for the - * given table starting at the given address. - */ -void intel_mocs_init_engine(struct intel_engine_cs *engine) +static void init_mocs_table(struct intel_engine_cs *engine, + const struct drm_i915_mocs_table *table) { - struct intel_gt *gt = engine->gt; - struct intel_uncore *uncore = gt->uncore; - struct drm_i915_mocs_table table; - unsigned int index; - u32 unused_value; - - /* Platforms with global MOCS do not need per-engine initialization. */ - if (HAS_GLOBAL_MOCS_REGISTERS(gt->i915)) - return; - - /* Called under a blanket forcewake */ - assert_forcewakes_active(uncore, FORCEWAKE_ALL); - - if (!get_mocs_settings(gt, &table)) - return; - - /* Set unused values to PTE */ - unused_value = table.table[I915_MOCS_PTE].control_value; - - for (index = 0; index < table.size; index++) { - u32 value = get_entry_control(&table, index); + struct intel_uncore *uncore = engine->uncore; + u32 unused_value = table->table[I915_MOCS_PTE].control_value; + unsigned int i; + for (i = 0; i < table->size; i++) intel_uncore_write_fw(uncore, - mocs_register(engine->id, index), - value); - } + mocs_register(engine, i), + get_entry_control(table, i)); - /* All remaining entries are also unused */ - for (; index < table.n_entries; index++) + /* All remaining entries are unused */ + for (; i < table->n_entries; i++) intel_uncore_write_fw(uncore, - mocs_register(engine->id, index), + mocs_register(engine, i), unused_value); } -static void intel_mocs_init_global(struct intel_gt *gt) -{ - struct intel_uncore *uncore = gt->uncore; - struct drm_i915_mocs_table table; - unsigned int index; - - GEM_BUG_ON(!HAS_GLOBAL_MOCS_REGISTERS(gt->i915)); - - if (!get_mocs_settings(gt, &table)) - return; - - if (GEM_DEBUG_WARN_ON(table.size > table.n_entries)) - return; - - for (index = 0; index < table.size; index++) - intel_uncore_write(uncore, - GEN12_GLOBAL_MOCS(index), - table.table[index].control_value); - - /* - * Ok, now set the unused entries to the invalid entry (index 0). These - * entries are officially undefined and no contract for the contents and - * settings is given for these entries. - */ - for (; index < table.n_entries; index++) - intel_uncore_write(uncore, - GEN12_GLOBAL_MOCS(index), - table.table[0].control_value); -} - -static int emit_mocs_control_table(struct i915_request *rq, - const struct drm_i915_mocs_table *table) -{ - enum intel_engine_id engine = rq->engine->id; - unsigned int index; - u32 unused_value; - u32 *cs; - - if (GEM_WARN_ON(table->size > table->n_entries)) - return -ENODEV; - - /* Set unused values to PTE */ - unused_value = table->table[I915_MOCS_PTE].control_value; - - cs = intel_ring_begin(rq, 2 + 2 * table->n_entries); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - *cs++ = MI_LOAD_REGISTER_IMM(table->n_entries); - - for (index = 0; index < table->size; index++) { - u32 value = get_entry_control(table, index); - - *cs++ = i915_mmio_reg_offset(mocs_register(engine, index)); - *cs++ = value; - } - - /* All remaining entries are also unused */ - for (; index < table->n_entries; index++) { - *cs++ = i915_mmio_reg_offset(mocs_register(engine, index)); - *cs++ = unused_value; - } - - *cs++ = MI_NOOP; - intel_ring_advance(rq, cs); - - return 0; -} - /* * Get l3cc_value from MOCS entry taking into account when it's not used: * I915_MOCS_PTE's value is returned in this case. @@ -494,141 +393,99 @@ static inline u32 l3cc_combine(const struct drm_i915_mocs_table *table, u16 low, u16 high) { - return low | high << 16; + return low | (u32)high << 16; } -static int emit_mocs_l3cc_table(struct i915_request *rq, - const struct drm_i915_mocs_table *table) +static void init_l3cc_table(struct intel_engine_cs *engine, + const struct drm_i915_mocs_table *table) { - u16 unused_value; + struct intel_uncore *uncore = engine->uncore; + u16 unused_value = table->table[I915_MOCS_PTE].l3cc_value; unsigned int i; - u32 *cs; - - if (GEM_WARN_ON(table->size > table->n_entries)) - return -ENODEV; - - /* Set unused values to PTE */ - unused_value = table->table[I915_MOCS_PTE].l3cc_value; - - cs = intel_ring_begin(rq, 2 + table->n_entries); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - *cs++ = MI_LOAD_REGISTER_IMM(table->n_entries / 2); for (i = 0; i < table->size / 2; i++) { u16 low = get_entry_l3cc(table, 2 * i); u16 high = get_entry_l3cc(table, 2 * i + 1); - *cs++ = i915_mmio_reg_offset(GEN9_LNCFCMOCS(i)); - *cs++ = l3cc_combine(table, low, high); + intel_uncore_write(uncore, + GEN9_LNCFCMOCS(i), + l3cc_combine(table, low, high)); } /* Odd table size - 1 left over */ - if (table->size & 0x01) { + if (table->size & 1) { u16 low = get_entry_l3cc(table, 2 * i); - *cs++ = i915_mmio_reg_offset(GEN9_LNCFCMOCS(i)); - *cs++ = l3cc_combine(table, low, unused_value); + intel_uncore_write(uncore, + GEN9_LNCFCMOCS(i), + l3cc_combine(table, low, unused_value)); i++; } /* All remaining entries are also unused */ - for (; i < table->n_entries / 2; i++) { - *cs++ = i915_mmio_reg_offset(GEN9_LNCFCMOCS(i)); - *cs++ = l3cc_combine(table, unused_value, unused_value); - } + for (; i < table->n_entries / 2; i++) + intel_uncore_write(uncore, + GEN9_LNCFCMOCS(i), + l3cc_combine(table, unused_value, + unused_value)); +} - *cs++ = MI_NOOP; - intel_ring_advance(rq, cs); +void intel_mocs_init_engine(struct intel_engine_cs *engine) +{ + struct drm_i915_mocs_table table; + + /* Called under a blanket forcewake */ + assert_forcewakes_active(engine->uncore, FORCEWAKE_ALL); + + if (!get_mocs_settings(engine->i915, &table)) + return; + + /* Platforms with global MOCS do not need per-engine initialization. */ + if (!HAS_GLOBAL_MOCS_REGISTERS(engine->i915)) + init_mocs_table(engine, &table); - return 0; + if (engine->class == RENDER_CLASS) + init_l3cc_table(engine, &table); } -static void intel_mocs_init_l3cc_table(struct intel_gt *gt) +static void intel_mocs_init_global(struct intel_gt *gt) { struct intel_uncore *uncore = gt->uncore; struct drm_i915_mocs_table table; - unsigned int i; - u16 unused_value; + unsigned int index; - if (!get_mocs_settings(gt, &table)) + /* + * LLC and eDRAM control values are not applicable to dgfx + */ + if (IS_DGFX(gt->i915)) return; - /* Set unused values to PTE */ - unused_value = table.table[I915_MOCS_PTE].l3cc_value; - - for (i = 0; i < table.size / 2; i++) { - u16 low = get_entry_l3cc(&table, 2 * i); - u16 high = get_entry_l3cc(&table, 2 * i + 1); + GEM_BUG_ON(!HAS_GLOBAL_MOCS_REGISTERS(gt->i915)); - intel_uncore_write(uncore, - GEN9_LNCFCMOCS(i), - l3cc_combine(&table, low, high)); - } + if (!get_mocs_settings(gt->i915, &table)) + return; - /* Odd table size - 1 left over */ - if (table.size & 0x01) { - u16 low = get_entry_l3cc(&table, 2 * i); + if (GEM_DEBUG_WARN_ON(table.size > table.n_entries)) + return; + for (index = 0; index < table.size; index++) intel_uncore_write(uncore, - GEN9_LNCFCMOCS(i), - l3cc_combine(&table, low, unused_value)); - i++; - } + GEN12_GLOBAL_MOCS(index), + table.table[index].control_value); - /* All remaining entries are also unused */ - for (; i < table.n_entries / 2; i++) + /* + * Ok, now set the unused entries to the invalid entry (index 0). These + * entries are officially undefined and no contract for the contents and + * settings is given for these entries. + */ + for (; index < table.n_entries; index++) intel_uncore_write(uncore, - GEN9_LNCFCMOCS(i), - l3cc_combine(&table, unused_value, - unused_value)); -} - -/** - * intel_mocs_emit() - program the MOCS register. - * @rq: Request to use to set up the MOCS tables. - * - * This function will emit a batch buffer with the values required for - * programming the MOCS register values for all the currently supported - * rings. - * - * These registers are partially stored in the RCS context, so they are - * emitted at the same time so that when a context is created these registers - * are set up. These registers have to be emitted into the start of the - * context as setting the ELSP will re-init some of these registers back - * to the hw values. - * - * Return: 0 on success, otherwise the error status. - */ -int intel_mocs_emit(struct i915_request *rq) -{ - struct drm_i915_mocs_table t; - int ret; - - if (HAS_GLOBAL_MOCS_REGISTERS(rq->i915) || - rq->engine->class != RENDER_CLASS) - return 0; - - if (get_mocs_settings(rq->engine->gt, &t)) { - /* Program the RCS control registers */ - ret = emit_mocs_control_table(rq, &t); - if (ret) - return ret; - - /* Now program the l3cc registers */ - ret = emit_mocs_l3cc_table(rq, &t); - if (ret) - return ret; - } - - return 0; + GEN12_GLOBAL_MOCS(index), + table.table[0].control_value); } void intel_mocs_init(struct intel_gt *gt) { - intel_mocs_init_l3cc_table(gt); - if (HAS_GLOBAL_MOCS_REGISTERS(gt->i915)) intel_mocs_init_global(gt); } diff --git a/drivers/gpu/drm/i915/gt/intel_mocs.h b/drivers/gpu/drm/i915/gt/intel_mocs.h index 2ae816b7ca19..83371f3e6ba1 100644 --- a/drivers/gpu/drm/i915/gt/intel_mocs.h +++ b/drivers/gpu/drm/i915/gt/intel_mocs.h @@ -49,13 +49,10 @@ * context handling keep the MOCS in step. */ -struct i915_request; struct intel_engine_cs; struct intel_gt; void intel_mocs_init(struct intel_gt *gt); void intel_mocs_init_engine(struct intel_engine_cs *engine); -int intel_mocs_emit(struct i915_request *rq); - #endif diff --git a/drivers/gpu/drm/i915/gt/intel_rc6.c b/drivers/gpu/drm/i915/gt/intel_rc6.c new file mode 100644 index 000000000000..700104b90163 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rc6.c @@ -0,0 +1,787 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include <linux/pm_runtime.h> + +#include "i915_drv.h" +#include "intel_gt.h" +#include "intel_gt_pm.h" +#include "intel_rc6.h" +#include "intel_sideband.h" + +/** + * DOC: RC6 + * + * RC6 is a special power stage which allows the GPU to enter an very + * low-voltage mode when idle, using down to 0V while at this stage. This + * stage is entered automatically when the GPU is idle when RC6 support is + * enabled, and as soon as new workload arises GPU wakes up automatically as + * well. + * + * There are different RC6 modes available in Intel GPU, which differentiate + * among each other with the latency required to enter and leave RC6 and + * voltage consumed by the GPU in different states. + * + * The combination of the following flags define which states GPU is allowed + * to enter, while RC6 is the normal RC6 state, RC6p is the deep RC6, and + * RC6pp is deepest RC6. Their support by hardware varies according to the + * GPU, BIOS, chipset and platform. RC6 is usually the safest one and the one + * which brings the most power savings; deeper states save more power, but + * require higher latency to switch to and wake up. + */ + +static struct intel_gt *rc6_to_gt(struct intel_rc6 *rc6) +{ + return container_of(rc6, struct intel_gt, rc6); +} + +static struct intel_uncore *rc6_to_uncore(struct intel_rc6 *rc) +{ + return rc6_to_gt(rc)->uncore; +} + +static struct drm_i915_private *rc6_to_i915(struct intel_rc6 *rc) +{ + return rc6_to_gt(rc)->i915; +} + +static inline void set(struct intel_uncore *uncore, i915_reg_t reg, u32 val) +{ + intel_uncore_write_fw(uncore, reg, val); +} + +static void gen11_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + /* 2b: Program RC6 thresholds.*/ + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85); + set(uncore, GEN10_MEDIA_WAKE_RATE_LIMIT, 150); + + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + for_each_engine(engine, rc6_to_gt(rc6), id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + + set(uncore, GUC_MAX_IDLE_COUNT, 0xA); + + set(uncore, GEN6_RC_SLEEP, 0); + + set(uncore, GEN6_RC6_THRESHOLD, 50000); /* 50/125ms per EI */ + + /* + * 2c: Program Coarse Power Gating Policies. + * + * Bspec's guidance is to use 25us (really 25 * 1280ns) here. What we + * use instead is a more conservative estimate for the maximum time + * it takes us to service a CS interrupt and submit a new ELSP - that + * is the time which the GPU is idle waiting for the CPU to select the + * next request to execute. If the idle hysteresis is less than that + * interrupt service latency, the hardware will automatically gate + * the power well and we will then incur the wake up cost on top of + * the service latency. A similar guide from plane_state is that we + * do not want the enable hysteresis to less than the wakeup latency. + * + * igt/gem_exec_nop/sequential provides a rough estimate for the + * service latency, and puts it around 10us for Broadwell (and other + * big core) and around 40us for Broxton (and other low power cores). + * [Note that for legacy ringbuffer submission, this is less than 1us!] + * However, the wakeup latency on Broxton is closer to 100us. To be + * conservative, we have to factor in a context switch on top (due + * to ksoftirqd). + */ + set(uncore, GEN9_MEDIA_PG_IDLE_HYSTERESIS, 250); + set(uncore, GEN9_RENDER_PG_IDLE_HYSTERESIS, 250); + + /* 3a: Enable RC6 */ + set(uncore, GEN6_RC_CONTROL, + GEN6_RC_CTL_HW_ENABLE | + GEN6_RC_CTL_RC6_ENABLE | + GEN6_RC_CTL_EI_MODE(1)); + + set(uncore, GEN9_PG_ENABLE, + GEN9_RENDER_PG_ENABLE | + GEN9_MEDIA_PG_ENABLE | + GEN11_MEDIA_SAMPLER_PG_ENABLE); +} + +static void gen9_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + u32 rc6_mode; + + /* 2b: Program RC6 thresholds.*/ + if (INTEL_GEN(rc6_to_i915(rc6)) >= 10) { + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85); + set(uncore, GEN10_MEDIA_WAKE_RATE_LIMIT, 150); + } else if (IS_SKYLAKE(rc6_to_i915(rc6))) { + /* + * WaRsDoubleRc6WrlWithCoarsePowerGating:skl Doubling WRL only + * when CPG is enabled + */ + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 108 << 16); + } else { + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16); + } + + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + for_each_engine(engine, rc6_to_gt(rc6), id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + + set(uncore, GUC_MAX_IDLE_COUNT, 0xA); + + set(uncore, GEN6_RC_SLEEP, 0); + + /* + * 2c: Program Coarse Power Gating Policies. + * + * Bspec's guidance is to use 25us (really 25 * 1280ns) here. What we + * use instead is a more conservative estimate for the maximum time + * it takes us to service a CS interrupt and submit a new ELSP - that + * is the time which the GPU is idle waiting for the CPU to select the + * next request to execute. If the idle hysteresis is less than that + * interrupt service latency, the hardware will automatically gate + * the power well and we will then incur the wake up cost on top of + * the service latency. A similar guide from plane_state is that we + * do not want the enable hysteresis to less than the wakeup latency. + * + * igt/gem_exec_nop/sequential provides a rough estimate for the + * service latency, and puts it around 10us for Broadwell (and other + * big core) and around 40us for Broxton (and other low power cores). + * [Note that for legacy ringbuffer submission, this is less than 1us!] + * However, the wakeup latency on Broxton is closer to 100us. To be + * conservative, we have to factor in a context switch on top (due + * to ksoftirqd). + */ + set(uncore, GEN9_MEDIA_PG_IDLE_HYSTERESIS, 250); + set(uncore, GEN9_RENDER_PG_IDLE_HYSTERESIS, 250); + + /* 3a: Enable RC6 */ + set(uncore, GEN6_RC6_THRESHOLD, 37500); /* 37.5/125ms per EI */ + + /* WaRsUseTimeoutMode:cnl (pre-prod) */ + if (IS_CNL_REVID(rc6_to_i915(rc6), CNL_REVID_A0, CNL_REVID_C0)) + rc6_mode = GEN7_RC_CTL_TO_MODE; + else + rc6_mode = GEN6_RC_CTL_EI_MODE(1); + + set(uncore, GEN6_RC_CONTROL, + GEN6_RC_CTL_HW_ENABLE | + GEN6_RC_CTL_RC6_ENABLE | + rc6_mode); + + /* + * WaRsDisableCoarsePowerGating:skl,cnl + * - Render/Media PG need to be disabled with RC6. + */ + if (!NEEDS_WaRsDisableCoarsePowerGating(rc6_to_i915(rc6))) + set(uncore, GEN9_PG_ENABLE, + GEN9_RENDER_PG_ENABLE | GEN9_MEDIA_PG_ENABLE); +} + +static void gen8_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + /* 2b: Program RC6 thresholds.*/ + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16); + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + for_each_engine(engine, rc6_to_gt(rc6), id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + set(uncore, GEN6_RC_SLEEP, 0); + set(uncore, GEN6_RC6_THRESHOLD, 625); /* 800us/1.28 for TO */ + + /* 3: Enable RC6 */ + set(uncore, GEN6_RC_CONTROL, + GEN6_RC_CTL_HW_ENABLE | + GEN7_RC_CTL_TO_MODE | + GEN6_RC_CTL_RC6_ENABLE); +} + +static void gen6_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + u32 rc6vids, rc6_mask; + int ret; + + set(uncore, GEN6_RC1_WAKE_RATE_LIMIT, 1000 << 16); + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16 | 30); + set(uncore, GEN6_RC6pp_WAKE_RATE_LIMIT, 30); + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); + + for_each_engine(engine, rc6_to_gt(rc6), id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + + set(uncore, GEN6_RC_SLEEP, 0); + set(uncore, GEN6_RC1e_THRESHOLD, 1000); + if (IS_IVYBRIDGE(i915)) + set(uncore, GEN6_RC6_THRESHOLD, 125000); + else + set(uncore, GEN6_RC6_THRESHOLD, 50000); + set(uncore, GEN6_RC6p_THRESHOLD, 150000); + set(uncore, GEN6_RC6pp_THRESHOLD, 64000); /* unused */ + + /* We don't use those on Haswell */ + rc6_mask = GEN6_RC_CTL_RC6_ENABLE; + if (HAS_RC6p(i915)) + rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE; + if (HAS_RC6pp(i915)) + rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE; + set(uncore, GEN6_RC_CONTROL, + rc6_mask | + GEN6_RC_CTL_EI_MODE(1) | + GEN6_RC_CTL_HW_ENABLE); + + rc6vids = 0; + ret = sandybridge_pcode_read(i915, GEN6_PCODE_READ_RC6VIDS, + &rc6vids, NULL); + if (IS_GEN(i915, 6) && ret) { + DRM_DEBUG_DRIVER("Couldn't check for BIOS workaround\n"); + } else if (IS_GEN(i915, 6) && + (GEN6_DECODE_RC6_VID(rc6vids & 0xff) < 450)) { + DRM_DEBUG_DRIVER("You should update your BIOS. Correcting minimum rc6 voltage (%dmV->%dmV)\n", + GEN6_DECODE_RC6_VID(rc6vids & 0xff), 450); + rc6vids &= 0xffff00; + rc6vids |= GEN6_ENCODE_RC6_VID(450); + ret = sandybridge_pcode_write(i915, GEN6_PCODE_WRITE_RC6VIDS, rc6vids); + if (ret) + DRM_ERROR("Couldn't fix incorrect rc6 voltage\n"); + } +} + +/* Check that the pcbr address is not empty. */ +static int chv_rc6_init(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + resource_size_t pctx_paddr, paddr; + resource_size_t pctx_size = 32 * SZ_1K; + u32 pcbr; + + pcbr = intel_uncore_read(uncore, VLV_PCBR); + if ((pcbr >> VLV_PCBR_ADDR_SHIFT) == 0) { + DRM_DEBUG_DRIVER("BIOS didn't set up PCBR, fixing up\n"); + paddr = rc6_to_i915(rc6)->dsm.end + 1 - pctx_size; + GEM_BUG_ON(paddr > U32_MAX); + + pctx_paddr = (paddr & ~4095); + intel_uncore_write(uncore, VLV_PCBR, pctx_paddr); + } + + return 0; +} + +static int vlv_rc6_init(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct drm_i915_gem_object *pctx; + resource_size_t pctx_paddr; + resource_size_t pctx_size = 24 * SZ_1K; + u32 pcbr; + + pcbr = intel_uncore_read(uncore, VLV_PCBR); + if (pcbr) { + /* BIOS set it up already, grab the pre-alloc'd space */ + resource_size_t pcbr_offset; + + pcbr_offset = (pcbr & ~4095) - i915->dsm.start; + pctx = i915_gem_object_create_stolen_for_preallocated(i915, + pcbr_offset, + I915_GTT_OFFSET_NONE, + pctx_size); + if (IS_ERR(pctx)) + return PTR_ERR(pctx); + + goto out; + } + + DRM_DEBUG_DRIVER("BIOS didn't set up PCBR, fixing up\n"); + + /* + * From the Gunit register HAS: + * The Gfx driver is expected to program this register and ensure + * proper allocation within Gfx stolen memory. For example, this + * register should be programmed such than the PCBR range does not + * overlap with other ranges, such as the frame buffer, protected + * memory, or any other relevant ranges. + */ + pctx = i915_gem_object_create_stolen(i915, pctx_size); + if (IS_ERR(pctx)) { + DRM_DEBUG("not enough stolen space for PCTX, disabling\n"); + return PTR_ERR(pctx); + } + + GEM_BUG_ON(range_overflows_t(u64, + i915->dsm.start, + pctx->stolen->start, + U32_MAX)); + pctx_paddr = i915->dsm.start + pctx->stolen->start; + intel_uncore_write(uncore, VLV_PCBR, pctx_paddr); + +out: + rc6->pctx = pctx; + return 0; +} + +static void chv_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + /* 2a: Program RC6 thresholds.*/ + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16); + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + + for_each_engine(engine, rc6_to_gt(rc6), id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + set(uncore, GEN6_RC_SLEEP, 0); + + /* TO threshold set to 500 us (0x186 * 1.28 us) */ + set(uncore, GEN6_RC6_THRESHOLD, 0x186); + + /* Allows RC6 residency counter to work */ + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH | + VLV_MEDIA_RC6_COUNT_EN | + VLV_RENDER_RC6_COUNT_EN)); + + /* 3: Enable RC6 */ + set(uncore, GEN6_RC_CONTROL, GEN7_RC_CTL_TO_MODE); +} + +static void vlv_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 0x00280000); + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); + + for_each_engine(engine, rc6_to_gt(rc6), id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + + set(uncore, GEN6_RC6_THRESHOLD, 0x557); + + /* Allows RC6 residency counter to work */ + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH | + VLV_MEDIA_RC0_COUNT_EN | + VLV_RENDER_RC0_COUNT_EN | + VLV_MEDIA_RC6_COUNT_EN | + VLV_RENDER_RC6_COUNT_EN)); + + set(uncore, GEN6_RC_CONTROL, + GEN7_RC_CTL_TO_MODE | VLV_RC_CTL_CTX_RST_PARALLEL); +} + +static bool bxt_check_bios_rc6_setup(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct drm_i915_private *i915 = rc6_to_i915(rc6); + u32 rc6_ctx_base, rc_ctl, rc_sw_target; + bool enable_rc6 = true; + + rc_ctl = intel_uncore_read(uncore, GEN6_RC_CONTROL); + rc_sw_target = intel_uncore_read(uncore, GEN6_RC_STATE); + rc_sw_target &= RC_SW_TARGET_STATE_MASK; + rc_sw_target >>= RC_SW_TARGET_STATE_SHIFT; + DRM_DEBUG_DRIVER("BIOS enabled RC states: " + "HW_CTRL %s HW_RC6 %s SW_TARGET_STATE %x\n", + onoff(rc_ctl & GEN6_RC_CTL_HW_ENABLE), + onoff(rc_ctl & GEN6_RC_CTL_RC6_ENABLE), + rc_sw_target); + + if (!(intel_uncore_read(uncore, RC6_LOCATION) & RC6_CTX_IN_DRAM)) { + DRM_DEBUG_DRIVER("RC6 Base location not set properly.\n"); + enable_rc6 = false; + } + + /* + * The exact context size is not known for BXT, so assume a page size + * for this check. + */ + rc6_ctx_base = + intel_uncore_read(uncore, RC6_CTX_BASE) & RC6_CTX_BASE_MASK; + if (!(rc6_ctx_base >= i915->dsm_reserved.start && + rc6_ctx_base + PAGE_SIZE < i915->dsm_reserved.end)) { + DRM_DEBUG_DRIVER("RC6 Base address not as expected.\n"); + enable_rc6 = false; + } + + if (!((intel_uncore_read(uncore, PWRCTX_MAXCNT_RCSUNIT) & IDLE_TIME_MASK) > 1 && + (intel_uncore_read(uncore, PWRCTX_MAXCNT_VCSUNIT0) & IDLE_TIME_MASK) > 1 && + (intel_uncore_read(uncore, PWRCTX_MAXCNT_BCSUNIT) & IDLE_TIME_MASK) > 1 && + (intel_uncore_read(uncore, PWRCTX_MAXCNT_VECSUNIT) & IDLE_TIME_MASK) > 1)) { + DRM_DEBUG_DRIVER("Engine Idle wait time not set properly.\n"); + enable_rc6 = false; + } + + if (!intel_uncore_read(uncore, GEN8_PUSHBUS_CONTROL) || + !intel_uncore_read(uncore, GEN8_PUSHBUS_ENABLE) || + !intel_uncore_read(uncore, GEN8_PUSHBUS_SHIFT)) { + DRM_DEBUG_DRIVER("Pushbus not setup properly.\n"); + enable_rc6 = false; + } + + if (!intel_uncore_read(uncore, GEN6_GFXPAUSE)) { + DRM_DEBUG_DRIVER("GFX pause not setup properly.\n"); + enable_rc6 = false; + } + + if (!intel_uncore_read(uncore, GEN8_MISC_CTRL0)) { + DRM_DEBUG_DRIVER("GPM control not setup properly.\n"); + enable_rc6 = false; + } + + return enable_rc6; +} + +static bool rc6_supported(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + + if (!HAS_RC6(i915)) + return false; + + if (intel_vgpu_active(i915)) + return false; + + if (is_mock_gt(rc6_to_gt(rc6))) + return false; + + if (IS_GEN9_LP(i915) && !bxt_check_bios_rc6_setup(rc6)) { + dev_notice(i915->drm.dev, + "RC6 and powersaving disabled by BIOS\n"); + return false; + } + + return true; +} + +static void rpm_get(struct intel_rc6 *rc6) +{ + GEM_BUG_ON(rc6->wakeref); + pm_runtime_get_sync(&rc6_to_i915(rc6)->drm.pdev->dev); + rc6->wakeref = true; +} + +static void rpm_put(struct intel_rc6 *rc6) +{ + GEM_BUG_ON(!rc6->wakeref); + pm_runtime_put(&rc6_to_i915(rc6)->drm.pdev->dev); + rc6->wakeref = false; +} + +static bool intel_rc6_ctx_corrupted(struct intel_rc6 *rc6) +{ + return !intel_uncore_read(rc6_to_uncore(rc6), GEN8_RC6_CTX_INFO); +} + +static void intel_rc6_ctx_wa_init(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + + if (!NEEDS_RC6_CTX_CORRUPTION_WA(i915)) + return; + + if (intel_rc6_ctx_corrupted(rc6)) { + DRM_INFO("RC6 context corrupted, disabling runtime power management\n"); + rc6->ctx_corrupted = true; + } +} + +/** + * intel_rc6_ctx_wa_resume - system resume sequence for the RC6 CTX WA + * @rc6: rc6 state + * + * Perform any steps needed to re-init the RC6 CTX WA after system resume. + */ +void intel_rc6_ctx_wa_resume(struct intel_rc6 *rc6) +{ + if (rc6->ctx_corrupted && !intel_rc6_ctx_corrupted(rc6)) { + DRM_INFO("RC6 context restored, re-enabling runtime power management\n"); + rc6->ctx_corrupted = false; + } +} + +/** + * intel_rc6_ctx_wa_check - check for a new RC6 CTX corruption + * @rc6: rc6 state + * + * Check if an RC6 CTX corruption has happened since the last check and if so + * disable RC6 and runtime power management. +*/ +void intel_rc6_ctx_wa_check(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + + if (!NEEDS_RC6_CTX_CORRUPTION_WA(i915)) + return; + + if (rc6->ctx_corrupted) + return; + + if (!intel_rc6_ctx_corrupted(rc6)) + return; + + DRM_NOTE("RC6 context corruption, disabling runtime power management\n"); + + intel_rc6_disable(rc6); + rc6->ctx_corrupted = true; + + return; +} + +static void __intel_rc6_disable(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_uncore *uncore = rc6_to_uncore(rc6); + + intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); + if (INTEL_GEN(i915) >= 9) + set(uncore, GEN9_PG_ENABLE, 0); + set(uncore, GEN6_RC_CONTROL, 0); + set(uncore, GEN6_RC_STATE, 0); + intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); +} + +void intel_rc6_init(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + int err; + + /* Disable runtime-pm until we can save the GPU state with rc6 pctx */ + rpm_get(rc6); + + if (!rc6_supported(rc6)) + return; + + intel_rc6_ctx_wa_init(rc6); + + if (IS_CHERRYVIEW(i915)) + err = chv_rc6_init(rc6); + else if (IS_VALLEYVIEW(i915)) + err = vlv_rc6_init(rc6); + else + err = 0; + + /* Sanitize rc6, ensure it is disabled before we are ready. */ + __intel_rc6_disable(rc6); + + rc6->supported = err == 0; +} + +void intel_rc6_sanitize(struct intel_rc6 *rc6) +{ + if (rc6->enabled) { /* unbalanced suspend/resume */ + rpm_get(rc6); + rc6->enabled = false; + } + + if (rc6->supported) + __intel_rc6_disable(rc6); +} + +void intel_rc6_enable(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_uncore *uncore = rc6_to_uncore(rc6); + + if (!rc6->supported) + return; + + GEM_BUG_ON(rc6->enabled); + + if (rc6->ctx_corrupted) + return; + + intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); + + if (IS_CHERRYVIEW(i915)) + chv_rc6_enable(rc6); + else if (IS_VALLEYVIEW(i915)) + vlv_rc6_enable(rc6); + else if (INTEL_GEN(i915) >= 11) + gen11_rc6_enable(rc6); + else if (INTEL_GEN(i915) >= 9) + gen9_rc6_enable(rc6); + else if (IS_BROADWELL(i915)) + gen8_rc6_enable(rc6); + else if (INTEL_GEN(i915) >= 6) + gen6_rc6_enable(rc6); + + intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); + + /* rc6 is ready, runtime-pm is go! */ + rpm_put(rc6); + rc6->enabled = true; +} + +void intel_rc6_disable(struct intel_rc6 *rc6) +{ + if (!rc6->enabled) + return; + + rpm_get(rc6); + rc6->enabled = false; + + __intel_rc6_disable(rc6); +} + +void intel_rc6_fini(struct intel_rc6 *rc6) +{ + struct drm_i915_gem_object *pctx; + + intel_rc6_disable(rc6); + + pctx = fetch_and_zero(&rc6->pctx); + if (pctx) + i915_gem_object_put(pctx); + + if (rc6->wakeref) + rpm_put(rc6); +} + +static u64 vlv_residency_raw(struct intel_uncore *uncore, const i915_reg_t reg) +{ + u32 lower, upper, tmp; + int loop = 2; + + /* + * The register accessed do not need forcewake. We borrow + * uncore lock to prevent concurrent access to range reg. + */ + lockdep_assert_held(&uncore->lock); + + /* + * vlv and chv residency counters are 40 bits in width. + * With a control bit, we can choose between upper or lower + * 32bit window into this counter. + * + * Although we always use the counter in high-range mode elsewhere, + * userspace may attempt to read the value before rc6 is initialised, + * before we have set the default VLV_COUNTER_CONTROL value. So always + * set the high bit to be safe. + */ + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH)); + upper = intel_uncore_read_fw(uncore, reg); + do { + tmp = upper; + + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_DISABLE(VLV_COUNT_RANGE_HIGH)); + lower = intel_uncore_read_fw(uncore, reg); + + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH)); + upper = intel_uncore_read_fw(uncore, reg); + } while (upper != tmp && --loop); + + /* + * Everywhere else we always use VLV_COUNTER_CONTROL with the + * VLV_COUNT_RANGE_HIGH bit set - so it is safe to leave it set + * now. + */ + + return lower | (u64)upper << 8; +} + +u64 intel_rc6_residency_ns(struct intel_rc6 *rc6, const i915_reg_t reg) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_uncore *uncore = rc6_to_uncore(rc6); + u64 time_hw, prev_hw, overflow_hw; + unsigned int fw_domains; + unsigned long flags; + unsigned int i; + u32 mul, div; + + if (!rc6->supported) + return 0; + + /* + * Store previous hw counter values for counter wrap-around handling. + * + * There are only four interesting registers and they live next to each + * other so we can use the relative address, compared to the smallest + * one as the index into driver storage. + */ + i = (i915_mmio_reg_offset(reg) - + i915_mmio_reg_offset(GEN6_GT_GFX_RC6_LOCKED)) / sizeof(u32); + if (WARN_ON_ONCE(i >= ARRAY_SIZE(rc6->cur_residency))) + return 0; + + fw_domains = intel_uncore_forcewake_for_reg(uncore, reg, FW_REG_READ); + + spin_lock_irqsave(&uncore->lock, flags); + intel_uncore_forcewake_get__locked(uncore, fw_domains); + + /* On VLV and CHV, residency time is in CZ units rather than 1.28us */ + if (IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915)) { + mul = 1000000; + div = i915->czclk_freq; + overflow_hw = BIT_ULL(40); + time_hw = vlv_residency_raw(uncore, reg); + } else { + /* 833.33ns units on Gen9LP, 1.28us elsewhere. */ + if (IS_GEN9_LP(i915)) { + mul = 10000; + div = 12; + } else { + mul = 1280; + div = 1; + } + + overflow_hw = BIT_ULL(32); + time_hw = intel_uncore_read_fw(uncore, reg); + } + + /* + * Counter wrap handling. + * + * But relying on a sufficient frequency of queries otherwise counters + * can still wrap. + */ + prev_hw = rc6->prev_hw_residency[i]; + rc6->prev_hw_residency[i] = time_hw; + + /* RC6 delta from last sample. */ + if (time_hw >= prev_hw) + time_hw -= prev_hw; + else + time_hw += overflow_hw - prev_hw; + + /* Add delta to RC6 extended raw driver copy. */ + time_hw += rc6->cur_residency[i]; + rc6->cur_residency[i] = time_hw; + + intel_uncore_forcewake_put__locked(uncore, fw_domains); + spin_unlock_irqrestore(&uncore->lock, flags); + + return mul_u64_u32_div(time_hw, mul, div); +} + +u64 intel_rc6_residency_us(struct intel_rc6 *rc6, i915_reg_t reg) +{ + return DIV_ROUND_UP_ULL(intel_rc6_residency_ns(rc6, reg), 1000); +} diff --git a/drivers/gpu/drm/i915/gt/intel_rc6.h b/drivers/gpu/drm/i915/gt/intel_rc6.h new file mode 100644 index 000000000000..1370f6834a4c --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rc6.h @@ -0,0 +1,28 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_RC6_H +#define INTEL_RC6_H + +#include "i915_reg.h" + +struct intel_engine_cs; +struct intel_rc6; + +void intel_rc6_init(struct intel_rc6 *rc6); +void intel_rc6_fini(struct intel_rc6 *rc6); + +void intel_rc6_sanitize(struct intel_rc6 *rc6); +void intel_rc6_enable(struct intel_rc6 *rc6); +void intel_rc6_disable(struct intel_rc6 *rc6); + +u64 intel_rc6_residency_ns(struct intel_rc6 *rc6, i915_reg_t reg); +u64 intel_rc6_residency_us(struct intel_rc6 *rc6, i915_reg_t reg); + +void intel_rc6_ctx_wa_check(struct intel_rc6 *rc6); +void intel_rc6_ctx_wa_resume(struct intel_rc6 *rc6); + +#endif /* INTEL_RC6_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_rc6_types.h b/drivers/gpu/drm/i915/gt/intel_rc6_types.h new file mode 100644 index 000000000000..89ad5697a8d4 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rc6_types.h @@ -0,0 +1,29 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_RC6_TYPES_H +#define INTEL_RC6_TYPES_H + +#include <linux/spinlock.h> +#include <linux/types.h> + +#include "intel_engine_types.h" + +struct drm_i915_gem_object; + +struct intel_rc6 { + u64 prev_hw_residency[4]; + u64 cur_residency[4]; + + struct drm_i915_gem_object *pctx; + + bool supported : 1; + bool enabled : 1; + bool wakeref : 1; + bool ctx_corrupted : 1; +}; + +#endif /* INTEL_RC6_TYPES_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_renderstate.c b/drivers/gpu/drm/i915/gt/intel_renderstate.c index 6d05f9c64178..c4edc35e7d89 100644 --- a/drivers/gpu/drm/i915/gt/intel_renderstate.c +++ b/drivers/gpu/drm/i915/gt/intel_renderstate.c @@ -27,6 +27,7 @@ #include "i915_drv.h" #include "intel_renderstate.h" +#include "intel_ring.h" struct intel_renderstate { const struct intel_renderstate_rodata *rodata; diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c index b9d84d52e986..c97423a76642 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.c +++ b/drivers/gpu/drm/i915/gt/intel_reset.c @@ -42,11 +42,10 @@ static void engine_skip_context(struct i915_request *rq) struct intel_engine_cs *engine = rq->engine; struct i915_gem_context *hung_ctx = rq->gem_context; - lockdep_assert_held(&engine->active.lock); - if (!i915_request_is_active(rq)) return; + lockdep_assert_held(&engine->active.lock); list_for_each_entry_continue(rq, &engine->active.requests, sched.link) if (rq->gem_context == hung_ctx) i915_request_skip(rq, -EIO); @@ -123,7 +122,6 @@ void __i915_request_reset(struct i915_request *rq, bool guilty) rq->fence.seqno, yesno(guilty)); - lockdep_assert_held(&rq->engine->active.lock); GEM_BUG_ON(i915_request_completed(rq)); if (guilty) { @@ -284,14 +282,14 @@ static int gen6_reset_engines(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { - struct intel_engine_cs *engine; - const u32 hw_engine_mask[] = { + static const u32 hw_engine_mask[] = { [RCS0] = GEN6_GRDOM_RENDER, [BCS0] = GEN6_GRDOM_BLT, [VCS0] = GEN6_GRDOM_MEDIA, [VCS1] = GEN8_GRDOM_MEDIA2, [VECS0] = GEN6_GRDOM_VECS, }; + struct intel_engine_cs *engine; u32 hw_mask; if (engine_mask == ALL_ENGINES) { @@ -300,7 +298,7 @@ static int gen6_reset_engines(struct intel_gt *gt, intel_engine_mask_t tmp; hw_mask = 0; - for_each_engine_masked(engine, gt->i915, engine_mask, tmp) { + for_each_engine_masked(engine, gt, engine_mask, tmp) { GEM_BUG_ON(engine->id >= ARRAY_SIZE(hw_engine_mask)); hw_mask |= hw_engine_mask[engine->id]; } @@ -309,7 +307,7 @@ static int gen6_reset_engines(struct intel_gt *gt, return gen6_hw_domain_reset(gt, hw_mask); } -static u32 gen11_lock_sfc(struct intel_engine_cs *engine) +static int gen11_lock_sfc(struct intel_engine_cs *engine, u32 *hw_mask) { struct intel_uncore *uncore = engine->uncore; u8 vdbox_sfc_access = RUNTIME_INFO(engine->i915)->vdbox_sfc_access; @@ -318,6 +316,7 @@ static u32 gen11_lock_sfc(struct intel_engine_cs *engine) i915_reg_t sfc_usage; u32 sfc_usage_bit; u32 sfc_reset_bit; + int ret; switch (engine->class) { case VIDEO_DECODE_CLASS: @@ -352,27 +351,33 @@ static u32 gen11_lock_sfc(struct intel_engine_cs *engine) } /* - * Tell the engine that a software reset is going to happen. The engine - * will then try to force lock the SFC (if currently locked, it will - * remain so until we tell the engine it is safe to unlock; if currently - * unlocked, it will ignore this and all new lock requests). If SFC - * ends up being locked to the engine we want to reset, we have to reset - * it as well (we will unlock it once the reset sequence is completed). + * If the engine is using a SFC, tell the engine that a software reset + * is going to happen. The engine will then try to force lock the SFC. + * If SFC ends up being locked to the engine we want to reset, we have + * to reset it as well (we will unlock it once the reset sequence is + * completed). */ + if (!(intel_uncore_read_fw(uncore, sfc_usage) & sfc_usage_bit)) + return 0; + rmw_set_fw(uncore, sfc_forced_lock, sfc_forced_lock_bit); - if (__intel_wait_for_register_fw(uncore, - sfc_forced_lock_ack, - sfc_forced_lock_ack_bit, - sfc_forced_lock_ack_bit, - 1000, 0, NULL)) { - DRM_DEBUG_DRIVER("Wait for SFC forced lock ack failed\n"); + ret = __intel_wait_for_register_fw(uncore, + sfc_forced_lock_ack, + sfc_forced_lock_ack_bit, + sfc_forced_lock_ack_bit, + 1000, 0, NULL); + + /* Was the SFC released while we were trying to lock it? */ + if (!(intel_uncore_read_fw(uncore, sfc_usage) & sfc_usage_bit)) return 0; - } - if (intel_uncore_read_fw(uncore, sfc_usage) & sfc_usage_bit) - return sfc_reset_bit; + if (ret) { + DRM_DEBUG_DRIVER("Wait for SFC forced lock ack failed\n"); + return ret; + } + *hw_mask |= sfc_reset_bit; return 0; } @@ -408,7 +413,7 @@ static int gen11_reset_engines(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { - const u32 hw_engine_mask[] = { + static const u32 hw_engine_mask[] = { [RCS0] = GEN11_GRDOM_RENDER, [BCS0] = GEN11_GRDOM_BLT, [VCS0] = GEN11_GRDOM_MEDIA, @@ -427,17 +432,26 @@ static int gen11_reset_engines(struct intel_gt *gt, hw_mask = GEN11_GRDOM_FULL; } else { hw_mask = 0; - for_each_engine_masked(engine, gt->i915, engine_mask, tmp) { + for_each_engine_masked(engine, gt, engine_mask, tmp) { GEM_BUG_ON(engine->id >= ARRAY_SIZE(hw_engine_mask)); hw_mask |= hw_engine_mask[engine->id]; - hw_mask |= gen11_lock_sfc(engine); + ret = gen11_lock_sfc(engine, &hw_mask); + if (ret) + goto sfc_unlock; } } ret = gen6_hw_domain_reset(gt, hw_mask); +sfc_unlock: + /* + * We unlock the SFC based on the lock status and not the result of + * gen11_lock_sfc to make sure that we clean properly if something + * wrong happened during the lock (e.g. lock acquired after timeout + * expiration). + */ if (engine_mask != ALL_ENGINES) - for_each_engine_masked(engine, gt->i915, engine_mask, tmp) + for_each_engine_masked(engine, gt, engine_mask, tmp) gen11_unlock_sfc(engine); return ret; @@ -496,7 +510,7 @@ static int gen8_reset_engines(struct intel_gt *gt, intel_engine_mask_t tmp; int ret; - for_each_engine_masked(engine, gt->i915, engine_mask, tmp) { + for_each_engine_masked(engine, gt, engine_mask, tmp) { ret = gen8_engine_reset_prepare(engine); if (ret && !reset_non_ready) goto skip_reset; @@ -522,19 +536,30 @@ static int gen8_reset_engines(struct intel_gt *gt, ret = gen6_reset_engines(gt, engine_mask, retry); skip_reset: - for_each_engine_masked(engine, gt->i915, engine_mask, tmp) + for_each_engine_masked(engine, gt, engine_mask, tmp) gen8_engine_reset_cancel(engine); return ret; } +static int mock_reset(struct intel_gt *gt, + intel_engine_mask_t mask, + unsigned int retry) +{ + return 0; +} + typedef int (*reset_func)(struct intel_gt *, intel_engine_mask_t engine_mask, unsigned int retry); -static reset_func intel_get_gpu_reset(struct drm_i915_private *i915) +static reset_func intel_get_gpu_reset(const struct intel_gt *gt) { - if (INTEL_GEN(i915) >= 8) + struct drm_i915_private *i915 = gt->i915; + + if (is_mock_gt(gt)) + return mock_reset; + else if (INTEL_GEN(i915) >= 8) return gen8_reset_engines; else if (INTEL_GEN(i915) >= 6) return gen6_reset_engines; @@ -557,7 +582,7 @@ int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask) int ret = -ETIMEDOUT; int retry; - reset = intel_get_gpu_reset(gt->i915); + reset = intel_get_gpu_reset(gt); if (!reset) return -ENODEV; @@ -577,17 +602,20 @@ int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask) return ret; } -bool intel_has_gpu_reset(struct drm_i915_private *i915) +bool intel_has_gpu_reset(const struct intel_gt *gt) { if (!i915_modparams.reset) return NULL; - return intel_get_gpu_reset(i915); + return intel_get_gpu_reset(gt); } -bool intel_has_reset_engine(struct drm_i915_private *i915) +bool intel_has_reset_engine(const struct intel_gt *gt) { - return INTEL_INFO(i915)->has_reset_engine && i915_modparams.reset >= 2; + if (i915_modparams.reset < 2) + return false; + + return INTEL_INFO(gt->i915)->has_reset_engine; } int intel_reset_guc(struct intel_gt *gt) @@ -654,7 +682,7 @@ static intel_engine_mask_t reset_prepare(struct intel_gt *gt) intel_engine_mask_t awake = 0; enum intel_engine_id id; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { if (intel_engine_pm_get_if_awake(engine)) awake |= engine->mask; reset_prepare_engine(engine); @@ -684,10 +712,10 @@ static int gt_reset(struct intel_gt *gt, intel_engine_mask_t stalled_mask) if (err) return err; - for_each_engine(engine, gt->i915, id) + for_each_engine(engine, gt, id) __intel_engine_reset(engine, stalled_mask & engine->mask); - i915_gem_restore_fences(gt->i915); + i915_gem_restore_fences(gt->ggtt); return err; } @@ -697,7 +725,7 @@ static void reset_finish_engine(struct intel_engine_cs *engine) engine->reset.finish(engine); intel_uncore_forcewake_put(engine->uncore, FORCEWAKE_ALL); - intel_engine_signal_breadcrumbs(engine); + intel_engine_breadcrumbs_irq(engine); } static void reset_finish(struct intel_gt *gt, intel_engine_mask_t awake) @@ -705,7 +733,7 @@ static void reset_finish(struct intel_gt *gt, intel_engine_mask_t awake) struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { reset_finish_engine(engine); if (awake & engine->mask) intel_engine_pm_put(engine); @@ -741,7 +769,7 @@ static void __intel_gt_set_wedged(struct intel_gt *gt) if (GEM_SHOW_DEBUG() && !intel_engines_are_idle(gt)) { struct drm_printer p = drm_debug_printer(__func__); - for_each_engine(engine, gt->i915, id) + for_each_engine(engine, gt, id) intel_engine_dump(engine, &p, "%s\n", engine->name); } @@ -758,7 +786,7 @@ static void __intel_gt_set_wedged(struct intel_gt *gt) if (!INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) __intel_gt_reset(gt, ALL_ENGINES); - for_each_engine(engine, gt->i915, id) + for_each_engine(engine, gt, id) engine->submit_request = nop_submit_request; /* @@ -770,7 +798,7 @@ static void __intel_gt_set_wedged(struct intel_gt *gt) set_bit(I915_WEDGED, >->reset.flags); /* Mark all executing requests as skipped */ - for_each_engine(engine, gt->i915, id) + for_each_engine(engine, gt, id) engine->cancel_requests(engine); reset_finish(gt, awake); @@ -783,7 +811,7 @@ void intel_gt_set_wedged(struct intel_gt *gt) intel_wakeref_t wakeref; mutex_lock(>->reset.mutex); - with_intel_runtime_pm(>->i915->runtime_pm, wakeref) + with_intel_runtime_pm(gt->uncore->rpm, wakeref) __intel_gt_set_wedged(gt); mutex_unlock(>->reset.mutex); } @@ -793,11 +821,13 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) struct intel_gt_timelines *timelines = >->timelines; struct intel_timeline *tl; unsigned long flags; + bool ok; if (!test_bit(I915_WEDGED, >->reset.flags)) return true; - if (!gt->scratch) /* Never full initialised, recovery impossible */ + /* Never fully initialised, recovery impossible */ + if (test_bit(I915_WEDGED_ON_INIT, >->reset.flags)) return false; GEM_TRACE("start\n"); @@ -814,10 +844,10 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) */ spin_lock_irqsave(&timelines->lock, flags); list_for_each_entry(tl, &timelines->active_list, link) { - struct i915_request *rq; + struct dma_fence *fence; - rq = i915_active_request_get_unlocked(&tl->last_request); - if (!rq) + fence = i915_active_fence_get(&tl->last_request); + if (!fence) continue; spin_unlock_irqrestore(&timelines->lock, flags); @@ -829,8 +859,8 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) * (I915_FENCE_TIMEOUT) so this wait should not be unbounded * in the worst case. */ - dma_fence_default_wait(&rq->fence, false, MAX_SCHEDULE_TIMEOUT); - i915_request_put(rq); + dma_fence_default_wait(fence, false, MAX_SCHEDULE_TIMEOUT); + dma_fence_put(fence); /* Restart iteration after droping lock */ spin_lock_irqsave(&timelines->lock, flags); @@ -838,7 +868,18 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) } spin_unlock_irqrestore(&timelines->lock, flags); - intel_gt_sanitize(gt, false); + /* We must reset pending GPU events before restoring our submission */ + ok = !HAS_EXECLISTS(gt->i915); /* XXX better agnosticism desired */ + if (!INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) + ok = __intel_gt_reset(gt, ALL_ENGINES) == 0; + if (!ok) { + /* + * Warn CI about the unrecoverable wedged condition. + * Time for a reboot. + */ + add_taint_for_CI(TAINT_WARN); + return false; + } /* * Undo nop_submit_request. We prevent all new i915 requests from @@ -893,7 +934,7 @@ static int resume(struct intel_gt *gt) enum intel_engine_id id; int ret; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { ret = engine->resume(engine); if (ret) return ret; @@ -943,7 +984,7 @@ void intel_gt_reset(struct intel_gt *gt, awake = reset_prepare(gt); - if (!intel_has_gpu_reset(gt->i915)) { + if (!intel_has_gpu_reset(gt)) { if (i915_modparams.reset) dev_err(gt->i915->drm.dev, "GPU reset not supported\n"); else @@ -972,7 +1013,7 @@ void intel_gt_reset(struct intel_gt *gt, * was running at the time of the reset (i.e. we weren't VT * switched away). */ - ret = i915_gem_init_hw(gt->i915); + ret = intel_gt_init_hw(gt); if (ret) { DRM_ERROR("Failed to initialise HW following reset (%d)\n", ret); @@ -983,8 +1024,6 @@ void intel_gt_reset(struct intel_gt *gt, if (ret) goto taint; - intel_gt_queue_hangcheck(gt); - finish: reset_finish(gt, awake); unlock: @@ -1075,7 +1114,7 @@ int intel_engine_reset(struct intel_engine_cs *engine, const char *msg) out: intel_engine_cancel_stop_cs(engine); reset_finish_engine(engine); - intel_engine_pm_put(engine); + intel_engine_pm_put_async(engine); return ret; } @@ -1151,7 +1190,7 @@ void intel_gt_handle_error(struct intel_gt *gt, * isn't the case at least when we get here by doing a * simulated reset via debugfs, so get an RPM reference. */ - wakeref = intel_runtime_pm_get(>->i915->runtime_pm); + wakeref = intel_runtime_pm_get(gt->uncore->rpm); engine_mask &= INTEL_INFO(gt->i915)->engine_mask; @@ -1164,8 +1203,8 @@ void intel_gt_handle_error(struct intel_gt *gt, * Try engine reset when available. We fall back to full reset if * single reset fails. */ - if (intel_has_reset_engine(gt->i915) && !intel_gt_is_wedged(gt)) { - for_each_engine_masked(engine, gt->i915, engine_mask, tmp) { + if (intel_has_reset_engine(gt) && !intel_gt_is_wedged(gt)) { + for_each_engine_masked(engine, gt, engine_mask, tmp) { BUILD_BUG_ON(I915_RESET_MODESET >= I915_RESET_ENGINE); if (test_and_set_bit(I915_RESET_ENGINE + engine->id, >->reset.flags)) @@ -1193,7 +1232,7 @@ void intel_gt_handle_error(struct intel_gt *gt, synchronize_rcu_expedited(); /* Prevent any other reset-engine attempt. */ - for_each_engine(engine, gt->i915, tmp) { + for_each_engine(engine, gt, tmp) { while (test_and_set_bit(I915_RESET_ENGINE + engine->id, >->reset.flags)) wait_on_bit(>->reset.flags, @@ -1203,7 +1242,7 @@ void intel_gt_handle_error(struct intel_gt *gt, intel_gt_reset_global(gt, engine_mask, msg); - for_each_engine(engine, gt->i915, tmp) + for_each_engine(engine, gt, tmp) clear_bit_unlock(I915_RESET_ENGINE + engine->id, >->reset.flags); clear_bit_unlock(I915_RESET_BACKOFF, >->reset.flags); @@ -1211,13 +1250,11 @@ void intel_gt_handle_error(struct intel_gt *gt, wake_up_all(>->reset.queue); out: - intel_runtime_pm_put(>->i915->runtime_pm, wakeref); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); } -int intel_gt_reset_trylock(struct intel_gt *gt) +int intel_gt_reset_trylock(struct intel_gt *gt, int *srcu) { - int srcu; - might_lock(>->reset.backoff_srcu); might_sleep(); @@ -1232,10 +1269,10 @@ int intel_gt_reset_trylock(struct intel_gt *gt) rcu_read_lock(); } - srcu = srcu_read_lock(>->reset.backoff_srcu); + *srcu = srcu_read_lock(>->reset.backoff_srcu); rcu_read_unlock(); - return srcu; + return 0; } void intel_gt_reset_unlock(struct intel_gt *gt, int tag) @@ -1255,10 +1292,6 @@ int intel_gt_terminally_wedged(struct intel_gt *gt) if (!test_bit(I915_RESET_BACKOFF, >->reset.flags)) return -EIO; - /* XXX intel_reset_finish() still takes struct_mutex!!! */ - if (mutex_is_locked(>->i915->drm.struct_mutex)) - return -EAGAIN; - if (wait_event_interruptible(gt->reset.queue, !test_bit(I915_RESET_BACKOFF, >->reset.flags))) @@ -1267,6 +1300,14 @@ int intel_gt_terminally_wedged(struct intel_gt *gt) return intel_gt_is_wedged(gt) ? -EIO : 0; } +void intel_gt_set_wedged_on_init(struct intel_gt *gt) +{ + BUILD_BUG_ON(I915_RESET_ENGINE + I915_NUM_ENGINES > + I915_WEDGED_ON_INIT); + intel_gt_set_wedged(gt); + set_bit(I915_WEDGED_ON_INIT, >->reset.flags); +} + void intel_gt_init_reset(struct intel_gt *gt) { init_waitqueue_head(>->reset.queue); @@ -1310,4 +1351,5 @@ void __intel_fini_wedge(struct intel_wedge_me *w) #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) #include "selftest_reset.c" +#include "selftest_hangcheck.c" #endif diff --git a/drivers/gpu/drm/i915/gt/intel_reset.h b/drivers/gpu/drm/i915/gt/intel_reset.h index 37a987b17108..8e8d5f761166 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.h +++ b/drivers/gpu/drm/i915/gt/intel_reset.h @@ -14,7 +14,6 @@ #include "intel_engine_types.h" #include "intel_reset_types.h" -struct drm_i915_private; struct i915_request; struct intel_engine_cs; struct intel_gt; @@ -38,13 +37,19 @@ int intel_engine_reset(struct intel_engine_cs *engine, void __i915_request_reset(struct i915_request *rq, bool guilty); -int __must_check intel_gt_reset_trylock(struct intel_gt *gt); +int __must_check intel_gt_reset_trylock(struct intel_gt *gt, int *srcu); void intel_gt_reset_unlock(struct intel_gt *gt, int tag); void intel_gt_set_wedged(struct intel_gt *gt); bool intel_gt_unset_wedged(struct intel_gt *gt); int intel_gt_terminally_wedged(struct intel_gt *gt); +/* + * There's no unset_wedged_on_init paired with this one. + * Once we're wedged on init, there's no going back. + */ +void intel_gt_set_wedged_on_init(struct intel_gt *gt); + int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask); int intel_reset_guc(struct intel_gt *gt); @@ -68,10 +73,13 @@ void __intel_fini_wedge(struct intel_wedge_me *w); static inline bool __intel_reset_failed(const struct intel_reset *reset) { + GEM_BUG_ON(test_bit(I915_WEDGED_ON_INIT, &reset->flags) ? + !test_bit(I915_WEDGED, &reset->flags) : false); + return unlikely(test_bit(I915_WEDGED, &reset->flags)); } -bool intel_has_gpu_reset(struct drm_i915_private *i915); -bool intel_has_reset_engine(struct drm_i915_private *i915); +bool intel_has_gpu_reset(const struct intel_gt *gt); +bool intel_has_reset_engine(const struct intel_gt *gt); #endif /* I915_RESET_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_reset_types.h b/drivers/gpu/drm/i915/gt/intel_reset_types.h index 31968356e0c0..f43bc3a0fe4f 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset_types.h +++ b/drivers/gpu/drm/i915/gt/intel_reset_types.h @@ -29,11 +29,17 @@ struct intel_reset { * we set the #I915_WEDGED bit. Prior to command submission, e.g. * i915_request_alloc(), this bit is checked and the sequence * aborted (with -EIO reported to userspace) if set. + * + * #I915_WEDGED_ON_INIT - If we fail to initialize the GPU we can no + * longer use the GPU - similar to #I915_WEDGED bit. The difference in + * in the way we're handling "forced" unwedged (e.g. through debugfs), + * which is not allowed in case we failed to initialize. */ unsigned long flags; #define I915_RESET_BACKOFF 0 #define I915_RESET_MODESET 1 #define I915_RESET_ENGINE 2 +#define I915_WEDGED_ON_INIT (BITS_PER_LONG - 2) #define I915_WEDGED (BITS_PER_LONG - 1) struct mutex mutex; /* serialises wedging/unwedging */ diff --git a/drivers/gpu/drm/i915/gt/intel_ring.c b/drivers/gpu/drm/i915/gt/intel_ring.c new file mode 100644 index 000000000000..374b28f13ca0 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_ring.c @@ -0,0 +1,318 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include "gem/i915_gem_object.h" +#include "i915_drv.h" +#include "i915_vma.h" +#include "intel_engine.h" +#include "intel_ring.h" +#include "intel_timeline.h" + +unsigned int intel_ring_update_space(struct intel_ring *ring) +{ + unsigned int space; + + space = __intel_ring_space(ring->head, ring->emit, ring->size); + + ring->space = space; + return space; +} + +int intel_ring_pin(struct intel_ring *ring) +{ + struct i915_vma *vma = ring->vma; + unsigned int flags; + void *addr; + int ret; + + if (atomic_fetch_inc(&ring->pin_count)) + return 0; + + flags = PIN_GLOBAL; + + /* Ring wraparound at offset 0 sometimes hangs. No idea why. */ + flags |= PIN_OFFSET_BIAS | i915_ggtt_pin_bias(vma); + + if (vma->obj->stolen) + flags |= PIN_MAPPABLE; + else + flags |= PIN_HIGH; + + ret = i915_vma_pin(vma, 0, 0, flags); + if (unlikely(ret)) + goto err_unpin; + + if (i915_vma_is_map_and_fenceable(vma)) + addr = (void __force *)i915_vma_pin_iomap(vma); + else + addr = i915_gem_object_pin_map(vma->obj, + i915_coherent_map_type(vma->vm->i915)); + if (IS_ERR(addr)) { + ret = PTR_ERR(addr); + goto err_ring; + } + + i915_vma_make_unshrinkable(vma); + + /* Discard any unused bytes beyond that submitted to hw. */ + intel_ring_reset(ring, ring->emit); + + ring->vaddr = addr; + return 0; + +err_ring: + i915_vma_unpin(vma); +err_unpin: + atomic_dec(&ring->pin_count); + return ret; +} + +void intel_ring_reset(struct intel_ring *ring, u32 tail) +{ + tail = intel_ring_wrap(ring, tail); + ring->tail = tail; + ring->head = tail; + ring->emit = tail; + intel_ring_update_space(ring); +} + +void intel_ring_unpin(struct intel_ring *ring) +{ + struct i915_vma *vma = ring->vma; + + if (!atomic_dec_and_test(&ring->pin_count)) + return; + + i915_vma_unset_ggtt_write(vma); + if (i915_vma_is_map_and_fenceable(vma)) + i915_vma_unpin_iomap(vma); + else + i915_gem_object_unpin_map(vma->obj); + + i915_vma_make_purgeable(vma); + i915_vma_unpin(vma); +} + +static struct i915_vma *create_ring_vma(struct i915_ggtt *ggtt, int size) +{ + struct i915_address_space *vm = &ggtt->vm; + struct drm_i915_private *i915 = vm->i915; + struct drm_i915_gem_object *obj; + struct i915_vma *vma; + + obj = ERR_PTR(-ENODEV); + if (i915_ggtt_has_aperture(ggtt)) + obj = i915_gem_object_create_stolen(i915, size); + if (IS_ERR(obj)) + obj = i915_gem_object_create_internal(i915, size); + if (IS_ERR(obj)) + return ERR_CAST(obj); + + /* + * Mark ring buffers as read-only from GPU side (so no stray overwrites) + * if supported by the platform's GGTT. + */ + if (vm->has_read_only) + i915_gem_object_set_readonly(obj); + + vma = i915_vma_instance(obj, vm, NULL); + if (IS_ERR(vma)) + goto err; + + return vma; + +err: + i915_gem_object_put(obj); + return vma; +} + +struct intel_ring * +intel_engine_create_ring(struct intel_engine_cs *engine, int size) +{ + struct drm_i915_private *i915 = engine->i915; + struct intel_ring *ring; + struct i915_vma *vma; + + GEM_BUG_ON(!is_power_of_2(size)); + GEM_BUG_ON(RING_CTL_SIZE(size) & ~RING_NR_PAGES); + + ring = kzalloc(sizeof(*ring), GFP_KERNEL); + if (!ring) + return ERR_PTR(-ENOMEM); + + kref_init(&ring->ref); + ring->size = size; + + /* + * Workaround an erratum on the i830 which causes a hang if + * the TAIL pointer points to within the last 2 cachelines + * of the buffer. + */ + ring->effective_size = size; + if (IS_I830(i915) || IS_I845G(i915)) + ring->effective_size -= 2 * CACHELINE_BYTES; + + intel_ring_update_space(ring); + + vma = create_ring_vma(engine->gt->ggtt, size); + if (IS_ERR(vma)) { + kfree(ring); + return ERR_CAST(vma); + } + ring->vma = vma; + + return ring; +} + +void intel_ring_free(struct kref *ref) +{ + struct intel_ring *ring = container_of(ref, typeof(*ring), ref); + + i915_vma_put(ring->vma); + kfree(ring); +} + +static noinline int +wait_for_space(struct intel_ring *ring, + struct intel_timeline *tl, + unsigned int bytes) +{ + struct i915_request *target; + long timeout; + + if (intel_ring_update_space(ring) >= bytes) + return 0; + + GEM_BUG_ON(list_empty(&tl->requests)); + list_for_each_entry(target, &tl->requests, link) { + if (target->ring != ring) + continue; + + /* Would completion of this request free enough space? */ + if (bytes <= __intel_ring_space(target->postfix, + ring->emit, ring->size)) + break; + } + + if (GEM_WARN_ON(&target->link == &tl->requests)) + return -ENOSPC; + + timeout = i915_request_wait(target, + I915_WAIT_INTERRUPTIBLE, + MAX_SCHEDULE_TIMEOUT); + if (timeout < 0) + return timeout; + + i915_request_retire_upto(target); + + intel_ring_update_space(ring); + GEM_BUG_ON(ring->space < bytes); + return 0; +} + +u32 *intel_ring_begin(struct i915_request *rq, unsigned int num_dwords) +{ + struct intel_ring *ring = rq->ring; + const unsigned int remain_usable = ring->effective_size - ring->emit; + const unsigned int bytes = num_dwords * sizeof(u32); + unsigned int need_wrap = 0; + unsigned int total_bytes; + u32 *cs; + + /* Packets must be qword aligned. */ + GEM_BUG_ON(num_dwords & 1); + + total_bytes = bytes + rq->reserved_space; + GEM_BUG_ON(total_bytes > ring->effective_size); + + if (unlikely(total_bytes > remain_usable)) { + const int remain_actual = ring->size - ring->emit; + + if (bytes > remain_usable) { + /* + * Not enough space for the basic request. So need to + * flush out the remainder and then wait for + * base + reserved. + */ + total_bytes += remain_actual; + need_wrap = remain_actual | 1; + } else { + /* + * The base request will fit but the reserved space + * falls off the end. So we don't need an immediate + * wrap and only need to effectively wait for the + * reserved size from the start of ringbuffer. + */ + total_bytes = rq->reserved_space + remain_actual; + } + } + + if (unlikely(total_bytes > ring->space)) { + int ret; + + /* + * Space is reserved in the ringbuffer for finalising the + * request, as that cannot be allowed to fail. During request + * finalisation, reserved_space is set to 0 to stop the + * overallocation and the assumption is that then we never need + * to wait (which has the risk of failing with EINTR). + * + * See also i915_request_alloc() and i915_request_add(). + */ + GEM_BUG_ON(!rq->reserved_space); + + ret = wait_for_space(ring, + i915_request_timeline(rq), + total_bytes); + if (unlikely(ret)) + return ERR_PTR(ret); + } + + if (unlikely(need_wrap)) { + need_wrap &= ~1; + GEM_BUG_ON(need_wrap > ring->space); + GEM_BUG_ON(ring->emit + need_wrap > ring->size); + GEM_BUG_ON(!IS_ALIGNED(need_wrap, sizeof(u64))); + + /* Fill the tail with MI_NOOP */ + memset64(ring->vaddr + ring->emit, 0, need_wrap / sizeof(u64)); + ring->space -= need_wrap; + ring->emit = 0; + } + + GEM_BUG_ON(ring->emit > ring->size - bytes); + GEM_BUG_ON(ring->space < bytes); + cs = ring->vaddr + ring->emit; + GEM_DEBUG_EXEC(memset32(cs, POISON_INUSE, bytes / sizeof(*cs))); + ring->emit += bytes; + ring->space -= bytes; + + return cs; +} + +/* Align the ring tail to a cacheline boundary */ +int intel_ring_cacheline_align(struct i915_request *rq) +{ + int num_dwords; + void *cs; + + num_dwords = (rq->ring->emit & (CACHELINE_BYTES - 1)) / sizeof(u32); + if (num_dwords == 0) + return 0; + + num_dwords = CACHELINE_DWORDS - num_dwords; + GEM_BUG_ON(num_dwords & 1); + + cs = intel_ring_begin(rq, num_dwords); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + memset64(cs, (u64)MI_NOOP << 32 | MI_NOOP, num_dwords / 2); + intel_ring_advance(rq, cs + num_dwords); + + GEM_BUG_ON(rq->ring->emit & (CACHELINE_BYTES - 1)); + return 0; +} diff --git a/drivers/gpu/drm/i915/gt/intel_ring.h b/drivers/gpu/drm/i915/gt/intel_ring.h new file mode 100644 index 000000000000..ea2839d9e044 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_ring.h @@ -0,0 +1,131 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_RING_H +#define INTEL_RING_H + +#include "i915_gem.h" /* GEM_BUG_ON */ +#include "i915_request.h" +#include "intel_ring_types.h" + +struct intel_engine_cs; + +struct intel_ring * +intel_engine_create_ring(struct intel_engine_cs *engine, int size); + +u32 *intel_ring_begin(struct i915_request *rq, unsigned int num_dwords); +int intel_ring_cacheline_align(struct i915_request *rq); + +unsigned int intel_ring_update_space(struct intel_ring *ring); + +int intel_ring_pin(struct intel_ring *ring); +void intel_ring_unpin(struct intel_ring *ring); +void intel_ring_reset(struct intel_ring *ring, u32 tail); + +void intel_ring_free(struct kref *ref); + +static inline struct intel_ring *intel_ring_get(struct intel_ring *ring) +{ + kref_get(&ring->ref); + return ring; +} + +static inline void intel_ring_put(struct intel_ring *ring) +{ + kref_put(&ring->ref, intel_ring_free); +} + +static inline void intel_ring_advance(struct i915_request *rq, u32 *cs) +{ + /* Dummy function. + * + * This serves as a placeholder in the code so that the reader + * can compare against the preceding intel_ring_begin() and + * check that the number of dwords emitted matches the space + * reserved for the command packet (i.e. the value passed to + * intel_ring_begin()). + */ + GEM_BUG_ON((rq->ring->vaddr + rq->ring->emit) != cs); +} + +static inline u32 intel_ring_wrap(const struct intel_ring *ring, u32 pos) +{ + return pos & (ring->size - 1); +} + +static inline bool +intel_ring_offset_valid(const struct intel_ring *ring, + unsigned int pos) +{ + if (pos & -ring->size) /* must be strictly within the ring */ + return false; + + if (!IS_ALIGNED(pos, 8)) /* must be qword aligned */ + return false; + + return true; +} + +static inline u32 intel_ring_offset(const struct i915_request *rq, void *addr) +{ + /* Don't write ring->size (equivalent to 0) as that hangs some GPUs. */ + u32 offset = addr - rq->ring->vaddr; + GEM_BUG_ON(offset > rq->ring->size); + return intel_ring_wrap(rq->ring, offset); +} + +static inline void +assert_ring_tail_valid(const struct intel_ring *ring, unsigned int tail) +{ + GEM_BUG_ON(!intel_ring_offset_valid(ring, tail)); + + /* + * "Ring Buffer Use" + * Gen2 BSpec "1. Programming Environment" / 1.4.4.6 + * Gen3 BSpec "1c Memory Interface Functions" / 2.3.4.5 + * Gen4+ BSpec "1c Memory Interface and Command Stream" / 5.3.4.5 + * "If the Ring Buffer Head Pointer and the Tail Pointer are on the + * same cacheline, the Head Pointer must not be greater than the Tail + * Pointer." + * + * We use ring->head as the last known location of the actual RING_HEAD, + * it may have advanced but in the worst case it is equally the same + * as ring->head and so we should never program RING_TAIL to advance + * into the same cacheline as ring->head. + */ +#define cacheline(a) round_down(a, CACHELINE_BYTES) + GEM_BUG_ON(cacheline(tail) == cacheline(ring->head) && + tail < ring->head); +#undef cacheline +} + +static inline unsigned int +intel_ring_set_tail(struct intel_ring *ring, unsigned int tail) +{ + /* Whilst writes to the tail are strictly order, there is no + * serialisation between readers and the writers. The tail may be + * read by i915_request_retire() just as it is being updated + * by execlists, as although the breadcrumb is complete, the context + * switch hasn't been seen. + */ + assert_ring_tail_valid(ring, tail); + ring->tail = tail; + return tail; +} + +static inline unsigned int +__intel_ring_space(unsigned int head, unsigned int tail, unsigned int size) +{ + /* + * "If the Ring Buffer Head Pointer and the Tail Pointer are on the + * same cacheline, the Head Pointer must not be greater than the Tail + * Pointer." + */ + GEM_BUG_ON(!is_power_of_2(size)); + return (head - tail - CACHELINE_BYTES) & (size - 1); +} + +#endif /* INTEL_RING_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ring_submission.c index 601c16239fdf..a47d5a7c32c9 100644 --- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/gt/intel_ring_submission.c @@ -40,6 +40,7 @@ #include "intel_gt_irq.h" #include "intel_gt_pm_irq.h" #include "intel_reset.h" +#include "intel_ring.h" #include "intel_workarounds.h" /* Rough estimate of the typical request size, performing a flush, @@ -47,16 +48,6 @@ */ #define LEGACY_REQUEST_SIZE 200 -unsigned int intel_ring_update_space(struct intel_ring *ring) -{ - unsigned int space; - - space = __intel_ring_space(ring->head, ring->emit, ring->size); - - ring->space = space; - return space; -} - static int gen2_render_ring_flush(struct i915_request *rq, u32 mode) { @@ -322,7 +313,8 @@ static u32 *gen6_rcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) PIPE_CONTROL_DC_FLUSH_ENABLE | PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_CS_STALL); - *cs++ = rq->timeline->hwsp_offset | PIPE_CONTROL_GLOBAL_GTT; + *cs++ = i915_request_active_timeline(rq)->hwsp_offset | + PIPE_CONTROL_GLOBAL_GTT; *cs++ = rq->fence.seqno; *cs++ = MI_USER_INTERRUPT; @@ -425,7 +417,7 @@ static u32 *gen7_rcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_GLOBAL_GTT_IVB | PIPE_CONTROL_CS_STALL); - *cs++ = rq->timeline->hwsp_offset; + *cs++ = i915_request_active_timeline(rq)->hwsp_offset; *cs++ = rq->fence.seqno; *cs++ = MI_USER_INTERRUPT; @@ -439,8 +431,8 @@ static u32 *gen7_rcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) static u32 *gen6_xcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) { - GEM_BUG_ON(rq->timeline->hwsp_ggtt != rq->engine->status_page.vma); - GEM_BUG_ON(offset_in_page(rq->timeline->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); + GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma); + GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); *cs++ = MI_FLUSH_DW | MI_FLUSH_DW_OP_STOREDW | MI_FLUSH_DW_STORE_INDEX; *cs++ = I915_GEM_HWS_SEQNO_ADDR | MI_FLUSH_DW_USE_GTT; @@ -459,8 +451,8 @@ static u32 *gen7_xcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) { int i; - GEM_BUG_ON(rq->timeline->hwsp_ggtt != rq->engine->status_page.vma); - GEM_BUG_ON(offset_in_page(rq->timeline->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); + GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma); + GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); *cs++ = MI_FLUSH_DW | MI_FLUSH_DW_OP_STOREDW | MI_FLUSH_DW_STORE_INDEX; *cs++ = I915_GEM_HWS_SEQNO_ADDR | MI_FLUSH_DW_USE_GTT; @@ -930,6 +922,7 @@ static void cancel_requests(struct intel_engine_cs *engine) static void i9xx_submit_request(struct i915_request *request) { i915_request_submit(request); + wmb(); /* paranoid flush writes out of the WCB before mmio */ ENGINE_WRITE(request->engine, RING_TAIL, intel_ring_set_tail(request->ring, request->tail)); @@ -937,8 +930,8 @@ static void i9xx_submit_request(struct i915_request *request) static u32 *i9xx_emit_breadcrumb(struct i915_request *rq, u32 *cs) { - GEM_BUG_ON(rq->timeline->hwsp_ggtt != rq->engine->status_page.vma); - GEM_BUG_ON(offset_in_page(rq->timeline->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); + GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma); + GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); *cs++ = MI_FLUSH; @@ -960,8 +953,8 @@ static u32 *gen5_emit_breadcrumb(struct i915_request *rq, u32 *cs) { int i; - GEM_BUG_ON(rq->timeline->hwsp_ggtt != rq->engine->status_page.vma); - GEM_BUG_ON(offset_in_page(rq->timeline->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); + GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma); + GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); *cs++ = MI_FLUSH; @@ -1184,167 +1177,9 @@ i915_emit_bb_start(struct i915_request *rq, return 0; } -int intel_ring_pin(struct intel_ring *ring) -{ - struct i915_vma *vma = ring->vma; - unsigned int flags; - void *addr; - int ret; - - if (atomic_fetch_inc(&ring->pin_count)) - return 0; - - flags = PIN_GLOBAL; - - /* Ring wraparound at offset 0 sometimes hangs. No idea why. */ - flags |= PIN_OFFSET_BIAS | i915_ggtt_pin_bias(vma); - - if (vma->obj->stolen) - flags |= PIN_MAPPABLE; - else - flags |= PIN_HIGH; - - ret = i915_vma_pin(vma, 0, 0, flags); - if (unlikely(ret)) - goto err_unpin; - - if (i915_vma_is_map_and_fenceable(vma)) - addr = (void __force *)i915_vma_pin_iomap(vma); - else - addr = i915_gem_object_pin_map(vma->obj, - i915_coherent_map_type(vma->vm->i915)); - if (IS_ERR(addr)) { - ret = PTR_ERR(addr); - goto err_ring; - } - - i915_vma_make_unshrinkable(vma); - - GEM_BUG_ON(ring->vaddr); - ring->vaddr = addr; - - return 0; - -err_ring: - i915_vma_unpin(vma); -err_unpin: - atomic_dec(&ring->pin_count); - return ret; -} - -void intel_ring_reset(struct intel_ring *ring, u32 tail) -{ - tail = intel_ring_wrap(ring, tail); - ring->tail = tail; - ring->head = tail; - ring->emit = tail; - intel_ring_update_space(ring); -} - -void intel_ring_unpin(struct intel_ring *ring) -{ - struct i915_vma *vma = ring->vma; - - if (!atomic_dec_and_test(&ring->pin_count)) - return; - - /* Discard any unused bytes beyond that submitted to hw. */ - intel_ring_reset(ring, ring->emit); - - i915_vma_unset_ggtt_write(vma); - if (i915_vma_is_map_and_fenceable(vma)) - i915_vma_unpin_iomap(vma); - else - i915_gem_object_unpin_map(vma->obj); - - GEM_BUG_ON(!ring->vaddr); - ring->vaddr = NULL; - - i915_vma_unpin(vma); - i915_vma_make_purgeable(vma); -} - -static struct i915_vma *create_ring_vma(struct i915_ggtt *ggtt, int size) -{ - struct i915_address_space *vm = &ggtt->vm; - struct drm_i915_private *i915 = vm->i915; - struct drm_i915_gem_object *obj; - struct i915_vma *vma; - - obj = i915_gem_object_create_stolen(i915, size); - if (!obj) - obj = i915_gem_object_create_internal(i915, size); - if (IS_ERR(obj)) - return ERR_CAST(obj); - - /* - * Mark ring buffers as read-only from GPU side (so no stray overwrites) - * if supported by the platform's GGTT. - */ - if (vm->has_read_only) - i915_gem_object_set_readonly(obj); - - vma = i915_vma_instance(obj, vm, NULL); - if (IS_ERR(vma)) - goto err; - - return vma; - -err: - i915_gem_object_put(obj); - return vma; -} - -struct intel_ring * -intel_engine_create_ring(struct intel_engine_cs *engine, int size) -{ - struct drm_i915_private *i915 = engine->i915; - struct intel_ring *ring; - struct i915_vma *vma; - - GEM_BUG_ON(!is_power_of_2(size)); - GEM_BUG_ON(RING_CTL_SIZE(size) & ~RING_NR_PAGES); - - ring = kzalloc(sizeof(*ring), GFP_KERNEL); - if (!ring) - return ERR_PTR(-ENOMEM); - - kref_init(&ring->ref); - - ring->size = size; - /* Workaround an erratum on the i830 which causes a hang if - * the TAIL pointer points to within the last 2 cachelines - * of the buffer. - */ - ring->effective_size = size; - if (IS_I830(i915) || IS_I845G(i915)) - ring->effective_size -= 2 * CACHELINE_BYTES; - - intel_ring_update_space(ring); - - vma = create_ring_vma(engine->gt->ggtt, size); - if (IS_ERR(vma)) { - kfree(ring); - return ERR_CAST(vma); - } - ring->vma = vma; - - return ring; -} - -void intel_ring_free(struct kref *ref) -{ - struct intel_ring *ring = container_of(ref, typeof(*ring), ref); - - i915_vma_close(ring->vma); - i915_vma_put(ring->vma); - - kfree(ring); -} - static void __ring_context_fini(struct intel_context *ce) { - i915_gem_object_put(ce->state->obj); + i915_vma_put(ce->state); } static void ring_context_destroy(struct kref *ref) @@ -1573,7 +1408,7 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags) struct intel_engine_cs *engine = rq->engine; enum intel_engine_id id; const int num_engines = - IS_HSW_GT1(i915) ? RUNTIME_INFO(i915)->num_engines - 1 : 0; + IS_HASWELL(i915) ? RUNTIME_INFO(i915)->num_engines - 1 : 0; bool force_restore = false; int len; u32 *cs; @@ -1609,7 +1444,7 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags) struct intel_engine_cs *signaller; *cs++ = MI_LOAD_REGISTER_IMM(num_engines); - for_each_engine(signaller, i915, id) { + for_each_engine(signaller, engine->gt, id) { if (signaller == engine) continue; @@ -1663,7 +1498,7 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags) i915_reg_t last_reg = {}; /* keep gcc quiet */ *cs++ = MI_LOAD_REGISTER_IMM(num_engines); - for_each_engine(signaller, i915, id) { + for_each_engine(signaller, engine->gt, id) { if (signaller == engine) continue; @@ -1676,7 +1511,7 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags) /* Insert a delay before the next switch! */ *cs++ = MI_STORE_REGISTER_MEM | MI_SRM_LRM_GLOBAL_GTT; *cs++ = i915_mmio_reg_offset(last_reg); - *cs++ = intel_gt_scratch_offset(rq->engine->gt, + *cs++ = intel_gt_scratch_offset(engine->gt, INTEL_GT_SCRATCH_FIELD_DEFAULT); *cs++ = MI_NOOP; } @@ -1741,46 +1576,22 @@ static int remap_l3(struct i915_request *rq) static int switch_context(struct i915_request *rq) { - struct intel_engine_cs *engine = rq->engine; - struct i915_address_space *vm = vm_alias(rq->hw_context); - unsigned int unwind_mm = 0; - u32 hw_flags = 0; + struct intel_context *ce = rq->hw_context; + struct i915_address_space *vm = vm_alias(ce); int ret; GEM_BUG_ON(HAS_EXECLISTS(rq->i915)); if (vm) { - struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); - int loops; - - /* - * Baytail takes a little more convincing that it really needs - * to reload the PD between contexts. It is not just a little - * longer, as adding more stalls after the load_pd_dir (i.e. - * adding a long loop around flush_pd_dir) is not as effective - * as reloading the PD umpteen times. 32 is derived from - * experimentation (gem_exec_parallel/fds) and has no good - * explanation. - */ - loops = 1; - if (engine->id == BCS0 && IS_VALLEYVIEW(engine->i915)) - loops = 32; - - do { - ret = load_pd_dir(rq, ppgtt); - if (ret) - goto err; - } while (--loops); - - if (ppgtt->pd_dirty_engines & engine->mask) { - unwind_mm = engine->mask; - ppgtt->pd_dirty_engines &= ~unwind_mm; - hw_flags = MI_FORCE_RESTORE; - } + ret = load_pd_dir(rq, i915_vm_to_ppgtt(vm)); + if (ret) + return ret; } - if (rq->hw_context->state) { - GEM_BUG_ON(engine->id != RCS0); + if (ce->state) { + u32 hw_flags; + + GEM_BUG_ON(rq->engine->id != RCS0); /* * The kernel context(s) is treated as pure scratch and is not @@ -1789,22 +1600,25 @@ static int switch_context(struct i915_request *rq) * as nothing actually executes using the kernel context; it * is purely used for flushing user contexts. */ + hw_flags = 0; if (i915_gem_context_is_kernel(rq->gem_context)) hw_flags = MI_RESTORE_INHIBIT; ret = mi_set_context(rq, hw_flags); if (ret) - goto err_mm; + return ret; } if (vm) { + struct intel_engine_cs *engine = rq->engine; + ret = engine->emit_flush(rq, EMIT_INVALIDATE); if (ret) - goto err_mm; + return ret; ret = flush_pd_dir(rq); if (ret) - goto err_mm; + return ret; /* * Not only do we need a full barrier (post-sync write) after @@ -1816,24 +1630,18 @@ static int switch_context(struct i915_request *rq) */ ret = engine->emit_flush(rq, EMIT_INVALIDATE); if (ret) - goto err_mm; + return ret; ret = engine->emit_flush(rq, EMIT_FLUSH); if (ret) - goto err_mm; + return ret; } ret = remap_l3(rq); if (ret) - goto err_mm; + return ret; return 0; - -err_mm: - if (unwind_mm) - i915_vm_to_ppgtt(vm)->pd_dirty_engines |= unwind_mm; -err: - return ret; } static int ring_request_alloc(struct i915_request *request) @@ -1841,7 +1649,7 @@ static int ring_request_alloc(struct i915_request *request) int ret; GEM_BUG_ON(!intel_context_is_pinned(request->hw_context)); - GEM_BUG_ON(request->timeline->has_initial_breadcrumb); + GEM_BUG_ON(i915_request_timeline(request)->has_initial_breadcrumb); /* * Flush enough space to reduce the likelihood of waiting after @@ -1863,146 +1671,6 @@ static int ring_request_alloc(struct i915_request *request) return 0; } -static noinline int -wait_for_space(struct intel_ring *ring, - struct intel_timeline *tl, - unsigned int bytes) -{ - struct i915_request *target; - long timeout; - - if (intel_ring_update_space(ring) >= bytes) - return 0; - - GEM_BUG_ON(list_empty(&tl->requests)); - list_for_each_entry(target, &tl->requests, link) { - if (target->ring != ring) - continue; - - /* Would completion of this request free enough space? */ - if (bytes <= __intel_ring_space(target->postfix, - ring->emit, ring->size)) - break; - } - - if (GEM_WARN_ON(&target->link == &tl->requests)) - return -ENOSPC; - - timeout = i915_request_wait(target, - I915_WAIT_INTERRUPTIBLE, - MAX_SCHEDULE_TIMEOUT); - if (timeout < 0) - return timeout; - - i915_request_retire_upto(target); - - intel_ring_update_space(ring); - GEM_BUG_ON(ring->space < bytes); - return 0; -} - -u32 *intel_ring_begin(struct i915_request *rq, unsigned int num_dwords) -{ - struct intel_ring *ring = rq->ring; - const unsigned int remain_usable = ring->effective_size - ring->emit; - const unsigned int bytes = num_dwords * sizeof(u32); - unsigned int need_wrap = 0; - unsigned int total_bytes; - u32 *cs; - - /* Packets must be qword aligned. */ - GEM_BUG_ON(num_dwords & 1); - - total_bytes = bytes + rq->reserved_space; - GEM_BUG_ON(total_bytes > ring->effective_size); - - if (unlikely(total_bytes > remain_usable)) { - const int remain_actual = ring->size - ring->emit; - - if (bytes > remain_usable) { - /* - * Not enough space for the basic request. So need to - * flush out the remainder and then wait for - * base + reserved. - */ - total_bytes += remain_actual; - need_wrap = remain_actual | 1; - } else { - /* - * The base request will fit but the reserved space - * falls off the end. So we don't need an immediate - * wrap and only need to effectively wait for the - * reserved size from the start of ringbuffer. - */ - total_bytes = rq->reserved_space + remain_actual; - } - } - - if (unlikely(total_bytes > ring->space)) { - int ret; - - /* - * Space is reserved in the ringbuffer for finalising the - * request, as that cannot be allowed to fail. During request - * finalisation, reserved_space is set to 0 to stop the - * overallocation and the assumption is that then we never need - * to wait (which has the risk of failing with EINTR). - * - * See also i915_request_alloc() and i915_request_add(). - */ - GEM_BUG_ON(!rq->reserved_space); - - ret = wait_for_space(ring, rq->timeline, total_bytes); - if (unlikely(ret)) - return ERR_PTR(ret); - } - - if (unlikely(need_wrap)) { - need_wrap &= ~1; - GEM_BUG_ON(need_wrap > ring->space); - GEM_BUG_ON(ring->emit + need_wrap > ring->size); - GEM_BUG_ON(!IS_ALIGNED(need_wrap, sizeof(u64))); - - /* Fill the tail with MI_NOOP */ - memset64(ring->vaddr + ring->emit, 0, need_wrap / sizeof(u64)); - ring->space -= need_wrap; - ring->emit = 0; - } - - GEM_BUG_ON(ring->emit > ring->size - bytes); - GEM_BUG_ON(ring->space < bytes); - cs = ring->vaddr + ring->emit; - GEM_DEBUG_EXEC(memset32(cs, POISON_INUSE, bytes / sizeof(*cs))); - ring->emit += bytes; - ring->space -= bytes; - - return cs; -} - -/* Align the ring tail to a cacheline boundary */ -int intel_ring_cacheline_align(struct i915_request *rq) -{ - int num_dwords; - void *cs; - - num_dwords = (rq->ring->emit & (CACHELINE_BYTES - 1)) / sizeof(u32); - if (num_dwords == 0) - return 0; - - num_dwords = CACHELINE_DWORDS - num_dwords; - GEM_BUG_ON(num_dwords & 1); - - cs = intel_ring_begin(rq, num_dwords); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - memset64(cs, (u64)MI_NOOP << 32 | MI_NOOP, num_dwords / 2); - intel_ring_advance(rq, cs); - - GEM_BUG_ON(rq->ring->emit & (CACHELINE_BYTES - 1)); - return 0; -} - static void gen6_bsd_submit_request(struct i915_request *request) { struct intel_uncore *uncore = request->engine->uncore; diff --git a/drivers/gpu/drm/i915/gt/intel_ring_types.h b/drivers/gpu/drm/i915/gt/intel_ring_types.h new file mode 100644 index 000000000000..d9f17f38e0cc --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_ring_types.h @@ -0,0 +1,51 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_RING_TYPES_H +#define INTEL_RING_TYPES_H + +#include <linux/atomic.h> +#include <linux/kref.h> +#include <linux/types.h> + +/* + * Early gen2 devices have a cacheline of just 32 bytes, using 64 is overkill, + * but keeps the logic simple. Indeed, the whole purpose of this macro is just + * to give some inclination as to some of the magic values used in the various + * workarounds! + */ +#define CACHELINE_BYTES 64 +#define CACHELINE_DWORDS (CACHELINE_BYTES / sizeof(u32)) + +struct i915_vma; + +struct intel_ring { + struct kref ref; + struct i915_vma *vma; + void *vaddr; + + /* + * As we have two types of rings, one global to the engine used + * by ringbuffer submission and those that are exclusive to a + * context used by execlists, we have to play safe and allow + * atomic updates to the pin_count. However, the actual pinning + * of the context is either done during initialisation for + * ringbuffer submission or serialised as part of the context + * pinning for execlists, and so we do not need a mutex ourselves + * to serialise intel_ring_pin/intel_ring_unpin. + */ + atomic_t pin_count; + + u32 head; + u32 tail; + u32 emit; + + u32 space; + u32 size; + u32 effective_size; +}; + +#endif /* INTEL_RING_TYPES_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_rps.c b/drivers/gpu/drm/i915/gt/intel_rps.c new file mode 100644 index 000000000000..20d6ee148afc --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rps.c @@ -0,0 +1,1872 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include "i915_drv.h" +#include "intel_gt.h" +#include "intel_gt_irq.h" +#include "intel_gt_pm_irq.h" +#include "intel_rps.h" +#include "intel_sideband.h" +#include "../../../platform/x86/intel_ips.h" + +/* + * Lock protecting IPS related data structures + */ +static DEFINE_SPINLOCK(mchdev_lock); + +static struct intel_gt *rps_to_gt(struct intel_rps *rps) +{ + return container_of(rps, struct intel_gt, rps); +} + +static struct drm_i915_private *rps_to_i915(struct intel_rps *rps) +{ + return rps_to_gt(rps)->i915; +} + +static struct intel_uncore *rps_to_uncore(struct intel_rps *rps) +{ + return rps_to_gt(rps)->uncore; +} + +static u32 rps_pm_sanitize_mask(struct intel_rps *rps, u32 mask) +{ + return mask & ~rps->pm_intrmsk_mbz; +} + +static u32 rps_pm_mask(struct intel_rps *rps, u8 val) +{ + u32 mask = 0; + + /* We use UP_EI_EXPIRED interrupts for both up/down in manual mode */ + if (val > rps->min_freq_softlimit) + mask |= (GEN6_PM_RP_UP_EI_EXPIRED | + GEN6_PM_RP_DOWN_THRESHOLD | + GEN6_PM_RP_DOWN_TIMEOUT); + + if (val < rps->max_freq_softlimit) + mask |= GEN6_PM_RP_UP_EI_EXPIRED | GEN6_PM_RP_UP_THRESHOLD; + + mask &= rps->pm_events; + + return rps_pm_sanitize_mask(rps, ~mask); +} + +static void rps_reset_ei(struct intel_rps *rps) +{ + memset(&rps->ei, 0, sizeof(rps->ei)); +} + +static void rps_enable_interrupts(struct intel_rps *rps) +{ + struct intel_gt *gt = rps_to_gt(rps); + + rps_reset_ei(rps); + + if (IS_VALLEYVIEW(gt->i915)) + /* WaGsvRC0ResidencyMethod:vlv */ + rps->pm_events = GEN6_PM_RP_UP_EI_EXPIRED; + else + rps->pm_events = (GEN6_PM_RP_UP_THRESHOLD | + GEN6_PM_RP_DOWN_THRESHOLD | + GEN6_PM_RP_DOWN_TIMEOUT); + + spin_lock_irq(>->irq_lock); + gen6_gt_pm_enable_irq(gt, rps->pm_events); + spin_unlock_irq(>->irq_lock); + + intel_uncore_write(gt->uncore, GEN6_PMINTRMSK, + rps_pm_mask(rps, rps->cur_freq)); +} + +static void gen6_rps_reset_interrupts(struct intel_rps *rps) +{ + gen6_gt_pm_reset_iir(rps_to_gt(rps), GEN6_PM_RPS_EVENTS); +} + +static void gen11_rps_reset_interrupts(struct intel_rps *rps) +{ + while (gen11_gt_reset_one_iir(rps_to_gt(rps), 0, GEN11_GTPM)) + ; +} + +static void rps_reset_interrupts(struct intel_rps *rps) +{ + struct intel_gt *gt = rps_to_gt(rps); + + spin_lock_irq(>->irq_lock); + if (INTEL_GEN(gt->i915) >= 11) + gen11_rps_reset_interrupts(rps); + else + gen6_rps_reset_interrupts(rps); + + rps->pm_iir = 0; + spin_unlock_irq(>->irq_lock); +} + +static void rps_disable_interrupts(struct intel_rps *rps) +{ + struct intel_gt *gt = rps_to_gt(rps); + + rps->pm_events = 0; + + intel_uncore_write(gt->uncore, GEN6_PMINTRMSK, + rps_pm_sanitize_mask(rps, ~0u)); + + spin_lock_irq(>->irq_lock); + gen6_gt_pm_disable_irq(gt, GEN6_PM_RPS_EVENTS); + spin_unlock_irq(>->irq_lock); + + intel_synchronize_irq(gt->i915); + + /* + * Now that we will not be generating any more work, flush any + * outstanding tasks. As we are called on the RPS idle path, + * we will reset the GPU to minimum frequencies, so the current + * state of the worker can be discarded. + */ + cancel_work_sync(&rps->work); + + rps_reset_interrupts(rps); +} + +static const struct cparams { + u16 i; + u16 t; + u16 m; + u16 c; +} cparams[] = { + { 1, 1333, 301, 28664 }, + { 1, 1066, 294, 24460 }, + { 1, 800, 294, 25192 }, + { 0, 1333, 276, 27605 }, + { 0, 1066, 276, 27605 }, + { 0, 800, 231, 23784 }, +}; + +static void gen5_rps_init(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + struct intel_uncore *uncore = rps_to_uncore(rps); + u8 fmax, fmin, fstart; + u32 rgvmodectl; + int c_m, i; + + if (i915->fsb_freq <= 3200) + c_m = 0; + else if (i915->fsb_freq <= 4800) + c_m = 1; + else + c_m = 2; + + for (i = 0; i < ARRAY_SIZE(cparams); i++) { + if (cparams[i].i == c_m && cparams[i].t == i915->mem_freq) { + rps->ips.m = cparams[i].m; + rps->ips.c = cparams[i].c; + break; + } + } + + rgvmodectl = intel_uncore_read(uncore, MEMMODECTL); + + /* Set up min, max, and cur for interrupt handling */ + fmax = (rgvmodectl & MEMMODE_FMAX_MASK) >> MEMMODE_FMAX_SHIFT; + fmin = (rgvmodectl & MEMMODE_FMIN_MASK); + fstart = (rgvmodectl & MEMMODE_FSTART_MASK) >> + MEMMODE_FSTART_SHIFT; + DRM_DEBUG_DRIVER("fmax: %d, fmin: %d, fstart: %d\n", + fmax, fmin, fstart); + + rps->min_freq = fmax; + rps->max_freq = fmin; + + rps->idle_freq = rps->min_freq; + rps->cur_freq = rps->idle_freq; +} + +static unsigned long +__ips_chipset_val(struct intel_ips *ips) +{ + struct intel_uncore *uncore = + rps_to_uncore(container_of(ips, struct intel_rps, ips)); + unsigned long now = jiffies_to_msecs(jiffies), dt; + unsigned long result; + u64 total, delta; + + lockdep_assert_held(&mchdev_lock); + + /* + * Prevent division-by-zero if we are asking too fast. + * Also, we don't get interesting results if we are polling + * faster than once in 10ms, so just return the saved value + * in such cases. + */ + dt = now - ips->last_time1; + if (dt <= 10) + return ips->chipset_power; + + /* FIXME: handle per-counter overflow */ + total = intel_uncore_read(uncore, DMIEC); + total += intel_uncore_read(uncore, DDREC); + total += intel_uncore_read(uncore, CSIEC); + + delta = total - ips->last_count1; + + result = div_u64(div_u64(ips->m * delta, dt) + ips->c, 10); + + ips->last_count1 = total; + ips->last_time1 = now; + + ips->chipset_power = result; + + return result; +} + +static unsigned long ips_mch_val(struct intel_uncore *uncore) +{ + unsigned int m, x, b; + u32 tsfs; + + tsfs = intel_uncore_read(uncore, TSFS); + x = intel_uncore_read8(uncore, TR1); + + b = tsfs & TSFS_INTR_MASK; + m = (tsfs & TSFS_SLOPE_MASK) >> TSFS_SLOPE_SHIFT; + + return m * x / 127 - b; +} + +static int _pxvid_to_vd(u8 pxvid) +{ + if (pxvid == 0) + return 0; + + if (pxvid >= 8 && pxvid < 31) + pxvid = 31; + + return (pxvid + 2) * 125; +} + +static u32 pvid_to_extvid(struct drm_i915_private *i915, u8 pxvid) +{ + const int vd = _pxvid_to_vd(pxvid); + + if (INTEL_INFO(i915)->is_mobile) + return max(vd - 1125, 0); + + return vd; +} + +static void __gen5_ips_update(struct intel_ips *ips) +{ + struct intel_uncore *uncore = + rps_to_uncore(container_of(ips, struct intel_rps, ips)); + u64 now, delta, dt; + u32 count; + + lockdep_assert_held(&mchdev_lock); + + now = ktime_get_raw_ns(); + dt = now - ips->last_time2; + do_div(dt, NSEC_PER_MSEC); + + /* Don't divide by 0 */ + if (dt <= 10) + return; + + count = intel_uncore_read(uncore, GFXEC); + delta = count - ips->last_count2; + + ips->last_count2 = count; + ips->last_time2 = now; + + /* More magic constants... */ + ips->gfx_power = div_u64(delta * 1181, dt * 10); +} + +static void gen5_rps_update(struct intel_rps *rps) +{ + spin_lock_irq(&mchdev_lock); + __gen5_ips_update(&rps->ips); + spin_unlock_irq(&mchdev_lock); +} + +static bool gen5_rps_set(struct intel_rps *rps, u8 val) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + u16 rgvswctl; + + lockdep_assert_held(&mchdev_lock); + + rgvswctl = intel_uncore_read16(uncore, MEMSWCTL); + if (rgvswctl & MEMCTL_CMD_STS) { + DRM_DEBUG("gpu busy, RCS change rejected\n"); + return false; /* still busy with another command */ + } + + /* Invert the frequency bin into an ips delay */ + val = rps->max_freq - val; + val = rps->min_freq + val; + + rgvswctl = + (MEMCTL_CMD_CHFREQ << MEMCTL_CMD_SHIFT) | + (val << MEMCTL_FREQ_SHIFT) | + MEMCTL_SFCAVM; + intel_uncore_write16(uncore, MEMSWCTL, rgvswctl); + intel_uncore_posting_read16(uncore, MEMSWCTL); + + rgvswctl |= MEMCTL_CMD_STS; + intel_uncore_write16(uncore, MEMSWCTL, rgvswctl); + + return true; +} + +static unsigned long intel_pxfreq(u32 vidfreq) +{ + int div = (vidfreq & 0x3f0000) >> 16; + int post = (vidfreq & 0x3000) >> 12; + int pre = (vidfreq & 0x7); + + if (!pre) + return 0; + + return div * 133333 / (pre << post); +} + +static unsigned int init_emon(struct intel_uncore *uncore) +{ + u8 pxw[16]; + int i; + + /* Disable to program */ + intel_uncore_write(uncore, ECR, 0); + intel_uncore_posting_read(uncore, ECR); + + /* Program energy weights for various events */ + intel_uncore_write(uncore, SDEW, 0x15040d00); + intel_uncore_write(uncore, CSIEW0, 0x007f0000); + intel_uncore_write(uncore, CSIEW1, 0x1e220004); + intel_uncore_write(uncore, CSIEW2, 0x04000004); + + for (i = 0; i < 5; i++) + intel_uncore_write(uncore, PEW(i), 0); + for (i = 0; i < 3; i++) + intel_uncore_write(uncore, DEW(i), 0); + + /* Program P-state weights to account for frequency power adjustment */ + for (i = 0; i < 16; i++) { + u32 pxvidfreq = intel_uncore_read(uncore, PXVFREQ(i)); + unsigned int freq = intel_pxfreq(pxvidfreq); + unsigned int vid = + (pxvidfreq & PXVFREQ_PX_MASK) >> PXVFREQ_PX_SHIFT; + unsigned int val; + + val = vid * vid * freq / 1000 * 255; + val /= 127 * 127 * 900; + + pxw[i] = val; + } + /* Render standby states get 0 weight */ + pxw[14] = 0; + pxw[15] = 0; + + for (i = 0; i < 4; i++) { + intel_uncore_write(uncore, PXW(i), + pxw[i * 4 + 0] << 24 | + pxw[i * 4 + 1] << 16 | + pxw[i * 4 + 2] << 8 | + pxw[i * 4 + 3] << 0); + } + + /* Adjust magic regs to magic values (more experimental results) */ + intel_uncore_write(uncore, OGW0, 0); + intel_uncore_write(uncore, OGW1, 0); + intel_uncore_write(uncore, EG0, 0x00007f00); + intel_uncore_write(uncore, EG1, 0x0000000e); + intel_uncore_write(uncore, EG2, 0x000e0000); + intel_uncore_write(uncore, EG3, 0x68000300); + intel_uncore_write(uncore, EG4, 0x42000000); + intel_uncore_write(uncore, EG5, 0x00140031); + intel_uncore_write(uncore, EG6, 0); + intel_uncore_write(uncore, EG7, 0); + + for (i = 0; i < 8; i++) + intel_uncore_write(uncore, PXWL(i), 0); + + /* Enable PMON + select events */ + intel_uncore_write(uncore, ECR, 0x80000019); + + return intel_uncore_read(uncore, LCFUSE02) & LCFUSE_HIV_MASK; +} + +static bool gen5_rps_enable(struct intel_rps *rps) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + u8 fstart, vstart; + u32 rgvmodectl; + + spin_lock_irq(&mchdev_lock); + + rgvmodectl = intel_uncore_read(uncore, MEMMODECTL); + + /* Enable temp reporting */ + intel_uncore_write16(uncore, PMMISC, + intel_uncore_read16(uncore, PMMISC) | MCPPCE_EN); + intel_uncore_write16(uncore, TSC1, + intel_uncore_read16(uncore, TSC1) | TSE); + + /* 100ms RC evaluation intervals */ + intel_uncore_write(uncore, RCUPEI, 100000); + intel_uncore_write(uncore, RCDNEI, 100000); + + /* Set max/min thresholds to 90ms and 80ms respectively */ + intel_uncore_write(uncore, RCBMAXAVG, 90000); + intel_uncore_write(uncore, RCBMINAVG, 80000); + + intel_uncore_write(uncore, MEMIHYST, 1); + + /* Set up min, max, and cur for interrupt handling */ + fstart = (rgvmodectl & MEMMODE_FSTART_MASK) >> + MEMMODE_FSTART_SHIFT; + + vstart = (intel_uncore_read(uncore, PXVFREQ(fstart)) & + PXVFREQ_PX_MASK) >> PXVFREQ_PX_SHIFT; + + intel_uncore_write(uncore, + MEMINTREN, + MEMINT_CX_SUPR_EN | MEMINT_EVAL_CHG_EN); + + intel_uncore_write(uncore, VIDSTART, vstart); + intel_uncore_posting_read(uncore, VIDSTART); + + rgvmodectl |= MEMMODE_SWMODE_EN; + intel_uncore_write(uncore, MEMMODECTL, rgvmodectl); + + if (wait_for_atomic((intel_uncore_read(uncore, MEMSWCTL) & + MEMCTL_CMD_STS) == 0, 10)) + DRM_ERROR("stuck trying to change perf mode\n"); + mdelay(1); + + gen5_rps_set(rps, rps->cur_freq); + + rps->ips.last_count1 = intel_uncore_read(uncore, DMIEC); + rps->ips.last_count1 += intel_uncore_read(uncore, DDREC); + rps->ips.last_count1 += intel_uncore_read(uncore, CSIEC); + rps->ips.last_time1 = jiffies_to_msecs(jiffies); + + rps->ips.last_count2 = intel_uncore_read(uncore, GFXEC); + rps->ips.last_time2 = ktime_get_raw_ns(); + + spin_unlock_irq(&mchdev_lock); + + rps->ips.corr = init_emon(uncore); + + return true; +} + +static void gen5_rps_disable(struct intel_rps *rps) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + u16 rgvswctl; + + spin_lock_irq(&mchdev_lock); + + rgvswctl = intel_uncore_read16(uncore, MEMSWCTL); + + /* Ack interrupts, disable EFC interrupt */ + intel_uncore_write(uncore, MEMINTREN, + intel_uncore_read(uncore, MEMINTREN) & + ~MEMINT_EVAL_CHG_EN); + intel_uncore_write(uncore, MEMINTRSTS, MEMINT_EVAL_CHG); + intel_uncore_write(uncore, DEIER, + intel_uncore_read(uncore, DEIER) & ~DE_PCU_EVENT); + intel_uncore_write(uncore, DEIIR, DE_PCU_EVENT); + intel_uncore_write(uncore, DEIMR, + intel_uncore_read(uncore, DEIMR) | DE_PCU_EVENT); + + /* Go back to the starting frequency */ + gen5_rps_set(rps, rps->idle_freq); + mdelay(1); + rgvswctl |= MEMCTL_CMD_STS; + intel_uncore_write(uncore, MEMSWCTL, rgvswctl); + mdelay(1); + + spin_unlock_irq(&mchdev_lock); +} + +static u32 rps_limits(struct intel_rps *rps, u8 val) +{ + u32 limits; + + /* + * Only set the down limit when we've reached the lowest level to avoid + * getting more interrupts, otherwise leave this clear. This prevents a + * race in the hw when coming out of rc6: There's a tiny window where + * the hw runs at the minimal clock before selecting the desired + * frequency, if the down threshold expires in that window we will not + * receive a down interrupt. + */ + if (INTEL_GEN(rps_to_i915(rps)) >= 9) { + limits = rps->max_freq_softlimit << 23; + if (val <= rps->min_freq_softlimit) + limits |= rps->min_freq_softlimit << 14; + } else { + limits = rps->max_freq_softlimit << 24; + if (val <= rps->min_freq_softlimit) + limits |= rps->min_freq_softlimit << 16; + } + + return limits; +} + +static void rps_set_power(struct intel_rps *rps, int new_power) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 threshold_up = 0, threshold_down = 0; /* in % */ + u32 ei_up = 0, ei_down = 0; + + lockdep_assert_held(&rps->power.mutex); + + if (new_power == rps->power.mode) + return; + + /* Note the units here are not exactly 1us, but 1280ns. */ + switch (new_power) { + case LOW_POWER: + /* Upclock if more than 95% busy over 16ms */ + ei_up = 16000; + threshold_up = 95; + + /* Downclock if less than 85% busy over 32ms */ + ei_down = 32000; + threshold_down = 85; + break; + + case BETWEEN: + /* Upclock if more than 90% busy over 13ms */ + ei_up = 13000; + threshold_up = 90; + + /* Downclock if less than 75% busy over 32ms */ + ei_down = 32000; + threshold_down = 75; + break; + + case HIGH_POWER: + /* Upclock if more than 85% busy over 10ms */ + ei_up = 10000; + threshold_up = 85; + + /* Downclock if less than 60% busy over 32ms */ + ei_down = 32000; + threshold_down = 60; + break; + } + + /* When byt can survive without system hang with dynamic + * sw freq adjustments, this restriction can be lifted. + */ + if (IS_VALLEYVIEW(i915)) + goto skip_hw_write; + + intel_uncore_write(uncore, GEN6_RP_UP_EI, + GT_INTERVAL_FROM_US(i915, ei_up)); + intel_uncore_write(uncore, GEN6_RP_UP_THRESHOLD, + GT_INTERVAL_FROM_US(i915, + ei_up * threshold_up / 100)); + + intel_uncore_write(uncore, GEN6_RP_DOWN_EI, + GT_INTERVAL_FROM_US(i915, ei_down)); + intel_uncore_write(uncore, GEN6_RP_DOWN_THRESHOLD, + GT_INTERVAL_FROM_US(i915, + ei_down * threshold_down / 100)); + + intel_uncore_write(uncore, GEN6_RP_CONTROL, + (INTEL_GEN(i915) > 9 ? 0 : GEN6_RP_MEDIA_TURBO) | + GEN6_RP_MEDIA_HW_NORMAL_MODE | + GEN6_RP_MEDIA_IS_GFX | + GEN6_RP_ENABLE | + GEN6_RP_UP_BUSY_AVG | + GEN6_RP_DOWN_IDLE_AVG); + +skip_hw_write: + rps->power.mode = new_power; + rps->power.up_threshold = threshold_up; + rps->power.down_threshold = threshold_down; +} + +static void gen6_rps_set_thresholds(struct intel_rps *rps, u8 val) +{ + int new_power; + + new_power = rps->power.mode; + switch (rps->power.mode) { + case LOW_POWER: + if (val > rps->efficient_freq + 1 && + val > rps->cur_freq) + new_power = BETWEEN; + break; + + case BETWEEN: + if (val <= rps->efficient_freq && + val < rps->cur_freq) + new_power = LOW_POWER; + else if (val >= rps->rp0_freq && + val > rps->cur_freq) + new_power = HIGH_POWER; + break; + + case HIGH_POWER: + if (val < (rps->rp1_freq + rps->rp0_freq) >> 1 && + val < rps->cur_freq) + new_power = BETWEEN; + break; + } + /* Max/min bins are special */ + if (val <= rps->min_freq_softlimit) + new_power = LOW_POWER; + if (val >= rps->max_freq_softlimit) + new_power = HIGH_POWER; + + mutex_lock(&rps->power.mutex); + if (rps->power.interactive) + new_power = HIGH_POWER; + rps_set_power(rps, new_power); + mutex_unlock(&rps->power.mutex); +} + +void intel_rps_mark_interactive(struct intel_rps *rps, bool interactive) +{ + mutex_lock(&rps->power.mutex); + if (interactive) { + if (!rps->power.interactive++ && rps->active) + rps_set_power(rps, HIGH_POWER); + } else { + GEM_BUG_ON(!rps->power.interactive); + rps->power.interactive--; + } + mutex_unlock(&rps->power.mutex); +} + +static int gen6_rps_set(struct intel_rps *rps, u8 val) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 swreq; + + if (INTEL_GEN(i915) >= 9) + swreq = GEN9_FREQUENCY(val); + else if (IS_HASWELL(i915) || IS_BROADWELL(i915)) + swreq = HSW_FREQUENCY(val); + else + swreq = (GEN6_FREQUENCY(val) | + GEN6_OFFSET(0) | + GEN6_AGGRESSIVE_TURBO); + intel_uncore_write(uncore, GEN6_RPNSWREQ, swreq); + + return 0; +} + +static int vlv_rps_set(struct intel_rps *rps, u8 val) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + int err; + + vlv_punit_get(i915); + err = vlv_punit_write(i915, PUNIT_REG_GPU_FREQ_REQ, val); + vlv_punit_put(i915); + + return err; +} + +static int rps_set(struct intel_rps *rps, u8 val) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + int err; + + if (INTEL_GEN(i915) < 6) + return 0; + + if (val == rps->last_freq) + return 0; + + if (IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915)) + err = vlv_rps_set(rps, val); + else + err = gen6_rps_set(rps, val); + if (err) + return err; + + gen6_rps_set_thresholds(rps, val); + rps->last_freq = val; + + return 0; +} + +void intel_rps_unpark(struct intel_rps *rps) +{ + u8 freq; + + if (!rps->enabled) + return; + + /* + * Use the user's desired frequency as a guide, but for better + * performance, jump directly to RPe as our starting frequency. + */ + mutex_lock(&rps->lock); + rps->active = true; + freq = max(rps->cur_freq, rps->efficient_freq), + freq = clamp(freq, rps->min_freq_softlimit, rps->max_freq_softlimit); + intel_rps_set(rps, freq); + rps->last_adj = 0; + mutex_unlock(&rps->lock); + + if (INTEL_GEN(rps_to_i915(rps)) >= 6) + rps_enable_interrupts(rps); + + if (IS_GEN(rps_to_i915(rps), 5)) + gen5_rps_update(rps); +} + +void intel_rps_park(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + + if (!rps->enabled) + return; + + if (INTEL_GEN(i915) >= 6) + rps_disable_interrupts(rps); + + rps->active = false; + if (rps->last_freq <= rps->idle_freq) + return; + + /* + * The punit delays the write of the frequency and voltage until it + * determines the GPU is awake. During normal usage we don't want to + * waste power changing the frequency if the GPU is sleeping (rc6). + * However, the GPU and driver is now idle and we do not want to delay + * switching to minimum voltage (reducing power whilst idle) as we do + * not expect to be woken in the near future and so must flush the + * change by waking the device. + * + * We choose to take the media powerwell (either would do to trick the + * punit into committing the voltage change) as that takes a lot less + * power than the render powerwell. + */ + intel_uncore_forcewake_get(rps_to_uncore(rps), FORCEWAKE_MEDIA); + rps_set(rps, rps->idle_freq); + intel_uncore_forcewake_put(rps_to_uncore(rps), FORCEWAKE_MEDIA); +} + +void intel_rps_boost(struct i915_request *rq) +{ + struct intel_rps *rps = &rq->engine->gt->rps; + unsigned long flags; + + if (i915_request_signaled(rq) || !rps->active) + return; + + /* Serializes with i915_request_retire() */ + spin_lock_irqsave(&rq->lock, flags); + if (!i915_request_has_waitboost(rq) && + !dma_fence_is_signaled_locked(&rq->fence)) { + rq->flags |= I915_REQUEST_WAITBOOST; + + if (!atomic_fetch_inc(&rps->num_waiters) && + READ_ONCE(rps->cur_freq) < rps->boost_freq) + schedule_work(&rps->work); + + atomic_inc(&rps->boosts); + } + spin_unlock_irqrestore(&rq->lock, flags); +} + +int intel_rps_set(struct intel_rps *rps, u8 val) +{ + int err = 0; + + lockdep_assert_held(&rps->lock); + GEM_BUG_ON(val > rps->max_freq); + GEM_BUG_ON(val < rps->min_freq); + + if (rps->active) { + err = rps_set(rps, val); + + /* + * Make sure we continue to get interrupts + * until we hit the minimum or maximum frequencies. + */ + if (INTEL_GEN(rps_to_i915(rps)) >= 6) { + struct intel_uncore *uncore = rps_to_uncore(rps); + + intel_uncore_write(uncore, GEN6_RP_INTERRUPT_LIMITS, + rps_limits(rps, val)); + + intel_uncore_write(uncore, GEN6_PMINTRMSK, + rps_pm_mask(rps, val)); + } + } + + if (err == 0) + rps->cur_freq = val; + + return err; +} + +static void gen6_rps_init(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + struct intel_uncore *uncore = rps_to_uncore(rps); + + /* All of these values are in units of 50MHz */ + + /* static values from HW: RP0 > RP1 > RPn (min_freq) */ + if (IS_GEN9_LP(i915)) { + u32 rp_state_cap = intel_uncore_read(uncore, BXT_RP_STATE_CAP); + + rps->rp0_freq = (rp_state_cap >> 16) & 0xff; + rps->rp1_freq = (rp_state_cap >> 8) & 0xff; + rps->min_freq = (rp_state_cap >> 0) & 0xff; + } else { + u32 rp_state_cap = intel_uncore_read(uncore, GEN6_RP_STATE_CAP); + + rps->rp0_freq = (rp_state_cap >> 0) & 0xff; + rps->rp1_freq = (rp_state_cap >> 8) & 0xff; + rps->min_freq = (rp_state_cap >> 16) & 0xff; + } + + /* hw_max = RP0 until we check for overclocking */ + rps->max_freq = rps->rp0_freq; + + rps->efficient_freq = rps->rp1_freq; + if (IS_HASWELL(i915) || IS_BROADWELL(i915) || + IS_GEN9_BC(i915) || INTEL_GEN(i915) >= 10) { + u32 ddcc_status = 0; + + if (sandybridge_pcode_read(i915, + HSW_PCODE_DYNAMIC_DUTY_CYCLE_CONTROL, + &ddcc_status, NULL) == 0) + rps->efficient_freq = + clamp_t(u8, + (ddcc_status >> 8) & 0xff, + rps->min_freq, + rps->max_freq); + } + + if (IS_GEN9_BC(i915) || INTEL_GEN(i915) >= 10) { + /* Store the frequency values in 16.66 MHZ units, which is + * the natural hardware unit for SKL + */ + rps->rp0_freq *= GEN9_FREQ_SCALER; + rps->rp1_freq *= GEN9_FREQ_SCALER; + rps->min_freq *= GEN9_FREQ_SCALER; + rps->max_freq *= GEN9_FREQ_SCALER; + rps->efficient_freq *= GEN9_FREQ_SCALER; + } +} + +static bool rps_reset(struct intel_rps *rps) +{ + /* force a reset */ + rps->power.mode = -1; + rps->last_freq = -1; + + if (rps_set(rps, rps->min_freq)) { + DRM_ERROR("Failed to reset RPS to initial values\n"); + return false; + } + + rps->cur_freq = rps->min_freq; + return true; +} + +/* See the Gen9_GT_PM_Programming_Guide doc for the below */ +static bool gen9_rps_enable(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + struct intel_uncore *uncore = rps_to_uncore(rps); + + /* Program defaults and thresholds for RPS */ + if (IS_GEN(i915, 9)) + intel_uncore_write_fw(uncore, GEN6_RC_VIDEO_FREQ, + GEN9_FREQUENCY(rps->rp1_freq)); + + /* 1 second timeout */ + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_TIMEOUT, + GT_INTERVAL_FROM_US(i915, 1000000)); + + intel_uncore_write_fw(uncore, GEN6_RP_IDLE_HYSTERSIS, 0xa); + + return rps_reset(rps); +} + +static bool gen8_rps_enable(struct intel_rps *rps) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + + intel_uncore_write_fw(uncore, GEN6_RC_VIDEO_FREQ, + HSW_FREQUENCY(rps->rp1_freq)); + + /* NB: Docs say 1s, and 1000000 - which aren't equivalent */ + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_TIMEOUT, + 100000000 / 128); /* 1 second timeout */ + + intel_uncore_write_fw(uncore, GEN6_RP_IDLE_HYSTERSIS, 10); + + return rps_reset(rps); +} + +static bool gen6_rps_enable(struct intel_rps *rps) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + + /* Power down if completely idle for over 50ms */ + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_TIMEOUT, 50000); + intel_uncore_write_fw(uncore, GEN6_RP_IDLE_HYSTERSIS, 10); + + return rps_reset(rps); +} + +static int chv_rps_max_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + val = vlv_punit_read(i915, FB_GFX_FMAX_AT_VMAX_FUSE); + + switch (RUNTIME_INFO(i915)->sseu.eu_total) { + case 8: + /* (2 * 4) config */ + val >>= FB_GFX_FMAX_AT_VMAX_2SS4EU_FUSE_SHIFT; + break; + case 12: + /* (2 * 6) config */ + val >>= FB_GFX_FMAX_AT_VMAX_2SS6EU_FUSE_SHIFT; + break; + case 16: + /* (2 * 8) config */ + default: + /* Setting (2 * 8) Min RP0 for any other combination */ + val >>= FB_GFX_FMAX_AT_VMAX_2SS8EU_FUSE_SHIFT; + break; + } + + return val & FB_GFX_FREQ_FUSE_MASK; +} + +static int chv_rps_rpe_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + val = vlv_punit_read(i915, PUNIT_GPU_DUTYCYCLE_REG); + val >>= PUNIT_GPU_DUTYCYCLE_RPE_FREQ_SHIFT; + + return val & PUNIT_GPU_DUTYCYCLE_RPE_FREQ_MASK; +} + +static int chv_rps_guar_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + val = vlv_punit_read(i915, FB_GFX_FMAX_AT_VMAX_FUSE); + + return val & FB_GFX_FREQ_FUSE_MASK; +} + +static u32 chv_rps_min_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + val = vlv_punit_read(i915, FB_GFX_FMIN_AT_VMIN_FUSE); + val >>= FB_GFX_FMIN_AT_VMIN_FUSE_SHIFT; + + return val & FB_GFX_FREQ_FUSE_MASK; +} + +static bool chv_rps_enable(struct intel_rps *rps) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + /* 1: Program defaults and thresholds for RPS*/ + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_TIMEOUT, 1000000); + intel_uncore_write_fw(uncore, GEN6_RP_UP_THRESHOLD, 59400); + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_THRESHOLD, 245000); + intel_uncore_write_fw(uncore, GEN6_RP_UP_EI, 66000); + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_EI, 350000); + + intel_uncore_write_fw(uncore, GEN6_RP_IDLE_HYSTERSIS, 10); + + /* 2: Enable RPS */ + intel_uncore_write_fw(uncore, GEN6_RP_CONTROL, + GEN6_RP_MEDIA_HW_NORMAL_MODE | + GEN6_RP_MEDIA_IS_GFX | + GEN6_RP_ENABLE | + GEN6_RP_UP_BUSY_AVG | + GEN6_RP_DOWN_IDLE_AVG); + + /* Setting Fixed Bias */ + vlv_punit_get(i915); + + val = VLV_OVERRIDE_EN | VLV_SOC_TDP_EN | CHV_BIAS_CPU_50_SOC_50; + vlv_punit_write(i915, VLV_TURBO_SOC_OVERRIDE, val); + + val = vlv_punit_read(i915, PUNIT_REG_GPU_FREQ_STS); + + vlv_punit_put(i915); + + /* RPS code assumes GPLL is used */ + WARN_ONCE((val & GPLLENABLE) == 0, "GPLL not enabled\n"); + + DRM_DEBUG_DRIVER("GPLL enabled? %s\n", yesno(val & GPLLENABLE)); + DRM_DEBUG_DRIVER("GPU status: 0x%08x\n", val); + + return rps_reset(rps); +} + +static int vlv_rps_guar_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val, rp1; + + val = vlv_nc_read(i915, IOSF_NC_FB_GFX_FREQ_FUSE); + + rp1 = val & FB_GFX_FGUARANTEED_FREQ_FUSE_MASK; + rp1 >>= FB_GFX_FGUARANTEED_FREQ_FUSE_SHIFT; + + return rp1; +} + +static int vlv_rps_max_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val, rp0; + + val = vlv_nc_read(i915, IOSF_NC_FB_GFX_FREQ_FUSE); + + rp0 = (val & FB_GFX_MAX_FREQ_FUSE_MASK) >> FB_GFX_MAX_FREQ_FUSE_SHIFT; + /* Clamp to max */ + rp0 = min_t(u32, rp0, 0xea); + + return rp0; +} + +static int vlv_rps_rpe_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val, rpe; + + val = vlv_nc_read(i915, IOSF_NC_FB_GFX_FMAX_FUSE_LO); + rpe = (val & FB_FMAX_VMIN_FREQ_LO_MASK) >> FB_FMAX_VMIN_FREQ_LO_SHIFT; + val = vlv_nc_read(i915, IOSF_NC_FB_GFX_FMAX_FUSE_HI); + rpe |= (val & FB_FMAX_VMIN_FREQ_HI_MASK) << 5; + + return rpe; +} + +static int vlv_rps_min_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + val = vlv_punit_read(i915, PUNIT_REG_GPU_LFM) & 0xff; + /* + * According to the BYT Punit GPU turbo HAS 1.1.6.3 the minimum value + * for the minimum frequency in GPLL mode is 0xc1. Contrary to this on + * a BYT-M B0 the above register contains 0xbf. Moreover when setting + * a frequency Punit will not allow values below 0xc0. Clamp it 0xc0 + * to make sure it matches what Punit accepts. + */ + return max_t(u32, val, 0xc0); +} + +static bool vlv_rps_enable(struct intel_rps *rps) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_TIMEOUT, 1000000); + intel_uncore_write_fw(uncore, GEN6_RP_UP_THRESHOLD, 59400); + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_THRESHOLD, 245000); + intel_uncore_write_fw(uncore, GEN6_RP_UP_EI, 66000); + intel_uncore_write_fw(uncore, GEN6_RP_DOWN_EI, 350000); + + intel_uncore_write_fw(uncore, GEN6_RP_IDLE_HYSTERSIS, 10); + + intel_uncore_write_fw(uncore, GEN6_RP_CONTROL, + GEN6_RP_MEDIA_TURBO | + GEN6_RP_MEDIA_HW_NORMAL_MODE | + GEN6_RP_MEDIA_IS_GFX | + GEN6_RP_ENABLE | + GEN6_RP_UP_BUSY_AVG | + GEN6_RP_DOWN_IDLE_CONT); + + vlv_punit_get(i915); + + /* Setting Fixed Bias */ + val = VLV_OVERRIDE_EN | VLV_SOC_TDP_EN | VLV_BIAS_CPU_125_SOC_875; + vlv_punit_write(i915, VLV_TURBO_SOC_OVERRIDE, val); + + val = vlv_punit_read(i915, PUNIT_REG_GPU_FREQ_STS); + + vlv_punit_put(i915); + + /* RPS code assumes GPLL is used */ + WARN_ONCE((val & GPLLENABLE) == 0, "GPLL not enabled\n"); + + DRM_DEBUG_DRIVER("GPLL enabled? %s\n", yesno(val & GPLLENABLE)); + DRM_DEBUG_DRIVER("GPU status: 0x%08x\n", val); + + return rps_reset(rps); +} + +static unsigned long __ips_gfx_val(struct intel_ips *ips) +{ + struct intel_rps *rps = container_of(ips, typeof(*rps), ips); + struct intel_uncore *uncore = rps_to_uncore(rps); + unsigned long t, corr, state1, corr2, state2; + u32 pxvid, ext_v; + + lockdep_assert_held(&mchdev_lock); + + pxvid = intel_uncore_read(uncore, PXVFREQ(rps->cur_freq)); + pxvid = (pxvid >> 24) & 0x7f; + ext_v = pvid_to_extvid(rps_to_i915(rps), pxvid); + + state1 = ext_v; + + /* Revel in the empirically derived constants */ + + /* Correction factor in 1/100000 units */ + t = ips_mch_val(uncore); + if (t > 80) + corr = t * 2349 + 135940; + else if (t >= 50) + corr = t * 964 + 29317; + else /* < 50 */ + corr = t * 301 + 1004; + + corr = corr * 150142 * state1 / 10000 - 78642; + corr /= 100000; + corr2 = corr * ips->corr; + + state2 = corr2 * state1 / 10000; + state2 /= 100; /* convert to mW */ + + __gen5_ips_update(ips); + + return ips->gfx_power + state2; +} + +void intel_rps_enable(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + struct intel_uncore *uncore = rps_to_uncore(rps); + + intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); + if (IS_CHERRYVIEW(i915)) + rps->enabled = chv_rps_enable(rps); + else if (IS_VALLEYVIEW(i915)) + rps->enabled = vlv_rps_enable(rps); + else if (INTEL_GEN(i915) >= 9) + rps->enabled = gen9_rps_enable(rps); + else if (INTEL_GEN(i915) >= 8) + rps->enabled = gen8_rps_enable(rps); + else if (INTEL_GEN(i915) >= 6) + rps->enabled = gen6_rps_enable(rps); + else if (IS_IRONLAKE_M(i915)) + rps->enabled = gen5_rps_enable(rps); + intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); + if (!rps->enabled) + return; + + WARN_ON(rps->max_freq < rps->min_freq); + WARN_ON(rps->idle_freq > rps->max_freq); + + WARN_ON(rps->efficient_freq < rps->min_freq); + WARN_ON(rps->efficient_freq > rps->max_freq); +} + +static void gen6_rps_disable(struct intel_rps *rps) +{ + intel_uncore_write(rps_to_uncore(rps), GEN6_RP_CONTROL, 0); +} + +void intel_rps_disable(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + + rps->enabled = false; + + if (INTEL_GEN(i915) >= 6) + gen6_rps_disable(rps); + else if (IS_IRONLAKE_M(i915)) + gen5_rps_disable(rps); +} + +static int byt_gpu_freq(struct intel_rps *rps, int val) +{ + /* + * N = val - 0xb7 + * Slow = Fast = GPLL ref * N + */ + return DIV_ROUND_CLOSEST(rps->gpll_ref_freq * (val - 0xb7), 1000); +} + +static int byt_freq_opcode(struct intel_rps *rps, int val) +{ + return DIV_ROUND_CLOSEST(1000 * val, rps->gpll_ref_freq) + 0xb7; +} + +static int chv_gpu_freq(struct intel_rps *rps, int val) +{ + /* + * N = val / 2 + * CU (slow) = CU2x (fast) / 2 = GPLL ref * N / 2 + */ + return DIV_ROUND_CLOSEST(rps->gpll_ref_freq * val, 2 * 2 * 1000); +} + +static int chv_freq_opcode(struct intel_rps *rps, int val) +{ + /* CHV needs even values */ + return DIV_ROUND_CLOSEST(2 * 1000 * val, rps->gpll_ref_freq) * 2; +} + +int intel_gpu_freq(struct intel_rps *rps, int val) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + + if (INTEL_GEN(i915) >= 9) + return DIV_ROUND_CLOSEST(val * GT_FREQUENCY_MULTIPLIER, + GEN9_FREQ_SCALER); + else if (IS_CHERRYVIEW(i915)) + return chv_gpu_freq(rps, val); + else if (IS_VALLEYVIEW(i915)) + return byt_gpu_freq(rps, val); + else + return val * GT_FREQUENCY_MULTIPLIER; +} + +int intel_freq_opcode(struct intel_rps *rps, int val) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + + if (INTEL_GEN(i915) >= 9) + return DIV_ROUND_CLOSEST(val * GEN9_FREQ_SCALER, + GT_FREQUENCY_MULTIPLIER); + else if (IS_CHERRYVIEW(i915)) + return chv_freq_opcode(rps, val); + else if (IS_VALLEYVIEW(i915)) + return byt_freq_opcode(rps, val); + else + return DIV_ROUND_CLOSEST(val, GT_FREQUENCY_MULTIPLIER); +} + +static void vlv_init_gpll_ref_freq(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + + rps->gpll_ref_freq = + vlv_get_cck_clock(i915, "GPLL ref", + CCK_GPLL_CLOCK_CONTROL, + i915->czclk_freq); + + DRM_DEBUG_DRIVER("GPLL reference freq: %d kHz\n", rps->gpll_ref_freq); +} + +static void vlv_rps_init(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + vlv_iosf_sb_get(i915, + BIT(VLV_IOSF_SB_PUNIT) | + BIT(VLV_IOSF_SB_NC) | + BIT(VLV_IOSF_SB_CCK)); + + vlv_init_gpll_ref_freq(rps); + + val = vlv_punit_read(i915, PUNIT_REG_GPU_FREQ_STS); + switch ((val >> 6) & 3) { + case 0: + case 1: + i915->mem_freq = 800; + break; + case 2: + i915->mem_freq = 1066; + break; + case 3: + i915->mem_freq = 1333; + break; + } + DRM_DEBUG_DRIVER("DDR speed: %d MHz\n", i915->mem_freq); + + rps->max_freq = vlv_rps_max_freq(rps); + rps->rp0_freq = rps->max_freq; + DRM_DEBUG_DRIVER("max GPU freq: %d MHz (%u)\n", + intel_gpu_freq(rps, rps->max_freq), + rps->max_freq); + + rps->efficient_freq = vlv_rps_rpe_freq(rps); + DRM_DEBUG_DRIVER("RPe GPU freq: %d MHz (%u)\n", + intel_gpu_freq(rps, rps->efficient_freq), + rps->efficient_freq); + + rps->rp1_freq = vlv_rps_guar_freq(rps); + DRM_DEBUG_DRIVER("RP1(Guar Freq) GPU freq: %d MHz (%u)\n", + intel_gpu_freq(rps, rps->rp1_freq), + rps->rp1_freq); + + rps->min_freq = vlv_rps_min_freq(rps); + DRM_DEBUG_DRIVER("min GPU freq: %d MHz (%u)\n", + intel_gpu_freq(rps, rps->min_freq), + rps->min_freq); + + vlv_iosf_sb_put(i915, + BIT(VLV_IOSF_SB_PUNIT) | + BIT(VLV_IOSF_SB_NC) | + BIT(VLV_IOSF_SB_CCK)); +} + +static void chv_rps_init(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 val; + + vlv_iosf_sb_get(i915, + BIT(VLV_IOSF_SB_PUNIT) | + BIT(VLV_IOSF_SB_NC) | + BIT(VLV_IOSF_SB_CCK)); + + vlv_init_gpll_ref_freq(rps); + + val = vlv_cck_read(i915, CCK_FUSE_REG); + + switch ((val >> 2) & 0x7) { + case 3: + i915->mem_freq = 2000; + break; + default: + i915->mem_freq = 1600; + break; + } + DRM_DEBUG_DRIVER("DDR speed: %d MHz\n", i915->mem_freq); + + rps->max_freq = chv_rps_max_freq(rps); + rps->rp0_freq = rps->max_freq; + DRM_DEBUG_DRIVER("max GPU freq: %d MHz (%u)\n", + intel_gpu_freq(rps, rps->max_freq), + rps->max_freq); + + rps->efficient_freq = chv_rps_rpe_freq(rps); + DRM_DEBUG_DRIVER("RPe GPU freq: %d MHz (%u)\n", + intel_gpu_freq(rps, rps->efficient_freq), + rps->efficient_freq); + + rps->rp1_freq = chv_rps_guar_freq(rps); + DRM_DEBUG_DRIVER("RP1(Guar) GPU freq: %d MHz (%u)\n", + intel_gpu_freq(rps, rps->rp1_freq), + rps->rp1_freq); + + rps->min_freq = chv_rps_min_freq(rps); + DRM_DEBUG_DRIVER("min GPU freq: %d MHz (%u)\n", + intel_gpu_freq(rps, rps->min_freq), + rps->min_freq); + + vlv_iosf_sb_put(i915, + BIT(VLV_IOSF_SB_PUNIT) | + BIT(VLV_IOSF_SB_NC) | + BIT(VLV_IOSF_SB_CCK)); + + WARN_ONCE((rps->max_freq | rps->efficient_freq | rps->rp1_freq | + rps->min_freq) & 1, + "Odd GPU freq values\n"); +} + +static void vlv_c0_read(struct intel_uncore *uncore, struct intel_rps_ei *ei) +{ + ei->ktime = ktime_get_raw(); + ei->render_c0 = intel_uncore_read(uncore, VLV_RENDER_C0_COUNT); + ei->media_c0 = intel_uncore_read(uncore, VLV_MEDIA_C0_COUNT); +} + +static u32 vlv_wa_c0_ei(struct intel_rps *rps, u32 pm_iir) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + const struct intel_rps_ei *prev = &rps->ei; + struct intel_rps_ei now; + u32 events = 0; + + if ((pm_iir & GEN6_PM_RP_UP_EI_EXPIRED) == 0) + return 0; + + vlv_c0_read(uncore, &now); + + if (prev->ktime) { + u64 time, c0; + u32 render, media; + + time = ktime_us_delta(now.ktime, prev->ktime); + + time *= rps_to_i915(rps)->czclk_freq; + + /* Workload can be split between render + media, + * e.g. SwapBuffers being blitted in X after being rendered in + * mesa. To account for this we need to combine both engines + * into our activity counter. + */ + render = now.render_c0 - prev->render_c0; + media = now.media_c0 - prev->media_c0; + c0 = max(render, media); + c0 *= 1000 * 100 << 8; /* to usecs and scale to threshold% */ + + if (c0 > time * rps->power.up_threshold) + events = GEN6_PM_RP_UP_THRESHOLD; + else if (c0 < time * rps->power.down_threshold) + events = GEN6_PM_RP_DOWN_THRESHOLD; + } + + rps->ei = now; + return events; +} + +static void rps_work(struct work_struct *work) +{ + struct intel_rps *rps = container_of(work, typeof(*rps), work); + struct intel_gt *gt = rps_to_gt(rps); + bool client_boost = false; + int new_freq, adj, min, max; + u32 pm_iir = 0; + + spin_lock_irq(>->irq_lock); + pm_iir = fetch_and_zero(&rps->pm_iir); + client_boost = atomic_read(&rps->num_waiters); + spin_unlock_irq(>->irq_lock); + + /* Make sure we didn't queue anything we're not going to process. */ + if ((pm_iir & rps->pm_events) == 0 && !client_boost) + goto out; + + mutex_lock(&rps->lock); + + pm_iir |= vlv_wa_c0_ei(rps, pm_iir); + + adj = rps->last_adj; + new_freq = rps->cur_freq; + min = rps->min_freq_softlimit; + max = rps->max_freq_softlimit; + if (client_boost) + max = rps->max_freq; + if (client_boost && new_freq < rps->boost_freq) { + new_freq = rps->boost_freq; + adj = 0; + } else if (pm_iir & GEN6_PM_RP_UP_THRESHOLD) { + if (adj > 0) + adj *= 2; + else /* CHV needs even encode values */ + adj = IS_CHERRYVIEW(gt->i915) ? 2 : 1; + + if (new_freq >= rps->max_freq_softlimit) + adj = 0; + } else if (client_boost) { + adj = 0; + } else if (pm_iir & GEN6_PM_RP_DOWN_TIMEOUT) { + if (rps->cur_freq > rps->efficient_freq) + new_freq = rps->efficient_freq; + else if (rps->cur_freq > rps->min_freq_softlimit) + new_freq = rps->min_freq_softlimit; + adj = 0; + } else if (pm_iir & GEN6_PM_RP_DOWN_THRESHOLD) { + if (adj < 0) + adj *= 2; + else /* CHV needs even encode values */ + adj = IS_CHERRYVIEW(gt->i915) ? -2 : -1; + + if (new_freq <= rps->min_freq_softlimit) + adj = 0; + } else { /* unknown event */ + adj = 0; + } + + rps->last_adj = adj; + + /* + * Limit deboosting and boosting to keep ourselves at the extremes + * when in the respective power modes (i.e. slowly decrease frequencies + * while in the HIGH_POWER zone and slowly increase frequencies while + * in the LOW_POWER zone). On idle, we will hit the timeout and drop + * to the next level quickly, and conversely if busy we expect to + * hit a waitboost and rapidly switch into max power. + */ + if ((adj < 0 && rps->power.mode == HIGH_POWER) || + (adj > 0 && rps->power.mode == LOW_POWER)) + rps->last_adj = 0; + + /* sysfs frequency interfaces may have snuck in while servicing the + * interrupt + */ + new_freq += adj; + new_freq = clamp_t(int, new_freq, min, max); + + if (intel_rps_set(rps, new_freq)) { + DRM_DEBUG_DRIVER("Failed to set new GPU frequency\n"); + rps->last_adj = 0; + } + + mutex_unlock(&rps->lock); + +out: + spin_lock_irq(>->irq_lock); + gen6_gt_pm_unmask_irq(gt, rps->pm_events); + spin_unlock_irq(>->irq_lock); +} + +void gen11_rps_irq_handler(struct intel_rps *rps, u32 pm_iir) +{ + struct intel_gt *gt = rps_to_gt(rps); + const u32 events = rps->pm_events & pm_iir; + + lockdep_assert_held(>->irq_lock); + + if (unlikely(!events)) + return; + + gen6_gt_pm_mask_irq(gt, events); + + rps->pm_iir |= events; + schedule_work(&rps->work); +} + +void gen6_rps_irq_handler(struct intel_rps *rps, u32 pm_iir) +{ + struct intel_gt *gt = rps_to_gt(rps); + + if (pm_iir & rps->pm_events) { + spin_lock(>->irq_lock); + gen6_gt_pm_mask_irq(gt, pm_iir & rps->pm_events); + rps->pm_iir |= pm_iir & rps->pm_events; + schedule_work(&rps->work); + spin_unlock(>->irq_lock); + } + + if (INTEL_GEN(gt->i915) >= 8) + return; + + if (pm_iir & PM_VEBOX_USER_INTERRUPT) + intel_engine_breadcrumbs_irq(gt->engine[VECS0]); + + if (pm_iir & PM_VEBOX_CS_ERROR_INTERRUPT) + DRM_DEBUG("Command parser error, pm_iir 0x%08x\n", pm_iir); +} + +void gen5_rps_irq_handler(struct intel_rps *rps) +{ + struct intel_uncore *uncore = rps_to_uncore(rps); + u32 busy_up, busy_down, max_avg, min_avg; + u8 new_freq; + + spin_lock(&mchdev_lock); + + intel_uncore_write16(uncore, + MEMINTRSTS, + intel_uncore_read(uncore, MEMINTRSTS)); + + intel_uncore_write16(uncore, MEMINTRSTS, MEMINT_EVAL_CHG); + busy_up = intel_uncore_read(uncore, RCPREVBSYTUPAVG); + busy_down = intel_uncore_read(uncore, RCPREVBSYTDNAVG); + max_avg = intel_uncore_read(uncore, RCBMAXAVG); + min_avg = intel_uncore_read(uncore, RCBMINAVG); + + /* Handle RCS change request from hw */ + new_freq = rps->cur_freq; + if (busy_up > max_avg) + new_freq++; + else if (busy_down < min_avg) + new_freq--; + new_freq = clamp(new_freq, + rps->min_freq_softlimit, + rps->max_freq_softlimit); + + if (new_freq != rps->cur_freq && gen5_rps_set(rps, new_freq)) + rps->cur_freq = new_freq; + + spin_unlock(&mchdev_lock); +} + +void intel_rps_init_early(struct intel_rps *rps) +{ + mutex_init(&rps->lock); + mutex_init(&rps->power.mutex); + + INIT_WORK(&rps->work, rps_work); + + atomic_set(&rps->num_waiters, 0); +} + +void intel_rps_init(struct intel_rps *rps) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + + if (IS_CHERRYVIEW(i915)) + chv_rps_init(rps); + else if (IS_VALLEYVIEW(i915)) + vlv_rps_init(rps); + else if (INTEL_GEN(i915) >= 6) + gen6_rps_init(rps); + else if (IS_IRONLAKE_M(i915)) + gen5_rps_init(rps); + + /* Derive initial user preferences/limits from the hardware limits */ + rps->max_freq_softlimit = rps->max_freq; + rps->min_freq_softlimit = rps->min_freq; + + /* After setting max-softlimit, find the overclock max freq */ + if (IS_GEN(i915, 6) || IS_IVYBRIDGE(i915) || IS_HASWELL(i915)) { + u32 params = 0; + + sandybridge_pcode_read(i915, GEN6_READ_OC_PARAMS, + ¶ms, NULL); + if (params & BIT(31)) { /* OC supported */ + DRM_DEBUG_DRIVER("Overclocking supported, max: %dMHz, overclock: %dMHz\n", + (rps->max_freq & 0xff) * 50, + (params & 0xff) * 50); + rps->max_freq = params & 0xff; + } + } + + /* Finally allow us to boost to max by default */ + rps->boost_freq = rps->max_freq; + rps->idle_freq = rps->min_freq; + rps->cur_freq = rps->idle_freq; + + rps->pm_intrmsk_mbz = 0; + + /* + * SNB,IVB,HSW can while VLV,CHV may hard hang on looping batchbuffer + * if GEN6_PM_UP_EI_EXPIRED is masked. + * + * TODO: verify if this can be reproduced on VLV,CHV. + */ + if (INTEL_GEN(i915) <= 7) + rps->pm_intrmsk_mbz |= GEN6_PM_RP_UP_EI_EXPIRED; + + if (INTEL_GEN(i915) >= 8) + rps->pm_intrmsk_mbz |= GEN8_PMINTR_DISABLE_REDIRECT_TO_GUC; +} + +u32 intel_get_cagf(struct intel_rps *rps, u32 rpstat) +{ + struct drm_i915_private *i915 = rps_to_i915(rps); + u32 cagf; + + if (INTEL_GEN(i915) >= 9) + cagf = (rpstat & GEN9_CAGF_MASK) >> GEN9_CAGF_SHIFT; + else if (IS_HASWELL(i915) || IS_BROADWELL(i915)) + cagf = (rpstat & HSW_CAGF_MASK) >> HSW_CAGF_SHIFT; + else + cagf = (rpstat & GEN6_CAGF_MASK) >> GEN6_CAGF_SHIFT; + + return cagf; +} + +/* External interface for intel_ips.ko */ + +static struct drm_i915_private __rcu *ips_mchdev; + +/** + * Tells the intel_ips driver that the i915 driver is now loaded, if + * IPS got loaded first. + * + * This awkward dance is so that neither module has to depend on the + * other in order for IPS to do the appropriate communication of + * GPU turbo limits to i915. + */ +static void +ips_ping_for_i915_load(void) +{ + void (*link)(void); + + link = symbol_get(ips_link_to_i915_driver); + if (link) { + link(); + symbol_put(ips_link_to_i915_driver); + } +} + +void intel_rps_driver_register(struct intel_rps *rps) +{ + struct intel_gt *gt = rps_to_gt(rps); + + /* + * We only register the i915 ips part with intel-ips once everything is + * set up, to avoid intel-ips sneaking in and reading bogus values. + */ + if (IS_GEN(gt->i915, 5)) { + rcu_assign_pointer(ips_mchdev, gt->i915); + ips_ping_for_i915_load(); + } +} + +void intel_rps_driver_unregister(struct intel_rps *rps) +{ + rcu_assign_pointer(ips_mchdev, NULL); +} + +static struct drm_i915_private *mchdev_get(void) +{ + struct drm_i915_private *i915; + + rcu_read_lock(); + i915 = rcu_dereference(ips_mchdev); + if (!kref_get_unless_zero(&i915->drm.ref)) + i915 = NULL; + rcu_read_unlock(); + + return i915; +} + +/** + * i915_read_mch_val - return value for IPS use + * + * Calculate and return a value for the IPS driver to use when deciding whether + * we have thermal and power headroom to increase CPU or GPU power budget. + */ +unsigned long i915_read_mch_val(void) +{ + struct drm_i915_private *i915; + unsigned long chipset_val = 0; + unsigned long graphics_val = 0; + intel_wakeref_t wakeref; + + i915 = mchdev_get(); + if (!i915) + return 0; + + with_intel_runtime_pm(&i915->runtime_pm, wakeref) { + struct intel_ips *ips = &i915->gt.rps.ips; + + spin_lock_irq(&mchdev_lock); + chipset_val = __ips_chipset_val(ips); + graphics_val = __ips_gfx_val(ips); + spin_unlock_irq(&mchdev_lock); + } + + drm_dev_put(&i915->drm); + return chipset_val + graphics_val; +} +EXPORT_SYMBOL_GPL(i915_read_mch_val); + +/** + * i915_gpu_raise - raise GPU frequency limit + * + * Raise the limit; IPS indicates we have thermal headroom. + */ +bool i915_gpu_raise(void) +{ + struct drm_i915_private *i915; + struct intel_rps *rps; + + i915 = mchdev_get(); + if (!i915) + return false; + + rps = &i915->gt.rps; + + spin_lock_irq(&mchdev_lock); + if (rps->max_freq_softlimit < rps->max_freq) + rps->max_freq_softlimit++; + spin_unlock_irq(&mchdev_lock); + + drm_dev_put(&i915->drm); + return true; +} +EXPORT_SYMBOL_GPL(i915_gpu_raise); + +/** + * i915_gpu_lower - lower GPU frequency limit + * + * IPS indicates we're close to a thermal limit, so throttle back the GPU + * frequency maximum. + */ +bool i915_gpu_lower(void) +{ + struct drm_i915_private *i915; + struct intel_rps *rps; + + i915 = mchdev_get(); + if (!i915) + return false; + + rps = &i915->gt.rps; + + spin_lock_irq(&mchdev_lock); + if (rps->max_freq_softlimit > rps->min_freq) + rps->max_freq_softlimit--; + spin_unlock_irq(&mchdev_lock); + + drm_dev_put(&i915->drm); + return true; +} +EXPORT_SYMBOL_GPL(i915_gpu_lower); + +/** + * i915_gpu_busy - indicate GPU business to IPS + * + * Tell the IPS driver whether or not the GPU is busy. + */ +bool i915_gpu_busy(void) +{ + struct drm_i915_private *i915; + bool ret; + + i915 = mchdev_get(); + if (!i915) + return false; + + ret = i915->gt.awake; + + drm_dev_put(&i915->drm); + return ret; +} +EXPORT_SYMBOL_GPL(i915_gpu_busy); + +/** + * i915_gpu_turbo_disable - disable graphics turbo + * + * Disable graphics turbo by resetting the max frequency and setting the + * current frequency to the default. + */ +bool i915_gpu_turbo_disable(void) +{ + struct drm_i915_private *i915; + struct intel_rps *rps; + bool ret; + + i915 = mchdev_get(); + if (!i915) + return false; + + rps = &i915->gt.rps; + + spin_lock_irq(&mchdev_lock); + rps->max_freq_softlimit = rps->min_freq; + ret = gen5_rps_set(&i915->gt.rps, rps->min_freq); + spin_unlock_irq(&mchdev_lock); + + drm_dev_put(&i915->drm); + return ret; +} +EXPORT_SYMBOL_GPL(i915_gpu_turbo_disable); diff --git a/drivers/gpu/drm/i915/gt/intel_rps.h b/drivers/gpu/drm/i915/gt/intel_rps.h new file mode 100644 index 000000000000..9518c66c9792 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rps.h @@ -0,0 +1,38 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_RPS_H +#define INTEL_RPS_H + +#include "intel_rps_types.h" + +struct i915_request; + +void intel_rps_init_early(struct intel_rps *rps); +void intel_rps_init(struct intel_rps *rps); + +void intel_rps_driver_register(struct intel_rps *rps); +void intel_rps_driver_unregister(struct intel_rps *rps); + +void intel_rps_enable(struct intel_rps *rps); +void intel_rps_disable(struct intel_rps *rps); + +void intel_rps_park(struct intel_rps *rps); +void intel_rps_unpark(struct intel_rps *rps); +void intel_rps_boost(struct i915_request *rq); + +int intel_rps_set(struct intel_rps *rps, u8 val); +void intel_rps_mark_interactive(struct intel_rps *rps, bool interactive); + +int intel_gpu_freq(struct intel_rps *rps, int val); +int intel_freq_opcode(struct intel_rps *rps, int val); +u32 intel_get_cagf(struct intel_rps *rps, u32 rpstat1); + +void gen5_rps_irq_handler(struct intel_rps *rps); +void gen6_rps_irq_handler(struct intel_rps *rps, u32 pm_iir); +void gen11_rps_irq_handler(struct intel_rps *rps, u32 pm_iir); + +#endif /* INTEL_RPS_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_rps_types.h b/drivers/gpu/drm/i915/gt/intel_rps_types.h new file mode 100644 index 000000000000..c2e279154bd5 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rps_types.h @@ -0,0 +1,93 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_RPS_TYPES_H +#define INTEL_RPS_TYPES_H + +#include <linux/atomic.h> +#include <linux/ktime.h> +#include <linux/mutex.h> +#include <linux/types.h> +#include <linux/workqueue.h> + +struct intel_ips { + u64 last_count1; + unsigned long last_time1; + unsigned long chipset_power; + u64 last_count2; + u64 last_time2; + unsigned long gfx_power; + u8 corr; + + int c, m; +}; + +struct intel_rps_ei { + ktime_t ktime; + u32 render_c0; + u32 media_c0; +}; + +struct intel_rps { + struct mutex lock; /* protects enabling and the worker */ + + /* + * work, interrupts_enabled and pm_iir are protected by + * dev_priv->irq_lock + */ + struct work_struct work; + bool enabled; + bool active; + u32 pm_iir; + + /* PM interrupt bits that should never be masked */ + u32 pm_intrmsk_mbz; + u32 pm_events; + + /* Frequencies are stored in potentially platform dependent multiples. + * In other words, *_freq needs to be multiplied by X to be interesting. + * Soft limits are those which are used for the dynamic reclocking done + * by the driver (raise frequencies under heavy loads, and lower for + * lighter loads). Hard limits are those imposed by the hardware. + * + * A distinction is made for overclocking, which is never enabled by + * default, and is considered to be above the hard limit if it's + * possible at all. + */ + u8 cur_freq; /* Current frequency (cached, may not == HW) */ + u8 last_freq; /* Last SWREQ frequency */ + u8 min_freq_softlimit; /* Minimum frequency permitted by the driver */ + u8 max_freq_softlimit; /* Max frequency permitted by the driver */ + u8 max_freq; /* Maximum frequency, RP0 if not overclocking */ + u8 min_freq; /* AKA RPn. Minimum frequency */ + u8 boost_freq; /* Frequency to request when wait boosting */ + u8 idle_freq; /* Frequency to request when we are idle */ + u8 efficient_freq; /* AKA RPe. Pre-determined balanced frequency */ + u8 rp1_freq; /* "less than" RP0 power/freqency */ + u8 rp0_freq; /* Non-overclocked max frequency. */ + u16 gpll_ref_freq; /* vlv/chv GPLL reference frequency */ + + int last_adj; + + struct { + struct mutex mutex; + + enum { LOW_POWER, BETWEEN, HIGH_POWER } mode; + unsigned int interactive; + + u8 up_threshold; /* Current %busy required to uplock */ + u8 down_threshold; /* Current %busy required to downclock */ + } power; + + atomic_t num_waiters; + atomic_t boosts; + + /* manual wa residency calculations */ + struct intel_rps_ei ei; + struct intel_ips ips; +}; + +#endif /* INTEL_RPS_TYPES_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_sseu.c b/drivers/gpu/drm/i915/gt/intel_sseu.c index 6bf2d87da109..74f793423231 100644 --- a/drivers/gpu/drm/i915/gt/intel_sseu.c +++ b/drivers/gpu/drm/i915/gt/intel_sseu.c @@ -8,6 +8,19 @@ #include "intel_lrc_reg.h" #include "intel_sseu.h" +void intel_sseu_set_info(struct sseu_dev_info *sseu, u8 max_slices, + u8 max_subslices, u8 max_eus_per_subslice) +{ + sseu->max_slices = max_slices; + sseu->max_subslices = max_subslices; + sseu->max_eus_per_subslice = max_eus_per_subslice; + + sseu->ss_stride = GEN_SSEU_STRIDE(sseu->max_subslices); + GEM_BUG_ON(sseu->ss_stride > GEN_MAX_SUBSLICE_STRIDE); + sseu->eu_stride = GEN_SSEU_STRIDE(sseu->max_eus_per_subslice); + GEM_BUG_ON(sseu->eu_stride > GEN_MAX_EU_STRIDE); +} + unsigned int intel_sseu_subslice_total(const struct sseu_dev_info *sseu) { @@ -19,10 +32,32 @@ intel_sseu_subslice_total(const struct sseu_dev_info *sseu) return total; } +u32 intel_sseu_get_subslices(const struct sseu_dev_info *sseu, u8 slice) +{ + int i, offset = slice * sseu->ss_stride; + u32 mask = 0; + + GEM_BUG_ON(slice >= sseu->max_slices); + + for (i = 0; i < sseu->ss_stride; i++) + mask |= (u32)sseu->subslice_mask[offset + i] << + i * BITS_PER_BYTE; + + return mask; +} + +void intel_sseu_set_subslices(struct sseu_dev_info *sseu, int slice, + u32 ss_mask) +{ + int offset = slice * sseu->ss_stride; + + memcpy(&sseu->subslice_mask[offset], &ss_mask, sseu->ss_stride); +} + unsigned int intel_sseu_subslices_per_slice(const struct sseu_dev_info *sseu, u8 slice) { - return hweight8(sseu->subslice_mask[slice]); + return hweight32(intel_sseu_get_subslices(sseu, slice)); } u32 intel_sseu_make_rpcs(struct drm_i915_private *i915, diff --git a/drivers/gpu/drm/i915/gt/intel_sseu.h b/drivers/gpu/drm/i915/gt/intel_sseu.h index b50d0401a4e2..d1d225204f09 100644 --- a/drivers/gpu/drm/i915/gt/intel_sseu.h +++ b/drivers/gpu/drm/i915/gt/intel_sseu.h @@ -10,15 +10,21 @@ #include <linux/types.h> #include <linux/kernel.h> +#include "i915_gem.h" + struct drm_i915_private; #define GEN_MAX_SLICES (6) /* CNL upper bound */ #define GEN_MAX_SUBSLICES (8) /* ICL upper bound */ #define GEN_SSEU_STRIDE(max_entries) DIV_ROUND_UP(max_entries, BITS_PER_BYTE) +#define GEN_MAX_SUBSLICE_STRIDE GEN_SSEU_STRIDE(GEN_MAX_SUBSLICES) +#define GEN_MAX_EUS (16) /* TGL upper bound */ +#define GEN_MAX_EU_STRIDE GEN_SSEU_STRIDE(GEN_MAX_EUS) struct sseu_dev_info { u8 slice_mask; - u8 subslice_mask[GEN_MAX_SLICES]; + u8 subslice_mask[GEN_MAX_SLICES * GEN_MAX_SUBSLICE_STRIDE]; + u8 eu_mask[GEN_MAX_SLICES * GEN_MAX_SUBSLICES * GEN_MAX_EU_STRIDE]; u16 eu_total; u8 eu_per_subslice; u8 min_eu_in_pool; @@ -33,11 +39,8 @@ struct sseu_dev_info { u8 max_subslices; u8 max_eus_per_subslice; - /* We don't have more than 8 eus per subslice at the moment and as we - * store eus enabled using bits, no need to multiply by eus per - * subslice. - */ - u8 eu_mask[GEN_MAX_SLICES * GEN_MAX_SUBSLICES]; + u8 ss_stride; + u8 eu_stride; }; /* @@ -63,12 +66,34 @@ intel_sseu_from_device_info(const struct sseu_dev_info *sseu) return value; } +static inline bool +intel_sseu_has_subslice(const struct sseu_dev_info *sseu, int slice, + int subslice) +{ + u8 mask; + int ss_idx = subslice / BITS_PER_BYTE; + + GEM_BUG_ON(ss_idx >= sseu->ss_stride); + + mask = sseu->subslice_mask[slice * sseu->ss_stride + ss_idx]; + + return mask & BIT(subslice % BITS_PER_BYTE); +} + +void intel_sseu_set_info(struct sseu_dev_info *sseu, u8 max_slices, + u8 max_subslices, u8 max_eus_per_subslice); + unsigned int intel_sseu_subslice_total(const struct sseu_dev_info *sseu); unsigned int intel_sseu_subslices_per_slice(const struct sseu_dev_info *sseu, u8 slice); +u32 intel_sseu_get_subslices(const struct sseu_dev_info *sseu, u8 slice); + +void intel_sseu_set_subslices(struct sseu_dev_info *sseu, int slice, + u32 ss_mask); + u32 intel_sseu_make_rpcs(struct drm_i915_private *i915, const struct intel_sseu *req_sseu); diff --git a/drivers/gpu/drm/i915/gt/intel_timeline.c b/drivers/gpu/drm/i915/gt/intel_timeline.c index 9cb01d9828f1..649798c184fb 100644 --- a/drivers/gpu/drm/i915/gt/intel_timeline.c +++ b/drivers/gpu/drm/i915/gt/intel_timeline.c @@ -4,13 +4,13 @@ * Copyright © 2016-2018 Intel Corporation */ -#include "gt/intel_gt_types.h" - #include "i915_drv.h" #include "i915_active.h" #include "i915_syncmap.h" -#include "gt/intel_timeline.h" +#include "intel_gt.h" +#include "intel_ring.h" +#include "intel_timeline.h" #define ptr_set_bit(ptr, bit) ((typeof(ptr))((unsigned long)(ptr) | BIT(bit))) #define ptr_test_bit(ptr, bit) ((unsigned long)(ptr) & BIT(bit)) @@ -136,6 +136,7 @@ static void __idle_cacheline_free(struct intel_timeline_cacheline *cl) kfree(cl); } +__i915_active_call static void __cacheline_retire(struct i915_active *active) { struct intel_timeline_cacheline *cl = @@ -177,8 +178,7 @@ cacheline_alloc(struct intel_timeline_hwsp *hwsp, unsigned int cacheline) cl->hwsp = hwsp; cl->vaddr = page_pack_bits(vaddr, cacheline); - i915_active_init(hwsp->gt->i915, &cl->active, - __cacheline_active, __cacheline_retire); + i915_active_init(&cl->active, __cacheline_active, __cacheline_retire); return cl; } @@ -254,7 +254,7 @@ int intel_timeline_init(struct intel_timeline *timeline, mutex_init(&timeline->mutex); - INIT_ACTIVE_REQUEST(&timeline->last_request, &timeline->mutex); + INIT_ACTIVE_FENCE(&timeline->last_request, &timeline->mutex); INIT_LIST_HEAD(&timeline->requests); i915_syncmap_init(&timeline->sync); @@ -282,6 +282,7 @@ void intel_timeline_fini(struct intel_timeline *timeline) { GEM_BUG_ON(atomic_read(&timeline->pin_count)); GEM_BUG_ON(!list_empty(&timeline->requests)); + GEM_BUG_ON(timeline->retire); if (timeline->hwsp_cacheline) cacheline_free(timeline->hwsp_cacheline); @@ -339,15 +340,33 @@ void intel_timeline_enter(struct intel_timeline *tl) struct intel_gt_timelines *timelines = &tl->gt->timelines; unsigned long flags; + /* + * Pretend we are serialised by the timeline->mutex. + * + * While generally true, there are a few exceptions to the rule + * for the engine->kernel_context being used to manage power + * transitions. As the engine_park may be called from under any + * timeline, it uses the power mutex as a global serialisation + * lock to prevent any other request entering its timeline. + * + * The rule is generally tl->mutex, otherwise engine->wakeref.mutex. + * + * However, intel_gt_retire_request() does not know which engine + * it is retiring along and so cannot partake in the engine-pm + * barrier, and there we use the tl->active_count as a means to + * pin the timeline in the active_list while the locks are dropped. + * Ergo, as that is outside of the engine-pm barrier, we need to + * use atomic to manipulate tl->active_count. + */ lockdep_assert_held(&tl->mutex); - GEM_BUG_ON(!atomic_read(&tl->pin_count)); - if (tl->active_count++) + + if (atomic_add_unless(&tl->active_count, 1, 0)) return; - GEM_BUG_ON(!tl->active_count); /* overflow? */ spin_lock_irqsave(&timelines->lock, flags); - list_add(&tl->link, &timelines->active_list); + if (!atomic_fetch_inc(&tl->active_count)) + list_add_tail(&tl->link, &timelines->active_list); spin_unlock_irqrestore(&timelines->lock, flags); } @@ -356,14 +375,16 @@ void intel_timeline_exit(struct intel_timeline *tl) struct intel_gt_timelines *timelines = &tl->gt->timelines; unsigned long flags; + /* See intel_timeline_enter() */ lockdep_assert_held(&tl->mutex); - GEM_BUG_ON(!tl->active_count); - if (--tl->active_count) + GEM_BUG_ON(!atomic_read(&tl->active_count)); + if (atomic_add_unless(&tl->active_count, -1, 1)) return; spin_lock_irqsave(&timelines->lock, flags); - list_del(&tl->link); + if (atomic_dec_and_test(&tl->active_count)) + list_del(&tl->link); spin_unlock_irqrestore(&timelines->lock, flags); /* @@ -442,7 +463,7 @@ __intel_timeline_get_seqno(struct intel_timeline *tl, * free it after the current request is retired, which ensures that * all writes into the cacheline from previous requests are complete. */ - err = i915_active_ref(&tl->hwsp_cacheline->active, tl, rq); + err = i915_active_ref(&tl->hwsp_cacheline->active, tl, &rq->fence); if (err) goto err_cacheline; @@ -493,24 +514,39 @@ int intel_timeline_get_seqno(struct intel_timeline *tl, static int cacheline_ref(struct intel_timeline_cacheline *cl, struct i915_request *rq) { - return i915_active_ref(&cl->active, rq->timeline, rq); + return i915_active_add_request(&cl->active, rq); } int intel_timeline_read_hwsp(struct i915_request *from, struct i915_request *to, u32 *hwsp) { - struct intel_timeline_cacheline *cl = from->hwsp_cacheline; - struct intel_timeline *tl = from->timeline; + struct intel_timeline *tl; int err; - GEM_BUG_ON(to->timeline == tl); + rcu_read_lock(); + tl = rcu_dereference(from->timeline); + if (i915_request_completed(from) || !kref_get_unless_zero(&tl->kref)) + tl = NULL; + rcu_read_unlock(); + if (!tl) /* already completed */ + return 1; + + GEM_BUG_ON(rcu_access_pointer(to->timeline) == tl); + + err = -EBUSY; + if (mutex_trylock(&tl->mutex)) { + struct intel_timeline_cacheline *cl = from->hwsp_cacheline; + + if (i915_request_completed(from)) { + err = 1; + goto unlock; + } - mutex_lock_nested(&tl->mutex, SINGLE_DEPTH_NESTING); - err = i915_request_completed(from); - if (!err) err = cacheline_ref(cl, to); - if (!err) { + if (err) + goto unlock; + if (likely(cl == tl->hwsp_cacheline)) { *hwsp = tl->hwsp_offset; } else { /* across a seqno wrap, recover the original offset */ @@ -518,8 +554,11 @@ int intel_timeline_read_hwsp(struct i915_request *from, ptr_unmask_bits(cl->vaddr, CACHELINE_BITS) * CACHELINE_BYTES; } + +unlock: + mutex_unlock(&tl->mutex); } - mutex_unlock(&tl->mutex); + intel_timeline_put(tl); return err; } @@ -541,7 +580,7 @@ void __intel_timeline_free(struct kref *kref) container_of(kref, typeof(*timeline), kref); intel_timeline_fini(timeline); - kfree(timeline); + kfree_rcu(timeline, rcu); } static void timelines_fini(struct intel_gt *gt) diff --git a/drivers/gpu/drm/i915/gt/intel_timeline_types.h b/drivers/gpu/drm/i915/gt/intel_timeline_types.h index 2b1baf2fcc8e..aaf15cbe1ce1 100644 --- a/drivers/gpu/drm/i915/gt/intel_timeline_types.h +++ b/drivers/gpu/drm/i915/gt/intel_timeline_types.h @@ -42,7 +42,7 @@ struct intel_timeline { * from the intel_context caller plus internal atomicity. */ atomic_t pin_count; - unsigned int active_count; + atomic_t active_count; const u32 *hwsp_seqno; struct i915_vma *hwsp_ggtt; @@ -58,12 +58,16 @@ struct intel_timeline { */ struct list_head requests; - /* Contains an RCU guarded pointer to the last request. No reference is + /* + * Contains an RCU guarded pointer to the last request. No reference is * held to the request, users must carefully acquire a reference to - * the request using i915_active_request_get_request_rcu(), or hold the - * struct_mutex. + * the request using i915_active_fence_get(), or manage the RCU + * protection themselves (cf the i915_active_fence API). */ - struct i915_active_request last_request; + struct i915_active_fence last_request; + + /** A chain of completed timelines ready for early retirement. */ + struct intel_timeline *retire; /** * We track the most recent seqno that we wait on in every context so @@ -80,6 +84,7 @@ struct intel_timeline { struct intel_gt *gt; struct kref kref; + struct rcu_head rcu; }; #endif /* __I915_TIMELINE_TYPES_H__ */ diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c index 45481eb1fa3c..e4bccc14602f 100644 --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c @@ -7,6 +7,7 @@ #include "i915_drv.h" #include "intel_context.h" #include "intel_gt.h" +#include "intel_ring.h" #include "intel_workarounds.h" /** @@ -567,6 +568,9 @@ static void icl_ctx_workarounds_init(struct intel_engine_cs *engine, static void tgl_ctx_workarounds_init(struct intel_engine_cs *engine, struct i915_wa_list *wal) { + /* Wa_1409142259:tgl */ + WA_SET_BIT_MASKED(GEN11_COMMON_SLICE_CHICKEN3, + GEN12_DISABLE_CPS_AWARE_COLOR_PIPE); } static void @@ -796,11 +800,10 @@ wa_init_mcr(struct drm_i915_private *i915, struct i915_wa_list *wal) } slice = fls(sseu->slice_mask) - 1; - GEM_BUG_ON(slice >= ARRAY_SIZE(sseu->subslice_mask)); - subslice = fls(l3_en & sseu->subslice_mask[slice]); + subslice = fls(l3_en & intel_sseu_get_subslices(sseu, slice)); if (!subslice) { DRM_WARN("No common index found between subslice mask %x and L3 bank mask %x!\n", - sseu->subslice_mask[slice], l3_en); + intel_sseu_get_subslices(sseu, slice), l3_en); subslice = fls(l3_en); WARN_ON(!subslice); } @@ -890,11 +893,27 @@ icl_gt_workarounds_init(struct drm_i915_private *i915, struct i915_wa_list *wal) wa_write_or(wal, GAMT_CHKN_BIT_REG, GAMT_CHKN_DISABLE_L3_COH_PIPE); + + /* Wa_1607087056:icl */ + wa_write_or(wal, + SLICE_UNIT_LEVEL_CLKGATE, + L3_CLKGATE_DIS | L3_CR2X_CLKGATE_DIS); } static void tgl_gt_workarounds_init(struct drm_i915_private *i915, struct i915_wa_list *wal) { + /* Wa_1409420604:tgl */ + if (IS_TGL_REVID(i915, TGL_REVID_A0, TGL_REVID_A0)) + wa_write_or(wal, + SUBSLICE_UNIT_LEVEL_CLKGATE2, + CPSSUNIT_CLKGATE_DIS); + + /* Wa_1409180338:tgl */ + if (IS_TGL_REVID(i915, TGL_REVID_A0, TGL_REVID_A0)) + wa_write_or(wal, + SLICE_UNIT_LEVEL_CLKGATE, + L3_CLKGATE_DIS | L3_CR2X_CLKGATE_DIS); } static void @@ -1063,6 +1082,9 @@ static void gen9_whitelist_build(struct i915_wa_list *w) /* WaAllowUMDToModifyHDCChicken1:skl,bxt,kbl,glk,cfl */ whitelist_reg(w, GEN8_HDC_CHICKEN1); + + /* WaSendPushConstantsFromMMIO:skl,bxt */ + whitelist_reg(w, COMMON_SLICE_CHICKEN2); } static void skl_whitelist_build(struct intel_engine_cs *engine) @@ -1194,6 +1216,26 @@ static void icl_whitelist_build(struct intel_engine_cs *engine) static void tgl_whitelist_build(struct intel_engine_cs *engine) { + struct i915_wa_list *w = &engine->whitelist; + + switch (engine->class) { + case RENDER_CLASS: + /* + * WaAllowPMDepthAndInvocationCountAccessFromUMD:tgl + * + * This covers 4 registers which are next to one another : + * - PS_INVOCATION_COUNT + * - PS_INVOCATION_COUNT_UDW + * - PS_DEPTH_COUNT + * - PS_DEPTH_COUNT_UDW + */ + whitelist_reg_ext(w, PS_INVOCATION_COUNT, + RING_FORCE_TO_NONPRIV_ACCESS_RD | + RING_FORCE_TO_NONPRIV_RANGE_4); + break; + default: + break; + } } void intel_engine_init_whitelist(struct intel_engine_cs *engine) @@ -1255,6 +1297,26 @@ rcs_engine_wa_init(struct intel_engine_cs *engine, struct i915_wa_list *wal) { struct drm_i915_private *i915 = engine->i915; + if (IS_TGL_REVID(i915, TGL_REVID_A0, TGL_REVID_A0)) { + /* Wa_1606700617:tgl */ + wa_masked_en(wal, + GEN9_CS_DEBUG_MODE1, + FF_DOP_CLOCK_GATE_DISABLE); + + /* Wa_1607138336:tgl */ + wa_write_or(wal, + GEN9_CTX_PREEMPT_REG, + GEN12_DISABLE_POSH_BUSY_FF_DOP_CG); + + /* Wa_1607030317:tgl */ + /* Wa_1607186500:tgl */ + /* Wa_1607297627:tgl */ + wa_masked_en(wal, + GEN6_RC_SLEEP_PSMI_CONTROL, + GEN12_WAIT_FOR_EVENT_POWER_DOWN_DISABLE | + GEN8_RC_SEMA_IDLE_MSG_DISABLE); + } + if (IS_GEN(i915, 11)) { /* This is not an Wa. Enable for better image quality */ wa_masked_en(wal, @@ -1449,7 +1511,7 @@ static bool mcr_range(struct drm_i915_private *i915, u32 offset) * which only controls CPU initiated MMIO. Routing does not * work for CS access so we cannot verify them on this path. */ - if (INTEL_GEN(i915) >= 8 && (offset >= 0xb100 && offset <= 0xb3ff)) + if (INTEL_GEN(i915) >= 8 && (offset >= 0xb000 && offset <= 0xb4ff)) return true; return false; diff --git a/drivers/gpu/drm/i915/gt/mock_engine.c b/drivers/gpu/drm/i915/gt/mock_engine.c index 5d43cbc3f345..83f549d203a0 100644 --- a/drivers/gpu/drm/i915/gt/mock_engine.c +++ b/drivers/gpu/drm/i915/gt/mock_engine.c @@ -23,6 +23,7 @@ */ #include "gem/i915_gem_context.h" +#include "gt/intel_ring.h" #include "i915_drv.h" #include "intel_context.h" @@ -240,6 +241,7 @@ struct intel_engine_cs *mock_engine(struct drm_i915_private *i915, struct mock_engine *engine; GEM_BUG_ON(id >= I915_NUM_ENGINES); + GEM_BUG_ON(!i915->gt.uncore); engine = kzalloc(sizeof(*engine) + PAGE_SIZE, GFP_KERNEL); if (!engine) @@ -248,9 +250,11 @@ struct intel_engine_cs *mock_engine(struct drm_i915_private *i915, /* minimal engine setup for requests */ engine->base.i915 = i915; engine->base.gt = &i915->gt; + engine->base.uncore = i915->gt.uncore; snprintf(engine->base.name, sizeof(engine->base.name), "%s", name); engine->base.id = id; engine->base.mask = BIT(id); + engine->base.legacy_idx = INVALID_ENGINE; engine->base.instance = id; engine->base.status_page.addr = (void *)(engine + 1); @@ -265,6 +269,9 @@ struct intel_engine_cs *mock_engine(struct drm_i915_private *i915, engine->base.reset.finish = mock_reset_finish; engine->base.cancel_requests = mock_cancel_requests; + i915->gt.engine[id] = &engine->base; + i915->gt.engine_class[0][id] = &engine->base; + /* fake hw queue */ spin_lock_init(&engine->hw_lock); timer_setup(&engine->hw_delay, hw_delay_complete, 0); diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c index 9d1ea26c7a2d..bc720defc6b8 100644 --- a/drivers/gpu/drm/i915/gt/selftest_context.c +++ b/drivers/gpu/drm/i915/gt/selftest_context.c @@ -14,22 +14,28 @@ static int request_sync(struct i915_request *rq) { + struct intel_timeline *tl = i915_request_timeline(rq); long timeout; int err = 0; + intel_timeline_get(tl); i915_request_get(rq); - i915_request_add(rq); + /* Opencode i915_request_add() so we can keep the timeline locked. */ + __i915_request_commit(rq); + __i915_request_queue(rq, NULL); + timeout = i915_request_wait(rq, 0, HZ / 10); - if (timeout < 0) { + if (timeout < 0) err = timeout; - } else { - mutex_lock(&rq->timeline->mutex); + else i915_request_retire_upto(rq); - mutex_unlock(&rq->timeline->mutex); - } + + lockdep_unpin_lock(&tl->mutex, rq->cookie); + mutex_unlock(&tl->mutex); i915_request_put(rq); + intel_timeline_put(tl); return err; } @@ -41,24 +47,20 @@ static int context_sync(struct intel_context *ce) mutex_lock(&tl->mutex); do { - struct i915_request *rq; + struct dma_fence *fence; long timeout; - rcu_read_lock(); - rq = rcu_dereference(tl->last_request.request); - if (rq) - rq = i915_request_get_rcu(rq); - rcu_read_unlock(); - if (!rq) + fence = i915_active_fence_get(&tl->last_request); + if (!fence) break; - timeout = i915_request_wait(rq, 0, HZ / 10); + timeout = dma_fence_wait_timeout(fence, false, HZ / 10); if (timeout < 0) err = timeout; else - i915_request_retire_upto(rq); + i915_request_retire_upto(to_request(fence)); - i915_request_put(rq); + dma_fence_put(fence); } while (!err); mutex_unlock(&tl->mutex); @@ -101,9 +103,6 @@ static int __live_context_size(struct intel_engine_cs *engine, * * TLDR; this overlaps with the execlists redzone. */ - if (HAS_EXECLISTS(engine->i915)) - vaddr += LRC_HEADER_PAGES * PAGE_SIZE; - vaddr += engine->context_size - I915_GTT_PAGE_SIZE; memset(vaddr, POISON_INUSE, I915_GTT_PAGE_SIZE); @@ -153,15 +152,11 @@ static int live_context_size(void *arg) * HW tries to write past the end of one. */ - mutex_lock(>->i915->drm.struct_mutex); - fixme = kernel_context(gt->i915); - if (IS_ERR(fixme)) { - err = PTR_ERR(fixme); - goto unlock; - } + if (IS_ERR(fixme)) + return PTR_ERR(fixme); - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { struct { struct drm_i915_gem_object *state; void *pinned; @@ -199,8 +194,6 @@ static int live_context_size(void *arg) } kernel_context_close(fixme); -unlock: - mutex_unlock(>->i915->drm.struct_mutex); return err; } @@ -303,26 +296,23 @@ static int live_active_context(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(>->i915->drm.struct_mutex); - fixme = live_context(gt->i915, file); if (IS_ERR(fixme)) { err = PTR_ERR(fixme); - goto unlock; + goto out_file; } - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { err = __live_active_context(engine, fixme); if (err) break; - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + err = igt_flush_test(gt->i915); if (err) break; } -unlock: - mutex_unlock(>->i915->drm.struct_mutex); +out_file: mock_file_free(gt->i915, file); return err; } @@ -416,26 +406,23 @@ static int live_remote_context(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(>->i915->drm.struct_mutex); - fixme = live_context(gt->i915, file); if (IS_ERR(fixme)) { err = PTR_ERR(fixme); - goto unlock; + goto out_file; } - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { err = __live_remote_context(engine, fixme); if (err) break; - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + err = igt_flush_test(gt->i915); if (err) break; } -unlock: - mutex_unlock(>->i915->drm.struct_mutex); +out_file: mock_file_free(gt->i915, file); return err; } diff --git a/drivers/gpu/drm/i915/gt/selftest_engine_heartbeat.c b/drivers/gpu/drm/i915/gt/selftest_engine_heartbeat.c new file mode 100644 index 000000000000..e864406bd2d9 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/selftest_engine_heartbeat.c @@ -0,0 +1,350 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2018 Intel Corporation + */ + +#include <linux/sort.h> + +#include "i915_drv.h" + +#include "intel_gt_requests.h" +#include "i915_selftest.h" + +struct pulse { + struct i915_active active; + struct kref kref; +}; + +static int pulse_active(struct i915_active *active) +{ + kref_get(&container_of(active, struct pulse, active)->kref); + return 0; +} + +static void pulse_free(struct kref *kref) +{ + kfree(container_of(kref, struct pulse, kref)); +} + +static void pulse_put(struct pulse *p) +{ + kref_put(&p->kref, pulse_free); +} + +static void pulse_retire(struct i915_active *active) +{ + pulse_put(container_of(active, struct pulse, active)); +} + +static struct pulse *pulse_create(void) +{ + struct pulse *p; + + p = kmalloc(sizeof(*p), GFP_KERNEL); + if (!p) + return p; + + kref_init(&p->kref); + i915_active_init(&p->active, pulse_active, pulse_retire); + + return p; +} + +static void pulse_unlock_wait(struct pulse *p) +{ + mutex_lock(&p->active.mutex); + mutex_unlock(&p->active.mutex); + flush_work(&p->active.work); +} + +static int __live_idle_pulse(struct intel_engine_cs *engine, + int (*fn)(struct intel_engine_cs *cs)) +{ + struct pulse *p; + int err; + + GEM_BUG_ON(!intel_engine_pm_is_awake(engine)); + + p = pulse_create(); + if (!p) + return -ENOMEM; + + err = i915_active_acquire(&p->active); + if (err) + goto out; + + err = i915_active_acquire_preallocate_barrier(&p->active, engine); + if (err) { + i915_active_release(&p->active); + goto out; + } + + i915_active_acquire_barrier(&p->active); + i915_active_release(&p->active); + + GEM_BUG_ON(i915_active_is_idle(&p->active)); + GEM_BUG_ON(llist_empty(&engine->barrier_tasks)); + + err = fn(engine); + if (err) + goto out; + + GEM_BUG_ON(!llist_empty(&engine->barrier_tasks)); + + if (intel_gt_retire_requests_timeout(engine->gt, HZ / 5)) { + err = -ETIME; + goto out; + } + + GEM_BUG_ON(READ_ONCE(engine->serial) != engine->wakeref_serial); + + pulse_unlock_wait(p); /* synchronize with the retirement callback */ + + if (!i915_active_is_idle(&p->active)) { + struct drm_printer m = drm_err_printer("pulse"); + + pr_err("%s: heartbeat pulse did not flush idle tasks\n", + engine->name); + i915_active_print(&p->active, &m); + + err = -EINVAL; + goto out; + } + +out: + pulse_put(p); + return err; +} + +static int live_idle_flush(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err = 0; + + /* Check that we can flush the idle barriers */ + + for_each_engine(engine, gt, id) { + intel_engine_pm_get(engine); + err = __live_idle_pulse(engine, intel_engine_flush_barriers); + intel_engine_pm_put(engine); + if (err) + break; + } + + return err; +} + +static int live_idle_pulse(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err = 0; + + /* Check that heartbeat pulses flush the idle barriers */ + + for_each_engine(engine, gt, id) { + intel_engine_pm_get(engine); + err = __live_idle_pulse(engine, intel_engine_pulse); + intel_engine_pm_put(engine); + if (err && err != -ENODEV) + break; + + err = 0; + } + + return err; +} + +static int cmp_u32(const void *_a, const void *_b) +{ + const u32 *a = _a, *b = _b; + + return *a - *b; +} + +static int __live_heartbeat_fast(struct intel_engine_cs *engine) +{ + struct intel_context *ce; + struct i915_request *rq; + ktime_t t0, t1; + u32 times[5]; + int err; + int i; + + ce = intel_context_create(engine->kernel_context->gem_context, + engine); + if (IS_ERR(ce)) + return PTR_ERR(ce); + + intel_engine_pm_get(engine); + + err = intel_engine_set_heartbeat(engine, 1); + if (err) + goto err_pm; + + for (i = 0; i < ARRAY_SIZE(times); i++) { + /* Manufacture a tick */ + do { + while (READ_ONCE(engine->heartbeat.systole)) + flush_delayed_work(&engine->heartbeat.work); + + engine->serial++; /* quick, pretend we are not idle! */ + flush_delayed_work(&engine->heartbeat.work); + if (!delayed_work_pending(&engine->heartbeat.work)) { + pr_err("%s: heartbeat did not start\n", + engine->name); + err = -EINVAL; + goto err_pm; + } + + rcu_read_lock(); + rq = READ_ONCE(engine->heartbeat.systole); + if (rq) + rq = i915_request_get_rcu(rq); + rcu_read_unlock(); + } while (!rq); + + t0 = ktime_get(); + while (rq == READ_ONCE(engine->heartbeat.systole)) + yield(); /* work is on the local cpu! */ + t1 = ktime_get(); + + i915_request_put(rq); + times[i] = ktime_us_delta(t1, t0); + } + + sort(times, ARRAY_SIZE(times), sizeof(times[0]), cmp_u32, NULL); + + pr_info("%s: Heartbeat delay: %uus [%u, %u]\n", + engine->name, + times[ARRAY_SIZE(times) / 2], + times[0], + times[ARRAY_SIZE(times) - 1]); + + /* Min work delay is 2 * 2 (worst), +1 for scheduling, +1 for slack */ + if (times[ARRAY_SIZE(times) / 2] > jiffies_to_usecs(6)) { + pr_err("%s: Heartbeat delay was %uus, expected less than %dus\n", + engine->name, + times[ARRAY_SIZE(times) / 2], + jiffies_to_usecs(6)); + err = -EINVAL; + } + + intel_engine_set_heartbeat(engine, CONFIG_DRM_I915_HEARTBEAT_INTERVAL); +err_pm: + intel_engine_pm_put(engine); + intel_context_put(ce); + return err; +} + +static int live_heartbeat_fast(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err = 0; + + /* Check that the heartbeat ticks at the desired rate. */ + if (!CONFIG_DRM_I915_HEARTBEAT_INTERVAL) + return 0; + + for_each_engine(engine, gt, id) { + err = __live_heartbeat_fast(engine); + if (err) + break; + } + + return err; +} + +static int __live_heartbeat_off(struct intel_engine_cs *engine) +{ + int err; + + intel_engine_pm_get(engine); + + engine->serial++; + flush_delayed_work(&engine->heartbeat.work); + if (!delayed_work_pending(&engine->heartbeat.work)) { + pr_err("%s: heartbeat not running\n", + engine->name); + err = -EINVAL; + goto err_pm; + } + + err = intel_engine_set_heartbeat(engine, 0); + if (err) + goto err_pm; + + engine->serial++; + flush_delayed_work(&engine->heartbeat.work); + if (delayed_work_pending(&engine->heartbeat.work)) { + pr_err("%s: heartbeat still running\n", + engine->name); + err = -EINVAL; + goto err_beat; + } + + if (READ_ONCE(engine->heartbeat.systole)) { + pr_err("%s: heartbeat still allocated\n", + engine->name); + err = -EINVAL; + goto err_beat; + } + +err_beat: + intel_engine_set_heartbeat(engine, CONFIG_DRM_I915_HEARTBEAT_INTERVAL); +err_pm: + intel_engine_pm_put(engine); + return err; +} + +static int live_heartbeat_off(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err = 0; + + /* Check that we can turn off heartbeat and not interrupt VIP */ + if (!CONFIG_DRM_I915_HEARTBEAT_INTERVAL) + return 0; + + for_each_engine(engine, gt, id) { + if (!intel_engine_has_preemption(engine)) + continue; + + err = __live_heartbeat_off(engine); + if (err) + break; + } + + return err; +} + +int intel_heartbeat_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(live_idle_flush), + SUBTEST(live_idle_pulse), + SUBTEST(live_heartbeat_fast), + SUBTEST(live_heartbeat_off), + }; + int saved_hangcheck; + int err; + + if (intel_gt_is_wedged(&i915->gt)) + return 0; + + saved_hangcheck = i915_modparams.enable_hangcheck; + i915_modparams.enable_hangcheck = INT_MAX; + + err = intel_gt_live_subtests(tests, &i915->gt); + + i915_modparams.enable_hangcheck = saved_hangcheck; + return err; +} diff --git a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c index 3a1419376912..cbf6b0735272 100644 --- a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c +++ b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c @@ -25,7 +25,7 @@ static int live_engine_pm(void *arg) } GEM_BUG_ON(intel_gt_pm_is_awake(gt)); - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { const typeof(*igt_atomic_phases) *p; for (p = igt_atomic_phases; p->name; p++) { @@ -51,11 +51,12 @@ static int live_engine_pm(void *arg) pr_err("intel_engine_pm_get_if_awake(%s) failed under %s\n", engine->name, p->name); else - intel_engine_pm_put(engine); - intel_engine_pm_put(engine); + intel_engine_pm_put_async(engine); + intel_engine_pm_put_async(engine); p->critical_section_end(); - /* engine wakeref is sync (instant) */ + intel_engine_pm_flush(engine); + if (intel_engine_pm_is_awake(engine)) { pr_err("%s is still awake after flushing pm\n", engine->name); diff --git a/drivers/gpu/drm/i915/gt/selftest_gt_pm.c b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c new file mode 100644 index 000000000000..d1752f15702a --- /dev/null +++ b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c @@ -0,0 +1,60 @@ + +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include "selftest_llc.h" + +static int live_gt_resume(void *arg) +{ + struct intel_gt *gt = arg; + IGT_TIMEOUT(end_time); + int err; + + /* Do several suspend/resume cycles to check we don't explode! */ + do { + intel_gt_suspend_prepare(gt); + intel_gt_suspend_late(gt); + + if (gt->rc6.enabled) { + pr_err("rc6 still enabled after suspend!\n"); + intel_gt_set_wedged_on_init(gt); + err = -EINVAL; + break; + } + + err = intel_gt_resume(gt); + if (err) + break; + + if (gt->rc6.supported && !gt->rc6.enabled) { + pr_err("rc6 not enabled upon resume!\n"); + intel_gt_set_wedged_on_init(gt); + err = -EINVAL; + break; + } + + err = st_llc_verify(>->llc); + if (err) { + pr_err("llc state not restored upon resume!\n"); + intel_gt_set_wedged_on_init(gt); + break; + } + } while (!__igt_timeout(end_time, NULL)); + + return err; +} + +int intel_gt_pm_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(live_gt_resume), + }; + + if (intel_gt_is_wedged(&i915->gt)) + return 0; + + return intel_gt_live_subtests(tests, &i915->gt); +} diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c index a0098fc35921..85e9ccf5c304 100644 --- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c +++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c @@ -131,7 +131,7 @@ static struct i915_request * hang_create_request(struct hang *h, struct intel_engine_cs *engine) { struct intel_gt *gt = h->gt; - struct i915_address_space *vm = h->ctx->vm ?: &engine->gt->ggtt->vm; + struct i915_address_space *vm = i915_gem_context_get_vm_rcu(h->ctx); struct drm_i915_gem_object *obj; struct i915_request *rq = NULL; struct i915_vma *hws, *vma; @@ -141,12 +141,15 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) int err; obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); - if (IS_ERR(obj)) + if (IS_ERR(obj)) { + i915_vm_put(vm); return ERR_CAST(obj); + } vaddr = i915_gem_object_pin_map(obj, i915_coherent_map_type(gt->i915)); if (IS_ERR(vaddr)) { i915_gem_object_put(obj); + i915_vm_put(vm); return ERR_CAST(vaddr); } @@ -157,16 +160,22 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) h->batch = vaddr; vma = i915_vma_instance(h->obj, vm, NULL); - if (IS_ERR(vma)) + if (IS_ERR(vma)) { + i915_vm_put(vm); return ERR_CAST(vma); + } hws = i915_vma_instance(h->hws, vm, NULL); - if (IS_ERR(hws)) + if (IS_ERR(hws)) { + i915_vm_put(vm); return ERR_CAST(hws); + } err = i915_vma_pin(vma, 0, 0, PIN_USER); - if (err) + if (err) { + i915_vm_put(vm); return ERR_PTR(err); + } err = i915_vma_pin(hws, 0, 0, PIN_USER); if (err) @@ -264,6 +273,7 @@ unpin_hws: i915_vma_unpin(hws); unpin_vma: i915_vma_unpin(vma); + i915_vm_put(vm); return err ? ERR_PTR(err) : rq; } @@ -285,7 +295,7 @@ static void hang_fini(struct hang *h) kernel_context_close(h->ctx); - igt_flush_test(h->gt->i915, I915_WAIT_LOCKED); + igt_flush_test(h->gt->i915); } static bool wait_until_running(struct hang *h, struct i915_request *rq) @@ -309,12 +319,11 @@ static int igt_hang_sanitycheck(void *arg) /* Basic check that we can execute our hanging batch */ - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); if (err) - goto unlock; + return err; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { struct intel_wedge_me w; long timeout; @@ -355,8 +364,6 @@ static int igt_hang_sanitycheck(void *arg) fini: hang_fini(&h); -unlock: - mutex_unlock(>->i915->drm.struct_mutex); return err; } @@ -383,9 +390,7 @@ static int igt_reset_nop(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(>->i915->drm.struct_mutex); ctx = live_context(gt->i915, file); - mutex_unlock(>->i915->drm.struct_mutex); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto out; @@ -395,9 +400,7 @@ static int igt_reset_nop(void *arg) reset_count = i915_reset_count(global); count = 0; do { - mutex_lock(>->i915->drm.struct_mutex); - - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { int i; for (i = 0; i < 16; i++) { @@ -417,7 +420,6 @@ static int igt_reset_nop(void *arg) intel_gt_reset(gt, ALL_ENGINES, NULL); igt_global_reset_unlock(gt); - mutex_unlock(>->i915->drm.struct_mutex); if (intel_gt_is_wedged(gt)) { err = -EIO; break; @@ -429,16 +431,13 @@ static int igt_reset_nop(void *arg) break; } - err = igt_flush_test(gt->i915, 0); + err = igt_flush_test(gt->i915); if (err) break; } while (time_before(jiffies, end_time)); pr_info("%s: %d resets\n", __func__, count); - mutex_lock(>->i915->drm.struct_mutex); - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); - mutex_unlock(>->i915->drm.struct_mutex); - + err = igt_flush_test(gt->i915); out: mock_file_free(gt->i915, file); if (intel_gt_is_wedged(gt)) @@ -458,23 +457,21 @@ static int igt_reset_nop_engine(void *arg) /* Check that we can engine-reset during non-user portions */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; file = mock_file(gt->i915); if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(>->i915->drm.struct_mutex); ctx = live_context(gt->i915, file); - mutex_unlock(>->i915->drm.struct_mutex); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto out; } i915_gem_context_clear_bannable(ctx); - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { unsigned int reset_count, reset_engine_count; unsigned int count; IGT_TIMEOUT(end_time); @@ -494,7 +491,6 @@ static int igt_reset_nop_engine(void *arg) break; } - mutex_lock(>->i915->drm.struct_mutex); for (i = 0; i < 16; i++) { struct i915_request *rq; @@ -507,7 +503,6 @@ static int igt_reset_nop_engine(void *arg) i915_request_add(rq); } err = intel_engine_reset(engine, NULL); - mutex_unlock(>->i915->drm.struct_mutex); if (err) { pr_err("i915_reset_engine failed\n"); break; @@ -533,15 +528,12 @@ static int igt_reset_nop_engine(void *arg) if (err) break; - err = igt_flush_test(gt->i915, 0); + err = igt_flush_test(gt->i915); if (err) break; } - mutex_lock(>->i915->drm.struct_mutex); - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); - mutex_unlock(>->i915->drm.struct_mutex); - + err = igt_flush_test(gt->i915); out: mock_file_free(gt->i915, file); if (intel_gt_is_wedged(gt)) @@ -559,18 +551,16 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active) /* Check that we can issue an engine reset on an idle engine (no-op) */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (active) { - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); - mutex_unlock(>->i915->drm.struct_mutex); if (err) return err; } - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { unsigned int reset_count, reset_engine_count; IGT_TIMEOUT(end_time); @@ -593,17 +583,14 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active) if (active) { struct i915_request *rq; - mutex_lock(>->i915->drm.struct_mutex); rq = hang_create_request(&h, engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); - mutex_unlock(>->i915->drm.struct_mutex); break; } i915_request_get(rq); i915_request_add(rq); - mutex_unlock(>->i915->drm.struct_mutex); if (!wait_until_running(&h, rq)) { struct drm_printer p = drm_info_printer(gt->i915->drm.dev); @@ -647,7 +634,7 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active) if (err) break; - err = igt_flush_test(gt->i915, 0); + err = igt_flush_test(gt->i915); if (err) break; } @@ -655,11 +642,8 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active) if (intel_gt_is_wedged(gt)) err = -EIO; - if (active) { - mutex_lock(>->i915->drm.struct_mutex); + if (active) hang_fini(&h); - mutex_unlock(>->i915->drm.struct_mutex); - } return err; } @@ -725,9 +709,7 @@ static int active_engine(void *data) return PTR_ERR(file); for (count = 0; count < ARRAY_SIZE(ctx); count++) { - mutex_lock(&engine->i915->drm.struct_mutex); ctx[count] = live_context(engine->i915, file); - mutex_unlock(&engine->i915->drm.struct_mutex); if (IS_ERR(ctx[count])) { err = PTR_ERR(ctx[count]); while (--count) @@ -741,10 +723,8 @@ static int active_engine(void *data) struct i915_request *old = rq[idx]; struct i915_request *new; - mutex_lock(&engine->i915->drm.struct_mutex); new = igt_request_alloc(ctx[idx], engine); if (IS_ERR(new)) { - mutex_unlock(&engine->i915->drm.struct_mutex); err = PTR_ERR(new); break; } @@ -755,7 +735,6 @@ static int active_engine(void *data) rq[idx] = i915_request_get(new); i915_request_add(new); - mutex_unlock(&engine->i915->drm.struct_mutex); err = active_request_put(old); if (err) @@ -791,13 +770,11 @@ static int __igt_reset_engines(struct intel_gt *gt, * with any other engine. */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (flags & TEST_ACTIVE) { - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); - mutex_unlock(>->i915->drm.struct_mutex); if (err) return err; @@ -805,7 +782,7 @@ static int __igt_reset_engines(struct intel_gt *gt, h.ctx->sched.priority = 1024; } - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { struct active_engine threads[I915_NUM_ENGINES] = {}; unsigned long device = i915_reset_count(global); unsigned long count = 0, reported; @@ -823,7 +800,7 @@ static int __igt_reset_engines(struct intel_gt *gt, } memset(threads, 0, sizeof(threads)); - for_each_engine(other, gt->i915, tmp) { + for_each_engine(other, gt, tmp) { struct task_struct *tsk; threads[tmp].resets = @@ -849,23 +826,22 @@ static int __igt_reset_engines(struct intel_gt *gt, get_task_struct(tsk); } + yield(); /* start all threads before we begin */ + intel_engine_pm_get(engine); set_bit(I915_RESET_ENGINE + id, >->reset.flags); do { struct i915_request *rq = NULL; if (flags & TEST_ACTIVE) { - mutex_lock(>->i915->drm.struct_mutex); rq = hang_create_request(&h, engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); - mutex_unlock(>->i915->drm.struct_mutex); break; } i915_request_get(rq); i915_request_add(rq); - mutex_unlock(>->i915->drm.struct_mutex); if (!wait_until_running(&h, rq)) { struct drm_printer p = drm_info_printer(gt->i915->drm.dev); @@ -940,7 +916,7 @@ static int __igt_reset_engines(struct intel_gt *gt, } unwind: - for_each_engine(other, gt->i915, tmp) { + for_each_engine(other, gt, tmp) { int ret; if (!threads[tmp].task) @@ -977,9 +953,7 @@ unwind: if (err) break; - mutex_lock(>->i915->drm.struct_mutex); - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); - mutex_unlock(>->i915->drm.struct_mutex); + err = igt_flush_test(gt->i915); if (err) break; } @@ -987,11 +961,8 @@ unwind: if (intel_gt_is_wedged(gt)) err = -EIO; - if (flags & TEST_ACTIVE) { - mutex_lock(>->i915->drm.struct_mutex); + if (flags & TEST_ACTIVE) hang_fini(&h); - mutex_unlock(>->i915->drm.struct_mutex); - } return err; } @@ -1047,7 +1018,7 @@ static int igt_reset_wait(void *arg) { struct intel_gt *gt = arg; struct i915_gpu_error *global = >->i915->gpu_error; - struct intel_engine_cs *engine = gt->i915->engine[RCS0]; + struct intel_engine_cs *engine = gt->engine[RCS0]; struct i915_request *rq; unsigned int reset_count; struct hang h; @@ -1061,7 +1032,6 @@ static int igt_reset_wait(void *arg) igt_global_reset_lock(gt); - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); if (err) goto unlock; @@ -1109,7 +1079,6 @@ out_rq: fini: hang_fini(&h); unlock: - mutex_unlock(>->i915->drm.struct_mutex); igt_global_reset_unlock(gt); if (intel_gt_is_wedged(gt)) @@ -1127,15 +1096,14 @@ static int evict_vma(void *data) { struct evict_vma *arg = data; struct i915_address_space *vm = arg->vma->vm; - struct drm_i915_private *i915 = vm->i915; struct drm_mm_node evict = arg->vma->node; int err; complete(&arg->completion); - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(&vm->mutex); err = i915_gem_evict_for_node(vm, &evict, 0); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(&vm->mutex); return err; } @@ -1143,39 +1111,33 @@ static int evict_vma(void *data) static int evict_fence(void *data) { struct evict_vma *arg = data; - struct drm_i915_private *i915 = arg->vma->vm->i915; int err; complete(&arg->completion); - mutex_lock(&i915->drm.struct_mutex); - /* Mark the fence register as dirty to force the mmio update. */ err = i915_gem_object_set_tiling(arg->vma->obj, I915_TILING_Y, 512); if (err) { pr_err("Invalid Y-tiling settings; err:%d\n", err); - goto out_unlock; + return err; } err = i915_vma_pin(arg->vma, 0, 0, PIN_GLOBAL | PIN_MAPPABLE); if (err) { pr_err("Unable to pin vma for Y-tiled fence; err:%d\n", err); - goto out_unlock; + return err; } err = i915_vma_pin_fence(arg->vma); i915_vma_unpin(arg->vma); if (err) { pr_err("Unable to pin Y-tiled fence; err:%d\n", err); - goto out_unlock; + return err; } i915_vma_unpin_fence(arg->vma); -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - - return err; + return 0; } static int __igt_reset_evict_vma(struct intel_gt *gt, @@ -1183,23 +1145,26 @@ static int __igt_reset_evict_vma(struct intel_gt *gt, int (*fn)(void *), unsigned int flags) { - struct intel_engine_cs *engine = gt->i915->engine[RCS0]; + struct intel_engine_cs *engine = gt->engine[RCS0]; struct drm_i915_gem_object *obj; struct task_struct *tsk = NULL; struct i915_request *rq; struct evict_vma arg; struct hang h; + unsigned int pin_flags; int err; + if (!gt->ggtt->num_fences && flags & EXEC_OBJECT_NEEDS_FENCE) + return 0; + if (!engine || !intel_engine_can_store_dword(engine)) return 0; /* Check that we can recover an unbind stuck on a hanging request */ - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); if (err) - goto unlock; + return err; obj = i915_gem_object_create_internal(gt->i915, SZ_1M); if (IS_ERR(obj)) { @@ -1227,10 +1192,12 @@ static int __igt_reset_evict_vma(struct intel_gt *gt, goto out_obj; } - err = i915_vma_pin(arg.vma, 0, 0, - i915_vma_is_ggtt(arg.vma) ? - PIN_GLOBAL | PIN_MAPPABLE : - PIN_USER); + pin_flags = i915_vma_is_ggtt(arg.vma) ? PIN_GLOBAL : PIN_USER; + + if (flags & EXEC_OBJECT_NEEDS_FENCE) + pin_flags |= PIN_MAPPABLE; + + err = i915_vma_pin(arg.vma, 0, 0, pin_flags); if (err) { i915_request_add(rq); goto out_obj; @@ -1262,8 +1229,6 @@ static int __igt_reset_evict_vma(struct intel_gt *gt, if (err) goto out_rq; - mutex_unlock(>->i915->drm.struct_mutex); - if (!wait_until_running(&h, rq)) { struct drm_printer p = drm_info_printer(gt->i915->drm.dev); @@ -1312,16 +1277,12 @@ out_reset: put_task_struct(tsk); } - mutex_lock(>->i915->drm.struct_mutex); out_rq: i915_request_put(rq); out_obj: i915_gem_object_put(obj); fini: hang_fini(&h); -unlock: - mutex_unlock(>->i915->drm.struct_mutex); - if (intel_gt_is_wedged(gt)) return -EIO; @@ -1340,6 +1301,7 @@ static int igt_reset_evict_ppgtt(void *arg) { struct intel_gt *gt = arg; struct i915_gem_context *ctx; + struct i915_address_space *vm; struct drm_file *file; int err; @@ -1347,18 +1309,20 @@ static int igt_reset_evict_ppgtt(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(>->i915->drm.struct_mutex); ctx = live_context(gt->i915, file); - mutex_unlock(>->i915->drm.struct_mutex); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto out; } err = 0; - if (ctx->vm) /* aliasing == global gtt locking, covered above */ - err = __igt_reset_evict_vma(gt, ctx->vm, + vm = i915_gem_context_get_vm_rcu(ctx); + if (!i915_is_ggtt(vm)) { + /* aliasing == global gtt locking, covered above */ + err = __igt_reset_evict_vma(gt, vm, evict_vma, EXEC_OBJECT_WRITE); + } + i915_vm_put(vm); out: mock_file_free(gt->i915, file); @@ -1379,7 +1343,7 @@ static int wait_for_others(struct intel_gt *gt, struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { if (engine == exclude) continue; @@ -1403,12 +1367,11 @@ static int igt_reset_queue(void *arg) igt_global_reset_lock(gt); - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); if (err) goto unlock; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { struct i915_request *prev; IGT_TIMEOUT(end_time); unsigned int count; @@ -1518,7 +1481,7 @@ static int igt_reset_queue(void *arg) i915_request_put(prev); - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + err = igt_flush_test(gt->i915); if (err) break; } @@ -1526,7 +1489,6 @@ static int igt_reset_queue(void *arg) fini: hang_fini(&h); unlock: - mutex_unlock(>->i915->drm.struct_mutex); igt_global_reset_unlock(gt); if (intel_gt_is_wedged(gt)) @@ -1539,7 +1501,7 @@ static int igt_handle_error(void *arg) { struct intel_gt *gt = arg; struct i915_gpu_error *global = >->i915->gpu_error; - struct intel_engine_cs *engine = gt->i915->engine[RCS0]; + struct intel_engine_cs *engine = gt->engine[RCS0]; struct hang h; struct i915_request *rq; struct i915_gpu_state *error; @@ -1547,17 +1509,15 @@ static int igt_handle_error(void *arg) /* Check that we can issue a global GPU and engine reset */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (!engine || !intel_engine_can_store_dword(engine)) return 0; - mutex_lock(>->i915->drm.struct_mutex); - err = hang_init(&h, gt); if (err) - goto err_unlock; + return err; rq = hang_create_request(&h, engine); if (IS_ERR(rq)) { @@ -1581,8 +1541,6 @@ static int igt_handle_error(void *arg) goto err_request; } - mutex_unlock(>->i915->drm.struct_mutex); - /* Temporarily disable error capture */ error = xchg(&global->first_error, (void *)-1); @@ -1590,8 +1548,6 @@ static int igt_handle_error(void *arg) xchg(&global->first_error, error); - mutex_lock(>->i915->drm.struct_mutex); - if (rq->fence.error != -EIO) { pr_err("Guilty request not identified!\n"); err = -EINVAL; @@ -1602,8 +1558,6 @@ err_request: i915_request_put(rq); err_fini: hang_fini(&h); -err_unlock: - mutex_unlock(>->i915->drm.struct_mutex); return err; } @@ -1617,7 +1571,7 @@ static int __igt_atomic_reset_engine(struct intel_engine_cs *engine, GEM_TRACE("i915_reset_engine(%s:%s) under %s\n", engine->name, mode, p->name); - tasklet_disable_nosync(t); + tasklet_disable(t); p->critical_section_begin(); err = intel_engine_reset(engine, NULL); @@ -1689,14 +1643,13 @@ static int igt_reset_engines_atomic(void *arg) /* Check that the engines resets are usable from atomic context */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (USES_GUC_SUBMISSION(gt->i915)) return 0; igt_global_reset_lock(gt); - mutex_lock(>->i915->drm.struct_mutex); /* Flush any requests before we get started and check basics */ if (!igt_force_reset(gt)) @@ -1706,7 +1659,7 @@ static int igt_reset_engines_atomic(void *arg) struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { err = igt_atomic_reset_engine(engine, p); if (err) goto out; @@ -1716,9 +1669,7 @@ static int igt_reset_engines_atomic(void *arg) out: /* As we poke around the guts, do a full reset before continuing. */ igt_force_reset(gt); - unlock: - mutex_unlock(>->i915->drm.struct_mutex); igt_global_reset_unlock(gt); return err; @@ -1743,27 +1694,19 @@ int intel_hangcheck_live_selftests(struct drm_i915_private *i915) }; struct intel_gt *gt = &i915->gt; intel_wakeref_t wakeref; - bool saved_hangcheck; int err; - if (!intel_has_gpu_reset(gt->i915)) + if (!intel_has_gpu_reset(gt)) return 0; if (intel_gt_is_wedged(gt)) return -EIO; /* we're long past hope of a successful reset */ - wakeref = intel_runtime_pm_get(>->i915->runtime_pm); - saved_hangcheck = fetch_and_zero(&i915_modparams.enable_hangcheck); - drain_delayed_work(>->hangcheck.work); /* flush param */ + wakeref = intel_runtime_pm_get(gt->uncore->rpm); err = intel_gt_live_subtests(tests, gt); - mutex_lock(>->i915->drm.struct_mutex); - igt_flush_test(gt->i915, I915_WAIT_LOCKED); - mutex_unlock(>->i915->drm.struct_mutex); - - i915_modparams.enable_hangcheck = saved_hangcheck; - intel_runtime_pm_put(>->i915->runtime_pm, wakeref); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); return err; } diff --git a/drivers/gpu/drm/i915/gt/selftest_llc.c b/drivers/gpu/drm/i915/gt/selftest_llc.c new file mode 100644 index 000000000000..fd3770e48ac7 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/selftest_llc.c @@ -0,0 +1,80 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include "intel_pm.h" /* intel_gpu_freq() */ +#include "selftest_llc.h" +#include "intel_rps.h" + +static int gen6_verify_ring_freq(struct intel_llc *llc) +{ + struct drm_i915_private *i915 = llc_to_gt(llc)->i915; + struct ia_constants consts; + intel_wakeref_t wakeref; + unsigned int gpu_freq; + int err = 0; + + wakeref = intel_runtime_pm_get(llc_to_gt(llc)->uncore->rpm); + + if (!get_ia_constants(llc, &consts)) { + err = -ENODEV; + goto out_rpm; + } + + for (gpu_freq = consts.min_gpu_freq; + gpu_freq <= consts.max_gpu_freq; + gpu_freq++) { + struct intel_rps *rps = &llc_to_gt(llc)->rps; + + unsigned int ia_freq, ring_freq, found; + u32 val; + + calc_ia_freq(llc, gpu_freq, &consts, &ia_freq, &ring_freq); + + val = gpu_freq; + if (sandybridge_pcode_read(i915, + GEN6_PCODE_READ_MIN_FREQ_TABLE, + &val, NULL)) { + pr_err("Failed to read freq table[%d], range [%d, %d]\n", + gpu_freq, consts.min_gpu_freq, consts.max_gpu_freq); + err = -ENXIO; + break; + } + + found = (val >> 0) & 0xff; + if (found != ia_freq) { + pr_err("Min freq table(%d/[%d, %d]):%dMHz did not match expected CPU freq, found %d, expected %d\n", + gpu_freq, consts.min_gpu_freq, consts.max_gpu_freq, + intel_gpu_freq(rps, gpu_freq * (INTEL_GEN(i915) >= 9 ? GEN9_FREQ_SCALER : 1)), + found, ia_freq); + err = -EINVAL; + break; + } + + found = (val >> 8) & 0xff; + if (found != ring_freq) { + pr_err("Min freq table(%d/[%d, %d]):%dMHz did not match expected ring freq, found %d, expected %d\n", + gpu_freq, consts.min_gpu_freq, consts.max_gpu_freq, + intel_gpu_freq(rps, gpu_freq * (INTEL_GEN(i915) >= 9 ? GEN9_FREQ_SCALER : 1)), + found, ring_freq); + err = -EINVAL; + break; + } + } + +out_rpm: + intel_runtime_pm_put(llc_to_gt(llc)->uncore->rpm, wakeref); + return err; +} + +int st_llc_verify(struct intel_llc *llc) +{ + int err = 0; + + if (HAS_LLC(llc_to_gt(llc)->i915)) + err = gen6_verify_ring_freq(llc); + + return err; +} diff --git a/drivers/gpu/drm/i915/gt/selftest_llc.h b/drivers/gpu/drm/i915/gt/selftest_llc.h new file mode 100644 index 000000000000..873f896e72f2 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/selftest_llc.h @@ -0,0 +1,14 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef SELFTEST_LLC_H +#define SELFTEST_LLC_H + +struct intel_llc; + +int st_llc_verify(struct intel_llc *llc); + +#endif /* SELFTEST_LLC_H */ diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c index d791158988d6..eb71ac2f992c 100644 --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c @@ -7,6 +7,7 @@ #include <linux/prime_numbers.h> #include "gem/i915_gem_pm.h" +#include "gt/intel_engine_heartbeat.h" #include "gt/intel_reset.h" #include "i915_selftest.h" @@ -19,26 +20,52 @@ #include "gem/selftests/igt_gem_utils.h" #include "gem/selftests/mock_context.h" +#define CS_GPR(engine, n) ((engine)->mmio_base + 0x600 + (n) * 4) +#define NUM_GPR_DW (16 * 2) /* each GPR is 2 dwords */ + +static struct i915_vma *create_scratch(struct intel_gt *gt) +{ + struct drm_i915_gem_object *obj; + struct i915_vma *vma; + int err; + + obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); + if (IS_ERR(obj)) + return ERR_CAST(obj); + + i915_gem_object_set_cache_coherency(obj, I915_CACHING_CACHED); + + vma = i915_vma_instance(obj, >->ggtt->vm, NULL); + if (IS_ERR(vma)) { + i915_gem_object_put(obj); + return vma; + } + + err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL); + if (err) { + i915_gem_object_put(obj); + return ERR_PTR(err); + } + + return vma; +} + static int live_sanitycheck(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_gem_engines_iter it; struct i915_gem_context *ctx; struct intel_context *ce; struct igt_spinner spin; - intel_wakeref_t wakeref; int err = -ENOMEM; - if (!HAS_LOGICAL_RING_CONTEXTS(i915)) + if (!HAS_LOGICAL_RING_CONTEXTS(gt->i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - if (igt_spinner_init(&spin, &i915->gt)) - goto err_unlock; + if (igt_spinner_init(&spin, gt)) + return -ENOMEM; - ctx = kernel_context(i915); + ctx = kernel_context(gt->i915); if (!ctx) goto err_spin; @@ -55,13 +82,13 @@ static int live_sanitycheck(void *arg) if (!igt_wait_for_spinner(&spin, rq)) { GEM_TRACE("spinner failed to start\n"); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_ctx; } igt_spinner_end(&spin); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) { + if (igt_flush_test(gt->i915)) { err = -EIO; goto err_ctx; } @@ -73,12 +100,175 @@ err_ctx: kernel_context_close(ctx); err_spin: igt_spinner_fini(&spin); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } +static int live_unlite_restore(struct intel_gt *gt, int prio) +{ + struct intel_engine_cs *engine; + struct i915_gem_context *ctx; + enum intel_engine_id id; + struct igt_spinner spin; + int err = -ENOMEM; + + /* + * Check that we can correctly context switch between 2 instances + * on the same engine from the same parent context. + */ + + if (igt_spinner_init(&spin, gt)) + return err; + + ctx = kernel_context(gt->i915); + if (!ctx) + goto err_spin; + + err = 0; + for_each_engine(engine, gt, id) { + struct intel_context *ce[2] = {}; + struct i915_request *rq[2]; + struct igt_live_test t; + int n; + + if (prio && !intel_engine_has_preemption(engine)) + continue; + + if (!intel_engine_can_store_dword(engine)) + continue; + + if (igt_live_test_begin(&t, gt->i915, __func__, engine->name)) { + err = -EIO; + break; + } + + for (n = 0; n < ARRAY_SIZE(ce); n++) { + struct intel_context *tmp; + + tmp = intel_context_create(ctx, engine); + if (IS_ERR(tmp)) { + err = PTR_ERR(tmp); + goto err_ce; + } + + err = intel_context_pin(tmp); + if (err) { + intel_context_put(tmp); + goto err_ce; + } + + /* + * Setup the pair of contexts such that if we + * lite-restore using the RING_TAIL from ce[1] it + * will execute garbage from ce[0]->ring. + */ + memset(tmp->ring->vaddr, + POISON_INUSE, /* IPEHR: 0x5a5a5a5a [hung!] */ + tmp->ring->vma->size); + + ce[n] = tmp; + } + GEM_BUG_ON(!ce[1]->ring->size); + intel_ring_reset(ce[1]->ring, ce[1]->ring->size / 2); + __execlists_update_reg_state(ce[1], engine); + + rq[0] = igt_spinner_create_request(&spin, ce[0], MI_ARB_CHECK); + if (IS_ERR(rq[0])) { + err = PTR_ERR(rq[0]); + goto err_ce; + } + + i915_request_get(rq[0]); + i915_request_add(rq[0]); + GEM_BUG_ON(rq[0]->postfix > ce[1]->ring->emit); + + if (!igt_wait_for_spinner(&spin, rq[0])) { + i915_request_put(rq[0]); + goto err_ce; + } + + rq[1] = i915_request_create(ce[1]); + if (IS_ERR(rq[1])) { + err = PTR_ERR(rq[1]); + i915_request_put(rq[0]); + goto err_ce; + } + + if (!prio) { + /* + * Ensure we do the switch to ce[1] on completion. + * + * rq[0] is already submitted, so this should reduce + * to a no-op (a wait on a request on the same engine + * uses the submit fence, not the completion fence), + * but it will install a dependency on rq[1] for rq[0] + * that will prevent the pair being reordered by + * timeslicing. + */ + i915_request_await_dma_fence(rq[1], &rq[0]->fence); + } + + i915_request_get(rq[1]); + i915_request_add(rq[1]); + GEM_BUG_ON(rq[1]->postfix <= rq[0]->postfix); + i915_request_put(rq[0]); + + if (prio) { + struct i915_sched_attr attr = { + .priority = prio, + }; + + /* Alternatively preempt the spinner with ce[1] */ + engine->schedule(rq[1], &attr); + } + + /* And switch back to ce[0] for good measure */ + rq[0] = i915_request_create(ce[0]); + if (IS_ERR(rq[0])) { + err = PTR_ERR(rq[0]); + i915_request_put(rq[1]); + goto err_ce; + } + + i915_request_await_dma_fence(rq[0], &rq[1]->fence); + i915_request_get(rq[0]); + i915_request_add(rq[0]); + GEM_BUG_ON(rq[0]->postfix > rq[1]->postfix); + i915_request_put(rq[1]); + i915_request_put(rq[0]); + +err_ce: + tasklet_kill(&engine->execlists.tasklet); /* flush submission */ + igt_spinner_end(&spin); + for (n = 0; n < ARRAY_SIZE(ce); n++) { + if (IS_ERR_OR_NULL(ce[n])) + break; + + intel_context_unpin(ce[n]); + intel_context_put(ce[n]); + } + + if (igt_live_test_end(&t)) + err = -EIO; + if (err) + break; + } + + kernel_context_close(ctx); +err_spin: + igt_spinner_fini(&spin); + return err; +} + +static int live_unlite_switch(void *arg) +{ + return live_unlite_restore(arg, 0); +} + +static int live_unlite_preempt(void *arg) +{ + return live_unlite_restore(arg, I915_USER_PRIORITY(I915_PRIORITY_MAX)); +} + static int emit_semaphore_chain(struct i915_request *rq, struct i915_vma *vma, int idx) { @@ -131,7 +321,13 @@ semaphore_queue(struct intel_engine_cs *engine, struct i915_vma *vma, int idx) if (IS_ERR(rq)) goto out_ctx; - err = emit_semaphore_chain(rq, vma, idx); + err = 0; + if (rq->engine->emit_init_breadcrumb) + err = rq->engine->emit_init_breadcrumb(rq); + if (err == 0) + err = emit_semaphore_chain(rq, vma, idx); + if (err == 0) + i915_request_get(rq); i915_request_add(rq); if (err) rq = ERR_PTR(err); @@ -144,10 +340,10 @@ out_ctx: static int release_queue(struct intel_engine_cs *engine, struct i915_vma *vma, - int idx) + int idx, int prio) { struct i915_sched_attr attr = { - .priority = I915_USER_PRIORITY(I915_PRIORITY_MAX), + .priority = prio, }; struct i915_request *rq; u32 *cs; @@ -168,9 +364,15 @@ release_queue(struct intel_engine_cs *engine, *cs++ = 1; intel_ring_advance(rq, cs); + + i915_request_get(rq); i915_request_add(rq); + local_bh_disable(); engine->schedule(rq, &attr); + local_bh_enable(); /* kick tasklet */ + + i915_request_put(rq); return 0; } @@ -189,8 +391,7 @@ slice_semaphore_queue(struct intel_engine_cs *outer, if (IS_ERR(head)) return PTR_ERR(head); - i915_request_get(head); - for_each_engine(engine, outer->i915, id) { + for_each_engine(engine, outer->gt, id) { for (i = 0; i < count; i++) { struct i915_request *rq; @@ -199,15 +400,16 @@ slice_semaphore_queue(struct intel_engine_cs *outer, err = PTR_ERR(rq); goto out; } + + i915_request_put(rq); } } - err = release_queue(outer, vma, n); + err = release_queue(outer, vma, n, INT_MAX); if (err) goto out; - if (i915_request_wait(head, - I915_WAIT_LOCKED, + if (i915_request_wait(head, 0, 2 * RUNTIME_INFO(outer->i915)->num_engines * (count + 2) * (count + 3)) < 0) { pr_err("Failed to slice along semaphore chain of length (%d, %d)!\n", count, n); @@ -223,9 +425,8 @@ out: static int live_timeslice_preempt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct drm_i915_gem_object *obj; - intel_wakeref_t wakeref; struct i915_vma *vma; void *vaddr; int err = 0; @@ -239,17 +440,14 @@ static int live_timeslice_preempt(void *arg) * need to preempt the current task and replace it with another * ready task. */ + if (!IS_ACTIVE(CONFIG_DRM_I915_TIMESLICE_DURATION)) + return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - obj = i915_gem_object_create_internal(i915, PAGE_SIZE); - if (IS_ERR(obj)) { - err = PTR_ERR(obj); - goto err_unlock; - } + obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); + if (IS_ERR(obj)) + return PTR_ERR(obj); - vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL); + vma = i915_vma_instance(obj, >->ggtt->vm, NULL); if (IS_ERR(vma)) { err = PTR_ERR(vma); goto err_obj; @@ -269,7 +467,7 @@ static int live_timeslice_preempt(void *arg) struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { if (!intel_engine_has_preemption(engine)) continue; @@ -279,7 +477,7 @@ static int live_timeslice_preempt(void *arg) if (err) goto err_pin; - if (igt_flush_test(i915, I915_WAIT_LOCKED)) { + if (igt_flush_test(gt->i915)) { err = -EIO; goto err_pin; } @@ -292,22 +490,168 @@ err_map: i915_gem_object_unpin_map(obj); err_obj: i915_gem_object_put(obj); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); + return err; +} + +static struct i915_request *nop_request(struct intel_engine_cs *engine) +{ + struct i915_request *rq; + + rq = i915_request_create(engine->kernel_context); + if (IS_ERR(rq)) + return rq; + + i915_request_get(rq); + i915_request_add(rq); + return rq; +} + +static void wait_for_submit(struct intel_engine_cs *engine, + struct i915_request *rq) +{ + do { + cond_resched(); + intel_engine_flush_submission(engine); + } while (!i915_request_is_active(rq)); +} + +static long timeslice_threshold(const struct intel_engine_cs *engine) +{ + return 2 * msecs_to_jiffies_timeout(timeslice(engine)) + 1; +} + +static int live_timeslice_queue(void *arg) +{ + struct intel_gt *gt = arg; + struct drm_i915_gem_object *obj; + struct intel_engine_cs *engine; + enum intel_engine_id id; + struct i915_vma *vma; + void *vaddr; + int err = 0; + + /* + * Make sure that even if ELSP[0] and ELSP[1] are filled with + * timeslicing between them disabled, we *do* enable timeslicing + * if the queue demands it. (Normally, we do not submit if + * ELSP[1] is already occupied, so must rely on timeslicing to + * eject ELSP[0] in favour of the queue.) + */ + if (!IS_ACTIVE(CONFIG_DRM_I915_TIMESLICE_DURATION)) + return 0; + + obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + vma = i915_vma_instance(obj, >->ggtt->vm, NULL); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto err_obj; + } + + vaddr = i915_gem_object_pin_map(obj, I915_MAP_WC); + if (IS_ERR(vaddr)) { + err = PTR_ERR(vaddr); + goto err_obj; + } + + err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL); + if (err) + goto err_map; + + for_each_engine(engine, gt, id) { + struct i915_sched_attr attr = { + .priority = I915_USER_PRIORITY(I915_PRIORITY_MAX), + }; + struct i915_request *rq, *nop; + + if (!intel_engine_has_preemption(engine)) + continue; + + memset(vaddr, 0, PAGE_SIZE); + + /* ELSP[0]: semaphore wait */ + rq = semaphore_queue(engine, vma, 0); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto err_pin; + } + engine->schedule(rq, &attr); + wait_for_submit(engine, rq); + + /* ELSP[1]: nop request */ + nop = nop_request(engine); + if (IS_ERR(nop)) { + err = PTR_ERR(nop); + i915_request_put(rq); + goto err_pin; + } + wait_for_submit(engine, nop); + i915_request_put(nop); + + GEM_BUG_ON(i915_request_completed(rq)); + GEM_BUG_ON(execlists_active(&engine->execlists) != rq); + + /* Queue: semaphore signal, matching priority as semaphore */ + err = release_queue(engine, vma, 1, effective_prio(rq)); + if (err) { + i915_request_put(rq); + goto err_pin; + } + + intel_engine_flush_submission(engine); + if (!READ_ONCE(engine->execlists.timer.expires) && + !i915_request_completed(rq)) { + struct drm_printer p = + drm_info_printer(gt->i915->drm.dev); + + GEM_TRACE_ERR("%s: Failed to enable timeslicing!\n", + engine->name); + intel_engine_dump(engine, &p, + "%s\n", engine->name); + GEM_TRACE_DUMP(); + + memset(vaddr, 0xff, PAGE_SIZE); + err = -EINVAL; + } + + /* Timeslice every jiffy, so within 2 we should signal */ + if (i915_request_wait(rq, 0, timeslice_threshold(engine)) < 0) { + struct drm_printer p = + drm_info_printer(gt->i915->drm.dev); + + pr_err("%s: Failed to timeslice into queue\n", + engine->name); + intel_engine_dump(engine, &p, + "%s\n", engine->name); + + memset(vaddr, 0xff, PAGE_SIZE); + err = -EIO; + } + i915_request_put(rq); + if (err) + break; + } + +err_pin: + i915_vma_unpin(vma); +err_map: + i915_gem_object_unpin_map(obj); +err_obj: + i915_gem_object_put(obj); return err; } static int live_busywait_preempt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_gem_context *ctx_hi, *ctx_lo; struct intel_engine_cs *engine; struct drm_i915_gem_object *obj; struct i915_vma *vma; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; u32 *map; @@ -316,22 +660,19 @@ static int live_busywait_preempt(void *arg) * preempt the busywaits used to synchronise between rings. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - ctx_hi = kernel_context(i915); + ctx_hi = kernel_context(gt->i915); if (!ctx_hi) - goto err_unlock; + return -ENOMEM; ctx_hi->sched.priority = I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY); - ctx_lo = kernel_context(i915); + ctx_lo = kernel_context(gt->i915); if (!ctx_lo) goto err_ctx_hi; ctx_lo->sched.priority = I915_USER_PRIORITY(I915_CONTEXT_MIN_USER_PRIORITY); - obj = i915_gem_object_create_internal(i915, PAGE_SIZE); + obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); if (IS_ERR(obj)) { err = PTR_ERR(obj); goto err_ctx_lo; @@ -343,7 +684,7 @@ static int live_busywait_preempt(void *arg) goto err_obj; } - vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL); + vma = i915_vma_instance(obj, >->ggtt->vm, NULL); if (IS_ERR(vma)) { err = PTR_ERR(vma); goto err_map; @@ -353,7 +694,7 @@ static int live_busywait_preempt(void *arg) if (err) goto err_map; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct i915_request *lo, *hi; struct igt_live_test t; u32 *cs; @@ -364,7 +705,7 @@ static int live_busywait_preempt(void *arg) if (!intel_engine_can_store_dword(engine)) continue; - if (igt_live_test_begin(&t, i915, __func__, engine->name)) { + if (igt_live_test_begin(&t, gt->i915, __func__, engine->name)) { err = -EIO; goto err_vma; } @@ -444,7 +785,7 @@ static int live_busywait_preempt(void *arg) i915_request_add(hi); if (i915_request_wait(lo, 0, HZ / 5) < 0) { - struct drm_printer p = drm_info_printer(i915->drm.dev); + struct drm_printer p = drm_info_printer(gt->i915->drm.dev); pr_err("%s: Failed to preempt semaphore busywait!\n", engine->name); @@ -452,7 +793,7 @@ static int live_busywait_preempt(void *arg) intel_engine_dump(engine, &p, "%s\n", engine->name); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_vma; } @@ -475,9 +816,6 @@ err_ctx_lo: kernel_context_close(ctx_lo); err_ctx_hi: kernel_context_close(ctx_hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -501,49 +839,45 @@ spinner_create_request(struct igt_spinner *spin, static int live_preempt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_gem_context *ctx_hi, *ctx_lo; struct igt_spinner spin_hi, spin_lo; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; - if (!HAS_LOGICAL_RING_PREEMPTION(i915)) + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) return 0; - if (!(i915->caps.scheduler & I915_SCHEDULER_CAP_PREEMPTION)) + if (!(gt->i915->caps.scheduler & I915_SCHEDULER_CAP_PREEMPTION)) pr_err("Logical preemption supported, but not exposed\n"); - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - if (igt_spinner_init(&spin_hi, &i915->gt)) - goto err_unlock; + if (igt_spinner_init(&spin_hi, gt)) + return -ENOMEM; - if (igt_spinner_init(&spin_lo, &i915->gt)) + if (igt_spinner_init(&spin_lo, gt)) goto err_spin_hi; - ctx_hi = kernel_context(i915); + ctx_hi = kernel_context(gt->i915); if (!ctx_hi) goto err_spin_lo; ctx_hi->sched.priority = I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY); - ctx_lo = kernel_context(i915); + ctx_lo = kernel_context(gt->i915); if (!ctx_lo) goto err_ctx_hi; ctx_lo->sched.priority = I915_USER_PRIORITY(I915_CONTEXT_MIN_USER_PRIORITY); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct igt_live_test t; struct i915_request *rq; if (!intel_engine_has_preemption(engine)) continue; - if (igt_live_test_begin(&t, i915, __func__, engine->name)) { + if (igt_live_test_begin(&t, gt->i915, __func__, engine->name)) { err = -EIO; goto err_ctx_lo; } @@ -559,7 +893,7 @@ static int live_preempt(void *arg) if (!igt_wait_for_spinner(&spin_lo, rq)) { GEM_TRACE("lo spinner failed to start\n"); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_ctx_lo; } @@ -576,7 +910,7 @@ static int live_preempt(void *arg) if (!igt_wait_for_spinner(&spin_hi, rq)) { GEM_TRACE("hi spinner failed to start\n"); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_ctx_lo; } @@ -599,54 +933,47 @@ err_spin_lo: igt_spinner_fini(&spin_lo); err_spin_hi: igt_spinner_fini(&spin_hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } static int live_late_preempt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_gem_context *ctx_hi, *ctx_lo; struct igt_spinner spin_hi, spin_lo; struct intel_engine_cs *engine; struct i915_sched_attr attr = {}; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; - if (!HAS_LOGICAL_RING_PREEMPTION(i915)) + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - if (igt_spinner_init(&spin_hi, &i915->gt)) - goto err_unlock; + if (igt_spinner_init(&spin_hi, gt)) + return -ENOMEM; - if (igt_spinner_init(&spin_lo, &i915->gt)) + if (igt_spinner_init(&spin_lo, gt)) goto err_spin_hi; - ctx_hi = kernel_context(i915); + ctx_hi = kernel_context(gt->i915); if (!ctx_hi) goto err_spin_lo; - ctx_lo = kernel_context(i915); + ctx_lo = kernel_context(gt->i915); if (!ctx_lo) goto err_ctx_hi; /* Make sure ctx_lo stays before ctx_hi until we trigger preemption. */ ctx_lo->sched.priority = I915_USER_PRIORITY(1); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct igt_live_test t; struct i915_request *rq; if (!intel_engine_has_preemption(engine)) continue; - if (igt_live_test_begin(&t, i915, __func__, engine->name)) { + if (igt_live_test_begin(&t, gt->i915, __func__, engine->name)) { err = -EIO; goto err_ctx_lo; } @@ -705,15 +1032,12 @@ err_spin_lo: igt_spinner_fini(&spin_lo); err_spin_hi: igt_spinner_fini(&spin_hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: igt_spinner_end(&spin_hi); igt_spinner_end(&spin_lo); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_ctx_lo; } @@ -723,14 +1047,13 @@ struct preempt_client { struct i915_gem_context *ctx; }; -static int preempt_client_init(struct drm_i915_private *i915, - struct preempt_client *c) +static int preempt_client_init(struct intel_gt *gt, struct preempt_client *c) { - c->ctx = kernel_context(i915); + c->ctx = kernel_context(gt->i915); if (!c->ctx) return -ENOMEM; - if (igt_spinner_init(&c->spin, &i915->gt)) + if (igt_spinner_init(&c->spin, gt)) goto err_ctx; return 0; @@ -748,11 +1071,10 @@ static void preempt_client_fini(struct preempt_client *c) static int live_nopreempt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *engine; struct preempt_client a, b; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; /* @@ -760,19 +1082,16 @@ static int live_nopreempt(void *arg) * that may be being observed and not want to be interrupted. */ - if (!HAS_LOGICAL_RING_PREEMPTION(i915)) + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - if (preempt_client_init(i915, &a)) - goto err_unlock; - if (preempt_client_init(i915, &b)) + if (preempt_client_init(gt, &a)) + return -ENOMEM; + if (preempt_client_init(gt, &b)) goto err_client_a; b.ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_MAX); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct i915_request *rq_a, *rq_b; if (!intel_engine_has_preemption(engine)) @@ -832,7 +1151,7 @@ static int live_nopreempt(void *arg) goto err_wedged; } - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) goto err_wedged; } @@ -841,29 +1160,344 @@ err_client_b: preempt_client_fini(&b); err_client_a: preempt_client_fini(&a); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: igt_spinner_end(&b.spin); igt_spinner_end(&a.spin); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_client_b; } +struct live_preempt_cancel { + struct intel_engine_cs *engine; + struct preempt_client a, b; +}; + +static int __cancel_active0(struct live_preempt_cancel *arg) +{ + struct i915_request *rq; + struct igt_live_test t; + int err; + + /* Preempt cancel of ELSP0 */ + GEM_TRACE("%s(%s)\n", __func__, arg->engine->name); + if (igt_live_test_begin(&t, arg->engine->i915, + __func__, arg->engine->name)) + return -EIO; + + clear_bit(CONTEXT_BANNED, &arg->a.ctx->flags); + rq = spinner_create_request(&arg->a.spin, + arg->a.ctx, arg->engine, + MI_ARB_CHECK); + if (IS_ERR(rq)) + return PTR_ERR(rq); + + i915_request_get(rq); + i915_request_add(rq); + if (!igt_wait_for_spinner(&arg->a.spin, rq)) { + err = -EIO; + goto out; + } + + i915_gem_context_set_banned(arg->a.ctx); + err = intel_engine_pulse(arg->engine); + if (err) + goto out; + + if (i915_request_wait(rq, 0, HZ / 5) < 0) { + err = -EIO; + goto out; + } + + if (rq->fence.error != -EIO) { + pr_err("Cancelled inflight0 request did not report -EIO\n"); + err = -EINVAL; + goto out; + } + +out: + i915_request_put(rq); + if (igt_live_test_end(&t)) + err = -EIO; + return err; +} + +static int __cancel_active1(struct live_preempt_cancel *arg) +{ + struct i915_request *rq[2] = {}; + struct igt_live_test t; + int err; + + /* Preempt cancel of ELSP1 */ + GEM_TRACE("%s(%s)\n", __func__, arg->engine->name); + if (igt_live_test_begin(&t, arg->engine->i915, + __func__, arg->engine->name)) + return -EIO; + + clear_bit(CONTEXT_BANNED, &arg->a.ctx->flags); + rq[0] = spinner_create_request(&arg->a.spin, + arg->a.ctx, arg->engine, + MI_NOOP); /* no preemption */ + if (IS_ERR(rq[0])) + return PTR_ERR(rq[0]); + + i915_request_get(rq[0]); + i915_request_add(rq[0]); + if (!igt_wait_for_spinner(&arg->a.spin, rq[0])) { + err = -EIO; + goto out; + } + + clear_bit(CONTEXT_BANNED, &arg->b.ctx->flags); + rq[1] = spinner_create_request(&arg->b.spin, + arg->b.ctx, arg->engine, + MI_ARB_CHECK); + if (IS_ERR(rq[1])) { + err = PTR_ERR(rq[1]); + goto out; + } + + i915_request_get(rq[1]); + err = i915_request_await_dma_fence(rq[1], &rq[0]->fence); + i915_request_add(rq[1]); + if (err) + goto out; + + i915_gem_context_set_banned(arg->b.ctx); + err = intel_engine_pulse(arg->engine); + if (err) + goto out; + + igt_spinner_end(&arg->a.spin); + if (i915_request_wait(rq[1], 0, HZ / 5) < 0) { + err = -EIO; + goto out; + } + + if (rq[0]->fence.error != 0) { + pr_err("Normal inflight0 request did not complete\n"); + err = -EINVAL; + goto out; + } + + if (rq[1]->fence.error != -EIO) { + pr_err("Cancelled inflight1 request did not report -EIO\n"); + err = -EINVAL; + goto out; + } + +out: + i915_request_put(rq[1]); + i915_request_put(rq[0]); + if (igt_live_test_end(&t)) + err = -EIO; + return err; +} + +static int __cancel_queued(struct live_preempt_cancel *arg) +{ + struct i915_request *rq[3] = {}; + struct igt_live_test t; + int err; + + /* Full ELSP and one in the wings */ + GEM_TRACE("%s(%s)\n", __func__, arg->engine->name); + if (igt_live_test_begin(&t, arg->engine->i915, + __func__, arg->engine->name)) + return -EIO; + + clear_bit(CONTEXT_BANNED, &arg->a.ctx->flags); + rq[0] = spinner_create_request(&arg->a.spin, + arg->a.ctx, arg->engine, + MI_ARB_CHECK); + if (IS_ERR(rq[0])) + return PTR_ERR(rq[0]); + + i915_request_get(rq[0]); + i915_request_add(rq[0]); + if (!igt_wait_for_spinner(&arg->a.spin, rq[0])) { + err = -EIO; + goto out; + } + + clear_bit(CONTEXT_BANNED, &arg->b.ctx->flags); + rq[1] = igt_request_alloc(arg->b.ctx, arg->engine); + if (IS_ERR(rq[1])) { + err = PTR_ERR(rq[1]); + goto out; + } + + i915_request_get(rq[1]); + err = i915_request_await_dma_fence(rq[1], &rq[0]->fence); + i915_request_add(rq[1]); + if (err) + goto out; + + rq[2] = spinner_create_request(&arg->b.spin, + arg->a.ctx, arg->engine, + MI_ARB_CHECK); + if (IS_ERR(rq[2])) { + err = PTR_ERR(rq[2]); + goto out; + } + + i915_request_get(rq[2]); + err = i915_request_await_dma_fence(rq[2], &rq[1]->fence); + i915_request_add(rq[2]); + if (err) + goto out; + + i915_gem_context_set_banned(arg->a.ctx); + err = intel_engine_pulse(arg->engine); + if (err) + goto out; + + if (i915_request_wait(rq[2], 0, HZ / 5) < 0) { + err = -EIO; + goto out; + } + + if (rq[0]->fence.error != -EIO) { + pr_err("Cancelled inflight0 request did not report -EIO\n"); + err = -EINVAL; + goto out; + } + + if (rq[1]->fence.error != 0) { + pr_err("Normal inflight1 request did not complete\n"); + err = -EINVAL; + goto out; + } + + if (rq[2]->fence.error != -EIO) { + pr_err("Cancelled queued request did not report -EIO\n"); + err = -EINVAL; + goto out; + } + +out: + i915_request_put(rq[2]); + i915_request_put(rq[1]); + i915_request_put(rq[0]); + if (igt_live_test_end(&t)) + err = -EIO; + return err; +} + +static int __cancel_hostile(struct live_preempt_cancel *arg) +{ + struct i915_request *rq; + int err; + + /* Preempt cancel non-preemptible spinner in ELSP0 */ + if (!IS_ACTIVE(CONFIG_DRM_I915_PREEMPT_TIMEOUT)) + return 0; + + GEM_TRACE("%s(%s)\n", __func__, arg->engine->name); + clear_bit(CONTEXT_BANNED, &arg->a.ctx->flags); + rq = spinner_create_request(&arg->a.spin, + arg->a.ctx, arg->engine, + MI_NOOP); /* preemption disabled */ + if (IS_ERR(rq)) + return PTR_ERR(rq); + + i915_request_get(rq); + i915_request_add(rq); + if (!igt_wait_for_spinner(&arg->a.spin, rq)) { + err = -EIO; + goto out; + } + + i915_gem_context_set_banned(arg->a.ctx); + err = intel_engine_pulse(arg->engine); /* force reset */ + if (err) + goto out; + + if (i915_request_wait(rq, 0, HZ / 5) < 0) { + err = -EIO; + goto out; + } + + if (rq->fence.error != -EIO) { + pr_err("Cancelled inflight0 request did not report -EIO\n"); + err = -EINVAL; + goto out; + } + +out: + i915_request_put(rq); + if (igt_flush_test(arg->engine->i915)) + err = -EIO; + return err; +} + +static int live_preempt_cancel(void *arg) +{ + struct intel_gt *gt = arg; + struct live_preempt_cancel data; + enum intel_engine_id id; + int err = -ENOMEM; + + /* + * To cancel an inflight context, we need to first remove it from the + * GPU. That sounds like preemption! Plus a little bit of bookkeeping. + */ + + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) + return 0; + + if (preempt_client_init(gt, &data.a)) + return -ENOMEM; + if (preempt_client_init(gt, &data.b)) + goto err_client_a; + + for_each_engine(data.engine, gt, id) { + if (!intel_engine_has_preemption(data.engine)) + continue; + + err = __cancel_active0(&data); + if (err) + goto err_wedged; + + err = __cancel_active1(&data); + if (err) + goto err_wedged; + + err = __cancel_queued(&data); + if (err) + goto err_wedged; + + err = __cancel_hostile(&data); + if (err) + goto err_wedged; + } + + err = 0; +err_client_b: + preempt_client_fini(&data.b); +err_client_a: + preempt_client_fini(&data.a); + return err; + +err_wedged: + GEM_TRACE_DUMP(); + igt_spinner_end(&data.b.spin); + igt_spinner_end(&data.a.spin); + intel_gt_set_wedged(gt); + goto err_client_b; +} + static int live_suppress_self_preempt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *engine; struct i915_sched_attr attr = { .priority = I915_USER_PRIORITY(I915_PRIORITY_MAX) }; struct preempt_client a, b; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; /* @@ -873,30 +1507,31 @@ static int live_suppress_self_preempt(void *arg) * completion event. */ - if (!HAS_LOGICAL_RING_PREEMPTION(i915)) + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) return 0; - if (USES_GUC_SUBMISSION(i915)) + if (USES_GUC_SUBMISSION(gt->i915)) return 0; /* presume black blox */ - if (intel_vgpu_active(i915)) + if (intel_vgpu_active(gt->i915)) return 0; /* GVT forces single port & request submission */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - if (preempt_client_init(i915, &a)) - goto err_unlock; - if (preempt_client_init(i915, &b)) + if (preempt_client_init(gt, &a)) + return -ENOMEM; + if (preempt_client_init(gt, &b)) goto err_client_a; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct i915_request *rq_a, *rq_b; int depth; if (!intel_engine_has_preemption(engine)) continue; + if (igt_flush_test(gt->i915)) + goto err_wedged; + + intel_engine_pm_get(engine); engine->execlists.preempt_hang.count = 0; rq_a = spinner_create_request(&a.spin, @@ -904,12 +1539,14 @@ static int live_suppress_self_preempt(void *arg) MI_NOOP); if (IS_ERR(rq_a)) { err = PTR_ERR(rq_a); + intel_engine_pm_put(engine); goto err_client_b; } i915_request_add(rq_a); if (!igt_wait_for_spinner(&a.spin, rq_a)) { pr_err("First client failed to start\n"); + intel_engine_pm_put(engine); goto err_wedged; } @@ -921,6 +1558,7 @@ static int live_suppress_self_preempt(void *arg) MI_NOOP); if (IS_ERR(rq_b)) { err = PTR_ERR(rq_b); + intel_engine_pm_put(engine); goto err_client_b; } i915_request_add(rq_b); @@ -931,6 +1569,7 @@ static int live_suppress_self_preempt(void *arg) if (!igt_wait_for_spinner(&b.spin, rq_b)) { pr_err("Second client failed to start\n"); + intel_engine_pm_put(engine); goto err_wedged; } @@ -944,11 +1583,13 @@ static int live_suppress_self_preempt(void *arg) engine->name, engine->execlists.preempt_hang.count, depth); + intel_engine_pm_put(engine); err = -EINVAL; goto err_client_b; } - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + intel_engine_pm_put(engine); + if (igt_flush_test(gt->i915)) goto err_wedged; } @@ -957,15 +1598,12 @@ err_client_b: preempt_client_fini(&b); err_client_a: preempt_client_fini(&a); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: igt_spinner_end(&b.spin); igt_spinner_end(&a.spin); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_client_b; } @@ -984,9 +1622,13 @@ static struct i915_request *dummy_request(struct intel_engine_cs *engine) if (!rq) return NULL; - INIT_LIST_HEAD(&rq->active_list); rq->engine = engine; + spin_lock_init(&rq->lock); + INIT_LIST_HEAD(&rq->fence.cb_list); + rq->fence.lock = &rq->lock; + rq->fence.ops = &i915_fence_ops; + i915_sched_node_init(&rq->sched); /* mark this request as permanently incomplete */ @@ -1021,11 +1663,10 @@ static void dummy_request_free(struct i915_request *dummy) static int live_suppress_wait_preempt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct preempt_client client[4]; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; int i; @@ -1035,22 +1676,19 @@ static int live_suppress_wait_preempt(void *arg) * not needlessly generate preempt-to-idle cycles. */ - if (!HAS_LOGICAL_RING_PREEMPTION(i915)) + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - if (preempt_client_init(i915, &client[0])) /* ELSP[0] */ - goto err_unlock; - if (preempt_client_init(i915, &client[1])) /* ELSP[1] */ + if (preempt_client_init(gt, &client[0])) /* ELSP[0] */ + return -ENOMEM; + if (preempt_client_init(gt, &client[1])) /* ELSP[1] */ goto err_client_0; - if (preempt_client_init(i915, &client[2])) /* head of queue */ + if (preempt_client_init(gt, &client[2])) /* head of queue */ goto err_client_1; - if (preempt_client_init(i915, &client[3])) /* bystander */ + if (preempt_client_init(gt, &client[3])) /* bystander */ goto err_client_2; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { int depth; if (!intel_engine_has_preemption(engine)) @@ -1079,8 +1717,8 @@ static int live_suppress_wait_preempt(void *arg) } /* Disable NEWCLIENT promotion */ - __i915_active_request_set(&rq[i]->timeline->last_request, - dummy); + __i915_active_fence_set(&i915_request_timeline(rq[i])->last_request, + &dummy->fence); i915_request_add(rq[i]); } @@ -1105,7 +1743,7 @@ static int live_suppress_wait_preempt(void *arg) for (i = 0; i < ARRAY_SIZE(client); i++) igt_spinner_end(&client[i].spin); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) goto err_wedged; if (engine->execlists.preempt_hang.count) { @@ -1128,26 +1766,22 @@ err_client_1: preempt_client_fini(&client[1]); err_client_0: preempt_client_fini(&client[0]); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: for (i = 0; i < ARRAY_SIZE(client); i++) igt_spinner_end(&client[i].spin); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_client_3; } static int live_chain_preempt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *engine; struct preempt_client hi, lo; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; /* @@ -1156,19 +1790,16 @@ static int live_chain_preempt(void *arg) * the previously submitted spinner in B. */ - if (!HAS_LOGICAL_RING_PREEMPTION(i915)) + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - if (preempt_client_init(i915, &hi)) - goto err_unlock; + if (preempt_client_init(gt, &hi)) + return -ENOMEM; - if (preempt_client_init(i915, &lo)) + if (preempt_client_init(gt, &lo)) goto err_client_hi; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct i915_sched_attr attr = { .priority = I915_USER_PRIORITY(I915_PRIORITY_MAX), }; @@ -1199,7 +1830,7 @@ static int live_chain_preempt(void *arg) goto err_wedged; } - if (igt_live_test_begin(&t, i915, __func__, engine->name)) { + if (igt_live_test_begin(&t, gt->i915, __func__, engine->name)) { err = -EIO; goto err_wedged; } @@ -1237,7 +1868,7 @@ static int live_chain_preempt(void *arg) igt_spinner_end(&hi.spin); if (i915_request_wait(rq, 0, HZ / 5) < 0) { struct drm_printer p = - drm_info_printer(i915->drm.dev); + drm_info_printer(gt->i915->drm.dev); pr_err("Failed to preempt over chain of %d\n", count); @@ -1253,7 +1884,7 @@ static int live_chain_preempt(void *arg) i915_request_add(rq); if (i915_request_wait(rq, 0, HZ / 5) < 0) { struct drm_printer p = - drm_info_printer(i915->drm.dev); + drm_info_printer(gt->i915->drm.dev); pr_err("Failed to flush low priority chain of %d requests\n", count); @@ -1274,57 +1905,50 @@ err_client_lo: preempt_client_fini(&lo); err_client_hi: preempt_client_fini(&hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: igt_spinner_end(&hi.spin); igt_spinner_end(&lo.spin); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_client_lo; } static int live_preempt_hang(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_gem_context *ctx_hi, *ctx_lo; struct igt_spinner spin_hi, spin_lo; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; - if (!HAS_LOGICAL_RING_PREEMPTION(i915)) + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) return 0; - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(gt)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - if (igt_spinner_init(&spin_hi, &i915->gt)) - goto err_unlock; + if (igt_spinner_init(&spin_hi, gt)) + return -ENOMEM; - if (igt_spinner_init(&spin_lo, &i915->gt)) + if (igt_spinner_init(&spin_lo, gt)) goto err_spin_hi; - ctx_hi = kernel_context(i915); + ctx_hi = kernel_context(gt->i915); if (!ctx_hi) goto err_spin_lo; ctx_hi->sched.priority = I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY); - ctx_lo = kernel_context(i915); + ctx_lo = kernel_context(gt->i915); if (!ctx_lo) goto err_ctx_hi; ctx_lo->sched.priority = I915_USER_PRIORITY(I915_CONTEXT_MIN_USER_PRIORITY); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct i915_request *rq; if (!intel_engine_has_preemption(engine)) @@ -1341,7 +1965,7 @@ static int live_preempt_hang(void *arg) if (!igt_wait_for_spinner(&spin_lo, rq)) { GEM_TRACE("lo spinner failed to start\n"); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_ctx_lo; } @@ -1363,28 +1987,28 @@ static int live_preempt_hang(void *arg) HZ / 10)) { pr_err("Preemption did not occur within timeout!"); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_ctx_lo; } - set_bit(I915_RESET_ENGINE + id, &i915->gt.reset.flags); + set_bit(I915_RESET_ENGINE + id, >->reset.flags); intel_engine_reset(engine, NULL); - clear_bit(I915_RESET_ENGINE + id, &i915->gt.reset.flags); + clear_bit(I915_RESET_ENGINE + id, >->reset.flags); engine->execlists.preempt_hang.inject_hang = false; if (!igt_wait_for_spinner(&spin_hi, rq)) { GEM_TRACE("hi spinner failed to start\n"); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto err_ctx_lo; } igt_spinner_end(&spin_hi); igt_spinner_end(&spin_lo); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) { + if (igt_flush_test(gt->i915)) { err = -EIO; goto err_ctx_lo; } @@ -1399,9 +2023,105 @@ err_spin_lo: igt_spinner_fini(&spin_lo); err_spin_hi: igt_spinner_fini(&spin_hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); + return err; +} + +static int live_preempt_timeout(void *arg) +{ + struct intel_gt *gt = arg; + struct i915_gem_context *ctx_hi, *ctx_lo; + struct igt_spinner spin_lo; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err = -ENOMEM; + + /* + * Check that we force preemption to occur by cancelling the previous + * context if it refuses to yield the GPU. + */ + if (!IS_ACTIVE(CONFIG_DRM_I915_PREEMPT_TIMEOUT)) + return 0; + + if (!HAS_LOGICAL_RING_PREEMPTION(gt->i915)) + return 0; + + if (!intel_has_reset_engine(gt)) + return 0; + + if (igt_spinner_init(&spin_lo, gt)) + return -ENOMEM; + + ctx_hi = kernel_context(gt->i915); + if (!ctx_hi) + goto err_spin_lo; + ctx_hi->sched.priority = + I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY); + + ctx_lo = kernel_context(gt->i915); + if (!ctx_lo) + goto err_ctx_hi; + ctx_lo->sched.priority = + I915_USER_PRIORITY(I915_CONTEXT_MIN_USER_PRIORITY); + + for_each_engine(engine, gt, id) { + unsigned long saved_timeout; + struct i915_request *rq; + + if (!intel_engine_has_preemption(engine)) + continue; + + rq = spinner_create_request(&spin_lo, ctx_lo, engine, + MI_NOOP); /* preemption disabled */ + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto err_ctx_lo; + } + + i915_request_add(rq); + if (!igt_wait_for_spinner(&spin_lo, rq)) { + intel_gt_set_wedged(gt); + err = -EIO; + goto err_ctx_lo; + } + + rq = igt_request_alloc(ctx_hi, engine); + if (IS_ERR(rq)) { + igt_spinner_end(&spin_lo); + err = PTR_ERR(rq); + goto err_ctx_lo; + } + + /* Flush the previous CS ack before changing timeouts */ + while (READ_ONCE(engine->execlists.pending[0])) + cpu_relax(); + + saved_timeout = engine->props.preempt_timeout_ms; + engine->props.preempt_timeout_ms = 1; /* in ms, -> 1 jiffie */ + + i915_request_get(rq); + i915_request_add(rq); + + intel_engine_flush_submission(engine); + engine->props.preempt_timeout_ms = saved_timeout; + + if (i915_request_wait(rq, 0, HZ / 10) < 0) { + intel_gt_set_wedged(gt); + i915_request_put(rq); + err = -ETIME; + goto err_ctx_lo; + } + + igt_spinner_end(&spin_lo); + i915_request_put(rq); + } + + err = 0; +err_ctx_lo: + kernel_context_close(ctx_lo); +err_ctx_hi: + kernel_context_close(ctx_hi); +err_spin_lo: + igt_spinner_fini(&spin_lo); return err; } @@ -1416,7 +2136,7 @@ static int random_priority(struct rnd_state *rnd) } struct preempt_smoke { - struct drm_i915_private *i915; + struct intel_gt *gt; struct i915_gem_context **contexts; struct intel_engine_cs *engine; struct drm_i915_gem_object *batch; @@ -1440,7 +2160,11 @@ static int smoke_submit(struct preempt_smoke *smoke, int err = 0; if (batch) { - vma = i915_vma_instance(batch, ctx->vm, NULL); + struct i915_address_space *vm; + + vm = i915_gem_context_get_vm_rcu(ctx); + vma = i915_vma_instance(batch, vm, NULL); + i915_vm_put(vm); if (IS_ERR(vma)) return PTR_ERR(vma); @@ -1489,11 +2213,9 @@ static int smoke_crescendo_thread(void *arg) struct i915_gem_context *ctx = smoke_context(smoke); int err; - mutex_lock(&smoke->i915->drm.struct_mutex); err = smoke_submit(smoke, ctx, count % I915_PRIORITY_MAX, smoke->batch); - mutex_unlock(&smoke->i915->drm.struct_mutex); if (err) return err; @@ -1514,9 +2236,7 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags) unsigned long count; int err = 0; - mutex_unlock(&smoke->i915->drm.struct_mutex); - - for_each_engine(engine, smoke->i915, id) { + for_each_engine(engine, smoke->gt, id) { arg[id] = *smoke; arg[id].engine = engine; if (!(flags & BATCH)) @@ -1532,8 +2252,10 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags) get_task_struct(tsk[id]); } + yield(); /* start all threads before we kthread_stop() */ + count = 0; - for_each_engine(engine, smoke->i915, id) { + for_each_engine(engine, smoke->gt, id) { int status; if (IS_ERR_OR_NULL(tsk[id])) @@ -1548,11 +2270,9 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags) put_task_struct(tsk[id]); } - mutex_lock(&smoke->i915->drm.struct_mutex); - pr_info("Submitted %lu crescendo:%x requests across %d engines and %d contexts\n", count, flags, - RUNTIME_INFO(smoke->i915)->num_engines, smoke->ncontext); + RUNTIME_INFO(smoke->gt->i915)->num_engines, smoke->ncontext); return 0; } @@ -1564,7 +2284,7 @@ static int smoke_random(struct preempt_smoke *smoke, unsigned int flags) count = 0; do { - for_each_engine(smoke->engine, smoke->i915, id) { + for_each_engine(smoke->engine, smoke->gt, id) { struct i915_gem_context *ctx = smoke_context(smoke); int err; @@ -1580,25 +2300,24 @@ static int smoke_random(struct preempt_smoke *smoke, unsigned int flags) pr_info("Submitted %lu random:%x requests across %d engines and %d contexts\n", count, flags, - RUNTIME_INFO(smoke->i915)->num_engines, smoke->ncontext); + RUNTIME_INFO(smoke->gt->i915)->num_engines, smoke->ncontext); return 0; } static int live_preempt_smoke(void *arg) { struct preempt_smoke smoke = { - .i915 = arg, + .gt = arg, .prng = I915_RND_STATE_INITIALIZER(i915_selftest.random_seed), .ncontext = 1024, }; const unsigned int phase[] = { 0, BATCH }; - intel_wakeref_t wakeref; struct igt_live_test t; int err = -ENOMEM; u32 *cs; int n; - if (!HAS_LOGICAL_RING_PREEMPTION(smoke.i915)) + if (!HAS_LOGICAL_RING_PREEMPTION(smoke.gt->i915)) return 0; smoke.contexts = kmalloc_array(smoke.ncontext, @@ -1607,13 +2326,11 @@ static int live_preempt_smoke(void *arg) if (!smoke.contexts) return -ENOMEM; - mutex_lock(&smoke.i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&smoke.i915->runtime_pm); - - smoke.batch = i915_gem_object_create_internal(smoke.i915, PAGE_SIZE); + smoke.batch = + i915_gem_object_create_internal(smoke.gt->i915, PAGE_SIZE); if (IS_ERR(smoke.batch)) { err = PTR_ERR(smoke.batch); - goto err_unlock; + goto err_free; } cs = i915_gem_object_pin_map(smoke.batch, I915_MAP_WB); @@ -1627,13 +2344,13 @@ static int live_preempt_smoke(void *arg) i915_gem_object_flush_map(smoke.batch); i915_gem_object_unpin_map(smoke.batch); - if (igt_live_test_begin(&t, smoke.i915, __func__, "all")) { + if (igt_live_test_begin(&t, smoke.gt->i915, __func__, "all")) { err = -EIO; goto err_batch; } for (n = 0; n < smoke.ncontext; n++) { - smoke.contexts[n] = kernel_context(smoke.i915); + smoke.contexts[n] = kernel_context(smoke.gt->i915); if (!smoke.contexts[n]) goto err_ctx; } @@ -1660,15 +2377,13 @@ err_ctx: err_batch: i915_gem_object_put(smoke.batch); -err_unlock: - intel_runtime_pm_put(&smoke.i915->runtime_pm, wakeref); - mutex_unlock(&smoke.i915->drm.struct_mutex); +err_free: kfree(smoke.contexts); return err; } -static int nop_virtual_engine(struct drm_i915_private *i915, +static int nop_virtual_engine(struct intel_gt *gt, struct intel_engine_cs **siblings, unsigned int nsibling, unsigned int nctx, @@ -1687,7 +2402,7 @@ static int nop_virtual_engine(struct drm_i915_private *i915, GEM_BUG_ON(!nctx || nctx > ARRAY_SIZE(ctx)); for (n = 0; n < nctx; n++) { - ctx[n] = kernel_context(i915); + ctx[n] = kernel_context(gt->i915); if (!ctx[n]) { err = -ENOMEM; nctx = n; @@ -1712,7 +2427,7 @@ static int nop_virtual_engine(struct drm_i915_private *i915, } } - err = igt_live_test_begin(&t, i915, __func__, ve[0]->engine->name); + err = igt_live_test_begin(&t, gt->i915, __func__, ve[0]->engine->name); if (err) goto out; @@ -1759,7 +2474,7 @@ static int nop_virtual_engine(struct drm_i915_private *i915, request[nc]->fence.context, request[nc]->fence.seqno); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); break; } } @@ -1781,7 +2496,7 @@ static int nop_virtual_engine(struct drm_i915_private *i915, prime, div64_u64(ktime_to_ns(times[1]), prime)); out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) err = -EIO; for (nc = 0; nc < nctx; nc++) { @@ -1794,25 +2509,22 @@ out: static int live_virtual_engine(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; struct intel_engine_cs *engine; - struct intel_gt *gt = &i915->gt; enum intel_engine_id id; unsigned int class, inst; - int err = -ENODEV; + int err; - if (USES_GUC_SUBMISSION(i915)) + if (USES_GUC_SUBMISSION(gt->i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - - for_each_engine(engine, i915, id) { - err = nop_virtual_engine(i915, &engine, 1, 1, 0); + for_each_engine(engine, gt, id) { + err = nop_virtual_engine(gt, &engine, 1, 1, 0); if (err) { pr_err("Failed to wrap engine %s: err=%d\n", engine->name, err); - goto out_unlock; + return err; } } @@ -1830,23 +2542,21 @@ static int live_virtual_engine(void *arg) continue; for (n = 1; n <= nsibling + 1; n++) { - err = nop_virtual_engine(i915, siblings, nsibling, + err = nop_virtual_engine(gt, siblings, nsibling, n, 0); if (err) - goto out_unlock; + return err; } - err = nop_virtual_engine(i915, siblings, nsibling, n, CHAIN); + err = nop_virtual_engine(gt, siblings, nsibling, n, CHAIN); if (err) - goto out_unlock; + return err; } -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - return err; + return 0; } -static int mask_virtual_engine(struct drm_i915_private *i915, +static int mask_virtual_engine(struct intel_gt *gt, struct intel_engine_cs **siblings, unsigned int nsibling) { @@ -1862,7 +2572,7 @@ static int mask_virtual_engine(struct drm_i915_private *i915, * restrict it to our desired engine within the virtual engine. */ - ctx = kernel_context(i915); + ctx = kernel_context(gt->i915); if (!ctx) return -ENOMEM; @@ -1876,7 +2586,7 @@ static int mask_virtual_engine(struct drm_i915_private *i915, if (err) goto out_put; - err = igt_live_test_begin(&t, i915, __func__, ve->engine->name); + err = igt_live_test_begin(&t, gt->i915, __func__, ve->engine->name); if (err) goto out_unpin; @@ -1907,7 +2617,7 @@ static int mask_virtual_engine(struct drm_i915_private *i915, request[n]->fence.context, request[n]->fence.seqno); GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); err = -EIO; goto out; } @@ -1922,11 +2632,8 @@ static int mask_virtual_engine(struct drm_i915_private *i915, } err = igt_live_test_end(&t); - if (err) - goto out; - out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) err = -EIO; for (n = 0; n < nsibling; n++) @@ -1943,17 +2650,14 @@ out_close: static int live_virtual_mask(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; - struct intel_gt *gt = &i915->gt; unsigned int class, inst; - int err = 0; + int err; - if (USES_GUC_SUBMISSION(i915)) + if (USES_GUC_SUBMISSION(gt->i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - for (class = 0; class <= MAX_ENGINE_CLASS; class++) { unsigned int nsibling; @@ -1967,17 +2671,166 @@ static int live_virtual_mask(void *arg) if (nsibling < 2) continue; - err = mask_virtual_engine(i915, siblings, nsibling); + err = mask_virtual_engine(gt, siblings, nsibling); if (err) - goto out_unlock; + return err; + } + + return 0; +} + +static int preserved_virtual_engine(struct intel_gt *gt, + struct intel_engine_cs **siblings, + unsigned int nsibling) +{ + struct i915_request *last = NULL; + struct i915_gem_context *ctx; + struct intel_context *ve; + struct i915_vma *scratch; + struct igt_live_test t; + unsigned int n; + int err = 0; + u32 *cs; + + ctx = kernel_context(gt->i915); + if (!ctx) + return -ENOMEM; + + scratch = create_scratch(siblings[0]->gt); + if (IS_ERR(scratch)) { + err = PTR_ERR(scratch); + goto out_close; } -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); + ve = intel_execlists_create_virtual(ctx, siblings, nsibling); + if (IS_ERR(ve)) { + err = PTR_ERR(ve); + goto out_scratch; + } + + err = intel_context_pin(ve); + if (err) + goto out_put; + + err = igt_live_test_begin(&t, gt->i915, __func__, ve->engine->name); + if (err) + goto out_unpin; + + for (n = 0; n < NUM_GPR_DW; n++) { + struct intel_engine_cs *engine = siblings[n % nsibling]; + struct i915_request *rq; + + rq = i915_request_create(ve); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto out_end; + } + + i915_request_put(last); + last = i915_request_get(rq); + + cs = intel_ring_begin(rq, 8); + if (IS_ERR(cs)) { + i915_request_add(rq); + err = PTR_ERR(cs); + goto out_end; + } + + *cs++ = MI_STORE_REGISTER_MEM_GEN8 | MI_USE_GGTT; + *cs++ = CS_GPR(engine, n); + *cs++ = i915_ggtt_offset(scratch) + n * sizeof(u32); + *cs++ = 0; + + *cs++ = MI_LOAD_REGISTER_IMM(1); + *cs++ = CS_GPR(engine, (n + 1) % NUM_GPR_DW); + *cs++ = n + 1; + + *cs++ = MI_NOOP; + intel_ring_advance(rq, cs); + + /* Restrict this request to run on a particular engine */ + rq->execution_mask = engine->mask; + i915_request_add(rq); + } + + if (i915_request_wait(last, 0, HZ / 5) < 0) { + err = -ETIME; + goto out_end; + } + + cs = i915_gem_object_pin_map(scratch->obj, I915_MAP_WB); + if (IS_ERR(cs)) { + err = PTR_ERR(cs); + goto out_end; + } + + for (n = 0; n < NUM_GPR_DW; n++) { + if (cs[n] != n) { + pr_err("Incorrect value[%d] found for GPR[%d]\n", + cs[n], n); + err = -EINVAL; + break; + } + } + + i915_gem_object_unpin_map(scratch->obj); + +out_end: + if (igt_live_test_end(&t)) + err = -EIO; + i915_request_put(last); +out_unpin: + intel_context_unpin(ve); +out_put: + intel_context_put(ve); +out_scratch: + i915_vma_unpin_and_release(&scratch, 0); +out_close: + kernel_context_close(ctx); return err; } -static int bond_virtual_engine(struct drm_i915_private *i915, +static int live_virtual_preserved(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; + unsigned int class, inst; + + /* + * Check that the context image retains non-privileged (user) registers + * from one engine to the next. For this we check that the CS_GPR + * are preserved. + */ + + if (USES_GUC_SUBMISSION(gt->i915)) + return 0; + + /* As we use CS_GPR we cannot run before they existed on all engines. */ + if (INTEL_GEN(gt->i915) < 9) + return 0; + + for (class = 0; class <= MAX_ENGINE_CLASS; class++) { + int nsibling, err; + + nsibling = 0; + for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) { + if (!gt->engine_class[class][inst]) + continue; + + siblings[nsibling++] = gt->engine_class[class][inst]; + } + if (nsibling < 2) + continue; + + err = preserved_virtual_engine(gt, siblings, nsibling); + if (err) + return err; + } + + return 0; +} + +static int bond_virtual_engine(struct intel_gt *gt, unsigned int class, struct intel_engine_cs **siblings, unsigned int nsibling, @@ -1993,13 +2846,13 @@ static int bond_virtual_engine(struct drm_i915_private *i915, GEM_BUG_ON(nsibling >= ARRAY_SIZE(rq) - 1); - ctx = kernel_context(i915); + ctx = kernel_context(gt->i915); if (!ctx) return -ENOMEM; err = 0; rq[0] = ERR_PTR(-ENOMEM); - for_each_engine(master, i915, id) { + for_each_engine(master, gt, id) { struct i915_sw_fence fence = {}; if (master->class == class) @@ -2104,7 +2957,7 @@ static int bond_virtual_engine(struct drm_i915_private *i915, out: for (n = 0; !IS_ERR(rq[n]); n++) i915_request_put(rq[n]); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) err = -EIO; kernel_context_close(ctx); @@ -2121,17 +2974,14 @@ static int live_virtual_bond(void *arg) { "schedule", BOND_SCHEDULE }, { }, }; - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; - struct intel_gt *gt = &i915->gt; unsigned int class, inst; - int err = 0; + int err; - if (USES_GUC_SUBMISSION(i915)) + if (USES_GUC_SUBMISSION(gt->i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - for (class = 0; class <= MAX_ENGINE_CLASS; class++) { const struct phase *p; int nsibling; @@ -2148,38 +2998,42 @@ static int live_virtual_bond(void *arg) continue; for (p = phases; p->name; p++) { - err = bond_virtual_engine(i915, + err = bond_virtual_engine(gt, class, siblings, nsibling, p->flags); if (err) { pr_err("%s(%s): failed class=%d, nsibling=%d, err=%d\n", __func__, p->name, class, nsibling, err); - goto out_unlock; + return err; } } } -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - return err; + return 0; } int intel_execlists_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(live_sanitycheck), + SUBTEST(live_unlite_switch), + SUBTEST(live_unlite_preempt), SUBTEST(live_timeslice_preempt), + SUBTEST(live_timeslice_queue), SUBTEST(live_busywait_preempt), SUBTEST(live_preempt), SUBTEST(live_late_preempt), SUBTEST(live_nopreempt), + SUBTEST(live_preempt_cancel), SUBTEST(live_suppress_self_preempt), SUBTEST(live_suppress_wait_preempt), SUBTEST(live_chain_preempt), SUBTEST(live_preempt_hang), + SUBTEST(live_preempt_timeout), SUBTEST(live_preempt_smoke), SUBTEST(live_virtual_engine), SUBTEST(live_virtual_mask), + SUBTEST(live_virtual_preserved), SUBTEST(live_virtual_bond), }; @@ -2189,5 +3043,512 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915) if (intel_gt_is_wedged(&i915->gt)) return 0; - return i915_live_subtests(tests, i915); + return intel_gt_live_subtests(tests, &i915->gt); +} + +static void hexdump(const void *buf, size_t len) +{ + const size_t rowsize = 8 * sizeof(u32); + const void *prev = NULL; + bool skip = false; + size_t pos; + + for (pos = 0; pos < len; pos += rowsize) { + char line[128]; + + if (prev && !memcmp(prev, buf + pos, rowsize)) { + if (!skip) { + pr_info("*\n"); + skip = true; + } + continue; + } + + WARN_ON_ONCE(hex_dump_to_buffer(buf + pos, len - pos, + rowsize, sizeof(u32), + line, sizeof(line), + false) >= sizeof(line)); + pr_info("[%04zx] %s\n", pos, line); + + prev = buf + pos; + skip = false; + } +} + +static int live_lrc_layout(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; + u32 *mem; + int err; + + /* + * Check the registers offsets we use to create the initial reg state + * match the layout saved by HW. + */ + + mem = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!mem) + return -ENOMEM; + + err = 0; + for_each_engine(engine, gt, id) { + u32 *hw, *lrc; + int dw; + + if (!engine->default_state) + continue; + + hw = i915_gem_object_pin_map(engine->default_state, + I915_MAP_WB); + if (IS_ERR(hw)) { + err = PTR_ERR(hw); + break; + } + hw += LRC_STATE_PN * PAGE_SIZE / sizeof(*hw); + + lrc = memset(mem, 0, PAGE_SIZE); + execlists_init_reg_state(lrc, + engine->kernel_context, + engine, + engine->kernel_context->ring, + true); + + dw = 0; + do { + u32 lri = hw[dw]; + + if (lri == 0) { + dw++; + continue; + } + + if ((lri & GENMASK(31, 23)) != MI_INSTR(0x22, 0)) { + pr_err("%s: Expected LRI command at dword %d, found %08x\n", + engine->name, dw, lri); + err = -EINVAL; + break; + } + + if (lrc[dw] != lri) { + pr_err("%s: LRI command mismatch at dword %d, expected %08x found %08x\n", + engine->name, dw, lri, lrc[dw]); + err = -EINVAL; + break; + } + + lri &= 0x7f; + lri++; + dw++; + + while (lri) { + if (hw[dw] != lrc[dw]) { + pr_err("%s: Different registers found at dword %d, expected %x, found %x\n", + engine->name, dw, hw[dw], lrc[dw]); + err = -EINVAL; + break; + } + + /* + * Skip over the actual register value as we + * expect that to differ. + */ + dw += 2; + lri -= 2; + } + } while ((lrc[dw] & ~BIT(0)) != MI_BATCH_BUFFER_END); + + if (err) { + pr_info("%s: HW register image:\n", engine->name); + hexdump(hw, PAGE_SIZE); + + pr_info("%s: SW register image:\n", engine->name); + hexdump(lrc, PAGE_SIZE); + } + + i915_gem_object_unpin_map(engine->default_state); + if (err) + break; + } + + kfree(mem); + return err; +} + +static int find_offset(const u32 *lri, u32 offset) +{ + int i; + + for (i = 0; i < PAGE_SIZE / sizeof(u32); i++) + if (lri[i] == offset) + return i; + + return -1; +} + +static int live_lrc_fixed(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err = 0; + + /* + * Check the assumed register offsets match the actual locations in + * the context image. + */ + + for_each_engine(engine, gt, id) { + const struct { + u32 reg; + u32 offset; + const char *name; + } tbl[] = { + { + i915_mmio_reg_offset(RING_START(engine->mmio_base)), + CTX_RING_BUFFER_START - 1, + "RING_START" + }, + { + i915_mmio_reg_offset(RING_CTL(engine->mmio_base)), + CTX_RING_BUFFER_CONTROL - 1, + "RING_CTL" + }, + { + i915_mmio_reg_offset(RING_HEAD(engine->mmio_base)), + CTX_RING_HEAD - 1, + "RING_HEAD" + }, + { + i915_mmio_reg_offset(RING_TAIL(engine->mmio_base)), + CTX_RING_TAIL - 1, + "RING_TAIL" + }, + { + i915_mmio_reg_offset(RING_MI_MODE(engine->mmio_base)), + lrc_ring_mi_mode(engine), + "RING_MI_MODE" + }, + { + engine->mmio_base + 0x110, + CTX_BB_STATE - 1, + "BB_STATE" + }, + { }, + }, *t; + u32 *hw; + + if (!engine->default_state) + continue; + + hw = i915_gem_object_pin_map(engine->default_state, + I915_MAP_WB); + if (IS_ERR(hw)) { + err = PTR_ERR(hw); + break; + } + hw += LRC_STATE_PN * PAGE_SIZE / sizeof(*hw); + + for (t = tbl; t->name; t++) { + int dw = find_offset(hw, t->reg); + + if (dw != t->offset) { + pr_err("%s: Offset for %s [0x%x] mismatch, found %x, expected %x\n", + engine->name, + t->name, + t->reg, + dw, + t->offset); + err = -EINVAL; + } + } + + i915_gem_object_unpin_map(engine->default_state); + } + + return err; +} + +static int __live_lrc_state(struct i915_gem_context *fixme, + struct intel_engine_cs *engine, + struct i915_vma *scratch) +{ + struct intel_context *ce; + struct i915_request *rq; + enum { + RING_START_IDX = 0, + RING_TAIL_IDX, + MAX_IDX + }; + u32 expected[MAX_IDX]; + u32 *cs; + int err; + int n; + + ce = intel_context_create(fixme, engine); + if (IS_ERR(ce)) + return PTR_ERR(ce); + + err = intel_context_pin(ce); + if (err) + goto err_put; + + rq = i915_request_create(ce); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto err_unpin; + } + + cs = intel_ring_begin(rq, 4 * MAX_IDX); + if (IS_ERR(cs)) { + err = PTR_ERR(cs); + i915_request_add(rq); + goto err_unpin; + } + + *cs++ = MI_STORE_REGISTER_MEM_GEN8 | MI_USE_GGTT; + *cs++ = i915_mmio_reg_offset(RING_START(engine->mmio_base)); + *cs++ = i915_ggtt_offset(scratch) + RING_START_IDX * sizeof(u32); + *cs++ = 0; + + expected[RING_START_IDX] = i915_ggtt_offset(ce->ring->vma); + + *cs++ = MI_STORE_REGISTER_MEM_GEN8 | MI_USE_GGTT; + *cs++ = i915_mmio_reg_offset(RING_TAIL(engine->mmio_base)); + *cs++ = i915_ggtt_offset(scratch) + RING_TAIL_IDX * sizeof(u32); + *cs++ = 0; + + i915_request_get(rq); + i915_request_add(rq); + + intel_engine_flush_submission(engine); + expected[RING_TAIL_IDX] = ce->ring->tail; + + if (i915_request_wait(rq, 0, HZ / 5) < 0) { + err = -ETIME; + goto err_rq; + } + + cs = i915_gem_object_pin_map(scratch->obj, I915_MAP_WB); + if (IS_ERR(cs)) { + err = PTR_ERR(cs); + goto err_rq; + } + + for (n = 0; n < MAX_IDX; n++) { + if (cs[n] != expected[n]) { + pr_err("%s: Stored register[%d] value[0x%x] did not match expected[0x%x]\n", + engine->name, n, cs[n], expected[n]); + err = -EINVAL; + break; + } + } + + i915_gem_object_unpin_map(scratch->obj); + +err_rq: + i915_request_put(rq); +err_unpin: + intel_context_unpin(ce); +err_put: + intel_context_put(ce); + return err; +} + +static int live_lrc_state(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + struct i915_gem_context *fixme; + struct i915_vma *scratch; + enum intel_engine_id id; + int err = 0; + + /* + * Check the live register state matches what we expect for this + * intel_context. + */ + + fixme = kernel_context(gt->i915); + if (!fixme) + return -ENOMEM; + + scratch = create_scratch(gt); + if (IS_ERR(scratch)) { + err = PTR_ERR(scratch); + goto out_close; + } + + for_each_engine(engine, gt, id) { + err = __live_lrc_state(fixme, engine, scratch); + if (err) + break; + } + + if (igt_flush_test(gt->i915)) + err = -EIO; + + i915_vma_unpin_and_release(&scratch, 0); +out_close: + kernel_context_close(fixme); + return err; +} + +static int gpr_make_dirty(struct intel_engine_cs *engine) +{ + struct i915_request *rq; + u32 *cs; + int n; + + rq = i915_request_create(engine->kernel_context); + if (IS_ERR(rq)) + return PTR_ERR(rq); + + cs = intel_ring_begin(rq, 2 * NUM_GPR_DW + 2); + if (IS_ERR(cs)) { + i915_request_add(rq); + return PTR_ERR(cs); + } + + *cs++ = MI_LOAD_REGISTER_IMM(NUM_GPR_DW); + for (n = 0; n < NUM_GPR_DW; n++) { + *cs++ = CS_GPR(engine, n); + *cs++ = STACK_MAGIC; + } + *cs++ = MI_NOOP; + + intel_ring_advance(rq, cs); + i915_request_add(rq); + + return 0; +} + +static int __live_gpr_clear(struct i915_gem_context *fixme, + struct intel_engine_cs *engine, + struct i915_vma *scratch) +{ + struct intel_context *ce; + struct i915_request *rq; + u32 *cs; + int err; + int n; + + if (INTEL_GEN(engine->i915) < 9 && engine->class != RENDER_CLASS) + return 0; /* GPR only on rcs0 for gen8 */ + + err = gpr_make_dirty(engine); + if (err) + return err; + + ce = intel_context_create(fixme, engine); + if (IS_ERR(ce)) + return PTR_ERR(ce); + + rq = intel_context_create_request(ce); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto err_put; + } + + cs = intel_ring_begin(rq, 4 * NUM_GPR_DW); + if (IS_ERR(cs)) { + err = PTR_ERR(cs); + i915_request_add(rq); + goto err_put; + } + + for (n = 0; n < NUM_GPR_DW; n++) { + *cs++ = MI_STORE_REGISTER_MEM_GEN8 | MI_USE_GGTT; + *cs++ = CS_GPR(engine, n); + *cs++ = i915_ggtt_offset(scratch) + n * sizeof(u32); + *cs++ = 0; + } + + i915_request_get(rq); + i915_request_add(rq); + + if (i915_request_wait(rq, 0, HZ / 5) < 0) { + err = -ETIME; + goto err_rq; + } + + cs = i915_gem_object_pin_map(scratch->obj, I915_MAP_WB); + if (IS_ERR(cs)) { + err = PTR_ERR(cs); + goto err_rq; + } + + for (n = 0; n < NUM_GPR_DW; n++) { + if (cs[n]) { + pr_err("%s: GPR[%d].%s was not zero, found 0x%08x!\n", + engine->name, + n / 2, n & 1 ? "udw" : "ldw", + cs[n]); + err = -EINVAL; + break; + } + } + + i915_gem_object_unpin_map(scratch->obj); + +err_rq: + i915_request_put(rq); +err_put: + intel_context_put(ce); + return err; +} + +static int live_gpr_clear(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + struct i915_gem_context *fixme; + struct i915_vma *scratch; + enum intel_engine_id id; + int err = 0; + + /* + * Check that GPR registers are cleared in new contexts as we need + * to avoid leaking any information from previous contexts. + */ + + fixme = kernel_context(gt->i915); + if (!fixme) + return -ENOMEM; + + scratch = create_scratch(gt); + if (IS_ERR(scratch)) { + err = PTR_ERR(scratch); + goto out_close; + } + + for_each_engine(engine, gt, id) { + err = __live_gpr_clear(fixme, engine, scratch); + if (err) + break; + } + + if (igt_flush_test(gt->i915)) + err = -EIO; + + i915_vma_unpin_and_release(&scratch, 0); +out_close: + kernel_context_close(fixme); + return err; +} + +int intel_lrc_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(live_lrc_layout), + SUBTEST(live_lrc_fixed), + SUBTEST(live_lrc_state), + SUBTEST(live_gpr_clear), + }; + + if (!HAS_LOGICAL_RING_CONTEXTS(i915)) + return 0; + + return intel_gt_live_subtests(tests, &i915->gt); } diff --git a/drivers/gpu/drm/i915/gt/selftest_reset.c b/drivers/gpu/drm/i915/gt/selftest_reset.c index 00a4f60cdfd5..6ad6aca315f6 100644 --- a/drivers/gpu/drm/i915/gt/selftest_reset.c +++ b/drivers/gpu/drm/i915/gt/selftest_reset.c @@ -17,7 +17,7 @@ static int igt_global_reset(void *arg) /* Check that we can issue a global GPU reset */ igt_global_reset_lock(gt); - wakeref = intel_runtime_pm_get(>->i915->runtime_pm); + wakeref = intel_runtime_pm_get(gt->uncore->rpm); reset_count = i915_reset_count(>->i915->gpu_error); @@ -28,7 +28,7 @@ static int igt_global_reset(void *arg) err = -EINVAL; } - intel_runtime_pm_put(>->i915->runtime_pm, wakeref); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); igt_global_reset_unlock(gt); if (intel_gt_is_wedged(gt)) @@ -45,14 +45,14 @@ static int igt_wedged_reset(void *arg) /* Check that we can recover a wedged device with a GPU reset */ igt_global_reset_lock(gt); - wakeref = intel_runtime_pm_get(>->i915->runtime_pm); + wakeref = intel_runtime_pm_get(gt->uncore->rpm); intel_gt_set_wedged(gt); GEM_BUG_ON(!intel_gt_is_wedged(gt)); intel_gt_reset(gt, ALL_ENGINES, NULL); - intel_runtime_pm_put(>->i915->runtime_pm, wakeref); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); igt_global_reset_unlock(gt); return intel_gt_is_wedged(gt) ? -EIO : 0; @@ -112,7 +112,7 @@ static int igt_atomic_engine_reset(void *arg) /* Check that the resets are usable from atomic context */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (USES_GUC_SUBMISSION(gt->i915)) @@ -125,8 +125,8 @@ static int igt_atomic_engine_reset(void *arg) if (!igt_force_reset(gt)) goto out_unlock; - for_each_engine(engine, gt->i915, id) { - tasklet_disable_nosync(&engine->execlists.tasklet); + for_each_engine(engine, gt, id) { + tasklet_disable(&engine->execlists.tasklet); intel_engine_pm_get(engine); for (p = igt_atomic_phases; p->name; p++) { @@ -170,7 +170,7 @@ int intel_reset_live_selftests(struct drm_i915_private *i915) }; struct intel_gt *gt = &i915->gt; - if (!intel_has_gpu_reset(gt->i915)) + if (!intel_has_gpu_reset(gt)) return 0; if (intel_gt_is_wedged(gt)) diff --git a/drivers/gpu/drm/i915/gt/selftest_timeline.c b/drivers/gpu/drm/i915/gt/selftest_timeline.c index 321481403165..f04a59fe5d2c 100644 --- a/drivers/gpu/drm/i915/gt/selftest_timeline.c +++ b/drivers/gpu/drm/i915/gt/selftest_timeline.c @@ -6,8 +6,10 @@ #include <linux/prime_numbers.h> -#include "gem/i915_gem_pm.h" +#include "intel_engine_pm.h" #include "intel_gt.h" +#include "intel_gt_requests.h" +#include "intel_ring.h" #include "../selftests/i915_random.h" #include "../i915_selftest.h" @@ -34,7 +36,7 @@ static unsigned long hwsp_cacheline(struct intel_timeline *tl) #define CACHELINES_PER_PAGE (PAGE_SIZE / CACHELINE_BYTES) struct mock_hwsp_freelist { - struct drm_i915_private *i915; + struct intel_gt *gt; struct radix_tree_root cachelines; struct intel_timeline **history; unsigned long count, max; @@ -67,7 +69,7 @@ static int __mock_hwsp_timeline(struct mock_hwsp_freelist *state, unsigned long cacheline; int err; - tl = intel_timeline_create(&state->i915->gt, NULL); + tl = intel_timeline_create(state->gt, NULL); if (IS_ERR(tl)) return PTR_ERR(tl); @@ -105,6 +107,7 @@ static int __mock_hwsp_timeline(struct mock_hwsp_freelist *state, static int mock_hwsp_freelist(void *arg) { struct mock_hwsp_freelist state; + struct drm_i915_private *i915; const struct { const char *name; unsigned int flags; @@ -116,12 +119,14 @@ static int mock_hwsp_freelist(void *arg) unsigned int na; int err = 0; + i915 = mock_gem_device(); + if (!i915) + return -ENOMEM; + INIT_RADIX_TREE(&state.cachelines, GFP_KERNEL); state.prng = I915_RND_STATE_INITIALIZER(i915_selftest.random_seed); - state.i915 = mock_gem_device(); - if (!state.i915) - return -ENOMEM; + state.gt = &i915->gt; /* * Create a bunch of timelines and check that their HWSP do not overlap. @@ -136,7 +141,6 @@ static int mock_hwsp_freelist(void *arg) goto err_put; } - mutex_lock(&state.i915->drm.struct_mutex); for (p = phases; p->name; p++) { pr_debug("%s(%s)\n", __func__, p->name); for_each_prime_number_from(na, 1, 2 * CACHELINES_PER_PAGE) { @@ -149,10 +153,9 @@ static int mock_hwsp_freelist(void *arg) out: for (na = 0; na < state.max; na++) __mock_hwsp_record(&state, na, NULL); - mutex_unlock(&state.i915->drm.struct_mutex); kfree(state.history); err_put: - drm_dev_put(&state.i915->drm); + drm_dev_put(&i915->drm); return err; } @@ -449,8 +452,6 @@ tl_write(struct intel_timeline *tl, struct intel_engine_cs *engine, u32 value) struct i915_request *rq; int err; - lockdep_assert_held(&tl->gt->i915->drm.struct_mutex); /* lazy rq refs */ - err = intel_timeline_pin(tl); if (err) { rq = ERR_PTR(err); @@ -461,10 +462,14 @@ tl_write(struct intel_timeline *tl, struct intel_engine_cs *engine, u32 value) if (IS_ERR(rq)) goto out_unpin; + i915_request_get(rq); + err = emit_ggtt_store_dw(rq, tl->hwsp_offset, value); i915_request_add(rq); - if (err) + if (err) { + i915_request_put(rq); rq = ERR_PTR(err); + } out_unpin: intel_timeline_unpin(tl); @@ -475,11 +480,11 @@ out: } static struct intel_timeline * -checked_intel_timeline_create(struct drm_i915_private *i915) +checked_intel_timeline_create(struct intel_gt *gt) { struct intel_timeline *tl; - tl = intel_timeline_create(&i915->gt, NULL); + tl = intel_timeline_create(gt, NULL); if (IS_ERR(tl)) return tl; @@ -496,11 +501,10 @@ checked_intel_timeline_create(struct drm_i915_private *i915) static int live_hwsp_engine(void *arg) { #define NUM_TIMELINES 4096 - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_timeline **timelines; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; unsigned long count, n; int err = 0; @@ -515,37 +519,40 @@ static int live_hwsp_engine(void *arg) if (!timelines) return -ENOMEM; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - count = 0; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { if (!intel_engine_can_store_dword(engine)) continue; + intel_engine_pm_get(engine); + for (n = 0; n < NUM_TIMELINES; n++) { struct intel_timeline *tl; struct i915_request *rq; - tl = checked_intel_timeline_create(i915); + tl = checked_intel_timeline_create(gt); if (IS_ERR(tl)) { err = PTR_ERR(tl); - goto out; + break; } rq = tl_write(tl, engine, count); if (IS_ERR(rq)) { intel_timeline_put(tl); err = PTR_ERR(rq); - goto out; + break; } timelines[count++] = tl; + i915_request_put(rq); } + + intel_engine_pm_put(engine); + if (err) + break; } -out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) err = -EIO; for (n = 0; n < count; n++) { @@ -559,11 +566,7 @@ out: intel_timeline_put(tl); } - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - kvfree(timelines); - return err; #undef NUM_TIMELINES } @@ -571,11 +574,10 @@ out: static int live_hwsp_alternate(void *arg) { #define NUM_TIMELINES 4096 - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_timeline **timelines; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; unsigned long count, n; int err = 0; @@ -591,25 +593,25 @@ static int live_hwsp_alternate(void *arg) if (!timelines) return -ENOMEM; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - count = 0; for (n = 0; n < NUM_TIMELINES; n++) { - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct intel_timeline *tl; struct i915_request *rq; if (!intel_engine_can_store_dword(engine)) continue; - tl = checked_intel_timeline_create(i915); + tl = checked_intel_timeline_create(gt); if (IS_ERR(tl)) { + intel_engine_pm_put(engine); err = PTR_ERR(tl); goto out; } + intel_engine_pm_get(engine); rq = tl_write(tl, engine, count); + intel_engine_pm_put(engine); if (IS_ERR(rq)) { intel_timeline_put(tl); err = PTR_ERR(rq); @@ -617,11 +619,12 @@ static int live_hwsp_alternate(void *arg) } timelines[count++] = tl; + i915_request_put(rq); } } out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) err = -EIO; for (n = 0; n < count; n++) { @@ -635,22 +638,17 @@ out: intel_timeline_put(tl); } - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - kvfree(timelines); - return err; #undef NUM_TIMELINES } static int live_hwsp_wrap(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *engine; struct intel_timeline *tl; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = 0; /* @@ -658,14 +656,10 @@ static int live_hwsp_wrap(void *arg) * foreign GPU references. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + tl = intel_timeline_create(gt, NULL); + if (IS_ERR(tl)) + return PTR_ERR(tl); - tl = intel_timeline_create(&i915->gt, NULL); - if (IS_ERR(tl)) { - err = PTR_ERR(tl); - goto out_rpm; - } if (!tl->has_initial_breadcrumb || !tl->hwsp_cacheline) goto out_free; @@ -673,7 +667,7 @@ static int live_hwsp_wrap(void *arg) if (err) goto out_free; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { const u32 *hwsp_seqno[2]; struct i915_request *rq; u32 seqno[2]; @@ -681,7 +675,9 @@ static int live_hwsp_wrap(void *arg) if (!intel_engine_can_store_dword(engine)) continue; + intel_engine_pm_get(engine); rq = i915_request_create(engine->kernel_context); + intel_engine_pm_put(engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); goto out; @@ -743,29 +739,24 @@ static int live_hwsp_wrap(void *arg) goto out; } - i915_retire_requests(i915); /* recycle HWSP */ + intel_gt_retire_requests(gt); /* recycle HWSP */ } out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) err = -EIO; intel_timeline_unpin(tl); out_free: intel_timeline_put(tl); -out_rpm: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - return err; } static int live_hwsp_recycle(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; unsigned long count; int err = 0; @@ -775,38 +766,38 @@ static int live_hwsp_recycle(void *arg) * want to confuse ourselves or the GPU. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - count = 0; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { IGT_TIMEOUT(end_time); if (!intel_engine_can_store_dword(engine)) continue; + intel_engine_pm_get(engine); + do { struct intel_timeline *tl; struct i915_request *rq; - tl = checked_intel_timeline_create(i915); + tl = checked_intel_timeline_create(gt); if (IS_ERR(tl)) { err = PTR_ERR(tl); - goto out; + break; } rq = tl_write(tl, engine, count); if (IS_ERR(rq)) { intel_timeline_put(tl); err = PTR_ERR(rq); - goto out; + break; } if (i915_request_wait(rq, 0, HZ / 5) < 0) { pr_err("Wait for timeline writes timed out!\n"); + i915_request_put(rq); intel_timeline_put(tl); err = -EIO; - goto out; + break; } if (*tl->hwsp_seqno != count) { @@ -815,17 +806,18 @@ static int live_hwsp_recycle(void *arg) err = -EINVAL; } + i915_request_put(rq); intel_timeline_put(tl); count++; if (err) - goto out; + break; } while (!__igt_timeout(end_time, NULL)); - } -out: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); + intel_engine_pm_put(engine); + if (err) + break; + } return err; } @@ -842,5 +834,5 @@ int intel_timeline_live_selftests(struct drm_i915_private *i915) if (intel_gt_is_wedged(&i915->gt)) return 0; - return i915_live_subtests(tests, i915); + return intel_gt_live_subtests(tests, &i915->gt); } diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c index d06d68ac2a3b..abce6e4ec9c0 100644 --- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c +++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c @@ -33,8 +33,32 @@ struct wa_lists { } engine[I915_NUM_ENGINES]; }; +static int request_add_sync(struct i915_request *rq, int err) +{ + i915_request_get(rq); + i915_request_add(rq); + if (i915_request_wait(rq, 0, HZ / 5) < 0) + err = -EIO; + i915_request_put(rq); + + return err; +} + +static int request_add_spin(struct i915_request *rq, struct igt_spinner *spin) +{ + int err = 0; + + i915_request_get(rq); + i915_request_add(rq); + if (spin && !igt_wait_for_spinner(spin, rq)) + err = -ETIMEDOUT; + i915_request_put(rq); + + return err; +} + static void -reference_lists_init(struct drm_i915_private *i915, struct wa_lists *lists) +reference_lists_init(struct intel_gt *gt, struct wa_lists *lists) { struct intel_engine_cs *engine; enum intel_engine_id id; @@ -42,10 +66,10 @@ reference_lists_init(struct drm_i915_private *i915, struct wa_lists *lists) memset(lists, 0, sizeof(*lists)); wa_init_start(&lists->gt_wa_list, "GT_REF", "global"); - gt_init_workarounds(i915, &lists->gt_wa_list); + gt_init_workarounds(gt->i915, &lists->gt_wa_list); wa_init_finish(&lists->gt_wa_list); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct i915_wa_list *wal = &lists->engine[id].wa_list; wa_init_start(wal, "REF", engine->name); @@ -59,12 +83,12 @@ reference_lists_init(struct drm_i915_private *i915, struct wa_lists *lists) } static void -reference_lists_fini(struct drm_i915_private *i915, struct wa_lists *lists) +reference_lists_fini(struct intel_gt *gt, struct wa_lists *lists) { struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) + for_each_engine(engine, gt, id) intel_wa_list_free(&lists->engine[id].wa_list); intel_wa_list_free(&lists->gt_wa_list); @@ -191,10 +215,10 @@ static int check_whitelist(struct i915_gem_context *ctx, err = 0; i915_gem_object_lock(results); - intel_wedge_on_timeout(&wedge, &ctx->i915->gt, HZ / 5) /* safety net! */ + intel_wedge_on_timeout(&wedge, engine->gt, HZ / 5) /* safety net! */ err = i915_gem_object_set_to_cpu_domain(results, false); i915_gem_object_unlock(results); - if (intel_gt_is_wedged(&ctx->i915->gt)) + if (intel_gt_is_wedged(engine->gt)) err = -EIO; if (err) goto out_put; @@ -243,7 +267,6 @@ switch_to_scratch_context(struct intel_engine_cs *engine, struct i915_gem_context *ctx; struct intel_context *ce; struct i915_request *rq; - intel_wakeref_t wakeref; int err = 0; ctx = kernel_context(engine->i915); @@ -255,12 +278,9 @@ switch_to_scratch_context(struct intel_engine_cs *engine, ce = i915_gem_context_get_engine(ctx, engine->legacy_idx); GEM_BUG_ON(IS_ERR(ce)); - rq = ERR_PTR(-ENODEV); - with_intel_runtime_pm(&engine->i915->runtime_pm, wakeref) - rq = igt_spinner_create_request(spin, ce, MI_NOOP); + rq = igt_spinner_create_request(spin, ce, MI_NOOP); intel_context_put(ce); - kernel_context_close(ctx); if (IS_ERR(rq)) { spin = NULL; @@ -268,17 +288,12 @@ switch_to_scratch_context(struct intel_engine_cs *engine, goto err; } - i915_request_add(rq); - - if (spin && !igt_wait_for_spinner(spin, rq)) { - pr_err("Spinner failed to start\n"); - err = -ETIMEDOUT; - } - + err = request_add_spin(rq, spin); err: if (err && spin) igt_spinner_end(spin); + kernel_context_close(ctx); return err; } @@ -313,7 +328,7 @@ static int check_whitelist_across_reset(struct intel_engine_cs *engine, if (err) goto out_spin; - with_intel_runtime_pm(&i915->runtime_pm, wakeref) + with_intel_runtime_pm(engine->uncore->rpm, wakeref) err = reset(engine); igt_spinner_end(&spin); @@ -355,6 +370,7 @@ out_ctx: static struct i915_vma *create_batch(struct i915_gem_context *ctx) { struct drm_i915_gem_object *obj; + struct i915_address_space *vm; struct i915_vma *vma; int err; @@ -362,7 +378,9 @@ static struct i915_vma *create_batch(struct i915_gem_context *ctx) if (IS_ERR(obj)) return ERR_CAST(obj); - vma = i915_vma_instance(obj, ctx->vm, NULL); + vm = i915_gem_context_get_vm_rcu(ctx); + vma = i915_vma_instance(obj, vm, NULL); + i915_vm_put(vm); if (IS_ERR(vma)) { err = PTR_ERR(vma); goto err_obj; @@ -463,12 +481,15 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx, 0xffff00ff, 0xffffffff, }; + struct i915_address_space *vm; struct i915_vma *scratch; struct i915_vma *batch; int err = 0, i, v; u32 *cs, *results; - scratch = create_scratch(ctx->vm, 2 * ARRAY_SIZE(values) + 1); + vm = i915_gem_context_get_vm_rcu(ctx); + scratch = create_scratch(vm, 2 * ARRAY_SIZE(values) + 1); + i915_vm_put(vm); if (IS_ERR(scratch)) return PTR_ERR(scratch); @@ -492,6 +513,9 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx, ro_reg = ro_register(reg); + /* Clear non priv flags */ + reg &= RING_FORCE_TO_NONPRIV_ADDRESS_MASK; + srm = MI_STORE_REGISTER_MEM; lrm = MI_LOAD_REGISTER_MEM; if (INTEL_GEN(ctx->i915) >= 8) @@ -565,6 +589,14 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx, goto err_request; } + i915_vma_lock(batch); + err = i915_request_await_object(rq, batch->obj, false); + if (err == 0) + err = i915_vma_move_to_active(batch, rq, 0); + i915_vma_unlock(batch); + if (err) + goto err_request; + err = engine->emit_bb_start(rq, batch->node.start, PAGE_SIZE, 0); @@ -572,15 +604,11 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx, goto err_request; err_request: - i915_request_add(rq); - if (err) - goto out_batch; - - if (i915_request_wait(rq, 0, HZ / 5) < 0) { + err = request_add_sync(rq, err); + if (err) { pr_err("%s: Futzing %x timedout; cancelling test\n", engine->name, reg); - intel_gt_set_wedged(&ctx->i915->gt); - err = -EIO; + intel_gt_set_wedged(engine->gt); goto out_batch; } @@ -668,7 +696,7 @@ out_unpin: break; } - if (igt_flush_test(ctx->i915, I915_WAIT_LOCKED)) + if (igt_flush_test(ctx->i915)) err = -EIO; out_batch: i915_vma_unpin_and_release(&batch, 0); @@ -679,36 +707,29 @@ out_scratch: static int live_dirty_whitelist(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_engine_cs *engine; struct i915_gem_context *ctx; enum intel_engine_id id; - intel_wakeref_t wakeref; struct drm_file *file; int err = 0; /* Can the user write to the whitelisted registers? */ - if (INTEL_GEN(i915) < 7) /* minimum requirement for LRI, SRM, LRM */ + if (INTEL_GEN(gt->i915) < 7) /* minimum requirement for LRI, SRM, LRM */ return 0; - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + file = mock_file(gt->i915); + if (IS_ERR(file)) + return PTR_ERR(file); - mutex_unlock(&i915->drm.struct_mutex); - file = mock_file(i915); - mutex_lock(&i915->drm.struct_mutex); - if (IS_ERR(file)) { - err = PTR_ERR(file); - goto out_rpm; - } - - ctx = live_context(i915, file); + ctx = live_context(gt->i915, file); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto out_file; } - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { if (engine->whitelist.count == 0) continue; @@ -718,45 +739,43 @@ static int live_dirty_whitelist(void *arg) } out_file: - mutex_unlock(&i915->drm.struct_mutex); - mock_file_free(i915, file); - mutex_lock(&i915->drm.struct_mutex); -out_rpm: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); + mock_file_free(gt->i915, file); return err; } static int live_reset_whitelist(void *arg) { - struct drm_i915_private *i915 = arg; - struct intel_engine_cs *engine = i915->engine[RCS0]; + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; int err = 0; /* If we reset the gpu, we should not lose the RING_NONPRIV */ + igt_global_reset_lock(gt); - if (!engine || engine->whitelist.count == 0) - return 0; - - igt_global_reset_lock(&i915->gt); + for_each_engine(engine, gt, id) { + if (engine->whitelist.count == 0) + continue; - if (intel_has_reset_engine(i915)) { - err = check_whitelist_across_reset(engine, - do_engine_reset, - "engine"); - if (err) - goto out; - } + if (intel_has_reset_engine(gt)) { + err = check_whitelist_across_reset(engine, + do_engine_reset, + "engine"); + if (err) + goto out; + } - if (intel_has_gpu_reset(i915)) { - err = check_whitelist_across_reset(engine, - do_device_reset, - "device"); - if (err) - goto out; + if (intel_has_gpu_reset(gt)) { + err = check_whitelist_across_reset(engine, + do_device_reset, + "device"); + if (err) + goto out; + } } out: - igt_global_reset_unlock(&i915->gt); + igt_global_reset_unlock(gt); return err; } @@ -772,6 +791,14 @@ static int read_whitelisted_registers(struct i915_gem_context *ctx, if (IS_ERR(rq)) return PTR_ERR(rq); + i915_vma_lock(results); + err = i915_request_await_object(rq, results->obj, true); + if (err == 0) + err = i915_vma_move_to_active(results, rq, EXEC_OBJECT_WRITE); + i915_vma_unlock(results); + if (err) + goto err_req; + srm = MI_STORE_REGISTER_MEM; if (INTEL_GEN(ctx->i915) >= 8) srm++; @@ -786,8 +813,8 @@ static int read_whitelisted_registers(struct i915_gem_context *ctx, u64 offset = results->node.start + sizeof(u32) * i; u32 reg = i915_mmio_reg_offset(engine->whitelist.list[i].reg); - /* Clear access permission field */ - reg &= ~RING_FORCE_TO_NONPRIV_ACCESS_MASK; + /* Clear non priv flags */ + reg &= RING_FORCE_TO_NONPRIV_ADDRESS_MASK; *cs++ = srm; *cs++ = reg; @@ -797,12 +824,7 @@ static int read_whitelisted_registers(struct i915_gem_context *ctx, intel_ring_advance(rq, cs); err_req: - i915_request_add(rq); - - if (i915_request_wait(rq, 0, HZ / 5) < 0) - err = -EIO; - - return err; + return request_add_sync(rq, err); } static int scrub_whitelisted_registers(struct i915_gem_context *ctx, @@ -830,6 +852,9 @@ static int scrub_whitelisted_registers(struct i915_gem_context *ctx, if (ro_register(reg)) continue; + /* Clear non priv flags */ + reg &= RING_FORCE_TO_NONPRIV_ADDRESS_MASK; + *cs++ = reg; *cs++ = 0xffffffff; } @@ -850,13 +875,19 @@ static int scrub_whitelisted_registers(struct i915_gem_context *ctx, goto err_request; } + i915_vma_lock(batch); + err = i915_request_await_object(rq, batch->obj, false); + if (err == 0) + err = i915_vma_move_to_active(batch, rq, 0); + i915_vma_unlock(batch); + if (err) + goto err_request; + /* Perform the writes from an unprivileged "user" batch */ err = engine->emit_bb_start(rq, batch->node.start, 0, 0); err_request: - i915_request_add(rq); - if (i915_request_wait(rq, 0, HZ / 5) < 0) - err = -EIO; + err = request_add_sync(rq, err); err_unpin: i915_gem_object_unpin_map(batch->obj); @@ -973,7 +1004,7 @@ err_a: static int live_isolated_whitelist(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct { struct i915_gem_context *ctx; struct i915_vma *scratch[2]; @@ -987,40 +1018,46 @@ static int live_isolated_whitelist(void *arg) * invisible to a second context. */ - if (!intel_engines_has_context_isolation(i915)) - return 0; - - if (!i915->kernel_context->vm) + if (!intel_engines_has_context_isolation(gt->i915)) return 0; for (i = 0; i < ARRAY_SIZE(client); i++) { + struct i915_address_space *vm; struct i915_gem_context *c; - c = kernel_context(i915); + c = kernel_context(gt->i915); if (IS_ERR(c)) { err = PTR_ERR(c); goto err; } - client[i].scratch[0] = create_scratch(c->vm, 1024); + vm = i915_gem_context_get_vm_rcu(c); + + client[i].scratch[0] = create_scratch(vm, 1024); if (IS_ERR(client[i].scratch[0])) { err = PTR_ERR(client[i].scratch[0]); + i915_vm_put(vm); kernel_context_close(c); goto err; } - client[i].scratch[1] = create_scratch(c->vm, 1024); + client[i].scratch[1] = create_scratch(vm, 1024); if (IS_ERR(client[i].scratch[1])) { err = PTR_ERR(client[i].scratch[1]); i915_vma_unpin_and_release(&client[i].scratch[0], 0); + i915_vm_put(vm); kernel_context_close(c); goto err; } client[i].ctx = c; + i915_vm_put(vm); } - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { + if (!engine->kernel_context->vm) + continue; + if (!whitelist_writable_count(engine)) continue; @@ -1074,7 +1111,7 @@ err: kernel_context_close(client[i].ctx); } - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) err = -EIO; return err; @@ -1109,16 +1146,16 @@ verify_wa_lists(struct i915_gem_context *ctx, struct wa_lists *lists, static int live_gpu_reset_workarounds(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_gem_context *ctx; intel_wakeref_t wakeref; struct wa_lists lists; bool ok; - if (!intel_has_gpu_reset(i915)) + if (!intel_has_gpu_reset(gt)) return 0; - ctx = kernel_context(i915); + ctx = kernel_context(gt->i915); if (IS_ERR(ctx)) return PTR_ERR(ctx); @@ -1126,25 +1163,25 @@ live_gpu_reset_workarounds(void *arg) pr_info("Verifying after GPU reset...\n"); - igt_global_reset_lock(&i915->gt); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + igt_global_reset_lock(gt); + wakeref = intel_runtime_pm_get(gt->uncore->rpm); - reference_lists_init(i915, &lists); + reference_lists_init(gt, &lists); ok = verify_wa_lists(ctx, &lists, "before reset"); if (!ok) goto out; - intel_gt_reset(&i915->gt, ALL_ENGINES, "live_workarounds"); + intel_gt_reset(gt, ALL_ENGINES, "live_workarounds"); ok = verify_wa_lists(ctx, &lists, "after reset"); out: i915_gem_context_unlock_engines(ctx); kernel_context_close(ctx); - reference_lists_fini(i915, &lists); - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - igt_global_reset_unlock(&i915->gt); + reference_lists_fini(gt, &lists); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); + igt_global_reset_unlock(gt); return ok ? 0 : -ESRCH; } @@ -1152,7 +1189,7 @@ out: static int live_engine_reset_workarounds(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_gem_engines_iter it; struct i915_gem_context *ctx; struct intel_context *ce; @@ -1162,17 +1199,17 @@ live_engine_reset_workarounds(void *arg) struct wa_lists lists; int ret = 0; - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(gt)) return 0; - ctx = kernel_context(i915); + ctx = kernel_context(gt->i915); if (IS_ERR(ctx)) return PTR_ERR(ctx); - igt_global_reset_lock(&i915->gt); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + igt_global_reset_lock(gt); + wakeref = intel_runtime_pm_get(gt->uncore->rpm); - reference_lists_init(i915, &lists); + reference_lists_init(gt, &lists); for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { struct intel_engine_cs *engine = ce->engine; @@ -1205,12 +1242,10 @@ live_engine_reset_workarounds(void *arg) goto err; } - i915_request_add(rq); - - if (!igt_wait_for_spinner(&spin, rq)) { + ret = request_add_spin(rq, &spin); + if (ret) { pr_err("Spinner failed to start\n"); igt_spinner_fini(&spin); - ret = -ETIMEDOUT; goto err; } @@ -1227,12 +1262,12 @@ live_engine_reset_workarounds(void *arg) } err: i915_gem_context_unlock_engines(ctx); - reference_lists_fini(i915, &lists); - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - igt_global_reset_unlock(&i915->gt); + reference_lists_fini(gt, &lists); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); + igt_global_reset_unlock(gt); kernel_context_close(ctx); - igt_flush_test(i915, I915_WAIT_LOCKED); + igt_flush_test(gt->i915); return ret; } @@ -1246,14 +1281,9 @@ int intel_workarounds_live_selftests(struct drm_i915_private *i915) SUBTEST(live_gpu_reset_workarounds), SUBTEST(live_engine_reset_workarounds), }; - int err; if (intel_gt_is_wedged(&i915->gt)) return 0; - mutex_lock(&i915->drm.struct_mutex); - err = i915_subtests(tests, i915); - mutex_unlock(&i915->drm.struct_mutex); - - return err; + return intel_gt_live_subtests(tests, &i915->gt); } diff --git a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c index 598170efcaf6..2a77c051f36a 100644 --- a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c +++ b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c @@ -15,7 +15,7 @@ void mock_timeline_init(struct intel_timeline *timeline, u64 context) mutex_init(&timeline->mutex); - INIT_ACTIVE_REQUEST(&timeline->last_request, &timeline->mutex); + INIT_ACTIVE_FENCE(&timeline->last_request, &timeline->mutex); INIT_LIST_HEAD(&timeline->requests); i915_syncmap_init(&timeline->sync); diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc.c b/drivers/gpu/drm/i915/gt/uc/intel_guc.c index 249c747e9756..3ee4a4e7689d 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc.c @@ -4,11 +4,34 @@ */ #include "gt/intel_gt.h" +#include "gt/intel_gt_irq.h" +#include "gt/intel_gt_pm_irq.h" #include "intel_guc.h" #include "intel_guc_ads.h" #include "intel_guc_submission.h" #include "i915_drv.h" +/** + * DOC: GuC + * + * The GuC is a microcontroller inside the GT HW, introduced in gen9. The GuC is + * designed to offload some of the functionality usually performed by the host + * driver; currently the main operations it can take care of are: + * + * - Authentication of the HuC, which is required to fully enable HuC usage. + * - Low latency graphics context scheduling (a.k.a. GuC submission). + * - GT Power management. + * + * The enable_guc module parameter can be used to select which of those + * operations to enable within GuC. Note that not all the operations are + * supported on all gen9+ platforms. + * + * Enabling the GuC is not mandatory and therefore the firmware is only loaded + * if at least one of the operations is selected. However, not loading the GuC + * might result in the loss of some features that do require the GuC (currently + * just the HuC, but more are expected to land in the future). + */ + static void gen8_guc_raise_irq(struct intel_guc *guc) { struct intel_gt *gt = guc_to_gt(guc); @@ -56,6 +79,93 @@ void intel_guc_init_send_regs(struct intel_guc *guc) guc->send_regs.fw_domains = fw_domains; } +static void gen9_reset_guc_interrupts(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + assert_rpm_wakelock_held(>->i915->runtime_pm); + + spin_lock_irq(>->irq_lock); + gen6_gt_pm_reset_iir(gt, gt->pm_guc_events); + spin_unlock_irq(>->irq_lock); +} + +static void gen9_enable_guc_interrupts(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + assert_rpm_wakelock_held(>->i915->runtime_pm); + + spin_lock_irq(>->irq_lock); + if (!guc->interrupts.enabled) { + WARN_ON_ONCE(intel_uncore_read(gt->uncore, GEN8_GT_IIR(2)) & + gt->pm_guc_events); + guc->interrupts.enabled = true; + gen6_gt_pm_enable_irq(gt, gt->pm_guc_events); + } + spin_unlock_irq(>->irq_lock); +} + +static void gen9_disable_guc_interrupts(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + assert_rpm_wakelock_held(>->i915->runtime_pm); + + spin_lock_irq(>->irq_lock); + guc->interrupts.enabled = false; + + gen6_gt_pm_disable_irq(gt, gt->pm_guc_events); + + spin_unlock_irq(>->irq_lock); + intel_synchronize_irq(gt->i915); + + gen9_reset_guc_interrupts(guc); +} + +static void gen11_reset_guc_interrupts(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + spin_lock_irq(>->irq_lock); + gen11_gt_reset_one_iir(gt, 0, GEN11_GUC); + spin_unlock_irq(>->irq_lock); +} + +static void gen11_enable_guc_interrupts(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + spin_lock_irq(>->irq_lock); + if (!guc->interrupts.enabled) { + u32 events = REG_FIELD_PREP(ENGINE1_MASK, GUC_INTR_GUC2HOST); + + WARN_ON_ONCE(gen11_gt_reset_one_iir(gt, 0, GEN11_GUC)); + intel_uncore_write(gt->uncore, + GEN11_GUC_SG_INTR_ENABLE, events); + intel_uncore_write(gt->uncore, + GEN11_GUC_SG_INTR_MASK, ~events); + guc->interrupts.enabled = true; + } + spin_unlock_irq(>->irq_lock); +} + +static void gen11_disable_guc_interrupts(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + spin_lock_irq(>->irq_lock); + guc->interrupts.enabled = false; + + intel_uncore_write(gt->uncore, GEN11_GUC_SG_INTR_MASK, ~0); + intel_uncore_write(gt->uncore, GEN11_GUC_SG_INTR_ENABLE, 0); + + spin_unlock_irq(>->irq_lock); + intel_synchronize_irq(gt->i915); + + gen11_reset_guc_interrupts(guc); +} + void intel_guc_init_early(struct intel_guc *guc) { struct drm_i915_private *i915 = guc_to_gt(guc)->i915; @@ -82,32 +192,6 @@ void intel_guc_init_early(struct intel_guc *guc) } } -static int guc_shared_data_create(struct intel_guc *guc) -{ - struct i915_vma *vma; - void *vaddr; - - vma = intel_guc_allocate_vma(guc, PAGE_SIZE); - if (IS_ERR(vma)) - return PTR_ERR(vma); - - vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); - if (IS_ERR(vaddr)) { - i915_vma_unpin_and_release(&vma, 0); - return PTR_ERR(vaddr); - } - - guc->shared_data = vma; - guc->shared_data_vaddr = vaddr; - - return 0; -} - -static void guc_shared_data_destroy(struct intel_guc *guc) -{ - i915_vma_unpin_and_release(&guc->shared_data, I915_VMA_RELEASE_MAP); -} - static u32 guc_ctl_debug_flags(struct intel_guc *guc) { u32 level = intel_guc_log_get_level(&guc->log); @@ -254,14 +338,9 @@ int intel_guc_init(struct intel_guc *guc) if (ret) goto err_fetch; - ret = guc_shared_data_create(guc); - if (ret) - goto err_fw; - GEM_BUG_ON(!guc->shared_data); - ret = intel_guc_log_create(&guc->log); if (ret) - goto err_shared; + goto err_fw; ret = intel_guc_ads_create(guc); if (ret) @@ -296,8 +375,6 @@ err_ads: intel_guc_ads_destroy(guc); err_log: intel_guc_log_destroy(&guc->log); -err_shared: - guc_shared_data_destroy(guc); err_fw: intel_uc_fw_fini(&guc->fw); err_fetch: @@ -322,7 +399,6 @@ void intel_guc_fini(struct intel_guc *guc) intel_guc_ads_destroy(guc); intel_guc_log_destroy(&guc->log); - guc_shared_data_destroy(guc); intel_uc_fw_fini(&guc->fw); intel_uc_fw_cleanup_fetch(&guc->fw); } @@ -478,6 +554,13 @@ int intel_guc_suspend(struct intel_guc *guc) }; /* + * If GuC communication is enabled but submission is not supported, + * we do not need to suspend the GuC. + */ + if (!intel_guc_submission_is_enabled(guc)) + return 0; + + /* * The ENTER_S_STATE action queues the save/restore operation in GuC FW * and then returns, so waiting on the H2G is not enough to guarantee * GuC is done. When all the processing is done, GuC writes @@ -518,19 +601,9 @@ int intel_guc_suspend(struct intel_guc *guc) int intel_guc_reset_engine(struct intel_guc *guc, struct intel_engine_cs *engine) { - u32 data[7]; - - GEM_BUG_ON(!guc->execbuf_client); - - data[0] = INTEL_GUC_ACTION_REQUEST_ENGINE_RESET; - data[1] = engine->guc_id; - data[2] = 0; - data[3] = 0; - data[4] = 0; - data[5] = guc->execbuf_client->stage_id; - data[6] = intel_guc_ggtt_offset(guc, guc->shared_data); + /* XXX: to be implemented with submission interface rework */ - return intel_guc_send(guc, data, ARRAY_SIZE(data)); + return -ENODEV; } /** @@ -544,13 +617,27 @@ int intel_guc_resume(struct intel_guc *guc) GUC_POWER_D0, }; + /* + * If GuC communication is enabled but submission is not supported, + * we do not need to resume the GuC but we do need to enable the + * GuC communication on resume (above). + */ + if (!intel_guc_submission_is_enabled(guc)) + return 0; + return intel_guc_send(guc, action, ARRAY_SIZE(action)); } /** - * DOC: GuC Address Space + * DOC: GuC Memory Management * - * The layout of GuC address space is shown below: + * GuC can't allocate any memory for its own usage, so all the allocations must + * be handled by the host driver. GuC accesses the memory via the GGTT, with the + * exception of the top and bottom parts of the 4GB address space, which are + * instead re-mapped by the GuC HW to memory location of the FW itself (WOPCM) + * or other parts of the HW. The driver must take care not to place objects that + * the GuC is going to access in these reserved ranges. The layout of the GuC + * address space is shown below: * * :: * diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc.h b/drivers/gpu/drm/i915/gt/uc/intel_guc.h index 2b2f046d3cc3..e6400204a2bd 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc.h +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc.h @@ -47,8 +47,6 @@ struct intel_guc { struct i915_vma *stage_desc_pool; void *stage_desc_pool_vaddr; struct ida stage_ids; - struct i915_vma *shared_data; - void *shared_data_vaddr; struct intel_guc_client *execbuf_client; diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h index 1d3cdd67ca2f..a26a85d50209 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h @@ -548,6 +548,7 @@ enum intel_guc_action { INTEL_GUC_ACTION_ALLOCATE_DOORBELL = 0x10, INTEL_GUC_ACTION_DEALLOCATE_DOORBELL = 0x20, INTEL_GUC_ACTION_LOG_BUFFER_FILE_FLUSH_COMPLETE = 0x30, + INTEL_GUC_ACTION_UK_LOG_ENABLE_LOGGING = 0x40, INTEL_GUC_ACTION_FORCE_LOG_BUFFER_FLUSH = 0x302, INTEL_GUC_ACTION_ENTER_S_STATE = 0x501, INTEL_GUC_ACTION_EXIT_S_STATE = 0x502, @@ -556,7 +557,6 @@ enum intel_guc_action { INTEL_GUC_ACTION_AUTHENTICATE_HUC = 0x4000, INTEL_GUC_ACTION_REGISTER_COMMAND_TRANSPORT_BUFFER = 0x4505, INTEL_GUC_ACTION_DEREGISTER_COMMAND_TRANSPORT_BUFFER = 0x4506, - INTEL_GUC_ACTION_UK_LOG_ENABLE_LOGGING = 0x0E000, INTEL_GUC_ACTION_LIMIT }; diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c index 36332064de9c..caed0d57e704 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c @@ -226,7 +226,7 @@ static void guc_read_update_log_buffer(struct intel_guc_log *log) mutex_lock(&log->relay.lock); - if (WARN_ON(!intel_guc_log_relay_enabled(log))) + if (WARN_ON(!intel_guc_log_relay_created(log))) goto out_unlock; /* Get the pointer to shared GuC log buffer */ @@ -361,6 +361,7 @@ void intel_guc_log_init_early(struct intel_guc_log *log) { mutex_init(&log->relay.lock); INIT_WORK(&log->relay.flush_work, capture_logs_work); + log->relay.started = false; } static int guc_log_relay_create(struct intel_guc_log *log) @@ -546,7 +547,7 @@ out_unlock: return ret; } -bool intel_guc_log_relay_enabled(const struct intel_guc_log *log) +bool intel_guc_log_relay_created(const struct intel_guc_log *log) { return log->relay.buf_addr; } @@ -560,7 +561,7 @@ int intel_guc_log_relay_open(struct intel_guc_log *log) mutex_lock(&log->relay.lock); - if (intel_guc_log_relay_enabled(log)) { + if (intel_guc_log_relay_created(log)) { ret = -EEXIST; goto out_unlock; } @@ -585,6 +586,21 @@ int intel_guc_log_relay_open(struct intel_guc_log *log) mutex_unlock(&log->relay.lock); + return 0; + +out_relay: + guc_log_relay_destroy(log); +out_unlock: + mutex_unlock(&log->relay.lock); + + return ret; +} + +int intel_guc_log_relay_start(struct intel_guc_log *log) +{ + if (log->relay.started) + return -EEXIST; + guc_log_enable_flush_events(log); /* @@ -594,47 +610,59 @@ int intel_guc_log_relay_open(struct intel_guc_log *log) */ queue_work(system_highpri_wq, &log->relay.flush_work); - return 0; + log->relay.started = true; -out_relay: - guc_log_relay_destroy(log); -out_unlock: - mutex_unlock(&log->relay.lock); - - return ret; + return 0; } void intel_guc_log_relay_flush(struct intel_guc_log *log) { struct intel_guc *guc = log_to_guc(log); - struct drm_i915_private *i915 = guc_to_gt(guc)->i915; intel_wakeref_t wakeref; + if (!log->relay.started) + return; + /* * Before initiating the forceful flush, wait for any pending/ongoing * flush to complete otherwise forceful flush may not actually happen. */ flush_work(&log->relay.flush_work); - with_intel_runtime_pm(&i915->runtime_pm, wakeref) + with_intel_runtime_pm(guc_to_gt(guc)->uncore->rpm, wakeref) guc_action_flush_log(guc); /* GuC would have updated log buffer by now, so capture it */ guc_log_capture_logs(log); } -void intel_guc_log_relay_close(struct intel_guc_log *log) +/* + * Stops the relay log. Called from intel_guc_log_relay_close(), so no + * possibility of race with start/flush since relay_write cannot race + * relay_close. + */ +static void guc_log_relay_stop(struct intel_guc_log *log) { struct intel_guc *guc = log_to_guc(log); struct drm_i915_private *i915 = guc_to_gt(guc)->i915; + if (!log->relay.started) + return; + guc_log_disable_flush_events(log); intel_synchronize_irq(i915); flush_work(&log->relay.flush_work); + log->relay.started = false; +} + +void intel_guc_log_relay_close(struct intel_guc_log *log) +{ + guc_log_relay_stop(log); + mutex_lock(&log->relay.lock); - GEM_BUG_ON(!intel_guc_log_relay_enabled(log)); + GEM_BUG_ON(!intel_guc_log_relay_created(log)); guc_log_unmap(log); guc_log_relay_destroy(log); mutex_unlock(&log->relay.lock); diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_log.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.h index 6f764879acb1..c252c022c5fc 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_log.h +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.h @@ -47,6 +47,7 @@ struct intel_guc_log { struct i915_vma *vma; struct { void *buf_addr; + bool started; struct work_struct flush_work; struct rchan *channel; struct mutex lock; @@ -65,8 +66,9 @@ int intel_guc_log_create(struct intel_guc_log *log); void intel_guc_log_destroy(struct intel_guc_log *log); int intel_guc_log_set_level(struct intel_guc_log *log, u32 level); -bool intel_guc_log_relay_enabled(const struct intel_guc_log *log); +bool intel_guc_log_relay_created(const struct intel_guc_log *log); int intel_guc_log_relay_open(struct intel_guc_log *log); +int intel_guc_log_relay_start(struct intel_guc_log *log); void intel_guc_log_relay_flush(struct intel_guc_log *log); void intel_guc_log_relay_close(struct intel_guc_log *log); diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h index edf194d23c6b..1949346e714e 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h @@ -83,6 +83,9 @@ #define GEN8_GTCR _MMIO(0x4274) #define GEN8_GTCR_INVALIDATE (1<<0) +#define GEN12_GUC_TLB_INV_CR _MMIO(0xcee8) +#define GEN12_GUC_TLB_INV_CR_INVALIDATE (1 << 0) + #define GUC_ARAT_C6DIS _MMIO(0xA178) #define GUC_SHIM_CONTROL _MMIO(0xc064) diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c index f325d3dd564f..2498c55e0ea5 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c @@ -6,12 +6,13 @@ #include <linux/circ_buf.h> #include "gem/i915_gem_context.h" - #include "gt/intel_context.h" #include "gt/intel_engine_pm.h" #include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" #include "gt/intel_lrc_reg.h" +#include "gt/intel_ring.h" + #include "intel_guc_submission.h" #include "i915_drv.h" @@ -29,6 +30,12 @@ enum { /** * DOC: GuC-based command submission * + * IMPORTANT NOTE: GuC submission is currently not supported in i915. The GuC + * firmware is moving to an updated submission interface and we plan to + * turn submission back on when that lands. The below documentation (and related + * code) matches the old submission model and will be updated as part of the + * upgrade to the new flow. + * * GuC client: * A intel_guc_client refers to a submission path through GuC. Currently, there * is only one client, which is charged with all submissions to the GuC. This @@ -1004,7 +1011,7 @@ void intel_guc_submission_fini(struct intel_guc *guc) static void guc_interrupts_capture(struct intel_gt *gt) { - struct intel_rps *rps = >->i915->gt_pm.rps; + struct intel_rps *rps = >->rps; struct intel_uncore *uncore = gt->uncore; struct intel_engine_cs *engine; enum intel_engine_id id; @@ -1014,7 +1021,7 @@ static void guc_interrupts_capture(struct intel_gt *gt) * to GuC */ irqs = _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING); - for_each_engine(engine, gt->i915, id) + for_each_engine(engine, gt, id) ENGINE_WRITE(engine, RING_MODE_GEN7, irqs); /* route USER_INTERRUPT to Host, all others are sent to GuC. */ @@ -1050,7 +1057,7 @@ static void guc_interrupts_capture(struct intel_gt *gt) static void guc_interrupts_release(struct intel_gt *gt) { - struct intel_rps *rps = >->i915->gt_pm.rps; + struct intel_rps *rps = >->rps; struct intel_uncore *uncore = gt->uncore; struct intel_engine_cs *engine; enum intel_engine_id id; @@ -1062,7 +1069,7 @@ static void guc_interrupts_release(struct intel_gt *gt) */ irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER); irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING); - for_each_engine(engine, gt->i915, id) + for_each_engine(engine, gt, id) ENGINE_WRITE(engine, RING_MODE_GEN7, irqs); /* route all GT interrupts to the host */ @@ -1119,7 +1126,7 @@ int intel_guc_submission_enable(struct intel_guc *guc) enum intel_engine_id id; int err; - err = i915_inject_load_error(gt->i915, -ENXIO); + err = i915_inject_probe_error(gt->i915, -ENXIO); if (err) return err; @@ -1145,7 +1152,7 @@ int intel_guc_submission_enable(struct intel_guc *guc) /* Take over from manual control of ELSP (execlists) */ guc_interrupts_capture(gt); - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { engine->set_default_submission = guc_set_default_submission; engine->set_default_submission(engine); } diff --git a/drivers/gpu/drm/i915/gt/uc/intel_huc.c b/drivers/gpu/drm/i915/gt/uc/intel_huc.c index d4625c97b4f9..32a069841c14 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_huc.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_huc.c @@ -9,6 +9,34 @@ #include "intel_huc.h" #include "i915_drv.h" +/** + * DOC: HuC + * + * The HuC is a dedicated microcontroller for usage in media HEVC (High + * Efficiency Video Coding) operations. Userspace can directly use the firmware + * capabilities by adding HuC specific commands to batch buffers. + * + * The kernel driver is only responsible for loading the HuC firmware and + * triggering its security authentication, which is performed by the GuC. For + * The GuC to correctly perform the authentication, the HuC binary must be + * loaded before the GuC one. Loading the HuC is optional; however, not using + * the HuC might negatively impact power usage and/or performance of media + * workloads, depending on the use-cases. + * + * See https://github.com/intel/media-driver for the latest details on HuC + * functionality. + */ + +/** + * DOC: HuC Memory Management + * + * Similarly to the GuC, the HuC can't do any memory allocations on its own, + * with the difference being that the allocations for HuC usage are handled by + * the userspace driver instead of the kernel one. The HuC accesses the memory + * via the PPGTT belonging to the context loaded on the VCS executing the + * HuC-specific commands. + */ + void intel_huc_init_early(struct intel_huc *huc) { struct drm_i915_private *i915 = huc_to_gt(huc)->i915; @@ -35,7 +63,7 @@ static int intel_huc_rsa_data_create(struct intel_huc *huc) void *vaddr; int err; - err = i915_inject_load_error(gt->i915, -ENXIO); + err = i915_inject_probe_error(gt->i915, -ENXIO); if (err) return err; @@ -118,10 +146,9 @@ void intel_huc_fini(struct intel_huc *huc) * * Called after HuC and GuC firmware loading during intel_uc_init_hw(). * - * This function pins HuC firmware image object into GGTT. - * Then it invokes GuC action to authenticate passing the offset to RSA - * signature through intel_guc_auth_huc(). It then waits for 50ms for - * firmware verification ACK and unpins the object. + * This function invokes the GuC action to authenticate the HuC firmware, + * passing the offset of the RSA signature to intel_guc_auth_huc(). It then + * waits for up to 50ms for firmware verification ACK. */ int intel_huc_auth(struct intel_huc *huc) { @@ -134,7 +161,7 @@ int intel_huc_auth(struct intel_huc *huc) if (!intel_uc_fw_is_loaded(&huc->fw)) return -ENOEXEC; - ret = i915_inject_load_error(gt->i915, -ENXIO); + ret = i915_inject_probe_error(gt->i915, -ENXIO); if (ret) goto fail; @@ -185,7 +212,7 @@ int intel_huc_check_status(struct intel_huc *huc) if (!intel_huc_is_supported(huc)) return -ENODEV; - with_intel_runtime_pm(>->i915->runtime_pm, wakeref) + with_intel_runtime_pm(gt->uncore->rpm, wakeref) status = intel_uncore_read(gt->uncore, huc->status.reg); return (status & huc->status.mask) == huc->status.value; diff --git a/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.c b/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.c index 74602487ed67..d654340d4d03 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.c @@ -8,21 +8,6 @@ #include "i915_drv.h" /** - * DOC: HuC Firmware - * - * Motivation: - * GEN9 introduces a new dedicated firmware for usage in media HEVC (High - * Efficiency Video Coding) operations. Userspace can use the firmware - * capabilities by adding HuC specific commands to batch buffers. - * - * Implementation: - * The same firmware loader is used as the GuC. However, the actual - * loading to HW is deferred until GEM initialization is done. - * - * Note that HuC firmware loading must be done before GuC loading. - */ - -/** * intel_huc_fw_init_early() - initializes HuC firmware struct * @huc: intel_huc struct * diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc.c b/drivers/gpu/drm/i915/gt/uc/intel_uc.c index 71ee7ab035cc..629b19377a29 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_uc.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc.c @@ -20,7 +20,7 @@ static int __intel_uc_reset_hw(struct intel_uc *uc) int ret; u32 guc_status; - ret = i915_inject_load_error(gt->i915, -ENXIO); + ret = i915_inject_probe_error(gt->i915, -ENXIO); if (ret) return ret; @@ -197,7 +197,7 @@ static int guc_enable_communication(struct intel_guc *guc) GEM_BUG_ON(guc_communication_enabled(guc)); - ret = i915_inject_load_error(i915, -ENXIO); + ret = i915_inject_probe_error(i915, -ENXIO); if (ret) return ret; @@ -224,17 +224,7 @@ static int guc_enable_communication(struct intel_guc *guc) return 0; } -static void guc_stop_communication(struct intel_guc *guc) -{ - intel_guc_ct_stop(&guc->ct); - - guc->send = intel_guc_send_nop; - guc->handler = intel_guc_to_host_event_handler_nop; - - guc_clear_mmio_msg(guc); -} - -static void guc_disable_communication(struct intel_guc *guc) +static void __guc_stop_communication(struct intel_guc *guc) { /* * Events generated during or after CT disable are logged by guc in @@ -247,6 +237,20 @@ static void guc_disable_communication(struct intel_guc *guc) guc->send = intel_guc_send_nop; guc->handler = intel_guc_to_host_event_handler_nop; +} + +static void guc_stop_communication(struct intel_guc *guc) +{ + intel_guc_ct_stop(&guc->ct); + + __guc_stop_communication(guc); + + DRM_INFO("GuC communication stopped\n"); +} + +static void guc_disable_communication(struct intel_guc *guc) +{ + __guc_stop_communication(guc); intel_guc_ct_disable(&guc->ct); @@ -368,7 +372,7 @@ static int uc_init_wopcm(struct intel_uc *uc) GEM_BUG_ON(!(size & GUC_WOPCM_SIZE_MASK)); GEM_BUG_ON(size & ~GUC_WOPCM_SIZE_MASK); - err = i915_inject_load_error(gt->i915, -ENXIO); + err = i915_inject_probe_error(gt->i915, -ENXIO); if (err) return err; @@ -537,7 +541,9 @@ void intel_uc_fini_hw(struct intel_uc *uc) if (intel_uc_supports_guc_submission(uc)) intel_guc_submission_disable(guc); - guc_disable_communication(guc); + if (guc_communication_enabled(guc)) + guc_disable_communication(guc); + __uc_sanitize(uc); } @@ -581,7 +587,7 @@ void intel_uc_suspend(struct intel_uc *uc) if (!intel_guc_is_running(guc)) return; - with_intel_runtime_pm(&uc_to_gt(uc)->i915->runtime_pm, wakeref) + with_intel_runtime_pm(uc_to_gt(uc)->uncore->rpm, wakeref) intel_uc_runtime_suspend(uc); } diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c index bd22bf11adad..66a30ab7044a 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c @@ -37,27 +37,34 @@ void intel_uc_fw_change_status(struct intel_uc_fw *uc_fw, /* * List of required GuC and HuC binaries per-platform. * Must be ordered based on platform + revid, from newer to older. + * + * TGL 35.2 is interface-compatible with 33.0 for previous Gens. The deltas + * between 33.0 and 35.2 are only related to new additions to support new Gen12 + * features. */ #define INTEL_UC_FIRMWARE_DEFS(fw_def, guc_def, huc_def) \ - fw_def(ICELAKE, 0, guc_def(icl, 33, 0, 0), huc_def(icl, 8, 4, 3238)) \ - fw_def(COFFEELAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 02, 00, 1810)) \ - fw_def(GEMINILAKE, 0, guc_def(glk, 33, 0, 0), huc_def(glk, 03, 01, 2893)) \ - fw_def(KABYLAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 02, 00, 1810)) \ - fw_def(BROXTON, 0, guc_def(bxt, 33, 0, 0), huc_def(bxt, 01, 8, 2893)) \ - fw_def(SKYLAKE, 0, guc_def(skl, 33, 0, 0), huc_def(skl, 01, 07, 1398)) - -#define __MAKE_UC_FW_PATH(prefix_, name_, separator_, major_, minor_, patch_) \ + fw_def(TIGERLAKE, 0, guc_def(tgl, 35, 2, 0), huc_def(tgl, 7, 0, 3)) \ + fw_def(ELKHARTLAKE, 0, guc_def(ehl, 33, 0, 4), huc_def(ehl, 9, 0, 0)) \ + fw_def(ICELAKE, 0, guc_def(icl, 33, 0, 0), huc_def(icl, 9, 0, 0)) \ + fw_def(COFFEELAKE, 5, guc_def(cml, 33, 0, 0), huc_def(cml, 4, 0, 0)) \ + fw_def(COFFEELAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 4, 0, 0)) \ + fw_def(GEMINILAKE, 0, guc_def(glk, 33, 0, 0), huc_def(glk, 4, 0, 0)) \ + fw_def(KABYLAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 4, 0, 0)) \ + fw_def(BROXTON, 0, guc_def(bxt, 33, 0, 0), huc_def(bxt, 2, 0, 0)) \ + fw_def(SKYLAKE, 0, guc_def(skl, 33, 0, 0), huc_def(skl, 2, 0, 0)) + +#define __MAKE_UC_FW_PATH(prefix_, name_, major_, minor_, patch_) \ "i915/" \ __stringify(prefix_) name_ \ - __stringify(major_) separator_ \ - __stringify(minor_) separator_ \ + __stringify(major_) "." \ + __stringify(minor_) "." \ __stringify(patch_) ".bin" #define MAKE_GUC_FW_PATH(prefix_, major_, minor_, patch_) \ - __MAKE_UC_FW_PATH(prefix_, "_guc_", ".", major_, minor_, patch_) + __MAKE_UC_FW_PATH(prefix_, "_guc_", major_, minor_, patch_) #define MAKE_HUC_FW_PATH(prefix_, major_, minor_, bld_num_) \ - __MAKE_UC_FW_PATH(prefix_, "_huc_ver", "_", major_, minor_, bld_num_) + __MAKE_UC_FW_PATH(prefix_, "_huc_", major_, minor_, bld_num_) /* All blobs need to be declared via MODULE_FIRMWARE() */ #define INTEL_UC_MODULE_FW(platform_, revid_, guc_, huc_) \ @@ -218,29 +225,31 @@ static void __force_fw_fetch_failures(struct intel_uc_fw *uc_fw, { bool user = e == -EINVAL; - if (i915_inject_load_error(i915, e)) { + if (i915_inject_probe_error(i915, e)) { /* non-existing blob */ uc_fw->path = "<invalid>"; uc_fw->user_overridden = user; - } else if (i915_inject_load_error(i915, e)) { + } else if (i915_inject_probe_error(i915, e)) { /* require next major version */ uc_fw->major_ver_wanted += 1; uc_fw->minor_ver_wanted = 0; uc_fw->user_overridden = user; - } else if (i915_inject_load_error(i915, e)) { + } else if (i915_inject_probe_error(i915, e)) { /* require next minor version */ uc_fw->minor_ver_wanted += 1; uc_fw->user_overridden = user; - } else if (uc_fw->major_ver_wanted && i915_inject_load_error(i915, e)) { + } else if (uc_fw->major_ver_wanted && + i915_inject_probe_error(i915, e)) { /* require prev major version */ uc_fw->major_ver_wanted -= 1; uc_fw->minor_ver_wanted = 0; uc_fw->user_overridden = user; - } else if (uc_fw->minor_ver_wanted && i915_inject_load_error(i915, e)) { + } else if (uc_fw->minor_ver_wanted && + i915_inject_probe_error(i915, e)) { /* require prev minor version - hey, this should work! */ uc_fw->minor_ver_wanted -= 1; uc_fw->user_overridden = user; - } else if (user && i915_inject_load_error(i915, e)) { + } else if (user && i915_inject_probe_error(i915, e)) { /* officially unsupported platform */ uc_fw->major_ver_wanted = 0; uc_fw->minor_ver_wanted = 0; @@ -269,7 +278,7 @@ int intel_uc_fw_fetch(struct intel_uc_fw *uc_fw, struct drm_i915_private *i915) GEM_BUG_ON(!i915->wopcm.size); GEM_BUG_ON(!intel_uc_fw_is_enabled(uc_fw)); - err = i915_inject_load_error(i915, -ENXIO); + err = i915_inject_probe_error(i915, -ENXIO); if (err) return err; @@ -337,25 +346,10 @@ int intel_uc_fw_fetch(struct intel_uc_fw *uc_fw, struct drm_i915_private *i915) } /* Get version numbers from the CSS header */ - switch (uc_fw->type) { - case INTEL_UC_FW_TYPE_GUC: - uc_fw->major_ver_found = FIELD_GET(CSS_SW_VERSION_GUC_MAJOR, - css->sw_version); - uc_fw->minor_ver_found = FIELD_GET(CSS_SW_VERSION_GUC_MINOR, - css->sw_version); - break; - - case INTEL_UC_FW_TYPE_HUC: - uc_fw->major_ver_found = FIELD_GET(CSS_SW_VERSION_HUC_MAJOR, - css->sw_version); - uc_fw->minor_ver_found = FIELD_GET(CSS_SW_VERSION_HUC_MINOR, - css->sw_version); - break; - - default: - MISSING_CASE(uc_fw->type); - break; - } + uc_fw->major_ver_found = FIELD_GET(CSS_SW_VERSION_UC_MAJOR, + css->sw_version); + uc_fw->minor_ver_found = FIELD_GET(CSS_SW_VERSION_UC_MINOR, + css->sw_version); if (uc_fw->major_ver_found != uc_fw->major_ver_wanted || uc_fw->minor_ver_found < uc_fw->minor_ver_wanted) { @@ -400,7 +394,7 @@ static u32 uc_fw_ggtt_offset(struct intel_uc_fw *uc_fw, struct i915_ggtt *ggtt) { struct drm_mm_node *node = &ggtt->uc_fw; - GEM_BUG_ON(!node->allocated); + GEM_BUG_ON(!drm_mm_node_allocated(node)); GEM_BUG_ON(upper_32_bits(node->start)); GEM_BUG_ON(upper_32_bits(node->start + node->size - 1)); @@ -445,7 +439,7 @@ static int uc_fw_xfer(struct intel_uc_fw *uc_fw, struct intel_gt *gt, u64 offset; int ret; - ret = i915_inject_load_error(gt->i915, -ETIMEDOUT); + ret = i915_inject_probe_error(gt->i915, -ETIMEDOUT); if (ret) return ret; @@ -506,7 +500,7 @@ int intel_uc_fw_upload(struct intel_uc_fw *uc_fw, struct intel_gt *gt, /* make sure the status was cleared the last time we reset the uc */ GEM_BUG_ON(intel_uc_fw_is_loaded(uc_fw)); - err = i915_inject_load_error(gt->i915, -ENOEXEC); + err = i915_inject_probe_error(gt->i915, -ENOEXEC); if (err) return err; diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h index ae58e8a8c53b..029214cdedd5 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h @@ -39,9 +39,6 @@ * 3. Length info of each component can be found in header, in dwords. * 4. Modulus and exponent key are not required by driver. They may not appear * in fw. So driver will load a truncated firmware in this case. - * - * The only difference between GuC and HuC firmwares is how the version - * information is saved. */ struct uc_css_header { @@ -69,11 +66,9 @@ struct uc_css_header { char username[8]; char buildnumber[12]; u32 sw_version; -#define CSS_SW_VERSION_GUC_MAJOR (0xFF << 16) -#define CSS_SW_VERSION_GUC_MINOR (0xFF << 8) -#define CSS_SW_VERSION_GUC_PATCH (0xFF << 0) -#define CSS_SW_VERSION_HUC_MAJOR (0xFFFF << 16) -#define CSS_SW_VERSION_HUC_MINOR (0xFFFF << 0) +#define CSS_SW_VERSION_UC_MAJOR (0xFF << 16) +#define CSS_SW_VERSION_UC_MINOR (0xFF << 8) +#define CSS_SW_VERSION_UC_PATCH (0xFF << 0) u32 reserved[14]; u32 header_info; } __packed; diff --git a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c index bba0eafe1cdb..d8a80388bd31 100644 --- a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c +++ b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c @@ -108,23 +108,15 @@ static bool client_doorbell_in_sync(struct intel_guc_client *client) * validating that the doorbells status expected by the driver matches what the * GuC/HW have. */ -static int igt_guc_clients(void *args) +static int igt_guc_clients(void *arg) { - struct drm_i915_private *dev_priv = args; + struct intel_gt *gt = arg; + struct intel_guc *guc = >->uc.guc; intel_wakeref_t wakeref; - struct intel_guc *guc; int err = 0; - GEM_BUG_ON(!HAS_GT_UC(dev_priv)); - mutex_lock(&dev_priv->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); - - guc = &dev_priv->gt.uc.guc; - if (!guc) { - pr_err("No guc object!\n"); - err = -EINVAL; - goto unlock; - } + GEM_BUG_ON(!HAS_GT_UC(gt->i915)); + wakeref = intel_runtime_pm_get(gt->uncore->rpm); err = check_all_doorbells(guc); if (err) @@ -189,8 +181,7 @@ out: guc_clients_create(guc); guc_clients_enable(guc); unlock: - intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); - mutex_unlock(&dev_priv->drm.struct_mutex); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); return err; } @@ -201,22 +192,14 @@ unlock: */ static int igt_guc_doorbells(void *arg) { - struct drm_i915_private *dev_priv = arg; + struct intel_gt *gt = arg; + struct intel_guc *guc = >->uc.guc; intel_wakeref_t wakeref; - struct intel_guc *guc; int i, err = 0; u16 db_id; - GEM_BUG_ON(!HAS_GT_UC(dev_priv)); - mutex_lock(&dev_priv->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); - - guc = &dev_priv->gt.uc.guc; - if (!guc) { - pr_err("No guc object!\n"); - err = -EINVAL; - goto unlock; - } + GEM_BUG_ON(!HAS_GT_UC(gt->i915)); + wakeref = intel_runtime_pm_get(gt->uncore->rpm); err = check_all_doorbells(guc); if (err) @@ -298,20 +281,19 @@ out: guc_client_free(clients[i]); } unlock: - intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); - mutex_unlock(&dev_priv->drm.struct_mutex); + intel_runtime_pm_put(gt->uncore->rpm, wakeref); return err; } -int intel_guc_live_selftest(struct drm_i915_private *dev_priv) +int intel_guc_live_selftest(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(igt_guc_clients), SUBTEST(igt_guc_doorbells), }; - if (!USES_GUC_SUBMISSION(dev_priv)) + if (!USES_GUC_SUBMISSION(i915)) return 0; - return i915_subtests(tests, dev_priv); + return intel_gt_live_subtests(tests, &i915->gt); } diff --git a/drivers/gpu/drm/i915/gvt/aperture_gm.c b/drivers/gpu/drm/i915/gvt/aperture_gm.c index 5ff2437b2998..771420453f82 100644 --- a/drivers/gpu/drm/i915/gvt/aperture_gm.c +++ b/drivers/gpu/drm/i915/gvt/aperture_gm.c @@ -61,14 +61,14 @@ static int alloc_gm(struct intel_vgpu *vgpu, bool high_gm) flags = PIN_MAPPABLE; } - mutex_lock(&dev_priv->drm.struct_mutex); + mutex_lock(&dev_priv->ggtt.vm.mutex); mmio_hw_access_pre(dev_priv); ret = i915_gem_gtt_insert(&dev_priv->ggtt.vm, node, size, I915_GTT_PAGE_SIZE, I915_COLOR_UNEVICTABLE, start, end, flags); mmio_hw_access_post(dev_priv); - mutex_unlock(&dev_priv->drm.struct_mutex); + mutex_unlock(&dev_priv->ggtt.vm.mutex); if (ret) gvt_err("fail to alloc %s gm space from host\n", high_gm ? "high" : "low"); @@ -98,9 +98,9 @@ static int alloc_vgpu_gm(struct intel_vgpu *vgpu) return 0; out_free_aperture: - mutex_lock(&dev_priv->drm.struct_mutex); + mutex_lock(&dev_priv->ggtt.vm.mutex); drm_mm_remove_node(&vgpu->gm.low_gm_node); - mutex_unlock(&dev_priv->drm.struct_mutex); + mutex_unlock(&dev_priv->ggtt.vm.mutex); return ret; } @@ -108,10 +108,10 @@ static void free_vgpu_gm(struct intel_vgpu *vgpu) { struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; - mutex_lock(&dev_priv->drm.struct_mutex); + mutex_lock(&dev_priv->ggtt.vm.mutex); drm_mm_remove_node(&vgpu->gm.low_gm_node); drm_mm_remove_node(&vgpu->gm.high_gm_node); - mutex_unlock(&dev_priv->drm.struct_mutex); + mutex_unlock(&dev_priv->ggtt.vm.mutex); } /** @@ -198,7 +198,7 @@ static int alloc_vgpu_fence(struct intel_vgpu *vgpu) mutex_lock(&dev_priv->ggtt.vm.mutex); for (i = 0; i < vgpu_fence_sz(vgpu); i++) { - reg = i915_reserve_fence(dev_priv); + reg = i915_reserve_fence(&dev_priv->ggtt); if (IS_ERR(reg)) goto out_free_fence; diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index e753b1e706e2..21a176cd8acc 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -35,7 +35,9 @@ */ #include <linux/slab.h> + #include "i915_drv.h" +#include "gt/intel_ring.h" #include "gvt.h" #include "i915_pvinfo.h" #include "trace.h" @@ -1597,9 +1599,9 @@ static int cmd_handler_mi_op_2f(struct parser_exec_state *s) if (!(cmd_val(s, 0) & (1 << 22))) return ret; - /* check if QWORD */ - if (DWORD_FIELD(0, 20, 19) == 1) - valid_len += 8; + /* check inline data */ + if (cmd_val(s, 0) & BIT(18)) + valid_len = CMD_LEN(9); ret = gvt_check_valid_cmd_length(cmd_length(s), valid_len); if (ret) diff --git a/drivers/gpu/drm/i915/gvt/dmabuf.c b/drivers/gpu/drm/i915/gvt/dmabuf.c index 13044c027f27..e451298d11c3 100644 --- a/drivers/gpu/drm/i915/gvt/dmabuf.c +++ b/drivers/gpu/drm/i915/gvt/dmabuf.c @@ -152,6 +152,7 @@ static const struct drm_i915_gem_object_ops intel_vgpu_gem_ops = { static struct drm_i915_gem_object *vgpu_create_gem(struct drm_device *dev, struct intel_vgpu_fb_info *info) { + static struct lock_class_key lock_class; struct drm_i915_private *dev_priv = to_i915(dev); struct drm_i915_gem_object *obj; @@ -161,7 +162,7 @@ static struct drm_i915_gem_object *vgpu_create_gem(struct drm_device *dev, drm_gem_private_object_init(dev, &obj->base, roundup(info->size, PAGE_SIZE)); - i915_gem_object_init(obj, &intel_vgpu_gem_ops); + i915_gem_object_init(obj, &intel_vgpu_gem_ops, &lock_class); obj->read_domains = I915_GEM_DOMAIN_GTT; obj->write_domain = 0; @@ -498,8 +499,6 @@ int intel_vgpu_get_dmabuf(struct intel_vgpu *vgpu, unsigned int dmabuf_id) goto out_free_gem; } - i915_gem_object_put(obj); - ret = dma_buf_fd(dmabuf, DRM_CLOEXEC | DRM_RDWR); if (ret < 0) { gvt_vgpu_err("create dma-buf fd failed ret:%d\n", ret); @@ -524,6 +523,8 @@ int intel_vgpu_get_dmabuf(struct intel_vgpu *vgpu, unsigned int dmabuf_id) file_count(dmabuf->file), kref_read(&obj->base.refcount)); + i915_gem_object_put(obj); + return dmabuf_fd; out_free_dmabuf: diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index f21b8fb5b37e..d6e7a1189bad 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -534,7 +534,7 @@ static void clean_execlist(struct intel_vgpu *vgpu, struct intel_vgpu_submission *s = &vgpu->submission; intel_engine_mask_t tmp; - for_each_engine_masked(engine, dev_priv, engine_mask, tmp) { + for_each_engine_masked(engine, &dev_priv->gt, engine_mask, tmp) { kfree(s->ring_scan_buffer[engine->id]); s->ring_scan_buffer[engine->id] = NULL; s->ring_scan_buffer_size[engine->id] = 0; @@ -548,7 +548,7 @@ static void reset_execlist(struct intel_vgpu *vgpu, struct intel_engine_cs *engine; intel_engine_mask_t tmp; - for_each_engine_masked(engine, dev_priv, engine_mask, tmp) + for_each_engine_masked(engine, &dev_priv->gt, engine_mask, tmp) init_vgpu_execlist(vgpu, engine->id); } diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 25f78196b964..bb9fe6bf5275 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -460,6 +460,7 @@ static int pipeconf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, static i915_reg_t force_nonpriv_white_list[] = { GEN9_CS_DEBUG_MODE1, //_MMIO(0x20ec) GEN9_CTX_PREEMPT_REG,//_MMIO(0x2248) + PS_INVOCATION_COUNT,//_MMIO(0x2348) GEN8_CS_CHICKEN1,//_MMIO(0x2580) _MMIO(0x2690), _MMIO(0x2694), @@ -508,7 +509,7 @@ static inline bool in_whitelist(unsigned int reg) static int force_nonpriv_write(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { - u32 reg_nonpriv = *(u32 *)p_data; + u32 reg_nonpriv = (*(u32 *)p_data) & REG_GENMASK(25, 2); int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset); u32 ring_base; struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; @@ -528,7 +529,7 @@ static int force_nonpriv_write(struct intel_vgpu *vgpu, bytes); } else gvt_err("vgpu(%d) Invalid FORCE_NONPRIV write %x at offset %x\n", - vgpu->id, reg_nonpriv, offset); + vgpu->id, *(u32 *)p_data, offset); return 0; } @@ -819,13 +820,16 @@ static int trigger_aux_channel_interrupt(struct intel_vgpu *vgpu, struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; enum intel_gvt_event_type event; - if (reg == _DPA_AUX_CH_CTL) + if (reg == i915_mmio_reg_offset(DP_AUX_CH_CTL(AUX_CH_A))) event = AUX_CHANNEL_A; - else if (reg == _PCH_DPB_AUX_CH_CTL || reg == _DPB_AUX_CH_CTL) + else if (reg == _PCH_DPB_AUX_CH_CTL || + reg == i915_mmio_reg_offset(DP_AUX_CH_CTL(AUX_CH_B))) event = AUX_CHANNEL_B; - else if (reg == _PCH_DPC_AUX_CH_CTL || reg == _DPC_AUX_CH_CTL) + else if (reg == _PCH_DPC_AUX_CH_CTL || + reg == i915_mmio_reg_offset(DP_AUX_CH_CTL(AUX_CH_C))) event = AUX_CHANNEL_C; - else if (reg == _PCH_DPD_AUX_CH_CTL || reg == _DPD_AUX_CH_CTL) + else if (reg == _PCH_DPD_AUX_CH_CTL || + reg == i915_mmio_reg_offset(DP_AUX_CH_CTL(AUX_CH_D))) event = AUX_CHANNEL_D; else { WARN_ON(true); @@ -2796,7 +2800,7 @@ static int init_broadwell_mmio_info(struct intel_gvt *gvt) MMIO_D(CHICKEN_PIPESL_1(PIPE_C), D_BDW_PLUS); MMIO_D(WM_MISC, D_BDW); - MMIO_D(_MMIO(BDW_EDP_PSR_BASE), D_BDW); + MMIO_D(_MMIO(_SRD_CTL_EDP), D_BDW); MMIO_D(_MMIO(0x6671c), D_BDW_PLUS); MMIO_D(_MMIO(0x66c00), D_BDW_PLUS); @@ -2872,11 +2876,11 @@ static int init_skl_mmio_info(struct intel_gvt *gvt) MMIO_DH(FORCEWAKE_MEDIA_GEN9, D_SKL_PLUS, NULL, mul_force_wake_write); MMIO_DH(FORCEWAKE_ACK_MEDIA_GEN9, D_SKL_PLUS, NULL, NULL); - MMIO_F(_MMIO(_DPB_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL, + MMIO_F(DP_AUX_CH_CTL(AUX_CH_B), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL, dp_aux_ch_ctl_mmio_write); - MMIO_F(_MMIO(_DPC_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL, + MMIO_F(DP_AUX_CH_CTL(AUX_CH_C), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL, dp_aux_ch_ctl_mmio_write); - MMIO_F(_MMIO(_DPD_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL, + MMIO_F(DP_AUX_CH_CTL(AUX_CH_D), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL, dp_aux_ch_ctl_mmio_write); MMIO_D(HSW_PWR_WELL_CTL1, D_SKL_PLUS); @@ -3417,6 +3421,10 @@ int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt, } for (i = 0; i < gvt->mmio.num_mmio_block; i++, block++) { + /* pvinfo data doesn't come from hw mmio */ + if (i915_mmio_reg_offset(block->offset) == VGT_PVINFO_PAGE) + continue; + for (j = 0; j < block->size; j += 4) { ret = handler(gvt, i915_mmio_reg_offset(block->offset) + j, diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c index 343d79c1cb7e..04a5a0d90823 100644 --- a/drivers/gpu/drm/i915/gvt/kvmgt.c +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c @@ -1564,27 +1564,10 @@ vgpu_id_show(struct device *dev, struct device_attribute *attr, return sprintf(buf, "\n"); } -static ssize_t -hw_id_show(struct device *dev, struct device_attribute *attr, - char *buf) -{ - struct mdev_device *mdev = mdev_from_dev(dev); - - if (mdev) { - struct intel_vgpu *vgpu = (struct intel_vgpu *) - mdev_get_drvdata(mdev); - return sprintf(buf, "%u\n", - vgpu->submission.shadow[0]->gem_context->hw_id); - } - return sprintf(buf, "\n"); -} - static DEVICE_ATTR_RO(vgpu_id); -static DEVICE_ATTR_RO(hw_id); static struct attribute *intel_vgpu_attrs[] = { &dev_attr_vgpu_id.attr, - &dev_attr_hw_id.attr, NULL }; diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c index 4208e40445b1..aaf15916d29a 100644 --- a/drivers/gpu/drm/i915/gvt/mmio_context.c +++ b/drivers/gpu/drm/i915/gvt/mmio_context.c @@ -35,6 +35,7 @@ #include "i915_drv.h" #include "gt/intel_context.h" +#include "gt/intel_ring.h" #include "gvt.h" #include "trace.h" diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 6c79d16b381e..5b2a7d072ec9 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -38,6 +38,7 @@ #include "gem/i915_gem_context.h" #include "gem/i915_gem_pm.h" #include "gt/intel_context.h" +#include "gt/intel_ring.h" #include "i915_drv.h" #include "gvt.h" @@ -194,7 +195,7 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload) return -EFAULT; } - page = i915_gem_object_get_page(ctx_obj, LRC_HEADER_PAGES + i); + page = i915_gem_object_get_page(ctx_obj, i); dst = kmap(page); intel_gvt_hypervisor_read_gpa(vgpu, context_gpa, dst, I915_GTT_PAGE_SIZE); @@ -365,7 +366,8 @@ static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload, struct i915_gem_context *ctx) { struct intel_vgpu_mm *mm = workload->shadow_mm; - struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ctx->vm); + struct i915_ppgtt *ppgtt = + i915_vm_to_ppgtt(i915_gem_context_get_vm_rcu(ctx)); int i = 0; if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) { @@ -378,6 +380,8 @@ static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload, px_dma(pd) = mm->ppgtt_mm.shadow_pdps[i]; } } + + i915_vm_put(&ppgtt->vm); } static int @@ -385,11 +389,8 @@ intel_gvt_workload_req_alloc(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; struct intel_vgpu_submission *s = &vgpu->submission; - struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; struct i915_request *rq; - lockdep_assert_held(&dev_priv->drm.struct_mutex); - if (workload->req) return 0; @@ -415,10 +416,9 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; struct intel_vgpu_submission *s = &vgpu->submission; - struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; int ret; - lockdep_assert_held(&dev_priv->drm.struct_mutex); + lockdep_assert_held(&vgpu->vgpu_lock); if (workload->shadow) return 0; @@ -580,8 +580,6 @@ static void update_vreg_in_ctx(struct intel_vgpu_workload *workload) static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) { - struct intel_vgpu *vgpu = workload->vgpu; - struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; struct intel_vgpu_shadow_bb *bb, *pos; if (list_empty(&workload->shadow_bb)) @@ -590,8 +588,6 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) bb = list_first_entry(&workload->shadow_bb, struct intel_vgpu_shadow_bb, list); - mutex_lock(&dev_priv->drm.struct_mutex); - list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) { if (bb->obj) { if (bb->accessing) @@ -609,8 +605,6 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) list_del(&bb->list); kfree(bb); } - - mutex_unlock(&dev_priv->drm.struct_mutex); } static int prepare_workload(struct intel_vgpu_workload *workload) @@ -685,7 +679,6 @@ err_unpin_mm: static int dispatch_workload(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; - struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; struct i915_request *rq; int ring_id = workload->ring_id; int ret; @@ -694,7 +687,6 @@ static int dispatch_workload(struct intel_vgpu_workload *workload) ring_id, workload); mutex_lock(&vgpu->vgpu_lock); - mutex_lock(&dev_priv->drm.struct_mutex); ret = intel_gvt_workload_req_alloc(workload); if (ret) @@ -729,7 +721,6 @@ out: err_req: if (ret) workload->status = ret; - mutex_unlock(&dev_priv->drm.struct_mutex); mutex_unlock(&vgpu->vgpu_lock); return ret; } @@ -844,7 +835,7 @@ static void update_guest_context(struct intel_vgpu_workload *workload) return; } - page = i915_gem_object_get_page(ctx_obj, LRC_HEADER_PAGES + i); + page = i915_gem_object_get_page(ctx_obj, i); src = kmap(page); intel_gvt_hypervisor_write_gpa(vgpu, context_gpa, src, I915_GTT_PAGE_SIZE); @@ -887,7 +878,7 @@ void intel_vgpu_clean_workloads(struct intel_vgpu *vgpu, intel_engine_mask_t tmp; /* free the unsubmited workloads in the queues. */ - for_each_engine_masked(engine, dev_priv, engine_mask, tmp) { + for_each_engine_masked(engine, &dev_priv->gt, engine_mask, tmp) { list_for_each_entry_safe(pos, n, &s->workload_q_head[engine->id], list) { list_del_init(&pos->list); @@ -1233,20 +1224,18 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) struct intel_vgpu_submission *s = &vgpu->submission; struct intel_engine_cs *engine; struct i915_gem_context *ctx; + struct i915_ppgtt *ppgtt; enum intel_engine_id i; int ret; - mutex_lock(&i915->drm.struct_mutex); - ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MAX); - if (IS_ERR(ctx)) { - ret = PTR_ERR(ctx); - goto out_unlock; - } + if (IS_ERR(ctx)) + return PTR_ERR(ctx); i915_gem_context_set_force_single_submission(ctx); - i915_context_ppgtt_root_save(s, i915_vm_to_ppgtt(ctx->vm)); + ppgtt = i915_vm_to_ppgtt(i915_gem_context_get_vm_rcu(ctx)); + i915_context_ppgtt_root_save(s, ppgtt); for_each_engine(engine, i915, i) { struct intel_context *ce; @@ -1291,12 +1280,12 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) atomic_set(&s->running_workload_num, 0); bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES); + i915_vm_put(&ppgtt->vm); i915_gem_context_put(ctx); - mutex_unlock(&i915->drm.struct_mutex); return 0; out_shadow_ctx: - i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(ctx->vm)); + i915_context_ppgtt_root_restore(s, ppgtt); for_each_engine(engine, i915, i) { if (IS_ERR(s->shadow[i])) break; @@ -1304,9 +1293,8 @@ out_shadow_ctx: intel_context_unpin(s->shadow[i]); intel_context_put(s->shadow[i]); } + i915_vm_put(&ppgtt->vm); i915_gem_context_put(ctx); -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); return ret; } @@ -1597,9 +1585,7 @@ intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, */ if (list_empty(workload_q_head(vgpu, ring_id))) { intel_runtime_pm_get(&dev_priv->runtime_pm); - mutex_lock(&dev_priv->drm.struct_mutex); ret = intel_gvt_scan_and_shadow_workload(workload); - mutex_unlock(&dev_priv->drm.struct_mutex); intel_runtime_pm_put_unchecked(&dev_priv->runtime_pm); } diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c index 48e16ad93bbd..a19e7d89bc8a 100644 --- a/drivers/gpu/drm/i915/i915_active.c +++ b/drivers/gpu/drm/i915/i915_active.c @@ -7,13 +7,12 @@ #include <linux/debugobjects.h> #include "gt/intel_engine_pm.h" +#include "gt/intel_ring.h" #include "i915_drv.h" #include "i915_active.h" #include "i915_globals.h" -#define BKL(ref) (&(ref)->i915->drm.struct_mutex) - /* * Active refs memory management * @@ -27,35 +26,35 @@ static struct i915_global_active { } global; struct active_node { - struct i915_active_request base; + struct i915_active_fence base; struct i915_active *ref; struct rb_node node; u64 timeline; }; static inline struct active_node * -node_from_active(struct i915_active_request *active) +node_from_active(struct i915_active_fence *active) { return container_of(active, struct active_node, base); } #define take_preallocated_barriers(x) llist_del_all(&(x)->preallocated_barriers) -static inline bool is_barrier(const struct i915_active_request *active) +static inline bool is_barrier(const struct i915_active_fence *active) { - return IS_ERR(rcu_access_pointer(active->request)); + return IS_ERR(rcu_access_pointer(active->fence)); } static inline struct llist_node *barrier_to_ll(struct active_node *node) { GEM_BUG_ON(!is_barrier(&node->base)); - return (struct llist_node *)&node->base.link; + return (struct llist_node *)&node->base.cb.node; } static inline struct intel_engine_cs * __barrier_to_engine(struct active_node *node) { - return (struct intel_engine_cs *)READ_ONCE(node->base.link.prev); + return (struct intel_engine_cs *)READ_ONCE(node->base.cb.node.prev); } static inline struct intel_engine_cs * @@ -68,7 +67,7 @@ barrier_to_engine(struct active_node *node) static inline struct active_node *barrier_from_ll(struct llist_node *x) { return container_of((struct list_head *)x, - struct active_node, base.link); + struct active_node, base.cb.node); } #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) && IS_ENABLED(CONFIG_DEBUG_OBJECTS) @@ -92,12 +91,17 @@ static void debug_active_init(struct i915_active *ref) static void debug_active_activate(struct i915_active *ref) { - debug_object_activate(ref, &active_debug_desc); + spin_lock_irq(&ref->tree_lock); + if (!atomic_read(&ref->count)) /* before the first inc */ + debug_object_activate(ref, &active_debug_desc); + spin_unlock_irq(&ref->tree_lock); } static void debug_active_deactivate(struct i915_active *ref) { - debug_object_deactivate(ref, &active_debug_desc); + lockdep_assert_held(&ref->tree_lock); + if (!atomic_read(&ref->count)) /* after the last dec */ + debug_object_deactivate(ref, &active_debug_desc); } static void debug_active_fini(struct i915_active *ref) @@ -125,31 +129,46 @@ __active_retire(struct i915_active *ref) { struct active_node *it, *n; struct rb_root root; - bool retire = false; + unsigned long flags; - lockdep_assert_held(&ref->mutex); + GEM_BUG_ON(i915_active_is_idle(ref)); /* return the unused nodes to our slabcache -- flushing the allocator */ - if (atomic_dec_and_test(&ref->count)) { - debug_active_deactivate(ref); - root = ref->tree; - ref->tree = RB_ROOT; - ref->cache = NULL; - retire = true; - } - - mutex_unlock(&ref->mutex); - if (!retire) + if (!atomic_dec_and_lock_irqsave(&ref->count, &ref->tree_lock, flags)) return; - rbtree_postorder_for_each_entry_safe(it, n, &root, node) { - GEM_BUG_ON(i915_active_request_isset(&it->base)); - kmem_cache_free(global.slab_cache, it); - } + GEM_BUG_ON(rcu_access_pointer(ref->excl.fence)); + debug_active_deactivate(ref); + + root = ref->tree; + ref->tree = RB_ROOT; + ref->cache = NULL; + + spin_unlock_irqrestore(&ref->tree_lock, flags); /* After the final retire, the entire struct may be freed */ if (ref->retire) ref->retire(ref); + + /* ... except if you wait on it, you must manage your own references! */ + wake_up_var(ref); + + rbtree_postorder_for_each_entry_safe(it, n, &root, node) { + GEM_BUG_ON(i915_active_fence_isset(&it->base)); + kmem_cache_free(global.slab_cache, it); + } +} + +static void +active_work(struct work_struct *wrk) +{ + struct i915_active *ref = container_of(wrk, typeof(*ref), work); + + GEM_BUG_ON(!atomic_read(&ref->count)); + if (atomic_add_unless(&ref->count, -1, 1)) + return; + + __active_retire(ref); } static void @@ -159,18 +178,29 @@ active_retire(struct i915_active *ref) if (atomic_add_unless(&ref->count, -1, 1)) return; - /* One active may be flushed from inside the acquire of another */ - mutex_lock_nested(&ref->mutex, SINGLE_DEPTH_NESTING); + if (ref->flags & I915_ACTIVE_RETIRE_SLEEPS) { + queue_work(system_unbound_wq, &ref->work); + return; + } + __active_retire(ref); } static void -node_retire(struct i915_active_request *base, struct i915_request *rq) +node_retire(struct dma_fence *fence, struct dma_fence_cb *cb) { - active_retire(node_from_active(base)->ref); + i915_active_fence_cb(fence, cb); + active_retire(container_of(cb, struct active_node, base.cb)->ref); } -static struct i915_active_request * +static void +excl_retire(struct dma_fence *fence, struct dma_fence_cb *cb) +{ + i915_active_fence_cb(fence, cb); + active_retire(container_of(cb, struct i915_active, excl.cb)); +} + +static struct i915_active_fence * active_instance(struct i915_active *ref, struct intel_timeline *tl) { struct active_node *node, *prealloc; @@ -193,7 +223,7 @@ active_instance(struct i915_active *ref, struct intel_timeline *tl) if (!prealloc) return NULL; - mutex_lock(&ref->mutex); + spin_lock_irq(&ref->tree_lock); GEM_BUG_ON(i915_active_is_idle(ref)); parent = NULL; @@ -214,7 +244,7 @@ active_instance(struct i915_active *ref, struct intel_timeline *tl) } node = prealloc; - i915_active_request_init(&node->base, &tl->mutex, NULL, node_retire); + __i915_active_fence_init(&node->base, &tl->mutex, NULL, node_retire); node->ref = ref; node->timeline = idx; @@ -223,29 +253,36 @@ active_instance(struct i915_active *ref, struct intel_timeline *tl) out: ref->cache = node; - mutex_unlock(&ref->mutex); + spin_unlock_irq(&ref->tree_lock); BUILD_BUG_ON(offsetof(typeof(*node), base)); return &node->base; } -void __i915_active_init(struct drm_i915_private *i915, - struct i915_active *ref, +void __i915_active_init(struct i915_active *ref, int (*active)(struct i915_active *ref), void (*retire)(struct i915_active *ref), struct lock_class_key *key) { + unsigned long bits; + debug_active_init(ref); - ref->i915 = i915; ref->flags = 0; ref->active = active; - ref->retire = retire; + ref->retire = ptr_unpack_bits(retire, &bits, 2); + if (bits & I915_ACTIVE_MAY_SLEEP) + ref->flags |= I915_ACTIVE_RETIRE_SLEEPS; + + spin_lock_init(&ref->tree_lock); ref->tree = RB_ROOT; ref->cache = NULL; + init_llist_head(&ref->preallocated_barriers); atomic_set(&ref->count, 0); __mutex_init(&ref->mutex, "i915_active", key); + __i915_active_fence_init(&ref->excl, &ref->mutex, NULL, excl_retire); + INIT_WORK(&ref->work, active_work); } static bool ____active_del_barrier(struct i915_active *ref, @@ -298,9 +335,9 @@ __active_del_barrier(struct i915_active *ref, struct active_node *node) int i915_active_ref(struct i915_active *ref, struct intel_timeline *tl, - struct i915_request *rq) + struct dma_fence *fence) { - struct i915_active_request *active; + struct i915_active_fence *active; int err; lockdep_assert_held(&tl->mutex); @@ -323,26 +360,44 @@ int i915_active_ref(struct i915_active *ref, * request that we want to emit on the kernel_context. */ __active_del_barrier(ref, node_from_active(active)); - RCU_INIT_POINTER(active->request, NULL); - INIT_LIST_HEAD(&active->link); - } else { - if (!i915_active_request_isset(active)) - atomic_inc(&ref->count); + RCU_INIT_POINTER(active->fence, NULL); + atomic_dec(&ref->count); } - GEM_BUG_ON(!atomic_read(&ref->count)); - __i915_active_request_set(active, rq); + if (!__i915_active_fence_set(active, fence)) + atomic_inc(&ref->count); out: i915_active_release(ref); return err; } +void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f) +{ + /* We expect the caller to manage the exclusive timeline ordering */ + GEM_BUG_ON(i915_active_is_idle(ref)); + + /* + * As we don't know which mutex the caller is using, we told a small + * lie to the debug code that it is using the i915_active.mutex; + * and now we must stick to that lie. + */ + mutex_acquire(&ref->mutex.dep_map, 0, 0, _THIS_IP_); + if (!__i915_active_fence_set(&ref->excl, f)) + atomic_inc(&ref->count); + mutex_release(&ref->mutex.dep_map, _THIS_IP_); +} + +bool i915_active_acquire_if_busy(struct i915_active *ref) +{ + debug_active_assert(ref); + return atomic_add_unless(&ref->count, 1, 0); +} + int i915_active_acquire(struct i915_active *ref) { int err; - debug_active_assert(ref); - if (atomic_add_unless(&ref->count, 1, 0)) + if (i915_active_acquire_if_busy(ref)) return 0; err = mutex_lock_interruptible(&ref->mutex); @@ -367,109 +422,66 @@ void i915_active_release(struct i915_active *ref) active_retire(ref); } -static void __active_ungrab(struct i915_active *ref) -{ - clear_and_wake_up_bit(I915_ACTIVE_GRAB_BIT, &ref->flags); -} - -bool i915_active_trygrab(struct i915_active *ref) +static void enable_signaling(struct i915_active_fence *active) { - debug_active_assert(ref); - - if (test_and_set_bit(I915_ACTIVE_GRAB_BIT, &ref->flags)) - return false; + struct dma_fence *fence; - if (!atomic_add_unless(&ref->count, 1, 0)) { - __active_ungrab(ref); - return false; - } + fence = i915_active_fence_get(active); + if (!fence) + return; - return true; -} - -void i915_active_ungrab(struct i915_active *ref) -{ - GEM_BUG_ON(!test_bit(I915_ACTIVE_GRAB_BIT, &ref->flags)); - - active_retire(ref); - __active_ungrab(ref); + dma_fence_enable_sw_signaling(fence); + dma_fence_put(fence); } int i915_active_wait(struct i915_active *ref) { struct active_node *it, *n; - int err; + int err = 0; might_sleep(); - might_lock(&ref->mutex); - if (i915_active_is_idle(ref)) + if (!i915_active_acquire_if_busy(ref)) return 0; - err = mutex_lock_interruptible(&ref->mutex); - if (err) - return err; - - if (!atomic_add_unless(&ref->count, 1, 0)) { - mutex_unlock(&ref->mutex); - return 0; - } - + /* Flush lazy signals */ + enable_signaling(&ref->excl); rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) { - if (is_barrier(&it->base)) { /* unconnected idle-barrier */ - err = -EBUSY; - break; - } + if (is_barrier(&it->base)) /* unconnected idle barrier */ + continue; - err = i915_active_request_retire(&it->base, BKL(ref)); - if (err) - break; + enable_signaling(&it->base); } + /* Any fence added after the wait begins will not be auto-signaled */ - __active_retire(ref); + i915_active_release(ref); if (err) return err; - if (wait_on_bit(&ref->flags, I915_ACTIVE_GRAB_BIT, TASK_KILLABLE)) + if (wait_var_event_interruptible(ref, i915_active_is_idle(ref))) return -EINTR; - if (!i915_active_is_idle(ref)) - return -EBUSY; - return 0; } -int i915_request_await_active_request(struct i915_request *rq, - struct i915_active_request *active) -{ - struct i915_request *barrier = - i915_active_request_raw(active, &rq->i915->drm.struct_mutex); - - return barrier ? i915_request_await_dma_fence(rq, &barrier->fence) : 0; -} - int i915_request_await_active(struct i915_request *rq, struct i915_active *ref) { - struct active_node *it, *n; - int err; - - if (RB_EMPTY_ROOT(&ref->tree)) - return 0; + int err = 0; - /* await allocates and so we need to avoid hitting the shrinker */ - err = i915_active_acquire(ref); - if (err) - return err; + if (rcu_access_pointer(ref->excl.fence)) { + struct dma_fence *fence; - mutex_lock(&ref->mutex); - rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) { - err = i915_request_await_active_request(rq, &it->base); - if (err) - break; + rcu_read_lock(); + fence = dma_fence_get_rcu_safe(&ref->excl.fence); + rcu_read_unlock(); + if (fence) { + err = i915_request_await_dma_fence(rq, fence); + dma_fence_put(fence); + } } - mutex_unlock(&ref->mutex); - i915_active_release(ref); + /* In the future we may choose to await on all fences */ + return err; } @@ -477,15 +489,16 @@ int i915_request_await_active(struct i915_request *rq, struct i915_active *ref) void i915_active_fini(struct i915_active *ref) { debug_active_fini(ref); - GEM_BUG_ON(!RB_EMPTY_ROOT(&ref->tree)); GEM_BUG_ON(atomic_read(&ref->count)); + GEM_BUG_ON(work_pending(&ref->work)); + GEM_BUG_ON(!RB_EMPTY_ROOT(&ref->tree)); mutex_destroy(&ref->mutex); } #endif static inline bool is_idle_barrier(struct active_node *node, u64 idx) { - return node->timeline == idx && !i915_active_request_isset(&node->base); + return node->timeline == idx && !i915_active_fence_isset(&node->base); } static struct active_node *reuse_idle_barrier(struct i915_active *ref, u64 idx) @@ -495,7 +508,7 @@ static struct active_node *reuse_idle_barrier(struct i915_active *ref, u64 idx) if (RB_EMPTY_ROOT(&ref->tree)) return NULL; - mutex_lock(&ref->mutex); + spin_lock_irq(&ref->tree_lock); GEM_BUG_ON(i915_active_is_idle(ref)); /* @@ -560,7 +573,7 @@ static struct active_node *reuse_idle_barrier(struct i915_active *ref, u64 idx) goto match; } - mutex_unlock(&ref->mutex); + spin_unlock_irq(&ref->tree_lock); return NULL; @@ -568,7 +581,7 @@ match: rb_erase(p, &ref->tree); /* Hide from waits and sibling allocations */ if (p == &ref->cache->node) ref->cache = NULL; - mutex_unlock(&ref->mutex); + spin_unlock_irq(&ref->tree_lock); return rb_entry(p, struct active_node, node); } @@ -576,11 +589,12 @@ match: int i915_active_acquire_preallocate_barrier(struct i915_active *ref, struct intel_engine_cs *engine) { - struct drm_i915_private *i915 = engine->i915; intel_engine_mask_t tmp, mask = engine->mask; + struct intel_gt *gt = engine->gt; struct llist_node *pos, *next; int err; + GEM_BUG_ON(i915_active_is_idle(ref)); GEM_BUG_ON(!llist_empty(&ref->preallocated_barriers)); /* @@ -589,7 +603,7 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref, * We can then use the preallocated nodes in * i915_active_acquire_barrier() */ - for_each_engine_masked(engine, i915, mask, tmp) { + for_each_engine_masked(engine, gt, mask, tmp) { u64 idx = engine->kernel_context->timeline->fence_context; struct active_node *node; @@ -605,13 +619,13 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref, node->base.lock = &engine->kernel_context->timeline->mutex; #endif - RCU_INIT_POINTER(node->base.request, NULL); - node->base.retire = node_retire; + RCU_INIT_POINTER(node->base.fence, NULL); + node->base.cb.func = node_retire; node->timeline = idx; node->ref = ref; } - if (!i915_active_request_isset(&node->base)) { + if (!i915_active_fence_isset(&node->base)) { /* * Mark this as being *our* unconnected proto-node. * @@ -621,8 +635,8 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref, * and then we can use the rb_node and list pointers * for our tracking of the pending barrier. */ - RCU_INIT_POINTER(node->base.request, ERR_PTR(-EAGAIN)); - node->base.link.prev = (void *)engine; + RCU_INIT_POINTER(node->base.fence, ERR_PTR(-EAGAIN)); + node->base.cb.node.prev = (void *)engine; atomic_inc(&ref->count); } @@ -648,6 +662,7 @@ unwind: void i915_active_acquire_barrier(struct i915_active *ref) { struct llist_node *pos, *next; + unsigned long flags; GEM_BUG_ON(i915_active_is_idle(ref)); @@ -657,12 +672,13 @@ void i915_active_acquire_barrier(struct i915_active *ref) * populated by i915_request_add_active_barriers() to point to the * request that will eventually release them. */ - mutex_lock_nested(&ref->mutex, SINGLE_DEPTH_NESTING); llist_for_each_safe(pos, next, take_preallocated_barriers(ref)) { struct active_node *node = barrier_from_ll(pos); struct intel_engine_cs *engine = barrier_to_engine(node); struct rb_node **p, *parent; + spin_lock_irqsave_nested(&ref->tree_lock, flags, + SINGLE_DEPTH_NESTING); parent = NULL; p = &ref->tree.rb_node; while (*p) { @@ -678,55 +694,125 @@ void i915_active_acquire_barrier(struct i915_active *ref) } rb_link_node(&node->node, parent, p); rb_insert_color(&node->node, &ref->tree); + spin_unlock_irqrestore(&ref->tree_lock, flags); + GEM_BUG_ON(!intel_engine_pm_is_awake(engine)); llist_add(barrier_to_ll(node), &engine->barrier_tasks); intel_engine_pm_put(engine); } - mutex_unlock(&ref->mutex); } void i915_request_add_active_barriers(struct i915_request *rq) { struct intel_engine_cs *engine = rq->engine; struct llist_node *node, *next; + unsigned long flags; GEM_BUG_ON(intel_engine_is_virtual(engine)); - GEM_BUG_ON(rq->timeline != engine->kernel_context->timeline); + GEM_BUG_ON(i915_request_timeline(rq) != engine->kernel_context->timeline); + node = llist_del_all(&engine->barrier_tasks); + if (!node) + return; /* * Attach the list of proto-fences to the in-flight request such * that the parent i915_active will be released when this request * is retired. */ - llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks)) { - RCU_INIT_POINTER(barrier_from_ll(node)->base.request, rq); + spin_lock_irqsave(&rq->lock, flags); + llist_for_each_safe(node, next, node) { + RCU_INIT_POINTER(barrier_from_ll(node)->base.fence, &rq->fence); smp_wmb(); /* serialise with reuse_idle_barrier */ - list_add_tail((struct list_head *)node, &rq->active_list); + list_add_tail((struct list_head *)node, &rq->fence.cb_list); } + spin_unlock_irqrestore(&rq->lock, flags); } -int i915_active_request_set(struct i915_active_request *active, - struct i915_request *rq) +#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) +#define active_is_held(active) lockdep_is_held((active)->lock) +#else +#define active_is_held(active) true +#endif + +/* + * __i915_active_fence_set: Update the last active fence along its timeline + * @active: the active tracker + * @fence: the new fence (under construction) + * + * Records the new @fence as the last active fence along its timeline in + * this active tracker, moving the tracking callbacks from the previous + * fence onto this one. Returns the previous fence (if not already completed), + * which the caller must ensure is executed before the new fence. To ensure + * that the order of fences within the timeline of the i915_active_fence is + * maintained, it must be locked by the caller. + */ +struct dma_fence * +__i915_active_fence_set(struct i915_active_fence *active, + struct dma_fence *fence) { - int err; + struct dma_fence *prev; + unsigned long flags; + + /* NB: must be serialised by an outer timeline mutex (active->lock) */ + spin_lock_irqsave(fence->lock, flags); + GEM_BUG_ON(test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)); + + prev = rcu_dereference_protected(active->fence, active_is_held(active)); + if (prev) { + GEM_BUG_ON(prev == fence); + spin_lock_nested(prev->lock, SINGLE_DEPTH_NESTING); + __list_del_entry(&active->cb.node); + spin_unlock(prev->lock); /* serialise with prev->cb_list */ + + /* + * active->fence is reset by the callback from inside + * interrupt context. We need to serialise our list + * manipulation with the fence->lock to prevent the prev + * being lost inside an interrupt (it can't be replaced as + * no other caller is allowed to enter __i915_active_fence_set + * as we hold the timeline lock). After serialising with + * the callback, we need to double check which ran first, + * our list_del() [decoupling prev from the callback] or + * the callback... + */ + prev = rcu_access_pointer(active->fence); + } + + rcu_assign_pointer(active->fence, fence); + list_add_tail(&active->cb.node, &fence->cb_list); + + spin_unlock_irqrestore(fence->lock, flags); + + return prev; +} + +int i915_active_fence_set(struct i915_active_fence *active, + struct i915_request *rq) +{ + struct dma_fence *fence; + int err = 0; #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) lockdep_assert_held(active->lock); #endif - /* Must maintain ordering wrt previous active requests */ - err = i915_request_await_active_request(rq, active); - if (err) - return err; + /* Must maintain timeline ordering wrt previous active requests */ + rcu_read_lock(); + fence = __i915_active_fence_set(active, &rq->fence); + if (fence) /* but the previous fence may not belong to that timeline! */ + fence = dma_fence_get_rcu(fence); + rcu_read_unlock(); + if (fence) { + err = i915_request_await_dma_fence(rq, fence); + dma_fence_put(fence); + } - __i915_active_request_set(active, rq); - return 0; + return err; } -void i915_active_retire_noop(struct i915_active_request *active, - struct i915_request *request) +void i915_active_noop(struct dma_fence *fence, struct dma_fence_cb *cb) { - /* Space left intentionally blank */ + i915_active_fence_cb(fence, cb); } #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) diff --git a/drivers/gpu/drm/i915/i915_active.h b/drivers/gpu/drm/i915/i915_active.h index f95058f99057..44859356ce97 100644 --- a/drivers/gpu/drm/i915/i915_active.h +++ b/drivers/gpu/drm/i915/i915_active.h @@ -12,6 +12,10 @@ #include "i915_active_types.h" #include "i915_request.h" +struct i915_request; +struct intel_engine_cs; +struct intel_timeline; + /* * We treat requests as fences. This is not be to confused with our * "fence registers" but pipeline synchronisation objects ala GL_ARB_sync. @@ -28,308 +32,108 @@ * write access so that we can perform concurrent read operations between * the CPU and GPU engines, as well as waiting for all rendering to * complete, or waiting for the last GPU user of a "fence register". The - * object then embeds a #i915_active_request to track the most recent (in + * object then embeds a #i915_active_fence to track the most recent (in * retirement order) request relevant for the desired mode of access. - * The #i915_active_request is updated with i915_active_request_set() to + * The #i915_active_fence is updated with i915_active_fence_set() to * track the most recent fence request, typically this is done as part of * i915_vma_move_to_active(). * - * When the #i915_active_request completes (is retired), it will + * When the #i915_active_fence completes (is retired), it will * signal its completion to the owner through a callback as well as mark - * itself as idle (i915_active_request.request == NULL). The owner + * itself as idle (i915_active_fence.request == NULL). The owner * can then perform any action, such as delayed freeing of an active * resource including itself. */ -void i915_active_retire_noop(struct i915_active_request *active, - struct i915_request *request); +void i915_active_noop(struct dma_fence *fence, struct dma_fence_cb *cb); /** - * i915_active_request_init - prepares the activity tracker for use + * __i915_active_fence_init - prepares the activity tracker for use * @active - the active tracker - * @rq - initial request to track, can be NULL + * @fence - initial fence to track, can be NULL * @func - a callback when then the tracker is retired (becomes idle), * can be NULL * - * i915_active_request_init() prepares the embedded @active struct for use as - * an activity tracker, that is for tracking the last known active request - * associated with it. When the last request becomes idle, when it is retired + * i915_active_fence_init() prepares the embedded @active struct for use as + * an activity tracker, that is for tracking the last known active fence + * associated with it. When the last fence becomes idle, when it is retired * after completion, the optional callback @func is invoked. */ static inline void -i915_active_request_init(struct i915_active_request *active, +__i915_active_fence_init(struct i915_active_fence *active, struct mutex *lock, - struct i915_request *rq, - i915_active_retire_fn retire) + void *fence, + dma_fence_func_t fn) { - RCU_INIT_POINTER(active->request, rq); - INIT_LIST_HEAD(&active->link); - active->retire = retire ?: i915_active_retire_noop; + RCU_INIT_POINTER(active->fence, fence); + active->cb.func = fn ?: i915_active_noop; #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) active->lock = lock; #endif } -#define INIT_ACTIVE_REQUEST(name, lock) \ - i915_active_request_init((name), (lock), NULL, NULL) - -/** - * i915_active_request_set - updates the tracker to watch the current request - * @active - the active tracker - * @request - the request to watch - * - * __i915_active_request_set() watches the given @request for completion. Whilst - * that @request is busy, the @active reports busy. When that @request is - * retired, the @active tracker is updated to report idle. - */ -static inline void -__i915_active_request_set(struct i915_active_request *active, - struct i915_request *request) -{ -#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) - lockdep_assert_held(active->lock); -#endif - list_move(&active->link, &request->active_list); - rcu_assign_pointer(active->request, request); -} +#define INIT_ACTIVE_FENCE(A, LOCK) \ + __i915_active_fence_init((A), (LOCK), NULL, NULL) -int __must_check -i915_active_request_set(struct i915_active_request *active, - struct i915_request *rq); +struct dma_fence * +__i915_active_fence_set(struct i915_active_fence *active, + struct dma_fence *fence); /** - * i915_active_request_raw - return the active request + * i915_active_fence_set - updates the tracker to watch the current fence * @active - the active tracker + * @rq - the request to watch * - * i915_active_request_raw() returns the current request being tracked, or NULL. - * It does not obtain a reference on the request for the caller, so the caller - * must hold struct_mutex. + * i915_active_fence_set() watches the given @rq for completion. While + * that @rq is busy, the @active reports busy. When that @rq is signaled + * (or else retired) the @active tracker is updated to report idle. */ -static inline struct i915_request * -i915_active_request_raw(const struct i915_active_request *active, - struct mutex *mutex) -{ - return rcu_dereference_protected(active->request, - lockdep_is_held(mutex)); -} - -/** - * i915_active_request_peek - report the active request being monitored - * @active - the active tracker - * - * i915_active_request_peek() returns the current request being tracked if - * still active, or NULL. It does not obtain a reference on the request - * for the caller, so the caller must hold struct_mutex. - */ -static inline struct i915_request * -i915_active_request_peek(const struct i915_active_request *active, - struct mutex *mutex) -{ - struct i915_request *request; - - request = i915_active_request_raw(active, mutex); - if (!request || i915_request_completed(request)) - return NULL; - - return request; -} - -/** - * i915_active_request_get - return a reference to the active request - * @active - the active tracker - * - * i915_active_request_get() returns a reference to the active request, or NULL - * if the active tracker is idle. The caller must hold struct_mutex. - */ -static inline struct i915_request * -i915_active_request_get(const struct i915_active_request *active, - struct mutex *mutex) -{ - return i915_request_get(i915_active_request_peek(active, mutex)); -} - -/** - * __i915_active_request_get_rcu - return a reference to the active request - * @active - the active tracker - * - * __i915_active_request_get() returns a reference to the active request, - * or NULL if the active tracker is idle. The caller must hold the RCU read - * lock, but the returned pointer is safe to use outside of RCU. - */ -static inline struct i915_request * -__i915_active_request_get_rcu(const struct i915_active_request *active) -{ - /* - * Performing a lockless retrieval of the active request is super - * tricky. SLAB_TYPESAFE_BY_RCU merely guarantees that the backing - * slab of request objects will not be freed whilst we hold the - * RCU read lock. It does not guarantee that the request itself - * will not be freed and then *reused*. Viz, - * - * Thread A Thread B - * - * rq = active.request - * retire(rq) -> free(rq); - * (rq is now first on the slab freelist) - * active.request = NULL - * - * rq = new submission on a new object - * ref(rq) - * - * To prevent the request from being reused whilst the caller - * uses it, we take a reference like normal. Whilst acquiring - * the reference we check that it is not in a destroyed state - * (refcnt == 0). That prevents the request being reallocated - * whilst the caller holds on to it. To check that the request - * was not reallocated as we acquired the reference we have to - * check that our request remains the active request across - * the lookup, in the same manner as a seqlock. The visibility - * of the pointer versus the reference counting is controlled - * by using RCU barriers (rcu_dereference and rcu_assign_pointer). - * - * In the middle of all that, we inspect whether the request is - * complete. Retiring is lazy so the request may be completed long - * before the active tracker is updated. Querying whether the - * request is complete is far cheaper (as it involves no locked - * instructions setting cachelines to exclusive) than acquiring - * the reference, so we do it first. The RCU read lock ensures the - * pointer dereference is valid, but does not ensure that the - * seqno nor HWS is the right one! However, if the request was - * reallocated, that means the active tracker's request was complete. - * If the new request is also complete, then both are and we can - * just report the active tracker is idle. If the new request is - * incomplete, then we acquire a reference on it and check that - * it remained the active request. - * - * It is then imperative that we do not zero the request on - * reallocation, so that we can chase the dangling pointers! - * See i915_request_alloc(). - */ - do { - struct i915_request *request; - - request = rcu_dereference(active->request); - if (!request || i915_request_completed(request)) - return NULL; - - /* - * An especially silly compiler could decide to recompute the - * result of i915_request_completed, more specifically - * re-emit the load for request->fence.seqno. A race would catch - * a later seqno value, which could flip the result from true to - * false. Which means part of the instructions below might not - * be executed, while later on instructions are executed. Due to - * barriers within the refcounting the inconsistency can't reach - * past the call to i915_request_get_rcu, but not executing - * that while still executing i915_request_put() creates - * havoc enough. Prevent this with a compiler barrier. - */ - barrier(); - - request = i915_request_get_rcu(request); - - /* - * What stops the following rcu_access_pointer() from occurring - * before the above i915_request_get_rcu()? If we were - * to read the value before pausing to get the reference to - * the request, we may not notice a change in the active - * tracker. - * - * The rcu_access_pointer() is a mere compiler barrier, which - * means both the CPU and compiler are free to perform the - * memory read without constraint. The compiler only has to - * ensure that any operations after the rcu_access_pointer() - * occur afterwards in program order. This means the read may - * be performed earlier by an out-of-order CPU, or adventurous - * compiler. - * - * The atomic operation at the heart of - * i915_request_get_rcu(), see dma_fence_get_rcu(), is - * atomic_inc_not_zero() which is only a full memory barrier - * when successful. That is, if i915_request_get_rcu() - * returns the request (and so with the reference counted - * incremented) then the following read for rcu_access_pointer() - * must occur after the atomic operation and so confirm - * that this request is the one currently being tracked. - * - * The corresponding write barrier is part of - * rcu_assign_pointer(). - */ - if (!request || request == rcu_access_pointer(active->request)) - return rcu_pointer_handoff(request); - - i915_request_put(request); - } while (1); -} - +int __must_check +i915_active_fence_set(struct i915_active_fence *active, + struct i915_request *rq); /** - * i915_active_request_get_unlocked - return a reference to the active request + * i915_active_fence_get - return a reference to the active fence * @active - the active tracker * - * i915_active_request_get_unlocked() returns a reference to the active request, + * i915_active_fence_get() returns a reference to the active fence, * or NULL if the active tracker is idle. The reference is obtained under RCU, * so no locking is required by the caller. * - * The reference should be freed with i915_request_put(). + * The reference should be freed with dma_fence_put(). */ -static inline struct i915_request * -i915_active_request_get_unlocked(const struct i915_active_request *active) +static inline struct dma_fence * +i915_active_fence_get(struct i915_active_fence *active) { - struct i915_request *request; + struct dma_fence *fence; rcu_read_lock(); - request = __i915_active_request_get_rcu(active); + fence = dma_fence_get_rcu_safe(&active->fence); rcu_read_unlock(); - return request; + return fence; } /** - * i915_active_request_isset - report whether the active tracker is assigned + * i915_active_fence_isset - report whether the active tracker is assigned * @active - the active tracker * - * i915_active_request_isset() returns true if the active tracker is currently - * assigned to a request. Due to the lazy retiring, that request may be idle + * i915_active_fence_isset() returns true if the active tracker is currently + * assigned to a fence. Due to the lazy retiring, that fence may be idle * and this may report stale information. */ static inline bool -i915_active_request_isset(const struct i915_active_request *active) +i915_active_fence_isset(const struct i915_active_fence *active) { - return rcu_access_pointer(active->request); + return rcu_access_pointer(active->fence); } -/** - * i915_active_request_retire - waits until the request is retired - * @active - the active request on which to wait - * - * i915_active_request_retire() waits until the request is completed, - * and then ensures that at least the retirement handler for this - * @active tracker is called before returning. If the @active - * tracker is idle, the function returns immediately. - */ -static inline int __must_check -i915_active_request_retire(struct i915_active_request *active, - struct mutex *mutex) +static inline void +i915_active_fence_cb(struct dma_fence *fence, struct dma_fence_cb *cb) { - struct i915_request *request; - long ret; - - request = i915_active_request_raw(active, mutex); - if (!request) - return 0; - - ret = i915_request_wait(request, - I915_WAIT_INTERRUPTIBLE, - MAX_SCHEDULE_TIMEOUT); - if (ret < 0) - return ret; + struct i915_active_fence *active = + container_of(cb, typeof(*active), cb); - list_del_init(&active->link); - RCU_INIT_POINTER(active->request, NULL); - - active->retire(active, request); - - return 0; + RCU_INIT_POINTER(active->fence, NULL); } /* @@ -358,34 +162,40 @@ i915_active_request_retire(struct i915_active_request *active, * synchronisation. */ -void __i915_active_init(struct drm_i915_private *i915, - struct i915_active *ref, +void __i915_active_init(struct i915_active *ref, int (*active)(struct i915_active *ref), void (*retire)(struct i915_active *ref), struct lock_class_key *key); -#define i915_active_init(i915, ref, active, retire) do { \ +#define i915_active_init(ref, active, retire) do { \ static struct lock_class_key __key; \ \ - __i915_active_init(i915, ref, active, retire, &__key); \ + __i915_active_init(ref, active, retire, &__key); \ } while (0) int i915_active_ref(struct i915_active *ref, struct intel_timeline *tl, - struct i915_request *rq); + struct dma_fence *fence); + +static inline int +i915_active_add_request(struct i915_active *ref, struct i915_request *rq) +{ + return i915_active_ref(ref, i915_request_timeline(rq), &rq->fence); +} + +void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f); + +static inline bool i915_active_has_exclusive(struct i915_active *ref) +{ + return rcu_access_pointer(ref->excl.fence); +} int i915_active_wait(struct i915_active *ref); -int i915_request_await_active(struct i915_request *rq, - struct i915_active *ref); -int i915_request_await_active_request(struct i915_request *rq, - struct i915_active_request *active); +int i915_request_await_active(struct i915_request *rq, struct i915_active *ref); int i915_active_acquire(struct i915_active *ref); +bool i915_active_acquire_if_busy(struct i915_active *ref); void i915_active_release(struct i915_active *ref); -void __i915_active_release_nested(struct i915_active *ref, int subclass); - -bool i915_active_trygrab(struct i915_active *ref); -void i915_active_ungrab(struct i915_active *ref); static inline bool i915_active_is_idle(const struct i915_active *ref) @@ -404,4 +214,6 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref, void i915_active_acquire_barrier(struct i915_active *ref); void i915_request_add_active_barriers(struct i915_request *rq); +void i915_active_print(struct i915_active *ref, struct drm_printer *m); + #endif /* _I915_ACTIVE_H_ */ diff --git a/drivers/gpu/drm/i915/i915_active_types.h b/drivers/gpu/drm/i915/i915_active_types.h index 1854e7d168c1..96aed0ee700a 100644 --- a/drivers/gpu/drm/i915/i915_active_types.h +++ b/drivers/gpu/drm/i915/i915_active_types.h @@ -8,22 +8,18 @@ #define _I915_ACTIVE_TYPES_H_ #include <linux/atomic.h> +#include <linux/dma-fence.h> #include <linux/llist.h> #include <linux/mutex.h> #include <linux/rbtree.h> #include <linux/rcupdate.h> +#include <linux/workqueue.h> -struct drm_i915_private; -struct i915_active_request; -struct i915_request; +#include "i915_utils.h" -typedef void (*i915_active_retire_fn)(struct i915_active_request *, - struct i915_request *); - -struct i915_active_request { - struct i915_request __rcu *request; - struct list_head link; - i915_active_retire_fn retire; +struct i915_active_fence { + struct dma_fence __rcu *fence; + struct dma_fence_cb cb; #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) /* * Incorporeal! @@ -43,20 +39,30 @@ struct i915_active_request { struct active_node; +#define I915_ACTIVE_MAY_SLEEP BIT(0) + +#define __i915_active_call __aligned(4) +#define i915_active_may_sleep(fn) ptr_pack_bits(&(fn), I915_ACTIVE_MAY_SLEEP, 2) + struct i915_active { - struct drm_i915_private *i915; + atomic_t count; + struct mutex mutex; + spinlock_t tree_lock; struct active_node *cache; struct rb_root tree; - struct mutex mutex; - atomic_t count; + + /* Preallocated "exclusive" node */ + struct i915_active_fence excl; unsigned long flags; -#define I915_ACTIVE_GRAB_BIT 0 +#define I915_ACTIVE_RETIRE_SLEEPS BIT(0) int (*active)(struct i915_active *ref); void (*retire)(struct i915_active *ref); + struct work_struct work; + struct llist_head preallocated_barriers; }; diff --git a/drivers/gpu/drm/i915/i915_buddy.c b/drivers/gpu/drm/i915/i915_buddy.c index fe1871d7c126..e9d4200ce3bc 100644 --- a/drivers/gpu/drm/i915/i915_buddy.c +++ b/drivers/gpu/drm/i915/i915_buddy.c @@ -38,6 +38,7 @@ int __init i915_global_buddy_init(void) if (!global.slab_blocks) return -ENOMEM; + i915_global_register(&global.base); return 0; } diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c index 24555102e198..f24096e27bef 100644 --- a/drivers/gpu/drm/i915/i915_cmd_parser.c +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c @@ -53,13 +53,11 @@ * granting userspace undue privileges. There are three categories of privilege. * * First, commands which are explicitly defined as privileged or which should - * only be used by the kernel driver. The parser generally rejects such - * commands, though it may allow some from the drm master process. + * only be used by the kernel driver. The parser rejects such commands * * Second, commands which access registers. To support correct/enhanced * userspace functionality, particularly certain OpenGL extensions, the parser - * provides a whitelist of registers which userspace may safely access (for both - * normal and drm master processes). + * provides a whitelist of registers which userspace may safely access * * Third, commands which access privileged memory (i.e. GGTT, HWS page, etc). * The parser always rejects such commands. @@ -84,9 +82,9 @@ * in the per-engine command tables. * * Other command table entries map fairly directly to high level categories - * mentioned above: rejected, master-only, register whitelist. The parser - * implements a number of checks, including the privileged memory checks, via a - * general bitmasking mechanism. + * mentioned above: rejected, register whitelist. The parser implements a number + * of checks, including the privileged memory checks, via a general bitmasking + * mechanism. */ /* @@ -104,8 +102,6 @@ struct drm_i915_cmd_descriptor { * CMD_DESC_REJECT: The command is never allowed * CMD_DESC_REGISTER: The command should be checked against the * register whitelist for the appropriate ring - * CMD_DESC_MASTER: The command is allowed if the submitting process - * is the DRM master */ u32 flags; #define CMD_DESC_FIXED (1<<0) @@ -113,7 +109,6 @@ struct drm_i915_cmd_descriptor { #define CMD_DESC_REJECT (1<<2) #define CMD_DESC_REGISTER (1<<3) #define CMD_DESC_BITMASK (1<<4) -#define CMD_DESC_MASTER (1<<5) /* * The command's unique identification bits and the bitmask to get them. @@ -194,7 +189,7 @@ struct drm_i915_cmd_table { #define CMD(op, opm, f, lm, fl, ...) \ { \ .flags = (fl) | ((f) ? CMD_DESC_FIXED : 0), \ - .cmd = { (op), ~0u << (opm) }, \ + .cmd = { (op & ~0u << (opm)), ~0u << (opm) }, \ .length = { (lm) }, \ __VA_ARGS__ \ } @@ -209,14 +204,13 @@ struct drm_i915_cmd_table { #define R CMD_DESC_REJECT #define W CMD_DESC_REGISTER #define B CMD_DESC_BITMASK -#define M CMD_DESC_MASTER /* Command Mask Fixed Len Action ---------------------------------------------------------- */ -static const struct drm_i915_cmd_descriptor common_cmds[] = { +static const struct drm_i915_cmd_descriptor gen7_common_cmds[] = { CMD( MI_NOOP, SMI, F, 1, S ), CMD( MI_USER_INTERRUPT, SMI, F, 1, R ), - CMD( MI_WAIT_FOR_EVENT, SMI, F, 1, M ), + CMD( MI_WAIT_FOR_EVENT, SMI, F, 1, R ), CMD( MI_ARB_CHECK, SMI, F, 1, S ), CMD( MI_REPORT_HEAD, SMI, F, 1, S ), CMD( MI_SUSPEND_FLUSH, SMI, F, 1, S ), @@ -246,7 +240,7 @@ static const struct drm_i915_cmd_descriptor common_cmds[] = { CMD( MI_BATCH_BUFFER_START, SMI, !F, 0xFF, S ), }; -static const struct drm_i915_cmd_descriptor render_cmds[] = { +static const struct drm_i915_cmd_descriptor gen7_render_cmds[] = { CMD( MI_FLUSH, SMI, F, 1, S ), CMD( MI_ARB_ON_OFF, SMI, F, 1, R ), CMD( MI_PREDICATE, SMI, F, 1, S ), @@ -313,7 +307,7 @@ static const struct drm_i915_cmd_descriptor hsw_render_cmds[] = { CMD( MI_URB_ATOMIC_ALLOC, SMI, F, 1, S ), CMD( MI_SET_APPID, SMI, F, 1, S ), CMD( MI_RS_CONTEXT, SMI, F, 1, S ), - CMD( MI_LOAD_SCAN_LINES_INCL, SMI, !F, 0x3F, M ), + CMD( MI_LOAD_SCAN_LINES_INCL, SMI, !F, 0x3F, R ), CMD( MI_LOAD_SCAN_LINES_EXCL, SMI, !F, 0x3F, R ), CMD( MI_LOAD_REGISTER_REG, SMI, !F, 0xFF, W, .reg = { .offset = 1, .mask = 0x007FFFFC, .step = 1 } ), @@ -330,7 +324,7 @@ static const struct drm_i915_cmd_descriptor hsw_render_cmds[] = { CMD( GFX_OP_3DSTATE_BINDING_TABLE_EDIT_PS, S3D, !F, 0x1FF, S ), }; -static const struct drm_i915_cmd_descriptor video_cmds[] = { +static const struct drm_i915_cmd_descriptor gen7_video_cmds[] = { CMD( MI_ARB_ON_OFF, SMI, F, 1, R ), CMD( MI_SET_APPID, SMI, F, 1, S ), CMD( MI_STORE_DWORD_IMM, SMI, !F, 0xFF, B, @@ -374,7 +368,7 @@ static const struct drm_i915_cmd_descriptor video_cmds[] = { CMD( MFX_WAIT, SMFX, F, 1, S ), }; -static const struct drm_i915_cmd_descriptor vecs_cmds[] = { +static const struct drm_i915_cmd_descriptor gen7_vecs_cmds[] = { CMD( MI_ARB_ON_OFF, SMI, F, 1, R ), CMD( MI_SET_APPID, SMI, F, 1, S ), CMD( MI_STORE_DWORD_IMM, SMI, !F, 0xFF, B, @@ -412,7 +406,7 @@ static const struct drm_i915_cmd_descriptor vecs_cmds[] = { }}, ), }; -static const struct drm_i915_cmd_descriptor blt_cmds[] = { +static const struct drm_i915_cmd_descriptor gen7_blt_cmds[] = { CMD( MI_DISPLAY_FLIP, SMI, !F, 0xFF, R ), CMD( MI_STORE_DWORD_IMM, SMI, !F, 0x3FF, B, .bits = {{ @@ -446,10 +440,64 @@ static const struct drm_i915_cmd_descriptor blt_cmds[] = { }; static const struct drm_i915_cmd_descriptor hsw_blt_cmds[] = { - CMD( MI_LOAD_SCAN_LINES_INCL, SMI, !F, 0x3F, M ), + CMD( MI_LOAD_SCAN_LINES_INCL, SMI, !F, 0x3F, R ), CMD( MI_LOAD_SCAN_LINES_EXCL, SMI, !F, 0x3F, R ), }; +/* + * For Gen9 we can still rely on the h/w to enforce cmd security, and only + * need to re-enforce the register access checks. We therefore only need to + * teach the cmdparser how to find the end of each command, and identify + * register accesses. The table doesn't need to reject any commands, and so + * the only commands listed here are: + * 1) Those that touch registers + * 2) Those that do not have the default 8-bit length + * + * Note that the default MI length mask chosen for this table is 0xFF, not + * the 0x3F used on older devices. This is because the vast majority of MI + * cmds on Gen9 use a standard 8-bit Length field. + * All the Gen9 blitter instructions are standard 0xFF length mask, and + * none allow access to non-general registers, so in fact no BLT cmds are + * included in the table at all. + * + */ +static const struct drm_i915_cmd_descriptor gen9_blt_cmds[] = { + CMD( MI_NOOP, SMI, F, 1, S ), + CMD( MI_USER_INTERRUPT, SMI, F, 1, S ), + CMD( MI_WAIT_FOR_EVENT, SMI, F, 1, S ), + CMD( MI_FLUSH, SMI, F, 1, S ), + CMD( MI_ARB_CHECK, SMI, F, 1, S ), + CMD( MI_REPORT_HEAD, SMI, F, 1, S ), + CMD( MI_ARB_ON_OFF, SMI, F, 1, S ), + CMD( MI_SUSPEND_FLUSH, SMI, F, 1, S ), + CMD( MI_LOAD_SCAN_LINES_INCL, SMI, !F, 0x3F, S ), + CMD( MI_LOAD_SCAN_LINES_EXCL, SMI, !F, 0x3F, S ), + CMD( MI_STORE_DWORD_IMM, SMI, !F, 0x3FF, S ), + CMD( MI_LOAD_REGISTER_IMM(1), SMI, !F, 0xFF, W, + .reg = { .offset = 1, .mask = 0x007FFFFC, .step = 2 } ), + CMD( MI_UPDATE_GTT, SMI, !F, 0x3FF, S ), + CMD( MI_STORE_REGISTER_MEM_GEN8, SMI, F, 4, W, + .reg = { .offset = 1, .mask = 0x007FFFFC } ), + CMD( MI_FLUSH_DW, SMI, !F, 0x3F, S ), + CMD( MI_LOAD_REGISTER_MEM_GEN8, SMI, F, 4, W, + .reg = { .offset = 1, .mask = 0x007FFFFC } ), + CMD( MI_LOAD_REGISTER_REG, SMI, !F, 0xFF, W, + .reg = { .offset = 1, .mask = 0x007FFFFC, .step = 1 } ), + + /* + * We allow BB_START but apply further checks. We just sanitize the + * basic fields here. + */ +#define MI_BB_START_OPERAND_MASK GENMASK(SMI-1, 0) +#define MI_BB_START_OPERAND_EXPECT (MI_BATCH_PPGTT_HSW | 1) + CMD( MI_BATCH_BUFFER_START_GEN8, SMI, !F, 0xFF, B, + .bits = {{ + .offset = 0, + .mask = MI_BB_START_OPERAND_MASK, + .expected = MI_BB_START_OPERAND_EXPECT, + }}, ), +}; + static const struct drm_i915_cmd_descriptor noop_desc = CMD(MI_NOOP, SMI, F, 1, S); @@ -463,40 +511,44 @@ static const struct drm_i915_cmd_descriptor noop_desc = #undef R #undef W #undef B -#undef M -static const struct drm_i915_cmd_table gen7_render_cmds[] = { - { common_cmds, ARRAY_SIZE(common_cmds) }, - { render_cmds, ARRAY_SIZE(render_cmds) }, +static const struct drm_i915_cmd_table gen7_render_cmd_table[] = { + { gen7_common_cmds, ARRAY_SIZE(gen7_common_cmds) }, + { gen7_render_cmds, ARRAY_SIZE(gen7_render_cmds) }, }; -static const struct drm_i915_cmd_table hsw_render_ring_cmds[] = { - { common_cmds, ARRAY_SIZE(common_cmds) }, - { render_cmds, ARRAY_SIZE(render_cmds) }, +static const struct drm_i915_cmd_table hsw_render_ring_cmd_table[] = { + { gen7_common_cmds, ARRAY_SIZE(gen7_common_cmds) }, + { gen7_render_cmds, ARRAY_SIZE(gen7_render_cmds) }, { hsw_render_cmds, ARRAY_SIZE(hsw_render_cmds) }, }; -static const struct drm_i915_cmd_table gen7_video_cmds[] = { - { common_cmds, ARRAY_SIZE(common_cmds) }, - { video_cmds, ARRAY_SIZE(video_cmds) }, +static const struct drm_i915_cmd_table gen7_video_cmd_table[] = { + { gen7_common_cmds, ARRAY_SIZE(gen7_common_cmds) }, + { gen7_video_cmds, ARRAY_SIZE(gen7_video_cmds) }, }; -static const struct drm_i915_cmd_table hsw_vebox_cmds[] = { - { common_cmds, ARRAY_SIZE(common_cmds) }, - { vecs_cmds, ARRAY_SIZE(vecs_cmds) }, +static const struct drm_i915_cmd_table hsw_vebox_cmd_table[] = { + { gen7_common_cmds, ARRAY_SIZE(gen7_common_cmds) }, + { gen7_vecs_cmds, ARRAY_SIZE(gen7_vecs_cmds) }, }; -static const struct drm_i915_cmd_table gen7_blt_cmds[] = { - { common_cmds, ARRAY_SIZE(common_cmds) }, - { blt_cmds, ARRAY_SIZE(blt_cmds) }, +static const struct drm_i915_cmd_table gen7_blt_cmd_table[] = { + { gen7_common_cmds, ARRAY_SIZE(gen7_common_cmds) }, + { gen7_blt_cmds, ARRAY_SIZE(gen7_blt_cmds) }, }; -static const struct drm_i915_cmd_table hsw_blt_ring_cmds[] = { - { common_cmds, ARRAY_SIZE(common_cmds) }, - { blt_cmds, ARRAY_SIZE(blt_cmds) }, +static const struct drm_i915_cmd_table hsw_blt_ring_cmd_table[] = { + { gen7_common_cmds, ARRAY_SIZE(gen7_common_cmds) }, + { gen7_blt_cmds, ARRAY_SIZE(gen7_blt_cmds) }, { hsw_blt_cmds, ARRAY_SIZE(hsw_blt_cmds) }, }; +static const struct drm_i915_cmd_table gen9_blt_cmd_table[] = { + { gen9_blt_cmds, ARRAY_SIZE(gen9_blt_cmds) }, +}; + + /* * Register whitelists, sorted by increasing register offset. */ @@ -612,17 +664,27 @@ static const struct drm_i915_reg_descriptor gen7_blt_regs[] = { REG64_IDX(RING_TIMESTAMP, BLT_RING_BASE), }; -static const struct drm_i915_reg_descriptor ivb_master_regs[] = { - REG32(FORCEWAKE_MT), - REG32(DERRMR), - REG32(GEN7_PIPE_DE_LOAD_SL(PIPE_A)), - REG32(GEN7_PIPE_DE_LOAD_SL(PIPE_B)), - REG32(GEN7_PIPE_DE_LOAD_SL(PIPE_C)), -}; - -static const struct drm_i915_reg_descriptor hsw_master_regs[] = { - REG32(FORCEWAKE_MT), - REG32(DERRMR), +static const struct drm_i915_reg_descriptor gen9_blt_regs[] = { + REG64_IDX(RING_TIMESTAMP, RENDER_RING_BASE), + REG64_IDX(RING_TIMESTAMP, BSD_RING_BASE), + REG32(BCS_SWCTRL), + REG64_IDX(RING_TIMESTAMP, BLT_RING_BASE), + REG64_IDX(BCS_GPR, 0), + REG64_IDX(BCS_GPR, 1), + REG64_IDX(BCS_GPR, 2), + REG64_IDX(BCS_GPR, 3), + REG64_IDX(BCS_GPR, 4), + REG64_IDX(BCS_GPR, 5), + REG64_IDX(BCS_GPR, 6), + REG64_IDX(BCS_GPR, 7), + REG64_IDX(BCS_GPR, 8), + REG64_IDX(BCS_GPR, 9), + REG64_IDX(BCS_GPR, 10), + REG64_IDX(BCS_GPR, 11), + REG64_IDX(BCS_GPR, 12), + REG64_IDX(BCS_GPR, 13), + REG64_IDX(BCS_GPR, 14), + REG64_IDX(BCS_GPR, 15), }; #undef REG64 @@ -631,28 +693,27 @@ static const struct drm_i915_reg_descriptor hsw_master_regs[] = { struct drm_i915_reg_table { const struct drm_i915_reg_descriptor *regs; int num_regs; - bool master; }; static const struct drm_i915_reg_table ivb_render_reg_tables[] = { - { gen7_render_regs, ARRAY_SIZE(gen7_render_regs), false }, - { ivb_master_regs, ARRAY_SIZE(ivb_master_regs), true }, + { gen7_render_regs, ARRAY_SIZE(gen7_render_regs) }, }; static const struct drm_i915_reg_table ivb_blt_reg_tables[] = { - { gen7_blt_regs, ARRAY_SIZE(gen7_blt_regs), false }, - { ivb_master_regs, ARRAY_SIZE(ivb_master_regs), true }, + { gen7_blt_regs, ARRAY_SIZE(gen7_blt_regs) }, }; static const struct drm_i915_reg_table hsw_render_reg_tables[] = { - { gen7_render_regs, ARRAY_SIZE(gen7_render_regs), false }, - { hsw_render_regs, ARRAY_SIZE(hsw_render_regs), false }, - { hsw_master_regs, ARRAY_SIZE(hsw_master_regs), true }, + { gen7_render_regs, ARRAY_SIZE(gen7_render_regs) }, + { hsw_render_regs, ARRAY_SIZE(hsw_render_regs) }, }; static const struct drm_i915_reg_table hsw_blt_reg_tables[] = { - { gen7_blt_regs, ARRAY_SIZE(gen7_blt_regs), false }, - { hsw_master_regs, ARRAY_SIZE(hsw_master_regs), true }, + { gen7_blt_regs, ARRAY_SIZE(gen7_blt_regs) }, +}; + +static const struct drm_i915_reg_table gen9_blt_reg_tables[] = { + { gen9_blt_regs, ARRAY_SIZE(gen9_blt_regs) }, }; static u32 gen7_render_get_cmd_length_mask(u32 cmd_header) @@ -710,6 +771,17 @@ static u32 gen7_blt_get_cmd_length_mask(u32 cmd_header) return 0; } +static u32 gen9_blt_get_cmd_length_mask(u32 cmd_header) +{ + u32 client = cmd_header >> INSTR_CLIENT_SHIFT; + + if (client == INSTR_MI_CLIENT || client == INSTR_BC_CLIENT) + return 0xFF; + + DRM_DEBUG_DRIVER("CMD: Abnormal blt cmd length! 0x%08X\n", cmd_header); + return 0; +} + static bool validate_cmds_sorted(const struct intel_engine_cs *engine, const struct drm_i915_cmd_table *cmd_tables, int cmd_table_count) @@ -867,18 +939,19 @@ void intel_engine_init_cmd_parser(struct intel_engine_cs *engine) int cmd_table_count; int ret; - if (!IS_GEN(engine->i915, 7)) + if (!IS_GEN(engine->i915, 7) && !(IS_GEN(engine->i915, 9) && + engine->class == COPY_ENGINE_CLASS)) return; switch (engine->class) { case RENDER_CLASS: if (IS_HASWELL(engine->i915)) { - cmd_tables = hsw_render_ring_cmds; + cmd_tables = hsw_render_ring_cmd_table; cmd_table_count = - ARRAY_SIZE(hsw_render_ring_cmds); + ARRAY_SIZE(hsw_render_ring_cmd_table); } else { - cmd_tables = gen7_render_cmds; - cmd_table_count = ARRAY_SIZE(gen7_render_cmds); + cmd_tables = gen7_render_cmd_table; + cmd_table_count = ARRAY_SIZE(gen7_render_cmd_table); } if (IS_HASWELL(engine->i915)) { @@ -888,36 +961,46 @@ void intel_engine_init_cmd_parser(struct intel_engine_cs *engine) engine->reg_tables = ivb_render_reg_tables; engine->reg_table_count = ARRAY_SIZE(ivb_render_reg_tables); } - engine->get_cmd_length_mask = gen7_render_get_cmd_length_mask; break; case VIDEO_DECODE_CLASS: - cmd_tables = gen7_video_cmds; - cmd_table_count = ARRAY_SIZE(gen7_video_cmds); + cmd_tables = gen7_video_cmd_table; + cmd_table_count = ARRAY_SIZE(gen7_video_cmd_table); engine->get_cmd_length_mask = gen7_bsd_get_cmd_length_mask; break; case COPY_ENGINE_CLASS: - if (IS_HASWELL(engine->i915)) { - cmd_tables = hsw_blt_ring_cmds; - cmd_table_count = ARRAY_SIZE(hsw_blt_ring_cmds); + engine->get_cmd_length_mask = gen7_blt_get_cmd_length_mask; + if (IS_GEN(engine->i915, 9)) { + cmd_tables = gen9_blt_cmd_table; + cmd_table_count = ARRAY_SIZE(gen9_blt_cmd_table); + engine->get_cmd_length_mask = + gen9_blt_get_cmd_length_mask; + + /* BCS Engine unsafe without parser */ + engine->flags |= I915_ENGINE_REQUIRES_CMD_PARSER; + } else if (IS_HASWELL(engine->i915)) { + cmd_tables = hsw_blt_ring_cmd_table; + cmd_table_count = ARRAY_SIZE(hsw_blt_ring_cmd_table); } else { - cmd_tables = gen7_blt_cmds; - cmd_table_count = ARRAY_SIZE(gen7_blt_cmds); + cmd_tables = gen7_blt_cmd_table; + cmd_table_count = ARRAY_SIZE(gen7_blt_cmd_table); } - if (IS_HASWELL(engine->i915)) { + if (IS_GEN(engine->i915, 9)) { + engine->reg_tables = gen9_blt_reg_tables; + engine->reg_table_count = + ARRAY_SIZE(gen9_blt_reg_tables); + } else if (IS_HASWELL(engine->i915)) { engine->reg_tables = hsw_blt_reg_tables; engine->reg_table_count = ARRAY_SIZE(hsw_blt_reg_tables); } else { engine->reg_tables = ivb_blt_reg_tables; engine->reg_table_count = ARRAY_SIZE(ivb_blt_reg_tables); } - - engine->get_cmd_length_mask = gen7_blt_get_cmd_length_mask; break; case VIDEO_ENHANCEMENT_CLASS: - cmd_tables = hsw_vebox_cmds; - cmd_table_count = ARRAY_SIZE(hsw_vebox_cmds); + cmd_tables = hsw_vebox_cmd_table; + cmd_table_count = ARRAY_SIZE(hsw_vebox_cmd_table); /* VECS can use the same length_mask function as VCS */ engine->get_cmd_length_mask = gen7_bsd_get_cmd_length_mask; break; @@ -943,7 +1026,7 @@ void intel_engine_init_cmd_parser(struct intel_engine_cs *engine) return; } - engine->flags |= I915_ENGINE_NEEDS_CMD_PARSER; + engine->flags |= I915_ENGINE_USING_CMD_PARSER; } /** @@ -955,7 +1038,7 @@ void intel_engine_init_cmd_parser(struct intel_engine_cs *engine) */ void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine) { - if (!intel_engine_needs_cmd_parser(engine)) + if (!intel_engine_using_cmd_parser(engine)) return; fini_hash_table(engine); @@ -1029,22 +1112,16 @@ __find_reg(const struct drm_i915_reg_descriptor *table, int count, u32 addr) } static const struct drm_i915_reg_descriptor * -find_reg(const struct intel_engine_cs *engine, bool is_master, u32 addr) +find_reg(const struct intel_engine_cs *engine, u32 addr) { const struct drm_i915_reg_table *table = engine->reg_tables; + const struct drm_i915_reg_descriptor *reg = NULL; int count = engine->reg_table_count; - for (; count > 0; ++table, --count) { - if (!table->master || is_master) { - const struct drm_i915_reg_descriptor *reg; + for (; !reg && (count > 0); ++table, --count) + reg = __find_reg(table->regs, table->num_regs, addr); - reg = __find_reg(table->regs, table->num_regs, addr); - if (reg != NULL) - return reg; - } - } - - return NULL; + return reg; } /* Returns a vmap'd pointer to dst_obj, which the caller must unmap */ @@ -1128,8 +1205,7 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj, static bool check_cmd(const struct intel_engine_cs *engine, const struct drm_i915_cmd_descriptor *desc, - const u32 *cmd, u32 length, - const bool is_master) + const u32 *cmd, u32 length) { if (desc->flags & CMD_DESC_SKIP) return true; @@ -1139,12 +1215,6 @@ static bool check_cmd(const struct intel_engine_cs *engine, return false; } - if ((desc->flags & CMD_DESC_MASTER) && !is_master) { - DRM_DEBUG_DRIVER("CMD: Rejected master-only command: 0x%08X\n", - *cmd); - return false; - } - if (desc->flags & CMD_DESC_REGISTER) { /* * Get the distance between individual register offset @@ -1158,7 +1228,7 @@ static bool check_cmd(const struct intel_engine_cs *engine, offset += step) { const u32 reg_addr = cmd[offset] & desc->reg.mask; const struct drm_i915_reg_descriptor *reg = - find_reg(engine, is_master, reg_addr); + find_reg(engine, reg_addr); if (!reg) { DRM_DEBUG_DRIVER("CMD: Rejected register 0x%08X in command: 0x%08X (%s)\n", @@ -1236,16 +1306,112 @@ static bool check_cmd(const struct intel_engine_cs *engine, return true; } +static int check_bbstart(const struct i915_gem_context *ctx, + u32 *cmd, u32 offset, u32 length, + u32 batch_len, + u64 batch_start, + u64 shadow_batch_start) +{ + u64 jump_offset, jump_target; + u32 target_cmd_offset, target_cmd_index; + + /* For igt compatibility on older platforms */ + if (CMDPARSER_USES_GGTT(ctx->i915)) { + DRM_DEBUG("CMD: Rejecting BB_START for ggtt based submission\n"); + return -EACCES; + } + + if (length != 3) { + DRM_DEBUG("CMD: Recursive BB_START with bad length(%u)\n", + length); + return -EINVAL; + } + + jump_target = *(u64*)(cmd+1); + jump_offset = jump_target - batch_start; + + /* + * Any underflow of jump_target is guaranteed to be outside the range + * of a u32, so >= test catches both too large and too small + */ + if (jump_offset >= batch_len) { + DRM_DEBUG("CMD: BB_START to 0x%llx jumps out of BB\n", + jump_target); + return -EINVAL; + } + + /* + * This cannot overflow a u32 because we already checked jump_offset + * is within the BB, and the batch_len is a u32 + */ + target_cmd_offset = lower_32_bits(jump_offset); + target_cmd_index = target_cmd_offset / sizeof(u32); + + *(u64*)(cmd + 1) = shadow_batch_start + target_cmd_offset; + + if (target_cmd_index == offset) + return 0; + + if (ctx->jump_whitelist_cmds <= target_cmd_index) { + DRM_DEBUG("CMD: Rejecting BB_START - truncated whitelist array\n"); + return -EINVAL; + } else if (!test_bit(target_cmd_index, ctx->jump_whitelist)) { + DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n", + jump_target); + return -EINVAL; + } + + return 0; +} + +static void init_whitelist(struct i915_gem_context *ctx, u32 batch_len) +{ + const u32 batch_cmds = DIV_ROUND_UP(batch_len, sizeof(u32)); + const u32 exact_size = BITS_TO_LONGS(batch_cmds); + u32 next_size = BITS_TO_LONGS(roundup_pow_of_two(batch_cmds)); + unsigned long *next_whitelist; + + if (CMDPARSER_USES_GGTT(ctx->i915)) + return; + + if (batch_cmds <= ctx->jump_whitelist_cmds) { + bitmap_zero(ctx->jump_whitelist, batch_cmds); + return; + } + +again: + next_whitelist = kcalloc(next_size, sizeof(long), GFP_KERNEL); + if (next_whitelist) { + kfree(ctx->jump_whitelist); + ctx->jump_whitelist = next_whitelist; + ctx->jump_whitelist_cmds = + next_size * BITS_PER_BYTE * sizeof(long); + return; + } + + if (next_size > exact_size) { + next_size = exact_size; + goto again; + } + + DRM_DEBUG("CMD: Failed to extend whitelist. BB_START may be disallowed\n"); + bitmap_zero(ctx->jump_whitelist, ctx->jump_whitelist_cmds); + + return; +} + #define LENGTH_BIAS 2 /** * i915_parse_cmds() - parse a submitted batch buffer for privilege violations + * @ctx: the context in which the batch is to execute * @engine: the engine on which the batch is to execute * @batch_obj: the batch buffer in question - * @shadow_batch_obj: copy of the batch buffer in question + * @batch_start: Canonical base address of batch * @batch_start_offset: byte offset in the batch at which execution starts * @batch_len: length of the commands in batch_obj - * @is_master: is the submitting process the drm master? + * @shadow_batch_obj: copy of the batch buffer in question + * @shadow_batch_start: Canonical base address of shadow_batch_obj * * Parses the specified batch buffer looking for privilege violations as * described in the overview. @@ -1253,14 +1419,17 @@ static bool check_cmd(const struct intel_engine_cs *engine, * Return: non-zero if the parser finds violations or otherwise fails; -EACCES * if the batch appears legal but should use hardware parsing */ -int intel_engine_cmd_parser(struct intel_engine_cs *engine, + +int intel_engine_cmd_parser(struct i915_gem_context *ctx, + struct intel_engine_cs *engine, struct drm_i915_gem_object *batch_obj, - struct drm_i915_gem_object *shadow_batch_obj, + u64 batch_start, u32 batch_start_offset, u32 batch_len, - bool is_master) + struct drm_i915_gem_object *shadow_batch_obj, + u64 shadow_batch_start) { - u32 *cmd, *batch_end; + u32 *cmd, *batch_end, offset = 0; struct drm_i915_cmd_descriptor default_desc = noop_desc; const struct drm_i915_cmd_descriptor *desc = &default_desc; bool needs_clflush_after = false; @@ -1274,6 +1443,8 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine, return PTR_ERR(cmd); } + init_whitelist(ctx, batch_len); + /* * We use the batch length as size because the shadow object is as * large or larger and copy_batch() will write MI_NOPs to the extra @@ -1283,31 +1454,15 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine, do { u32 length; - if (*cmd == MI_BATCH_BUFFER_END) { - if (needs_clflush_after) { - void *ptr = page_mask_bits(shadow_batch_obj->mm.mapping); - drm_clflush_virt_range(ptr, - (void *)(cmd + 1) - ptr); - } + if (*cmd == MI_BATCH_BUFFER_END) break; - } desc = find_cmd(engine, *cmd, desc, &default_desc); if (!desc) { DRM_DEBUG_DRIVER("CMD: Unrecognized command: 0x%08X\n", *cmd); ret = -EINVAL; - break; - } - - /* - * If the batch buffer contains a chained batch, return an - * error that tells the caller to abort and dispatch the - * workload as a non-secure batch. - */ - if (desc->cmd.value == MI_BATCH_BUFFER_START) { - ret = -EACCES; - break; + goto err; } if (desc->flags & CMD_DESC_FIXED) @@ -1321,22 +1476,43 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine, length, batch_end - cmd); ret = -EINVAL; - break; + goto err; } - if (!check_cmd(engine, desc, cmd, length, is_master)) { + if (!check_cmd(engine, desc, cmd, length)) { ret = -EACCES; + goto err; + } + + if (desc->cmd.value == MI_BATCH_BUFFER_START) { + ret = check_bbstart(ctx, cmd, offset, length, + batch_len, batch_start, + shadow_batch_start); + + if (ret) + goto err; break; } + if (ctx->jump_whitelist_cmds > offset) + set_bit(offset, ctx->jump_whitelist); + cmd += length; + offset += length; if (cmd >= batch_end) { DRM_DEBUG_DRIVER("CMD: Got to the end of the buffer w/o a BBE cmd!\n"); ret = -EINVAL; - break; + goto err; } } while (1); + if (needs_clflush_after) { + void *ptr = page_mask_bits(shadow_batch_obj->mm.mapping); + + drm_clflush_virt_range(ptr, (void *)(cmd + 1) - ptr); + } + +err: i915_gem_object_unpin_map(shadow_batch_obj); return ret; } @@ -1357,7 +1533,7 @@ int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv) /* If the command parser is not enabled, report 0 - unsupported */ for_each_uabi_engine(engine, dev_priv) { - if (intel_engine_needs_cmd_parser(engine)) { + if (intel_engine_using_cmd_parser(engine)) { active = true; break; } @@ -1382,6 +1558,7 @@ int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv) * the parser enabled. * 9. Don't whitelist or handle oacontrol specially, as ownership * for oacontrol state is moving to i915-perf. + * 10. Support for Gen9 BCS Parsing */ - return 9; + return 10; } diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index b0f51591f2e4..8016484ebcd3 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -41,7 +41,10 @@ #include "gem/i915_gem_context.h" #include "gt/intel_gt_pm.h" +#include "gt/intel_gt_requests.h" #include "gt/intel_reset.h" +#include "gt/intel_rc6.h" +#include "gt/intel_rps.h" #include "gt/uc/intel_guc_submission.h" #include "i915_debugfs.h" @@ -61,11 +64,18 @@ static int i915_capabilities(struct seq_file *m, void *data) struct drm_i915_private *dev_priv = node_to_i915(m->private); const struct intel_device_info *info = INTEL_INFO(dev_priv); struct drm_printer p = drm_seq_file_printer(m); + const char *msg; seq_printf(m, "gen: %d\n", INTEL_GEN(dev_priv)); seq_printf(m, "platform: %s\n", intel_platform_name(info->platform)); seq_printf(m, "pch: %d\n", INTEL_PCH_TYPE(dev_priv)); + msg = "n/a"; +#ifdef CONFIG_INTEL_IOMMU + msg = enableddisabled(intel_iommu_gfx_mapped); +#endif + seq_printf(m, "iommu: %s\n", msg); + intel_device_info_dump_flags(info, &p); intel_device_info_dump_runtime(RUNTIME_INFO(dev_priv), &p); intel_driver_caps_print(&dev_priv->caps, &p); @@ -77,11 +87,6 @@ static int i915_capabilities(struct seq_file *m, void *data) return 0; } -static char get_pin_flag(struct drm_i915_gem_object *obj) -{ - return obj->pin_global ? 'p' : ' '; -} - static char get_tiling_flag(struct drm_i915_gem_object *obj) { switch (i915_gem_object_get_tiling(obj)) { @@ -140,9 +145,8 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj) struct i915_vma *vma; int pin_count = 0; - seq_printf(m, "%pK: %c%c%c%c %8zdKiB %02x %02x %s%s%s", + seq_printf(m, "%pK: %c%c%c %8zdKiB %02x %02x %s%s%s", &obj->base, - get_pin_flag(obj), get_tiling_flag(obj), get_global_flag(obj), get_pin_mapped_flag(obj), @@ -221,8 +225,8 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj) seq_printf(m, " (pinned x %d)", pin_count); if (obj->stolen) seq_printf(m, " (stolen: %08llx)", obj->stolen->start); - if (obj->pin_global) - seq_printf(m, " (global)"); + if (i915_gem_object_is_framebuffer(obj)) + seq_printf(m, " (fb)"); engine = i915_gem_object_last_write_engine(obj); if (engine) @@ -243,6 +247,9 @@ static int per_file_stats(int id, void *ptr, void *data) struct file_stats *stats = data; struct i915_vma *vma; + if (!kref_get_unless_zero(&obj->base.refcount)) + return 0; + stats->count++; stats->total += obj->base.size; if (!atomic_read(&obj->bind_count)) @@ -290,6 +297,7 @@ static int per_file_stats(int id, void *ptr, void *data) } spin_unlock(&obj->vma.lock); + i915_gem_object_put(obj); return 0; } @@ -309,34 +317,44 @@ static void print_context_stats(struct seq_file *m, struct drm_i915_private *i915) { struct file_stats kstats = {}; - struct i915_gem_context *ctx; + struct i915_gem_context *ctx, *cn; - list_for_each_entry(ctx, &i915->contexts.list, link) { + spin_lock(&i915->gem.contexts.lock); + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) { struct i915_gem_engines_iter it; struct intel_context *ce; + if (!kref_get_unless_zero(&ctx->ref)) + continue; + + spin_unlock(&i915->gem.contexts.lock); + for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { intel_context_lock_pinned(ce); if (intel_context_is_pinned(ce)) { + rcu_read_lock(); if (ce->state) per_file_stats(0, ce->state->obj, &kstats); per_file_stats(0, ce->ring->vma->obj, &kstats); + rcu_read_unlock(); } intel_context_unlock_pinned(ce); } i915_gem_context_unlock_engines(ctx); if (!IS_ERR_OR_NULL(ctx->file_priv)) { - struct file_stats stats = { .vm = ctx->vm, }; + struct file_stats stats = { + .vm = rcu_access_pointer(ctx->vm), + }; struct drm_file *file = ctx->file_priv->file; struct task_struct *task; char name[80]; - spin_lock(&file->table_lock); + rcu_read_lock(); idr_for_each(&file->object_idr, per_file_stats, &stats); - spin_unlock(&file->table_lock); + rcu_read_unlock(); rcu_read_lock(); task = pid_task(ctx->pid ?: file->pid, PIDTYPE_PID); @@ -346,7 +364,12 @@ static void print_context_stats(struct seq_file *m, print_file_stats(m, name, stats); } + + spin_lock(&i915->gem.contexts.lock); + list_safe_reset_next(ctx, cn, link); + i915_gem_context_put(ctx); } + spin_unlock(&i915->gem.contexts.lock); print_file_stats(m, "[k]contexts", kstats); } @@ -354,7 +377,6 @@ static void print_context_stats(struct seq_file *m, static int i915_gem_object_info(struct seq_file *m, void *data) { struct drm_i915_private *i915 = node_to_i915(m->private); - int ret; seq_printf(m, "%u shrinkable [%u free] objects, %llu bytes\n", i915->mm.shrink_count, @@ -363,12 +385,7 @@ static int i915_gem_object_info(struct seq_file *m, void *data) seq_putc(m, '\n'); - ret = mutex_lock_interruptible(&i915->drm.struct_mutex); - if (ret) - return ret; - print_context_stats(m, i915); - mutex_unlock(&i915->drm.struct_mutex); return 0; } @@ -376,7 +393,7 @@ static int i915_gem_object_info(struct seq_file *m, void *data) static void gen8_display_interrupt_info(struct seq_file *m) { struct drm_i915_private *dev_priv = node_to_i915(m->private); - int pipe; + enum pipe pipe; for_each_pipe(dev_priv, pipe) { enum intel_display_power_domain power_domain; @@ -527,6 +544,8 @@ static int i915_interrupt_info(struct seq_file *m, void *data) gen8_display_interrupt_info(m); } else if (IS_VALLEYVIEW(dev_priv)) { + intel_wakeref_t pref; + seq_printf(m, "Display IER:\t%08x\n", I915_READ(VLV_IER)); seq_printf(m, "Display IIR:\t%08x\n", @@ -537,7 +556,6 @@ static int i915_interrupt_info(struct seq_file *m, void *data) I915_READ(VLV_IMR)); for_each_pipe(dev_priv, pipe) { enum intel_display_power_domain power_domain; - intel_wakeref_t pref; power_domain = POWER_DOMAIN_PIPE(pipe); pref = intel_display_power_get_if_enabled(dev_priv, @@ -571,12 +589,14 @@ static int i915_interrupt_info(struct seq_file *m, void *data) seq_printf(m, "PM IMR:\t\t%08x\n", I915_READ(GEN6_PMIMR)); + pref = intel_display_power_get(dev_priv, POWER_DOMAIN_INIT); seq_printf(m, "Port hotplug:\t%08x\n", I915_READ(PORT_HOTPLUG_EN)); seq_printf(m, "DPFLIPSTAT:\t%08x\n", I915_READ(VLV_DPFLIPSTAT)); seq_printf(m, "DPINVGTT:\t%08x\n", I915_READ(DPINVGTT)); + intel_display_power_put(dev_priv, POWER_DOMAIN_INIT, pref); } else if (!HAS_PCH_SPLIT(dev_priv)) { seq_printf(m, "Interrupt enable: %08x\n", @@ -772,7 +792,7 @@ static int i915_frequency_info(struct seq_file *m, void *unused) { struct drm_i915_private *dev_priv = node_to_i915(m->private); struct intel_uncore *uncore = &dev_priv->uncore; - struct intel_rps *rps = &dev_priv->gt_pm.rps; + struct intel_rps *rps = &dev_priv->gt.rps; intel_wakeref_t wakeref; int ret = 0; @@ -808,23 +828,23 @@ static int i915_frequency_info(struct seq_file *m, void *unused) seq_printf(m, "DDR freq: %d MHz\n", dev_priv->mem_freq); seq_printf(m, "actual GPU freq: %d MHz\n", - intel_gpu_freq(dev_priv, (freq_sts >> 8) & 0xff)); + intel_gpu_freq(rps, (freq_sts >> 8) & 0xff)); seq_printf(m, "current GPU freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->cur_freq)); + intel_gpu_freq(rps, rps->cur_freq)); seq_printf(m, "max GPU freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->max_freq)); + intel_gpu_freq(rps, rps->max_freq)); seq_printf(m, "min GPU freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->min_freq)); + intel_gpu_freq(rps, rps->min_freq)); seq_printf(m, "idle GPU freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->idle_freq)); + intel_gpu_freq(rps, rps->idle_freq)); seq_printf(m, "efficient (RPe) frequency: %d MHz\n", - intel_gpu_freq(dev_priv, rps->efficient_freq)); + intel_gpu_freq(rps, rps->efficient_freq)); } else if (INTEL_GEN(dev_priv) >= 6) { u32 rp_state_limits; u32 gt_perf_status; @@ -858,7 +878,7 @@ static int i915_frequency_info(struct seq_file *m, void *unused) else reqf >>= 25; } - reqf = intel_gpu_freq(dev_priv, reqf); + reqf = intel_gpu_freq(rps, reqf); rpmodectl = I915_READ(GEN6_RP_CONTROL); rpinclimit = I915_READ(GEN6_RP_UP_THRESHOLD); @@ -871,8 +891,7 @@ static int i915_frequency_info(struct seq_file *m, void *unused) rpdownei = I915_READ(GEN6_RP_CUR_DOWN_EI) & GEN6_CURIAVG_MASK; rpcurdown = I915_READ(GEN6_RP_CUR_DOWN) & GEN6_CURBSYTAVG_MASK; rpprevdown = I915_READ(GEN6_RP_PREV_DOWN) & GEN6_CURBSYTAVG_MASK; - cagf = intel_gpu_freq(dev_priv, - intel_get_cagf(dev_priv, rpstat)); + cagf = intel_gpu_freq(rps, intel_get_cagf(rps, rpstat)); intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); @@ -949,37 +968,37 @@ static int i915_frequency_info(struct seq_file *m, void *unused) max_freq *= (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10 ? GEN9_FREQ_SCALER : 1); seq_printf(m, "Lowest (RPN) frequency: %dMHz\n", - intel_gpu_freq(dev_priv, max_freq)); + intel_gpu_freq(rps, max_freq)); max_freq = (rp_state_cap & 0xff00) >> 8; max_freq *= (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10 ? GEN9_FREQ_SCALER : 1); seq_printf(m, "Nominal (RP1) frequency: %dMHz\n", - intel_gpu_freq(dev_priv, max_freq)); + intel_gpu_freq(rps, max_freq)); max_freq = (IS_GEN9_LP(dev_priv) ? rp_state_cap >> 16 : rp_state_cap >> 0) & 0xff; max_freq *= (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10 ? GEN9_FREQ_SCALER : 1); seq_printf(m, "Max non-overclocked (RP0) frequency: %dMHz\n", - intel_gpu_freq(dev_priv, max_freq)); + intel_gpu_freq(rps, max_freq)); seq_printf(m, "Max overclocked frequency: %dMHz\n", - intel_gpu_freq(dev_priv, rps->max_freq)); + intel_gpu_freq(rps, rps->max_freq)); seq_printf(m, "Current freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->cur_freq)); + intel_gpu_freq(rps, rps->cur_freq)); seq_printf(m, "Actual freq: %d MHz\n", cagf); seq_printf(m, "Idle freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->idle_freq)); + intel_gpu_freq(rps, rps->idle_freq)); seq_printf(m, "Min freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->min_freq)); + intel_gpu_freq(rps, rps->min_freq)); seq_printf(m, "Boost freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->boost_freq)); + intel_gpu_freq(rps, rps->boost_freq)); seq_printf(m, "Max freq: %d MHz\n", - intel_gpu_freq(dev_priv, rps->max_freq)); + intel_gpu_freq(rps, rps->max_freq)); seq_printf(m, "efficient (RPe) frequency: %d MHz\n", - intel_gpu_freq(dev_priv, rps->efficient_freq)); + intel_gpu_freq(rps, rps->efficient_freq)); } else { seq_puts(m, "no P-state info available\n"); } @@ -992,91 +1011,6 @@ static int i915_frequency_info(struct seq_file *m, void *unused) return ret; } -static void i915_instdone_info(struct drm_i915_private *dev_priv, - struct seq_file *m, - struct intel_instdone *instdone) -{ - int slice; - int subslice; - - seq_printf(m, "\t\tINSTDONE: 0x%08x\n", - instdone->instdone); - - if (INTEL_GEN(dev_priv) <= 3) - return; - - seq_printf(m, "\t\tSC_INSTDONE: 0x%08x\n", - instdone->slice_common); - - if (INTEL_GEN(dev_priv) <= 6) - return; - - for_each_instdone_slice_subslice(dev_priv, slice, subslice) - seq_printf(m, "\t\tSAMPLER_INSTDONE[%d][%d]: 0x%08x\n", - slice, subslice, instdone->sampler[slice][subslice]); - - for_each_instdone_slice_subslice(dev_priv, slice, subslice) - seq_printf(m, "\t\tROW_INSTDONE[%d][%d]: 0x%08x\n", - slice, subslice, instdone->row[slice][subslice]); -} - -static int i915_hangcheck_info(struct seq_file *m, void *unused) -{ - struct drm_i915_private *i915 = node_to_i915(m->private); - struct intel_gt *gt = &i915->gt; - struct intel_engine_cs *engine; - intel_wakeref_t wakeref; - enum intel_engine_id id; - - seq_printf(m, "Reset flags: %lx\n", gt->reset.flags); - if (test_bit(I915_WEDGED, >->reset.flags)) - seq_puts(m, "\tWedged\n"); - if (test_bit(I915_RESET_BACKOFF, >->reset.flags)) - seq_puts(m, "\tDevice (global) reset in progress\n"); - - if (!i915_modparams.enable_hangcheck) { - seq_puts(m, "Hangcheck disabled\n"); - return 0; - } - - if (timer_pending(>->hangcheck.work.timer)) - seq_printf(m, "Hangcheck active, timer fires in %dms\n", - jiffies_to_msecs(gt->hangcheck.work.timer.expires - - jiffies)); - else if (delayed_work_pending(>->hangcheck.work)) - seq_puts(m, "Hangcheck active, work pending\n"); - else - seq_puts(m, "Hangcheck inactive\n"); - - seq_printf(m, "GT active? %s\n", yesno(gt->awake)); - - with_intel_runtime_pm(&i915->runtime_pm, wakeref) { - for_each_engine(engine, i915, id) { - struct intel_instdone instdone; - - seq_printf(m, "%s: %d ms ago\n", - engine->name, - jiffies_to_msecs(jiffies - - engine->hangcheck.action_timestamp)); - - seq_printf(m, "\tACTHD = 0x%08llx [current 0x%08llx]\n", - (long long)engine->hangcheck.acthd, - intel_engine_get_active_head(engine)); - - intel_engine_get_instdone(engine, &instdone); - - seq_puts(m, "\tinstdone read =\n"); - i915_instdone_info(i915, m, &instdone); - - seq_puts(m, "\tinstdone accu =\n"); - i915_instdone_info(i915, m, - &engine->hangcheck.instdone); - } - } - - return 0; -} - static int ironlake_drpc_info(struct seq_file *m) { struct drm_i915_private *i915 = node_to_i915(m->private); @@ -1157,11 +1091,13 @@ static void print_rc6_res(struct seq_file *m, const char *title, const i915_reg_t reg) { - struct drm_i915_private *dev_priv = node_to_i915(m->private); + struct drm_i915_private *i915 = node_to_i915(m->private); + intel_wakeref_t wakeref; - seq_printf(m, "%s %u (%llu us)\n", - title, I915_READ(reg), - intel_rc6_residency_us(dev_priv, reg)); + with_intel_runtime_pm(&i915->runtime_pm, wakeref) + seq_printf(m, "%s %u (%llu us)\n", title, + intel_uncore_read(&i915->uncore, reg), + intel_rc6_residency_us(&i915->gt.rc6, reg)); } static int vlv_drpc_info(struct seq_file *m) @@ -1439,7 +1375,7 @@ static int i915_sr_status(struct seq_file *m, void *unused) static int i915_ring_freq_table(struct seq_file *m, void *unused) { struct drm_i915_private *dev_priv = node_to_i915(m->private); - struct intel_rps *rps = &dev_priv->gt_pm.rps; + struct intel_rps *rps = &dev_priv->gt.rps; unsigned int max_gpu_freq, min_gpu_freq; intel_wakeref_t wakeref; int gpu_freq, ia_freq; @@ -1464,10 +1400,11 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused) GEN6_PCODE_READ_MIN_FREQ_TABLE, &ia_freq, NULL); seq_printf(m, "%d\t\t%d\t\t\t\t%d\n", - intel_gpu_freq(dev_priv, (gpu_freq * - (IS_GEN9_BC(dev_priv) || - INTEL_GEN(dev_priv) >= 10 ? - GEN9_FREQ_SCALER : 1))), + intel_gpu_freq(rps, + (gpu_freq * + (IS_GEN9_BC(dev_priv) || + INTEL_GEN(dev_priv) >= 10 ? + GEN9_FREQ_SCALER : 1))), ((ia_freq >> 0) & 0xff) * 100, ((ia_freq >> 8) & 0xff) * 100); } @@ -1478,21 +1415,11 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused) static int i915_opregion(struct seq_file *m, void *unused) { - struct drm_i915_private *dev_priv = node_to_i915(m->private); - struct drm_device *dev = &dev_priv->drm; - struct intel_opregion *opregion = &dev_priv->opregion; - int ret; - - ret = mutex_lock_interruptible(&dev->struct_mutex); - if (ret) - goto out; + struct intel_opregion *opregion = &node_to_i915(m->private)->opregion; if (opregion->header) seq_write(m, opregion->header, OPREGION_SIZE); - mutex_unlock(&dev->struct_mutex); - -out: return 0; } @@ -1512,11 +1439,6 @@ static int i915_gem_framebuffer_info(struct seq_file *m, void *data) struct drm_device *dev = &dev_priv->drm; struct intel_framebuffer *fbdev_fb = NULL; struct drm_framebuffer *drm_fb; - int ret; - - ret = mutex_lock_interruptible(&dev->struct_mutex); - if (ret) - return ret; #ifdef CONFIG_DRM_FBDEV_EMULATION if (dev_priv->fbdev && dev_priv->fbdev->helper.fb) { @@ -1551,7 +1473,6 @@ static int i915_gem_framebuffer_info(struct seq_file *m, void *data) seq_putc(m, '\n'); } mutex_unlock(&dev->mode_config.fb_lock); - mutex_unlock(&dev->struct_mutex); return 0; } @@ -1564,23 +1485,20 @@ static void describe_ctx_ring(struct seq_file *m, struct intel_ring *ring) static int i915_context_status(struct seq_file *m, void *unused) { - struct drm_i915_private *dev_priv = node_to_i915(m->private); - struct drm_device *dev = &dev_priv->drm; - struct i915_gem_context *ctx; - int ret; - - ret = mutex_lock_interruptible(&dev->struct_mutex); - if (ret) - return ret; + struct drm_i915_private *i915 = node_to_i915(m->private); + struct i915_gem_context *ctx, *cn; - list_for_each_entry(ctx, &dev_priv->contexts.list, link) { + spin_lock(&i915->gem.contexts.lock); + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) { struct i915_gem_engines_iter it; struct intel_context *ce; + if (!kref_get_unless_zero(&ctx->ref)) + continue; + + spin_unlock(&i915->gem.contexts.lock); + seq_puts(m, "HW context "); - if (!list_empty(&ctx->hw_id_link)) - seq_printf(m, "%x [pin %u]", ctx->hw_id, - atomic_read(&ctx->hw_id_pin_count)); if (ctx->pid) { struct task_struct *task; @@ -1614,9 +1532,12 @@ static int i915_context_status(struct seq_file *m, void *unused) i915_gem_context_unlock_engines(ctx); seq_putc(m, '\n'); - } - mutex_unlock(&dev->struct_mutex); + spin_lock(&i915->gem.contexts.lock); + list_safe_reset_next(ctx, cn, link); + i915_gem_context_put(ctx); + } + spin_unlock(&i915->gem.contexts.lock); return 0; } @@ -1654,9 +1575,9 @@ static int i915_swizzle_info(struct seq_file *m, void *data) wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); seq_printf(m, "bit6 swizzle for X-tiling = %s\n", - swizzle_string(dev_priv->mm.bit_6_swizzle_x)); + swizzle_string(dev_priv->ggtt.bit_6_swizzle_x)); seq_printf(m, "bit6 swizzle for Y-tiling = %s\n", - swizzle_string(dev_priv->mm.bit_6_swizzle_y)); + swizzle_string(dev_priv->ggtt.bit_6_swizzle_y)); if (IS_GEN_RANGE(dev_priv, 3, 4)) { seq_printf(m, "DDC = 0x%08x\n", @@ -1711,7 +1632,7 @@ static const char *rps_power_to_str(unsigned int power) static int i915_rps_boost_info(struct seq_file *m, void *data) { struct drm_i915_private *dev_priv = node_to_i915(m->private); - struct intel_rps *rps = &dev_priv->gt_pm.rps; + struct intel_rps *rps = &dev_priv->gt.rps; u32 act_freq = rps->cur_freq; intel_wakeref_t wakeref; @@ -1723,7 +1644,7 @@ static int i915_rps_boost_info(struct seq_file *m, void *data) vlv_punit_put(dev_priv); act_freq = (act_freq >> 8) & 0xff; } else { - act_freq = intel_get_cagf(dev_priv, + act_freq = intel_get_cagf(rps, I915_READ(GEN6_RPSTAT1)); } } @@ -1734,17 +1655,17 @@ static int i915_rps_boost_info(struct seq_file *m, void *data) atomic_read(&rps->num_waiters)); seq_printf(m, "Interactive? %d\n", READ_ONCE(rps->power.interactive)); seq_printf(m, "Frequency requested %d, actual %d\n", - intel_gpu_freq(dev_priv, rps->cur_freq), - intel_gpu_freq(dev_priv, act_freq)); + intel_gpu_freq(rps, rps->cur_freq), + intel_gpu_freq(rps, act_freq)); seq_printf(m, " min hard:%d, soft:%d; max soft:%d, hard:%d\n", - intel_gpu_freq(dev_priv, rps->min_freq), - intel_gpu_freq(dev_priv, rps->min_freq_softlimit), - intel_gpu_freq(dev_priv, rps->max_freq_softlimit), - intel_gpu_freq(dev_priv, rps->max_freq)); + intel_gpu_freq(rps, rps->min_freq), + intel_gpu_freq(rps, rps->min_freq_softlimit), + intel_gpu_freq(rps, rps->max_freq_softlimit), + intel_gpu_freq(rps, rps->max_freq)); seq_printf(m, " idle:%d, efficient:%d, boost:%d\n", - intel_gpu_freq(dev_priv, rps->idle_freq), - intel_gpu_freq(dev_priv, rps->efficient_freq), - intel_gpu_freq(dev_priv, rps->boost_freq)); + intel_gpu_freq(rps, rps->idle_freq), + intel_gpu_freq(rps, rps->efficient_freq), + intel_gpu_freq(rps, rps->boost_freq)); seq_printf(m, "Wait boosts: %d\n", atomic_read(&rps->boosts)); @@ -1860,8 +1781,8 @@ static void i915_guc_log_info(struct seq_file *m, struct intel_guc_log *log = &dev_priv->gt.uc.guc.log; enum guc_log_buffer_type type; - if (!intel_guc_log_relay_enabled(log)) { - seq_puts(m, "GuC log relay disabled\n"); + if (!intel_guc_log_relay_created(log)) { + seq_puts(m, "GuC log relay not created\n"); return; } @@ -2048,9 +1969,23 @@ i915_guc_log_relay_write(struct file *filp, loff_t *ppos) { struct intel_guc_log *log = filp->private_data; + int val; + int ret; - intel_guc_log_relay_flush(log); - return cnt; + ret = kstrtoint_from_user(ubuf, cnt, 0, &val); + if (ret < 0) + return ret; + + /* + * Enable and start the guc log relay on value of 1. + * Flush log relay for any other value. + */ + if (val == 1) + ret = intel_guc_log_relay_start(log); + else + intel_guc_log_relay_flush(log); + + return ret ?: cnt; } static int i915_guc_log_relay_release(struct inode *inode, struct file *file) @@ -2133,7 +2068,7 @@ psr_source_status(struct drm_i915_private *dev_priv, struct seq_file *m) "BUF_ON", "TG_ON" }; - val = I915_READ(EDP_PSR2_STATUS); + val = I915_READ(EDP_PSR2_STATUS(dev_priv->psr.transcoder)); status_val = (val & EDP_PSR2_STATUS_STATE_MASK) >> EDP_PSR2_STATUS_STATE_SHIFT; if (status_val < ARRAY_SIZE(live_status)) @@ -2149,7 +2084,7 @@ psr_source_status(struct drm_i915_private *dev_priv, struct seq_file *m) "SRDOFFACK", "SRDENT_ON", }; - val = I915_READ(EDP_PSR_STATUS); + val = I915_READ(EDP_PSR_STATUS(dev_priv->psr.transcoder)); status_val = (val & EDP_PSR_STATUS_STATE_MASK) >> EDP_PSR_STATUS_STATE_SHIFT; if (status_val < ARRAY_SIZE(live_status)) @@ -2188,14 +2123,18 @@ static int i915_edp_psr_status(struct seq_file *m, void *data) status = "disabled"; seq_printf(m, "PSR mode: %s\n", status); - if (!psr->enabled) + if (!psr->enabled) { + seq_printf(m, "PSR sink not reliable: %s\n", + yesno(psr->sink_not_reliable)); + goto unlock; + } if (psr->psr2_enabled) { - val = I915_READ(EDP_PSR2_CTL); + val = I915_READ(EDP_PSR2_CTL(dev_priv->psr.transcoder)); enabled = val & EDP_PSR2_ENABLE; } else { - val = I915_READ(EDP_PSR_CTL); + val = I915_READ(EDP_PSR_CTL(dev_priv->psr.transcoder)); enabled = val & EDP_PSR_ENABLE; } seq_printf(m, "Source PSR ctl: %s [0x%08x]\n", @@ -2208,7 +2147,8 @@ static int i915_edp_psr_status(struct seq_file *m, void *data) * SKL+ Perf counter is reset to 0 everytime DC state is entered */ if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) { - val = I915_READ(EDP_PSR_PERF_CNT) & EDP_PSR_PERF_CNT_MASK; + val = I915_READ(EDP_PSR_PERF_CNT(dev_priv->psr.transcoder)); + val &= EDP_PSR_PERF_CNT_MASK; seq_printf(m, "Performance counter: %u\n", val); } @@ -2226,8 +2166,11 @@ static int i915_edp_psr_status(struct seq_file *m, void *data) * Reading all 3 registers before hand to minimize crossing a * frame boundary between register reads */ - for (frame = 0; frame < PSR2_SU_STATUS_FRAMES; frame += 3) - su_frames_val[frame / 3] = I915_READ(PSR2_SU_STATUS(frame)); + for (frame = 0; frame < PSR2_SU_STATUS_FRAMES; frame += 3) { + val = I915_READ(PSR2_SU_STATUS(dev_priv->psr.transcoder, + frame)); + su_frames_val[frame / 3] = val; + } seq_puts(m, "Frame:\tPSR2 SU blocks:\n"); @@ -2360,8 +2303,7 @@ static int i915_power_domain_info(struct seq_file *m, void *unused) for_each_power_domain(power_domain, power_well->desc->domains) seq_printf(m, " %-23s %d\n", - intel_display_power_domain_str(dev_priv, - power_domain), + intel_display_power_domain_str(power_domain), power_domains->domain_use_count[power_domain]); } @@ -2396,6 +2338,13 @@ static int i915_dmc_info(struct seq_file *m, void *unused) if (INTEL_GEN(dev_priv) >= 12) { dc5_reg = TGL_DMC_DEBUG_DC5_COUNT; dc6_reg = TGL_DMC_DEBUG_DC6_COUNT; + /* + * NOTE: DMC_DEBUG3 is a general purpose reg. + * According to B.Specs:49196 DMC f/w reuses DC5/6 counter + * reg for DC3CO debugging and validation, + * but TGL DMC f/w is using DMC_DEBUG3 reg for DC3CO counter. + */ + seq_printf(m, "DC3CO count: %d\n", I915_READ(DMC_DEBUG3)); } else { dc5_reg = IS_BROXTON(dev_priv) ? BXT_CSR_DC3_DC5_COUNT : SKL_CSR_DC3_DC5_COUNT; @@ -3110,8 +3059,9 @@ static int i915_dp_mst_info(struct seq_file *m, void *unused) if (!intel_dig_port->dp.can_mst) continue; - seq_printf(m, "MST Source Port %c\n", - port_name(intel_dig_port->base.port)); + seq_printf(m, "MST Source Port [ENCODER:%d:%s]\n", + intel_dig_port->base.base.base.id, + intel_dig_port->base.base.name); drm_dp_mst_dump_topology(m, &intel_dig_port->dp.mst_mgr); } drm_connector_list_iter_end(&conn_iter); @@ -3573,6 +3523,37 @@ DEFINE_SIMPLE_ATTRIBUTE(i915_wedged_fops, i915_wedged_get, i915_wedged_set, "%llu\n"); +static int +i915_perf_noa_delay_set(void *data, u64 val) +{ + struct drm_i915_private *i915 = data; + const u32 clk = RUNTIME_INFO(i915)->cs_timestamp_frequency_khz; + + /* + * This would lead to infinite waits as we're doing timestamp + * difference on the CS with only 32bits. + */ + if (val > mul_u32_u32(U32_MAX, clk)) + return -EINVAL; + + atomic64_set(&i915->perf.noa_programming_delay, val); + return 0; +} + +static int +i915_perf_noa_delay_get(void *data, u64 *val) +{ + struct drm_i915_private *i915 = data; + + *val = atomic64_read(&i915->perf.noa_programming_delay); + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(i915_perf_noa_delay_fops, + i915_perf_noa_delay_get, + i915_perf_noa_delay_set, + "%llu\n"); + #define DROP_UNBOUND BIT(0) #define DROP_BOUND BIT(1) #define DROP_RETIRE BIT(2) @@ -3582,6 +3563,7 @@ DEFINE_SIMPLE_ATTRIBUTE(i915_wedged_fops, #define DROP_IDLE BIT(6) #define DROP_RESET_ACTIVE BIT(7) #define DROP_RESET_SEQNO BIT(8) +#define DROP_RCU BIT(9) #define DROP_ALL (DROP_UNBOUND | \ DROP_BOUND | \ DROP_RETIRE | \ @@ -3590,7 +3572,8 @@ DEFINE_SIMPLE_ATTRIBUTE(i915_wedged_fops, DROP_SHRINK_ALL |\ DROP_IDLE | \ DROP_RESET_ACTIVE | \ - DROP_RESET_SEQNO) + DROP_RESET_SEQNO | \ + DROP_RCU) static int i915_drop_caches_get(void *data, u64 *val) { @@ -3598,58 +3581,48 @@ i915_drop_caches_get(void *data, u64 *val) return 0; } - static int -i915_drop_caches_set(void *data, u64 val) +gt_drop_caches(struct intel_gt *gt, u64 val) { - struct drm_i915_private *i915 = data; - - DRM_DEBUG("Dropping caches: 0x%08llx [0x%08llx]\n", - val, val & DROP_ALL); + int ret; if (val & DROP_RESET_ACTIVE && - wait_for(intel_engines_are_idle(&i915->gt), - I915_IDLE_ENGINES_TIMEOUT)) - intel_gt_set_wedged(&i915->gt); + wait_for(intel_engines_are_idle(gt), I915_IDLE_ENGINES_TIMEOUT)) + intel_gt_set_wedged(gt); - /* No need to check and wait for gpu resets, only libdrm auto-restarts - * on ioctls on -EAGAIN. */ - if (val & (DROP_ACTIVE | DROP_IDLE | DROP_RETIRE | DROP_RESET_SEQNO)) { - int ret; + if (val & DROP_RETIRE) + intel_gt_retire_requests(gt); - ret = mutex_lock_interruptible(&i915->drm.struct_mutex); + if (val & (DROP_IDLE | DROP_ACTIVE)) { + ret = intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT); if (ret) return ret; + } - /* - * To finish the flush of the idle_worker, we must complete - * the switch-to-kernel-context, which requires a double - * pass through wait_for_idle: first queues the switch, - * second waits for the switch. - */ - if (ret == 0 && val & (DROP_IDLE | DROP_ACTIVE)) - ret = i915_gem_wait_for_idle(i915, - I915_WAIT_INTERRUPTIBLE | - I915_WAIT_LOCKED, - MAX_SCHEDULE_TIMEOUT); + if (val & DROP_IDLE) { + ret = intel_gt_pm_wait_for_idle(gt); + if (ret) + return ret; + } - if (ret == 0 && val & DROP_IDLE) - ret = i915_gem_wait_for_idle(i915, - I915_WAIT_INTERRUPTIBLE | - I915_WAIT_LOCKED, - MAX_SCHEDULE_TIMEOUT); + if (val & DROP_RESET_ACTIVE && intel_gt_terminally_wedged(gt)) + intel_gt_handle_error(gt, ALL_ENGINES, 0, NULL); - if (val & DROP_RETIRE) - i915_retire_requests(i915); + return 0; +} - mutex_unlock(&i915->drm.struct_mutex); +static int +i915_drop_caches_set(void *data, u64 val) +{ + struct drm_i915_private *i915 = data; + int ret; - if (ret == 0 && val & DROP_IDLE) - ret = intel_gt_pm_wait_for_idle(&i915->gt); - } + DRM_DEBUG("Dropping caches: 0x%08llx [0x%08llx]\n", + val, val & DROP_ALL); - if (val & DROP_RESET_ACTIVE && intel_gt_terminally_wedged(&i915->gt)) - intel_gt_handle_error(&i915->gt, ALL_ENGINES, 0, NULL); + ret = gt_drop_caches(&i915->gt, val); + if (ret) + return ret; fs_reclaim_acquire(GFP_KERNEL); if (val & DROP_BOUND) @@ -3662,10 +3635,8 @@ i915_drop_caches_set(void *data, u64 val) i915_gem_shrink_all(i915); fs_reclaim_release(GFP_KERNEL); - if (val & DROP_IDLE) { - flush_delayed_work(&i915->gem.retire_work); - flush_work(&i915->gem.idle_work); - } + if (val & DROP_RCU) + rcu_barrier(); if (val & DROP_FREED) i915_gem_drain_freed_objects(i915); @@ -3721,6 +3692,15 @@ i915_cache_sharing_set(void *data, u64 val) return 0; } +static void +intel_sseu_copy_subslices(const struct sseu_dev_info *sseu, int slice, + u8 *to_mask) +{ + int offset = slice * sseu->ss_stride; + + memcpy(&to_mask[offset], &sseu->subslice_mask[offset], sseu->ss_stride); +} + DEFINE_SIMPLE_ATTRIBUTE(i915_cache_sharing_fops, i915_cache_sharing_get, i915_cache_sharing_set, "%llu\n"); @@ -3794,12 +3774,13 @@ static void gen10_sseu_device_status(struct drm_i915_private *dev_priv, continue; sseu->slice_mask |= BIT(s); - sseu->subslice_mask[s] = info->sseu.subslice_mask[s]; + intel_sseu_copy_subslices(&info->sseu, s, sseu->subslice_mask); for (ss = 0; ss < info->sseu.max_subslices; ss++) { unsigned int eu_cnt; - if (!(s_reg[s] & (GEN9_PGCTL_SS_ACK(ss)))) + if (info->sseu.has_subslice_pg && + !(s_reg[s] & (GEN9_PGCTL_SS_ACK(ss)))) /* skip disabled subslice */ continue; @@ -3845,18 +3826,21 @@ static void gen9_sseu_device_status(struct drm_i915_private *dev_priv, sseu->slice_mask |= BIT(s); if (IS_GEN9_BC(dev_priv)) - sseu->subslice_mask[s] = - RUNTIME_INFO(dev_priv)->sseu.subslice_mask[s]; + intel_sseu_copy_subslices(&info->sseu, s, + sseu->subslice_mask); for (ss = 0; ss < info->sseu.max_subslices; ss++) { unsigned int eu_cnt; + u8 ss_idx = s * info->sseu.ss_stride + + ss / BITS_PER_BYTE; if (IS_GEN9_LP(dev_priv)) { if (!(s_reg[s] & (GEN9_PGCTL_SS_ACK(ss)))) /* skip disabled subslice */ continue; - sseu->subslice_mask[s] |= BIT(ss); + sseu->subslice_mask[ss_idx] |= + BIT(ss % BITS_PER_BYTE); } eu_cnt = 2 * hweight32(eu_reg[2*s + ss/2] & @@ -3873,25 +3857,23 @@ static void gen9_sseu_device_status(struct drm_i915_private *dev_priv, static void broadwell_sseu_device_status(struct drm_i915_private *dev_priv, struct sseu_dev_info *sseu) { + const struct intel_runtime_info *info = RUNTIME_INFO(dev_priv); u32 slice_info = I915_READ(GEN8_GT_SLICE_INFO); int s; sseu->slice_mask = slice_info & GEN8_LSLICESTAT_MASK; if (sseu->slice_mask) { - sseu->eu_per_subslice = - RUNTIME_INFO(dev_priv)->sseu.eu_per_subslice; - for (s = 0; s < fls(sseu->slice_mask); s++) { - sseu->subslice_mask[s] = - RUNTIME_INFO(dev_priv)->sseu.subslice_mask[s]; - } + sseu->eu_per_subslice = info->sseu.eu_per_subslice; + for (s = 0; s < fls(sseu->slice_mask); s++) + intel_sseu_copy_subslices(&info->sseu, s, + sseu->subslice_mask); sseu->eu_total = sseu->eu_per_subslice * intel_sseu_subslice_total(sseu); /* subtract fused off EU(s) from enabled slice(s) */ for (s = 0; s < fls(sseu->slice_mask); s++) { - u8 subslice_7eu = - RUNTIME_INFO(dev_priv)->sseu.subslice_7eu[s]; + u8 subslice_7eu = info->sseu.subslice_7eu[s]; sseu->eu_total -= hweight8(subslice_7eu); } @@ -3938,6 +3920,7 @@ static void i915_print_sseu_info(struct seq_file *m, bool is_available_info, static int i915_sseu_status(struct seq_file *m, void *unused) { struct drm_i915_private *dev_priv = node_to_i915(m->private); + const struct intel_runtime_info *info = RUNTIME_INFO(dev_priv); struct sseu_dev_info sseu; intel_wakeref_t wakeref; @@ -3945,14 +3928,13 @@ static int i915_sseu_status(struct seq_file *m, void *unused) return -ENODEV; seq_puts(m, "SSEU Device Info\n"); - i915_print_sseu_info(m, true, &RUNTIME_INFO(dev_priv)->sseu); + i915_print_sseu_info(m, true, &info->sseu); seq_puts(m, "SSEU Device Status\n"); memset(&sseu, 0, sizeof(sseu)); - sseu.max_slices = RUNTIME_INFO(dev_priv)->sseu.max_slices; - sseu.max_subslices = RUNTIME_INFO(dev_priv)->sseu.max_subslices; - sseu.max_eus_per_subslice = - RUNTIME_INFO(dev_priv)->sseu.max_eus_per_subslice; + intel_sseu_set_info(&sseu, info->sseu.max_slices, + info->sseu.max_subslices, + info->sseu.max_eus_per_subslice); with_intel_runtime_pm(&dev_priv->runtime_pm, wakeref) { if (IS_CHERRYVIEW(dev_priv)) @@ -3973,13 +3955,12 @@ static int i915_sseu_status(struct seq_file *m, void *unused) static int i915_forcewake_open(struct inode *inode, struct file *file) { struct drm_i915_private *i915 = inode->i_private; + struct intel_gt *gt = &i915->gt; - if (INTEL_GEN(i915) < 6) - return 0; - - file->private_data = - (void *)(uintptr_t)intel_runtime_pm_get(&i915->runtime_pm); - intel_uncore_forcewake_user_get(&i915->uncore); + atomic_inc(>->user_wakeref); + intel_gt_pm_get(gt); + if (INTEL_GEN(i915) >= 6) + intel_uncore_forcewake_user_get(gt->uncore); return 0; } @@ -3987,13 +3968,12 @@ static int i915_forcewake_open(struct inode *inode, struct file *file) static int i915_forcewake_release(struct inode *inode, struct file *file) { struct drm_i915_private *i915 = inode->i_private; + struct intel_gt *gt = &i915->gt; - if (INTEL_GEN(i915) < 6) - return 0; - - intel_uncore_forcewake_user_put(&i915->uncore); - intel_runtime_pm_put(&i915->runtime_pm, - (intel_wakeref_t)(uintptr_t)file->private_data); + if (INTEL_GEN(i915) >= 6) + intel_uncore_forcewake_user_put(&i915->uncore); + intel_gt_pm_put(gt); + atomic_dec(>->user_wakeref); return 0; } @@ -4302,7 +4282,6 @@ static const struct drm_info_list i915_debugfs_list[] = { {"i915_guc_stage_pool", i915_guc_stage_pool, 0}, {"i915_huc_load_status", i915_huc_load_status_info, 0}, {"i915_frequency_info", i915_frequency_info, 0}, - {"i915_hangcheck_info", i915_hangcheck_info, 0}, {"i915_drpc_info", i915_drpc_info, 0}, {"i915_ring_freq_table", i915_ring_freq_table, 0}, {"i915_frontbuffer_tracking", i915_frontbuffer_tracking, 0}, @@ -4339,6 +4318,7 @@ static const struct i915_debugfs_files { const char *name; const struct file_operations *fops; } i915_debugfs_files[] = { + {"i915_perf_noa_delay", &i915_perf_noa_delay_fops}, {"i915_wedged", &i915_wedged_fops}, {"i915_cache_sharing", &i915_cache_sharing_fops}, {"i915_gem_drop_caches", &i915_drop_caches_fops}, @@ -4528,7 +4508,7 @@ static int i915_dsc_fec_support_show(struct seq_file *m, void *data) intel_dp = enc_to_intel_dp(&intel_attached_encoder(connector)->base); crtc_state = to_intel_crtc_state(crtc->state); seq_printf(m, "DSC_Enabled: %s\n", - yesno(crtc_state->dsc_params.compression_enable)); + yesno(crtc_state->dsc.compression_enable)); seq_printf(m, "DSC_Sink_Support: %s\n", yesno(drm_dp_sink_supports_dsc(intel_dp->dsc_dpcd))); seq_printf(m, "Force_DSC_Enable: %s\n", diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 020696726f9e..3c512c571e60 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -36,7 +36,6 @@ #include <linux/pm_runtime.h> #include <linux/pnp.h> #include <linux/slab.h> -#include <linux/vgaarb.h> #include <linux/vga_switcheroo.h> #include <linux/vt.h> #include <acpi/video.h> @@ -54,16 +53,17 @@ #include "display/intel_display_types.h" #include "display/intel_dp.h" #include "display/intel_fbdev.h" -#include "display/intel_gmbus.h" #include "display/intel_hotplug.h" #include "display/intel_overlay.h" #include "display/intel_pipe_crc.h" #include "display/intel_sprite.h" +#include "display/intel_vga.h" #include "gem/i915_gem_context.h" #include "gem/i915_gem_ioctls.h" #include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" +#include "gt/intel_rc6.h" #include "i915_debugfs.h" #include "i915_drv.h" @@ -72,10 +72,12 @@ #include "i915_perf.h" #include "i915_query.h" #include "i915_suspend.h" +#include "i915_switcheroo.h" #include "i915_sysfs.h" #include "i915_trace.h" #include "i915_vgpu.h" #include "intel_csr.h" +#include "intel_memory_region.h" #include "intel_pm.h" static struct drm_driver driver; @@ -269,182 +271,97 @@ intel_teardown_mchbar(struct drm_i915_private *dev_priv) release_resource(&dev_priv->mch_res); } -/* true = enable decode, false = disable decoder */ -static unsigned int i915_vga_set_decode(void *cookie, bool state) +static int i915_driver_modeset_probe(struct drm_i915_private *i915) { - struct drm_i915_private *dev_priv = cookie; - - intel_modeset_vga_set_state(dev_priv, state); - if (state) - return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM | - VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM; - else - return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM; -} - -static int i915_resume_switcheroo(struct drm_i915_private *i915); -static int i915_suspend_switcheroo(struct drm_i915_private *i915, - pm_message_t state); - -static void i915_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_state state) -{ - struct drm_i915_private *i915 = pdev_to_i915(pdev); - pm_message_t pmm = { .event = PM_EVENT_SUSPEND }; - - if (!i915) { - dev_err(&pdev->dev, "DRM not initialized, aborting switch.\n"); - return; - } - - if (state == VGA_SWITCHEROO_ON) { - pr_info("switched on\n"); - i915->drm.switch_power_state = DRM_SWITCH_POWER_CHANGING; - /* i915 resume handler doesn't set to D0 */ - pci_set_power_state(pdev, PCI_D0); - i915_resume_switcheroo(i915); - i915->drm.switch_power_state = DRM_SWITCH_POWER_ON; - } else { - pr_info("switched off\n"); - i915->drm.switch_power_state = DRM_SWITCH_POWER_CHANGING; - i915_suspend_switcheroo(i915, pmm); - i915->drm.switch_power_state = DRM_SWITCH_POWER_OFF; - } -} - -static bool i915_switcheroo_can_switch(struct pci_dev *pdev) -{ - struct drm_i915_private *i915 = pdev_to_i915(pdev); - - /* - * FIXME: open_count is protected by drm_global_mutex but that would lead to - * locking inversion with the driver load path. And the access here is - * completely racy anyway. So don't bother with locking for now. - */ - return i915 && i915->drm.open_count == 0; -} - -static const struct vga_switcheroo_client_ops i915_switcheroo_ops = { - .set_gpu_state = i915_switcheroo_set_state, - .reprobe = NULL, - .can_switch = i915_switcheroo_can_switch, -}; - -static int i915_driver_modeset_probe(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = to_i915(dev); - struct pci_dev *pdev = dev_priv->drm.pdev; int ret; - if (i915_inject_probe_failure(dev_priv)) + if (i915_inject_probe_failure(i915)) return -ENODEV; - if (HAS_DISPLAY(dev_priv)) { - ret = drm_vblank_init(&dev_priv->drm, - INTEL_INFO(dev_priv)->num_pipes); + if (HAS_DISPLAY(i915) && INTEL_DISPLAY_ENABLED(i915)) { + ret = drm_vblank_init(&i915->drm, + INTEL_NUM_PIPES(i915)); if (ret) goto out; } - intel_bios_init(dev_priv); + intel_bios_init(i915); - /* If we have > 1 VGA cards, then we need to arbitrate access - * to the common VGA resources. - * - * If we are a secondary display controller (!PCI_DISPLAY_CLASS_VGA), - * then we do not take part in VGA arbitration and the - * vga_client_register() fails with -ENODEV. - */ - ret = vga_client_register(pdev, dev_priv, NULL, i915_vga_set_decode); - if (ret && ret != -ENODEV) + ret = intel_vga_register(i915); + if (ret) goto out; intel_register_dsm_handler(); - ret = vga_switcheroo_register_client(pdev, &i915_switcheroo_ops, false); + ret = i915_switcheroo_register(i915); if (ret) goto cleanup_vga_client; - /* must happen before intel_power_domains_init_hw() on VLV/CHV */ - intel_update_rawclk(dev_priv); - - intel_power_domains_init_hw(dev_priv, false); + intel_power_domains_init_hw(i915, false); - intel_csr_ucode_init(dev_priv); + intel_csr_ucode_init(i915); - ret = intel_irq_install(dev_priv); + ret = intel_irq_install(i915); if (ret) goto cleanup_csr; - intel_gmbus_setup(dev_priv); - /* Important: The output setup functions called by modeset_init need * working irqs for e.g. gmbus and dp aux transfers. */ - ret = intel_modeset_init(dev); + ret = intel_modeset_init(i915); if (ret) goto cleanup_irq; - ret = i915_gem_init(dev_priv); + ret = i915_gem_init(i915); if (ret) goto cleanup_modeset; - intel_overlay_setup(dev_priv); + intel_overlay_setup(i915); - if (!HAS_DISPLAY(dev_priv)) + if (!HAS_DISPLAY(i915) || !INTEL_DISPLAY_ENABLED(i915)) return 0; - ret = intel_fbdev_init(dev); + ret = intel_fbdev_init(&i915->drm); if (ret) goto cleanup_gem; /* Only enable hotplug handling once the fbdev is fully set up. */ - intel_hpd_init(dev_priv); + intel_hpd_init(i915); - intel_init_ipc(dev_priv); + intel_init_ipc(i915); return 0; cleanup_gem: - i915_gem_suspend(dev_priv); - i915_gem_driver_remove(dev_priv); - i915_gem_driver_release(dev_priv); + i915_gem_suspend(i915); + i915_gem_driver_remove(i915); + i915_gem_driver_release(i915); cleanup_modeset: - intel_modeset_driver_remove(dev); + intel_modeset_driver_remove(i915); cleanup_irq: - intel_irq_uninstall(dev_priv); - intel_gmbus_teardown(dev_priv); + intel_irq_uninstall(i915); cleanup_csr: - intel_csr_ucode_fini(dev_priv); - intel_power_domains_driver_remove(dev_priv); - vga_switcheroo_unregister_client(pdev); + intel_csr_ucode_fini(i915); + intel_power_domains_driver_remove(i915); + i915_switcheroo_unregister(i915); cleanup_vga_client: - vga_client_register(pdev, NULL, NULL, NULL); + intel_vga_unregister(i915); out: return ret; } -static int i915_kick_out_firmware_fb(struct drm_i915_private *dev_priv) +static void i915_driver_modeset_remove(struct drm_i915_private *i915) { - struct apertures_struct *ap; - struct pci_dev *pdev = dev_priv->drm.pdev; - struct i915_ggtt *ggtt = &dev_priv->ggtt; - bool primary; - int ret; + intel_modeset_driver_remove(i915); - ap = alloc_apertures(1); - if (!ap) - return -ENOMEM; + intel_irq_uninstall(i915); - ap->ranges[0].base = ggtt->gmadr.start; - ap->ranges[0].size = ggtt->mappable_end; - - primary = - pdev->resource[PCI_ROM_RESOURCE].flags & IORESOURCE_ROM_SHADOW; + intel_bios_driver_remove(i915); - ret = drm_fb_helper_remove_conflicting_framebuffers(ap, "inteldrmfb", primary); + i915_switcheroo_unregister(i915); - kfree(ap); + intel_vga_unregister(i915); - return ret; + intel_csr_ucode_fini(i915); } static void intel_init_dpio(struct drm_i915_private *dev_priv) @@ -601,9 +518,7 @@ static int i915_driver_early_probe(struct drm_i915_private *dev_priv) intel_gt_init_early(&dev_priv->gt, dev_priv); - ret = i915_gem_init_early(dev_priv); - if (ret < 0) - goto err_gt; + i915_gem_init_early(dev_priv); /* This must be called before any calls to HAS_PCH_* */ intel_detect_pch(dev_priv); @@ -625,7 +540,6 @@ static int i915_driver_early_probe(struct drm_i915_private *dev_priv) err_gem: i915_gem_cleanup_early(dev_priv); -err_gt: intel_gt_driver_late_release(&dev_priv->gt); vlv_free_s0ix_state(dev_priv); err_workqueues: @@ -683,12 +597,10 @@ static int i915_driver_mmio_probe(struct drm_i915_private *dev_priv) intel_uc_init_mmio(&dev_priv->gt.uc); - ret = intel_engines_init_mmio(dev_priv); + ret = intel_engines_init_mmio(&dev_priv->gt); if (ret) goto err_uncore; - i915_gem_init_mmio(dev_priv); - return 0; err_uncore: @@ -706,7 +618,7 @@ err_bridge: */ static void i915_driver_mmio_release(struct drm_i915_private *dev_priv) { - intel_engines_cleanup(dev_priv); + intel_engines_cleanup(&dev_priv->gt); intel_teardown_mchbar(dev_priv); intel_uncore_fini_mmio(&dev_priv->uncore); pci_dev_put(dev_priv->bridge_dev); @@ -1160,8 +1072,8 @@ intel_get_dram_info(struct drm_i915_private *dev_priv) static u32 gen9_edram_size_mb(struct drm_i915_private *dev_priv, u32 cap) { - const unsigned int ways[8] = { 4, 8, 12, 16, 16, 16, 16, 16 }; - const unsigned int sets[4] = { 1, 1, 2, 2 }; + static const u8 ways[8] = { 4, 8, 12, 16, 16, 16, 16, 16 }; + static const u8 sets[4] = { 1, 1, 2, 2 }; return EDRAM_NUM_BANKS(cap) * ways[EDRAM_WAYS_IDX(cap)] * @@ -1249,32 +1161,24 @@ static int i915_driver_hw_probe(struct drm_i915_private *dev_priv) if (ret) goto err_perf; - /* - * WARNING: Apparently we must kick fbdev drivers before vgacon, - * otherwise the vga fbdev driver falls over. - */ - ret = i915_kick_out_firmware_fb(dev_priv); - if (ret) { - DRM_ERROR("failed to remove conflicting framebuffer drivers\n"); + ret = drm_fb_helper_remove_conflicting_pci_framebuffers(pdev, "inteldrmfb"); + if (ret) goto err_ggtt; - } - ret = vga_remove_vgacon(pdev); - if (ret) { - DRM_ERROR("failed to remove conflicting VGA console\n"); + ret = i915_ggtt_init_hw(dev_priv); + if (ret) goto err_ggtt; - } - ret = i915_ggtt_init_hw(dev_priv); + ret = intel_memory_regions_hw_probe(dev_priv); if (ret) goto err_ggtt; - intel_gt_init_hw(dev_priv); + intel_gt_init_hw_early(&dev_priv->gt, &dev_priv->ggtt); ret = i915_ggtt_enable_hw(dev_priv); if (ret) { DRM_ERROR("failed to enable GGTT\n"); - goto err_ggtt; + goto err_mem_regions; } pci_set_master(pdev); @@ -1291,7 +1195,7 @@ static int i915_driver_hw_probe(struct drm_i915_private *dev_priv) if (ret) { DRM_ERROR("failed to set DMA mask\n"); - goto err_ggtt; + goto err_mem_regions; } } @@ -1309,16 +1213,13 @@ static int i915_driver_hw_probe(struct drm_i915_private *dev_priv) if (ret) { DRM_ERROR("failed to set DMA mask\n"); - goto err_ggtt; + goto err_mem_regions; } } pm_qos_add_request(&dev_priv->pm_qos, PM_QOS_CPU_DMA_LATENCY, PM_QOS_DEFAULT_VALUE); - /* BIOS often leaves RC6 enabled, but disable it for hw init */ - intel_sanitize_gt_powersave(dev_priv); - intel_gt_init_workarounds(dev_priv); /* On the 945G/GM, the chipset reports the MSI capability on the @@ -1364,6 +1265,8 @@ err_msi: if (pdev->msi_enabled) pci_disable_msi(pdev); pm_qos_remove_request(&dev_priv->pm_qos); +err_mem_regions: + intel_memory_regions_driver_release(dev_priv); err_ggtt: i915_ggtt_driver_release(dev_priv); err_perf: @@ -1418,14 +1321,13 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) } else DRM_ERROR("Failed to register driver for userspace access!\n"); - if (HAS_DISPLAY(dev_priv)) { + if (HAS_DISPLAY(dev_priv) && INTEL_DISPLAY_ENABLED(dev_priv)) { /* Must be done after probing outputs */ intel_opregion_register(dev_priv); acpi_video_register(); } - if (IS_GEN(dev_priv, 5)) - intel_gpu_ips_init(dev_priv); + intel_gt_driver_register(&dev_priv->gt); intel_audio_init(dev_priv); @@ -1442,7 +1344,7 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) * We need to coordinate the hotplugs with the asynchronous fbdev * configuration, for which we use the fbdev->async_cookie. */ - if (HAS_DISPLAY(dev_priv)) + if (HAS_DISPLAY(dev_priv) && INTEL_DISPLAY_ENABLED(dev_priv)) drm_kms_helper_poll_init(dev); intel_power_domains_enable(dev_priv); @@ -1468,7 +1370,7 @@ static void i915_driver_unregister(struct drm_i915_private *dev_priv) */ drm_kms_helper_poll_fini(&dev_priv->drm); - intel_gpu_ips_teardown(); + intel_gt_driver_unregister(&dev_priv->gt); acpi_video_unregister(); intel_opregion_unregister(dev_priv); @@ -1577,6 +1479,23 @@ int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent) if (!i915_modparams.nuclear_pageflip && match_info->gen < 5) dev_priv->drm.driver_features &= ~DRIVER_ATOMIC; + /* + * Check if we support fake LMEM -- for now we only unleash this for + * the live selftests(test-and-exit). + */ +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) + if (IS_ENABLED(CONFIG_DRM_I915_UNSTABLE_FAKE_LMEM)) { + if (INTEL_GEN(dev_priv) >= 9 && i915_selftest.live < 0 && + i915_modparams.fake_lmem_start) { + mkwrite_device_info(dev_priv)->memory_regions = + REGION_SMEM | REGION_LMEM | REGION_STOLEN; + mkwrite_device_info(dev_priv)->is_dgfx = true; + GEM_BUG_ON(!HAS_LMEM(dev_priv)); + GEM_BUG_ON(!IS_DGFX(dev_priv)); + } + } +#endif + ret = pci_enable_device(pdev); if (ret) goto out_fini; @@ -1597,7 +1516,7 @@ int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent) if (ret < 0) goto out_cleanup_mmio; - ret = i915_driver_modeset_probe(&dev_priv->drm); + ret = i915_driver_modeset_probe(dev_priv); if (ret < 0) goto out_cleanup_hw; @@ -1611,10 +1530,8 @@ int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent) out_cleanup_hw: i915_driver_hw_remove(dev_priv); + intel_memory_regions_driver_release(dev_priv); i915_ggtt_driver_release(dev_priv); - - /* Paranoia: make sure we have disabled everything before we exit. */ - intel_sanitize_gt_powersave(dev_priv); out_cleanup_mmio: i915_driver_mmio_release(dev_priv); out_runtime_pm_put: @@ -1630,8 +1547,6 @@ out_fini: void i915_driver_remove(struct drm_i915_private *i915) { - struct pci_dev *pdev = i915->drm.pdev; - disable_rpm_wakeref_asserts(&i915->runtime_pm); i915_driver_unregister(i915); @@ -1652,19 +1567,9 @@ void i915_driver_remove(struct drm_i915_private *i915) intel_gvt_driver_remove(i915); - intel_modeset_driver_remove(&i915->drm); + i915_driver_modeset_remove(i915); - intel_bios_driver_remove(i915); - - vga_switcheroo_unregister_client(pdev); - vga_client_register(pdev, NULL, NULL, NULL); - - intel_csr_ucode_fini(i915); - - /* Free error state after interrupts are fully disabled. */ - cancel_delayed_work_sync(&i915->gt.hangcheck.work); i915_reset_error_state(i915); - i915_gem_driver_remove(i915); intel_power_domains_driver_remove(i915); @@ -1683,11 +1588,9 @@ static void i915_driver_release(struct drm_device *dev) i915_gem_driver_release(dev_priv); + intel_memory_regions_driver_release(dev_priv); i915_ggtt_driver_release(dev_priv); - /* Paranoia: make sure we have disabled everything before we exit. */ - intel_sanitize_gt_powersave(dev_priv); - i915_driver_mmio_release(dev_priv); enable_rpm_wakeref_asserts(rpm); @@ -1731,12 +1634,10 @@ static void i915_driver_postclose(struct drm_device *dev, struct drm_file *file) { struct drm_i915_file_private *file_priv = file->driver_priv; - mutex_lock(&dev->struct_mutex); i915_gem_context_close(file); i915_gem_release(dev, file); - mutex_unlock(&dev->struct_mutex); - kfree(file_priv); + kfree_rcu(file_priv, rcu); /* Catch up with all the deferred frees from "this" client */ i915_gem_flush_free_objects(to_i915(dev)); @@ -1891,8 +1792,7 @@ out: return ret; } -static int -i915_suspend_switcheroo(struct drm_i915_private *i915, pm_message_t state) +int i915_suspend_switcheroo(struct drm_i915_private *i915, pm_message_t state) { int error; @@ -1916,14 +1816,18 @@ static int i915_drm_resume(struct drm_device *dev) int ret; disable_rpm_wakeref_asserts(&dev_priv->runtime_pm); - intel_sanitize_gt_powersave(dev_priv); - i915_gem_sanitize(dev_priv); + intel_rc6_ctx_wa_resume(&dev_priv->gt.rc6); + + intel_gt_sanitize(&dev_priv->gt, true); ret = i915_ggtt_enable_hw(dev_priv); if (ret) DRM_ERROR("failed to re-enable GGTT\n"); + i915_gem_restore_gtt_mappings(dev_priv); + i915_gem_restore_fences(&dev_priv->ggtt); + intel_csr_ucode_resume(dev_priv); i915_restore_state(dev_priv); @@ -1947,7 +1851,7 @@ static int i915_drm_resume(struct drm_device *dev) i915_gem_resume(dev_priv); - intel_modeset_init_hw(dev); + intel_modeset_init_hw(dev_priv); intel_init_clock_gating(dev_priv); spin_lock_irq(&dev_priv->irq_lock); @@ -2044,18 +1948,14 @@ static int i915_drm_resume_early(struct drm_device *dev) intel_display_power_resume_early(dev_priv); - intel_sanitize_gt_powersave(dev_priv); - intel_power_domains_resume(dev_priv); - intel_gt_sanitize(&dev_priv->gt, true); - enable_rpm_wakeref_asserts(&dev_priv->runtime_pm); return ret; } -static int i915_resume_switcheroo(struct drm_i915_private *i915) +int i915_resume_switcheroo(struct drm_i915_private *i915) { int ret; @@ -2588,9 +2488,6 @@ static int intel_runtime_suspend(struct device *kdev) struct intel_runtime_pm *rpm = &dev_priv->runtime_pm; int ret = 0; - if (WARN_ON_ONCE(!(dev_priv->gt_pm.rc6.enabled && HAS_RC6(dev_priv)))) - return -ENODEV; - if (WARN_ON_ONCE(!HAS_RUNTIME_PM(dev_priv))) return -ENODEV; @@ -2623,7 +2520,7 @@ static int intel_runtime_suspend(struct device *kdev) intel_gt_runtime_resume(&dev_priv->gt); - i915_gem_restore_fences(dev_priv); + i915_gem_restore_fences(&dev_priv->ggtt); enable_rpm_wakeref_asserts(rpm); @@ -2703,7 +2600,7 @@ static int intel_runtime_resume(struct device *kdev) * we can do is to hope that things will still work (and disable RPM). */ intel_gt_runtime_resume(&dev_priv->gt); - i915_gem_restore_fences(dev_priv); + i915_gem_restore_fences(&dev_priv->ggtt); /* * On VLV/CHV display interrupts are part of the display diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 772154e4073e..e29bc137e7ba 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -67,6 +67,7 @@ #include "display/intel_display.h" #include "display/intel_display_power.h" #include "display/intel_dpll_mgr.h" +#include "display/intel_dsb.h" #include "display/intel_frontbuffer.h" #include "display/intel_gmbus.h" #include "display/intel_opregion.h" @@ -84,6 +85,7 @@ #include "intel_device_info.h" #include "intel_pch.h" #include "intel_runtime_pm.h" +#include "intel_memory_region.h" #include "intel_uncore.h" #include "intel_wakeref.h" #include "intel_wopcm.h" @@ -92,12 +94,15 @@ #include "i915_gem_fence_reg.h" #include "i915_gem_gtt.h" #include "i915_gpu_error.h" +#include "i915_perf_types.h" #include "i915_request.h" #include "i915_scheduler.h" #include "gt/intel_timeline.h" #include "i915_vma.h" #include "i915_irq.h" +#include "intel_region_lmem.h" + #include "intel_gvt.h" /* General customization: @@ -105,8 +110,8 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20190822" -#define DRIVER_TIMESTAMP 1566477988 +#define DRIVER_DATE "20191101" +#define DRIVER_TIMESTAMP 1572604873 struct drm_i915_gem_object; @@ -185,7 +190,11 @@ struct i915_mmu_object; struct drm_i915_file_private { struct drm_i915_private *dev_priv; - struct drm_file *file; + + union { + struct drm_file *file; + struct rcu_head rcu; + }; struct { spinlock_t lock; @@ -272,6 +281,7 @@ struct drm_i915_display_funcs { int (*compute_global_watermarks)(struct intel_atomic_state *state); void (*update_wm)(struct intel_crtc *crtc); int (*modeset_calc_cdclk)(struct intel_atomic_state *state); + u8 (*calc_voltage_level)(int cdclk); /* Returns the active state of the crtc, and if the crtc is active, * fills out the pipe-config with the hw state. */ bool (*get_pipe_config)(struct intel_crtc *, @@ -284,7 +294,8 @@ struct drm_i915_display_funcs { struct intel_atomic_state *old_state); void (*crtc_disable)(struct intel_crtc_state *old_crtc_state, struct intel_atomic_state *old_state); - void (*update_crtcs)(struct intel_atomic_state *state); + void (*commit_modeset_enables)(struct intel_atomic_state *state); + void (*commit_modeset_disables)(struct intel_atomic_state *state); void (*audio_codec_enable)(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state, const struct drm_connector_state *conn_state); @@ -331,6 +342,7 @@ struct intel_csr { i915_reg_t mmioaddr[20]; u32 mmiodata[20]; u32 dc_state; + u32 target_dc_state; u32 allowed_dc_mask; intel_wakeref_t wakeref; }; @@ -479,6 +491,7 @@ struct i915_psr { bool enabled; struct intel_dp *dp; enum pipe pipe; + enum transcoder transcoder; bool active; struct work_struct work; unsigned busy_frontbuffer_bits; @@ -492,6 +505,9 @@ struct i915_psr { bool sink_not_reliable; bool irq_aux_error; u16 su_x_granularity; + bool dc3co_enabled; + u32 dc3co_exit_delay; + struct delayed_work idle_work; }; #define QUIRK_LVDS_SSC_DISABLE (1<<1) @@ -529,106 +545,6 @@ struct i915_suspend_saved_registers { struct vlv_s0ix_state; -struct intel_rps_ei { - ktime_t ktime; - u32 render_c0; - u32 media_c0; -}; - -struct intel_rps { - struct mutex lock; /* protects enabling and the worker */ - - /* - * work, interrupts_enabled and pm_iir are protected by - * dev_priv->irq_lock - */ - struct work_struct work; - bool interrupts_enabled; - u32 pm_iir; - - /* PM interrupt bits that should never be masked */ - u32 pm_intrmsk_mbz; - - /* Frequencies are stored in potentially platform dependent multiples. - * In other words, *_freq needs to be multiplied by X to be interesting. - * Soft limits are those which are used for the dynamic reclocking done - * by the driver (raise frequencies under heavy loads, and lower for - * lighter loads). Hard limits are those imposed by the hardware. - * - * A distinction is made for overclocking, which is never enabled by - * default, and is considered to be above the hard limit if it's - * possible at all. - */ - u8 cur_freq; /* Current frequency (cached, may not == HW) */ - u8 min_freq_softlimit; /* Minimum frequency permitted by the driver */ - u8 max_freq_softlimit; /* Max frequency permitted by the driver */ - u8 max_freq; /* Maximum frequency, RP0 if not overclocking */ - u8 min_freq; /* AKA RPn. Minimum frequency */ - u8 boost_freq; /* Frequency to request when wait boosting */ - u8 idle_freq; /* Frequency to request when we are idle */ - u8 efficient_freq; /* AKA RPe. Pre-determined balanced frequency */ - u8 rp1_freq; /* "less than" RP0 power/freqency */ - u8 rp0_freq; /* Non-overclocked max frequency. */ - u16 gpll_ref_freq; /* vlv/chv GPLL reference frequency */ - - int last_adj; - - struct { - struct mutex mutex; - - enum { LOW_POWER, BETWEEN, HIGH_POWER } mode; - unsigned int interactive; - - u8 up_threshold; /* Current %busy required to uplock */ - u8 down_threshold; /* Current %busy required to downclock */ - } power; - - bool enabled; - atomic_t num_waiters; - atomic_t boosts; - - /* manual wa residency calculations */ - struct intel_rps_ei ei; -}; - -struct intel_rc6 { - bool enabled; - u64 prev_hw_residency[4]; - u64 cur_residency[4]; -}; - -struct intel_llc_pstate { - bool enabled; -}; - -struct intel_gen6_power_mgmt { - struct intel_rps rps; - struct intel_rc6 rc6; - struct intel_llc_pstate llc_pstate; -}; - -/* defined intel_pm.c */ -extern spinlock_t mchdev_lock; - -struct intel_ilk_power_mgmt { - u8 cur_delay; - u8 min_delay; - u8 max_delay; - u8 fmax; - u8 fstart; - - u64 last_count1; - unsigned long last_time1; - unsigned long chipset_power; - u64 last_count2; - u64 last_time2; - unsigned long gfx_power; - u8 corr; - - int c_m; - int r_t; -}; - #define MAX_L3_SLICES 2 struct intel_l3_parity { u32 *remap_info[MAX_L3_SLICES]; @@ -677,6 +593,8 @@ struct i915_gem_mm { */ struct vfsmount *gemfs; + struct intel_memory_region *regions[INTEL_REGION_UNKNOWN]; + struct notifier_block oom_notifier; struct notifier_block vmap_notifier; struct shrinker shrinker; @@ -688,11 +606,6 @@ struct i915_gem_mm { */ struct workqueue_struct *userptr_wq; - /** Bit 6 swizzling required for X tiling */ - u32 bit_6_swizzle_x; - /** Bit 6 swizzling required for Y tiling */ - u32 bit_6_swizzle_y; - /* shrinker accounting, also useful for userland debugging */ u64 shrink_memory; u32 shrink_count; @@ -973,305 +886,6 @@ struct intel_wm_config { bool sprites_scaled; }; -struct i915_oa_format { - u32 format; - int size; -}; - -struct i915_oa_reg { - i915_reg_t addr; - u32 value; -}; - -struct i915_oa_config { - char uuid[UUID_STRING_LEN + 1]; - int id; - - const struct i915_oa_reg *mux_regs; - u32 mux_regs_len; - const struct i915_oa_reg *b_counter_regs; - u32 b_counter_regs_len; - const struct i915_oa_reg *flex_regs; - u32 flex_regs_len; - - struct attribute_group sysfs_metric; - struct attribute *attrs[2]; - struct device_attribute sysfs_metric_id; - - atomic_t ref_count; -}; - -struct i915_perf_stream; - -/** - * struct i915_perf_stream_ops - the OPs to support a specific stream type - */ -struct i915_perf_stream_ops { - /** - * @enable: Enables the collection of HW samples, either in response to - * `I915_PERF_IOCTL_ENABLE` or implicitly called when stream is opened - * without `I915_PERF_FLAG_DISABLED`. - */ - void (*enable)(struct i915_perf_stream *stream); - - /** - * @disable: Disables the collection of HW samples, either in response - * to `I915_PERF_IOCTL_DISABLE` or implicitly called before destroying - * the stream. - */ - void (*disable)(struct i915_perf_stream *stream); - - /** - * @poll_wait: Call poll_wait, passing a wait queue that will be woken - * once there is something ready to read() for the stream - */ - void (*poll_wait)(struct i915_perf_stream *stream, - struct file *file, - poll_table *wait); - - /** - * @wait_unlocked: For handling a blocking read, wait until there is - * something to ready to read() for the stream. E.g. wait on the same - * wait queue that would be passed to poll_wait(). - */ - int (*wait_unlocked)(struct i915_perf_stream *stream); - - /** - * @read: Copy buffered metrics as records to userspace - * **buf**: the userspace, destination buffer - * **count**: the number of bytes to copy, requested by userspace - * **offset**: zero at the start of the read, updated as the read - * proceeds, it represents how many bytes have been copied so far and - * the buffer offset for copying the next record. - * - * Copy as many buffered i915 perf samples and records for this stream - * to userspace as will fit in the given buffer. - * - * Only write complete records; returning -%ENOSPC if there isn't room - * for a complete record. - * - * Return any error condition that results in a short read such as - * -%ENOSPC or -%EFAULT, even though these may be squashed before - * returning to userspace. - */ - int (*read)(struct i915_perf_stream *stream, - char __user *buf, - size_t count, - size_t *offset); - - /** - * @destroy: Cleanup any stream specific resources. - * - * The stream will always be disabled before this is called. - */ - void (*destroy)(struct i915_perf_stream *stream); -}; - -/** - * struct i915_perf_stream - state for a single open stream FD - */ -struct i915_perf_stream { - /** - * @dev_priv: i915 drm device - */ - struct drm_i915_private *dev_priv; - - /** - * @link: Links the stream into ``&drm_i915_private->streams`` - */ - struct list_head link; - - /** - * @wakeref: As we keep the device awake while the perf stream is - * active, we track our runtime pm reference for later release. - */ - intel_wakeref_t wakeref; - - /** - * @sample_flags: Flags representing the `DRM_I915_PERF_PROP_SAMPLE_*` - * properties given when opening a stream, representing the contents - * of a single sample as read() by userspace. - */ - u32 sample_flags; - - /** - * @sample_size: Considering the configured contents of a sample - * combined with the required header size, this is the total size - * of a single sample record. - */ - int sample_size; - - /** - * @ctx: %NULL if measuring system-wide across all contexts or a - * specific context that is being monitored. - */ - struct i915_gem_context *ctx; - - /** - * @enabled: Whether the stream is currently enabled, considering - * whether the stream was opened in a disabled state and based - * on `I915_PERF_IOCTL_ENABLE` and `I915_PERF_IOCTL_DISABLE` calls. - */ - bool enabled; - - /** - * @ops: The callbacks providing the implementation of this specific - * type of configured stream. - */ - const struct i915_perf_stream_ops *ops; - - /** - * @oa_config: The OA configuration used by the stream. - */ - struct i915_oa_config *oa_config; - - /** - * The OA context specific information. - */ - struct intel_context *pinned_ctx; - u32 specific_ctx_id; - u32 specific_ctx_id_mask; - - struct hrtimer poll_check_timer; - wait_queue_head_t poll_wq; - bool pollin; - - bool periodic; - int period_exponent; - - /** - * State of the OA buffer. - */ - struct { - struct i915_vma *vma; - u8 *vaddr; - u32 last_ctx_id; - int format; - int format_size; - int size_exponent; - - /** - * Locks reads and writes to all head/tail state - * - * Consider: the head and tail pointer state needs to be read - * consistently from a hrtimer callback (atomic context) and - * read() fop (user context) with tail pointer updates happening - * in atomic context and head updates in user context and the - * (unlikely) possibility of read() errors needing to reset all - * head/tail state. - * - * Note: Contention/performance aren't currently a significant - * concern here considering the relatively low frequency of - * hrtimer callbacks (5ms period) and that reads typically only - * happen in response to a hrtimer event and likely complete - * before the next callback. - * - * Note: This lock is not held *while* reading and copying data - * to userspace so the value of head observed in htrimer - * callbacks won't represent any partial consumption of data. - */ - spinlock_t ptr_lock; - - /** - * One 'aging' tail pointer and one 'aged' tail pointer ready to - * used for reading. - * - * Initial values of 0xffffffff are invalid and imply that an - * update is required (and should be ignored by an attempted - * read) - */ - struct { - u32 offset; - } tails[2]; - - /** - * Index for the aged tail ready to read() data up to. - */ - unsigned int aged_tail_idx; - - /** - * A monotonic timestamp for when the current aging tail pointer - * was read; used to determine when it is old enough to trust. - */ - u64 aging_timestamp; - - /** - * Although we can always read back the head pointer register, - * we prefer to avoid trusting the HW state, just to avoid any - * risk that some hardware condition could * somehow bump the - * head pointer unpredictably and cause us to forward the wrong - * OA buffer data to userspace. - */ - u32 head; - } oa_buffer; -}; - -/** - * struct i915_oa_ops - Gen specific implementation of an OA unit stream - */ -struct i915_oa_ops { - /** - * @is_valid_b_counter_reg: Validates register's address for - * programming boolean counters for a particular platform. - */ - bool (*is_valid_b_counter_reg)(struct drm_i915_private *dev_priv, - u32 addr); - - /** - * @is_valid_mux_reg: Validates register's address for programming mux - * for a particular platform. - */ - bool (*is_valid_mux_reg)(struct drm_i915_private *dev_priv, u32 addr); - - /** - * @is_valid_flex_reg: Validates register's address for programming - * flex EU filtering for a particular platform. - */ - bool (*is_valid_flex_reg)(struct drm_i915_private *dev_priv, u32 addr); - - /** - * @enable_metric_set: Selects and applies any MUX configuration to set - * up the Boolean and Custom (B/C) counters that are part of the - * counter reports being sampled. May apply system constraints such as - * disabling EU clock gating as required. - */ - int (*enable_metric_set)(struct i915_perf_stream *stream); - - /** - * @disable_metric_set: Remove system constraints associated with using - * the OA unit. - */ - void (*disable_metric_set)(struct i915_perf_stream *stream); - - /** - * @oa_enable: Enable periodic sampling - */ - void (*oa_enable)(struct i915_perf_stream *stream); - - /** - * @oa_disable: Disable periodic sampling - */ - void (*oa_disable)(struct i915_perf_stream *stream); - - /** - * @read: Copy data from the circular OA buffer into a given userspace - * buffer. - */ - int (*read)(struct i915_perf_stream *stream, - char __user *buf, - size_t count, - size_t *offset); - - /** - * @oa_hw_tail_read: read the OA tail pointer register - * - * In particular this enables us to share all the fiddly code for - * handling the OA unit tail pointer race that affects multiple - * generations. - */ - u32 (*oa_hw_tail_read)(struct i915_perf_stream *stream); -}; - struct intel_cdclk_state { unsigned int cdclk, vco, ref, bypass; u8 voltage_level; @@ -1331,11 +945,11 @@ struct drm_i915_private { */ u32 gpio_mmio_base; + u32 hsw_psr_mmio_adjust; + /* MMIO base address for MIPI regs */ u32 mipi_mmio_base; - u32 psr_mmio_base; - u32 pps_mmio_base; wait_queue_head_t gmbus_wait_queue; @@ -1367,7 +981,6 @@ struct drm_i915_private { u32 irq_mask; u32 de_irq_mask[I915_MAX_PIPES]; }; - u32 pm_rps_events; u32 pipestat_irq_mask[I915_MAX_PIPES]; struct i915_hotplug hotplug; @@ -1397,13 +1010,14 @@ struct drm_i915_private { unsigned int fdi_pll_freq; unsigned int czclk_freq; + /* + * For reading holding any crtc lock is sufficient, + * for writing must hold all of them. + */ struct { /* * The current logical cdclk state. * See intel_atomic_state.cdclk.logical - * - * For reading holding any crtc lock is sufficient, - * for writing must hold all of them. */ struct intel_cdclk_state logical; /* @@ -1414,6 +1028,9 @@ struct drm_i915_private { /* The current hardware cdclk state */ struct intel_cdclk_state hw; + /* cdclk, divider, and ratio table from bspec */ + const struct intel_cdclk_vals *table; + int force_min_cdclk; } cdclk; @@ -1428,6 +1045,8 @@ struct drm_i915_private { /* ordered wq for modesets */ struct workqueue_struct *modeset_wq; + /* unbound hipri wq for page flips/plane updates */ + struct workqueue_struct *flip_wq; /* Display functions */ struct drm_i915_display_funcs display; @@ -1468,7 +1087,11 @@ struct drm_i915_private { */ struct mutex dpll_lock; - unsigned int active_crtcs; + /* + * For reading active_pipes, min_cdclk, min_voltage_level holding + * any crtc lock is sufficient, for writing must hold all of them. + */ + u8 active_pipes; /* minimum acceptable cdclk for each pipe */ int min_cdclk[I915_MAX_PIPES]; /* minimum acceptable voltage level for each pipe */ @@ -1497,13 +1120,6 @@ struct drm_i915_private { */ u32 edram_size_mb; - /* gen6+ GT PM state */ - struct intel_gen6_power_mgmt gt_pm; - - /* ilk-only ips/rps state. Everything in here is protected by the global - * mchdev_lock in intel_pm.c */ - struct intel_ilk_power_mgmt ips; - struct i915_power_domains power_domains; struct i915_psr psr; @@ -1528,25 +1144,7 @@ struct drm_i915_private { */ struct mutex av_mutex; int audio_power_refcount; - - struct { - struct mutex mutex; - struct list_head list; - struct llist_head free_list; - struct work_struct free_work; - - /* The hw wants to have a stable context identifier for the - * lifetime of the context (for OA, PASID, faults, etc). - * This is limited in execlists to 21 bits. - */ - struct ida hw_ida; -#define MAX_CONTEXT_HW_ID (1<<21) /* exclusive */ -#define MAX_GUC_CONTEXT_HW_ID (1 << 20) /* exclusive */ -#define GEN11_MAX_CONTEXT_HW_ID (1<<11) /* exclusive */ -/* in Gen12 ID 0x7FF is reserved to indicate idle */ -#define GEN12_MAX_CONTEXT_HW_ID (GEN11_MAX_CONTEXT_HW_ID - 1) - struct list_head hw_id_list; - } contexts; + u32 audio_freq_cntrl; u32 fdi_rx_config; @@ -1572,6 +1170,8 @@ struct drm_i915_private { I915_SAGV_NOT_CONTROLLED } sagv_status; + u32 sagv_block_time_us; + struct { /* * Raw watermark latency values: @@ -1642,61 +1242,7 @@ struct drm_i915_private { struct intel_runtime_pm runtime_pm; - struct { - bool initialized; - - struct kobject *metrics_kobj; - struct ctl_table_header *sysctl_header; - - /* - * Lock associated with adding/modifying/removing OA configs - * in dev_priv->perf.metrics_idr. - */ - struct mutex metrics_lock; - - /* - * List of dynamic configurations, you need to hold - * dev_priv->perf.metrics_lock to access it. - */ - struct idr metrics_idr; - - /* - * Lock associated with anything below within this structure - * except exclusive_stream. - */ - struct mutex lock; - struct list_head streams; - - /* - * The stream currently using the OA unit. If accessed - * outside a syscall associated to its file - * descriptor, you need to hold - * dev_priv->drm.struct_mutex. - */ - struct i915_perf_stream *exclusive_stream; - - /** - * For rate limiting any notifications of spurious - * invalid OA reports - */ - struct ratelimit_state spurious_report_rs; - - struct i915_oa_config test_config; - - u32 gen7_latched_oastatus1; - u32 ctx_oactxctrl_offset; - u32 ctx_flexeu0_offset; - - /** - * The RPT_ID/reason field for Gen8+ includes a bit - * to determine if the CTX ID in the report is valid - * but the specific bit differs between Gen 8 and 9 - */ - u32 gen8_valid_ctx_bit; - - struct i915_oa_ops ops; - const struct i915_oa_format *oa_formats; - } perf; + struct i915_perf perf; /* Abstract the submission mechanism (legacy ringbuffer or execlists) away */ struct intel_gt gt; @@ -1704,32 +1250,19 @@ struct drm_i915_private { struct { struct notifier_block pm_notifier; - /** - * We leave the user IRQ off as much as possible, - * but this means that requests will finish and never - * be retired once the system goes idle. Set a timer to - * fire periodically while the ring is running. When it - * fires, go retire requests. - */ - struct delayed_work retire_work; - - /** - * When we detect an idle GPU, we want to turn on - * powersaving features. So once we see that there - * are no more requests outstanding and no more - * arrive within a small period of time, we fire - * off the idle_work. - */ - struct work_struct idle_work; + struct i915_gem_contexts { + spinlock_t lock; /* locks list */ + struct list_head list; + + struct llist_head free_list; + struct work_struct free_work; + } contexts; } gem; - /* For i945gm vblank irq vs. C3 workaround */ - struct { - struct work_struct work; - struct pm_qos_request pm_qos; - u8 c3_disable_latency; - u8 enabled; - } i945gm_vblank; + u8 pch_ssc_use; + + /* For i915gm/i945gm vblank irq workaround */ + u8 vblank_enabled; /* perform PHY state sanity checks? */ bool chv_phy_assert[2]; @@ -1792,10 +1325,10 @@ static inline struct drm_i915_private *pdev_to_i915(struct pci_dev *pdev) for_each_if ((engine__) = (dev_priv__)->engine[(id__)]) /* Iterator over subset of engines selected by mask */ -#define for_each_engine_masked(engine__, dev_priv__, mask__, tmp__) \ - for ((tmp__) = (mask__) & INTEL_INFO(dev_priv__)->engine_mask; \ +#define for_each_engine_masked(engine__, gt__, mask__, tmp__) \ + for ((tmp__) = (mask__) & INTEL_INFO((gt__)->i915)->engine_mask; \ (tmp__) ? \ - ((engine__) = (dev_priv__)->engine[__mask_next_bit(tmp__)]), 1 : \ + ((engine__) = (gt__)->engine[__mask_next_bit(tmp__)]), 1 : \ 0;) #define rb_to_uabi_engine(rb) \ @@ -1851,6 +1384,8 @@ static inline struct drm_i915_private *pdev_to_i915(struct pci_dev *pdev) (BUILD_BUG_ON_ZERO(!__builtin_constant_p(n)) + \ INTEL_INFO(dev_priv)->gen == (n)) +#define HAS_DSB(dev_priv) (INTEL_INFO(dev_priv)->display.has_dsb) + /* * Return true if revision is in range [since,until] inclusive. * @@ -1922,6 +1457,7 @@ IS_SUBPLATFORM(const struct drm_i915_private *i915, } #define IS_MOBILE(dev_priv) (INTEL_INFO(dev_priv)->is_mobile) +#define IS_DGFX(dev_priv) (INTEL_INFO(dev_priv)->is_dgfx) #define IS_I830(dev_priv) IS_PLATFORM(dev_priv, INTEL_I830) #define IS_I845G(dev_priv) IS_PLATFORM(dev_priv, INTEL_I845G) @@ -2056,6 +1592,11 @@ IS_SUBPLATFORM(const struct drm_i915_private *i915, #define IS_ICL_REVID(p, since, until) \ (IS_ICELAKE(p) && IS_REVID(p, since, until)) +#define TGL_REVID_A0 0x0 + +#define IS_TGL_REVID(p, since, until) \ + (IS_TIGERLAKE(p) && IS_REVID(p, since, until)) + #define IS_LP(dev_priv) (INTEL_INFO(dev_priv)->is_lp) #define IS_GEN9_LP(dev_priv) (IS_GEN(dev_priv, 9) && IS_LP(dev_priv)) #define IS_GEN9_BC(dev_priv) (IS_GEN(dev_priv, 9) && !IS_LP(dev_priv)) @@ -2073,9 +1614,16 @@ IS_SUBPLATFORM(const struct drm_i915_private *i915, #define VEBOX_MASK(dev_priv) \ ENGINE_INSTANCES_MASK(dev_priv, VECS0, I915_MAX_VECS) +/* + * The Gen7 cmdparser copies the scanned buffer to the ggtt for execution + * All later gens can run the final buffer from the ppgtt + */ +#define CMDPARSER_USES_GGTT(dev_priv) IS_GEN(dev_priv, 7) + #define HAS_LLC(dev_priv) (INTEL_INFO(dev_priv)->has_llc) #define HAS_SNOOP(dev_priv) (INTEL_INFO(dev_priv)->has_snoop) #define HAS_EDRAM(dev_priv) ((dev_priv)->edram_size_mb) +#define HAS_SECURE_BATCHES(dev_priv) (INTEL_GEN(dev_priv) < 6) #define HAS_WT(dev_priv) ((IS_HASWELL(dev_priv) || \ IS_BROADWELL(dev_priv)) && HAS_EDRAM(dev_priv)) @@ -2108,10 +1656,12 @@ IS_SUBPLATFORM(const struct drm_i915_private *i915, /* Early gen2 have a totally busted CS tlb and require pinned batches. */ #define HAS_BROKEN_CS_TLB(dev_priv) (IS_I830(dev_priv) || IS_I845G(dev_priv)) +#define NEEDS_RC6_CTX_CORRUPTION_WA(dev_priv) \ + (IS_BROADWELL(dev_priv) || IS_GEN(dev_priv, 9)) + /* WaRsDisableCoarsePowerGating:skl,cnl */ #define NEEDS_WaRsDisableCoarsePowerGating(dev_priv) \ - (IS_CANNONLAKE(dev_priv) || \ - IS_SKL_GT3(dev_priv) || IS_SKL_GT4(dev_priv)) + (IS_CANNONLAKE(dev_priv) || IS_GEN(dev_priv, 9)) #define HAS_GMBUS_IRQ(dev_priv) (INTEL_GEN(dev_priv) >= 4) #define HAS_GMBUS_BURST_READ(dev_priv) (INTEL_GEN(dev_priv) >= 10 || \ @@ -2153,6 +1703,9 @@ IS_SUBPLATFORM(const struct drm_i915_private *i915, #define HAS_IPC(dev_priv) (INTEL_INFO(dev_priv)->display.has_ipc) +#define HAS_REGION(i915, i) (INTEL_INFO(i915)->memory_regions & (i)) +#define HAS_LMEM(i915) HAS_REGION(i915, REGION_LMEM) + #define HAS_GT_UC(dev_priv) (INTEL_INFO(dev_priv)->has_gt_uc) /* Having GuC is not the same as using GuC */ @@ -2176,7 +1729,12 @@ IS_SUBPLATFORM(const struct drm_i915_private *i915, #define GT_FREQUENCY_MULTIPLIER 50 #define GEN9_FREQ_SCALER 3 -#define HAS_DISPLAY(dev_priv) (INTEL_INFO(dev_priv)->num_pipes > 0) +#define INTEL_NUM_PIPES(dev_priv) (hweight8(INTEL_INFO(dev_priv)->pipe_mask)) + +#define HAS_DISPLAY(dev_priv) (INTEL_INFO(dev_priv)->pipe_mask != 0) + +/* Only valid when HAS_DISPLAY() is true */ +#define INTEL_DISPLAY_ENABLED(dev_priv) (WARN_ON(!HAS_DISPLAY(dev_priv)), !i915_modparams.disable_display) static inline bool intel_vtd_active(void) { @@ -2209,7 +1767,9 @@ extern const struct dev_pm_ops i915_pm_ops; int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent); void i915_driver_remove(struct drm_i915_private *i915); -void intel_engine_init_hangcheck(struct intel_engine_cs *engine); +int i915_resume_switcheroo(struct drm_i915_private *i915); +int i915_suspend_switcheroo(struct drm_i915_private *i915, pm_message_t state); + int vlv_force_gfx_clock(struct drm_i915_private *dev_priv, bool on); static inline bool intel_gvt_active(struct drm_i915_private *dev_priv) @@ -2228,12 +1788,13 @@ int i915_getparam_ioctl(struct drm_device *dev, void *data, /* i915_gem.c */ int i915_gem_init_userptr(struct drm_i915_private *dev_priv); void i915_gem_cleanup_userptr(struct drm_i915_private *dev_priv); -void i915_gem_sanitize(struct drm_i915_private *i915); -int i915_gem_init_early(struct drm_i915_private *dev_priv); +void i915_gem_init_early(struct drm_i915_private *dev_priv); void i915_gem_cleanup_early(struct drm_i915_private *dev_priv); int i915_gem_freeze(struct drm_i915_private *dev_priv); int i915_gem_freeze_late(struct drm_i915_private *dev_priv); +struct intel_memory_region *i915_gem_shmem_setup(struct drm_i915_private *i915); + static inline void i915_gem_drain_freed_objects(struct drm_i915_private *i915) { /* @@ -2282,6 +1843,14 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj, unsigned long flags); #define I915_GEM_OBJECT_UNBIND_ACTIVE BIT(0) +struct i915_vma * __must_check +i915_gem_object_pin(struct drm_i915_gem_object *obj, + struct i915_address_space *vm, + const struct i915_ggtt_view *view, + u64 size, + u64 alignment, + u64 flags); + void i915_gem_runtime_suspend(struct drm_i915_private *dev_priv); static inline int __must_check @@ -2310,15 +1879,11 @@ static inline u32 i915_reset_engine_count(struct i915_gpu_error *error, return atomic_read(&error->reset_engine_count[engine->uabi_class]); } -void i915_gem_init_mmio(struct drm_i915_private *i915); int __must_check i915_gem_init(struct drm_i915_private *dev_priv); -int __must_check i915_gem_init_hw(struct drm_i915_private *dev_priv); void i915_gem_driver_register(struct drm_i915_private *i915); void i915_gem_driver_unregister(struct drm_i915_private *i915); void i915_gem_driver_remove(struct drm_i915_private *dev_priv); void i915_gem_driver_release(struct drm_i915_private *dev_priv); -int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv, - unsigned int flags, long timeout); void i915_gem_suspend(struct drm_i915_private *dev_priv); void i915_gem_suspend_late(struct drm_i915_private *dev_priv); void i915_gem_resume(struct drm_i915_private *dev_priv); @@ -2358,7 +1923,7 @@ i915_gem_context_lookup(struct drm_i915_file_private *file_priv, u32 id) /* i915_gem_evict.c */ int __must_check i915_gem_evict_something(struct i915_address_space *vm, u64 min_size, u64 alignment, - unsigned cache_level, + unsigned long color, u64 start, u64 end, unsigned flags); int __must_check i915_gem_evict_for_node(struct i915_address_space *vm, @@ -2374,9 +1939,9 @@ i915_gem_object_create_internal(struct drm_i915_private *dev_priv, /* i915_gem_tiling.c */ static inline bool i915_gem_object_needs_bit17_swizzle(struct drm_i915_gem_object *obj) { - struct drm_i915_private *dev_priv = to_i915(obj->base.dev); + struct drm_i915_private *i915 = to_i915(obj->base.dev); - return dev_priv->mm.bit_6_swizzle_x == I915_BIT_6_SWIZZLE_9_10_17 && + return i915->ggtt.bit_6_swizzle_x == I915_BIT_6_SWIZZLE_9_10_17 && i915_gem_object_is_tiled(obj); } @@ -2391,12 +1956,14 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type); int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv); void intel_engine_init_cmd_parser(struct intel_engine_cs *engine); void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine); -int intel_engine_cmd_parser(struct intel_engine_cs *engine, +int intel_engine_cmd_parser(struct i915_gem_context *cxt, + struct intel_engine_cs *engine, struct drm_i915_gem_object *batch_obj, - struct drm_i915_gem_object *shadow_batch_obj, + u64 user_batch_start, u32 batch_start_offset, u32 batch_len, - bool is_master); + struct drm_i915_gem_object *shadow_batch_obj, + u64 shadow_batch_start); /* intel_device_info.c */ static inline struct intel_device_info * @@ -2478,4 +2045,10 @@ i915_coherent_map_type(struct drm_i915_private *i915) return HAS_LLC(i915) ? I915_MAP_WB : I915_MAP_WC; } +static inline bool intel_guc_submission_is_enabled(struct intel_guc *guc) +{ + return intel_guc_is_submission_supported(guc) && + intel_guc_is_running(guc); +} + #endif diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 95e7c52cf8ed..b9eb6b3149b7 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -45,13 +45,14 @@ #include "gem/i915_gem_context.h" #include "gem/i915_gem_ioctls.h" #include "gem/i915_gem_pm.h" -#include "gem/i915_gemfs.h" #include "gt/intel_engine_user.h" #include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" +#include "gt/intel_gt_requests.h" #include "gt/intel_mocs.h" #include "gt/intel_reset.h" #include "gt/intel_renderstate.h" +#include "gt/intel_rps.h" #include "gt/intel_workarounds.h" #include "i915_drv.h" @@ -62,20 +63,31 @@ #include "intel_pm.h" static int -insert_mappable_node(struct i915_ggtt *ggtt, - struct drm_mm_node *node, u32 size) +insert_mappable_node(struct i915_ggtt *ggtt, struct drm_mm_node *node, u32 size) { + int err; + + err = mutex_lock_interruptible(&ggtt->vm.mutex); + if (err) + return err; + memset(node, 0, sizeof(*node)); - return drm_mm_insert_node_in_range(&ggtt->vm.mm, node, - size, 0, I915_COLOR_UNEVICTABLE, - 0, ggtt->mappable_end, - DRM_MM_INSERT_LOW); + err = drm_mm_insert_node_in_range(&ggtt->vm.mm, node, + size, 0, I915_COLOR_UNEVICTABLE, + 0, ggtt->mappable_end, + DRM_MM_INSERT_LOW); + + mutex_unlock(&ggtt->vm.mutex); + + return err; } static void -remove_mappable_node(struct drm_mm_node *node) +remove_mappable_node(struct i915_ggtt *ggtt, struct drm_mm_node *node) { + mutex_lock(&ggtt->vm.mutex); drm_mm_remove_node(node); + mutex_unlock(&ggtt->vm.mutex); } int @@ -87,7 +99,8 @@ i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data, struct i915_vma *vma; u64 pinned; - mutex_lock(&ggtt->vm.mutex); + if (mutex_lock_interruptible(&ggtt->vm.mutex)) + return -EINTR; pinned = ggtt->vm.reserved; list_for_each_entry(vma, &ggtt->vm.bound_list, vm_link) @@ -109,20 +122,24 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj, LIST_HEAD(still_in_list); int ret = 0; - lockdep_assert_held(&obj->base.dev->struct_mutex); - spin_lock(&obj->vma.lock); while (!ret && (vma = list_first_entry_or_null(&obj->vma.list, struct i915_vma, obj_link))) { + struct i915_address_space *vm = vma->vm; + + ret = -EBUSY; + if (!i915_vm_tryopen(vm)) + break; + list_move_tail(&vma->obj_link, &still_in_list); spin_unlock(&obj->vma.lock); - ret = -EBUSY; if (flags & I915_GEM_OBJECT_UNBIND_ACTIVE || !i915_vma_is_active(vma)) ret = i915_vma_unbind(vma); + i915_vm_close(vm); spin_lock(&obj->vma.lock); } list_splice(&still_in_list, &obj->vma.list); @@ -338,10 +355,6 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj, u64 remain, offset; int ret; - ret = mutex_lock_interruptible(&i915->drm.struct_mutex); - if (ret) - return ret; - wakeref = intel_runtime_pm_get(&i915->runtime_pm); vma = ERR_PTR(-ENODEV); if (!i915_gem_object_is_tiled(obj)) @@ -351,16 +364,14 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj, PIN_NOEVICT); if (!IS_ERR(vma)) { node.start = i915_ggtt_offset(vma); - node.allocated = false; + node.flags = 0; } else { ret = insert_mappable_node(ggtt, &node, PAGE_SIZE); if (ret) - goto out_unlock; - GEM_BUG_ON(!node.allocated); + goto out_rpm; + GEM_BUG_ON(!drm_mm_node_allocated(&node)); } - mutex_unlock(&i915->drm.struct_mutex); - ret = i915_gem_object_lock_interruptible(obj); if (ret) goto out_unpin; @@ -393,7 +404,7 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj, unsigned page_offset = offset_in_page(offset); unsigned page_length = PAGE_SIZE - page_offset; page_length = remain < page_length ? remain : page_length; - if (node.allocated) { + if (drm_mm_node_allocated(&node)) { ggtt->vm.insert_page(&ggtt->vm, i915_gem_object_get_dma_address(obj, offset >> PAGE_SHIFT), node.start, I915_CACHE_NONE, 0); @@ -414,17 +425,14 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj, i915_gem_object_unlock_fence(obj, fence); out_unpin: - mutex_lock(&i915->drm.struct_mutex); - if (node.allocated) { + if (drm_mm_node_allocated(&node)) { ggtt->vm.clear_range(&ggtt->vm, node.start, node.size); - remove_mappable_node(&node); + remove_mappable_node(ggtt, &node); } else { i915_vma_unpin(vma); } -out_unlock: +out_rpm: intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - return ret; } @@ -531,10 +539,6 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj, void __user *user_data; int ret; - ret = mutex_lock_interruptible(&i915->drm.struct_mutex); - if (ret) - return ret; - if (i915_gem_object_has_struct_page(obj)) { /* * Avoid waking the device up if we can fallback, as @@ -544,10 +548,8 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj, * using the cache bypass of indirect GGTT access. */ wakeref = intel_runtime_pm_get_if_in_use(rpm); - if (!wakeref) { - ret = -EFAULT; - goto out_unlock; - } + if (!wakeref) + return -EFAULT; } else { /* No backing pages, no fallback, we must force GGTT access */ wakeref = intel_runtime_pm_get(rpm); @@ -561,16 +563,14 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj, PIN_NOEVICT); if (!IS_ERR(vma)) { node.start = i915_ggtt_offset(vma); - node.allocated = false; + node.flags = 0; } else { ret = insert_mappable_node(ggtt, &node, PAGE_SIZE); if (ret) goto out_rpm; - GEM_BUG_ON(!node.allocated); + GEM_BUG_ON(!drm_mm_node_allocated(&node)); } - mutex_unlock(&i915->drm.struct_mutex); - ret = i915_gem_object_lock_interruptible(obj); if (ret) goto out_unpin; @@ -604,7 +604,7 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj, unsigned int page_offset = offset_in_page(offset); unsigned int page_length = PAGE_SIZE - page_offset; page_length = remain < page_length ? remain : page_length; - if (node.allocated) { + if (drm_mm_node_allocated(&node)) { /* flush the write before we modify the GGTT */ intel_gt_flush_ggtt_writes(ggtt->vm.gt); ggtt->vm.insert_page(&ggtt->vm, @@ -634,18 +634,15 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj, i915_gem_object_unlock_fence(obj, fence); out_unpin: - mutex_lock(&i915->drm.struct_mutex); intel_gt_flush_ggtt_writes(ggtt->vm.gt); - if (node.allocated) { + if (drm_mm_node_allocated(&node)) { ggtt->vm.clear_range(&ggtt->vm, node.start, node.size); - remove_mappable_node(&node); + remove_mappable_node(ggtt, &node); } else { i915_vma_unpin(vma); } out_rpm: intel_runtime_pm_put(rpm, wakeref); -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); return ret; } @@ -887,74 +884,6 @@ void i915_gem_runtime_suspend(struct drm_i915_private *i915) } } -static long -wait_for_timelines(struct drm_i915_private *i915, - unsigned int wait, long timeout) -{ - struct intel_gt_timelines *timelines = &i915->gt.timelines; - struct intel_timeline *tl; - unsigned long flags; - - spin_lock_irqsave(&timelines->lock, flags); - list_for_each_entry(tl, &timelines->active_list, link) { - struct i915_request *rq; - - rq = i915_active_request_get_unlocked(&tl->last_request); - if (!rq) - continue; - - spin_unlock_irqrestore(&timelines->lock, flags); - - /* - * "Race-to-idle". - * - * Switching to the kernel context is often used a synchronous - * step prior to idling, e.g. in suspend for flushing all - * current operations to memory before sleeping. These we - * want to complete as quickly as possible to avoid prolonged - * stalls, so allow the gpu to boost to maximum clocks. - */ - if (wait & I915_WAIT_FOR_IDLE_BOOST) - gen6_rps_boost(rq); - - timeout = i915_request_wait(rq, wait, timeout); - i915_request_put(rq); - if (timeout < 0) - return timeout; - - /* restart after reacquiring the lock */ - spin_lock_irqsave(&timelines->lock, flags); - tl = list_entry(&timelines->active_list, typeof(*tl), link); - } - spin_unlock_irqrestore(&timelines->lock, flags); - - return timeout; -} - -int i915_gem_wait_for_idle(struct drm_i915_private *i915, - unsigned int flags, long timeout) -{ - /* If the device is asleep, we have no requests outstanding */ - if (!intel_gt_pm_is_awake(&i915->gt)) - return 0; - - GEM_TRACE("flags=%x (%s), timeout=%ld%s\n", - flags, flags & I915_WAIT_LOCKED ? "locked" : "unlocked", - timeout, timeout == MAX_SCHEDULE_TIMEOUT ? " (forever)" : ""); - - timeout = wait_for_timelines(i915, flags, timeout); - if (timeout < 0) - return timeout; - - if (flags & I915_WAIT_LOCKED) { - lockdep_assert_held(&i915->drm.struct_mutex); - - i915_retire_requests(i915); - } - - return 0; -} - struct i915_vma * i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj, const struct i915_ggtt_view *view, @@ -964,10 +893,25 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj, { struct drm_i915_private *dev_priv = to_i915(obj->base.dev); struct i915_address_space *vm = &dev_priv->ggtt.vm; + + return i915_gem_object_pin(obj, vm, view, size, alignment, + flags | PIN_GLOBAL); +} + +struct i915_vma * +i915_gem_object_pin(struct drm_i915_gem_object *obj, + struct i915_address_space *vm, + const struct i915_ggtt_view *view, + u64 size, + u64 alignment, + u64 flags) +{ + struct drm_i915_private *dev_priv = to_i915(obj->base.dev); struct i915_vma *vma; int ret; - lockdep_assert_held(&obj->base.dev->struct_mutex); + if (i915_gem_object_never_bind_ggtt(obj)) + return ERR_PTR(-ENODEV); if (flags & PIN_MAPPABLE && (!view || view->type == I915_GGTT_VIEW_NORMAL)) { @@ -1015,13 +959,6 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj, return ERR_PTR(-ENOSPC); } - WARN(i915_vma_is_pinned(vma), - "bo is already pinned in ggtt with incorrect alignment:" - " offset=%08x, req.alignment=%llx," - " req.map_and_fenceable=%d, vma->map_and_fenceable=%d\n", - i915_ggtt_offset(vma), alignment, - !!(flags & PIN_MAPPABLE), - i915_vma_is_map_and_fenceable(vma)); ret = i915_vma_unbind(vma); if (ret) return ERR_PTR(ret); @@ -1035,7 +972,7 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj, return ERR_PTR(ret); } - ret = i915_vma_pin(vma, size, alignment, flags | PIN_GLOBAL); + ret = i915_vma_pin(vma, size, alignment, flags); if (ret) return ERR_PTR(ret); @@ -1116,128 +1053,7 @@ out: return err; } -void i915_gem_sanitize(struct drm_i915_private *i915) -{ - intel_wakeref_t wakeref; - - GEM_TRACE("\n"); - - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - intel_uncore_forcewake_get(&i915->uncore, FORCEWAKE_ALL); - - /* - * As we have just resumed the machine and woken the device up from - * deep PCI sleep (presumably D3_cold), assume the HW has been reset - * back to defaults, recovering from whatever wedged state we left it - * in and so worth trying to use the device once more. - */ - if (intel_gt_is_wedged(&i915->gt)) - intel_gt_unset_wedged(&i915->gt); - - /* - * If we inherit context state from the BIOS or earlier occupants - * of the GPU, the GPU may be in an inconsistent state when we - * try to take over. The only way to remove the earlier state - * is by resetting. However, resetting on earlier gen is tricky as - * it may impact the display and we are uncertain about the stability - * of the reset, so this could be applied to even earlier gen. - */ - intel_gt_sanitize(&i915->gt, false); - - intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL); - intel_runtime_pm_put(&i915->runtime_pm, wakeref); -} - -static void init_unused_ring(struct intel_gt *gt, u32 base) -{ - struct intel_uncore *uncore = gt->uncore; - - intel_uncore_write(uncore, RING_CTL(base), 0); - intel_uncore_write(uncore, RING_HEAD(base), 0); - intel_uncore_write(uncore, RING_TAIL(base), 0); - intel_uncore_write(uncore, RING_START(base), 0); -} - -static void init_unused_rings(struct intel_gt *gt) -{ - struct drm_i915_private *i915 = gt->i915; - - if (IS_I830(i915)) { - init_unused_ring(gt, PRB1_BASE); - init_unused_ring(gt, SRB0_BASE); - init_unused_ring(gt, SRB1_BASE); - init_unused_ring(gt, SRB2_BASE); - init_unused_ring(gt, SRB3_BASE); - } else if (IS_GEN(i915, 2)) { - init_unused_ring(gt, SRB0_BASE); - init_unused_ring(gt, SRB1_BASE); - } else if (IS_GEN(i915, 3)) { - init_unused_ring(gt, PRB1_BASE); - init_unused_ring(gt, PRB2_BASE); - } -} - -int i915_gem_init_hw(struct drm_i915_private *i915) -{ - struct intel_uncore *uncore = &i915->uncore; - struct intel_gt *gt = &i915->gt; - int ret; - - BUG_ON(!i915->kernel_context); - ret = intel_gt_terminally_wedged(gt); - if (ret) - return ret; - - gt->last_init_time = ktime_get(); - - /* Double layer security blanket, see i915_gem_init() */ - intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); - - if (HAS_EDRAM(i915) && INTEL_GEN(i915) < 9) - intel_uncore_rmw(uncore, HSW_IDICR, 0, IDIHASHMSK(0xf)); - - if (IS_HASWELL(i915)) - intel_uncore_write(uncore, - MI_PREDICATE_RESULT_2, - IS_HSW_GT3(i915) ? - LOWER_SLICE_ENABLED : LOWER_SLICE_DISABLED); - - /* Apply the GT workarounds... */ - intel_gt_apply_workarounds(gt); - /* ...and determine whether they are sticking. */ - intel_gt_verify_workarounds(gt, "init"); - - intel_gt_init_swizzling(gt); - - /* - * At least 830 can leave some of the unused rings - * "active" (ie. head != tail) after resume which - * will prevent c3 entry. Makes sure all unused rings - * are totally idle. - */ - init_unused_rings(gt); - - ret = i915_ppgtt_init_hw(gt); - if (ret) { - DRM_ERROR("Enabling PPGTT failed (%d)\n", ret); - goto out; - } - - /* We can't enable contexts until all firmware is loaded */ - ret = intel_uc_init_hw(>->uc); - if (ret) { - i915_probe_error(i915, "Enabling uc failed (%d)\n", ret); - goto out; - } - - intel_mocs_init(gt); - -out: - intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); - return ret; -} - -static int __intel_engines_record_defaults(struct drm_i915_private *i915) +static int __intel_engines_record_defaults(struct intel_gt *gt) { struct i915_request *requests[I915_NUM_ENGINES] = {}; struct intel_engine_cs *engine; @@ -1253,7 +1069,7 @@ static int __intel_engines_record_defaults(struct drm_i915_private *i915) * from the same default HW values. */ - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct intel_context *ce; struct i915_request *rq; @@ -1261,7 +1077,8 @@ static int __intel_engines_record_defaults(struct drm_i915_private *i915) GEM_BUG_ON(!engine->kernel_context); engine->serial++; /* force the kernel context switch */ - ce = intel_context_create(i915->kernel_context, engine); + ce = intel_context_create(engine->kernel_context->gem_context, + engine); if (IS_ERR(ce)) { err = PTR_ERR(ce); goto out; @@ -1278,15 +1095,6 @@ static int __intel_engines_record_defaults(struct drm_i915_private *i915) if (err) goto err_rq; - /* - * Failing to program the MOCS is non-fatal.The system will not - * run at peak performance. So warn the user and carry on. - */ - err = intel_mocs_emit(rq); - if (err) - dev_notice(i915->drm.dev, - "Failed to program MOCS registers; expect performance issues.\n"); - err = intel_renderstate_emit(rq); if (err) goto err_rq; @@ -1299,7 +1107,7 @@ err_rq: } /* Flush the default context image to memory, and enable powersaving. */ - if (!i915_gem_load_power_context(i915)) { + if (intel_gt_wait_for_idle(gt, I915_GEM_IDLE_TIMEOUT) == -ETIME) { err = -EIO; goto out; } @@ -1358,7 +1166,7 @@ out: * this is by declaring ourselves wedged. */ if (err) - intel_gt_set_wedged(&i915->gt); + intel_gt_set_wedged(gt); for (id = 0; id < ARRAY_SIZE(requests); id++) { struct intel_context *ce; @@ -1375,18 +1183,7 @@ out: return err; } -static int -i915_gem_init_scratch(struct drm_i915_private *i915, unsigned int size) -{ - return intel_gt_init_scratch(&i915->gt, size); -} - -static void i915_gem_fini_scratch(struct drm_i915_private *i915) -{ - intel_gt_fini_scratch(&i915->gt); -} - -static int intel_engines_verify_workarounds(struct drm_i915_private *i915) +static int intel_engines_verify_workarounds(struct intel_gt *gt) { struct intel_engine_cs *engine; enum intel_engine_id id; @@ -1395,7 +1192,7 @@ static int intel_engines_verify_workarounds(struct drm_i915_private *i915) if (!IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) return 0; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { if (intel_engine_verify_workarounds(engine, "load")) err = -EIO; } @@ -1427,7 +1224,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv) * we hold the forcewake during initialisation these problems * just magically go away. */ - mutex_lock(&dev_priv->drm.struct_mutex); intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); ret = i915_init_ggtt(dev_priv); @@ -1436,36 +1232,29 @@ int i915_gem_init(struct drm_i915_private *dev_priv) goto err_unlock; } - ret = i915_gem_init_scratch(dev_priv, - IS_GEN(dev_priv, 2) ? SZ_256K : PAGE_SIZE); - if (ret) { - GEM_BUG_ON(ret == -EIO); - goto err_ggtt; - } + intel_gt_init(&dev_priv->gt); - ret = intel_engines_setup(dev_priv); + ret = intel_engines_setup(&dev_priv->gt); if (ret) { GEM_BUG_ON(ret == -EIO); goto err_unlock; } - ret = i915_gem_contexts_init(dev_priv); + ret = i915_gem_init_contexts(dev_priv); if (ret) { GEM_BUG_ON(ret == -EIO); goto err_scratch; } - ret = intel_engines_init(dev_priv); + ret = intel_engines_init(&dev_priv->gt); if (ret) { GEM_BUG_ON(ret == -EIO); goto err_context; } - intel_init_gt_powersave(dev_priv); - intel_uc_init(&dev_priv->gt.uc); - ret = i915_gem_init_hw(dev_priv); + ret = intel_gt_init_hw(&dev_priv->gt); if (ret) goto err_uc_init; @@ -1485,24 +1274,23 @@ int i915_gem_init(struct drm_i915_private *dev_priv) */ intel_init_clock_gating(dev_priv); - ret = intel_engines_verify_workarounds(dev_priv); + ret = intel_engines_verify_workarounds(&dev_priv->gt); if (ret) goto err_gt; - ret = __intel_engines_record_defaults(dev_priv); + ret = __intel_engines_record_defaults(&dev_priv->gt); if (ret) goto err_gt; - ret = i915_inject_load_error(dev_priv, -ENODEV); + ret = i915_inject_probe_error(dev_priv, -ENODEV); if (ret) goto err_gt; - ret = i915_inject_load_error(dev_priv, -EIO); + ret = i915_inject_probe_error(dev_priv, -EIO); if (ret) goto err_gt; intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); - mutex_unlock(&dev_priv->drm.struct_mutex); return 0; @@ -1513,32 +1301,25 @@ int i915_gem_init(struct drm_i915_private *dev_priv) * driver doesn't explode during runtime. */ err_gt: - mutex_unlock(&dev_priv->drm.struct_mutex); - - intel_gt_set_wedged(&dev_priv->gt); + intel_gt_set_wedged_on_init(&dev_priv->gt); i915_gem_suspend(dev_priv); i915_gem_suspend_late(dev_priv); i915_gem_drain_workqueue(dev_priv); - - mutex_lock(&dev_priv->drm.struct_mutex); err_init_hw: intel_uc_fini_hw(&dev_priv->gt.uc); err_uc_init: if (ret != -EIO) { intel_uc_fini(&dev_priv->gt.uc); - intel_cleanup_gt_powersave(dev_priv); - intel_engines_cleanup(dev_priv); + intel_engines_cleanup(&dev_priv->gt); } err_context: if (ret != -EIO) - i915_gem_contexts_fini(dev_priv); + i915_gem_driver_release__contexts(dev_priv); err_scratch: - i915_gem_fini_scratch(dev_priv); -err_ggtt: + intel_gt_driver_release(&dev_priv->gt); err_unlock: intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); - mutex_unlock(&dev_priv->drm.struct_mutex); if (ret != -EIO) { intel_uc_cleanup_firmwares(&dev_priv->gt.uc); @@ -1547,8 +1328,6 @@ err_unlock: } if (ret == -EIO) { - mutex_lock(&dev_priv->drm.struct_mutex); - /* * Allow engines or uC initialisation to fail by marking the GPU * as wedged. But we only want to do this when the GPU is angry, @@ -1563,10 +1342,8 @@ err_unlock: /* Minimal basic recovery for KMS */ ret = i915_ggtt_enable_hw(dev_priv); i915_gem_restore_gtt_mappings(dev_priv); - i915_gem_restore_fences(dev_priv); + i915_gem_restore_fences(&dev_priv->ggtt); intel_init_clock_gating(dev_priv); - - mutex_unlock(&dev_priv->drm.struct_mutex); } i915_gem_drain_freed_objects(dev_priv); @@ -1587,48 +1364,35 @@ void i915_gem_driver_unregister(struct drm_i915_private *i915) void i915_gem_driver_remove(struct drm_i915_private *dev_priv) { - GEM_BUG_ON(dev_priv->gt.awake); - intel_wakeref_auto_fini(&dev_priv->ggtt.userfault_wakeref); i915_gem_suspend_late(dev_priv); - intel_disable_gt_powersave(dev_priv); + intel_gt_driver_remove(&dev_priv->gt); /* Flush any outstanding unpin_work. */ i915_gem_drain_workqueue(dev_priv); - mutex_lock(&dev_priv->drm.struct_mutex); intel_uc_fini_hw(&dev_priv->gt.uc); intel_uc_fini(&dev_priv->gt.uc); - mutex_unlock(&dev_priv->drm.struct_mutex); i915_gem_drain_freed_objects(dev_priv); } void i915_gem_driver_release(struct drm_i915_private *dev_priv) { - mutex_lock(&dev_priv->drm.struct_mutex); - intel_engines_cleanup(dev_priv); - i915_gem_contexts_fini(dev_priv); - i915_gem_fini_scratch(dev_priv); - mutex_unlock(&dev_priv->drm.struct_mutex); + intel_engines_cleanup(&dev_priv->gt); + i915_gem_driver_release__contexts(dev_priv); + intel_gt_driver_release(&dev_priv->gt); intel_wa_list_free(&dev_priv->gt_wa_list); - intel_cleanup_gt_powersave(dev_priv); - intel_uc_cleanup_firmwares(&dev_priv->gt.uc); i915_gem_cleanup_userptr(dev_priv); intel_timelines_fini(dev_priv); i915_gem_drain_freed_objects(dev_priv); - WARN_ON(!list_empty(&dev_priv->contexts.list)); -} - -void i915_gem_init_mmio(struct drm_i915_private *i915) -{ - i915_gem_sanitize(i915); + WARN_ON(!list_empty(&dev_priv->gem.contexts.list)); } static void i915_gem_init__mm(struct drm_i915_private *i915) @@ -1643,20 +1407,11 @@ static void i915_gem_init__mm(struct drm_i915_private *i915) i915_gem_init__objects(i915); } -int i915_gem_init_early(struct drm_i915_private *dev_priv) +void i915_gem_init_early(struct drm_i915_private *dev_priv) { - int err; - i915_gem_init__mm(dev_priv); - i915_gem_init__pm(dev_priv); spin_lock_init(&dev_priv->fb_tracking.lock); - - err = i915_gemfs_init(dev_priv); - if (err) - DRM_NOTE("Unable to create a private tmpfs mount, hugepage support will be disabled(%d).\n", err); - - return 0; } void i915_gem_cleanup_early(struct drm_i915_private *dev_priv) @@ -1665,8 +1420,6 @@ void i915_gem_cleanup_early(struct drm_i915_private *dev_priv) GEM_BUG_ON(!llist_empty(&dev_priv->mm.free_list)); GEM_BUG_ON(atomic_read(&dev_priv->mm.free_count)); WARN_ON(dev_priv->mm.shrink_count); - - i915_gemfs_fini(dev_priv); } int i915_gem_freeze(struct drm_i915_private *dev_priv) diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h index 167a7b56ed5b..f6f9675848b8 100644 --- a/drivers/gpu/drm/i915/i915_gem.h +++ b/drivers/gpu/drm/i915/i915_gem.h @@ -37,10 +37,8 @@ struct drm_i915_private; #define GEM_SHOW_DEBUG() (drm_debug & DRM_UT_DRIVER) #define GEM_BUG_ON(condition) do { if (unlikely((condition))) { \ - pr_err("%s:%d GEM_BUG_ON(%s)\n", \ - __func__, __LINE__, __stringify(condition)); \ - GEM_TRACE("%s:%d GEM_BUG_ON(%s)\n", \ - __func__, __LINE__, __stringify(condition)); \ + GEM_TRACE_ERR("%s:%d GEM_BUG_ON(%s)\n", \ + __func__, __LINE__, __stringify(condition)); \ BUG(); \ } \ } while(0) @@ -66,17 +64,33 @@ struct drm_i915_private; #if IS_ENABLED(CONFIG_DRM_I915_TRACE_GEM) #define GEM_TRACE(...) trace_printk(__VA_ARGS__) +#define GEM_TRACE_ERR(...) do { \ + pr_err(__VA_ARGS__); \ + trace_printk(__VA_ARGS__); \ +} while (0) #define GEM_TRACE_DUMP() ftrace_dump(DUMP_ALL) #define GEM_TRACE_DUMP_ON(expr) \ do { if (expr) ftrace_dump(DUMP_ALL); } while (0) #else #define GEM_TRACE(...) do { } while (0) +#define GEM_TRACE_ERR(...) do { } while (0) #define GEM_TRACE_DUMP() do { } while (0) #define GEM_TRACE_DUMP_ON(expr) BUILD_BUG_ON_INVALID(expr) #endif #define I915_GEM_IDLE_TIMEOUT (HZ / 5) +static inline void tasklet_lock(struct tasklet_struct *t) +{ + while (!tasklet_trylock(t)) + cpu_relax(); +} + +static inline bool tasklet_is_locked(const struct tasklet_struct *t) +{ + return test_bit(TASKLET_STATE_RUN, &t->state); +} + static inline void __tasklet_disable_sync_once(struct tasklet_struct *t) { if (!atomic_fetch_inc(&t->count)) diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c index 52c86c6e0673..7e62c310290f 100644 --- a/drivers/gpu/drm/i915/i915_gem_evict.c +++ b/drivers/gpu/drm/i915/i915_gem_evict.c @@ -29,6 +29,7 @@ #include <drm/i915_drm.h> #include "gem/i915_gem_context.h" +#include "gt/intel_gt_requests.h" #include "i915_drv.h" #include "i915_trace.h" @@ -37,7 +38,7 @@ I915_SELFTEST_DECLARE(static struct igt_evict_ctl { bool fail_if_busy:1; } igt_evict_ctl;) -static int ggtt_flush(struct drm_i915_private *i915) +static int ggtt_flush(struct intel_gt *gt) { /* * Not everything in the GGTT is tracked via vma (otherwise we @@ -46,10 +47,7 @@ static int ggtt_flush(struct drm_i915_private *i915) * the hopes that we can then remove contexts and the like only * bound by their active reference. */ - return i915_gem_wait_for_idle(i915, - I915_WAIT_INTERRUPTIBLE | - I915_WAIT_LOCKED, - MAX_SCHEDULE_TIMEOUT); + return intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT); } static bool @@ -70,7 +68,7 @@ mark_free(struct drm_mm_scan *scan, * @vm: address space to evict from * @min_size: size of the desired free space * @alignment: alignment constraint of the desired free space - * @cache_level: cache_level for the desired space + * @color: color for the desired space * @start: start (inclusive) of the range from which to evict objects * @end: end (exclusive) of the range from which to evict objects * @flags: additional flags to control the eviction algorithm @@ -91,11 +89,10 @@ mark_free(struct drm_mm_scan *scan, int i915_gem_evict_something(struct i915_address_space *vm, u64 min_size, u64 alignment, - unsigned cache_level, + unsigned long color, u64 start, u64 end, unsigned flags) { - struct drm_i915_private *dev_priv = vm->i915; struct drm_mm_scan scan; struct list_head eviction_list; struct i915_vma *vma, *next; @@ -104,7 +101,7 @@ i915_gem_evict_something(struct i915_address_space *vm, struct i915_vma *active; int ret; - lockdep_assert_held(&vm->i915->drm.struct_mutex); + lockdep_assert_held(&vm->mutex); trace_i915_gem_evict(vm, min_size, alignment, flags); /* @@ -124,17 +121,10 @@ i915_gem_evict_something(struct i915_address_space *vm, if (flags & PIN_MAPPABLE) mode = DRM_MM_INSERT_LOW; drm_mm_scan_init_with_range(&scan, &vm->mm, - min_size, alignment, cache_level, + min_size, alignment, color, start, end, mode); - /* - * Retire before we search the active list. Although we have - * reasonable accuracy in our retirement lists, we may have - * a stray pin (preventing eviction) that can only be resolved by - * retiring. - */ - if (!(flags & PIN_NONBLOCK)) - i915_retire_requests(dev_priv); + intel_gt_retire_requests(vm->gt); search_again: active = NULL; @@ -207,7 +197,7 @@ search_again: if (I915_SELFTEST_ONLY(igt_evict_ctl.fail_if_busy)) return -EBUSY; - ret = ggtt_flush(dev_priv); + ret = ggtt_flush(vm->gt); if (ret) return ret; @@ -235,12 +225,12 @@ found: list_for_each_entry_safe(vma, next, &eviction_list, evict_link) { __i915_vma_unpin(vma); if (ret == 0) - ret = i915_vma_unbind(vma); + ret = __i915_vma_unbind(vma); } while (ret == 0 && (node = drm_mm_scan_color_evict(&scan))) { vma = container_of(node, struct i915_vma, node); - ret = i915_vma_unbind(vma); + ret = __i915_vma_unbind(vma); } return ret; @@ -266,25 +256,23 @@ int i915_gem_evict_for_node(struct i915_address_space *vm, u64 start = target->start; u64 end = start + target->size; struct i915_vma *vma, *next; - bool check_color; int ret = 0; - lockdep_assert_held(&vm->i915->drm.struct_mutex); + lockdep_assert_held(&vm->mutex); GEM_BUG_ON(!IS_ALIGNED(start, I915_GTT_PAGE_SIZE)); GEM_BUG_ON(!IS_ALIGNED(end, I915_GTT_PAGE_SIZE)); trace_i915_gem_evict_node(vm, target, flags); - /* Retire before we search the active list. Although we have + /* + * Retire before we search the active list. Although we have * reasonable accuracy in our retirement lists, we may have * a stray pin (preventing eviction) that can only be resolved by * retiring. */ - if (!(flags & PIN_NONBLOCK)) - i915_retire_requests(vm->i915); + intel_gt_retire_requests(vm->gt); - check_color = vm->mm.color_adjust; - if (check_color) { + if (i915_vm_has_cache_coloring(vm)) { /* Expand search to cover neighbouring guard pages (or lack!) */ if (start) start -= I915_GTT_PAGE_SIZE; @@ -301,7 +289,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm, break; } - GEM_BUG_ON(!node->allocated); + GEM_BUG_ON(!drm_mm_node_allocated(node)); vma = container_of(node, typeof(*vma), node); /* If we are using coloring to insert guard pages between @@ -310,7 +298,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm, * abutt and conflict. If they are in conflict, then we evict * those as well to make room for our guard pages. */ - if (check_color) { + if (i915_vm_has_cache_coloring(vm)) { if (node->start + node->size == target->start) { if (node->color == target->color) continue; @@ -351,7 +339,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm, list_for_each_entry_safe(vma, next, &eviction_list, evict_link) { __i915_vma_unpin(vma); if (ret == 0) - ret = i915_vma_unbind(vma); + ret = __i915_vma_unbind(vma); } return ret; @@ -375,7 +363,7 @@ int i915_gem_evict_vm(struct i915_address_space *vm) struct i915_vma *vma, *next; int ret; - lockdep_assert_held(&vm->i915->drm.struct_mutex); + lockdep_assert_held(&vm->mutex); trace_i915_gem_evict_vm(vm); /* Switch back to the default context in order to unpin @@ -384,13 +372,12 @@ int i915_gem_evict_vm(struct i915_address_space *vm) * switch otherwise is ineffective. */ if (i915_is_ggtt(vm)) { - ret = ggtt_flush(vm->i915); + ret = ggtt_flush(vm->gt); if (ret) return ret; } INIT_LIST_HEAD(&eviction_list); - mutex_lock(&vm->mutex); list_for_each_entry(vma, &vm->bound_list, vm_link) { if (i915_vma_is_pinned(vma)) continue; @@ -398,13 +385,12 @@ int i915_gem_evict_vm(struct i915_address_space *vm) __i915_vma_pin(vma); list_add(&vma->evict_link, &eviction_list); } - mutex_unlock(&vm->mutex); ret = 0; list_for_each_entry_safe(vma, next, &eviction_list, evict_link) { __i915_vma_unpin(vma); if (ret == 0) - ret = i915_vma_unbind(vma); + ret = __i915_vma_unbind(vma); } return ret; } diff --git a/drivers/gpu/drm/i915/i915_gem_fence_reg.c b/drivers/gpu/drm/i915/i915_gem_fence_reg.c index 615a9f4ef30c..71efccfde122 100644 --- a/drivers/gpu/drm/i915/i915_gem_fence_reg.c +++ b/drivers/gpu/drm/i915/i915_gem_fence_reg.c @@ -59,6 +59,16 @@ #define pipelined 0 +static struct drm_i915_private *fence_to_i915(struct i915_fence_reg *fence) +{ + return fence->ggtt->vm.i915; +} + +static struct intel_uncore *fence_to_uncore(struct i915_fence_reg *fence) +{ + return fence->ggtt->vm.gt->uncore; +} + static void i965_write_fence_reg(struct i915_fence_reg *fence, struct i915_vma *vma) { @@ -66,7 +76,7 @@ static void i965_write_fence_reg(struct i915_fence_reg *fence, int fence_pitch_shift; u64 val; - if (INTEL_GEN(fence->i915) >= 6) { + if (INTEL_GEN(fence_to_i915(fence)) >= 6) { fence_reg_lo = FENCE_REG_GEN6_LO(fence->id); fence_reg_hi = FENCE_REG_GEN6_HI(fence->id); fence_pitch_shift = GEN6_FENCE_PITCH_SHIFT; @@ -95,7 +105,7 @@ static void i965_write_fence_reg(struct i915_fence_reg *fence, } if (!pipelined) { - struct intel_uncore *uncore = &fence->i915->uncore; + struct intel_uncore *uncore = fence_to_uncore(fence); /* * To w/a incoherency with non-atomic 64-bit register updates, @@ -132,7 +142,7 @@ static void i915_write_fence_reg(struct i915_fence_reg *fence, GEM_BUG_ON(!is_power_of_2(vma->fence_size)); GEM_BUG_ON(!IS_ALIGNED(vma->node.start, vma->fence_size)); - if (is_y_tiled && HAS_128_BYTE_Y_TILING(fence->i915)) + if (is_y_tiled && HAS_128_BYTE_Y_TILING(fence_to_i915(fence))) stride /= 128; else stride /= 512; @@ -148,7 +158,7 @@ static void i915_write_fence_reg(struct i915_fence_reg *fence, } if (!pipelined) { - struct intel_uncore *uncore = &fence->i915->uncore; + struct intel_uncore *uncore = fence_to_uncore(fence); i915_reg_t reg = FENCE_REG(fence->id); intel_uncore_write_fw(uncore, reg, val); @@ -180,7 +190,7 @@ static void i830_write_fence_reg(struct i915_fence_reg *fence, } if (!pipelined) { - struct intel_uncore *uncore = &fence->i915->uncore; + struct intel_uncore *uncore = fence_to_uncore(fence); i915_reg_t reg = FENCE_REG(fence->id); intel_uncore_write_fw(uncore, reg, val); @@ -191,15 +201,17 @@ static void i830_write_fence_reg(struct i915_fence_reg *fence, static void fence_write(struct i915_fence_reg *fence, struct i915_vma *vma) { + struct drm_i915_private *i915 = fence_to_i915(fence); + /* * Previous access through the fence register is marshalled by * the mb() inside the fault handlers (i915_gem_release_mmaps) * and explicitly managed for internal users. */ - if (IS_GEN(fence->i915, 2)) + if (IS_GEN(i915, 2)) i830_write_fence_reg(fence, vma); - else if (IS_GEN(fence->i915, 3)) + else if (IS_GEN(i915, 3)) i915_write_fence_reg(fence, vma); else i965_write_fence_reg(fence, vma); @@ -215,6 +227,8 @@ static void fence_write(struct i915_fence_reg *fence, static int fence_update(struct i915_fence_reg *fence, struct i915_vma *vma) { + struct i915_ggtt *ggtt = fence->ggtt; + struct intel_uncore *uncore = fence_to_uncore(fence); intel_wakeref_t wakeref; struct i915_vma *old; int ret; @@ -230,14 +244,15 @@ static int fence_update(struct i915_fence_reg *fence, i915_gem_object_get_tiling(vma->obj))) return -EINVAL; - ret = i915_active_wait(&vma->active); + ret = i915_vma_sync(vma); if (ret) return ret; } old = xchg(&fence->vma, NULL); if (old) { - ret = i915_active_wait(&old->active); + /* XXX Ideally we would move the waiting to outside the mutex */ + ret = i915_vma_sync(old); if (ret) { fence->vma = old; return ret; @@ -255,7 +270,7 @@ static int fence_update(struct i915_fence_reg *fence, old->fence = NULL; } - list_move(&fence->link, &fence->i915->ggtt.fence_list); + list_move(&fence->link, &ggtt->fence_list); } /* @@ -268,7 +283,7 @@ static int fence_update(struct i915_fence_reg *fence, * be cleared before we can use any other fences to ensure that * the new fences do not overlap the elided clears, confusing HW. */ - wakeref = intel_runtime_pm_get_if_in_use(&fence->i915->runtime_pm); + wakeref = intel_runtime_pm_get_if_in_use(uncore->rpm); if (!wakeref) { GEM_BUG_ON(vma); return 0; @@ -279,10 +294,10 @@ static int fence_update(struct i915_fence_reg *fence, if (vma) { vma->fence = fence; - list_move_tail(&fence->link, &fence->i915->ggtt.fence_list); + list_move_tail(&fence->link, &ggtt->fence_list); } - intel_runtime_pm_put(&fence->i915->runtime_pm, wakeref); + intel_runtime_pm_put(uncore->rpm, wakeref); return 0; } @@ -311,11 +326,11 @@ int i915_vma_revoke_fence(struct i915_vma *vma) return fence_update(fence, NULL); } -static struct i915_fence_reg *fence_find(struct drm_i915_private *i915) +static struct i915_fence_reg *fence_find(struct i915_ggtt *ggtt) { struct i915_fence_reg *fence; - list_for_each_entry(fence, &i915->ggtt.fence_list, link) { + list_for_each_entry(fence, &ggtt->fence_list, link) { GEM_BUG_ON(fence->vma && fence->vma->fence != fence); if (atomic_read(&fence->pin_count)) @@ -325,19 +340,21 @@ static struct i915_fence_reg *fence_find(struct drm_i915_private *i915) } /* Wait for completion of pending flips which consume fences */ - if (intel_has_pending_fb_unpin(i915)) + if (intel_has_pending_fb_unpin(ggtt->vm.i915)) return ERR_PTR(-EAGAIN); return ERR_PTR(-EDEADLK); } -static int __i915_vma_pin_fence(struct i915_vma *vma) +int __i915_vma_pin_fence(struct i915_vma *vma) { struct i915_ggtt *ggtt = i915_vm_to_ggtt(vma->vm); struct i915_fence_reg *fence; struct i915_vma *set = i915_gem_object_is_tiled(vma->obj) ? vma : NULL; int err; + lockdep_assert_held(&vma->vm->mutex); + /* Just update our place in the LRU if our fence is getting reused. */ if (vma->fence) { fence = vma->fence; @@ -348,7 +365,7 @@ static int __i915_vma_pin_fence(struct i915_vma *vma) return 0; } } else if (set) { - fence = fence_find(vma->vm->i915); + fence = fence_find(ggtt); if (IS_ERR(fence)) return PTR_ERR(fence); @@ -399,7 +416,7 @@ int i915_vma_pin_fence(struct i915_vma *vma) * Note that we revoke fences on runtime suspend. Therefore the user * must keep the device awake whilst using the fence. */ - assert_rpm_wakelock_held(&vma->vm->i915->runtime_pm); + assert_rpm_wakelock_held(vma->vm->gt->uncore->rpm); GEM_BUG_ON(!i915_vma_is_pinned(vma)); GEM_BUG_ON(!i915_vma_is_ggtt(vma)); @@ -415,14 +432,13 @@ int i915_vma_pin_fence(struct i915_vma *vma) /** * i915_reserve_fence - Reserve a fence for vGPU - * @i915: i915 device private + * @ggtt: Global GTT * * This function walks the fence regs looking for a free one and remove * it from the fence_list. It is used to reserve fence for vGPU to use. */ -struct i915_fence_reg *i915_reserve_fence(struct drm_i915_private *i915) +struct i915_fence_reg *i915_reserve_fence(struct i915_ggtt *ggtt) { - struct i915_ggtt *ggtt = &i915->ggtt; struct i915_fence_reg *fence; int count; int ret; @@ -436,7 +452,7 @@ struct i915_fence_reg *i915_reserve_fence(struct drm_i915_private *i915) if (count <= 1) return ERR_PTR(-ENOSPC); - fence = fence_find(i915); + fence = fence_find(ggtt); if (IS_ERR(fence)) return fence; @@ -460,7 +476,7 @@ struct i915_fence_reg *i915_reserve_fence(struct drm_i915_private *i915) */ void i915_unreserve_fence(struct i915_fence_reg *fence) { - struct i915_ggtt *ggtt = &fence->i915->ggtt; + struct i915_ggtt *ggtt = fence->ggtt; lockdep_assert_held(&ggtt->vm.mutex); @@ -469,19 +485,19 @@ void i915_unreserve_fence(struct i915_fence_reg *fence) /** * i915_gem_restore_fences - restore fence state - * @i915: i915 device private + * @ggtt: Global GTT * * Restore the hw fence state to match the software tracking again, to be called * after a gpu reset and on resume. Note that on runtime suspend we only cancel * the fences, to be reacquired by the user later. */ -void i915_gem_restore_fences(struct drm_i915_private *i915) +void i915_gem_restore_fences(struct i915_ggtt *ggtt) { int i; rcu_read_lock(); /* keep obj alive as we dereference */ - for (i = 0; i < i915->ggtt.num_fences; i++) { - struct i915_fence_reg *reg = &i915->ggtt.fence_regs[i]; + for (i = 0; i < ggtt->num_fences; i++) { + struct i915_fence_reg *reg = &ggtt->fence_regs[i]; struct i915_vma *vma = READ_ONCE(reg->vma); GEM_BUG_ON(vma && vma->fence != reg); @@ -547,15 +563,16 @@ void i915_gem_restore_fences(struct drm_i915_private *i915) */ /** - * i915_gem_detect_bit_6_swizzle - detect bit 6 swizzling pattern - * @i915: i915 device private + * detect_bit_6_swizzle - detect bit 6 swizzling pattern + * @ggtt: Global GGTT * * Detects bit 6 swizzling of address lookup between IGD access and CPU * access through main memory. */ -static void detect_bit_6_swizzle(struct drm_i915_private *i915) +static void detect_bit_6_swizzle(struct i915_ggtt *ggtt) { - struct intel_uncore *uncore = &i915->uncore; + struct intel_uncore *uncore = ggtt->vm.gt->uncore; + struct drm_i915_private *i915 = ggtt->vm.i915; u32 swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN; u32 swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN; @@ -717,8 +734,8 @@ static void detect_bit_6_swizzle(struct drm_i915_private *i915) swizzle_y = I915_BIT_6_SWIZZLE_NONE; } - i915->mm.bit_6_swizzle_x = swizzle_x; - i915->mm.bit_6_swizzle_y = swizzle_y; + i915->ggtt.bit_6_swizzle_x = swizzle_x; + i915->ggtt.bit_6_swizzle_y = swizzle_y; } /* @@ -819,17 +836,20 @@ i915_gem_object_save_bit_17_swizzle(struct drm_i915_gem_object *obj, void i915_ggtt_init_fences(struct i915_ggtt *ggtt) { struct drm_i915_private *i915 = ggtt->vm.i915; + struct intel_uncore *uncore = ggtt->vm.gt->uncore; int num_fences; int i; INIT_LIST_HEAD(&ggtt->fence_list); INIT_LIST_HEAD(&ggtt->userfault_list); - intel_wakeref_auto_init(&ggtt->userfault_wakeref, &i915->runtime_pm); + intel_wakeref_auto_init(&ggtt->userfault_wakeref, uncore->rpm); - detect_bit_6_swizzle(i915); + detect_bit_6_swizzle(ggtt); - if (INTEL_GEN(i915) >= 7 && - !(IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915))) + if (!i915_ggtt_has_aperture(ggtt)) + num_fences = 0; + else if (INTEL_GEN(i915) >= 7 && + !(IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915))) num_fences = 32; else if (INTEL_GEN(i915) >= 4 || IS_I945G(i915) || IS_I945GM(i915) || @@ -839,20 +859,20 @@ void i915_ggtt_init_fences(struct i915_ggtt *ggtt) num_fences = 8; if (intel_vgpu_active(i915)) - num_fences = intel_uncore_read(&i915->uncore, + num_fences = intel_uncore_read(uncore, vgtif_reg(avail_rs.fence_num)); /* Initialize fence registers to zero */ for (i = 0; i < num_fences; i++) { struct i915_fence_reg *fence = &ggtt->fence_regs[i]; - fence->i915 = i915; + fence->ggtt = ggtt; fence->id = i; list_add_tail(&fence->link, &ggtt->fence_list); } ggtt->num_fences = num_fences; - i915_gem_restore_fences(i915); + i915_gem_restore_fences(ggtt); } void intel_gt_init_swizzling(struct intel_gt *gt) @@ -861,7 +881,7 @@ void intel_gt_init_swizzling(struct intel_gt *gt) struct intel_uncore *uncore = gt->uncore; if (INTEL_GEN(i915) < 5 || - i915->mm.bit_6_swizzle_x == I915_BIT_6_SWIZZLE_NONE) + i915->ggtt.bit_6_swizzle_x == I915_BIT_6_SWIZZLE_NONE) return; intel_uncore_rmw(uncore, DISP_ARB_CTL, 0, DISP_TILE_SURFACE_SWIZZLING); diff --git a/drivers/gpu/drm/i915/i915_gem_fence_reg.h b/drivers/gpu/drm/i915/i915_gem_fence_reg.h index 99866fb9d94f..7bd521cd7cd7 100644 --- a/drivers/gpu/drm/i915/i915_gem_fence_reg.h +++ b/drivers/gpu/drm/i915/i915_gem_fence_reg.h @@ -29,7 +29,6 @@ #include <linux/types.h> struct drm_i915_gem_object; -struct drm_i915_private; struct i915_ggtt; struct i915_vma; struct intel_gt; @@ -39,7 +38,7 @@ struct sg_table; struct i915_fence_reg { struct list_head link; - struct drm_i915_private *i915; + struct i915_ggtt *ggtt; struct i915_vma *vma; atomic_t pin_count; int id; @@ -55,10 +54,10 @@ struct i915_fence_reg { }; /* i915_gem_fence_reg.c */ -struct i915_fence_reg *i915_reserve_fence(struct drm_i915_private *i915); +struct i915_fence_reg *i915_reserve_fence(struct i915_ggtt *ggtt); void i915_unreserve_fence(struct i915_fence_reg *fence); -void i915_gem_restore_fences(struct drm_i915_private *i915); +void i915_gem_restore_fences(struct i915_ggtt *ggtt); void i915_gem_object_do_bit_17_swizzle(struct drm_i915_gem_object *obj, struct sg_table *pages); diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index b1a7a8b9b46a..6239a9adbf14 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -38,6 +38,7 @@ #include "display/intel_frontbuffer.h" #include "gt/intel_gt.h" +#include "gt/intel_gt_requests.h" #include "i915_drv.h" #include "i915_scatterlist.h" @@ -132,9 +133,15 @@ static void gen6_ggtt_invalidate(struct i915_ggtt *ggtt) static void guc_ggtt_invalidate(struct i915_ggtt *ggtt) { struct intel_uncore *uncore = ggtt->vm.gt->uncore; + struct drm_i915_private *i915 = ggtt->vm.i915; gen6_ggtt_invalidate(ggtt); - intel_uncore_write_fw(uncore, GEN8_GTCR, GEN8_GTCR_INVALIDATE); + + if (INTEL_GEN(i915) >= 12) + intel_uncore_write_fw(uncore, GEN12_GUC_TLB_INV_CR, + GEN12_GUC_TLB_INV_CR_INVALIDATE); + else + intel_uncore_write_fw(uncore, GEN8_GTCR, GEN8_GTCR_INVALIDATE); } static void gmch_ggtt_invalidate(struct i915_ggtt *ggtt) @@ -144,16 +151,18 @@ static void gmch_ggtt_invalidate(struct i915_ggtt *ggtt) static int ppgtt_bind_vma(struct i915_vma *vma, enum i915_cache_level cache_level, - u32 unused) + u32 flags) { u32 pte_flags; int err; - if (!(vma->flags & I915_VMA_LOCAL_BIND)) { + if (flags & I915_VMA_ALLOC) { err = vma->vm->allocate_va_range(vma->vm, vma->node.start, vma->size); if (err) return err; + + set_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma)); } /* Applicable to VLV, and gen8+ */ @@ -161,14 +170,17 @@ static int ppgtt_bind_vma(struct i915_vma *vma, if (i915_gem_object_is_readonly(vma->obj)) pte_flags |= PTE_READ_ONLY; + GEM_BUG_ON(!test_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma))); vma->vm->insert_entries(vma->vm, vma, cache_level, pte_flags); + wmb(); return 0; } static void ppgtt_unbind_vma(struct i915_vma *vma) { - vma->vm->clear_range(vma->vm, vma->node.start, vma->size); + if (test_and_clear_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma))) + vma->vm->clear_range(vma->vm, vma->node.start, vma->size); } static int ppgtt_set_pages(struct i915_vma *vma) @@ -496,22 +508,26 @@ static void i915_address_space_fini(struct i915_address_space *vm) mutex_destroy(&vm->mutex); } -static void ppgtt_destroy_vma(struct i915_address_space *vm) +void __i915_vm_close(struct i915_address_space *vm) { - struct list_head *phases[] = { - &vm->bound_list, - &vm->unbound_list, - NULL, - }, **phase; + struct i915_vma *vma, *vn; + + mutex_lock(&vm->mutex); + list_for_each_entry_safe(vma, vn, &vm->bound_list, vm_link) { + struct drm_i915_gem_object *obj = vma->obj; - mutex_lock(&vm->i915->drm.struct_mutex); - for (phase = phases; *phase; phase++) { - struct i915_vma *vma, *vn; + /* Keep the obj (and hence the vma) alive as _we_ destroy it */ + if (!kref_get_unless_zero(&obj->base.refcount)) + continue; + + atomic_and(~I915_VMA_PIN_MASK, &vma->flags); + WARN_ON(__i915_vma_unbind(vma)); + i915_vma_destroy(vma); - list_for_each_entry_safe(vma, vn, *phase, vm_link) - i915_vma_destroy(vma); + i915_gem_object_put(obj); } - mutex_unlock(&vm->i915->drm.struct_mutex); + GEM_BUG_ON(!list_empty(&vm->bound_list)); + mutex_unlock(&vm->mutex); } static void __i915_vm_release(struct work_struct *work) @@ -519,11 +535,6 @@ static void __i915_vm_release(struct work_struct *work) struct i915_address_space *vm = container_of(work, struct i915_address_space, rcu.work); - ppgtt_destroy_vma(vm); - - GEM_BUG_ON(!list_empty(&vm->bound_list)); - GEM_BUG_ON(!list_empty(&vm->unbound_list)); - vm->cleanup(vm); i915_address_space_fini(vm); @@ -538,7 +549,6 @@ void i915_vm_release(struct kref *kref) GEM_BUG_ON(i915_is_ggtt(vm)); trace_i915_ppgtt_release(vm); - vm->closed = true; queue_rcu_work(vm->i915->wq, &vm->rcu); } @@ -546,6 +556,7 @@ static void i915_address_space_init(struct i915_address_space *vm, int subclass) { kref_init(&vm->ref); INIT_RCU_WORK(&vm->rcu, __i915_vm_release); + atomic_set(&vm->open, 1); /* * The vm->mutex must be reclaim safe (for use in the shrinker). @@ -562,7 +573,6 @@ static void i915_address_space_init(struct i915_address_space *vm, int subclass) stash_init(&vm->free_pages); - INIT_LIST_HEAD(&vm->unbound_list); INIT_LIST_HEAD(&vm->bound_list); } @@ -816,17 +826,6 @@ release_pd_entry(struct i915_page_directory * const pd, return free; } -/* - * PDE TLBs are a pain to invalidate on GEN8+. When we modify - * the page table structures, we mark them dirty so that - * context switching/execlist queuing code takes extra steps - * to ensure that tlbs are flushed. - */ -static void mark_tlbs_dirty(struct i915_ppgtt *ppgtt) -{ - ppgtt->pd_dirty_engines = ALL_ENGINES; -} - static void gen8_ppgtt_notify_vgt(struct i915_ppgtt *ppgtt, bool create) { struct drm_i915_private *dev_priv = ppgtt->vm.i915; @@ -1367,7 +1366,9 @@ static int gen8_init_scratch(struct i915_address_space *vm) if (vm->has_read_only && vm->i915->kernel_context && vm->i915->kernel_context->vm) { - struct i915_address_space *clone = vm->i915->kernel_context->vm; + struct i915_address_space *clone = + rcu_dereference_protected(vm->i915->kernel_context->vm, + true); /* static */ GEM_BUG_ON(!clone->has_read_only); @@ -1422,6 +1423,7 @@ static int gen8_preallocate_top_level_pdp(struct i915_ppgtt *ppgtt) set_pd_entry(pd, idx, pde); atomic_inc(px_used(pde)); /* keep pinned */ } + wmb(); return 0; } @@ -1489,8 +1491,10 @@ static struct i915_ppgtt *gen8_ppgtt_create(struct drm_i915_private *i915) * * Gen11 has HSDES#:1807136187 unresolved. Disable ro support * for now. + * + * Gen12 has inherited the same read-only fault issue from gen11. */ - ppgtt->vm.has_read_only = INTEL_GEN(i915) != 11; + ppgtt->vm.has_read_only = !IS_GEN_RANGE(i915, 11, 12); /* There are only few exceptions for gen >=6. chv and bxt. * And we are not sure about the latter so play safe for now. @@ -1509,13 +1513,12 @@ static struct i915_ppgtt *gen8_ppgtt_create(struct drm_i915_private *i915) } if (!i915_vm_is_4lvl(&ppgtt->vm)) { - if (intel_vgpu_active(i915)) { - err = gen8_preallocate_top_level_pdp(ppgtt); - if (err) - goto err_free_pd; - } + err = gen8_preallocate_top_level_pdp(ppgtt); + if (err) + goto err_free_pd; } + ppgtt->vm.bind_async_flags = I915_VMA_LOCAL_BIND; ppgtt->vm.insert_entries = gen8_ppgtt_insert; ppgtt->vm.allocate_va_range = gen8_ppgtt_alloc; ppgtt->vm.clear_range = gen8_ppgtt_clear; @@ -1566,7 +1569,7 @@ static void gen7_ppgtt_enable(struct intel_gt *gt) } intel_uncore_write(uncore, GAM_ECOCHK, ecochk); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { /* GFX_MODE is per-ring on gen7+ */ ENGINE_WRITE(engine, RING_MODE_GEN7, @@ -1729,10 +1732,8 @@ static int gen6_alloc_va_range(struct i915_address_space *vm, } spin_unlock(&pd->lock); - if (flush) { - mark_tlbs_dirty(&ppgtt->base); + if (flush) gen6_ggtt_invalidate(vm->gt->ggtt); - } goto out; @@ -1786,15 +1787,13 @@ static void gen6_ppgtt_free_pd(struct gen6_ppgtt *ppgtt) static void gen6_ppgtt_cleanup(struct i915_address_space *vm) { struct gen6_ppgtt *ppgtt = to_gen6_ppgtt(i915_vm_to_ppgtt(vm)); - struct drm_i915_private *i915 = vm->i915; - /* FIXME remove the struct_mutex to bring the locking under control */ - mutex_lock(&i915->drm.struct_mutex); i915_vma_destroy(ppgtt->vma); - mutex_unlock(&i915->drm.struct_mutex); gen6_ppgtt_free_pd(ppgtt); free_scratch(vm); + + mutex_destroy(&ppgtt->pin_mutex); kfree(ppgtt->base.pd); } @@ -1827,7 +1826,6 @@ static int pd_vma_bind(struct i915_vma *vma, gen6_for_all_pdes(pt, ppgtt->base.pd, pde) gen6_write_pde(ppgtt, pde, pt); - mark_tlbs_dirty(&ppgtt->base); gen6_ggtt_invalidate(ggtt); return 0; @@ -1866,7 +1864,6 @@ static const struct i915_vma_ops pd_vma_ops = { static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size) { - struct drm_i915_private *i915 = ppgtt->base.vm.i915; struct i915_ggtt *ggtt = ppgtt->base.vm.gt->ggtt; struct i915_vma *vma; @@ -1877,33 +1874,30 @@ static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size) if (!vma) return ERR_PTR(-ENOMEM); - i915_active_init(i915, &vma->active, NULL, NULL); + i915_active_init(&vma->active, NULL, NULL); - vma->vm = &ggtt->vm; + mutex_init(&vma->pages_mutex); + vma->vm = i915_vm_get(&ggtt->vm); vma->ops = &pd_vma_ops; vma->private = ppgtt; vma->size = size; vma->fence_size = size; - vma->flags = I915_VMA_GGTT; + atomic_set(&vma->flags, I915_VMA_GGTT); vma->ggtt_view.type = I915_GGTT_VIEW_ROTATED; /* prevent fencing */ INIT_LIST_HEAD(&vma->obj_link); INIT_LIST_HEAD(&vma->closed_link); - mutex_lock(&vma->vm->mutex); - list_add(&vma->vm_link, &vma->vm->unbound_list); - mutex_unlock(&vma->vm->mutex); - return vma; } int gen6_ppgtt_pin(struct i915_ppgtt *base) { struct gen6_ppgtt *ppgtt = to_gen6_ppgtt(base); - int err; + int err = 0; - GEM_BUG_ON(ppgtt->base.vm.closed); + GEM_BUG_ON(!atomic_read(&ppgtt->base.vm.open)); /* * Workaround the limited maximum vma->pin_count and the aliasing_ppgtt @@ -1911,24 +1905,26 @@ int gen6_ppgtt_pin(struct i915_ppgtt *base) * (When vma->pin_count becomes atomic, I expect we will naturally * need a larger, unpacked, type and kill this redundancy.) */ - if (ppgtt->pin_count++) + if (atomic_add_unless(&ppgtt->pin_count, 1, 0)) return 0; + if (mutex_lock_interruptible(&ppgtt->pin_mutex)) + return -EINTR; + /* * PPGTT PDEs reside in the GGTT and consists of 512 entries. The * allocator works in address space sizes, so it's multiplied by page * size. We allocate at the top of the GTT to avoid fragmentation. */ - err = i915_vma_pin(ppgtt->vma, - 0, GEN6_PD_ALIGN, - PIN_GLOBAL | PIN_HIGH); - if (err) - goto unpin; - - return 0; + if (!atomic_read(&ppgtt->pin_count)) { + err = i915_vma_pin(ppgtt->vma, + 0, GEN6_PD_ALIGN, + PIN_GLOBAL | PIN_HIGH); + } + if (!err) + atomic_inc(&ppgtt->pin_count); + mutex_unlock(&ppgtt->pin_mutex); -unpin: - ppgtt->pin_count = 0; return err; } @@ -1936,22 +1932,20 @@ void gen6_ppgtt_unpin(struct i915_ppgtt *base) { struct gen6_ppgtt *ppgtt = to_gen6_ppgtt(base); - GEM_BUG_ON(!ppgtt->pin_count); - if (--ppgtt->pin_count) - return; - - i915_vma_unpin(ppgtt->vma); + GEM_BUG_ON(!atomic_read(&ppgtt->pin_count)); + if (atomic_dec_and_test(&ppgtt->pin_count)) + i915_vma_unpin(ppgtt->vma); } void gen6_ppgtt_unpin_all(struct i915_ppgtt *base) { struct gen6_ppgtt *ppgtt = to_gen6_ppgtt(base); - if (!ppgtt->pin_count) + if (!atomic_read(&ppgtt->pin_count)) return; - ppgtt->pin_count = 0; i915_vma_unpin(ppgtt->vma); + atomic_set(&ppgtt->pin_count, 0); } static struct i915_ppgtt *gen6_ppgtt_create(struct drm_i915_private *i915) @@ -1964,9 +1958,12 @@ static struct i915_ppgtt *gen6_ppgtt_create(struct drm_i915_private *i915) if (!ppgtt) return ERR_PTR(-ENOMEM); + mutex_init(&ppgtt->pin_mutex); + ppgtt_init(&ppgtt->base, &i915->gt); ppgtt->base.vm.top = 1; + ppgtt->base.vm.bind_async_flags = I915_VMA_LOCAL_BIND; ppgtt->base.vm.allocate_va_range = gen6_alloc_va_range; ppgtt->base.vm.clear_range = gen6_ppgtt_clear_range; ppgtt->base.vm.insert_entries = gen6_ppgtt_insert_entries; @@ -2023,7 +2020,7 @@ static void gtt_write_workarounds(struct intel_gt *gt) intel_uncore_write(uncore, GEN8_L3_LRA_1_GPGPU, GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_BXT); - else if (INTEL_GEN(i915) >= 9) + else if (INTEL_GEN(i915) >= 9 && INTEL_GEN(i915) <= 11) intel_uncore_write(uncore, GEN8_L3_LRA_1_GPGPU, GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_SKL); @@ -2202,7 +2199,7 @@ static void gen8_ggtt_insert_entries(struct i915_address_space *vm, gtt_entries = (gen8_pte_t __iomem *)ggtt->gsm; gtt_entries += vma->node.start / I915_GTT_PAGE_SIZE; - for_each_sgt_dma(addr, sgt_iter, vma->pages) + for_each_sgt_daddr(addr, sgt_iter, vma->pages) gen8_set_pte(gtt_entries++, pte_encode | addr); /* @@ -2243,7 +2240,7 @@ static void gen6_ggtt_insert_entries(struct i915_address_space *vm, unsigned int i = vma->node.start / I915_GTT_PAGE_SIZE; struct sgt_iter iter; dma_addr_t addr; - for_each_sgt_dma(addr, iter, vma->pages) + for_each_sgt_daddr(addr, iter, vma->pages) iowrite32(vm->pte_encode(addr, level, flags), &entries[i++]); /* @@ -2448,7 +2445,7 @@ static int ggtt_bind_vma(struct i915_vma *vma, * GLOBAL/LOCAL_BIND, it's all the same ptes. Hence unconditionally * upgrade to both bound if we bind either to avoid double-binding. */ - vma->flags |= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND; + atomic_or(I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND, &vma->flags); return 0; } @@ -2478,14 +2475,18 @@ static int aliasing_gtt_bind_vma(struct i915_vma *vma, if (flags & I915_VMA_LOCAL_BIND) { struct i915_ppgtt *alias = i915_vm_to_ggtt(vma->vm)->alias; - if (!(vma->flags & I915_VMA_LOCAL_BIND)) { + if (flags & I915_VMA_ALLOC) { ret = alias->vm.allocate_va_range(&alias->vm, vma->node.start, vma->size); if (ret) return ret; + + set_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma)); } + GEM_BUG_ON(!test_bit(I915_VMA_ALLOC_BIT, + __i915_vma_flags(vma))); alias->vm.insert_entries(&alias->vm, vma, cache_level, pte_flags); } @@ -2506,7 +2507,7 @@ static void aliasing_gtt_unbind_vma(struct i915_vma *vma) { struct drm_i915_private *i915 = vma->vm->i915; - if (vma->flags & I915_VMA_GLOBAL_BIND) { + if (i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND)) { struct i915_address_space *vm = vma->vm; intel_wakeref_t wakeref; @@ -2514,7 +2515,7 @@ static void aliasing_gtt_unbind_vma(struct i915_vma *vma) vm->clear_range(vm, vma->node.start, vma->size); } - if (vma->flags & I915_VMA_LOCAL_BIND) { + if (test_and_clear_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma))) { struct i915_address_space *vm = &i915_vm_to_ggtt(vma->vm)->alias->vm; @@ -2530,7 +2531,9 @@ void i915_gem_gtt_finish_pages(struct drm_i915_gem_object *obj, struct i915_ggtt *ggtt = &dev_priv->ggtt; if (unlikely(ggtt->do_idle_maps)) { - if (i915_gem_wait_for_idle(dev_priv, 0, MAX_SCHEDULE_TIMEOUT)) { + /* XXX This does not prevent more requests being submitted! */ + if (intel_gt_retire_requests_timeout(ggtt->vm.gt, + -MAX_SCHEDULE_TIMEOUT)) { DRM_ERROR("Failed to wait for idle; VT'd may hang.\n"); /* Wait a bit, in hopes it avoids the hang */ udelay(10); @@ -2555,12 +2558,12 @@ static int ggtt_set_pages(struct i915_vma *vma) return 0; } -static void i915_gtt_color_adjust(const struct drm_mm_node *node, - unsigned long color, - u64 *start, - u64 *end) +static void i915_ggtt_color_adjust(const struct drm_mm_node *node, + unsigned long color, + u64 *start, + u64 *end) { - if (node->allocated && node->color != color) + if (i915_node_color_differs(node, color)) *start += I915_GTT_PAGE_SIZE; /* Also leave a space between the unallocated reserved node after the @@ -2598,6 +2601,7 @@ static int init_aliasing_ppgtt(struct i915_ggtt *ggtt) goto err_ppgtt; ggtt->alias = ppgtt; + ggtt->vm.bind_async_flags |= ppgtt->vm.bind_async_flags; GEM_BUG_ON(ggtt->vm.vma_ops.bind_vma != ggtt_bind_vma); ggtt->vm.vma_ops.bind_vma = aliasing_gtt_bind_vma; @@ -2614,22 +2618,16 @@ err_ppgtt: static void fini_aliasing_ppgtt(struct i915_ggtt *ggtt) { - struct drm_i915_private *i915 = ggtt->vm.i915; struct i915_ppgtt *ppgtt; - mutex_lock(&i915->drm.struct_mutex); - ppgtt = fetch_and_zero(&ggtt->alias); if (!ppgtt) - goto out; + return; i915_vm_put(&ppgtt->vm); ggtt->vm.vma_ops.bind_vma = ggtt_bind_vma; ggtt->vm.vma_ops.unbind_vma = ggtt_unbind_vma; - -out: - mutex_unlock(&i915->drm.struct_mutex); } static int ggtt_reserve_guc_top(struct i915_ggtt *ggtt) @@ -2661,7 +2659,8 @@ static void ggtt_release_guc_top(struct i915_ggtt *ggtt) static void cleanup_init_ggtt(struct i915_ggtt *ggtt) { ggtt_release_guc_top(ggtt); - drm_mm_remove_node(&ggtt->error_capture); + if (drm_mm_node_allocated(&ggtt->error_capture)) + drm_mm_remove_node(&ggtt->error_capture); } static int init_ggtt(struct i915_ggtt *ggtt) @@ -2692,13 +2691,15 @@ static int init_ggtt(struct i915_ggtt *ggtt) if (ret) return ret; - /* Reserve a mappable slot for our lockless error capture */ - ret = drm_mm_insert_node_in_range(&ggtt->vm.mm, &ggtt->error_capture, - PAGE_SIZE, 0, I915_COLOR_UNEVICTABLE, - 0, ggtt->mappable_end, - DRM_MM_INSERT_LOW); - if (ret) - return ret; + if (ggtt->mappable_end) { + /* Reserve a mappable slot for our lockless error capture */ + ret = drm_mm_insert_node_in_range(&ggtt->vm.mm, &ggtt->error_capture, + PAGE_SIZE, 0, I915_COLOR_UNEVICTABLE, + 0, ggtt->mappable_end, + DRM_MM_INSERT_LOW); + if (ret) + return ret; + } /* * The upper portion of the GuC address space has a sizeable hole @@ -2746,35 +2747,33 @@ int i915_init_ggtt(struct drm_i915_private *i915) static void ggtt_cleanup_hw(struct i915_ggtt *ggtt) { - struct drm_i915_private *i915 = ggtt->vm.i915; struct i915_vma *vma, *vn; - ggtt->vm.closed = true; + atomic_set(&ggtt->vm.open, 0); rcu_barrier(); /* flush the RCU'ed__i915_vm_release */ - flush_workqueue(i915->wq); + flush_workqueue(ggtt->vm.i915->wq); - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(&ggtt->vm.mutex); list_for_each_entry_safe(vma, vn, &ggtt->vm.bound_list, vm_link) - WARN_ON(i915_vma_unbind(vma)); + WARN_ON(__i915_vma_unbind(vma)); if (drm_mm_node_allocated(&ggtt->error_capture)) drm_mm_remove_node(&ggtt->error_capture); ggtt_release_guc_top(ggtt); - - if (drm_mm_initialized(&ggtt->vm.mm)) { - intel_vgt_deballoon(ggtt); - i915_address_space_fini(&ggtt->vm); - } + intel_vgt_deballoon(ggtt); ggtt->vm.cleanup(&ggtt->vm); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(&ggtt->vm.mutex); + i915_address_space_fini(&ggtt->vm); arch_phys_wc_del(ggtt->mtrr); - io_mapping_fini(&ggtt->iomap); + + if (ggtt->iomap.size) + io_mapping_fini(&ggtt->iomap); } /** @@ -2794,8 +2793,6 @@ void i915_ggtt_driver_release(struct drm_i915_private *i915) set_pages_array_wb(pvec->pages, pvec->nr); __pagevec_release(pvec); } - - i915_gem_cleanup_stolen(i915); } static unsigned int gen6_get_total_gtt_size(u16 snb_gmch_ctl) @@ -2873,35 +2870,51 @@ static int ggtt_probe_common(struct i915_ggtt *ggtt, u64 size) return 0; } -static void tgl_setup_private_ppat(struct drm_i915_private *dev_priv) +static void tgl_setup_private_ppat(struct intel_uncore *uncore) { /* TGL doesn't support LLC or AGE settings */ - I915_WRITE(GEN12_PAT_INDEX(0), GEN8_PPAT_WB); - I915_WRITE(GEN12_PAT_INDEX(1), GEN8_PPAT_WC); - I915_WRITE(GEN12_PAT_INDEX(2), GEN8_PPAT_WT); - I915_WRITE(GEN12_PAT_INDEX(3), GEN8_PPAT_UC); - I915_WRITE(GEN12_PAT_INDEX(4), GEN8_PPAT_WB); - I915_WRITE(GEN12_PAT_INDEX(5), GEN8_PPAT_WB); - I915_WRITE(GEN12_PAT_INDEX(6), GEN8_PPAT_WB); - I915_WRITE(GEN12_PAT_INDEX(7), GEN8_PPAT_WB); -} - -static void cnl_setup_private_ppat(struct drm_i915_private *dev_priv) -{ - I915_WRITE(GEN10_PAT_INDEX(0), GEN8_PPAT_WB | GEN8_PPAT_LLC); - I915_WRITE(GEN10_PAT_INDEX(1), GEN8_PPAT_WC | GEN8_PPAT_LLCELLC); - I915_WRITE(GEN10_PAT_INDEX(2), GEN8_PPAT_WT | GEN8_PPAT_LLCELLC); - I915_WRITE(GEN10_PAT_INDEX(3), GEN8_PPAT_UC); - I915_WRITE(GEN10_PAT_INDEX(4), GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(0)); - I915_WRITE(GEN10_PAT_INDEX(5), GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(1)); - I915_WRITE(GEN10_PAT_INDEX(6), GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(2)); - I915_WRITE(GEN10_PAT_INDEX(7), GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(3)); + intel_uncore_write(uncore, GEN12_PAT_INDEX(0), GEN8_PPAT_WB); + intel_uncore_write(uncore, GEN12_PAT_INDEX(1), GEN8_PPAT_WC); + intel_uncore_write(uncore, GEN12_PAT_INDEX(2), GEN8_PPAT_WT); + intel_uncore_write(uncore, GEN12_PAT_INDEX(3), GEN8_PPAT_UC); + intel_uncore_write(uncore, GEN12_PAT_INDEX(4), GEN8_PPAT_WB); + intel_uncore_write(uncore, GEN12_PAT_INDEX(5), GEN8_PPAT_WB); + intel_uncore_write(uncore, GEN12_PAT_INDEX(6), GEN8_PPAT_WB); + intel_uncore_write(uncore, GEN12_PAT_INDEX(7), GEN8_PPAT_WB); +} + +static void cnl_setup_private_ppat(struct intel_uncore *uncore) +{ + intel_uncore_write(uncore, + GEN10_PAT_INDEX(0), + GEN8_PPAT_WB | GEN8_PPAT_LLC); + intel_uncore_write(uncore, + GEN10_PAT_INDEX(1), + GEN8_PPAT_WC | GEN8_PPAT_LLCELLC); + intel_uncore_write(uncore, + GEN10_PAT_INDEX(2), + GEN8_PPAT_WT | GEN8_PPAT_LLCELLC); + intel_uncore_write(uncore, + GEN10_PAT_INDEX(3), + GEN8_PPAT_UC); + intel_uncore_write(uncore, + GEN10_PAT_INDEX(4), + GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(0)); + intel_uncore_write(uncore, + GEN10_PAT_INDEX(5), + GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(1)); + intel_uncore_write(uncore, + GEN10_PAT_INDEX(6), + GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(2)); + intel_uncore_write(uncore, + GEN10_PAT_INDEX(7), + GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(3)); } /* The GGTT and PPGTT need a private PPAT setup in order to handle cacheability * bits. When using advanced contexts each context stores its own PAT, but * writing this data shouldn't be harmful even in those cases. */ -static void bdw_setup_private_ppat(struct drm_i915_private *dev_priv) +static void bdw_setup_private_ppat(struct intel_uncore *uncore) { u64 pat; @@ -2914,11 +2927,11 @@ static void bdw_setup_private_ppat(struct drm_i915_private *dev_priv) GEN8_PPAT(6, GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(2)) | GEN8_PPAT(7, GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(3)); - I915_WRITE(GEN8_PRIVATE_PAT_LO, lower_32_bits(pat)); - I915_WRITE(GEN8_PRIVATE_PAT_HI, upper_32_bits(pat)); + intel_uncore_write(uncore, GEN8_PRIVATE_PAT_LO, lower_32_bits(pat)); + intel_uncore_write(uncore, GEN8_PRIVATE_PAT_HI, upper_32_bits(pat)); } -static void chv_setup_private_ppat(struct drm_i915_private *dev_priv) +static void chv_setup_private_ppat(struct intel_uncore *uncore) { u64 pat; @@ -2950,8 +2963,8 @@ static void chv_setup_private_ppat(struct drm_i915_private *dev_priv) GEN8_PPAT(6, CHV_PPAT_SNOOP) | GEN8_PPAT(7, CHV_PPAT_SNOOP); - I915_WRITE(GEN8_PRIVATE_PAT_LO, lower_32_bits(pat)); - I915_WRITE(GEN8_PRIVATE_PAT_HI, upper_32_bits(pat)); + intel_uncore_write(uncore, GEN8_PRIVATE_PAT_LO, lower_32_bits(pat)); + intel_uncore_write(uncore, GEN8_PRIVATE_PAT_HI, upper_32_bits(pat)); } static void gen6_gmch_remove(struct i915_address_space *vm) @@ -2962,18 +2975,26 @@ static void gen6_gmch_remove(struct i915_address_space *vm) cleanup_scratch_page(vm); } -static void setup_private_pat(struct drm_i915_private *dev_priv) +static void setup_private_pat(struct intel_uncore *uncore) { - GEM_BUG_ON(INTEL_GEN(dev_priv) < 8); + struct drm_i915_private *i915 = uncore->i915; - if (INTEL_GEN(dev_priv) >= 12) - tgl_setup_private_ppat(dev_priv); - else if (INTEL_GEN(dev_priv) >= 10) - cnl_setup_private_ppat(dev_priv); - else if (IS_CHERRYVIEW(dev_priv) || IS_GEN9_LP(dev_priv)) - chv_setup_private_ppat(dev_priv); + GEM_BUG_ON(INTEL_GEN(i915) < 8); + + if (INTEL_GEN(i915) >= 12) + tgl_setup_private_ppat(uncore); + else if (INTEL_GEN(i915) >= 10) + cnl_setup_private_ppat(uncore); + else if (IS_CHERRYVIEW(i915) || IS_GEN9_LP(i915)) + chv_setup_private_ppat(uncore); else - bdw_setup_private_ppat(dev_priv); + bdw_setup_private_ppat(uncore); +} + +static struct resource pci_resource(struct pci_dev *pdev, int bar) +{ + return (struct resource)DEFINE_RES_MEM(pci_resource_start(pdev, bar), + pci_resource_len(pdev, bar)); } static int gen8_gmch_probe(struct i915_ggtt *ggtt) @@ -2985,10 +3006,10 @@ static int gen8_gmch_probe(struct i915_ggtt *ggtt) int err; /* TODO: We're not aware of mappable constraints on gen8 yet */ - ggtt->gmadr = - (struct resource) DEFINE_RES_MEM(pci_resource_start(pdev, 2), - pci_resource_len(pdev, 2)); - ggtt->mappable_end = resource_size(&ggtt->gmadr); + if (!IS_DGFX(dev_priv)) { + ggtt->gmadr = pci_resource(pdev, 2); + ggtt->mappable_end = resource_size(&ggtt->gmadr); + } err = pci_set_dma_mask(pdev, DMA_BIT_MASK(39)); if (!err) @@ -3029,7 +3050,7 @@ static int gen8_gmch_probe(struct i915_ggtt *ggtt) ggtt->vm.pte_encode = gen8_pte_encode; - setup_private_pat(dev_priv); + setup_private_pat(ggtt->vm.gt->uncore); return ggtt_probe_common(ggtt, size); } @@ -3200,9 +3221,6 @@ int i915_ggtt_probe_hw(struct drm_i915_private *i915) static int ggtt_init_hw(struct i915_ggtt *ggtt) { struct drm_i915_private *i915 = ggtt->vm.i915; - int ret = 0; - - mutex_lock(&i915->drm.struct_mutex); i915_address_space_init(&ggtt->vm, VM_CLASS_GGTT); @@ -3212,24 +3230,23 @@ static int ggtt_init_hw(struct i915_ggtt *ggtt) ggtt->vm.has_read_only = IS_VALLEYVIEW(i915); if (!HAS_LLC(i915) && !HAS_PPGTT(i915)) - ggtt->vm.mm.color_adjust = i915_gtt_color_adjust; - - if (!io_mapping_init_wc(&ggtt->iomap, - ggtt->gmadr.start, - ggtt->mappable_end)) { - ggtt->vm.cleanup(&ggtt->vm); - ret = -EIO; - goto out; - } + ggtt->vm.mm.color_adjust = i915_ggtt_color_adjust; + + if (ggtt->mappable_end) { + if (!io_mapping_init_wc(&ggtt->iomap, + ggtt->gmadr.start, + ggtt->mappable_end)) { + ggtt->vm.cleanup(&ggtt->vm); + return -EIO; + } - ggtt->mtrr = arch_phys_wc_add(ggtt->gmadr.start, ggtt->mappable_end); + ggtt->mtrr = arch_phys_wc_add(ggtt->gmadr.start, + ggtt->mappable_end); + } i915_ggtt_init_fences(ggtt); -out: - mutex_unlock(&i915->drm.struct_mutex); - - return ret; + return 0; } /** @@ -3251,19 +3268,7 @@ int i915_ggtt_init_hw(struct drm_i915_private *dev_priv) if (ret) return ret; - /* - * Initialise stolen early so that we may reserve preallocated - * objects for the BIOS to KMS transition. - */ - ret = i915_gem_init_stolen(dev_priv); - if (ret) - goto out_gtt_cleanup; - return 0; - -out_gtt_cleanup: - dev_priv->ggtt.vm.cleanup(&dev_priv->ggtt.vm); - return ret; } int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv) @@ -3301,6 +3306,7 @@ static void ggtt_restore_mappings(struct i915_ggtt *ggtt) { struct i915_vma *vma, *vn; bool flush = false; + int open; intel_gt_check_and_clear_faults(ggtt->vm.gt); @@ -3308,33 +3314,31 @@ static void ggtt_restore_mappings(struct i915_ggtt *ggtt) /* First fill our portion of the GTT with scratch pages */ ggtt->vm.clear_range(&ggtt->vm, 0, ggtt->vm.total); - ggtt->vm.closed = true; /* skip rewriting PTE on VMA unbind */ + + /* Skip rewriting PTE on VMA unbind. */ + open = atomic_xchg(&ggtt->vm.open, 0); /* clflush objects bound into the GGTT and rebind them. */ list_for_each_entry_safe(vma, vn, &ggtt->vm.bound_list, vm_link) { struct drm_i915_gem_object *obj = vma->obj; - if (!(vma->flags & I915_VMA_GLOBAL_BIND)) + if (!i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND)) continue; - mutex_unlock(&ggtt->vm.mutex); - - if (!i915_vma_unbind(vma)) - goto lock; + if (!__i915_vma_unbind(vma)) + continue; + clear_bit(I915_VMA_GLOBAL_BIND_BIT, __i915_vma_flags(vma)); WARN_ON(i915_vma_bind(vma, obj ? obj->cache_level : 0, - PIN_UPDATE)); + PIN_GLOBAL, NULL)); if (obj) { /* only used during resume => exclusive access */ flush |= fetch_and_zero(&obj->write_domain); obj->read_domains |= I915_GEM_DOMAIN_GTT; } - -lock: - mutex_lock(&ggtt->vm.mutex); } - ggtt->vm.closed = false; + atomic_set(&ggtt->vm.open, open); ggtt->invalidate(ggtt); mutex_unlock(&ggtt->vm.mutex); @@ -3345,10 +3349,12 @@ lock: void i915_gem_restore_gtt_mappings(struct drm_i915_private *i915) { - ggtt_restore_mappings(&i915->ggtt); + struct i915_ggtt *ggtt = &i915->ggtt; + + ggtt_restore_mappings(ggtt); if (INTEL_GEN(i915) >= 8) - setup_private_pat(i915); + setup_private_pat(ggtt->vm.gt->uncore); } static struct scatterlist * @@ -3726,7 +3732,8 @@ int i915_gem_gtt_insert(struct i915_address_space *vm, u64 offset; int err; - lockdep_assert_held(&vm->i915->drm.struct_mutex); + lockdep_assert_held(&vm->mutex); + GEM_BUG_ON(!size); GEM_BUG_ON(!IS_ALIGNED(size, I915_GTT_PAGE_SIZE)); GEM_BUG_ON(alignment && !is_power_of_2(alignment)); diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h index b97a47fc7a68..402283ce2864 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.h +++ b/drivers/gpu/drm/i915/i915_gem_gtt.h @@ -148,8 +148,8 @@ typedef u64 gen8_pte_t; #define GEN8_PDE_IPS_64K BIT(11) #define GEN8_PDE_PS_2M BIT(7) -#define for_each_sgt_dma(__dmap, __iter, __sgt) \ - __for_each_sgt_dma(__dmap, __iter, __sgt, I915_GTT_PAGE_SIZE) +#define for_each_sgt_daddr(__dp, __iter, __sgt) \ + __for_each_sgt_daddr(__dp, __iter, __sgt, I915_GTT_PAGE_SIZE) struct intel_remapped_plane_info { /* in gtt pages */ @@ -305,7 +305,16 @@ struct i915_address_space { u64 total; /* size addr space maps (ex. 2GB for ggtt) */ u64 reserved; /* size addr space reserved */ - bool closed; + unsigned int bind_async_flags; + + /* + * Each active user context has its own address space (in full-ppgtt). + * Since the vm may be shared between multiple contexts, we count how + * many contexts keep us "open". Once open hits zero, we are closed + * and do not allow any new attachments, and proceed to shutdown our + * vma and page directories. + */ + atomic_t open; struct mutex mutex; /* protects vma and our lists */ #define VM_CLASS_GGTT 0 @@ -320,11 +329,6 @@ struct i915_address_space { */ struct list_head bound_list; - /** - * List of vma that are not unbound. - */ - struct list_head unbound_list; - struct pagestash free_pages; /* Global GTT */ @@ -376,6 +380,12 @@ i915_vm_has_scratch_64K(struct i915_address_space *vm) return vm->scratch_order == get_order(I915_GTT_PAGE_SIZE_64K); } +static inline bool +i915_vm_has_cache_coloring(struct i915_address_space *vm) +{ + return i915_is_ggtt(vm) && vm->mm.color_adjust; +} + /* The Graphics Translation Table is the way in which GEN hardware translates a * Graphics Virtual Address into a Physical Address. In addition to the normal * collateral associated with any va->pa translations GEN hardware also has a @@ -401,6 +411,11 @@ struct i915_ggtt { int mtrr; + /** Bit 6 swizzling required for X tiling */ + u32 bit_6_swizzle_x; + /** Bit 6 swizzling required for Y tiling */ + u32 bit_6_swizzle_y; + u32 pin_bias; unsigned int num_fences; @@ -422,7 +437,6 @@ struct i915_ggtt { struct i915_ppgtt { struct i915_address_space vm; - intel_engine_mask_t pd_dirty_engines; struct i915_page_directory *pd; }; @@ -432,7 +446,9 @@ struct gen6_ppgtt { struct i915_vma *vma; gen6_pte_t __iomem *pd_addr; - unsigned int pin_count; + atomic_t pin_count; + struct mutex pin_mutex; + bool scan_for_unused_pt; }; @@ -559,6 +575,11 @@ void i915_ggtt_disable_guc(struct i915_ggtt *ggtt); int i915_init_ggtt(struct drm_i915_private *dev_priv); void i915_ggtt_driver_release(struct drm_i915_private *dev_priv); +static inline bool i915_ggtt_has_aperture(const struct i915_ggtt *ggtt) +{ + return ggtt->mappable_end > 0; +} + int i915_ppgtt_init_hw(struct intel_gt *gt); struct i915_ppgtt *i915_ppgtt_create(struct drm_i915_private *dev_priv); @@ -577,6 +598,35 @@ static inline void i915_vm_put(struct i915_address_space *vm) kref_put(&vm->ref, i915_vm_release); } +static inline struct i915_address_space * +i915_vm_open(struct i915_address_space *vm) +{ + GEM_BUG_ON(!atomic_read(&vm->open)); + atomic_inc(&vm->open); + return i915_vm_get(vm); +} + +static inline bool +i915_vm_tryopen(struct i915_address_space *vm) +{ + if (atomic_add_unless(&vm->open, 1, 0)) + return i915_vm_get(vm); + + return false; +} + +void __i915_vm_close(struct i915_address_space *vm); + +static inline void +i915_vm_close(struct i915_address_space *vm) +{ + GEM_BUG_ON(!atomic_read(&vm->open)); + if (atomic_dec_and_test(&vm->open)) + __i915_vm_close(vm); + + i915_vm_put(vm); +} + int gen6_ppgtt_pin(struct i915_ppgtt *base); void gen6_ppgtt_unpin(struct i915_ppgtt *base); void gen6_ppgtt_unpin_all(struct i915_ppgtt *base); @@ -609,10 +659,9 @@ int i915_gem_gtt_insert(struct i915_address_space *vm, #define PIN_OFFSET_BIAS BIT_ULL(6) #define PIN_OFFSET_FIXED BIT_ULL(7) -#define PIN_MBZ BIT_ULL(8) /* I915_VMA_PIN_OVERFLOW */ -#define PIN_GLOBAL BIT_ULL(9) /* I915_VMA_GLOBAL_BIND */ -#define PIN_USER BIT_ULL(10) /* I915_VMA_LOCAL_BIND */ -#define PIN_UPDATE BIT_ULL(11) +#define PIN_UPDATE BIT_ULL(9) +#define PIN_GLOBAL BIT_ULL(10) /* I915_VMA_GLOBAL_BIND */ +#define PIN_USER BIT_ULL(11) /* I915_VMA_LOCAL_BIND */ #define PIN_OFFSET_MASK (-I915_GTT_PAGE_SIZE) diff --git a/drivers/gpu/drm/i915/i915_getparam.c b/drivers/gpu/drm/i915/i915_getparam.c index 5d9101376a3d..cf8a8c3ef047 100644 --- a/drivers/gpu/drm/i915/i915_getparam.c +++ b/drivers/gpu/drm/i915/i915_getparam.c @@ -5,6 +5,7 @@ #include "gt/intel_engine_user.h" #include "i915_drv.h" +#include "i915_perf.h" int i915_getparam_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -62,7 +63,7 @@ int i915_getparam_ioctl(struct drm_device *dev, void *data, value = !!(i915->caps.scheduler & I915_SCHEDULER_CAP_SEMAPHORES); break; case I915_PARAM_HAS_SECURE_BATCHES: - value = capable(CAP_SYS_ADMIN); + value = HAS_SECURE_BATCHES(i915) && capable(CAP_SYS_ADMIN); break; case I915_PARAM_CMD_PARSER_VERSION: value = i915_cmd_parser_get_version(i915); @@ -79,8 +80,8 @@ int i915_getparam_ioctl(struct drm_device *dev, void *data, break; case I915_PARAM_HAS_GPU_RESET: value = i915_modparams.enable_hangcheck && - intel_has_gpu_reset(i915); - if (value && intel_has_reset_engine(i915)) + intel_has_gpu_reset(&i915->gt); + if (value && intel_has_reset_engine(&i915->gt)) value = 2; break; case I915_PARAM_HAS_RESOURCE_STREAMER: @@ -156,6 +157,9 @@ int i915_getparam_ioctl(struct drm_device *dev, void *data, case I915_PARAM_MMAP_GTT_COHERENT: value = INTEL_INFO(i915)->has_coherent_ggtt; break; + case I915_PARAM_PERF_REVISION: + value = i915_perf_ioctl_version(); + break; default: DRM_DEBUG("Unknown parameter %d\n", param->param); return -EINVAL; diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index e284bd76fa86..3c85cb0ee99f 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -40,6 +40,7 @@ #include "display/intel_overlay.h" #include "gem/i915_gem_context.h" +#include "gem/i915_gem_lmem.h" #include "i915_drv.h" #include "i915_gpu_error.h" @@ -235,6 +236,7 @@ struct compress { struct pagevec pool; struct z_stream_s zstream; void *tmp; + bool wc; }; static bool compress_init(struct compress *c) @@ -292,7 +294,7 @@ static int compress_page(struct compress *c, struct z_stream_s *zstream = &c->zstream; zstream->next_in = src; - if (c->tmp && i915_memcpy_from_wc(c->tmp, src, PAGE_SIZE)) + if (c->wc && c->tmp && i915_memcpy_from_wc(c->tmp, src, PAGE_SIZE)) zstream->next_in = c->tmp; zstream->avail_in = PAGE_SIZE; @@ -367,6 +369,7 @@ static void err_compression_marker(struct drm_i915_error_state_buf *m) struct compress { struct pagevec pool; + bool wc; }; static bool compress_init(struct compress *c) @@ -389,7 +392,7 @@ static int compress_page(struct compress *c, if (!ptr) return -ENOMEM; - if (!i915_memcpy_from_wc(ptr, src, PAGE_SIZE)) + if (!(c->wc && i915_memcpy_from_wc(ptr, src, PAGE_SIZE))) memcpy(ptr, src, PAGE_SIZE); dst->pages[dst->page_count++] = ptr; @@ -421,6 +424,7 @@ static void err_compression_marker(struct drm_i915_error_state_buf *m) static void error_print_instdone(struct drm_i915_error_state_buf *m, const struct drm_i915_error_engine *ee) { + const struct sseu_dev_info *sseu = &RUNTIME_INFO(m->i915)->sseu; int slice; int subslice; @@ -436,12 +440,12 @@ static void error_print_instdone(struct drm_i915_error_state_buf *m, if (INTEL_GEN(m->i915) <= 6) return; - for_each_instdone_slice_subslice(m->i915, slice, subslice) + for_each_instdone_slice_subslice(m->i915, sseu, slice, subslice) err_printf(m, " SAMPLER_INSTDONE[%d][%d]: 0x%08x\n", slice, subslice, ee->instdone.sampler[slice][subslice]); - for_each_instdone_slice_subslice(m->i915, slice, subslice) + for_each_instdone_slice_subslice(m->i915, sseu, slice, subslice) err_printf(m, " ROW_INSTDONE[%d][%d]: 0x%08x\n", slice, subslice, ee->instdone.row[slice][subslice]); @@ -470,9 +474,9 @@ static void error_print_context(struct drm_i915_error_state_buf *m, const char *header, const struct drm_i915_error_context *ctx) { - err_printf(m, "%s%s[%d] hw_id %d, prio %d, guilty %d active %d\n", - header, ctx->comm, ctx->pid, ctx->hw_id, - ctx->sched_attr.priority, ctx->guilty, ctx->active); + err_printf(m, "%s%s[%d] prio %d, guilty %d active %d\n", + header, ctx->comm, ctx->pid, ctx->sched_attr.priority, + ctx->guilty, ctx->active); } static void error_print_engine(struct drm_i915_error_state_buf *m, @@ -533,10 +537,6 @@ static void error_print_engine(struct drm_i915_error_state_buf *m, } err_printf(m, " ring->head: 0x%08x\n", ee->cpu_ring_head); err_printf(m, " ring->tail: 0x%08x\n", ee->cpu_ring_tail); - err_printf(m, " hangcheck timestamp: %dms (%lu%s)\n", - jiffies_to_msecs(ee->hangcheck_timestamp - epoch), - ee->hangcheck_timestamp, - ee->hangcheck_timestamp == epoch ? "; epoch" : ""); err_printf(m, " engine reset count: %u\n", ee->reset_count); for (n = 0; n < ee->num_ports; n++) { @@ -574,6 +574,9 @@ static void print_error_obj(struct drm_i915_error_state_buf *m, lower_32_bits(obj->gtt_offset)); } + if (obj->gtt_page_sizes > I915_GTT_PAGE_SIZE_4K) + err_printf(m, "gtt_page_sizes = 0x%08x\n", obj->gtt_page_sizes); + err_compression_marker(m); for (page = 0; page < obj->page_count; page++) { int i, len; @@ -675,11 +678,8 @@ static void __err_print_to_sgl(struct drm_i915_error_state_buf *m, ts = ktime_to_timespec64(error->uptime); err_printf(m, "Uptime: %lld s %ld us\n", (s64)ts.tv_sec, ts.tv_nsec / NSEC_PER_USEC); - err_printf(m, "Epoch: %lu jiffies (%u HZ)\n", error->epoch, HZ); - err_printf(m, "Capture: %lu jiffies; %d ms ago, %d ms after epoch\n", - error->capture, - jiffies_to_msecs(jiffies - error->capture), - jiffies_to_msecs(error->capture - error->epoch)); + err_printf(m, "Capture: %lu jiffies; %d ms ago\n", + error->capture, jiffies_to_msecs(jiffies - error->capture)); for (ee = error->engine; ee; ee = ee->next) err_printf(m, "Active process (on ring %s): %s [%d]\n", @@ -734,8 +734,24 @@ static void __err_print_to_sgl(struct drm_i915_error_state_buf *m, if (IS_GEN(m->i915, 7)) err_printf(m, "ERR_INT: 0x%08x\n", error->err_int); + if (IS_GEN_RANGE(m->i915, 8, 11)) + err_printf(m, "GTT_CACHE_EN: 0x%08x\n", error->gtt_cache); + + if (IS_GEN(m->i915, 12)) + err_printf(m, "AUX_ERR_DBG: 0x%08x\n", error->aux_err); + + if (INTEL_GEN(m->i915) >= 12) { + int i; + + for (i = 0; i < GEN12_SFC_DONE_MAX; i++) + err_printf(m, " SFC_DONE[%d]: 0x%08x\n", i, + error->sfc_done[i]); + + err_printf(m, " GAM_DONE: 0x%08x\n", error->gam_done); + } + for (ee = error->engine; ee; ee = ee->next) - error_print_engine(m, ee, error->epoch); + error_print_engine(m, ee, error->capture); for (ee = error->engine; ee; ee = ee->next) { const struct drm_i915_error_object *obj; @@ -763,7 +779,7 @@ static void __err_print_to_sgl(struct drm_i915_error_state_buf *m, for (j = 0; j < ee->num_requests; j++) error_print_request(m, " ", &ee->requests[j], - error->epoch); + error->capture); } print_error_obj(m, ee->engine, "ringbuffer", ee->ringbuffer); @@ -963,7 +979,6 @@ i915_error_object_create(struct drm_i915_private *i915, struct drm_i915_error_object *dst; unsigned long num_pages; struct sgt_iter iter; - dma_addr_t dma; int ret; might_sleep(); @@ -984,21 +999,59 @@ i915_error_object_create(struct drm_i915_private *i915, dst->gtt_offset = vma->node.start; dst->gtt_size = vma->node.size; + dst->gtt_page_sizes = vma->page_sizes.gtt; dst->num_pages = num_pages; dst->page_count = 0; dst->unused = 0; + compress->wc = i915_gem_object_is_lmem(vma->obj) || + drm_mm_node_allocated(&ggtt->error_capture); + ret = -EINVAL; - for_each_sgt_dma(dma, iter, vma->pages) { + if (drm_mm_node_allocated(&ggtt->error_capture)) { void __iomem *s; + dma_addr_t dma; - ggtt->vm.insert_page(&ggtt->vm, dma, slot, I915_CACHE_NONE, 0); + for_each_sgt_daddr(dma, iter, vma->pages) { + ggtt->vm.insert_page(&ggtt->vm, dma, slot, + I915_CACHE_NONE, 0); - s = io_mapping_map_wc(&ggtt->iomap, slot, PAGE_SIZE); - ret = compress_page(compress, (void __force *)s, dst); - io_mapping_unmap(s); - if (ret) - break; + s = io_mapping_map_wc(&ggtt->iomap, slot, PAGE_SIZE); + ret = compress_page(compress, (void __force *)s, dst); + io_mapping_unmap(s); + if (ret) + break; + } + } else if (i915_gem_object_is_lmem(vma->obj)) { + struct intel_memory_region *mem = vma->obj->mm.region; + dma_addr_t dma; + + for_each_sgt_daddr(dma, iter, vma->pages) { + void __iomem *s; + + s = io_mapping_map_wc(&mem->iomap, dma, PAGE_SIZE); + ret = compress_page(compress, (void __force *)s, dst); + io_mapping_unmap(s); + if (ret) + break; + } + } else { + struct page *page; + + for_each_sgt_page(page, iter, vma->pages) { + void *s; + + drm_clflush_pages(&page, 1); + + s = kmap(page); + ret = compress_page(compress, s, dst); + kunmap(s); + + drm_clflush_pages(&page, 1); + + if (ret) + break; + } } if (ret || compress_flush(compress, dst)) { @@ -1136,8 +1189,6 @@ static void error_record_engine_registers(struct i915_gpu_state *error, } ee->idle = intel_engine_is_idle(engine); - if (!ee->idle) - ee->hangcheck_timestamp = engine->hangcheck.action_timestamp; ee->reset_count = i915_reset_engine_count(&dev_priv->gpu_error, engine); @@ -1263,7 +1314,6 @@ static bool record_context(struct drm_i915_error_context *e, rcu_read_unlock(); } - e->hw_id = ctx->hw_id; e->sched_attr = ctx->sched; e->guilty = atomic_read(&ctx->guilty_count); e->active = atomic_read(&ctx->active_count); @@ -1291,7 +1341,7 @@ capture_vma(struct capture_vma *next, if (!c) return next; - if (!i915_active_trygrab(&vma->active)) { + if (!i915_active_acquire_if_busy(&vma->active)) { kfree(c); return next; } @@ -1431,7 +1481,7 @@ gem_record_rings(struct i915_gpu_state *error, struct compress *compress) *this->slot = i915_error_object_create(i915, vma, compress); - i915_active_ungrab(&vma->active); + i915_active_release(&vma->active); i915_vma_put(vma); capture = this->next; @@ -1553,6 +1603,21 @@ static void capture_reg_state(struct i915_gpu_state *error) error->gac_eco = intel_uncore_read(uncore, GAC_ECO_BITS); } + if (IS_GEN_RANGE(i915, 8, 11)) + error->gtt_cache = intel_uncore_read(uncore, HSW_GTT_CACHE_EN); + + if (IS_GEN(i915, 12)) + error->aux_err = intel_uncore_read(uncore, GEN12_AUX_ERR_DBG); + + if (INTEL_GEN(i915) >= 12) { + for (i = 0; i < GEN12_SFC_DONE_MAX; i++) { + error->sfc_done[i] = + intel_uncore_read(uncore, GEN12_SFC_DONE(i)); + } + + error->gam_done = intel_uncore_read(uncore, GEN12_GAM_DONE); + } + /* 4: Everything else */ if (INTEL_GEN(i915) >= 11) { error->ier = intel_uncore_read(uncore, GEN8_DE_MISC_IER); @@ -1647,26 +1712,15 @@ static void capture_params(struct i915_gpu_state *error) i915_params_copy(&error->params, &i915_modparams); } -static unsigned long capture_find_epoch(const struct i915_gpu_state *error) -{ - const struct drm_i915_error_engine *ee; - unsigned long epoch = error->capture; - - for (ee = error->engine; ee; ee = ee->next) { - if (ee->hangcheck_timestamp && - time_before(ee->hangcheck_timestamp, epoch)) - epoch = ee->hangcheck_timestamp; - } - - return epoch; -} - static void capture_finish(struct i915_gpu_state *error) { struct i915_ggtt *ggtt = &error->i915->ggtt; - const u64 slot = ggtt->error_capture.start; - ggtt->vm.clear_range(&ggtt->vm, slot, PAGE_SIZE); + if (drm_mm_node_allocated(&ggtt->error_capture)) { + const u64 slot = ggtt->error_capture.start; + + ggtt->vm.clear_range(&ggtt->vm, slot, PAGE_SIZE); + } } #define DAY_AS_SECONDS(x) (24 * 60 * 60 * (x)) @@ -1712,8 +1766,6 @@ i915_capture_gpu_state(struct drm_i915_private *i915) error->overlay = intel_overlay_capture_error_state(i915); error->display = intel_display_capture_error_state(i915); - error->epoch = capture_find_epoch(error); - capture_finish(error); compress_fini(&compress); diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h b/drivers/gpu/drm/i915/i915_gpu_error.h index df9f57766626..5d2c3372ff99 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.h +++ b/drivers/gpu/drm/i915/i915_gpu_error.h @@ -34,7 +34,6 @@ struct i915_gpu_state { ktime_t boottime; ktime_t uptime; unsigned long capture; - unsigned long epoch; struct drm_i915_private *i915; @@ -74,6 +73,10 @@ struct i915_gpu_state { u32 gam_ecochk; u32 gab_ctl; u32 gfx_mode; + u32 gtt_cache; + u32 aux_err; /* gen12 */ + u32 sfc_done[GEN12_SFC_DONE_MAX]; /* gen12 */ + u32 gam_done; /* gen12 */ u32 nfence; u64 fence[I915_MAX_NUM_FENCES]; @@ -85,7 +88,6 @@ struct i915_gpu_state { /* Software tracked state */ bool idle; - unsigned long hangcheck_timestamp; int num_requests; u32 reset_count; @@ -118,7 +120,6 @@ struct i915_gpu_state { struct drm_i915_error_context { char comm[TASK_COMM_LEN]; pid_t pid; - u32 hw_id; int active; int guilty; struct i915_sched_attr sched_attr; @@ -127,6 +128,7 @@ struct i915_gpu_state { struct drm_i915_error_object { u64 gtt_offset; u64 gtt_size; + u32 gtt_page_sizes; int num_pages; int page_count; int unused; diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 37e3dd3c1a9d..dae00f7dd7df 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -29,7 +29,6 @@ #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt #include <linux/circ_buf.h> -#include <linux/cpuidle.h> #include <linux/slab.h> #include <linux/sysrq.h> @@ -46,6 +45,7 @@ #include "gt/intel_gt.h" #include "gt/intel_gt_irq.h" #include "gt/intel_gt_pm_irq.h" +#include "gt/intel_rps.h" #include "i915_drv.h" #include "i915_irq.h" @@ -149,30 +149,24 @@ static const u32 hpd_gen12[HPD_NUM_PINS] = { }; static const u32 hpd_icp[HPD_NUM_PINS] = { - [HPD_PORT_A] = SDE_DDIA_HOTPLUG_ICP, - [HPD_PORT_B] = SDE_DDIB_HOTPLUG_ICP, - [HPD_PORT_C] = SDE_TC1_HOTPLUG_ICP, - [HPD_PORT_D] = SDE_TC2_HOTPLUG_ICP, - [HPD_PORT_E] = SDE_TC3_HOTPLUG_ICP, - [HPD_PORT_F] = SDE_TC4_HOTPLUG_ICP -}; - -static const u32 hpd_mcc[HPD_NUM_PINS] = { - [HPD_PORT_A] = SDE_DDIA_HOTPLUG_ICP, - [HPD_PORT_B] = SDE_DDIB_HOTPLUG_ICP, - [HPD_PORT_C] = SDE_TC1_HOTPLUG_ICP + [HPD_PORT_A] = SDE_DDI_HOTPLUG_ICP(PORT_A), + [HPD_PORT_B] = SDE_DDI_HOTPLUG_ICP(PORT_B), + [HPD_PORT_C] = SDE_TC_HOTPLUG_ICP(PORT_TC1), + [HPD_PORT_D] = SDE_TC_HOTPLUG_ICP(PORT_TC2), + [HPD_PORT_E] = SDE_TC_HOTPLUG_ICP(PORT_TC3), + [HPD_PORT_F] = SDE_TC_HOTPLUG_ICP(PORT_TC4), }; static const u32 hpd_tgp[HPD_NUM_PINS] = { - [HPD_PORT_A] = SDE_DDIA_HOTPLUG_ICP, - [HPD_PORT_B] = SDE_DDIB_HOTPLUG_ICP, - [HPD_PORT_C] = SDE_DDIC_HOTPLUG_TGP, - [HPD_PORT_D] = SDE_TC1_HOTPLUG_ICP, - [HPD_PORT_E] = SDE_TC2_HOTPLUG_ICP, - [HPD_PORT_F] = SDE_TC3_HOTPLUG_ICP, - [HPD_PORT_G] = SDE_TC4_HOTPLUG_ICP, - [HPD_PORT_H] = SDE_TC5_HOTPLUG_TGP, - [HPD_PORT_I] = SDE_TC6_HOTPLUG_TGP, + [HPD_PORT_A] = SDE_DDI_HOTPLUG_ICP(PORT_A), + [HPD_PORT_B] = SDE_DDI_HOTPLUG_ICP(PORT_B), + [HPD_PORT_C] = SDE_DDI_HOTPLUG_ICP(PORT_C), + [HPD_PORT_D] = SDE_TC_HOTPLUG_ICP(PORT_TC1), + [HPD_PORT_E] = SDE_TC_HOTPLUG_ICP(PORT_TC2), + [HPD_PORT_F] = SDE_TC_HOTPLUG_ICP(PORT_TC3), + [HPD_PORT_G] = SDE_TC_HOTPLUG_ICP(PORT_TC4), + [HPD_PORT_H] = SDE_TC_HOTPLUG_ICP(PORT_TC5), + [HPD_PORT_I] = SDE_TC_HOTPLUG_ICP(PORT_TC6), }; void gen3_irq_reset(struct intel_uncore *uncore, i915_reg_t imr, @@ -327,180 +321,6 @@ void ilk_update_display_irq(struct drm_i915_private *dev_priv, } } -static i915_reg_t gen6_pm_iir(struct drm_i915_private *dev_priv) -{ - WARN_ON_ONCE(INTEL_GEN(dev_priv) >= 11); - - return INTEL_GEN(dev_priv) >= 8 ? GEN8_GT_IIR(2) : GEN6_PMIIR; -} - -void gen11_reset_rps_interrupts(struct drm_i915_private *dev_priv) -{ - struct intel_gt *gt = &dev_priv->gt; - - spin_lock_irq(>->irq_lock); - - while (gen11_gt_reset_one_iir(gt, 0, GEN11_GTPM)) - ; - - dev_priv->gt_pm.rps.pm_iir = 0; - - spin_unlock_irq(>->irq_lock); -} - -void gen6_reset_rps_interrupts(struct drm_i915_private *dev_priv) -{ - struct intel_gt *gt = &dev_priv->gt; - - spin_lock_irq(>->irq_lock); - gen6_gt_pm_reset_iir(gt, GEN6_PM_RPS_EVENTS); - dev_priv->gt_pm.rps.pm_iir = 0; - spin_unlock_irq(>->irq_lock); -} - -void gen6_enable_rps_interrupts(struct drm_i915_private *dev_priv) -{ - struct intel_gt *gt = &dev_priv->gt; - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - if (READ_ONCE(rps->interrupts_enabled)) - return; - - spin_lock_irq(>->irq_lock); - WARN_ON_ONCE(rps->pm_iir); - - if (INTEL_GEN(dev_priv) >= 11) - WARN_ON_ONCE(gen11_gt_reset_one_iir(gt, 0, GEN11_GTPM)); - else - WARN_ON_ONCE(I915_READ(gen6_pm_iir(dev_priv)) & dev_priv->pm_rps_events); - - rps->interrupts_enabled = true; - gen6_gt_pm_enable_irq(gt, dev_priv->pm_rps_events); - - spin_unlock_irq(>->irq_lock); -} - -u32 gen6_sanitize_rps_pm_mask(const struct drm_i915_private *i915, u32 mask) -{ - return mask & ~i915->gt_pm.rps.pm_intrmsk_mbz; -} - -void gen6_disable_rps_interrupts(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - struct intel_gt *gt = &dev_priv->gt; - - if (!READ_ONCE(rps->interrupts_enabled)) - return; - - spin_lock_irq(>->irq_lock); - rps->interrupts_enabled = false; - - I915_WRITE(GEN6_PMINTRMSK, gen6_sanitize_rps_pm_mask(dev_priv, ~0u)); - - gen6_gt_pm_disable_irq(gt, GEN6_PM_RPS_EVENTS); - - spin_unlock_irq(>->irq_lock); - intel_synchronize_irq(dev_priv); - - /* Now that we will not be generating any more work, flush any - * outstanding tasks. As we are called on the RPS idle path, - * we will reset the GPU to minimum frequencies, so the current - * state of the worker can be discarded. - */ - cancel_work_sync(&rps->work); - if (INTEL_GEN(dev_priv) >= 11) - gen11_reset_rps_interrupts(dev_priv); - else - gen6_reset_rps_interrupts(dev_priv); -} - -void gen9_reset_guc_interrupts(struct intel_guc *guc) -{ - struct intel_gt *gt = guc_to_gt(guc); - - assert_rpm_wakelock_held(>->i915->runtime_pm); - - spin_lock_irq(>->irq_lock); - gen6_gt_pm_reset_iir(gt, gt->pm_guc_events); - spin_unlock_irq(>->irq_lock); -} - -void gen9_enable_guc_interrupts(struct intel_guc *guc) -{ - struct intel_gt *gt = guc_to_gt(guc); - - assert_rpm_wakelock_held(>->i915->runtime_pm); - - spin_lock_irq(>->irq_lock); - if (!guc->interrupts.enabled) { - WARN_ON_ONCE(intel_uncore_read(gt->uncore, - gen6_pm_iir(gt->i915)) & - gt->pm_guc_events); - guc->interrupts.enabled = true; - gen6_gt_pm_enable_irq(gt, gt->pm_guc_events); - } - spin_unlock_irq(>->irq_lock); -} - -void gen9_disable_guc_interrupts(struct intel_guc *guc) -{ - struct intel_gt *gt = guc_to_gt(guc); - - assert_rpm_wakelock_held(>->i915->runtime_pm); - - spin_lock_irq(>->irq_lock); - guc->interrupts.enabled = false; - - gen6_gt_pm_disable_irq(gt, gt->pm_guc_events); - - spin_unlock_irq(>->irq_lock); - intel_synchronize_irq(gt->i915); - - gen9_reset_guc_interrupts(guc); -} - -void gen11_reset_guc_interrupts(struct intel_guc *guc) -{ - struct intel_gt *gt = guc_to_gt(guc); - - spin_lock_irq(>->irq_lock); - gen11_gt_reset_one_iir(gt, 0, GEN11_GUC); - spin_unlock_irq(>->irq_lock); -} - -void gen11_enable_guc_interrupts(struct intel_guc *guc) -{ - struct intel_gt *gt = guc_to_gt(guc); - - spin_lock_irq(>->irq_lock); - if (!guc->interrupts.enabled) { - u32 events = REG_FIELD_PREP(ENGINE1_MASK, GUC_INTR_GUC2HOST); - - WARN_ON_ONCE(gen11_gt_reset_one_iir(gt, 0, GEN11_GUC)); - intel_uncore_write(gt->uncore, GEN11_GUC_SG_INTR_ENABLE, events); - intel_uncore_write(gt->uncore, GEN11_GUC_SG_INTR_MASK, ~events); - guc->interrupts.enabled = true; - } - spin_unlock_irq(>->irq_lock); -} - -void gen11_disable_guc_interrupts(struct intel_guc *guc) -{ - struct intel_gt *gt = guc_to_gt(guc); - - spin_lock_irq(>->irq_lock); - guc->interrupts.enabled = false; - - intel_uncore_write(gt->uncore, GEN11_GUC_SG_INTR_MASK, ~0); - intel_uncore_write(gt->uncore, GEN11_GUC_SG_INTR_ENABLE, 0); - - spin_unlock_irq(>->irq_lock); - intel_synchronize_irq(gt->i915); - - gen11_reset_guc_interrupts(guc); -} - /** * bdw_update_port_irq - update DE port interrupt * @dev_priv: driver private @@ -942,14 +762,14 @@ static int __intel_get_crtc_scanline(struct intel_crtc *crtc) return (position + crtc->scanline_offset) % vtotal; } -bool i915_get_crtc_scanoutpos(struct drm_device *dev, unsigned int pipe, +bool i915_get_crtc_scanoutpos(struct drm_device *dev, unsigned int index, bool in_vblank_irq, int *vpos, int *hpos, ktime_t *stime, ktime_t *etime, const struct drm_display_mode *mode) { struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *intel_crtc = intel_get_crtc_for_pipe(dev_priv, - pipe); + struct intel_crtc *crtc = to_intel_crtc(drm_crtc_from_index(dev, index)); + enum pipe pipe = crtc->pipe; int position; int vbl_start, vbl_end, hsync_start, htotal, vtotal; unsigned long irqflags; @@ -992,7 +812,7 @@ bool i915_get_crtc_scanoutpos(struct drm_device *dev, unsigned int pipe, /* No obvious pixelcount register. Only query vertical * scanout position from Display scan line register. */ - position = __intel_get_crtc_scanline(intel_crtc); + position = __intel_get_crtc_scanline(crtc); } else { /* Have access to pixelcount since start of frame. * We can split this into vertical and horizontal @@ -1072,199 +892,6 @@ int intel_get_crtc_scanline(struct intel_crtc *crtc) return position; } -static void ironlake_rps_change_irq_handler(struct drm_i915_private *dev_priv) -{ - struct intel_uncore *uncore = &dev_priv->uncore; - u32 busy_up, busy_down, max_avg, min_avg; - u8 new_delay; - - spin_lock(&mchdev_lock); - - intel_uncore_write16(uncore, - MEMINTRSTS, - intel_uncore_read(uncore, MEMINTRSTS)); - - new_delay = dev_priv->ips.cur_delay; - - intel_uncore_write16(uncore, MEMINTRSTS, MEMINT_EVAL_CHG); - busy_up = intel_uncore_read(uncore, RCPREVBSYTUPAVG); - busy_down = intel_uncore_read(uncore, RCPREVBSYTDNAVG); - max_avg = intel_uncore_read(uncore, RCBMAXAVG); - min_avg = intel_uncore_read(uncore, RCBMINAVG); - - /* Handle RCS change request from hw */ - if (busy_up > max_avg) { - if (dev_priv->ips.cur_delay != dev_priv->ips.max_delay) - new_delay = dev_priv->ips.cur_delay - 1; - if (new_delay < dev_priv->ips.max_delay) - new_delay = dev_priv->ips.max_delay; - } else if (busy_down < min_avg) { - if (dev_priv->ips.cur_delay != dev_priv->ips.min_delay) - new_delay = dev_priv->ips.cur_delay + 1; - if (new_delay > dev_priv->ips.min_delay) - new_delay = dev_priv->ips.min_delay; - } - - if (ironlake_set_drps(dev_priv, new_delay)) - dev_priv->ips.cur_delay = new_delay; - - spin_unlock(&mchdev_lock); - - return; -} - -static void vlv_c0_read(struct drm_i915_private *dev_priv, - struct intel_rps_ei *ei) -{ - ei->ktime = ktime_get_raw(); - ei->render_c0 = I915_READ(VLV_RENDER_C0_COUNT); - ei->media_c0 = I915_READ(VLV_MEDIA_C0_COUNT); -} - -void gen6_rps_reset_ei(struct drm_i915_private *dev_priv) -{ - memset(&dev_priv->gt_pm.rps.ei, 0, sizeof(dev_priv->gt_pm.rps.ei)); -} - -static u32 vlv_wa_c0_ei(struct drm_i915_private *dev_priv, u32 pm_iir) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - const struct intel_rps_ei *prev = &rps->ei; - struct intel_rps_ei now; - u32 events = 0; - - if ((pm_iir & GEN6_PM_RP_UP_EI_EXPIRED) == 0) - return 0; - - vlv_c0_read(dev_priv, &now); - - if (prev->ktime) { - u64 time, c0; - u32 render, media; - - time = ktime_us_delta(now.ktime, prev->ktime); - - time *= dev_priv->czclk_freq; - - /* Workload can be split between render + media, - * e.g. SwapBuffers being blitted in X after being rendered in - * mesa. To account for this we need to combine both engines - * into our activity counter. - */ - render = now.render_c0 - prev->render_c0; - media = now.media_c0 - prev->media_c0; - c0 = max(render, media); - c0 *= 1000 * 100 << 8; /* to usecs and scale to threshold% */ - - if (c0 > time * rps->power.up_threshold) - events = GEN6_PM_RP_UP_THRESHOLD; - else if (c0 < time * rps->power.down_threshold) - events = GEN6_PM_RP_DOWN_THRESHOLD; - } - - rps->ei = now; - return events; -} - -static void gen6_pm_rps_work(struct work_struct *work) -{ - struct drm_i915_private *dev_priv = - container_of(work, struct drm_i915_private, gt_pm.rps.work); - struct intel_gt *gt = &dev_priv->gt; - struct intel_rps *rps = &dev_priv->gt_pm.rps; - bool client_boost = false; - int new_delay, adj, min, max; - u32 pm_iir = 0; - - spin_lock_irq(>->irq_lock); - if (rps->interrupts_enabled) { - pm_iir = fetch_and_zero(&rps->pm_iir); - client_boost = atomic_read(&rps->num_waiters); - } - spin_unlock_irq(>->irq_lock); - - /* Make sure we didn't queue anything we're not going to process. */ - WARN_ON(pm_iir & ~dev_priv->pm_rps_events); - if ((pm_iir & dev_priv->pm_rps_events) == 0 && !client_boost) - goto out; - - mutex_lock(&rps->lock); - - pm_iir |= vlv_wa_c0_ei(dev_priv, pm_iir); - - adj = rps->last_adj; - new_delay = rps->cur_freq; - min = rps->min_freq_softlimit; - max = rps->max_freq_softlimit; - if (client_boost) - max = rps->max_freq; - if (client_boost && new_delay < rps->boost_freq) { - new_delay = rps->boost_freq; - adj = 0; - } else if (pm_iir & GEN6_PM_RP_UP_THRESHOLD) { - if (adj > 0) - adj *= 2; - else /* CHV needs even encode values */ - adj = IS_CHERRYVIEW(dev_priv) ? 2 : 1; - - if (new_delay >= rps->max_freq_softlimit) - adj = 0; - } else if (client_boost) { - adj = 0; - } else if (pm_iir & GEN6_PM_RP_DOWN_TIMEOUT) { - if (rps->cur_freq > rps->efficient_freq) - new_delay = rps->efficient_freq; - else if (rps->cur_freq > rps->min_freq_softlimit) - new_delay = rps->min_freq_softlimit; - adj = 0; - } else if (pm_iir & GEN6_PM_RP_DOWN_THRESHOLD) { - if (adj < 0) - adj *= 2; - else /* CHV needs even encode values */ - adj = IS_CHERRYVIEW(dev_priv) ? -2 : -1; - - if (new_delay <= rps->min_freq_softlimit) - adj = 0; - } else { /* unknown event */ - adj = 0; - } - - rps->last_adj = adj; - - /* - * Limit deboosting and boosting to keep ourselves at the extremes - * when in the respective power modes (i.e. slowly decrease frequencies - * while in the HIGH_POWER zone and slowly increase frequencies while - * in the LOW_POWER zone). On idle, we will hit the timeout and drop - * to the next level quickly, and conversely if busy we expect to - * hit a waitboost and rapidly switch into max power. - */ - if ((adj < 0 && rps->power.mode == HIGH_POWER) || - (adj > 0 && rps->power.mode == LOW_POWER)) - rps->last_adj = 0; - - /* sysfs frequency interfaces may have snuck in while servicing the - * interrupt - */ - new_delay += adj; - new_delay = clamp_t(int, new_delay, min, max); - - if (intel_set_rps(dev_priv, new_delay)) { - DRM_DEBUG_DRIVER("Failed to set new GPU frequency\n"); - rps->last_adj = 0; - } - - mutex_unlock(&rps->lock); - -out: - /* Make sure not to corrupt PMIMR state used by ringbuffer on GEN6 */ - spin_lock_irq(>->irq_lock); - if (rps->interrupts_enabled) - gen6_gt_pm_unmask_irq(gt, dev_priv->pm_rps_events); - spin_unlock_irq(>->irq_lock); -} - - /** * ivybridge_parity_work - Workqueue called when a parity error interrupt * occurred. @@ -1401,11 +1028,11 @@ static bool icp_ddi_port_hotplug_long_detect(enum hpd_pin pin, u32 val) { switch (pin) { case HPD_PORT_A: - return val & ICP_DDIA_HPD_LONG_DETECT; + return val & SHOTPLUG_CTL_DDI_HPD_LONG_DETECT(PORT_A); case HPD_PORT_B: - return val & ICP_DDIB_HPD_LONG_DETECT; + return val & SHOTPLUG_CTL_DDI_HPD_LONG_DETECT(PORT_B); case HPD_PORT_C: - return val & TGP_DDIC_HPD_LONG_DETECT; + return val & SHOTPLUG_CTL_DDI_HPD_LONG_DETECT(PORT_C); default: return false; } @@ -1427,20 +1054,6 @@ static bool icp_tc_port_hotplug_long_detect(enum hpd_pin pin, u32 val) } } -static bool tgp_ddi_port_hotplug_long_detect(enum hpd_pin pin, u32 val) -{ - switch (pin) { - case HPD_PORT_A: - return val & ICP_DDIA_HPD_LONG_DETECT; - case HPD_PORT_B: - return val & ICP_DDIB_HPD_LONG_DETECT; - case HPD_PORT_C: - return val & TGP_DDIC_HPD_LONG_DETECT; - default: - return false; - } -} - static bool tgp_tc_port_hotplug_long_detect(enum hpd_pin pin, u32 val) { switch (pin) { @@ -1652,54 +1265,6 @@ static void i9xx_pipe_crc_irq_handler(struct drm_i915_private *dev_priv, res1, res2); } -/* The RPS events need forcewake, so we add them to a work queue and mask their - * IMR bits until the work is done. Other interrupts can be processed without - * the work queue. */ -void gen11_rps_irq_handler(struct intel_gt *gt, u32 pm_iir) -{ - struct drm_i915_private *i915 = gt->i915; - struct intel_rps *rps = &i915->gt_pm.rps; - const u32 events = i915->pm_rps_events & pm_iir; - - lockdep_assert_held(>->irq_lock); - - if (unlikely(!events)) - return; - - gen6_gt_pm_mask_irq(gt, events); - - if (!rps->interrupts_enabled) - return; - - rps->pm_iir |= events; - schedule_work(&rps->work); -} - -void gen6_rps_irq_handler(struct drm_i915_private *dev_priv, u32 pm_iir) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - struct intel_gt *gt = &dev_priv->gt; - - if (pm_iir & dev_priv->pm_rps_events) { - spin_lock(>->irq_lock); - gen6_gt_pm_mask_irq(gt, pm_iir & dev_priv->pm_rps_events); - if (rps->interrupts_enabled) { - rps->pm_iir |= pm_iir & dev_priv->pm_rps_events; - schedule_work(&rps->work); - } - spin_unlock(>->irq_lock); - } - - if (INTEL_GEN(dev_priv) >= 8) - return; - - if (pm_iir & PM_VEBOX_USER_INTERRUPT) - intel_engine_breadcrumbs_irq(dev_priv->engine[VECS0]); - - if (pm_iir & PM_VEBOX_CS_ERROR_INTERRUPT) - DRM_DEBUG("Command parser error, pm_iir 0x%08x\n", pm_iir); -} - static void i9xx_pipestat_irq_reset(struct drm_i915_private *dev_priv) { enum pipe pipe; @@ -1716,7 +1281,7 @@ static void i9xx_pipestat_irq_reset(struct drm_i915_private *dev_priv) static void i9xx_pipestat_irq_ack(struct drm_i915_private *dev_priv, u32 iir, u32 pipe_stats[I915_MAX_PIPES]) { - int pipe; + enum pipe pipe; spin_lock(&dev_priv->irq_lock); @@ -1741,6 +1306,7 @@ static void i9xx_pipestat_irq_ack(struct drm_i915_private *dev_priv, status_mask = PIPE_FIFO_UNDERRUN_STATUS; switch (pipe) { + default: case PIPE_A: iir_bit = I915_DISPLAY_PIPE_A_EVENT_INTERRUPT; break; @@ -2009,7 +1575,7 @@ static irqreturn_t valleyview_irq_handler(int irq, void *arg) if (gt_iir) gen6_gt_irq_handler(&dev_priv->gt, gt_iir); if (pm_iir) - gen6_rps_irq_handler(dev_priv, pm_iir); + gen6_rps_irq_handler(&dev_priv->gt.rps, pm_iir); if (hotplug_status) i9xx_hpd_irq_handler(dev_priv, hotplug_status); @@ -2136,7 +1702,7 @@ static void ibx_hpd_irq_handler(struct drm_i915_private *dev_priv, static void ibx_irq_handler(struct drm_i915_private *dev_priv, u32 pch_iir) { - int pipe; + enum pipe pipe; u32 hotplug_trigger = pch_iir & SDE_HOTPLUG_MASK; ibx_hpd_irq_handler(dev_priv, hotplug_trigger, hpd_ibx); @@ -2222,7 +1788,7 @@ static void cpt_serr_int_handler(struct drm_i915_private *dev_priv) static void cpt_irq_handler(struct drm_i915_private *dev_priv, u32 pch_iir) { - int pipe; + enum pipe pipe; u32 hotplug_trigger = pch_iir & SDE_HOTPLUG_MASK_CPT; ibx_hpd_irq_handler(dev_priv, hotplug_trigger, hpd_cpt); @@ -2256,19 +1822,35 @@ static void cpt_irq_handler(struct drm_i915_private *dev_priv, u32 pch_iir) cpt_serr_int_handler(dev_priv); } -static void icp_irq_handler(struct drm_i915_private *dev_priv, u32 pch_iir, - const u32 *pins) +static void icp_irq_handler(struct drm_i915_private *dev_priv, u32 pch_iir) { - u32 ddi_hotplug_trigger; - u32 tc_hotplug_trigger; + u32 ddi_hotplug_trigger, tc_hotplug_trigger; u32 pin_mask = 0, long_mask = 0; + bool (*tc_port_hotplug_long_detect)(enum hpd_pin pin, u32 val); + const u32 *pins; - if (HAS_PCH_MCC(dev_priv)) { + if (HAS_PCH_TGP(dev_priv)) { + ddi_hotplug_trigger = pch_iir & SDE_DDI_MASK_TGP; + tc_hotplug_trigger = pch_iir & SDE_TC_MASK_TGP; + tc_port_hotplug_long_detect = tgp_tc_port_hotplug_long_detect; + pins = hpd_tgp; + } else if (HAS_PCH_JSP(dev_priv)) { ddi_hotplug_trigger = pch_iir & SDE_DDI_MASK_TGP; tc_hotplug_trigger = 0; + pins = hpd_tgp; + } else if (HAS_PCH_MCC(dev_priv)) { + ddi_hotplug_trigger = pch_iir & SDE_DDI_MASK_ICP; + tc_hotplug_trigger = pch_iir & SDE_TC_HOTPLUG_ICP(PORT_TC1); + tc_port_hotplug_long_detect = icp_tc_port_hotplug_long_detect; + pins = hpd_icp; } else { + WARN(!HAS_PCH_ICP(dev_priv), + "Unrecognized PCH type 0x%x\n", INTEL_PCH_TYPE(dev_priv)); + ddi_hotplug_trigger = pch_iir & SDE_DDI_MASK_ICP; tc_hotplug_trigger = pch_iir & SDE_TC_MASK_ICP; + tc_port_hotplug_long_detect = icp_tc_port_hotplug_long_detect; + pins = hpd_icp; } if (ddi_hotplug_trigger) { @@ -2292,44 +1874,7 @@ static void icp_irq_handler(struct drm_i915_private *dev_priv, u32 pch_iir, intel_get_hpd_pins(dev_priv, &pin_mask, &long_mask, tc_hotplug_trigger, dig_hotplug_reg, pins, - icp_tc_port_hotplug_long_detect); - } - - if (pin_mask) - intel_hpd_irq_handler(dev_priv, pin_mask, long_mask); - - if (pch_iir & SDE_GMBUS_ICP) - gmbus_irq_handler(dev_priv); -} - -static void tgp_irq_handler(struct drm_i915_private *dev_priv, u32 pch_iir) -{ - u32 ddi_hotplug_trigger = pch_iir & SDE_DDI_MASK_TGP; - u32 tc_hotplug_trigger = pch_iir & SDE_TC_MASK_TGP; - u32 pin_mask = 0, long_mask = 0; - - if (ddi_hotplug_trigger) { - u32 dig_hotplug_reg; - - dig_hotplug_reg = I915_READ(SHOTPLUG_CTL_DDI); - I915_WRITE(SHOTPLUG_CTL_DDI, dig_hotplug_reg); - - intel_get_hpd_pins(dev_priv, &pin_mask, &long_mask, - ddi_hotplug_trigger, - dig_hotplug_reg, hpd_tgp, - tgp_ddi_port_hotplug_long_detect); - } - - if (tc_hotplug_trigger) { - u32 dig_hotplug_reg; - - dig_hotplug_reg = I915_READ(SHOTPLUG_CTL_TC); - I915_WRITE(SHOTPLUG_CTL_TC, dig_hotplug_reg); - - intel_get_hpd_pins(dev_priv, &pin_mask, &long_mask, - tc_hotplug_trigger, - dig_hotplug_reg, hpd_tgp, - tgp_tc_port_hotplug_long_detect); + tc_port_hotplug_long_detect); } if (pin_mask) @@ -2434,7 +1979,7 @@ static void ilk_display_irq_handler(struct drm_i915_private *dev_priv, } if (IS_GEN(dev_priv, 5) && de_iir & DE_PCU_EVENT) - ironlake_rps_change_irq_handler(dev_priv); + gen5_rps_irq_handler(&dev_priv->gt.rps); } static void ivb_display_irq_handler(struct drm_i915_private *dev_priv, @@ -2539,7 +2084,7 @@ static irqreturn_t ironlake_irq_handler(int irq, void *arg) if (pm_iir) { I915_WRITE(GEN6_PMIIR, pm_iir); ret = IRQ_HANDLED; - gen6_rps_irq_handler(dev_priv, pm_iir); + gen6_rps_irq_handler(&dev_priv->gt.rps, pm_iir); } } @@ -2616,10 +2161,16 @@ static u32 gen8_de_port_aux_mask(struct drm_i915_private *dev_priv) u32 mask; if (INTEL_GEN(dev_priv) >= 12) - /* TODO: Add AUX entries for USBC */ return TGL_DE_PORT_AUX_DDIA | TGL_DE_PORT_AUX_DDIB | - TGL_DE_PORT_AUX_DDIC; + TGL_DE_PORT_AUX_DDIC | + TGL_DE_PORT_AUX_USBC1 | + TGL_DE_PORT_AUX_USBC2 | + TGL_DE_PORT_AUX_USBC3 | + TGL_DE_PORT_AUX_USBC4 | + TGL_DE_PORT_AUX_USBC5 | + TGL_DE_PORT_AUX_USBC6; + mask = GEN8_AUX_CHANNEL_A; if (INTEL_GEN(dev_priv) >= 9) @@ -2638,7 +2189,9 @@ static u32 gen8_de_port_aux_mask(struct drm_i915_private *dev_priv) static u32 gen8_de_pipe_fault_mask(struct drm_i915_private *dev_priv) { - if (INTEL_GEN(dev_priv) >= 9) + if (INTEL_GEN(dev_priv) >= 11) + return GEN11_DE_PIPE_IRQ_FAULT_ERRORS; + else if (INTEL_GEN(dev_priv) >= 9) return GEN9_DE_PIPE_IRQ_FAULT_ERRORS; else return GEN8_DE_PIPE_IRQ_FAULT_ERRORS; @@ -2655,11 +2208,21 @@ gen8_de_misc_irq_handler(struct drm_i915_private *dev_priv, u32 iir) } if (iir & GEN8_DE_EDP_PSR) { - u32 psr_iir = I915_READ(EDP_PSR_IIR); + u32 psr_iir; + i915_reg_t iir_reg; + + if (INTEL_GEN(dev_priv) >= 12) + iir_reg = TRANS_PSR_IIR(dev_priv->psr.transcoder); + else + iir_reg = EDP_PSR_IIR; + + psr_iir = I915_READ(iir_reg); + I915_WRITE(iir_reg, psr_iir); + + if (psr_iir) + found = true; intel_psr_irq_handler(dev_priv, psr_iir); - I915_WRITE(EDP_PSR_IIR, psr_iir); - found = true; } if (!found) @@ -2780,12 +2343,8 @@ gen8_de_irq_handler(struct drm_i915_private *dev_priv, u32 master_ctl) I915_WRITE(SDEIIR, iir); ret = IRQ_HANDLED; - if (INTEL_PCH_TYPE(dev_priv) >= PCH_TGP) - tgp_irq_handler(dev_priv, iir); - else if (INTEL_PCH_TYPE(dev_priv) >= PCH_MCC) - icp_irq_handler(dev_priv, iir, hpd_mcc); - else if (INTEL_PCH_TYPE(dev_priv) >= PCH_ICP) - icp_irq_handler(dev_priv, iir, hpd_icp); + if (INTEL_PCH_TYPE(dev_priv) >= PCH_ICP) + icp_irq_handler(dev_priv, iir); else if (INTEL_PCH_TYPE(dev_priv) >= PCH_SPT) spt_irq_handler(dev_priv, iir); else @@ -2894,9 +2453,11 @@ static inline void gen11_master_intr_enable(void __iomem * const regs) raw_reg_write(regs, GEN11_GFX_MSTR_IRQ, GEN11_MASTER_IRQ); } -static irqreturn_t gen11_irq_handler(int irq, void *arg) +static __always_inline irqreturn_t +__gen11_irq_handler(struct drm_i915_private * const i915, + u32 (*intr_disable)(void __iomem * const regs), + void (*intr_enable)(void __iomem * const regs)) { - struct drm_i915_private * const i915 = arg; void __iomem * const regs = i915->uncore.regs; struct intel_gt *gt = &i915->gt; u32 master_ctl; @@ -2905,9 +2466,9 @@ static irqreturn_t gen11_irq_handler(int irq, void *arg) if (!intel_irqs_enabled(i915)) return IRQ_NONE; - master_ctl = gen11_master_intr_disable(regs); + master_ctl = intr_disable(regs); if (!master_ctl) { - gen11_master_intr_enable(regs); + intr_enable(regs); return IRQ_NONE; } @@ -2929,13 +2490,20 @@ static irqreturn_t gen11_irq_handler(int irq, void *arg) gu_misc_iir = gen11_gu_misc_irq_ack(gt, master_ctl); - gen11_master_intr_enable(regs); + intr_enable(regs); gen11_gu_misc_irq_handler(gt, gu_misc_iir); return IRQ_HANDLED; } +static irqreturn_t gen11_irq_handler(int irq, void *arg) +{ + return __gen11_irq_handler(arg, + gen11_master_intr_disable, + gen11_master_intr_enable); +} + /* Called from drm generic code, passed 'crtc' which * we use as a pipe index */ @@ -2952,12 +2520,18 @@ int i8xx_enable_vblank(struct drm_crtc *crtc) return 0; } -int i945gm_enable_vblank(struct drm_crtc *crtc) +int i915gm_enable_vblank(struct drm_crtc *crtc) { struct drm_i915_private *dev_priv = to_i915(crtc->dev); - if (dev_priv->i945gm_vblank.enabled++ == 0) - schedule_work(&dev_priv->i945gm_vblank.work); + /* + * Vblank interrupts fail to wake the device up from C2+. + * Disabling render clock gating during C-states avoids + * the problem. There is a small power cost so we do this + * only when vblank interrupts are actually enabled. + */ + if (dev_priv->vblank_enabled++ == 0) + I915_WRITE(SCPD0, _MASKED_BIT_ENABLE(CSTATE_RENDER_CLOCK_GATE_DISABLE)); return i8xx_enable_vblank(crtc); } @@ -3030,14 +2604,14 @@ void i8xx_disable_vblank(struct drm_crtc *crtc) spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); } -void i945gm_disable_vblank(struct drm_crtc *crtc) +void i915gm_disable_vblank(struct drm_crtc *crtc) { struct drm_i915_private *dev_priv = to_i915(crtc->dev); i8xx_disable_vblank(crtc); - if (--dev_priv->i945gm_vblank.enabled == 0) - schedule_work(&dev_priv->i945gm_vblank.work); + if (--dev_priv->vblank_enabled == 0) + I915_WRITE(SCPD0, _MASKED_BIT_DISABLE(CSTATE_RENDER_CLOCK_GATE_DISABLE)); } void i965_disable_vblank(struct drm_crtc *crtc) @@ -3076,60 +2650,6 @@ void bdw_disable_vblank(struct drm_crtc *crtc) spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags); } -static void i945gm_vblank_work_func(struct work_struct *work) -{ - struct drm_i915_private *dev_priv = - container_of(work, struct drm_i915_private, i945gm_vblank.work); - - /* - * Vblank interrupts fail to wake up the device from C3, - * hence we want to prevent C3 usage while vblank interrupts - * are enabled. - */ - pm_qos_update_request(&dev_priv->i945gm_vblank.pm_qos, - READ_ONCE(dev_priv->i945gm_vblank.enabled) ? - dev_priv->i945gm_vblank.c3_disable_latency : - PM_QOS_DEFAULT_VALUE); -} - -static int cstate_disable_latency(const char *name) -{ - const struct cpuidle_driver *drv; - int i; - - drv = cpuidle_get_driver(); - if (!drv) - return 0; - - for (i = 0; i < drv->state_count; i++) { - const struct cpuidle_state *state = &drv->states[i]; - - if (!strcmp(state->name, name)) - return state->exit_latency ? - state->exit_latency - 1 : 0; - } - - return 0; -} - -static void i945gm_vblank_work_init(struct drm_i915_private *dev_priv) -{ - INIT_WORK(&dev_priv->i945gm_vblank.work, - i945gm_vblank_work_func); - - dev_priv->i945gm_vblank.c3_disable_latency = - cstate_disable_latency("C3"); - pm_qos_add_request(&dev_priv->i945gm_vblank.pm_qos, - PM_QOS_CPU_DMA_LATENCY, - PM_QOS_DEFAULT_VALUE); -} - -static void i945gm_vblank_work_fini(struct drm_i915_private *dev_priv) -{ - cancel_work_sync(&dev_priv->i945gm_vblank.work); - pm_qos_remove_request(&dev_priv->i945gm_vblank.pm_qos); -} - static void ibx_irq_reset(struct drm_i915_private *dev_priv) { struct intel_uncore *uncore = &dev_priv->uncore; @@ -3246,7 +2766,7 @@ static void valleyview_irq_reset(struct drm_i915_private *dev_priv) static void gen8_irq_reset(struct drm_i915_private *dev_priv) { struct intel_uncore *uncore = &dev_priv->uncore; - int pipe; + enum pipe pipe; gen8_master_intr_disable(dev_priv->uncore.regs); @@ -3271,7 +2791,7 @@ static void gen8_irq_reset(struct drm_i915_private *dev_priv) static void gen11_irq_reset(struct drm_i915_private *dev_priv) { struct intel_uncore *uncore = &dev_priv->uncore; - int pipe; + enum pipe pipe; gen11_master_intr_disable(dev_priv->uncore.regs); @@ -3279,8 +2799,23 @@ static void gen11_irq_reset(struct drm_i915_private *dev_priv) intel_uncore_write(uncore, GEN11_DISPLAY_INT_CTL, 0); - intel_uncore_write(uncore, EDP_PSR_IMR, 0xffffffff); - intel_uncore_write(uncore, EDP_PSR_IIR, 0xffffffff); + if (INTEL_GEN(dev_priv) >= 12) { + enum transcoder trans; + + for (trans = TRANSCODER_A; trans <= TRANSCODER_D; trans++) { + enum intel_display_power_domain domain; + + domain = POWER_DOMAIN_TRANSCODER(trans); + if (!intel_display_power_is_enabled(dev_priv, domain)) + continue; + + intel_uncore_write(uncore, TRANS_PSR_IMR(trans), 0xffffffff); + intel_uncore_write(uncore, TRANS_PSR_IIR(trans), 0xffffffff); + } + } else { + intel_uncore_write(uncore, EDP_PSR_IMR, 0xffffffff); + intel_uncore_write(uncore, EDP_PSR_IIR, 0xffffffff); + } for_each_pipe(dev_priv, pipe) if (intel_display_power_is_enabled(dev_priv, @@ -3431,42 +2966,44 @@ static void icp_hpd_detection_setup(struct drm_i915_private *dev_priv, } } -static void icp_hpd_irq_setup(struct drm_i915_private *dev_priv) +static void icp_hpd_irq_setup(struct drm_i915_private *dev_priv, + u32 sde_ddi_mask, u32 sde_tc_mask, + u32 ddi_enable_mask, u32 tc_enable_mask, + const u32 *pins) { u32 hotplug_irqs, enabled_irqs; - hotplug_irqs = SDE_DDI_MASK_ICP | SDE_TC_MASK_ICP; - enabled_irqs = intel_hpd_enabled_irqs(dev_priv, hpd_icp); + hotplug_irqs = sde_ddi_mask | sde_tc_mask; + enabled_irqs = intel_hpd_enabled_irqs(dev_priv, pins); ibx_display_interrupt_update(dev_priv, hotplug_irqs, enabled_irqs); - icp_hpd_detection_setup(dev_priv, ICP_DDI_HPD_ENABLE_MASK, - ICP_TC_HPD_ENABLE_MASK); + icp_hpd_detection_setup(dev_priv, ddi_enable_mask, tc_enable_mask); } +/* + * EHL doesn't need most of gen11_hpd_irq_setup, it's handling only the + * equivalent of SDE. + */ static void mcc_hpd_irq_setup(struct drm_i915_private *dev_priv) { - u32 hotplug_irqs, enabled_irqs; - - hotplug_irqs = SDE_DDI_MASK_TGP; - enabled_irqs = intel_hpd_enabled_irqs(dev_priv, hpd_mcc); - - ibx_display_interrupt_update(dev_priv, hotplug_irqs, enabled_irqs); - - icp_hpd_detection_setup(dev_priv, TGP_DDI_HPD_ENABLE_MASK, 0); + icp_hpd_irq_setup(dev_priv, + SDE_DDI_MASK_ICP, SDE_TC_HOTPLUG_ICP(PORT_TC1), + ICP_DDI_HPD_ENABLE_MASK, ICP_TC_HPD_ENABLE(PORT_TC1), + hpd_icp); } -static void tgp_hpd_irq_setup(struct drm_i915_private *dev_priv) +/* + * JSP behaves exactly the same as MCC above except that port C is mapped to + * the DDI-C pins instead of the TC1 pins. This means we should follow TGP's + * masks & tables rather than ICP's masks & tables. + */ +static void jsp_hpd_irq_setup(struct drm_i915_private *dev_priv) { - u32 hotplug_irqs, enabled_irqs; - - hotplug_irqs = SDE_DDI_MASK_TGP | SDE_TC_MASK_TGP; - enabled_irqs = intel_hpd_enabled_irqs(dev_priv, hpd_tgp); - - ibx_display_interrupt_update(dev_priv, hotplug_irqs, enabled_irqs); - - icp_hpd_detection_setup(dev_priv, TGP_DDI_HPD_ENABLE_MASK, - TGP_TC_HPD_ENABLE_MASK); + icp_hpd_irq_setup(dev_priv, + SDE_DDI_MASK_TGP, 0, + TGP_DDI_HPD_ENABLE_MASK, 0, + hpd_tgp); } static void gen11_hpd_detection_setup(struct drm_i915_private *dev_priv) @@ -3506,9 +3043,13 @@ static void gen11_hpd_irq_setup(struct drm_i915_private *dev_priv) gen11_hpd_detection_setup(dev_priv); if (INTEL_PCH_TYPE(dev_priv) >= PCH_TGP) - tgp_hpd_irq_setup(dev_priv); + icp_hpd_irq_setup(dev_priv, SDE_DDI_MASK_TGP, SDE_TC_MASK_TGP, + TGP_DDI_HPD_ENABLE_MASK, + TGP_TC_HPD_ENABLE_MASK, hpd_tgp); else if (INTEL_PCH_TYPE(dev_priv) >= PCH_ICP) - icp_hpd_irq_setup(dev_priv); + icp_hpd_irq_setup(dev_priv, SDE_DDI_MASK_ICP, SDE_TC_MASK_ICP, + ICP_DDI_HPD_ENABLE_MASK, + ICP_TC_HPD_ENABLE_MASK, hpd_icp); } static void spt_hpd_detection_setup(struct drm_i915_private *dev_priv) @@ -3684,7 +3225,6 @@ static void ironlake_irq_postinstall(struct drm_i915_private *dev_priv) if (IS_HASWELL(dev_priv)) { gen3_assert_iir_is_zero(uncore, EDP_PSR_IIR); - intel_psr_irq_control(dev_priv, dev_priv->psr.debug); display_mask |= DE_EDP_PSR_INT_HSW; } @@ -3794,8 +3334,21 @@ static void gen8_de_irq_postinstall(struct drm_i915_private *dev_priv) else if (IS_BROADWELL(dev_priv)) de_port_enables |= GEN8_PORT_DP_A_HOTPLUG; - gen3_assert_iir_is_zero(uncore, EDP_PSR_IIR); - intel_psr_irq_control(dev_priv, dev_priv->psr.debug); + if (INTEL_GEN(dev_priv) >= 12) { + enum transcoder trans; + + for (trans = TRANSCODER_A; trans <= TRANSCODER_D; trans++) { + enum intel_display_power_domain domain; + + domain = POWER_DOMAIN_TRANSCODER(trans); + if (!intel_display_power_is_enabled(dev_priv, domain)) + continue; + + gen3_assert_iir_is_zero(uncore, TRANS_PSR_IIR(trans)); + } + } else { + gen3_assert_iir_is_zero(uncore, EDP_PSR_IIR); + } for_each_pipe(dev_priv, pipe) { dev_priv->de_irq_mask[pipe] = ~de_pipe_masked; @@ -3853,8 +3406,11 @@ static void icp_irq_postinstall(struct drm_i915_private *dev_priv) if (HAS_PCH_TGP(dev_priv)) icp_hpd_detection_setup(dev_priv, TGP_DDI_HPD_ENABLE_MASK, TGP_TC_HPD_ENABLE_MASK); - else if (HAS_PCH_MCC(dev_priv)) + else if (HAS_PCH_JSP(dev_priv)) icp_hpd_detection_setup(dev_priv, TGP_DDI_HPD_ENABLE_MASK, 0); + else if (HAS_PCH_MCC(dev_priv)) + icp_hpd_detection_setup(dev_priv, ICP_DDI_HPD_ENABLE_MASK, + ICP_TC_HPD_ENABLE(PORT_TC1)); else icp_hpd_detection_setup(dev_priv, ICP_DDI_HPD_ENABLE_MASK, ICP_TC_HPD_ENABLE_MASK); @@ -4317,16 +3873,10 @@ static irqreturn_t i965_irq_handler(int irq, void *arg) void intel_irq_init(struct drm_i915_private *dev_priv) { struct drm_device *dev = &dev_priv->drm; - struct intel_rps *rps = &dev_priv->gt_pm.rps; int i; - if (IS_I945GM(dev_priv)) - i945gm_vblank_work_init(dev_priv); - intel_hpd_init_work(dev_priv); - INIT_WORK(&rps->work, gen6_pm_rps_work); - INIT_WORK(&dev_priv->l3_parity.error_work, ivybridge_parity_work); for (i = 0; i < MAX_L3_SLICES; ++i) dev_priv->l3_parity.remap_info[i] = NULL; @@ -4335,33 +3885,6 @@ void intel_irq_init(struct drm_i915_private *dev_priv) if (HAS_GT_UC(dev_priv) && INTEL_GEN(dev_priv) < 11) dev_priv->gt.pm_guc_events = GUC_INTR_GUC2HOST << 16; - /* Let's track the enabled rps events */ - if (IS_VALLEYVIEW(dev_priv)) - /* WaGsvRC0ResidencyMethod:vlv */ - dev_priv->pm_rps_events = GEN6_PM_RP_UP_EI_EXPIRED; - else - dev_priv->pm_rps_events = (GEN6_PM_RP_UP_THRESHOLD | - GEN6_PM_RP_DOWN_THRESHOLD | - GEN6_PM_RP_DOWN_TIMEOUT); - - /* We share the register with other engine */ - if (INTEL_GEN(dev_priv) > 9) - GEM_WARN_ON(dev_priv->pm_rps_events & 0xffff0000); - - rps->pm_intrmsk_mbz = 0; - - /* - * SNB,IVB,HSW can while VLV,CHV may hard hang on looping batchbuffer - * if GEN6_PM_UP_EI_EXPIRED is masked. - * - * TODO: verify if this can be reproduced on VLV,CHV. - */ - if (INTEL_GEN(dev_priv) <= 7) - rps->pm_intrmsk_mbz |= GEN6_PM_RP_UP_EI_EXPIRED; - - if (INTEL_GEN(dev_priv) >= 8) - rps->pm_intrmsk_mbz |= GEN8_PMINTR_DISABLE_REDIRECT_TO_GUC; - dev->vblank_disable_immediate = true; /* Most platforms treat the display irq block as an always-on @@ -4387,8 +3910,9 @@ void intel_irq_init(struct drm_i915_private *dev_priv) if (I915_HAS_HOTPLUG(dev_priv)) dev_priv->display.hpd_irq_setup = i915_hpd_irq_setup; } else { - if (HAS_PCH_MCC(dev_priv)) - /* EHL doesn't need most of gen11_hpd_irq_setup */ + if (HAS_PCH_JSP(dev_priv)) + dev_priv->display.hpd_irq_setup = jsp_hpd_irq_setup; + else if (HAS_PCH_MCC(dev_priv)) dev_priv->display.hpd_irq_setup = mcc_hpd_irq_setup; else if (INTEL_GEN(dev_priv) >= 11) dev_priv->display.hpd_irq_setup = gen11_hpd_irq_setup; @@ -4411,9 +3935,6 @@ void intel_irq_fini(struct drm_i915_private *i915) { int i; - if (IS_I945GM(i915)) - i945gm_vblank_work_fini(i915); - for (i = 0; i < MAX_L3_SLICES; ++i) kfree(i915->l3_parity.remap_info[i]); } @@ -4538,10 +4059,10 @@ void intel_irq_uninstall(struct drm_i915_private *dev_priv) int irq = dev_priv->drm.pdev->irq; /* - * FIXME we can get called twice during driver load - * error handling due to intel_modeset_cleanup() - * calling us out of sequence. Would be nice if - * it didn't do that... + * FIXME we can get called twice during driver probe + * error handling as well as during driver remove due to + * intel_modeset_driver_remove() calling us out of sequence. + * Would be nice if it didn't do that... */ if (!dev_priv->drm.irq_enabled) return; diff --git a/drivers/gpu/drm/i915/i915_irq.h b/drivers/gpu/drm/i915/i915_irq.h index 8e7e6071777e..812c47a9c2d6 100644 --- a/drivers/gpu/drm/i915/i915_irq.h +++ b/drivers/gpu/drm/i915/i915_irq.h @@ -17,14 +17,8 @@ struct drm_device; struct drm_display_mode; struct drm_i915_private; struct intel_crtc; -struct intel_crtc; -struct intel_gt; -struct intel_guc; struct intel_uncore; -void gen11_rps_irq_handler(struct intel_gt *gt, u32 pm_iir); -void gen6_rps_irq_handler(struct drm_i915_private *dev_priv, u32 pm_iir); - void intel_irq_init(struct drm_i915_private *dev_priv); void intel_irq_fini(struct drm_i915_private *dev_priv); int intel_irq_install(struct drm_i915_private *dev_priv); @@ -106,12 +100,6 @@ void gen8_irq_power_well_post_enable(struct drm_i915_private *dev_priv, u8 pipe_mask); void gen8_irq_power_well_pre_disable(struct drm_i915_private *dev_priv, u8 pipe_mask); -void gen9_reset_guc_interrupts(struct intel_guc *guc); -void gen9_enable_guc_interrupts(struct intel_guc *guc); -void gen9_disable_guc_interrupts(struct intel_guc *guc); -void gen11_reset_guc_interrupts(struct intel_guc *guc); -void gen11_enable_guc_interrupts(struct intel_guc *guc); -void gen11_disable_guc_interrupts(struct intel_guc *guc); bool i915_get_crtc_scanoutpos(struct drm_device *dev, unsigned int pipe, bool in_vblank_irq, int *vpos, int *hpos, @@ -122,12 +110,12 @@ u32 i915_get_vblank_counter(struct drm_crtc *crtc); u32 g4x_get_vblank_counter(struct drm_crtc *crtc); int i8xx_enable_vblank(struct drm_crtc *crtc); -int i945gm_enable_vblank(struct drm_crtc *crtc); +int i915gm_enable_vblank(struct drm_crtc *crtc); int i965_enable_vblank(struct drm_crtc *crtc); int ilk_enable_vblank(struct drm_crtc *crtc); int bdw_enable_vblank(struct drm_crtc *crtc); void i8xx_disable_vblank(struct drm_crtc *crtc); -void i945gm_disable_vblank(struct drm_crtc *crtc); +void i915gm_disable_vblank(struct drm_crtc *crtc); void i965_disable_vblank(struct drm_crtc *crtc); void ilk_disable_vblank(struct drm_crtc *crtc); void bdw_disable_vblank(struct drm_crtc *crtc); diff --git a/drivers/gpu/drm/i915/i915_params.c b/drivers/gpu/drm/i915/i915_params.c index 296452f9efe4..1dd1f3652795 100644 --- a/drivers/gpu/drm/i915/i915_params.c +++ b/drivers/gpu/drm/i915/i915_params.c @@ -46,7 +46,8 @@ i915_param_named(modeset, int, 0400, i915_param_named_unsafe(enable_dc, int, 0400, "Enable power-saving display C-states. " - "(-1=auto [default]; 0=disable; 1=up to DC5; 2=up to DC6)"); + "(-1=auto [default]; 0=disable; 1=up to DC5; 2=up to DC6; " + "3=up to DC5 with DC3CO; 4=up to DC6 with DC3CO)"); i915_param_named_unsafe(enable_fbc, int, 0600, "Enable frame buffer compression for power savings " @@ -165,7 +166,7 @@ i915_param_named_unsafe(enable_dp_mst, bool, 0600, "Enable multi-stream transport (MST) for new DisplayPort sinks. (default: true)"); #if IS_ENABLED(CONFIG_DRM_I915_DEBUG) -i915_param_named_unsafe(inject_load_failure, uint, 0400, +i915_param_named_unsafe(inject_probe_failure, uint, 0400, "Force an error after a number of failure check points (0:disabled (default), N:force failure at the Nth failure check point)"); #endif @@ -178,6 +179,11 @@ i915_param_named(enable_gvt, bool, 0400, "Enable support for Intel GVT-g graphics virtualization host support(default:false)"); #endif +#if IS_ENABLED(CONFIG_DRM_I915_UNSTABLE_FAKE_LMEM) +i915_param_named_unsafe(fake_lmem_start, ulong, 0600, + "Fake LMEM start offset (default: 0)"); +#endif + static __always_inline void _print_param(struct drm_printer *p, const char *name, const char *type, @@ -189,6 +195,8 @@ static __always_inline void _print_param(struct drm_printer *p, drm_printf(p, "i915.%s=%d\n", name, *(const int *)x); else if (!__builtin_strcmp(type, "unsigned int")) drm_printf(p, "i915.%s=%u\n", name, *(const unsigned int *)x); + else if (!__builtin_strcmp(type, "unsigned long")) + drm_printf(p, "i915.%s=%lu\n", name, *(const unsigned long *)x); else if (!__builtin_strcmp(type, "char *")) drm_printf(p, "i915.%s=%s\n", name, *(const char **)x); else diff --git a/drivers/gpu/drm/i915/i915_params.h b/drivers/gpu/drm/i915/i915_params.h index d29ade3b7de6..31b88f297fbc 100644 --- a/drivers/gpu/drm/i915/i915_params.h +++ b/drivers/gpu/drm/i915/i915_params.h @@ -61,11 +61,12 @@ struct drm_printer; param(char *, dmc_firmware_path, NULL) \ param(int, mmio_debug, -IS_ENABLED(CONFIG_DRM_I915_DEBUG_MMIO)) \ param(int, edp_vswing, 0) \ - param(int, reset, 2) \ - param(unsigned int, inject_load_failure, 0) \ + param(int, reset, 3) \ + param(unsigned int, inject_probe_failure, 0) \ param(int, fastboot, -1) \ param(int, enable_dpcd_backlight, 0) \ param(char *, force_probe, CONFIG_DRM_I915_FORCE_PROBE) \ + param(unsigned long, fake_lmem_start, 0) \ /* leave bools at the end to not create holes */ \ param(bool, alpha_support, IS_ENABLED(CONFIG_DRM_I915_ALPHA_SUPPORT)) \ param(bool, enable_hangcheck, true) \ diff --git a/drivers/gpu/drm/i915/i915_pci.c b/drivers/gpu/drm/i915/i915_pci.c index 1974e4c78a43..1bb701d32a5d 100644 --- a/drivers/gpu/drm/i915/i915_pci.c +++ b/drivers/gpu/drm/i915/i915_pci.c @@ -23,7 +23,6 @@ */ #include <linux/console.h> -#include <linux/vgaarb.h> #include <linux/vga_switcheroo.h> #include <drm/drm_drv.h> @@ -118,6 +117,14 @@ [PIPE_C] = IVB_CURSOR_C_OFFSET, \ } +#define TGL_CURSOR_OFFSETS \ + .cursor_offsets = { \ + [PIPE_A] = CURSOR_A_OFFSET, \ + [PIPE_B] = IVB_CURSOR_B_OFFSET, \ + [PIPE_C] = IVB_CURSOR_C_OFFSET, \ + [PIPE_D] = TGL_CURSOR_D_OFFSET, \ + } + #define I9XX_COLORS \ .color = { .gamma_lut_size = 256 } #define I965_COLORS \ @@ -144,10 +151,13 @@ #define GEN_DEFAULT_PAGE_SIZES \ .page_sizes = I915_GTT_PAGE_SIZE_4K +#define GEN_DEFAULT_REGIONS \ + .memory_regions = REGION_SMEM | REGION_STOLEN + #define I830_FEATURES \ GEN(2), \ .is_mobile = 1, \ - .num_pipes = 2, \ + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B), \ .display.has_overlay = 1, \ .display.cursor_needs_physical = 1, \ .display.overlay_needs_physical = 1, \ @@ -161,11 +171,12 @@ I9XX_PIPE_OFFSETS, \ I9XX_CURSOR_OFFSETS, \ I9XX_COLORS, \ - GEN_DEFAULT_PAGE_SIZES + GEN_DEFAULT_PAGE_SIZES, \ + GEN_DEFAULT_REGIONS #define I845_FEATURES \ GEN(2), \ - .num_pipes = 1, \ + .pipe_mask = BIT(PIPE_A), \ .display.has_overlay = 1, \ .display.overlay_needs_physical = 1, \ .display.has_gmch = 1, \ @@ -178,7 +189,8 @@ I845_PIPE_OFFSETS, \ I845_CURSOR_OFFSETS, \ I9XX_COLORS, \ - GEN_DEFAULT_PAGE_SIZES + GEN_DEFAULT_PAGE_SIZES, \ + GEN_DEFAULT_REGIONS static const struct intel_device_info intel_i830_info = { I830_FEATURES, @@ -203,7 +215,7 @@ static const struct intel_device_info intel_i865g_info = { #define GEN3_FEATURES \ GEN(3), \ - .num_pipes = 2, \ + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B), \ .display.has_gmch = 1, \ .gpu_reset_clobbers_display = true, \ .engine_mask = BIT(RCS0), \ @@ -212,7 +224,8 @@ static const struct intel_device_info intel_i865g_info = { I9XX_PIPE_OFFSETS, \ I9XX_CURSOR_OFFSETS, \ I9XX_COLORS, \ - GEN_DEFAULT_PAGE_SIZES + GEN_DEFAULT_PAGE_SIZES, \ + GEN_DEFAULT_REGIONS static const struct intel_device_info intel_i915g_info = { GEN3_FEATURES, @@ -287,7 +300,7 @@ static const struct intel_device_info intel_pineview_m_info = { #define GEN4_FEATURES \ GEN(4), \ - .num_pipes = 2, \ + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B), \ .display.has_hotplug = 1, \ .display.has_gmch = 1, \ .gpu_reset_clobbers_display = true, \ @@ -297,7 +310,8 @@ static const struct intel_device_info intel_pineview_m_info = { I9XX_PIPE_OFFSETS, \ I9XX_CURSOR_OFFSETS, \ I965_COLORS, \ - GEN_DEFAULT_PAGE_SIZES + GEN_DEFAULT_PAGE_SIZES, \ + GEN_DEFAULT_REGIONS static const struct intel_device_info intel_i965g_info = { GEN4_FEATURES, @@ -337,7 +351,7 @@ static const struct intel_device_info intel_gm45_info = { #define GEN5_FEATURES \ GEN(5), \ - .num_pipes = 2, \ + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B), \ .display.has_hotplug = 1, \ .engine_mask = BIT(RCS0) | BIT(VCS0), \ .has_snoop = true, \ @@ -347,7 +361,8 @@ static const struct intel_device_info intel_gm45_info = { I9XX_PIPE_OFFSETS, \ I9XX_CURSOR_OFFSETS, \ ILK_COLORS, \ - GEN_DEFAULT_PAGE_SIZES + GEN_DEFAULT_PAGE_SIZES, \ + GEN_DEFAULT_REGIONS static const struct intel_device_info intel_ironlake_d_info = { GEN5_FEATURES, @@ -363,7 +378,7 @@ static const struct intel_device_info intel_ironlake_m_info = { #define GEN6_FEATURES \ GEN(6), \ - .num_pipes = 2, \ + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B), \ .display.has_hotplug = 1, \ .display.has_fbc = 1, \ .engine_mask = BIT(RCS0) | BIT(VCS0) | BIT(BCS0), \ @@ -377,7 +392,8 @@ static const struct intel_device_info intel_ironlake_m_info = { I9XX_PIPE_OFFSETS, \ I9XX_CURSOR_OFFSETS, \ ILK_COLORS, \ - GEN_DEFAULT_PAGE_SIZES + GEN_DEFAULT_PAGE_SIZES, \ + GEN_DEFAULT_REGIONS #define SNB_D_PLATFORM \ GEN6_FEATURES, \ @@ -411,7 +427,7 @@ static const struct intel_device_info intel_sandybridge_m_gt2_info = { #define GEN7_FEATURES \ GEN(7), \ - .num_pipes = 3, \ + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B) | BIT(PIPE_C), \ .display.has_hotplug = 1, \ .display.has_fbc = 1, \ .engine_mask = BIT(RCS0) | BIT(VCS0) | BIT(BCS0), \ @@ -420,12 +436,13 @@ static const struct intel_device_info intel_sandybridge_m_gt2_info = { .has_rc6 = 1, \ .has_rc6p = 1, \ .has_rps = true, \ - .ppgtt_type = INTEL_PPGTT_FULL, \ + .ppgtt_type = INTEL_PPGTT_ALIASING, \ .ppgtt_size = 31, \ IVB_PIPE_OFFSETS, \ IVB_CURSOR_OFFSETS, \ IVB_COLORS, \ - GEN_DEFAULT_PAGE_SIZES + GEN_DEFAULT_PAGE_SIZES, \ + GEN_DEFAULT_REGIONS #define IVB_D_PLATFORM \ GEN7_FEATURES, \ @@ -462,7 +479,7 @@ static const struct intel_device_info intel_ivybridge_q_info = { GEN7_FEATURES, PLATFORM(INTEL_IVYBRIDGE), .gt = 2, - .num_pipes = 0, /* legal, last one wins */ + .pipe_mask = 0, /* legal, last one wins */ .has_l3_dpf = 1, }; @@ -470,13 +487,13 @@ static const struct intel_device_info intel_valleyview_info = { PLATFORM(INTEL_VALLEYVIEW), GEN(7), .is_lp = 1, - .num_pipes = 2, + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B), .has_runtime_pm = 1, .has_rc6 = 1, .has_rps = true, .display.has_gmch = 1, .display.has_hotplug = 1, - .ppgtt_type = INTEL_PPGTT_FULL, + .ppgtt_type = INTEL_PPGTT_ALIASING, .ppgtt_size = 31, .has_snoop = true, .has_coherent_ggtt = false, @@ -486,6 +503,7 @@ static const struct intel_device_info intel_valleyview_info = { I9XX_CURSOR_OFFSETS, I965_COLORS, GEN_DEFAULT_PAGE_SIZES, + GEN_DEFAULT_REGIONS, }; #define G75_FEATURES \ @@ -560,7 +578,7 @@ static const struct intel_device_info intel_broadwell_gt3_info = { static const struct intel_device_info intel_cherryview_info = { PLATFORM(INTEL_CHERRYVIEW), GEN(8), - .num_pipes = 3, + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B) | BIT(PIPE_C), .display.has_hotplug = 1, .is_lp = 1, .engine_mask = BIT(RCS0) | BIT(VCS0) | BIT(BCS0) | BIT(VECS0), @@ -570,7 +588,7 @@ static const struct intel_device_info intel_cherryview_info = { .has_rps = true, .has_logical_ring_contexts = 1, .display.has_gmch = 1, - .ppgtt_type = INTEL_PPGTT_FULL, + .ppgtt_type = INTEL_PPGTT_ALIASING, .ppgtt_size = 32, .has_reset_engine = 1, .has_snoop = true, @@ -580,6 +598,7 @@ static const struct intel_device_info intel_cherryview_info = { CHV_CURSOR_OFFSETS, CHV_COLORS, GEN_DEFAULT_PAGE_SIZES, + GEN_DEFAULT_REGIONS, }; #define GEN9_DEFAULT_PAGE_SIZES \ @@ -593,6 +612,7 @@ static const struct intel_device_info intel_cherryview_info = { .has_logical_ring_preemption = 1, \ .display.has_csr = 1, \ .has_gt_uc = 1, \ + .display.has_hdcp = 1, \ .display.has_ipc = 1, \ .ddb_size = 896 @@ -631,11 +651,12 @@ static const struct intel_device_info intel_skylake_gt4_info = { .is_lp = 1, \ .display.has_hotplug = 1, \ .engine_mask = BIT(RCS0) | BIT(VCS0) | BIT(BCS0) | BIT(VECS0), \ - .num_pipes = 3, \ + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B) | BIT(PIPE_C), \ .has_64bit_reloc = 1, \ .display.has_ddi = 1, \ .has_fpga_dbg = 1, \ .display.has_fbc = 1, \ + .display.has_hdcp = 1, \ .display.has_psr = 1, \ .has_runtime_pm = 1, \ .display.has_csr = 1, \ @@ -654,7 +675,8 @@ static const struct intel_device_info intel_skylake_gt4_info = { HSW_PIPE_OFFSETS, \ IVB_CURSOR_OFFSETS, \ IVB_COLORS, \ - GEN9_DEFAULT_PAGE_SIZES + GEN9_DEFAULT_PAGE_SIZES, \ + GEN_DEFAULT_REGIONS static const struct intel_device_info intel_broxton_info = { GEN9_LP_FEATURES, @@ -715,6 +737,7 @@ static const struct intel_device_info intel_coffeelake_gt3_info = { GEN9_FEATURES, \ GEN(10), \ .ddb_size = 1024, \ + .display.has_dsc = 1, \ .has_coherent_ggtt = false, \ GLK_COLORS @@ -787,18 +810,25 @@ static const struct intel_device_info intel_elkhartlake_info = { [TRANSCODER_DSI_0] = TRANSCODER_DSI0_OFFSET, \ [TRANSCODER_DSI_1] = TRANSCODER_DSI1_OFFSET, \ }, \ - .has_global_mocs = 1 + TGL_CURSOR_OFFSETS, \ + .has_global_mocs = 1, \ + .display.has_dsb = 1 static const struct intel_device_info intel_tigerlake_12_info = { GEN12_FEATURES, PLATFORM(INTEL_TIGERLAKE), - .num_pipes = 4, + .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B) | BIT(PIPE_C) | BIT(PIPE_D), .require_force_probe = 1, .display.has_modular_fia = 1, .engine_mask = BIT(RCS0) | BIT(BCS0) | BIT(VECS0) | BIT(VCS0) | BIT(VCS2), + .has_rps = false, /* XXX disabled for debugging */ }; +#define GEN12_DGFX_FEATURES \ + GEN12_FEATURES, \ + .is_dgfx = 1 + #undef GEN #undef PLATFORM diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index e42b86827d6b..65d7c2e599de 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -196,8 +196,11 @@ #include <linux/uuid.h> #include "gem/i915_gem_context.h" -#include "gem/i915_gem_pm.h" +#include "gt/intel_engine_pm.h" +#include "gt/intel_engine_user.h" +#include "gt/intel_gt.h" #include "gt/intel_lrc_reg.h" +#include "gt/intel_ring.h" #include "i915_drv.h" #include "i915_perf.h" @@ -215,6 +218,7 @@ #include "oa/i915_oa_cflgt3.h" #include "oa/i915_oa_cnl.h" #include "oa/i915_oa_icl.h" +#include "oa/i915_oa_tgl.h" /* HW requires this to be a power of two, between 128k and 16M, though driver * is currently generally designed assuming the largest 16M size is used such @@ -291,6 +295,7 @@ static u32 i915_perf_stream_paranoid = true; /* On Gen8+ automatically triggered OA reports include a 'reason' field... */ #define OAREPORT_REASON_MASK 0x3f +#define OAREPORT_REASON_MASK_EXTENDED 0x7f #define OAREPORT_REASON_SHIFT 19 #define OAREPORT_REASON_TIMER (1<<0) #define OAREPORT_REASON_CTX_SWITCH (1<<3) @@ -336,17 +341,24 @@ static const struct i915_oa_format gen8_plus_oa_formats[I915_OA_FORMAT_MAX] = { [I915_OA_FORMAT_C4_B8] = { 7, 64 }, }; +static const struct i915_oa_format gen12_oa_formats[I915_OA_FORMAT_MAX] = { + [I915_OA_FORMAT_A32u40_A4u32_B8_C8] = { 5, 256 }, +}; + #define SAMPLE_OA_REPORT (1<<0) /** * struct perf_open_properties - for validated properties given to open a stream * @sample_flags: `DRM_I915_PERF_PROP_SAMPLE_*` properties are tracked as flags * @single_context: Whether a single or all gpu contexts should be monitored + * @hold_preemption: Whether the preemption is disabled for the filtered + * context * @ctx_handle: A gem ctx handle for use with @single_context * @metrics_set: An ID for an OA unit metric set advertised via sysfs * @oa_format: An OA unit HW report format * @oa_periodic: Whether to enable periodic OA unit sampling * @oa_period_exponent: The OA unit sampling period is derived from this + * @engine: The engine (typically rcs0) being monitored by the OA unit * * As read_properties_unlocked() enumerates and validates the properties given * to open a stream of metrics the configuration is built up in the structure @@ -356,6 +368,7 @@ struct perf_open_properties { u32 sample_flags; u64 single_context:1; + u64 hold_preemption:1; u64 ctx_handle; /* OA sampling state */ @@ -363,69 +376,74 @@ struct perf_open_properties { int oa_format; bool oa_periodic; int oa_period_exponent; + + struct intel_engine_cs *engine; +}; + +struct i915_oa_config_bo { + struct llist_node node; + + struct i915_oa_config *oa_config; + struct i915_vma *vma; }; static enum hrtimer_restart oa_poll_check_timer_cb(struct hrtimer *hrtimer); -static void free_oa_config(struct drm_i915_private *dev_priv, - struct i915_oa_config *oa_config) +void i915_oa_config_release(struct kref *ref) { - if (!PTR_ERR(oa_config->flex_regs)) - kfree(oa_config->flex_regs); - if (!PTR_ERR(oa_config->b_counter_regs)) - kfree(oa_config->b_counter_regs); - if (!PTR_ERR(oa_config->mux_regs)) - kfree(oa_config->mux_regs); - kfree(oa_config); -} + struct i915_oa_config *oa_config = + container_of(ref, typeof(*oa_config), ref); -static void put_oa_config(struct drm_i915_private *dev_priv, - struct i915_oa_config *oa_config) -{ - if (!atomic_dec_and_test(&oa_config->ref_count)) - return; + kfree(oa_config->flex_regs); + kfree(oa_config->b_counter_regs); + kfree(oa_config->mux_regs); - free_oa_config(dev_priv, oa_config); + kfree_rcu(oa_config, rcu); } -static int get_oa_config(struct drm_i915_private *dev_priv, - int metrics_set, - struct i915_oa_config **out_config) +struct i915_oa_config * +i915_perf_get_oa_config(struct i915_perf *perf, int metrics_set) { - int ret; + struct i915_oa_config *oa_config; - if (metrics_set == 1) { - *out_config = &dev_priv->perf.test_config; - atomic_inc(&dev_priv->perf.test_config.ref_count); - return 0; - } + rcu_read_lock(); + if (metrics_set == 1) + oa_config = &perf->test_config; + else + oa_config = idr_find(&perf->metrics_idr, metrics_set); + if (oa_config) + oa_config = i915_oa_config_get(oa_config); + rcu_read_unlock(); - ret = mutex_lock_interruptible(&dev_priv->perf.metrics_lock); - if (ret) - return ret; + return oa_config; +} - *out_config = idr_find(&dev_priv->perf.metrics_idr, metrics_set); - if (!*out_config) - ret = -EINVAL; - else - atomic_inc(&(*out_config)->ref_count); +static void free_oa_config_bo(struct i915_oa_config_bo *oa_bo) +{ + i915_oa_config_put(oa_bo->oa_config); + i915_vma_put(oa_bo->vma); + kfree(oa_bo); +} - mutex_unlock(&dev_priv->perf.metrics_lock); +static u32 gen12_oa_hw_tail_read(struct i915_perf_stream *stream) +{ + struct intel_uncore *uncore = stream->uncore; - return ret; + return intel_uncore_read(uncore, GEN12_OAG_OATAILPTR) & + GEN12_OAG_OATAILPTR_MASK; } static u32 gen8_oa_hw_tail_read(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; - return I915_READ(GEN8_OATAILPTR) & GEN8_OATAILPTR_MASK; + return intel_uncore_read(uncore, GEN8_OATAILPTR) & GEN8_OATAILPTR_MASK; } static u32 gen7_oa_hw_tail_read(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; - u32 oastatus1 = I915_READ(GEN7_OASTATUS1); + struct intel_uncore *uncore = stream->uncore; + u32 oastatus1 = intel_uncore_read(uncore, GEN7_OASTATUS1); return oastatus1 & GEN7_OASTATUS1_TAIL_MASK; } @@ -456,7 +474,6 @@ static u32 gen7_oa_hw_tail_read(struct i915_perf_stream *stream) */ static bool oa_buffer_check_unlocked(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; int report_size = stream->oa_buffer.format_size; unsigned long flags; unsigned int aged_idx; @@ -479,7 +496,7 @@ static bool oa_buffer_check_unlocked(struct i915_perf_stream *stream) aged_tail = stream->oa_buffer.tails[aged_idx].offset; aging_tail = stream->oa_buffer.tails[!aged_idx].offset; - hw_tail = dev_priv->perf.ops.oa_hw_tail_read(stream); + hw_tail = stream->perf->ops.oa_hw_tail_read(stream); /* The tail pointer increases in 64 byte increments, * not in report_size steps... @@ -536,7 +553,7 @@ static bool oa_buffer_check_unlocked(struct i915_perf_stream *stream) aging_tail = hw_tail; stream->oa_buffer.aging_timestamp = now; } else { - DRM_ERROR("Ignoring spurious out of range OA buffer tail pointer = %u\n", + DRM_ERROR("Ignoring spurious out of range OA buffer tail pointer = %x\n", hw_tail); } } @@ -655,7 +672,7 @@ static int gen8_append_oa_reports(struct i915_perf_stream *stream, size_t count, size_t *offset) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; int report_size = stream->oa_buffer.format_size; u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = i915_ggtt_offset(stream->oa_buffer.vma); @@ -738,9 +755,11 @@ static int gen8_append_oa_reports(struct i915_perf_stream *stream, * it to userspace... */ reason = ((report32[0] >> OAREPORT_REASON_SHIFT) & - OAREPORT_REASON_MASK); + (IS_GEN(stream->perf->i915, 12) ? + OAREPORT_REASON_MASK_EXTENDED : + OAREPORT_REASON_MASK)); if (reason == 0) { - if (__ratelimit(&dev_priv->perf.spurious_report_rs)) + if (__ratelimit(&stream->perf->spurious_report_rs)) DRM_NOTE("Skipping spurious, invalid OA report\n"); continue; } @@ -755,7 +774,8 @@ static int gen8_append_oa_reports(struct i915_perf_stream *stream, * Note: that we don't clear the valid_ctx_bit so userspace can * understand that the ID has been squashed by the kernel. */ - if (!(report32[0] & dev_priv->perf.gen8_valid_ctx_bit)) + if (!(report32[0] & stream->perf->gen8_valid_ctx_bit) && + INTEL_GEN(stream->perf->i915) <= 11) ctx_id = report32[2] = INVALID_CTX_ID; /* @@ -789,7 +809,7 @@ static int gen8_append_oa_reports(struct i915_perf_stream *stream, * switches since it's not-uncommon for periodic samples to * identify a switch before any 'context switch' report. */ - if (!dev_priv->perf.exclusive_stream->ctx || + if (!stream->perf->exclusive_stream->ctx || stream->specific_ctx_id == ctx_id || stream->oa_buffer.last_ctx_id == stream->specific_ctx_id || reason & OAREPORT_REASON_CTX_SWITCH) { @@ -798,7 +818,7 @@ static int gen8_append_oa_reports(struct i915_perf_stream *stream, * While filtering for a single context we avoid * leaking the IDs of other contexts. */ - if (dev_priv->perf.exclusive_stream->ctx && + if (stream->perf->exclusive_stream->ctx && stream->specific_ctx_id != ctx_id) { report32[2] = INVALID_CTX_ID; } @@ -822,6 +842,11 @@ static int gen8_append_oa_reports(struct i915_perf_stream *stream, } if (start_offset != *offset) { + i915_reg_t oaheadptr; + + oaheadptr = IS_GEN(stream->perf->i915, 12) ? + GEN12_OAG_OAHEADPTR : GEN8_OAHEADPTR; + spin_lock_irqsave(&stream->oa_buffer.ptr_lock, flags); /* @@ -829,8 +854,8 @@ static int gen8_append_oa_reports(struct i915_perf_stream *stream, * relative to oa_buf_base so put back here... */ head += gtt_offset; - - I915_WRITE(GEN8_OAHEADPTR, head & GEN8_OAHEADPTR_MASK); + intel_uncore_write(uncore, oaheadptr, + head & GEN12_OAG_OAHEADPTR_MASK); stream->oa_buffer.head = head; spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); @@ -864,14 +889,18 @@ static int gen8_oa_read(struct i915_perf_stream *stream, size_t count, size_t *offset) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; u32 oastatus; + i915_reg_t oastatus_reg; int ret; if (WARN_ON(!stream->oa_buffer.vaddr)) return -EIO; - oastatus = I915_READ(GEN8_OASTATUS); + oastatus_reg = IS_GEN(stream->perf->i915, 12) ? + GEN12_OAG_OASTATUS : GEN8_OASTATUS; + + oastatus = intel_uncore_read(uncore, oastatus_reg); /* * We treat OABUFFER_OVERFLOW as a significant error: @@ -896,14 +925,14 @@ static int gen8_oa_read(struct i915_perf_stream *stream, DRM_DEBUG("OA buffer overflow (exponent = %d): force restart\n", stream->period_exponent); - dev_priv->perf.ops.oa_disable(stream); - dev_priv->perf.ops.oa_enable(stream); + stream->perf->ops.oa_disable(stream); + stream->perf->ops.oa_enable(stream); /* * Note: .oa_enable() is expected to re-init the oabuffer and * reset GEN8_OASTATUS for us */ - oastatus = I915_READ(GEN8_OASTATUS); + oastatus = intel_uncore_read(uncore, oastatus_reg); } if (oastatus & GEN8_OASTATUS_REPORT_LOST) { @@ -911,8 +940,8 @@ static int gen8_oa_read(struct i915_perf_stream *stream, DRM_I915_PERF_RECORD_OA_REPORT_LOST); if (ret) return ret; - I915_WRITE(GEN8_OASTATUS, - oastatus & ~GEN8_OASTATUS_REPORT_LOST); + intel_uncore_write(uncore, oastatus_reg, + oastatus & ~GEN8_OASTATUS_REPORT_LOST); } return gen8_append_oa_reports(stream, buf, count, offset); @@ -943,7 +972,7 @@ static int gen7_append_oa_reports(struct i915_perf_stream *stream, size_t count, size_t *offset) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; int report_size = stream->oa_buffer.format_size; u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = i915_ggtt_offset(stream->oa_buffer.vma); @@ -1017,7 +1046,7 @@ static int gen7_append_oa_reports(struct i915_perf_stream *stream, * copying it to userspace... */ if (report32[0] == 0) { - if (__ratelimit(&dev_priv->perf.spurious_report_rs)) + if (__ratelimit(&stream->perf->spurious_report_rs)) DRM_NOTE("Skipping spurious, invalid OA report\n"); continue; } @@ -1043,9 +1072,9 @@ static int gen7_append_oa_reports(struct i915_perf_stream *stream, */ head += gtt_offset; - I915_WRITE(GEN7_OASTATUS2, - ((head & GEN7_OASTATUS2_HEAD_MASK) | - GEN7_OASTATUS2_MEM_SELECT_GGTT)); + intel_uncore_write(uncore, GEN7_OASTATUS2, + (head & GEN7_OASTATUS2_HEAD_MASK) | + GEN7_OASTATUS2_MEM_SELECT_GGTT); stream->oa_buffer.head = head; spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); @@ -1075,21 +1104,21 @@ static int gen7_oa_read(struct i915_perf_stream *stream, size_t count, size_t *offset) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; u32 oastatus1; int ret; if (WARN_ON(!stream->oa_buffer.vaddr)) return -EIO; - oastatus1 = I915_READ(GEN7_OASTATUS1); + oastatus1 = intel_uncore_read(uncore, GEN7_OASTATUS1); /* XXX: On Haswell we don't have a safe way to clear oastatus1 * bits while the OA unit is enabled (while the tail pointer * may be updated asynchronously) so we ignore status bits * that have already been reported to userspace. */ - oastatus1 &= ~dev_priv->perf.gen7_latched_oastatus1; + oastatus1 &= ~stream->perf->gen7_latched_oastatus1; /* We treat OABUFFER_OVERFLOW as a significant error: * @@ -1120,10 +1149,10 @@ static int gen7_oa_read(struct i915_perf_stream *stream, DRM_DEBUG("OA buffer overflow (exponent = %d): force restart\n", stream->period_exponent); - dev_priv->perf.ops.oa_disable(stream); - dev_priv->perf.ops.oa_enable(stream); + stream->perf->ops.oa_disable(stream); + stream->perf->ops.oa_enable(stream); - oastatus1 = I915_READ(GEN7_OASTATUS1); + oastatus1 = intel_uncore_read(uncore, GEN7_OASTATUS1); } if (unlikely(oastatus1 & GEN7_OASTATUS1_REPORT_LOST)) { @@ -1131,7 +1160,7 @@ static int gen7_oa_read(struct i915_perf_stream *stream, DRM_I915_PERF_RECORD_OA_REPORT_LOST); if (ret) return ret; - dev_priv->perf.gen7_latched_oastatus1 |= + stream->perf->gen7_latched_oastatus1 |= GEN7_OASTATUS1_REPORT_LOST; } @@ -1196,25 +1225,18 @@ static int i915_oa_read(struct i915_perf_stream *stream, size_t count, size_t *offset) { - struct drm_i915_private *dev_priv = stream->dev_priv; - - return dev_priv->perf.ops.read(stream, buf, count, offset); + return stream->perf->ops.read(stream, buf, count, offset); } static struct intel_context *oa_pin_context(struct i915_perf_stream *stream) { struct i915_gem_engines_iter it; - struct drm_i915_private *i915 = stream->dev_priv; struct i915_gem_context *ctx = stream->ctx; struct intel_context *ce; int err; - err = i915_mutex_lock_interruptible(&i915->drm); - if (err) - return ERR_PTR(err); - for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { - if (ce->engine->class != RENDER_CLASS) + if (ce->engine != stream->engine) /* first match! */ continue; /* @@ -1229,10 +1251,6 @@ static struct intel_context *oa_pin_context(struct i915_perf_stream *stream) } i915_gem_context_unlock_engines(ctx); - mutex_unlock(&i915->drm.struct_mutex); - if (err) - return ERR_PTR(err); - return stream->pinned_ctx; } @@ -1248,14 +1266,13 @@ static struct intel_context *oa_pin_context(struct i915_perf_stream *stream) */ static int oa_get_render_ctx_id(struct i915_perf_stream *stream) { - struct drm_i915_private *i915 = stream->dev_priv; struct intel_context *ce; ce = oa_pin_context(stream); if (IS_ERR(ce)) return PTR_ERR(ce); - switch (INTEL_GEN(i915)) { + switch (INTEL_GEN(ce->engine->i915)) { case 7: { /* * On Haswell we don't do any post processing of the reports @@ -1269,7 +1286,11 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream) case 8: case 9: case 10: - if (USES_GUC_SUBMISSION(i915)) { + if (intel_engine_in_execlists_submission_mode(ce->engine)) { + stream->specific_ctx_id_mask = + (1U << GEN8_CTX_ID_WIDTH) - 1; + stream->specific_ctx_id = stream->specific_ctx_id_mask; + } else { /* * When using GuC, the context descriptor we write in * i915 is read by GuC and rewritten before it's @@ -1289,31 +1310,23 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream) */ stream->specific_ctx_id_mask = (1U << (GEN8_CTX_ID_WIDTH - 1)) - 1; - } else { - stream->specific_ctx_id_mask = - (1U << GEN8_CTX_ID_WIDTH) - 1; - stream->specific_ctx_id = - upper_32_bits(ce->lrc_desc); - stream->specific_ctx_id &= - stream->specific_ctx_id_mask; } break; - case 11: { + case 11: + case 12: { stream->specific_ctx_id_mask = - ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32) | - ((1U << GEN11_ENGINE_INSTANCE_WIDTH) - 1) << (GEN11_ENGINE_INSTANCE_SHIFT - 32) | - ((1 << GEN11_ENGINE_CLASS_WIDTH) - 1) << (GEN11_ENGINE_CLASS_SHIFT - 32); - stream->specific_ctx_id = upper_32_bits(ce->lrc_desc); - stream->specific_ctx_id &= - stream->specific_ctx_id_mask; + ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32); + stream->specific_ctx_id = stream->specific_ctx_id_mask; break; } default: - MISSING_CASE(INTEL_GEN(i915)); + MISSING_CASE(INTEL_GEN(ce->engine->i915)); } + ce->tag = stream->specific_ctx_id_mask; + DRM_DEBUG_DRIVER("filtering on ctx_id=0x%x ctx_id_mask=0x%x\n", stream->specific_ctx_id, stream->specific_ctx_id_mask); @@ -1330,69 +1343,76 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream) */ static void oa_put_render_ctx_id(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; struct intel_context *ce; - stream->specific_ctx_id = INVALID_CTX_ID; - stream->specific_ctx_id_mask = 0; - ce = fetch_and_zero(&stream->pinned_ctx); if (ce) { - mutex_lock(&dev_priv->drm.struct_mutex); + ce->tag = 0; /* recomputed on next submission after parking */ intel_context_unpin(ce); - mutex_unlock(&dev_priv->drm.struct_mutex); } + + stream->specific_ctx_id = INVALID_CTX_ID; + stream->specific_ctx_id_mask = 0; } static void free_oa_buffer(struct i915_perf_stream *stream) { - struct drm_i915_private *i915 = stream->dev_priv; - - mutex_lock(&i915->drm.struct_mutex); - i915_vma_unpin_and_release(&stream->oa_buffer.vma, I915_VMA_RELEASE_MAP); - mutex_unlock(&i915->drm.struct_mutex); - stream->oa_buffer.vaddr = NULL; } +static void +free_oa_configs(struct i915_perf_stream *stream) +{ + struct i915_oa_config_bo *oa_bo, *tmp; + + i915_oa_config_put(stream->oa_config); + llist_for_each_entry_safe(oa_bo, tmp, stream->oa_config_bos.first, node) + free_oa_config_bo(oa_bo); +} + +static void +free_noa_wait(struct i915_perf_stream *stream) +{ + i915_vma_unpin_and_release(&stream->noa_wait, 0); +} + static void i915_oa_stream_destroy(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct i915_perf *perf = stream->perf; - BUG_ON(stream != dev_priv->perf.exclusive_stream); + BUG_ON(stream != perf->exclusive_stream); /* * Unset exclusive_stream first, it will be checked while disabling * the metric set on gen8+. */ - mutex_lock(&dev_priv->drm.struct_mutex); - dev_priv->perf.exclusive_stream = NULL; - dev_priv->perf.ops.disable_metric_set(stream); - mutex_unlock(&dev_priv->drm.struct_mutex); + perf->exclusive_stream = NULL; + perf->ops.disable_metric_set(stream); free_oa_buffer(stream); - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); - intel_runtime_pm_put(&dev_priv->runtime_pm, stream->wakeref); + intel_uncore_forcewake_put(stream->uncore, FORCEWAKE_ALL); + intel_engine_pm_put(stream->engine); if (stream->ctx) oa_put_render_ctx_id(stream); - put_oa_config(dev_priv, stream->oa_config); + free_oa_configs(stream); + free_noa_wait(stream); - if (dev_priv->perf.spurious_report_rs.missed) { + if (perf->spurious_report_rs.missed) { DRM_NOTE("%d spurious OA report notices suppressed due to ratelimiting\n", - dev_priv->perf.spurious_report_rs.missed); + perf->spurious_report_rs.missed); } } static void gen7_init_oa_buffer(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; u32 gtt_offset = i915_ggtt_offset(stream->oa_buffer.vma); unsigned long flags; @@ -1401,13 +1421,14 @@ static void gen7_init_oa_buffer(struct i915_perf_stream *stream) /* Pre-DevBDW: OABUFFER must be set with counters off, * before OASTATUS1, but after OASTATUS2 */ - I915_WRITE(GEN7_OASTATUS2, - gtt_offset | GEN7_OASTATUS2_MEM_SELECT_GGTT); /* head */ + intel_uncore_write(uncore, GEN7_OASTATUS2, /* head */ + gtt_offset | GEN7_OASTATUS2_MEM_SELECT_GGTT); stream->oa_buffer.head = gtt_offset; - I915_WRITE(GEN7_OABUFFER, gtt_offset); + intel_uncore_write(uncore, GEN7_OABUFFER, gtt_offset); - I915_WRITE(GEN7_OASTATUS1, gtt_offset | OABUFFER_SIZE_16M); /* tail */ + intel_uncore_write(uncore, GEN7_OASTATUS1, /* tail */ + gtt_offset | OABUFFER_SIZE_16M); /* Mark that we need updated tail pointers to read from... */ stream->oa_buffer.tails[0].offset = INVALID_TAIL_PTR; @@ -1419,7 +1440,7 @@ static void gen7_init_oa_buffer(struct i915_perf_stream *stream) * already seen since they can't be cleared while periodic * sampling is enabled. */ - dev_priv->perf.gen7_latched_oastatus1 = 0; + stream->perf->gen7_latched_oastatus1 = 0; /* NB: although the OA buffer will initially be allocated * zeroed via shmfs (and so this memset is redundant when @@ -1434,25 +1455,22 @@ static void gen7_init_oa_buffer(struct i915_perf_stream *stream) */ memset(stream->oa_buffer.vaddr, 0, OA_BUFFER_SIZE); - /* Maybe make ->pollin per-stream state if we support multiple - * concurrent streams in the future. - */ stream->pollin = false; } static void gen8_init_oa_buffer(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; u32 gtt_offset = i915_ggtt_offset(stream->oa_buffer.vma); unsigned long flags; spin_lock_irqsave(&stream->oa_buffer.ptr_lock, flags); - I915_WRITE(GEN8_OASTATUS, 0); - I915_WRITE(GEN8_OAHEADPTR, gtt_offset); + intel_uncore_write(uncore, GEN8_OASTATUS, 0); + intel_uncore_write(uncore, GEN8_OAHEADPTR, gtt_offset); stream->oa_buffer.head = gtt_offset; - I915_WRITE(GEN8_OABUFFER_UDW, 0); + intel_uncore_write(uncore, GEN8_OABUFFER_UDW, 0); /* * PRM says: @@ -1462,9 +1480,9 @@ static void gen8_init_oa_buffer(struct i915_perf_stream *stream) * to enable proper functionality of the overflow * bit." */ - I915_WRITE(GEN8_OABUFFER, gtt_offset | + intel_uncore_write(uncore, GEN8_OABUFFER, gtt_offset | OABUFFER_SIZE_16M | GEN8_OABUFFER_MEM_SELECT_GGTT); - I915_WRITE(GEN8_OATAILPTR, gtt_offset & GEN8_OATAILPTR_MASK); + intel_uncore_write(uncore, GEN8_OATAILPTR, gtt_offset & GEN8_OATAILPTR_MASK); /* Mark that we need updated tail pointers to read from... */ stream->oa_buffer.tails[0].offset = INVALID_TAIL_PTR; @@ -1493,35 +1511,82 @@ static void gen8_init_oa_buffer(struct i915_perf_stream *stream) */ memset(stream->oa_buffer.vaddr, 0, OA_BUFFER_SIZE); + stream->pollin = false; +} + +static void gen12_init_oa_buffer(struct i915_perf_stream *stream) +{ + struct intel_uncore *uncore = stream->uncore; + u32 gtt_offset = i915_ggtt_offset(stream->oa_buffer.vma); + unsigned long flags; + + spin_lock_irqsave(&stream->oa_buffer.ptr_lock, flags); + + intel_uncore_write(uncore, GEN12_OAG_OASTATUS, 0); + intel_uncore_write(uncore, GEN12_OAG_OAHEADPTR, + gtt_offset & GEN12_OAG_OAHEADPTR_MASK); + stream->oa_buffer.head = gtt_offset; + + /* + * PRM says: + * + * "This MMIO must be set before the OATAILPTR + * register and after the OAHEADPTR register. This is + * to enable proper functionality of the overflow + * bit." + */ + intel_uncore_write(uncore, GEN12_OAG_OABUFFER, gtt_offset | + OABUFFER_SIZE_16M | GEN8_OABUFFER_MEM_SELECT_GGTT); + intel_uncore_write(uncore, GEN12_OAG_OATAILPTR, + gtt_offset & GEN12_OAG_OATAILPTR_MASK); + + /* Mark that we need updated tail pointers to read from... */ + stream->oa_buffer.tails[0].offset = INVALID_TAIL_PTR; + stream->oa_buffer.tails[1].offset = INVALID_TAIL_PTR; + + /* + * Reset state used to recognise context switches, affecting which + * reports we will forward to userspace while filtering for a single + * context. + */ + stream->oa_buffer.last_ctx_id = INVALID_CTX_ID; + + spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); + /* - * Maybe make ->pollin per-stream state if we support multiple - * concurrent streams in the future. + * NB: although the OA buffer will initially be allocated + * zeroed via shmfs (and so this memset is redundant when + * first allocating), we may re-init the OA buffer, either + * when re-enabling a stream or in error/reset paths. + * + * The reason we clear the buffer for each re-init is for the + * sanity check in gen8_append_oa_reports() that looks at the + * reason field to make sure it's non-zero which relies on + * the assumption that new reports are being written to zeroed + * memory... */ + memset(stream->oa_buffer.vaddr, 0, + stream->oa_buffer.vma->size); + stream->pollin = false; } static int alloc_oa_buffer(struct i915_perf_stream *stream) { struct drm_i915_gem_object *bo; - struct drm_i915_private *dev_priv = stream->dev_priv; struct i915_vma *vma; int ret; if (WARN_ON(stream->oa_buffer.vma)) return -ENODEV; - ret = i915_mutex_lock_interruptible(&dev_priv->drm); - if (ret) - return ret; - BUILD_BUG_ON_NOT_POWER_OF_2(OA_BUFFER_SIZE); BUILD_BUG_ON(OA_BUFFER_SIZE < SZ_128K || OA_BUFFER_SIZE > SZ_16M); - bo = i915_gem_object_create_shmem(dev_priv, OA_BUFFER_SIZE); + bo = i915_gem_object_create_shmem(stream->perf->i915, OA_BUFFER_SIZE); if (IS_ERR(bo)) { DRM_ERROR("Failed to allocate OA buffer\n"); - ret = PTR_ERR(bo); - goto unlock; + return PTR_ERR(bo); } i915_gem_object_set_cache_coherency(bo, I915_CACHE_LLC); @@ -1541,11 +1606,7 @@ static int alloc_oa_buffer(struct i915_perf_stream *stream) goto err_unpin; } - DRM_DEBUG_DRIVER("OA Buffer initialized, gtt offset = 0x%x, vaddr = %p\n", - i915_ggtt_offset(stream->oa_buffer.vma), - stream->oa_buffer.vaddr); - - goto unlock; + return 0; err_unpin: __i915_vma_unpin(vma); @@ -1556,55 +1617,389 @@ err_unref: stream->oa_buffer.vaddr = NULL; stream->oa_buffer.vma = NULL; -unlock: - mutex_unlock(&dev_priv->drm.struct_mutex); return ret; } -static void config_oa_regs(struct drm_i915_private *dev_priv, - const struct i915_oa_reg *regs, - u32 n_regs) +static u32 *save_restore_register(struct i915_perf_stream *stream, u32 *cs, + bool save, i915_reg_t reg, u32 offset, + u32 dword_count) +{ + u32 cmd; + u32 d; + + cmd = save ? MI_STORE_REGISTER_MEM : MI_LOAD_REGISTER_MEM; + if (INTEL_GEN(stream->perf->i915) >= 8) + cmd++; + + for (d = 0; d < dword_count; d++) { + *cs++ = cmd; + *cs++ = i915_mmio_reg_offset(reg) + 4 * d; + *cs++ = intel_gt_scratch_offset(stream->engine->gt, + offset) + 4 * d; + *cs++ = 0; + } + + return cs; +} + +static int alloc_noa_wait(struct i915_perf_stream *stream) +{ + struct drm_i915_private *i915 = stream->perf->i915; + struct drm_i915_gem_object *bo; + struct i915_vma *vma; + const u64 delay_ticks = 0xffffffffffffffff - + DIV64_U64_ROUND_UP( + atomic64_read(&stream->perf->noa_programming_delay) * + RUNTIME_INFO(i915)->cs_timestamp_frequency_khz, + 1000000ull); + const u32 base = stream->engine->mmio_base; +#define CS_GPR(x) GEN8_RING_CS_GPR(base, x) + u32 *batch, *ts0, *cs, *jump; + int ret, i; + enum { + START_TS, + NOW_TS, + DELTA_TS, + JUMP_PREDICATE, + DELTA_TARGET, + N_CS_GPR + }; + + bo = i915_gem_object_create_internal(i915, 4096); + if (IS_ERR(bo)) { + DRM_ERROR("Failed to allocate NOA wait batchbuffer\n"); + return PTR_ERR(bo); + } + + /* + * We pin in GGTT because we jump into this buffer now because + * multiple OA config BOs will have a jump to this address and it + * needs to be fixed during the lifetime of the i915/perf stream. + */ + vma = i915_gem_object_ggtt_pin(bo, NULL, 0, 0, PIN_HIGH); + if (IS_ERR(vma)) { + ret = PTR_ERR(vma); + goto err_unref; + } + + batch = cs = i915_gem_object_pin_map(bo, I915_MAP_WB); + if (IS_ERR(batch)) { + ret = PTR_ERR(batch); + goto err_unpin; + } + + /* Save registers. */ + for (i = 0; i < N_CS_GPR; i++) + cs = save_restore_register( + stream, cs, true /* save */, CS_GPR(i), + INTEL_GT_SCRATCH_FIELD_PERF_CS_GPR + 8 * i, 2); + cs = save_restore_register( + stream, cs, true /* save */, MI_PREDICATE_RESULT_1, + INTEL_GT_SCRATCH_FIELD_PERF_PREDICATE_RESULT_1, 1); + + /* First timestamp snapshot location. */ + ts0 = cs; + + /* + * Initial snapshot of the timestamp register to implement the wait. + * We work with 32b values, so clear out the top 32b bits of the + * register because the ALU works 64bits. + */ + *cs++ = MI_LOAD_REGISTER_IMM(1); + *cs++ = i915_mmio_reg_offset(CS_GPR(START_TS)) + 4; + *cs++ = 0; + *cs++ = MI_LOAD_REGISTER_REG | (3 - 2); + *cs++ = i915_mmio_reg_offset(RING_TIMESTAMP(base)); + *cs++ = i915_mmio_reg_offset(CS_GPR(START_TS)); + + /* + * This is the location we're going to jump back into until the + * required amount of time has passed. + */ + jump = cs; + + /* + * Take another snapshot of the timestamp register. Take care to clear + * up the top 32bits of CS_GPR(1) as we're using it for other + * operations below. + */ + *cs++ = MI_LOAD_REGISTER_IMM(1); + *cs++ = i915_mmio_reg_offset(CS_GPR(NOW_TS)) + 4; + *cs++ = 0; + *cs++ = MI_LOAD_REGISTER_REG | (3 - 2); + *cs++ = i915_mmio_reg_offset(RING_TIMESTAMP(base)); + *cs++ = i915_mmio_reg_offset(CS_GPR(NOW_TS)); + + /* + * Do a diff between the 2 timestamps and store the result back into + * CS_GPR(1). + */ + *cs++ = MI_MATH(5); + *cs++ = MI_MATH_LOAD(MI_MATH_REG_SRCA, MI_MATH_REG(NOW_TS)); + *cs++ = MI_MATH_LOAD(MI_MATH_REG_SRCB, MI_MATH_REG(START_TS)); + *cs++ = MI_MATH_SUB; + *cs++ = MI_MATH_STORE(MI_MATH_REG(DELTA_TS), MI_MATH_REG_ACCU); + *cs++ = MI_MATH_STORE(MI_MATH_REG(JUMP_PREDICATE), MI_MATH_REG_CF); + + /* + * Transfer the carry flag (set to 1 if ts1 < ts0, meaning the + * timestamp have rolled over the 32bits) into the predicate register + * to be used for the predicated jump. + */ + *cs++ = MI_LOAD_REGISTER_REG | (3 - 2); + *cs++ = i915_mmio_reg_offset(CS_GPR(JUMP_PREDICATE)); + *cs++ = i915_mmio_reg_offset(MI_PREDICATE_RESULT_1); + + /* Restart from the beginning if we had timestamps roll over. */ + *cs++ = (INTEL_GEN(i915) < 8 ? + MI_BATCH_BUFFER_START : + MI_BATCH_BUFFER_START_GEN8) | + MI_BATCH_PREDICATE; + *cs++ = i915_ggtt_offset(vma) + (ts0 - batch) * 4; + *cs++ = 0; + + /* + * Now add the diff between to previous timestamps and add it to : + * (((1 * << 64) - 1) - delay_ns) + * + * When the Carry Flag contains 1 this means the elapsed time is + * longer than the expected delay, and we can exit the wait loop. + */ + *cs++ = MI_LOAD_REGISTER_IMM(2); + *cs++ = i915_mmio_reg_offset(CS_GPR(DELTA_TARGET)); + *cs++ = lower_32_bits(delay_ticks); + *cs++ = i915_mmio_reg_offset(CS_GPR(DELTA_TARGET)) + 4; + *cs++ = upper_32_bits(delay_ticks); + + *cs++ = MI_MATH(4); + *cs++ = MI_MATH_LOAD(MI_MATH_REG_SRCA, MI_MATH_REG(DELTA_TS)); + *cs++ = MI_MATH_LOAD(MI_MATH_REG_SRCB, MI_MATH_REG(DELTA_TARGET)); + *cs++ = MI_MATH_ADD; + *cs++ = MI_MATH_STOREINV(MI_MATH_REG(JUMP_PREDICATE), MI_MATH_REG_CF); + + /* + * Transfer the result into the predicate register to be used for the + * predicated jump. + */ + *cs++ = MI_LOAD_REGISTER_REG | (3 - 2); + *cs++ = i915_mmio_reg_offset(CS_GPR(JUMP_PREDICATE)); + *cs++ = i915_mmio_reg_offset(MI_PREDICATE_RESULT_1); + + /* Predicate the jump. */ + *cs++ = (INTEL_GEN(i915) < 8 ? + MI_BATCH_BUFFER_START : + MI_BATCH_BUFFER_START_GEN8) | + MI_BATCH_PREDICATE; + *cs++ = i915_ggtt_offset(vma) + (jump - batch) * 4; + *cs++ = 0; + + /* Restore registers. */ + for (i = 0; i < N_CS_GPR; i++) + cs = save_restore_register( + stream, cs, false /* restore */, CS_GPR(i), + INTEL_GT_SCRATCH_FIELD_PERF_CS_GPR + 8 * i, 2); + cs = save_restore_register( + stream, cs, false /* restore */, MI_PREDICATE_RESULT_1, + INTEL_GT_SCRATCH_FIELD_PERF_PREDICATE_RESULT_1, 1); + + /* And return to the ring. */ + *cs++ = MI_BATCH_BUFFER_END; + + GEM_BUG_ON(cs - batch > PAGE_SIZE / sizeof(*batch)); + + i915_gem_object_flush_map(bo); + i915_gem_object_unpin_map(bo); + + stream->noa_wait = vma; + return 0; + +err_unpin: + i915_vma_unpin_and_release(&vma, 0); +err_unref: + i915_gem_object_put(bo); + return ret; +} + +static u32 *write_cs_mi_lri(u32 *cs, + const struct i915_oa_reg *reg_data, + u32 n_regs) { u32 i; for (i = 0; i < n_regs; i++) { - const struct i915_oa_reg *reg = regs + i; + if ((i % MI_LOAD_REGISTER_IMM_MAX_REGS) == 0) { + u32 n_lri = min_t(u32, + n_regs - i, + MI_LOAD_REGISTER_IMM_MAX_REGS); - I915_WRITE(reg->addr, reg->value); + *cs++ = MI_LOAD_REGISTER_IMM(n_lri); + } + *cs++ = i915_mmio_reg_offset(reg_data[i].addr); + *cs++ = reg_data[i].value; } + + return cs; } -static void delay_after_mux(void) +static int num_lri_dwords(int num_regs) { + int count = 0; + + if (num_regs > 0) { + count += DIV_ROUND_UP(num_regs, MI_LOAD_REGISTER_IMM_MAX_REGS); + count += num_regs * 2; + } + + return count; +} + +static struct i915_oa_config_bo * +alloc_oa_config_buffer(struct i915_perf_stream *stream, + struct i915_oa_config *oa_config) +{ + struct drm_i915_gem_object *obj; + struct i915_oa_config_bo *oa_bo; + size_t config_length = 0; + u32 *cs; + int err; + + oa_bo = kzalloc(sizeof(*oa_bo), GFP_KERNEL); + if (!oa_bo) + return ERR_PTR(-ENOMEM); + + config_length += num_lri_dwords(oa_config->mux_regs_len); + config_length += num_lri_dwords(oa_config->b_counter_regs_len); + config_length += num_lri_dwords(oa_config->flex_regs_len); + config_length += 3; /* MI_BATCH_BUFFER_START */ + config_length = ALIGN(sizeof(u32) * config_length, I915_GTT_PAGE_SIZE); + + obj = i915_gem_object_create_shmem(stream->perf->i915, config_length); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + goto err_free; + } + + cs = i915_gem_object_pin_map(obj, I915_MAP_WB); + if (IS_ERR(cs)) { + err = PTR_ERR(cs); + goto err_oa_bo; + } + + cs = write_cs_mi_lri(cs, + oa_config->mux_regs, + oa_config->mux_regs_len); + cs = write_cs_mi_lri(cs, + oa_config->b_counter_regs, + oa_config->b_counter_regs_len); + cs = write_cs_mi_lri(cs, + oa_config->flex_regs, + oa_config->flex_regs_len); + + /* Jump into the active wait. */ + *cs++ = (INTEL_GEN(stream->perf->i915) < 8 ? + MI_BATCH_BUFFER_START : + MI_BATCH_BUFFER_START_GEN8); + *cs++ = i915_ggtt_offset(stream->noa_wait); + *cs++ = 0; + + i915_gem_object_flush_map(obj); + i915_gem_object_unpin_map(obj); + + oa_bo->vma = i915_vma_instance(obj, + &stream->engine->gt->ggtt->vm, + NULL); + if (IS_ERR(oa_bo->vma)) { + err = PTR_ERR(oa_bo->vma); + goto err_oa_bo; + } + + oa_bo->oa_config = i915_oa_config_get(oa_config); + llist_add(&oa_bo->node, &stream->oa_config_bos); + + return oa_bo; + +err_oa_bo: + i915_gem_object_put(obj); +err_free: + kfree(oa_bo); + return ERR_PTR(err); +} + +static struct i915_vma * +get_oa_vma(struct i915_perf_stream *stream, struct i915_oa_config *oa_config) +{ + struct i915_oa_config_bo *oa_bo; + /* - * It apparently takes a fairly long time for a new MUX - * configuration to be be applied after these register writes. - * This delay duration was derived empirically based on the - * render_basic config but hopefully it covers the maximum - * configuration latency. - * - * As a fallback, the checks in _append_oa_reports() to skip - * invalid OA reports do also seem to work to discard reports - * generated before this config has completed - albeit not - * silently. - * - * Unfortunately this is essentially a magic number, since we - * don't currently know of a reliable mechanism for predicting - * how long the MUX config will take to apply and besides - * seeing invalid reports we don't know of a reliable way to - * explicitly check that the MUX config has landed. - * - * It's even possible we've miss characterized the underlying - * problem - it just seems like the simplest explanation why - * a delay at this location would mitigate any invalid reports. + * Look for the buffer in the already allocated BOs attached + * to the stream. */ - usleep_range(15000, 20000); + llist_for_each_entry(oa_bo, stream->oa_config_bos.first, node) { + if (oa_bo->oa_config == oa_config && + memcmp(oa_bo->oa_config->uuid, + oa_config->uuid, + sizeof(oa_config->uuid)) == 0) + goto out; + } + + oa_bo = alloc_oa_config_buffer(stream, oa_config); + if (IS_ERR(oa_bo)) + return ERR_CAST(oa_bo); + +out: + return i915_vma_get(oa_bo->vma); +} + +static int emit_oa_config(struct i915_perf_stream *stream, + struct i915_oa_config *oa_config, + struct intel_context *ce) +{ + struct i915_request *rq; + struct i915_vma *vma; + int err; + + vma = get_oa_vma(stream, oa_config); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL | PIN_HIGH); + if (err) + goto err_vma_put; + + rq = i915_request_create(ce); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto err_vma_unpin; + } + + i915_vma_lock(vma); + err = i915_request_await_object(rq, vma->obj, 0); + if (!err) + err = i915_vma_move_to_active(vma, rq, 0); + i915_vma_unlock(vma); + if (err) + goto err_add_request; + + err = rq->engine->emit_bb_start(rq, + vma->node.start, 0, + I915_DISPATCH_SECURE); +err_add_request: + i915_request_add(rq); +err_vma_unpin: + i915_vma_unpin(vma); +err_vma_put: + i915_vma_put(vma); + return err; +} + +static struct intel_context *oa_context(struct i915_perf_stream *stream) +{ + return stream->pinned_ctx ?: stream->engine->kernel_context; } static int hsw_enable_metric_set(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; - const struct i915_oa_config *oa_config = stream->oa_config; + struct intel_uncore *uncore = stream->uncore; /* * PRM: @@ -1616,31 +2011,24 @@ static int hsw_enable_metric_set(struct i915_perf_stream *stream) * count the events from non-render domain. Unit level clock * gating for RCS should also be disabled. */ - I915_WRITE(GEN7_MISCCPCTL, (I915_READ(GEN7_MISCCPCTL) & - ~GEN7_DOP_CLOCK_GATE_ENABLE)); - I915_WRITE(GEN6_UCGCTL1, (I915_READ(GEN6_UCGCTL1) | - GEN6_CSUNIT_CLOCK_GATE_DISABLE)); - - config_oa_regs(dev_priv, oa_config->mux_regs, oa_config->mux_regs_len); - delay_after_mux(); - - config_oa_regs(dev_priv, oa_config->b_counter_regs, - oa_config->b_counter_regs_len); + intel_uncore_rmw(uncore, GEN7_MISCCPCTL, + GEN7_DOP_CLOCK_GATE_ENABLE, 0); + intel_uncore_rmw(uncore, GEN6_UCGCTL1, + 0, GEN6_CSUNIT_CLOCK_GATE_DISABLE); - return 0; + return emit_oa_config(stream, stream->oa_config, oa_context(stream)); } static void hsw_disable_metric_set(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; - I915_WRITE(GEN6_UCGCTL1, (I915_READ(GEN6_UCGCTL1) & - ~GEN6_CSUNIT_CLOCK_GATE_DISABLE)); - I915_WRITE(GEN7_MISCCPCTL, (I915_READ(GEN7_MISCCPCTL) | - GEN7_DOP_CLOCK_GATE_ENABLE)); + intel_uncore_rmw(uncore, GEN6_UCGCTL1, + GEN6_CSUNIT_CLOCK_GATE_DISABLE, 0); + intel_uncore_rmw(uncore, GEN7_MISCCPCTL, + 0, GEN7_DOP_CLOCK_GATE_ENABLE); - I915_WRITE(GDT_CHICKEN_BITS, (I915_READ(GDT_CHICKEN_BITS) & - ~GT_NOA_ENABLE)); + intel_uncore_rmw(uncore, GDT_CHICKEN_BITS, GT_NOA_ENABLE, 0); } static u32 oa_config_flex_reg(const struct i915_oa_config *oa_config, @@ -1672,14 +2060,11 @@ static u32 oa_config_flex_reg(const struct i915_oa_config *oa_config, * in the case that the OA unit has been disabled. */ static void -gen8_update_reg_state_unlocked(struct i915_perf_stream *stream, - struct intel_context *ce, - u32 *reg_state, - const struct i915_oa_config *oa_config) -{ - struct drm_i915_private *i915 = ce->engine->i915; - u32 ctx_oactxctrl = i915->perf.ctx_oactxctrl_offset; - u32 ctx_flexeu0 = i915->perf.ctx_flexeu0_offset; +gen8_update_reg_state_unlocked(const struct intel_context *ce, + const struct i915_perf_stream *stream) +{ + u32 ctx_oactxctrl = stream->perf->ctx_oactxctrl_offset; + u32 ctx_flexeu0 = stream->perf->ctx_flexeu0_offset; /* The MMIO offsets for Flex EU registers aren't contiguous */ i915_reg_t flex_regs[] = { EU_PERF_CNTL0, @@ -1690,21 +2075,28 @@ gen8_update_reg_state_unlocked(struct i915_perf_stream *stream, EU_PERF_CNTL5, EU_PERF_CNTL6, }; + u32 *reg_state = ce->lrc_reg_state; int i; - CTX_REG(reg_state, ctx_oactxctrl, GEN8_OACTXCONTROL, - (stream->period_exponent << GEN8_OA_TIMER_PERIOD_SHIFT) | - (stream->periodic ? GEN8_OA_TIMER_ENABLE : 0) | - GEN8_OA_COUNTER_RESUME); + if (IS_GEN(stream->perf->i915, 12)) { + u32 format = stream->oa_buffer.format; - for (i = 0; i < ARRAY_SIZE(flex_regs); i++) { - CTX_REG(reg_state, ctx_flexeu0 + i * 2, flex_regs[i], - oa_config_flex_reg(oa_config, flex_regs[i])); + reg_state[ctx_oactxctrl + 1] = + (format << GEN12_OAR_OACONTROL_COUNTER_FORMAT_SHIFT) | + (stream->oa_config ? GEN12_OAR_OACONTROL_COUNTER_ENABLE : 0); + } else { + reg_state[ctx_oactxctrl + 1] = + (stream->period_exponent << GEN8_OA_TIMER_PERIOD_SHIFT) | + (stream->periodic ? GEN8_OA_TIMER_ENABLE : 0) | + GEN8_OA_COUNTER_RESUME; } - CTX_REG(reg_state, - CTX_R_PWR_CLK_STATE, GEN8_R_PWR_CLK_STATE, - intel_sseu_make_rpcs(i915, &ce->sseu)); + for (i = 0; !!ctx_flexeu0 && i < ARRAY_SIZE(flex_regs); i++) + reg_state[ctx_flexeu0 + i * 2 + 1] = + oa_config_flex_reg(stream->oa_config, flex_regs[i]); + + reg_state[CTX_R_PWR_CLK_STATE] = + intel_sseu_make_rpcs(ce->engine->i915, &ce->sseu); } struct flex { @@ -1728,7 +2120,7 @@ gen8_store_flex(struct i915_request *rq, offset = i915_ggtt_offset(ce->state) + LRC_STATE_PN * PAGE_SIZE; do { *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; - *cs++ = offset + (flex->offset + 1) * sizeof(u32); + *cs++ = offset + flex->offset * sizeof(u32); *cs++ = 0; *cs++ = flex->value; } while (flex++, --count); @@ -1832,6 +2224,36 @@ static int gen8_configure_context(struct i915_gem_context *ctx, return err; } +static int gen12_emit_oar_config(struct intel_context *ce, bool enable) +{ + struct i915_request *rq; + u32 *cs; + int err = 0; + + rq = i915_request_create(ce); + if (IS_ERR(rq)) + return PTR_ERR(rq); + + cs = intel_ring_begin(rq, 4); + if (IS_ERR(cs)) { + err = PTR_ERR(cs); + goto out; + } + + *cs++ = MI_LOAD_REGISTER_IMM(1); + *cs++ = i915_mmio_reg_offset(RING_CONTEXT_CONTROL(ce->engine->mmio_base)); + *cs++ = _MASKED_FIELD(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE, + enable ? GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE : 0); + *cs++ = MI_NOOP; + + intel_ring_advance(rq, cs); + +out: + i915_request_add(rq); + + return err; +} + /* * Manages updating the per-context aspects of the OA stream * configuration across all contexts. @@ -1856,24 +2278,22 @@ static int gen8_configure_context(struct i915_gem_context *ctx, * * Note: it's only the RCS/Render context that has any OA state. */ -static int gen8_configure_all_contexts(struct i915_perf_stream *stream, - const struct i915_oa_config *oa_config) +static int lrc_configure_all_contexts(struct i915_perf_stream *stream, + const struct i915_oa_config *oa_config) { - struct drm_i915_private *i915 = stream->dev_priv; + struct drm_i915_private *i915 = stream->perf->i915; /* The MMIO offsets for Flex EU registers aren't contiguous */ - const u32 ctx_flexeu0 = i915->perf.ctx_flexeu0_offset; -#define ctx_flexeuN(N) (ctx_flexeu0 + 2 * (N)) + const u32 ctx_flexeu0 = stream->perf->ctx_flexeu0_offset; +#define ctx_flexeuN(N) (ctx_flexeu0 + 2 * (N) + 1) struct flex regs[] = { { GEN8_R_PWR_CLK_STATE, CTX_R_PWR_CLK_STATE, }, { - GEN8_OACTXCONTROL, - i915->perf.ctx_oactxctrl_offset, - ((stream->period_exponent << GEN8_OA_TIMER_PERIOD_SHIFT) | - (stream->periodic ? GEN8_OA_TIMER_ENABLE : 0) | - GEN8_OA_COUNTER_RESUME) + IS_GEN(i915, 12) ? + GEN12_OAR_OACONTROL : GEN8_OACTXCONTROL, + stream->perf->ctx_oactxctrl_offset + 1, }, { EU_PERF_CNTL0, ctx_flexeuN(0) }, { EU_PERF_CNTL1, ctx_flexeuN(1) }, @@ -1885,13 +2305,27 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream, }; #undef ctx_flexeuN struct intel_engine_cs *engine; - struct i915_gem_context *ctx; - int i; + struct i915_gem_context *ctx, *cn; + size_t array_size = IS_GEN(i915, 12) ? 2 : ARRAY_SIZE(regs); + int i, err; - for (i = 2; i < ARRAY_SIZE(regs); i++) + if (IS_GEN(i915, 12)) { + u32 format = stream->oa_buffer.format; + + regs[1].value = + (format << GEN12_OAR_OACONTROL_COUNTER_FORMAT_SHIFT) | + (oa_config ? GEN12_OAR_OACONTROL_COUNTER_ENABLE : 0); + } else { + regs[1].value = + (stream->period_exponent << GEN8_OA_TIMER_PERIOD_SHIFT) | + (stream->periodic ? GEN8_OA_TIMER_ENABLE : 0) | + GEN8_OA_COUNTER_RESUME; + } + + for (i = 2; !!ctx_flexeu0 && i < array_size; i++) regs[i].value = oa_config_flex_reg(oa_config, regs[i].reg); - lockdep_assert_held(&i915->drm.struct_mutex); + lockdep_assert_held(&stream->perf->lock); /* * The OA register config is setup through the context image. This image @@ -1909,16 +2343,27 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream, * context. Contexts idle at the time of reconfiguration are not * trapped behind the barrier. */ - list_for_each_entry(ctx, &i915->contexts.list, link) { - int err; - + spin_lock(&i915->gem.contexts.lock); + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) { if (ctx == i915->kernel_context) continue; - err = gen8_configure_context(ctx, regs, ARRAY_SIZE(regs)); - if (err) + if (!kref_get_unless_zero(&ctx->ref)) + continue; + + spin_unlock(&i915->gem.contexts.lock); + + err = gen8_configure_context(ctx, regs, array_size); + if (err) { + i915_gem_context_put(ctx); return err; + } + + spin_lock(&i915->gem.contexts.lock); + list_safe_reset_next(ctx, cn, link); + i915_gem_context_put(ctx); } + spin_unlock(&i915->gem.contexts.lock); /* * After updating all other contexts, we need to modify ourselves. @@ -1927,14 +2372,13 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream, */ for_each_uabi_engine(engine, i915) { struct intel_context *ce = engine->kernel_context; - int err; if (engine->class != RENDER_CLASS) continue; regs[0].value = intel_sseu_make_rpcs(i915, &ce->sseu); - err = gen8_modify_self(ce, regs, ARRAY_SIZE(regs)); + err = gen8_modify_self(ce, regs, array_size); if (err) return err; } @@ -1944,8 +2388,8 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream, static int gen8_enable_metric_set(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; - const struct i915_oa_config *oa_config = stream->oa_config; + struct intel_uncore *uncore = stream->uncore; + struct i915_oa_config *oa_config = stream->oa_config; int ret; /* @@ -1971,10 +2415,10 @@ static int gen8_enable_metric_set(struct i915_perf_stream *stream) * be read back from automatically triggered reports, as part of the * RPT_ID field. */ - if (IS_GEN_RANGE(dev_priv, 9, 11)) { - I915_WRITE(GEN8_OA_DEBUG, - _MASKED_BIT_ENABLE(GEN9_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS | - GEN9_OA_DEBUG_INCLUDE_CLK_RATIO)); + if (IS_GEN_RANGE(stream->perf->i915, 9, 11)) { + intel_uncore_write(uncore, GEN8_OA_DEBUG, + _MASKED_BIT_ENABLE(GEN9_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS | + GEN9_OA_DEBUG_INCLUDE_CLK_RATIO)); } /* @@ -1982,45 +2426,102 @@ static int gen8_enable_metric_set(struct i915_perf_stream *stream) * to make sure all slices/subslices are ON before writing to NOA * registers. */ - ret = gen8_configure_all_contexts(stream, oa_config); + ret = lrc_configure_all_contexts(stream, oa_config); if (ret) return ret; - config_oa_regs(dev_priv, oa_config->mux_regs, oa_config->mux_regs_len); - delay_after_mux(); + return emit_oa_config(stream, oa_config, oa_context(stream)); +} + +static int gen12_enable_metric_set(struct i915_perf_stream *stream) +{ + struct intel_uncore *uncore = stream->uncore; + struct i915_oa_config *oa_config = stream->oa_config; + bool periodic = stream->periodic; + u32 period_exponent = stream->period_exponent; + int ret; - config_oa_regs(dev_priv, oa_config->b_counter_regs, - oa_config->b_counter_regs_len); + intel_uncore_write(uncore, GEN12_OAG_OA_DEBUG, + /* Disable clk ratio reports, like previous Gens. */ + _MASKED_BIT_ENABLE(GEN12_OAG_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS | + GEN12_OAG_OA_DEBUG_INCLUDE_CLK_RATIO) | + /* + * If the user didn't require OA reports, instruct the + * hardware not to emit ctx switch reports. + */ + !(stream->sample_flags & SAMPLE_OA_REPORT) ? + _MASKED_BIT_ENABLE(GEN12_OAG_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS) : + _MASKED_BIT_DISABLE(GEN12_OAG_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS)); + + intel_uncore_write(uncore, GEN12_OAG_OAGLBCTXCTRL, periodic ? + (GEN12_OAG_OAGLBCTXCTRL_COUNTER_RESUME | + GEN12_OAG_OAGLBCTXCTRL_TIMER_ENABLE | + (period_exponent << GEN12_OAG_OAGLBCTXCTRL_TIMER_PERIOD_SHIFT)) + : 0); - return 0; + /* + * Update all contexts prior writing the mux configurations as we need + * to make sure all slices/subslices are ON before writing to NOA + * registers. + */ + ret = lrc_configure_all_contexts(stream, oa_config); + if (ret) + return ret; + + /* + * For Gen12, performance counters are context + * saved/restored. Only enable it for the context that + * requested this. + */ + if (stream->ctx) { + ret = gen12_emit_oar_config(stream->pinned_ctx, + oa_config != NULL); + if (ret) + return ret; + } + + return emit_oa_config(stream, oa_config, oa_context(stream)); } static void gen8_disable_metric_set(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; /* Reset all contexts' slices/subslices configurations. */ - gen8_configure_all_contexts(stream, NULL); + lrc_configure_all_contexts(stream, NULL); - I915_WRITE(GDT_CHICKEN_BITS, (I915_READ(GDT_CHICKEN_BITS) & - ~GT_NOA_ENABLE)); + intel_uncore_rmw(uncore, GDT_CHICKEN_BITS, GT_NOA_ENABLE, 0); } static void gen10_disable_metric_set(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; + + /* Reset all contexts' slices/subslices configurations. */ + lrc_configure_all_contexts(stream, NULL); + + /* Make sure we disable noa to save power. */ + intel_uncore_rmw(uncore, RPM_CONFIG1, GEN10_GT_NOA_ENABLE, 0); +} + +static void gen12_disable_metric_set(struct i915_perf_stream *stream) +{ + struct intel_uncore *uncore = stream->uncore; /* Reset all contexts' slices/subslices configurations. */ - gen8_configure_all_contexts(stream, NULL); + lrc_configure_all_contexts(stream, NULL); + + /* disable the context save/restore or OAR counters */ + if (stream->ctx) + gen12_emit_oar_config(stream->pinned_ctx, false); /* Make sure we disable noa to save power. */ - I915_WRITE(RPM_CONFIG1, - I915_READ(RPM_CONFIG1) & ~GEN10_GT_NOA_ENABLE); + intel_uncore_rmw(uncore, RPM_CONFIG1, GEN10_GT_NOA_ENABLE, 0); } static void gen7_oa_enable(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; struct i915_gem_context *ctx = stream->ctx; u32 ctx_id = stream->specific_ctx_id; bool periodic = stream->periodic; @@ -2038,19 +2539,19 @@ static void gen7_oa_enable(struct i915_perf_stream *stream) */ gen7_init_oa_buffer(stream); - I915_WRITE(GEN7_OACONTROL, - (ctx_id & GEN7_OACONTROL_CTX_MASK) | - (period_exponent << - GEN7_OACONTROL_TIMER_PERIOD_SHIFT) | - (periodic ? GEN7_OACONTROL_TIMER_ENABLE : 0) | - (report_format << GEN7_OACONTROL_FORMAT_SHIFT) | - (ctx ? GEN7_OACONTROL_PER_CTX_ENABLE : 0) | - GEN7_OACONTROL_ENABLE); + intel_uncore_write(uncore, GEN7_OACONTROL, + (ctx_id & GEN7_OACONTROL_CTX_MASK) | + (period_exponent << + GEN7_OACONTROL_TIMER_PERIOD_SHIFT) | + (periodic ? GEN7_OACONTROL_TIMER_ENABLE : 0) | + (report_format << GEN7_OACONTROL_FORMAT_SHIFT) | + (ctx ? GEN7_OACONTROL_PER_CTX_ENABLE : 0) | + GEN7_OACONTROL_ENABLE); } static void gen8_oa_enable(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct intel_uncore *uncore = stream->uncore; u32 report_format = stream->oa_buffer.format; /* @@ -2069,9 +2570,28 @@ static void gen8_oa_enable(struct i915_perf_stream *stream) * filtering and instead filter on the cpu based on the context-id * field of reports */ - I915_WRITE(GEN8_OACONTROL, (report_format << - GEN8_OA_REPORT_FORMAT_SHIFT) | - GEN8_OA_COUNTER_ENABLE); + intel_uncore_write(uncore, GEN8_OACONTROL, + (report_format << GEN8_OA_REPORT_FORMAT_SHIFT) | + GEN8_OA_COUNTER_ENABLE); +} + +static void gen12_oa_enable(struct i915_perf_stream *stream) +{ + struct intel_uncore *uncore = stream->uncore; + u32 report_format = stream->oa_buffer.format; + + /* + * If we don't want OA reports from the OA buffer, then we don't even + * need to program the OAG unit. + */ + if (!(stream->sample_flags & SAMPLE_OA_REPORT)) + return; + + gen12_init_oa_buffer(stream); + + intel_uncore_write(uncore, GEN12_OAG_OACONTROL, + (report_format << GEN12_OAG_OACONTROL_OA_COUNTER_FORMAT_SHIFT) | + GEN12_OAG_OACONTROL_OA_COUNTER_ENABLE); } /** @@ -2085,9 +2605,7 @@ static void gen8_oa_enable(struct i915_perf_stream *stream) */ static void i915_oa_stream_enable(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; - - dev_priv->perf.ops.oa_enable(stream); + stream->perf->ops.oa_enable(stream); if (stream->periodic) hrtimer_start(&stream->poll_check_timer, @@ -2097,7 +2615,7 @@ static void i915_oa_stream_enable(struct i915_perf_stream *stream) static void gen7_oa_disable(struct i915_perf_stream *stream) { - struct intel_uncore *uncore = &stream->dev_priv->uncore; + struct intel_uncore *uncore = stream->uncore; intel_uncore_write(uncore, GEN7_OACONTROL, 0); if (intel_wait_for_register(uncore, @@ -2108,7 +2626,7 @@ static void gen7_oa_disable(struct i915_perf_stream *stream) static void gen8_oa_disable(struct i915_perf_stream *stream) { - struct intel_uncore *uncore = &stream->dev_priv->uncore; + struct intel_uncore *uncore = stream->uncore; intel_uncore_write(uncore, GEN8_OACONTROL, 0); if (intel_wait_for_register(uncore, @@ -2117,6 +2635,18 @@ static void gen8_oa_disable(struct i915_perf_stream *stream) DRM_ERROR("wait for OA to be disabled timed out\n"); } +static void gen12_oa_disable(struct i915_perf_stream *stream) +{ + struct intel_uncore *uncore = stream->uncore; + + intel_uncore_write(uncore, GEN12_OAG_OACONTROL, 0); + if (intel_wait_for_register(uncore, + GEN12_OAG_OACONTROL, + GEN12_OAG_OACONTROL_OA_COUNTER_ENABLE, 0, + 50)) + DRM_ERROR("wait for OA to be disabled timed out\n"); +} + /** * i915_oa_stream_disable - handle `I915_PERF_IOCTL_DISABLE` for OA stream * @stream: An i915 perf stream opened for OA metrics @@ -2127,9 +2657,7 @@ static void gen8_oa_disable(struct i915_perf_stream *stream) */ static void i915_oa_stream_disable(struct i915_perf_stream *stream) { - struct drm_i915_private *dev_priv = stream->dev_priv; - - dev_priv->perf.ops.oa_disable(stream); + stream->perf->ops.oa_disable(stream); if (stream->periodic) hrtimer_cancel(&stream->poll_check_timer); @@ -2166,15 +2694,21 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, struct drm_i915_perf_open_param *param, struct perf_open_properties *props) { - struct drm_i915_private *dev_priv = stream->dev_priv; + struct i915_perf *perf = stream->perf; int format_size; int ret; - /* If the sysfs metrics/ directory wasn't registered for some + if (!props->engine) { + DRM_DEBUG("OA engine not specified\n"); + return -EINVAL; + } + + /* + * If the sysfs metrics/ directory wasn't registered for some * reason then don't let userspace try their luck with config * IDs */ - if (!dev_priv->perf.metrics_kobj) { + if (!perf->metrics_kobj) { DRM_DEBUG("OA metrics weren't advertised via sysfs\n"); return -EINVAL; } @@ -2184,16 +2718,17 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, return -EINVAL; } - if (!dev_priv->perf.ops.enable_metric_set) { + if (!perf->ops.enable_metric_set) { DRM_DEBUG("OA unit not supported\n"); return -ENODEV; } - /* To avoid the complexity of having to accurately filter + /* + * To avoid the complexity of having to accurately filter * counter reports and marshal to the appropriate client * we currently only allow exclusive access */ - if (dev_priv->perf.exclusive_stream) { + if (perf->exclusive_stream) { DRM_DEBUG("OA unit already in use\n"); return -EBUSY; } @@ -2203,9 +2738,12 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, return -EINVAL; } + stream->engine = props->engine; + stream->uncore = stream->engine->gt->uncore; + stream->sample_size = sizeof(struct drm_i915_perf_record_header); - format_size = dev_priv->perf.oa_formats[props->oa_format].size; + format_size = perf->oa_formats[props->oa_format].size; stream->sample_flags |= SAMPLE_OA_REPORT; stream->sample_size += format_size; @@ -2214,8 +2752,10 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, if (WARN_ON(stream->oa_buffer.format_size == 0)) return -EINVAL; + stream->hold_preemption = props->hold_preemption; + stream->oa_buffer.format = - dev_priv->perf.oa_formats[props->oa_format].format; + perf->oa_formats[props->oa_format].format; stream->periodic = props->oa_periodic; if (stream->periodic) @@ -2229,9 +2769,16 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, } } - ret = get_oa_config(dev_priv, props->metrics_set, &stream->oa_config); + ret = alloc_noa_wait(stream); if (ret) { + DRM_DEBUG("Unable to allocate NOA wait batch buffer\n"); + goto err_noa_wait_alloc; + } + + stream->oa_config = i915_perf_get_oa_config(perf, props->metrics_set); + if (!stream->oa_config) { DRM_DEBUG("Invalid OA config id=%i\n", props->metrics_set); + ret = -EINVAL; goto err_config; } @@ -2247,27 +2794,24 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, * In our case we are expecting that taking pm + FORCEWAKE * references will effectively disable RC6. */ - stream->wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); + intel_engine_pm_get(stream->engine); + intel_uncore_forcewake_get(stream->uncore, FORCEWAKE_ALL); ret = alloc_oa_buffer(stream); if (ret) goto err_oa_buf_alloc; - ret = i915_mutex_lock_interruptible(&dev_priv->drm); - if (ret) - goto err_lock; - stream->ops = &i915_oa_stream_ops; - dev_priv->perf.exclusive_stream = stream; + perf->exclusive_stream = stream; - ret = dev_priv->perf.ops.enable_metric_set(stream); + ret = perf->ops.enable_metric_set(stream); if (ret) { DRM_DEBUG("Unable to enable metric set\n"); goto err_enable; } - mutex_unlock(&dev_priv->drm.struct_mutex); + DRM_DEBUG("opening stream oa config uuid=%s\n", + stream->oa_config->uuid); hrtimer_init(&stream->poll_check_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); @@ -2278,38 +2822,40 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, return 0; err_enable: - dev_priv->perf.exclusive_stream = NULL; - dev_priv->perf.ops.disable_metric_set(stream); - mutex_unlock(&dev_priv->drm.struct_mutex); + perf->exclusive_stream = NULL; + perf->ops.disable_metric_set(stream); -err_lock: free_oa_buffer(stream); err_oa_buf_alloc: - put_oa_config(dev_priv, stream->oa_config); + free_oa_configs(stream); - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); - intel_runtime_pm_put(&dev_priv->runtime_pm, stream->wakeref); + intel_uncore_forcewake_put(stream->uncore, FORCEWAKE_ALL); + intel_engine_pm_put(stream->engine); err_config: + free_noa_wait(stream); + +err_noa_wait_alloc: if (stream->ctx) oa_put_render_ctx_id(stream); return ret; } -void i915_oa_init_reg_state(struct intel_engine_cs *engine, - struct intel_context *ce, - u32 *regs) +void i915_oa_init_reg_state(const struct intel_context *ce, + const struct intel_engine_cs *engine) { struct i915_perf_stream *stream; + /* perf.exclusive_stream serialised by lrc_configure_all_contexts() */ + if (engine->class != RENDER_CLASS) return; stream = engine->i915->perf.exclusive_stream; if (stream) - gen8_update_reg_state_unlocked(stream, ce, regs, stream->oa_config); + gen8_update_reg_state_unlocked(ce, stream); } /** @@ -2379,7 +2925,7 @@ static ssize_t i915_perf_read(struct file *file, loff_t *ppos) { struct i915_perf_stream *stream = file->private_data; - struct drm_i915_private *dev_priv = stream->dev_priv; + struct i915_perf *perf = stream->perf; ssize_t ret; /* To ensure it's handled consistently we simply treat all reads of a @@ -2402,15 +2948,15 @@ static ssize_t i915_perf_read(struct file *file, if (ret) return ret; - mutex_lock(&dev_priv->perf.lock); + mutex_lock(&perf->lock); ret = i915_perf_read_locked(stream, file, buf, count, ppos); - mutex_unlock(&dev_priv->perf.lock); + mutex_unlock(&perf->lock); } while (ret == -EAGAIN); } else { - mutex_lock(&dev_priv->perf.lock); + mutex_lock(&perf->lock); ret = i915_perf_read_locked(stream, file, buf, count, ppos); - mutex_unlock(&dev_priv->perf.lock); + mutex_unlock(&perf->lock); } /* We allow the poll checking to sometimes report false positive EPOLLIN @@ -2448,7 +2994,6 @@ static enum hrtimer_restart oa_poll_check_timer_cb(struct hrtimer *hrtimer) /** * i915_perf_poll_locked - poll_wait() with a suitable wait queue for stream - * @dev_priv: i915 device instance * @stream: An i915 perf stream * @file: An i915 perf stream file * @wait: poll() state table @@ -2457,15 +3002,14 @@ static enum hrtimer_restart oa_poll_check_timer_cb(struct hrtimer *hrtimer) * &i915_perf_stream_ops->poll_wait to call poll_wait() with a wait queue that * will be woken for new stream data. * - * Note: The &drm_i915_private->perf.lock mutex has been taken to serialize + * Note: The &perf->lock mutex has been taken to serialize * with any non-file-operation driver hooks. * * Returns: any poll events that are ready without sleeping */ -static __poll_t i915_perf_poll_locked(struct drm_i915_private *dev_priv, - struct i915_perf_stream *stream, - struct file *file, - poll_table *wait) +static __poll_t i915_perf_poll_locked(struct i915_perf_stream *stream, + struct file *file, + poll_table *wait) { __poll_t events = 0; @@ -2499,12 +3043,12 @@ static __poll_t i915_perf_poll_locked(struct drm_i915_private *dev_priv, static __poll_t i915_perf_poll(struct file *file, poll_table *wait) { struct i915_perf_stream *stream = file->private_data; - struct drm_i915_private *dev_priv = stream->dev_priv; + struct i915_perf *perf = stream->perf; __poll_t ret; - mutex_lock(&dev_priv->perf.lock); - ret = i915_perf_poll_locked(dev_priv, stream, file, wait); - mutex_unlock(&dev_priv->perf.lock); + mutex_lock(&perf->lock); + ret = i915_perf_poll_locked(stream, file, wait); + mutex_unlock(&perf->lock); return ret; } @@ -2529,6 +3073,9 @@ static void i915_perf_enable_locked(struct i915_perf_stream *stream) if (stream->ops->enable) stream->ops->enable(stream); + + if (stream->hold_preemption) + i915_gem_context_set_nopreempt(stream->ctx); } /** @@ -2553,17 +3100,54 @@ static void i915_perf_disable_locked(struct i915_perf_stream *stream) /* Allow stream->ops->disable() to refer to this */ stream->enabled = false; + if (stream->hold_preemption) + i915_gem_context_clear_nopreempt(stream->ctx); + if (stream->ops->disable) stream->ops->disable(stream); } +static long i915_perf_config_locked(struct i915_perf_stream *stream, + unsigned long metrics_set) +{ + struct i915_oa_config *config; + long ret = stream->oa_config->id; + + config = i915_perf_get_oa_config(stream->perf, metrics_set); + if (!config) + return -EINVAL; + + if (config != stream->oa_config) { + int err; + + /* + * If OA is bound to a specific context, emit the + * reconfiguration inline from that context. The update + * will then be ordered with respect to submission on that + * context. + * + * When set globally, we use a low priority kernel context, + * so it will effectively take effect when idle. + */ + err = emit_oa_config(stream, config, oa_context(stream)); + if (err == 0) + config = xchg(&stream->oa_config, config); + else + ret = err; + } + + i915_oa_config_put(config); + + return ret; +} + /** * i915_perf_ioctl - support ioctl() usage with i915 perf stream FDs * @stream: An i915 perf stream * @cmd: the ioctl request * @arg: the ioctl data * - * Note: The &drm_i915_private->perf.lock mutex has been taken to serialize + * Note: The &perf->lock mutex has been taken to serialize * with any non-file-operation driver hooks. * * Returns: zero on success or a negative error code. Returns -EINVAL for @@ -2580,6 +3164,8 @@ static long i915_perf_ioctl_locked(struct i915_perf_stream *stream, case I915_PERF_IOCTL_DISABLE: i915_perf_disable_locked(stream); return 0; + case I915_PERF_IOCTL_CONFIG: + return i915_perf_config_locked(stream, arg); } return -EINVAL; @@ -2601,12 +3187,12 @@ static long i915_perf_ioctl(struct file *file, unsigned long arg) { struct i915_perf_stream *stream = file->private_data; - struct drm_i915_private *dev_priv = stream->dev_priv; + struct i915_perf *perf = stream->perf; long ret; - mutex_lock(&dev_priv->perf.lock); + mutex_lock(&perf->lock); ret = i915_perf_ioctl_locked(stream, cmd, arg); - mutex_unlock(&dev_priv->perf.lock); + mutex_unlock(&perf->lock); return ret; } @@ -2618,7 +3204,7 @@ static long i915_perf_ioctl(struct file *file, * Frees all resources associated with the given i915 perf @stream, disabling * any associated data capture in the process. * - * Note: The &drm_i915_private->perf.lock mutex has been taken to serialize + * Note: The &perf->lock mutex has been taken to serialize * with any non-file-operation driver hooks. */ static void i915_perf_destroy_locked(struct i915_perf_stream *stream) @@ -2629,8 +3215,6 @@ static void i915_perf_destroy_locked(struct i915_perf_stream *stream) if (stream->ops->destroy) stream->ops->destroy(stream); - list_del(&stream->link); - if (stream->ctx) i915_gem_context_put(stream->ctx); @@ -2651,14 +3235,14 @@ static void i915_perf_destroy_locked(struct i915_perf_stream *stream) static int i915_perf_release(struct inode *inode, struct file *file) { struct i915_perf_stream *stream = file->private_data; - struct drm_i915_private *dev_priv = stream->dev_priv; + struct i915_perf *perf = stream->perf; - mutex_lock(&dev_priv->perf.lock); + mutex_lock(&perf->lock); i915_perf_destroy_locked(stream); - mutex_unlock(&dev_priv->perf.lock); + mutex_unlock(&perf->lock); /* Release the reference the perf stream kept on the driver. */ - drm_dev_put(&dev_priv->drm); + drm_dev_put(&perf->i915->drm); return 0; } @@ -2680,7 +3264,7 @@ static const struct file_operations fops = { /** * i915_perf_open_ioctl_locked - DRM ioctl() for userspace to open a stream FD - * @dev_priv: i915 device instance + * @perf: i915 perf instance * @param: The open parameters passed to 'DRM_I915_PERF_OPEN` * @props: individually validated u64 property value pairs * @file: drm file @@ -2688,7 +3272,7 @@ static const struct file_operations fops = { * See i915_perf_ioctl_open() for interface details. * * Implements further stream config validation and stream initialization on - * behalf of i915_perf_open_ioctl() with the &drm_i915_private->perf.lock mutex + * behalf of i915_perf_open_ioctl() with the &perf->lock mutex * taken to serialize with any non-file-operation driver hooks. * * Note: at this point the @props have only been validated in isolation and @@ -2703,7 +3287,7 @@ static const struct file_operations fops = { * Returns: zero on success or a negative error code. */ static int -i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, +i915_perf_open_ioctl_locked(struct i915_perf *perf, struct drm_i915_perf_open_param *param, struct perf_open_properties *props, struct drm_file *file) @@ -2734,17 +3318,34 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, * rest of the system, which we consider acceptable for a * non-privileged client. * - * For Gen8+ the OA unit no longer supports clock gating off for a + * For Gen8->11 the OA unit no longer supports clock gating off for a * specific context and the kernel can't securely stop the counters * from updating as system-wide / global values. Even though we can * filter reports based on the included context ID we can't block * clients from seeing the raw / global counter values via * MI_REPORT_PERF_COUNT commands and so consider it a privileged op to * enable the OA unit by default. + * + * For Gen12+ we gain a new OAR unit that only monitors the RCS on a + * per context basis. So we can relax requirements there if the user + * doesn't request global stream access (i.e. query based sampling + * using MI_RECORD_PERF_COUNT. */ - if (IS_HASWELL(dev_priv) && specific_ctx) + if (IS_HASWELL(perf->i915) && specific_ctx) + privileged_op = false; + else if (IS_GEN(perf->i915, 12) && specific_ctx && + (props->sample_flags & SAMPLE_OA_REPORT) == 0) privileged_op = false; + if (props->hold_preemption) { + if (!props->single_context) { + DRM_DEBUG("preemption disable with no context\n"); + ret = -EINVAL; + goto err; + } + privileged_op = true; + } + /* Similar to perf's kernel.perf_paranoid_cpu sysctl option * we check a dev.i915.perf_stream_paranoid sysctl option * to determine if it's ok to access system wide OA counters @@ -2752,7 +3353,7 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, */ if (privileged_op && i915_perf_stream_paranoid && !capable(CAP_SYS_ADMIN)) { - DRM_DEBUG("Insufficient privileges to open system-wide i915 perf stream\n"); + DRM_DEBUG("Insufficient privileges to open i915 perf stream\n"); ret = -EACCES; goto err_ctx; } @@ -2763,7 +3364,7 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, goto err_ctx; } - stream->dev_priv = dev_priv; + stream->perf = perf; stream->ctx = specific_ctx; ret = i915_oa_stream_init(stream, param, props); @@ -2779,8 +3380,6 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, goto err_flags; } - list_add(&stream->link, &dev_priv->perf.streams); - if (param->flags & I915_PERF_FLAG_FD_CLOEXEC) f_flags |= O_CLOEXEC; if (param->flags & I915_PERF_FLAG_FD_NONBLOCK) @@ -2789,7 +3388,7 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, stream_fd = anon_inode_getfd("[i915_perf]", &fops, stream, f_flags); if (stream_fd < 0) { ret = stream_fd; - goto err_open; + goto err_flags; } if (!(param->flags & I915_PERF_FLAG_DISABLED)) @@ -2798,12 +3397,10 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, /* Take a reference on the driver that will be kept with stream_fd * until its release. */ - drm_dev_get(&dev_priv->drm); + drm_dev_get(&perf->i915->drm); return stream_fd; -err_open: - list_del(&stream->link); err_flags: if (stream->ops->destroy) stream->ops->destroy(stream); @@ -2816,15 +3413,15 @@ err: return ret; } -static u64 oa_exponent_to_ns(struct drm_i915_private *dev_priv, int exponent) +static u64 oa_exponent_to_ns(struct i915_perf *perf, int exponent) { return div64_u64(1000000000ULL * (2ULL << exponent), - 1000ULL * RUNTIME_INFO(dev_priv)->cs_timestamp_frequency_khz); + 1000ULL * RUNTIME_INFO(perf->i915)->cs_timestamp_frequency_khz); } /** * read_properties_unlocked - validate + copy userspace stream open properties - * @dev_priv: i915 device instance + * @perf: i915 perf instance * @uprops: The array of u64 key value pairs given by userspace * @n_props: The number of key value pairs expected in @uprops * @props: The stream configuration built up while validating properties @@ -2837,7 +3434,7 @@ static u64 oa_exponent_to_ns(struct drm_i915_private *dev_priv, int exponent) * we shouldn't validate or assume anything about ordering here. This doesn't * rule out defining new properties with ordering requirements in the future. */ -static int read_properties_unlocked(struct drm_i915_private *dev_priv, +static int read_properties_unlocked(struct i915_perf *perf, u64 __user *uprops, u32 n_props, struct perf_open_properties *props) @@ -2852,6 +3449,15 @@ static int read_properties_unlocked(struct drm_i915_private *dev_priv, return -EINVAL; } + /* At the moment we only support using i915-perf on the RCS. */ + props->engine = intel_engine_lookup_user(perf->i915, + I915_ENGINE_CLASS_RENDER, + 0); + if (!props->engine) { + DRM_DEBUG("No RENDER-capable engines\n"); + return -EINVAL; + } + /* Considering that ID = 0 is reserved and assuming that we don't * (currently) expect any configurations to ever specify duplicate * values for a particular property ID then the last _PROP_MAX value is @@ -2903,7 +3509,7 @@ static int read_properties_unlocked(struct drm_i915_private *dev_priv, value); return -EINVAL; } - if (!dev_priv->perf.oa_formats[value].size) { + if (!perf->oa_formats[value].size) { DRM_DEBUG("Unsupported OA report format %llu\n", value); return -EINVAL; @@ -2924,7 +3530,7 @@ static int read_properties_unlocked(struct drm_i915_private *dev_priv, */ BUILD_BUG_ON(sizeof(oa_period) != 8); - oa_period = oa_exponent_to_ns(dev_priv, value); + oa_period = oa_exponent_to_ns(perf, value); /* This check is primarily to ensure that oa_period <= * UINT32_MAX (before passing to do_div which only @@ -2949,6 +3555,9 @@ static int read_properties_unlocked(struct drm_i915_private *dev_priv, props->oa_periodic = true; props->oa_period_exponent = value; break; + case DRM_I915_PERF_PROP_HOLD_PREEMPTION: + props->hold_preemption = !!value; + break; case DRM_I915_PERF_PROP_MAX: MISSING_CASE(id); return -EINVAL; @@ -2978,7 +3587,7 @@ static int read_properties_unlocked(struct drm_i915_private *dev_priv, * mutex to avoid an awkward lockdep with mmap_sem. * * Most of the implementation details are handled by - * i915_perf_open_ioctl_locked() after taking the &drm_i915_private->perf.lock + * i915_perf_open_ioctl_locked() after taking the &perf->lock * mutex for serializing with any non-file-operation driver hooks. * * Return: A newly opened i915 Perf stream file descriptor or negative @@ -2987,13 +3596,13 @@ static int read_properties_unlocked(struct drm_i915_private *dev_priv, int i915_perf_open_ioctl(struct drm_device *dev, void *data, struct drm_file *file) { - struct drm_i915_private *dev_priv = dev->dev_private; + struct i915_perf *perf = &to_i915(dev)->perf; struct drm_i915_perf_open_param *param = data; struct perf_open_properties props; u32 known_open_flags; int ret; - if (!dev_priv->perf.initialized) { + if (!perf->i915) { DRM_DEBUG("i915 perf interface not available for this system\n"); return -ENOTSUPP; } @@ -3006,124 +3615,130 @@ int i915_perf_open_ioctl(struct drm_device *dev, void *data, return -EINVAL; } - ret = read_properties_unlocked(dev_priv, + ret = read_properties_unlocked(perf, u64_to_user_ptr(param->properties_ptr), param->num_properties, &props); if (ret) return ret; - mutex_lock(&dev_priv->perf.lock); - ret = i915_perf_open_ioctl_locked(dev_priv, param, &props, file); - mutex_unlock(&dev_priv->perf.lock); + mutex_lock(&perf->lock); + ret = i915_perf_open_ioctl_locked(perf, param, &props, file); + mutex_unlock(&perf->lock); return ret; } /** * i915_perf_register - exposes i915-perf to userspace - * @dev_priv: i915 device instance + * @i915: i915 device instance * * In particular OA metric sets are advertised under a sysfs metrics/ * directory allowing userspace to enumerate valid IDs that can be * used to open an i915-perf stream. */ -void i915_perf_register(struct drm_i915_private *dev_priv) +void i915_perf_register(struct drm_i915_private *i915) { + struct i915_perf *perf = &i915->perf; int ret; - if (!dev_priv->perf.initialized) + if (!perf->i915) return; /* To be sure we're synchronized with an attempted * i915_perf_open_ioctl(); considering that we register after * being exposed to userspace. */ - mutex_lock(&dev_priv->perf.lock); + mutex_lock(&perf->lock); - dev_priv->perf.metrics_kobj = + perf->metrics_kobj = kobject_create_and_add("metrics", - &dev_priv->drm.primary->kdev->kobj); - if (!dev_priv->perf.metrics_kobj) + &i915->drm.primary->kdev->kobj); + if (!perf->metrics_kobj) goto exit; - sysfs_attr_init(&dev_priv->perf.test_config.sysfs_metric_id.attr); - - if (INTEL_GEN(dev_priv) >= 11) { - i915_perf_load_test_config_icl(dev_priv); - } else if (IS_CANNONLAKE(dev_priv)) { - i915_perf_load_test_config_cnl(dev_priv); - } else if (IS_COFFEELAKE(dev_priv)) { - if (IS_CFL_GT2(dev_priv)) - i915_perf_load_test_config_cflgt2(dev_priv); - if (IS_CFL_GT3(dev_priv)) - i915_perf_load_test_config_cflgt3(dev_priv); - } else if (IS_GEMINILAKE(dev_priv)) { - i915_perf_load_test_config_glk(dev_priv); - } else if (IS_KABYLAKE(dev_priv)) { - if (IS_KBL_GT2(dev_priv)) - i915_perf_load_test_config_kblgt2(dev_priv); - else if (IS_KBL_GT3(dev_priv)) - i915_perf_load_test_config_kblgt3(dev_priv); - } else if (IS_BROXTON(dev_priv)) { - i915_perf_load_test_config_bxt(dev_priv); - } else if (IS_SKYLAKE(dev_priv)) { - if (IS_SKL_GT2(dev_priv)) - i915_perf_load_test_config_sklgt2(dev_priv); - else if (IS_SKL_GT3(dev_priv)) - i915_perf_load_test_config_sklgt3(dev_priv); - else if (IS_SKL_GT4(dev_priv)) - i915_perf_load_test_config_sklgt4(dev_priv); - } else if (IS_CHERRYVIEW(dev_priv)) { - i915_perf_load_test_config_chv(dev_priv); - } else if (IS_BROADWELL(dev_priv)) { - i915_perf_load_test_config_bdw(dev_priv); - } else if (IS_HASWELL(dev_priv)) { - i915_perf_load_test_config_hsw(dev_priv); -} - - if (dev_priv->perf.test_config.id == 0) + sysfs_attr_init(&perf->test_config.sysfs_metric_id.attr); + + if (IS_TIGERLAKE(i915)) { + i915_perf_load_test_config_tgl(i915); + } else if (INTEL_GEN(i915) >= 11) { + i915_perf_load_test_config_icl(i915); + } else if (IS_CANNONLAKE(i915)) { + i915_perf_load_test_config_cnl(i915); + } else if (IS_COFFEELAKE(i915)) { + if (IS_CFL_GT2(i915)) + i915_perf_load_test_config_cflgt2(i915); + if (IS_CFL_GT3(i915)) + i915_perf_load_test_config_cflgt3(i915); + } else if (IS_GEMINILAKE(i915)) { + i915_perf_load_test_config_glk(i915); + } else if (IS_KABYLAKE(i915)) { + if (IS_KBL_GT2(i915)) + i915_perf_load_test_config_kblgt2(i915); + else if (IS_KBL_GT3(i915)) + i915_perf_load_test_config_kblgt3(i915); + } else if (IS_BROXTON(i915)) { + i915_perf_load_test_config_bxt(i915); + } else if (IS_SKYLAKE(i915)) { + if (IS_SKL_GT2(i915)) + i915_perf_load_test_config_sklgt2(i915); + else if (IS_SKL_GT3(i915)) + i915_perf_load_test_config_sklgt3(i915); + else if (IS_SKL_GT4(i915)) + i915_perf_load_test_config_sklgt4(i915); + } else if (IS_CHERRYVIEW(i915)) { + i915_perf_load_test_config_chv(i915); + } else if (IS_BROADWELL(i915)) { + i915_perf_load_test_config_bdw(i915); + } else if (IS_HASWELL(i915)) { + i915_perf_load_test_config_hsw(i915); + } + + if (perf->test_config.id == 0) goto sysfs_error; - ret = sysfs_create_group(dev_priv->perf.metrics_kobj, - &dev_priv->perf.test_config.sysfs_metric); + ret = sysfs_create_group(perf->metrics_kobj, + &perf->test_config.sysfs_metric); if (ret) goto sysfs_error; - atomic_set(&dev_priv->perf.test_config.ref_count, 1); + perf->test_config.perf = perf; + kref_init(&perf->test_config.ref); goto exit; sysfs_error: - kobject_put(dev_priv->perf.metrics_kobj); - dev_priv->perf.metrics_kobj = NULL; + kobject_put(perf->metrics_kobj); + perf->metrics_kobj = NULL; exit: - mutex_unlock(&dev_priv->perf.lock); + mutex_unlock(&perf->lock); } /** * i915_perf_unregister - hide i915-perf from userspace - * @dev_priv: i915 device instance + * @i915: i915 device instance * * i915-perf state cleanup is split up into an 'unregister' and * 'deinit' phase where the interface is first hidden from * userspace by i915_perf_unregister() before cleaning up * remaining state in i915_perf_fini(). */ -void i915_perf_unregister(struct drm_i915_private *dev_priv) +void i915_perf_unregister(struct drm_i915_private *i915) { - if (!dev_priv->perf.metrics_kobj) + struct i915_perf *perf = &i915->perf; + + if (!perf->metrics_kobj) return; - sysfs_remove_group(dev_priv->perf.metrics_kobj, - &dev_priv->perf.test_config.sysfs_metric); + sysfs_remove_group(perf->metrics_kobj, + &perf->test_config.sysfs_metric); - kobject_put(dev_priv->perf.metrics_kobj); - dev_priv->perf.metrics_kobj = NULL; + kobject_put(perf->metrics_kobj); + perf->metrics_kobj = NULL; } -static bool gen8_is_valid_flex_addr(struct drm_i915_private *dev_priv, u32 addr) +static bool gen8_is_valid_flex_addr(struct i915_perf *perf, u32 addr) { static const i915_reg_t flex_eu_regs[] = { EU_PERF_CNTL0, @@ -3143,56 +3758,80 @@ static bool gen8_is_valid_flex_addr(struct drm_i915_private *dev_priv, u32 addr) return false; } -static bool gen7_is_valid_b_counter_addr(struct drm_i915_private *dev_priv, u32 addr) +#define ADDR_IN_RANGE(addr, start, end) \ + ((addr) >= (start) && \ + (addr) <= (end)) + +#define REG_IN_RANGE(addr, start, end) \ + ((addr) >= i915_mmio_reg_offset(start) && \ + (addr) <= i915_mmio_reg_offset(end)) + +#define REG_EQUAL(addr, mmio) \ + ((addr) == i915_mmio_reg_offset(mmio)) + +static bool gen7_is_valid_b_counter_addr(struct i915_perf *perf, u32 addr) +{ + return REG_IN_RANGE(addr, OASTARTTRIG1, OASTARTTRIG8) || + REG_IN_RANGE(addr, OAREPORTTRIG1, OAREPORTTRIG8) || + REG_IN_RANGE(addr, OACEC0_0, OACEC7_1); +} + +static bool gen7_is_valid_mux_addr(struct i915_perf *perf, u32 addr) +{ + return REG_EQUAL(addr, HALF_SLICE_CHICKEN2) || + REG_IN_RANGE(addr, MICRO_BP0_0, NOA_WRITE) || + REG_IN_RANGE(addr, OA_PERFCNT1_LO, OA_PERFCNT2_HI) || + REG_IN_RANGE(addr, OA_PERFMATRIX_LO, OA_PERFMATRIX_HI); +} + +static bool gen8_is_valid_mux_addr(struct i915_perf *perf, u32 addr) { - return (addr >= i915_mmio_reg_offset(OASTARTTRIG1) && - addr <= i915_mmio_reg_offset(OASTARTTRIG8)) || - (addr >= i915_mmio_reg_offset(OAREPORTTRIG1) && - addr <= i915_mmio_reg_offset(OAREPORTTRIG8)) || - (addr >= i915_mmio_reg_offset(OACEC0_0) && - addr <= i915_mmio_reg_offset(OACEC7_1)); + return gen7_is_valid_mux_addr(perf, addr) || + REG_EQUAL(addr, WAIT_FOR_RC6_EXIT) || + REG_IN_RANGE(addr, RPM_CONFIG0, NOA_CONFIG(8)); } -static bool gen7_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) +static bool gen10_is_valid_mux_addr(struct i915_perf *perf, u32 addr) { - return addr == i915_mmio_reg_offset(HALF_SLICE_CHICKEN2) || - (addr >= i915_mmio_reg_offset(MICRO_BP0_0) && - addr <= i915_mmio_reg_offset(NOA_WRITE)) || - (addr >= i915_mmio_reg_offset(OA_PERFCNT1_LO) && - addr <= i915_mmio_reg_offset(OA_PERFCNT2_HI)) || - (addr >= i915_mmio_reg_offset(OA_PERFMATRIX_LO) && - addr <= i915_mmio_reg_offset(OA_PERFMATRIX_HI)); + return gen8_is_valid_mux_addr(perf, addr) || + REG_EQUAL(addr, GEN10_NOA_WRITE_HIGH) || + REG_IN_RANGE(addr, OA_PERFCNT3_LO, OA_PERFCNT4_HI); } -static bool gen8_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) +static bool hsw_is_valid_mux_addr(struct i915_perf *perf, u32 addr) { - return gen7_is_valid_mux_addr(dev_priv, addr) || - addr == i915_mmio_reg_offset(WAIT_FOR_RC6_EXIT) || - (addr >= i915_mmio_reg_offset(RPM_CONFIG0) && - addr <= i915_mmio_reg_offset(NOA_CONFIG(8))); + return gen7_is_valid_mux_addr(perf, addr) || + ADDR_IN_RANGE(addr, 0x25100, 0x2FF90) || + REG_IN_RANGE(addr, HSW_MBVID2_NOA0, HSW_MBVID2_NOA9) || + REG_EQUAL(addr, HSW_MBVID2_MISR0); } -static bool gen10_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) +static bool chv_is_valid_mux_addr(struct i915_perf *perf, u32 addr) { - return gen8_is_valid_mux_addr(dev_priv, addr) || - addr == i915_mmio_reg_offset(GEN10_NOA_WRITE_HIGH) || - (addr >= i915_mmio_reg_offset(OA_PERFCNT3_LO) && - addr <= i915_mmio_reg_offset(OA_PERFCNT4_HI)); + return gen7_is_valid_mux_addr(perf, addr) || + ADDR_IN_RANGE(addr, 0x182300, 0x1823A4); } -static bool hsw_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) +static bool gen12_is_valid_b_counter_addr(struct i915_perf *perf, u32 addr) { - return gen7_is_valid_mux_addr(dev_priv, addr) || - (addr >= 0x25100 && addr <= 0x2FF90) || - (addr >= i915_mmio_reg_offset(HSW_MBVID2_NOA0) && - addr <= i915_mmio_reg_offset(HSW_MBVID2_NOA9)) || - addr == i915_mmio_reg_offset(HSW_MBVID2_MISR0); + return REG_IN_RANGE(addr, GEN12_OAG_OASTARTTRIG1, GEN12_OAG_OASTARTTRIG8) || + REG_IN_RANGE(addr, GEN12_OAG_OAREPORTTRIG1, GEN12_OAG_OAREPORTTRIG8) || + REG_IN_RANGE(addr, GEN12_OAG_CEC0_0, GEN12_OAG_CEC7_1) || + REG_IN_RANGE(addr, GEN12_OAG_SCEC0_0, GEN12_OAG_SCEC7_1) || + REG_EQUAL(addr, GEN12_OAA_DBG_REG) || + REG_EQUAL(addr, GEN12_OAG_OA_PESS) || + REG_EQUAL(addr, GEN12_OAG_SPCTR_CNF); } -static bool chv_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) +static bool gen12_is_valid_mux_addr(struct i915_perf *perf, u32 addr) { - return gen7_is_valid_mux_addr(dev_priv, addr) || - (addr >= 0x182300 && addr <= 0x1823A4); + return REG_EQUAL(addr, NOA_WRITE) || + REG_EQUAL(addr, GEN10_NOA_WRITE_HIGH) || + REG_EQUAL(addr, GDT_CHICKEN_BITS) || + REG_EQUAL(addr, WAIT_FOR_RC6_EXIT) || + REG_EQUAL(addr, RPM_CONFIG0) || + REG_EQUAL(addr, RPM_CONFIG1) || + REG_IN_RANGE(addr, NOA_CONFIG(0), NOA_CONFIG(8)); } static u32 mask_reg_value(u32 reg, u32 val) @@ -3201,21 +3840,21 @@ static u32 mask_reg_value(u32 reg, u32 val) * WaDisableSTUnitPowerOptimization workaround. Make sure the value * programmed by userspace doesn't change this. */ - if (i915_mmio_reg_offset(HALF_SLICE_CHICKEN2) == reg) + if (REG_EQUAL(reg, HALF_SLICE_CHICKEN2)) val = val & ~_MASKED_BIT_ENABLE(GEN8_ST_PO_DISABLE); /* WAIT_FOR_RC6_EXIT has only one bit fullfilling the function * indicated by its name and a bunch of selection fields used by OA * configs. */ - if (i915_mmio_reg_offset(WAIT_FOR_RC6_EXIT) == reg) + if (REG_EQUAL(reg, WAIT_FOR_RC6_EXIT)) val = val & ~_MASKED_BIT_ENABLE(HSW_WAIT_FOR_RC6_EXIT_ENABLE); return val; } -static struct i915_oa_reg *alloc_oa_regs(struct drm_i915_private *dev_priv, - bool (*is_valid)(struct drm_i915_private *dev_priv, u32 addr), +static struct i915_oa_reg *alloc_oa_regs(struct i915_perf *perf, + bool (*is_valid)(struct i915_perf *perf, u32 addr), u32 __user *regs, u32 n_regs) { @@ -3245,7 +3884,7 @@ static struct i915_oa_reg *alloc_oa_regs(struct drm_i915_private *dev_priv, if (err) goto addr_err; - if (!is_valid(dev_priv, addr)) { + if (!is_valid(perf, addr)) { DRM_DEBUG("Invalid oa_reg address: %X\n", addr); err = -EINVAL; goto addr_err; @@ -3278,7 +3917,7 @@ static ssize_t show_dynamic_id(struct device *dev, return sprintf(buf, "%d\n", oa_config->id); } -static int create_dynamic_oa_sysfs_entry(struct drm_i915_private *dev_priv, +static int create_dynamic_oa_sysfs_entry(struct i915_perf *perf, struct i915_oa_config *oa_config) { sysfs_attr_init(&oa_config->sysfs_metric_id.attr); @@ -3293,7 +3932,7 @@ static int create_dynamic_oa_sysfs_entry(struct drm_i915_private *dev_priv, oa_config->sysfs_metric.name = oa_config->uuid; oa_config->sysfs_metric.attrs = oa_config->attrs; - return sysfs_create_group(dev_priv->perf.metrics_kobj, + return sysfs_create_group(perf->metrics_kobj, &oa_config->sysfs_metric); } @@ -3313,17 +3952,18 @@ static int create_dynamic_oa_sysfs_entry(struct drm_i915_private *dev_priv, int i915_perf_add_config_ioctl(struct drm_device *dev, void *data, struct drm_file *file) { - struct drm_i915_private *dev_priv = dev->dev_private; + struct i915_perf *perf = &to_i915(dev)->perf; struct drm_i915_perf_oa_config *args = data; struct i915_oa_config *oa_config, *tmp; + static struct i915_oa_reg *regs; int err, id; - if (!dev_priv->perf.initialized) { + if (!perf->i915) { DRM_DEBUG("i915 perf interface not available for this system\n"); return -ENOTSUPP; } - if (!dev_priv->perf.metrics_kobj) { + if (!perf->metrics_kobj) { DRM_DEBUG("OA metrics weren't advertised via sysfs\n"); return -EINVAL; } @@ -3346,7 +3986,8 @@ int i915_perf_add_config_ioctl(struct drm_device *dev, void *data, return -ENOMEM; } - atomic_set(&oa_config->ref_count, 1); + oa_config->perf = perf; + kref_init(&oa_config->ref); if (!uuid_is_valid(args->uuid)) { DRM_DEBUG("Invalid uuid format for OA config\n"); @@ -3360,59 +4001,59 @@ int i915_perf_add_config_ioctl(struct drm_device *dev, void *data, memcpy(oa_config->uuid, args->uuid, sizeof(args->uuid)); oa_config->mux_regs_len = args->n_mux_regs; - oa_config->mux_regs = - alloc_oa_regs(dev_priv, - dev_priv->perf.ops.is_valid_mux_reg, - u64_to_user_ptr(args->mux_regs_ptr), - args->n_mux_regs); + regs = alloc_oa_regs(perf, + perf->ops.is_valid_mux_reg, + u64_to_user_ptr(args->mux_regs_ptr), + args->n_mux_regs); - if (IS_ERR(oa_config->mux_regs)) { + if (IS_ERR(regs)) { DRM_DEBUG("Failed to create OA config for mux_regs\n"); - err = PTR_ERR(oa_config->mux_regs); + err = PTR_ERR(regs); goto reg_err; } + oa_config->mux_regs = regs; oa_config->b_counter_regs_len = args->n_boolean_regs; - oa_config->b_counter_regs = - alloc_oa_regs(dev_priv, - dev_priv->perf.ops.is_valid_b_counter_reg, - u64_to_user_ptr(args->boolean_regs_ptr), - args->n_boolean_regs); + regs = alloc_oa_regs(perf, + perf->ops.is_valid_b_counter_reg, + u64_to_user_ptr(args->boolean_regs_ptr), + args->n_boolean_regs); - if (IS_ERR(oa_config->b_counter_regs)) { + if (IS_ERR(regs)) { DRM_DEBUG("Failed to create OA config for b_counter_regs\n"); - err = PTR_ERR(oa_config->b_counter_regs); + err = PTR_ERR(regs); goto reg_err; } + oa_config->b_counter_regs = regs; - if (INTEL_GEN(dev_priv) < 8) { + if (INTEL_GEN(perf->i915) < 8) { if (args->n_flex_regs != 0) { err = -EINVAL; goto reg_err; } } else { oa_config->flex_regs_len = args->n_flex_regs; - oa_config->flex_regs = - alloc_oa_regs(dev_priv, - dev_priv->perf.ops.is_valid_flex_reg, - u64_to_user_ptr(args->flex_regs_ptr), - args->n_flex_regs); + regs = alloc_oa_regs(perf, + perf->ops.is_valid_flex_reg, + u64_to_user_ptr(args->flex_regs_ptr), + args->n_flex_regs); - if (IS_ERR(oa_config->flex_regs)) { + if (IS_ERR(regs)) { DRM_DEBUG("Failed to create OA config for flex_regs\n"); - err = PTR_ERR(oa_config->flex_regs); + err = PTR_ERR(regs); goto reg_err; } + oa_config->flex_regs = regs; } - err = mutex_lock_interruptible(&dev_priv->perf.metrics_lock); + err = mutex_lock_interruptible(&perf->metrics_lock); if (err) goto reg_err; /* We shouldn't have too many configs, so this iteration shouldn't be * too costly. */ - idr_for_each_entry(&dev_priv->perf.metrics_idr, tmp, id) { + idr_for_each_entry(&perf->metrics_idr, tmp, id) { if (!strcmp(tmp->uuid, oa_config->uuid)) { DRM_DEBUG("OA config already exists with this uuid\n"); err = -EADDRINUSE; @@ -3420,14 +4061,14 @@ int i915_perf_add_config_ioctl(struct drm_device *dev, void *data, } } - err = create_dynamic_oa_sysfs_entry(dev_priv, oa_config); + err = create_dynamic_oa_sysfs_entry(perf, oa_config); if (err) { DRM_DEBUG("Failed to create sysfs entry for OA config\n"); goto sysfs_err; } /* Config id 0 is invalid, id 1 for kernel stored test config. */ - oa_config->id = idr_alloc(&dev_priv->perf.metrics_idr, + oa_config->id = idr_alloc(&perf->metrics_idr, oa_config, 2, 0, GFP_KERNEL); if (oa_config->id < 0) { @@ -3436,16 +4077,16 @@ int i915_perf_add_config_ioctl(struct drm_device *dev, void *data, goto sysfs_err; } - mutex_unlock(&dev_priv->perf.metrics_lock); + mutex_unlock(&perf->metrics_lock); DRM_DEBUG("Added config %s id=%i\n", oa_config->uuid, oa_config->id); return oa_config->id; sysfs_err: - mutex_unlock(&dev_priv->perf.metrics_lock); + mutex_unlock(&perf->metrics_lock); reg_err: - put_oa_config(dev_priv, oa_config); + i915_oa_config_put(oa_config); DRM_DEBUG("Failed to add new OA config\n"); return err; } @@ -3464,12 +4105,12 @@ reg_err: int i915_perf_remove_config_ioctl(struct drm_device *dev, void *data, struct drm_file *file) { - struct drm_i915_private *dev_priv = dev->dev_private; + struct i915_perf *perf = &to_i915(dev)->perf; u64 *arg = data; struct i915_oa_config *oa_config; int ret; - if (!dev_priv->perf.initialized) { + if (!perf->i915) { DRM_DEBUG("i915 perf interface not available for this system\n"); return -ENOTSUPP; } @@ -3479,31 +4120,33 @@ int i915_perf_remove_config_ioctl(struct drm_device *dev, void *data, return -EACCES; } - ret = mutex_lock_interruptible(&dev_priv->perf.metrics_lock); + ret = mutex_lock_interruptible(&perf->metrics_lock); if (ret) - goto lock_err; + return ret; - oa_config = idr_find(&dev_priv->perf.metrics_idr, *arg); + oa_config = idr_find(&perf->metrics_idr, *arg); if (!oa_config) { DRM_DEBUG("Failed to remove unknown OA config\n"); ret = -ENOENT; - goto config_err; + goto err_unlock; } GEM_BUG_ON(*arg != oa_config->id); - sysfs_remove_group(dev_priv->perf.metrics_kobj, - &oa_config->sysfs_metric); + sysfs_remove_group(perf->metrics_kobj, &oa_config->sysfs_metric); + + idr_remove(&perf->metrics_idr, *arg); - idr_remove(&dev_priv->perf.metrics_idr, *arg); + mutex_unlock(&perf->metrics_lock); DRM_DEBUG("Removed config %s id=%i\n", oa_config->uuid, oa_config->id); - put_oa_config(dev_priv, oa_config); + i915_oa_config_put(oa_config); -config_err: - mutex_unlock(&dev_priv->perf.metrics_lock); -lock_err: + return 0; + +err_unlock: + mutex_unlock(&perf->metrics_lock); return ret; } @@ -3551,103 +4194,126 @@ static struct ctl_table dev_root[] = { /** * i915_perf_init - initialize i915-perf state on module load - * @dev_priv: i915 device instance + * @i915: i915 device instance * * Initializes i915-perf state without exposing anything to userspace. * * Note: i915-perf initialization is split into an 'init' and 'register' * phase with the i915_perf_register() exposing state to userspace. */ -void i915_perf_init(struct drm_i915_private *dev_priv) -{ - if (IS_HASWELL(dev_priv)) { - dev_priv->perf.ops.is_valid_b_counter_reg = - gen7_is_valid_b_counter_addr; - dev_priv->perf.ops.is_valid_mux_reg = - hsw_is_valid_mux_addr; - dev_priv->perf.ops.is_valid_flex_reg = NULL; - dev_priv->perf.ops.enable_metric_set = hsw_enable_metric_set; - dev_priv->perf.ops.disable_metric_set = hsw_disable_metric_set; - dev_priv->perf.ops.oa_enable = gen7_oa_enable; - dev_priv->perf.ops.oa_disable = gen7_oa_disable; - dev_priv->perf.ops.read = gen7_oa_read; - dev_priv->perf.ops.oa_hw_tail_read = - gen7_oa_hw_tail_read; - - dev_priv->perf.oa_formats = hsw_oa_formats; - } else if (HAS_LOGICAL_RING_CONTEXTS(dev_priv)) { +void i915_perf_init(struct drm_i915_private *i915) +{ + struct i915_perf *perf = &i915->perf; + + /* XXX const struct i915_perf_ops! */ + + if (IS_HASWELL(i915)) { + perf->ops.is_valid_b_counter_reg = gen7_is_valid_b_counter_addr; + perf->ops.is_valid_mux_reg = hsw_is_valid_mux_addr; + perf->ops.is_valid_flex_reg = NULL; + perf->ops.enable_metric_set = hsw_enable_metric_set; + perf->ops.disable_metric_set = hsw_disable_metric_set; + perf->ops.oa_enable = gen7_oa_enable; + perf->ops.oa_disable = gen7_oa_disable; + perf->ops.read = gen7_oa_read; + perf->ops.oa_hw_tail_read = gen7_oa_hw_tail_read; + + perf->oa_formats = hsw_oa_formats; + } else if (HAS_LOGICAL_RING_CONTEXTS(i915)) { /* Note: that although we could theoretically also support the * legacy ringbuffer mode on BDW (and earlier iterations of * this driver, before upstreaming did this) it didn't seem * worth the complexity to maintain now that BDW+ enable * execlist mode by default. */ - dev_priv->perf.oa_formats = gen8_plus_oa_formats; + perf->ops.read = gen8_oa_read; - dev_priv->perf.ops.oa_enable = gen8_oa_enable; - dev_priv->perf.ops.oa_disable = gen8_oa_disable; - dev_priv->perf.ops.read = gen8_oa_read; - dev_priv->perf.ops.oa_hw_tail_read = gen8_oa_hw_tail_read; + if (IS_GEN_RANGE(i915, 8, 9)) { + perf->oa_formats = gen8_plus_oa_formats; - if (IS_GEN_RANGE(dev_priv, 8, 9)) { - dev_priv->perf.ops.is_valid_b_counter_reg = + perf->ops.is_valid_b_counter_reg = gen7_is_valid_b_counter_addr; - dev_priv->perf.ops.is_valid_mux_reg = + perf->ops.is_valid_mux_reg = gen8_is_valid_mux_addr; - dev_priv->perf.ops.is_valid_flex_reg = + perf->ops.is_valid_flex_reg = gen8_is_valid_flex_addr; - if (IS_CHERRYVIEW(dev_priv)) { - dev_priv->perf.ops.is_valid_mux_reg = + if (IS_CHERRYVIEW(i915)) { + perf->ops.is_valid_mux_reg = chv_is_valid_mux_addr; } - dev_priv->perf.ops.enable_metric_set = gen8_enable_metric_set; - dev_priv->perf.ops.disable_metric_set = gen8_disable_metric_set; + perf->ops.oa_enable = gen8_oa_enable; + perf->ops.oa_disable = gen8_oa_disable; + perf->ops.enable_metric_set = gen8_enable_metric_set; + perf->ops.disable_metric_set = gen8_disable_metric_set; + perf->ops.oa_hw_tail_read = gen8_oa_hw_tail_read; - if (IS_GEN(dev_priv, 8)) { - dev_priv->perf.ctx_oactxctrl_offset = 0x120; - dev_priv->perf.ctx_flexeu0_offset = 0x2ce; + if (IS_GEN(i915, 8)) { + perf->ctx_oactxctrl_offset = 0x120; + perf->ctx_flexeu0_offset = 0x2ce; - dev_priv->perf.gen8_valid_ctx_bit = BIT(25); + perf->gen8_valid_ctx_bit = BIT(25); } else { - dev_priv->perf.ctx_oactxctrl_offset = 0x128; - dev_priv->perf.ctx_flexeu0_offset = 0x3de; + perf->ctx_oactxctrl_offset = 0x128; + perf->ctx_flexeu0_offset = 0x3de; - dev_priv->perf.gen8_valid_ctx_bit = BIT(16); + perf->gen8_valid_ctx_bit = BIT(16); } - } else if (IS_GEN_RANGE(dev_priv, 10, 11)) { - dev_priv->perf.ops.is_valid_b_counter_reg = + } else if (IS_GEN_RANGE(i915, 10, 11)) { + perf->oa_formats = gen8_plus_oa_formats; + + perf->ops.is_valid_b_counter_reg = gen7_is_valid_b_counter_addr; - dev_priv->perf.ops.is_valid_mux_reg = + perf->ops.is_valid_mux_reg = gen10_is_valid_mux_addr; - dev_priv->perf.ops.is_valid_flex_reg = + perf->ops.is_valid_flex_reg = gen8_is_valid_flex_addr; - dev_priv->perf.ops.enable_metric_set = gen8_enable_metric_set; - dev_priv->perf.ops.disable_metric_set = gen10_disable_metric_set; + perf->ops.oa_enable = gen8_oa_enable; + perf->ops.oa_disable = gen8_oa_disable; + perf->ops.enable_metric_set = gen8_enable_metric_set; + perf->ops.disable_metric_set = gen10_disable_metric_set; + perf->ops.oa_hw_tail_read = gen8_oa_hw_tail_read; - if (IS_GEN(dev_priv, 10)) { - dev_priv->perf.ctx_oactxctrl_offset = 0x128; - dev_priv->perf.ctx_flexeu0_offset = 0x3de; + if (IS_GEN(i915, 10)) { + perf->ctx_oactxctrl_offset = 0x128; + perf->ctx_flexeu0_offset = 0x3de; } else { - dev_priv->perf.ctx_oactxctrl_offset = 0x124; - dev_priv->perf.ctx_flexeu0_offset = 0x78e; + perf->ctx_oactxctrl_offset = 0x124; + perf->ctx_flexeu0_offset = 0x78e; } - dev_priv->perf.gen8_valid_ctx_bit = BIT(16); + perf->gen8_valid_ctx_bit = BIT(16); + } else if (IS_GEN(i915, 12)) { + perf->oa_formats = gen12_oa_formats; + + perf->ops.is_valid_b_counter_reg = + gen12_is_valid_b_counter_addr; + perf->ops.is_valid_mux_reg = + gen12_is_valid_mux_addr; + perf->ops.is_valid_flex_reg = + gen8_is_valid_flex_addr; + + perf->ops.oa_enable = gen12_oa_enable; + perf->ops.oa_disable = gen12_oa_disable; + perf->ops.enable_metric_set = gen12_enable_metric_set; + perf->ops.disable_metric_set = gen12_disable_metric_set; + perf->ops.oa_hw_tail_read = gen12_oa_hw_tail_read; + + perf->ctx_flexeu0_offset = 0; + perf->ctx_oactxctrl_offset = 0x144; } } - if (dev_priv->perf.ops.enable_metric_set) { - INIT_LIST_HEAD(&dev_priv->perf.streams); - mutex_init(&dev_priv->perf.lock); + if (perf->ops.enable_metric_set) { + mutex_init(&perf->lock); oa_sample_rate_hard_limit = 1000 * - (RUNTIME_INFO(dev_priv)->cs_timestamp_frequency_khz / 2); - dev_priv->perf.sysctl_header = register_sysctl_table(dev_root); + (RUNTIME_INFO(i915)->cs_timestamp_frequency_khz / 2); + perf->sysctl_header = register_sysctl_table(dev_root); - mutex_init(&dev_priv->perf.metrics_lock); - idr_init(&dev_priv->perf.metrics_idr); + mutex_init(&perf->metrics_lock); + idr_init(&perf->metrics_idr); /* We set up some ratelimit state to potentially throttle any * _NOTES about spurious, invalid OA reports which we don't @@ -3659,44 +4325,70 @@ void i915_perf_init(struct drm_i915_private *dev_priv) * * Using the same limiting factors as printk_ratelimit() */ - ratelimit_state_init(&dev_priv->perf.spurious_report_rs, - 5 * HZ, 10); + ratelimit_state_init(&perf->spurious_report_rs, 5 * HZ, 10); /* Since we use a DRM_NOTE for spurious reports it would be * inconsistent to let __ratelimit() automatically print a * warning for throttling. */ - ratelimit_set_flags(&dev_priv->perf.spurious_report_rs, + ratelimit_set_flags(&perf->spurious_report_rs, RATELIMIT_MSG_ON_RELEASE); - dev_priv->perf.initialized = true; + atomic64_set(&perf->noa_programming_delay, + 500 * 1000 /* 500us */); + + perf->i915 = i915; } } static int destroy_config(int id, void *p, void *data) { - struct drm_i915_private *dev_priv = data; - struct i915_oa_config *oa_config = p; - - put_oa_config(dev_priv, oa_config); - + i915_oa_config_put(p); return 0; } /** * i915_perf_fini - Counter part to i915_perf_init() - * @dev_priv: i915 device instance + * @i915: i915 device instance */ -void i915_perf_fini(struct drm_i915_private *dev_priv) +void i915_perf_fini(struct drm_i915_private *i915) { - if (!dev_priv->perf.initialized) + struct i915_perf *perf = &i915->perf; + + if (!perf->i915) return; - idr_for_each(&dev_priv->perf.metrics_idr, destroy_config, dev_priv); - idr_destroy(&dev_priv->perf.metrics_idr); + idr_for_each(&perf->metrics_idr, destroy_config, perf); + idr_destroy(&perf->metrics_idr); - unregister_sysctl_table(dev_priv->perf.sysctl_header); + unregister_sysctl_table(perf->sysctl_header); - memset(&dev_priv->perf.ops, 0, sizeof(dev_priv->perf.ops)); + memset(&perf->ops, 0, sizeof(perf->ops)); + perf->i915 = NULL; +} - dev_priv->perf.initialized = false; +/** + * i915_perf_ioctl_version - Version of the i915-perf subsystem + * + * This version number is used by userspace to detect available features. + */ +int i915_perf_ioctl_version(void) +{ + /* + * 1: Initial version + * I915_PERF_IOCTL_ENABLE + * I915_PERF_IOCTL_DISABLE + * + * 2: Added runtime modification of OA config. + * I915_PERF_IOCTL_CONFIG + * + * 3: Add DRM_I915_PERF_PROP_HOLD_PREEMPTION parameter to hold + * preemption on a particular context so that performance data is + * accessible from a delta of MI_RPC reports without looking at the + * OA buffer. + */ + return 3; } + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "selftests/i915_perf.c" +#endif diff --git a/drivers/gpu/drm/i915/i915_perf.h b/drivers/gpu/drm/i915/i915_perf.h index a412b16d9ffc..4ceebce72060 100644 --- a/drivers/gpu/drm/i915/i915_perf.h +++ b/drivers/gpu/drm/i915/i915_perf.h @@ -6,11 +6,15 @@ #ifndef __I915_PERF_H__ #define __I915_PERF_H__ +#include <linux/kref.h> #include <linux/types.h> +#include "i915_perf_types.h" + struct drm_device; struct drm_file; struct drm_i915_private; +struct i915_oa_config; struct intel_context; struct intel_engine_cs; @@ -18,6 +22,7 @@ void i915_perf_init(struct drm_i915_private *i915); void i915_perf_fini(struct drm_i915_private *i915); void i915_perf_register(struct drm_i915_private *i915); void i915_perf_unregister(struct drm_i915_private *i915); +int i915_perf_ioctl_version(void); int i915_perf_open_ioctl(struct drm_device *dev, void *data, struct drm_file *file); @@ -25,8 +30,29 @@ int i915_perf_add_config_ioctl(struct drm_device *dev, void *data, struct drm_file *file); int i915_perf_remove_config_ioctl(struct drm_device *dev, void *data, struct drm_file *file); -void i915_oa_init_reg_state(struct intel_engine_cs *engine, - struct intel_context *ce, - u32 *reg_state); + +void i915_oa_init_reg_state(const struct intel_context *ce, + const struct intel_engine_cs *engine); + +struct i915_oa_config * +i915_perf_get_oa_config(struct i915_perf *perf, int metrics_set); + +static inline struct i915_oa_config * +i915_oa_config_get(struct i915_oa_config *oa_config) +{ + if (kref_get_unless_zero(&oa_config->ref)) + return oa_config; + else + return NULL; +} + +void i915_oa_config_release(struct kref *ref); +static inline void i915_oa_config_put(struct i915_oa_config *oa_config) +{ + if (!oa_config) + return; + + kref_put(&oa_config->ref, i915_oa_config_release); +} #endif /* __I915_PERF_H__ */ diff --git a/drivers/gpu/drm/i915/i915_perf_types.h b/drivers/gpu/drm/i915/i915_perf_types.h new file mode 100644 index 000000000000..74ddc20a0d37 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_perf_types.h @@ -0,0 +1,435 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef _I915_PERF_TYPES_H_ +#define _I915_PERF_TYPES_H_ + +#include <linux/atomic.h> +#include <linux/device.h> +#include <linux/hrtimer.h> +#include <linux/llist.h> +#include <linux/poll.h> +#include <linux/sysfs.h> +#include <linux/types.h> +#include <linux/uuid.h> +#include <linux/wait.h> + +#include "i915_reg.h" +#include "intel_wakeref.h" + +struct drm_i915_private; +struct file; +struct i915_gem_context; +struct i915_perf; +struct i915_vma; +struct intel_context; +struct intel_engine_cs; + +struct i915_oa_format { + u32 format; + int size; +}; + +struct i915_oa_reg { + i915_reg_t addr; + u32 value; +}; + +struct i915_oa_config { + struct i915_perf *perf; + + char uuid[UUID_STRING_LEN + 1]; + int id; + + const struct i915_oa_reg *mux_regs; + u32 mux_regs_len; + const struct i915_oa_reg *b_counter_regs; + u32 b_counter_regs_len; + const struct i915_oa_reg *flex_regs; + u32 flex_regs_len; + + struct attribute_group sysfs_metric; + struct attribute *attrs[2]; + struct device_attribute sysfs_metric_id; + + struct kref ref; + struct rcu_head rcu; +}; + +struct i915_perf_stream; + +/** + * struct i915_perf_stream_ops - the OPs to support a specific stream type + */ +struct i915_perf_stream_ops { + /** + * @enable: Enables the collection of HW samples, either in response to + * `I915_PERF_IOCTL_ENABLE` or implicitly called when stream is opened + * without `I915_PERF_FLAG_DISABLED`. + */ + void (*enable)(struct i915_perf_stream *stream); + + /** + * @disable: Disables the collection of HW samples, either in response + * to `I915_PERF_IOCTL_DISABLE` or implicitly called before destroying + * the stream. + */ + void (*disable)(struct i915_perf_stream *stream); + + /** + * @poll_wait: Call poll_wait, passing a wait queue that will be woken + * once there is something ready to read() for the stream + */ + void (*poll_wait)(struct i915_perf_stream *stream, + struct file *file, + poll_table *wait); + + /** + * @wait_unlocked: For handling a blocking read, wait until there is + * something to ready to read() for the stream. E.g. wait on the same + * wait queue that would be passed to poll_wait(). + */ + int (*wait_unlocked)(struct i915_perf_stream *stream); + + /** + * @read: Copy buffered metrics as records to userspace + * **buf**: the userspace, destination buffer + * **count**: the number of bytes to copy, requested by userspace + * **offset**: zero at the start of the read, updated as the read + * proceeds, it represents how many bytes have been copied so far and + * the buffer offset for copying the next record. + * + * Copy as many buffered i915 perf samples and records for this stream + * to userspace as will fit in the given buffer. + * + * Only write complete records; returning -%ENOSPC if there isn't room + * for a complete record. + * + * Return any error condition that results in a short read such as + * -%ENOSPC or -%EFAULT, even though these may be squashed before + * returning to userspace. + */ + int (*read)(struct i915_perf_stream *stream, + char __user *buf, + size_t count, + size_t *offset); + + /** + * @destroy: Cleanup any stream specific resources. + * + * The stream will always be disabled before this is called. + */ + void (*destroy)(struct i915_perf_stream *stream); +}; + +/** + * struct i915_perf_stream - state for a single open stream FD + */ +struct i915_perf_stream { + /** + * @perf: i915_perf backpointer + */ + struct i915_perf *perf; + + /** + * @uncore: mmio access path + */ + struct intel_uncore *uncore; + + /** + * @engine: Engine associated with this performance stream. + */ + struct intel_engine_cs *engine; + + /** + * @sample_flags: Flags representing the `DRM_I915_PERF_PROP_SAMPLE_*` + * properties given when opening a stream, representing the contents + * of a single sample as read() by userspace. + */ + u32 sample_flags; + + /** + * @sample_size: Considering the configured contents of a sample + * combined with the required header size, this is the total size + * of a single sample record. + */ + int sample_size; + + /** + * @ctx: %NULL if measuring system-wide across all contexts or a + * specific context that is being monitored. + */ + struct i915_gem_context *ctx; + + /** + * @enabled: Whether the stream is currently enabled, considering + * whether the stream was opened in a disabled state and based + * on `I915_PERF_IOCTL_ENABLE` and `I915_PERF_IOCTL_DISABLE` calls. + */ + bool enabled; + + /** + * @hold_preemption: Whether preemption is put on hold for command + * submissions done on the @ctx. This is useful for some drivers that + * cannot easily post process the OA buffer context to subtract delta + * of performance counters not associated with @ctx. + */ + bool hold_preemption; + + /** + * @ops: The callbacks providing the implementation of this specific + * type of configured stream. + */ + const struct i915_perf_stream_ops *ops; + + /** + * @oa_config: The OA configuration used by the stream. + */ + struct i915_oa_config *oa_config; + + /** + * @oa_config_bos: A list of struct i915_oa_config_bo allocated lazily + * each time @oa_config changes. + */ + struct llist_head oa_config_bos; + + /** + * @pinned_ctx: The OA context specific information. + */ + struct intel_context *pinned_ctx; + + /** + * @specific_ctx_id: The id of the specific context. + */ + u32 specific_ctx_id; + + /** + * @specific_ctx_id_mask: The mask used to masking specific_ctx_id bits. + */ + u32 specific_ctx_id_mask; + + /** + * @poll_check_timer: High resolution timer that will periodically + * check for data in the circular OA buffer for notifying userspace + * (e.g. during a read() or poll()). + */ + struct hrtimer poll_check_timer; + + /** + * @poll_wq: The wait queue that hrtimer callback wakes when it + * sees data ready to read in the circular OA buffer. + */ + wait_queue_head_t poll_wq; + + /** + * @pollin: Whether there is data available to read. + */ + bool pollin; + + /** + * @periodic: Whether periodic sampling is currently enabled. + */ + bool periodic; + + /** + * @period_exponent: The OA unit sampling frequency is derived from this. + */ + int period_exponent; + + /** + * @oa_buffer: State of the OA buffer. + */ + struct { + struct i915_vma *vma; + u8 *vaddr; + u32 last_ctx_id; + int format; + int format_size; + int size_exponent; + + /** + * @ptr_lock: Locks reads and writes to all head/tail state + * + * Consider: the head and tail pointer state needs to be read + * consistently from a hrtimer callback (atomic context) and + * read() fop (user context) with tail pointer updates happening + * in atomic context and head updates in user context and the + * (unlikely) possibility of read() errors needing to reset all + * head/tail state. + * + * Note: Contention/performance aren't currently a significant + * concern here considering the relatively low frequency of + * hrtimer callbacks (5ms period) and that reads typically only + * happen in response to a hrtimer event and likely complete + * before the next callback. + * + * Note: This lock is not held *while* reading and copying data + * to userspace so the value of head observed in htrimer + * callbacks won't represent any partial consumption of data. + */ + spinlock_t ptr_lock; + + /** + * @tails: One 'aging' tail pointer and one 'aged' tail pointer ready to + * used for reading. + * + * Initial values of 0xffffffff are invalid and imply that an + * update is required (and should be ignored by an attempted + * read) + */ + struct { + u32 offset; + } tails[2]; + + /** + * @aged_tail_idx: Index for the aged tail ready to read() data up to. + */ + unsigned int aged_tail_idx; + + /** + * @aging_timestamp: A monotonic timestamp for when the current aging tail pointer + * was read; used to determine when it is old enough to trust. + */ + u64 aging_timestamp; + + /** + * @head: Although we can always read back the head pointer register, + * we prefer to avoid trusting the HW state, just to avoid any + * risk that some hardware condition could * somehow bump the + * head pointer unpredictably and cause us to forward the wrong + * OA buffer data to userspace. + */ + u32 head; + } oa_buffer; + + /** + * @noa_wait: A batch buffer doing a wait on the GPU for the NOA logic to be + * reprogrammed. + */ + struct i915_vma *noa_wait; +}; + +/** + * struct i915_oa_ops - Gen specific implementation of an OA unit stream + */ +struct i915_oa_ops { + /** + * @is_valid_b_counter_reg: Validates register's address for + * programming boolean counters for a particular platform. + */ + bool (*is_valid_b_counter_reg)(struct i915_perf *perf, u32 addr); + + /** + * @is_valid_mux_reg: Validates register's address for programming mux + * for a particular platform. + */ + bool (*is_valid_mux_reg)(struct i915_perf *perf, u32 addr); + + /** + * @is_valid_flex_reg: Validates register's address for programming + * flex EU filtering for a particular platform. + */ + bool (*is_valid_flex_reg)(struct i915_perf *perf, u32 addr); + + /** + * @enable_metric_set: Selects and applies any MUX configuration to set + * up the Boolean and Custom (B/C) counters that are part of the + * counter reports being sampled. May apply system constraints such as + * disabling EU clock gating as required. + */ + int (*enable_metric_set)(struct i915_perf_stream *stream); + + /** + * @disable_metric_set: Remove system constraints associated with using + * the OA unit. + */ + void (*disable_metric_set)(struct i915_perf_stream *stream); + + /** + * @oa_enable: Enable periodic sampling + */ + void (*oa_enable)(struct i915_perf_stream *stream); + + /** + * @oa_disable: Disable periodic sampling + */ + void (*oa_disable)(struct i915_perf_stream *stream); + + /** + * @read: Copy data from the circular OA buffer into a given userspace + * buffer. + */ + int (*read)(struct i915_perf_stream *stream, + char __user *buf, + size_t count, + size_t *offset); + + /** + * @oa_hw_tail_read: read the OA tail pointer register + * + * In particular this enables us to share all the fiddly code for + * handling the OA unit tail pointer race that affects multiple + * generations. + */ + u32 (*oa_hw_tail_read)(struct i915_perf_stream *stream); +}; + +struct i915_perf { + struct drm_i915_private *i915; + + struct kobject *metrics_kobj; + struct ctl_table_header *sysctl_header; + + /* + * Lock associated with adding/modifying/removing OA configs + * in perf->metrics_idr. + */ + struct mutex metrics_lock; + + /* + * List of dynamic configurations (struct i915_oa_config), you + * need to hold perf->metrics_lock to access it. + */ + struct idr metrics_idr; + + /* + * Lock associated with anything below within this structure + * except exclusive_stream. + */ + struct mutex lock; + + /* + * The stream currently using the OA unit. If accessed + * outside a syscall associated to its file + * descriptor. + */ + struct i915_perf_stream *exclusive_stream; + + /** + * For rate limiting any notifications of spurious + * invalid OA reports + */ + struct ratelimit_state spurious_report_rs; + + struct i915_oa_config test_config; + + u32 gen7_latched_oastatus1; + u32 ctx_oactxctrl_offset; + u32 ctx_flexeu0_offset; + + /** + * The RPT_ID/reason field for Gen8+ includes a bit + * to determine if the CTX ID in the report is valid + * but the specific bit differs between Gen 8 and 9 + */ + u32 gen8_valid_ctx_bit; + + struct i915_oa_ops ops; + const struct i915_oa_format *oa_formats; + + atomic64_t noa_programming_delay; +}; + +#endif /* _I915_PERF_TYPES_H_ */ diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 8e251e719390..2814218c5ba1 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -11,6 +11,8 @@ #include "gt/intel_engine_pm.h" #include "gt/intel_engine_user.h" #include "gt/intel_gt_pm.h" +#include "gt/intel_rc6.h" +#include "gt/intel_rps.h" #include "i915_drv.h" #include "i915_pmu.h" @@ -116,22 +118,124 @@ static bool pmu_needs_timer(struct i915_pmu *pmu, bool gpu_active) return enable; } -void i915_pmu_gt_parked(struct drm_i915_private *i915) +static u64 __get_rc6(struct intel_gt *gt) { - struct i915_pmu *pmu = &i915->pmu; + struct drm_i915_private *i915 = gt->i915; + u64 val; - if (!pmu->base.event_init) - return; + val = intel_rc6_residency_ns(>->rc6, + IS_VALLEYVIEW(i915) ? + VLV_GT_RENDER_RC6 : + GEN6_GT_GFX_RC6); + + if (HAS_RC6p(i915)) + val += intel_rc6_residency_ns(>->rc6, GEN6_GT_GFX_RC6p); + + if (HAS_RC6pp(i915)) + val += intel_rc6_residency_ns(>->rc6, GEN6_GT_GFX_RC6pp); + + return val; +} + +#if IS_ENABLED(CONFIG_PM) + +static inline s64 ktime_since(const ktime_t kt) +{ + return ktime_to_ns(ktime_sub(ktime_get(), kt)); +} + +static u64 __pmu_estimate_rc6(struct i915_pmu *pmu) +{ + u64 val; - spin_lock_irq(&pmu->lock); /* - * Signal sampling timer to stop if only engine events are enabled and - * GPU went idle. + * We think we are runtime suspended. + * + * Report the delta from when the device was suspended to now, + * on top of the last known real value, as the approximated RC6 + * counter value. */ - pmu->timer_enabled = pmu_needs_timer(pmu, false); - spin_unlock_irq(&pmu->lock); + val = ktime_since(pmu->sleep_last); + val += pmu->sample[__I915_SAMPLE_RC6].cur; + + pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur = val; + + return val; } +static u64 __pmu_update_rc6(struct i915_pmu *pmu, u64 val) +{ + /* + * If we are coming back from being runtime suspended we must + * be careful not to report a larger value than returned + * previously. + */ + if (val >= pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur) { + pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur = 0; + pmu->sample[__I915_SAMPLE_RC6].cur = val; + } else { + val = pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur; + } + + return val; +} + +static u64 get_rc6(struct intel_gt *gt) +{ + struct drm_i915_private *i915 = gt->i915; + struct i915_pmu *pmu = &i915->pmu; + unsigned long flags; + u64 val; + + val = 0; + if (intel_gt_pm_get_if_awake(gt)) { + val = __get_rc6(gt); + intel_gt_pm_put_async(gt); + } + + spin_lock_irqsave(&pmu->lock, flags); + + if (val) + val = __pmu_update_rc6(pmu, val); + else + val = __pmu_estimate_rc6(pmu); + + spin_unlock_irqrestore(&pmu->lock, flags); + + return val; +} + +static void park_rc6(struct drm_i915_private *i915) +{ + struct i915_pmu *pmu = &i915->pmu; + + if (pmu->enable & config_enabled_mask(I915_PMU_RC6_RESIDENCY)) + __pmu_update_rc6(pmu, __get_rc6(&i915->gt)); + + pmu->sleep_last = ktime_get(); +} + +static void unpark_rc6(struct drm_i915_private *i915) +{ + struct i915_pmu *pmu = &i915->pmu; + + /* Estimate how long we slept and accumulate that into rc6 counters */ + if (pmu->enable & config_enabled_mask(I915_PMU_RC6_RESIDENCY)) + __pmu_estimate_rc6(pmu); +} + +#else + +static u64 get_rc6(struct intel_gt *gt) +{ + return __get_rc6(gt); +} + +static void park_rc6(struct drm_i915_private *i915) {} +static void unpark_rc6(struct drm_i915_private *i915) {} + +#endif + static void __i915_pmu_maybe_start_timer(struct i915_pmu *pmu) { if (!pmu->timer_enabled && pmu_needs_timer(pmu, true)) { @@ -143,6 +247,26 @@ static void __i915_pmu_maybe_start_timer(struct i915_pmu *pmu) } } +void i915_pmu_gt_parked(struct drm_i915_private *i915) +{ + struct i915_pmu *pmu = &i915->pmu; + + if (!pmu->base.event_init) + return; + + spin_lock_irq(&pmu->lock); + + park_rc6(i915); + + /* + * Signal sampling timer to stop if only engine events are enabled and + * GPU went idle. + */ + pmu->timer_enabled = pmu_needs_timer(pmu, false); + + spin_unlock_irq(&pmu->lock); +} + void i915_pmu_gt_unparked(struct drm_i915_private *i915) { struct i915_pmu *pmu = &i915->pmu; @@ -151,10 +275,14 @@ void i915_pmu_gt_unparked(struct drm_i915_private *i915) return; spin_lock_irq(&pmu->lock); + /* * Re-enable sampling timer when GPU goes active. */ __i915_pmu_maybe_start_timer(pmu); + + unpark_rc6(i915); + spin_unlock_irq(&pmu->lock); } @@ -174,7 +302,7 @@ engines_sample(struct intel_gt *gt, unsigned int period_ns) if ((i915->pmu.enable & ENGINE_SAMPLE_MASK) == 0) return; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct intel_engine_pmu *pmu = &engine->pmu; unsigned long flags; bool busy; @@ -194,6 +322,10 @@ engines_sample(struct intel_gt *gt, unsigned int period_ns) if (val & RING_WAIT_SEMAPHORE) add_sample(&pmu->sample[I915_SAMPLE_SEMA], period_ns); + /* No need to sample when busy stats are supported. */ + if (intel_engine_supports_stats(engine)) + goto skip; + /* * While waiting on a semaphore or event, MI_MODE reports the * ring as idle. However, previously using the seqno, and with @@ -211,7 +343,7 @@ engines_sample(struct intel_gt *gt, unsigned int period_ns) skip: spin_unlock_irqrestore(&engine->uncore->lock, flags); - intel_engine_pm_put(engine); + intel_engine_pm_put_async(engine); } } @@ -227,25 +359,26 @@ frequency_sample(struct intel_gt *gt, unsigned int period_ns) struct drm_i915_private *i915 = gt->i915; struct intel_uncore *uncore = gt->uncore; struct i915_pmu *pmu = &i915->pmu; + struct intel_rps *rps = >->rps; if (pmu->enable & config_enabled_mask(I915_PMU_ACTUAL_FREQUENCY)) { u32 val; - val = i915->gt_pm.rps.cur_freq; + val = rps->cur_freq; if (intel_gt_pm_get_if_awake(gt)) { val = intel_uncore_read_notrace(uncore, GEN6_RPSTAT1); - val = intel_get_cagf(i915, val); - intel_gt_pm_put(gt); + val = intel_get_cagf(rps, val); + intel_gt_pm_put_async(gt); } add_sample_mult(&pmu->sample[__I915_SAMPLE_FREQ_ACT], - intel_gpu_freq(i915, val), + intel_gpu_freq(rps, val), period_ns / 1000); } if (pmu->enable & config_enabled_mask(I915_PMU_REQUESTED_FREQUENCY)) { add_sample_mult(&pmu->sample[__I915_SAMPLE_FREQ_REQ], - intel_gpu_freq(i915, i915->gt_pm.rps.cur_freq), + intel_gpu_freq(rps, rps->cur_freq), period_ns / 1000); } } @@ -426,104 +559,6 @@ static int i915_pmu_event_init(struct perf_event *event) return 0; } -static u64 __get_rc6(struct intel_gt *gt) -{ - struct drm_i915_private *i915 = gt->i915; - u64 val; - - val = intel_rc6_residency_ns(i915, - IS_VALLEYVIEW(i915) ? - VLV_GT_RENDER_RC6 : - GEN6_GT_GFX_RC6); - - if (HAS_RC6p(i915)) - val += intel_rc6_residency_ns(i915, GEN6_GT_GFX_RC6p); - - if (HAS_RC6pp(i915)) - val += intel_rc6_residency_ns(i915, GEN6_GT_GFX_RC6pp); - - return val; -} - -static u64 get_rc6(struct intel_gt *gt) -{ -#if IS_ENABLED(CONFIG_PM) - struct drm_i915_private *i915 = gt->i915; - struct intel_runtime_pm *rpm = &i915->runtime_pm; - struct i915_pmu *pmu = &i915->pmu; - intel_wakeref_t wakeref; - unsigned long flags; - u64 val; - - wakeref = intel_runtime_pm_get_if_in_use(rpm); - if (wakeref) { - val = __get_rc6(gt); - intel_runtime_pm_put(rpm, wakeref); - - /* - * If we are coming back from being runtime suspended we must - * be careful not to report a larger value than returned - * previously. - */ - - spin_lock_irqsave(&pmu->lock, flags); - - if (val >= pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur) { - pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur = 0; - pmu->sample[__I915_SAMPLE_RC6].cur = val; - } else { - val = pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur; - } - - spin_unlock_irqrestore(&pmu->lock, flags); - } else { - struct device *kdev = rpm->kdev; - - /* - * We are runtime suspended. - * - * Report the delta from when the device was suspended to now, - * on top of the last known real value, as the approximated RC6 - * counter value. - */ - spin_lock_irqsave(&pmu->lock, flags); - - /* - * After the above branch intel_runtime_pm_get_if_in_use failed - * to get the runtime PM reference we cannot assume we are in - * runtime suspend since we can either: a) race with coming out - * of it before we took the power.lock, or b) there are other - * states than suspended which can bring us here. - * - * We need to double-check that we are indeed currently runtime - * suspended and if not we cannot do better than report the last - * known RC6 value. - */ - if (pm_runtime_status_suspended(kdev)) { - val = pm_runtime_suspended_time(kdev); - - if (!pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur) - pmu->suspended_time_last = val; - - val -= pmu->suspended_time_last; - val += pmu->sample[__I915_SAMPLE_RC6].cur; - - pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur = val; - } else if (pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur) { - val = pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur; - } else { - val = pmu->sample[__I915_SAMPLE_RC6].cur; - } - - spin_unlock_irqrestore(&pmu->lock, flags); - } - - return val; -#else - return __get_rc6(gt); -#endif -} - static u64 __i915_pmu_event_read(struct perf_event *event) { struct drm_i915_private *i915 = @@ -843,8 +878,8 @@ create_event_attributes(struct i915_pmu *pmu) const char *name; const char *unit; } events[] = { - __event(I915_PMU_ACTUAL_FREQUENCY, "actual-frequency", "MHz"), - __event(I915_PMU_REQUESTED_FREQUENCY, "requested-frequency", "MHz"), + __event(I915_PMU_ACTUAL_FREQUENCY, "actual-frequency", "M"), + __event(I915_PMU_REQUESTED_FREQUENCY, "requested-frequency", "M"), __event(I915_PMU_INTERRUPTS, "interrupts", NULL), __event(I915_PMU_RC6_RESIDENCY, "rc6-residency", "ns"), }; @@ -1047,21 +1082,43 @@ static void i915_pmu_unregister_cpuhp_state(struct i915_pmu *pmu) cpuhp_remove_multi_state(cpuhp_slot); } +static bool is_igp(struct drm_i915_private *i915) +{ + struct pci_dev *pdev = i915->drm.pdev; + + /* IGP is 0000:00:02.0 */ + return pci_domain_nr(pdev->bus) == 0 && + pdev->bus->number == 0 && + PCI_SLOT(pdev->devfn) == 2 && + PCI_FUNC(pdev->devfn) == 0; +} + void i915_pmu_register(struct drm_i915_private *i915) { struct i915_pmu *pmu = &i915->pmu; - int ret; + int ret = -ENOMEM; if (INTEL_GEN(i915) <= 2) { dev_info(i915->drm.dev, "PMU not supported for this GPU."); return; } - i915_pmu_events_attr_group.attrs = create_event_attributes(pmu); - if (!i915_pmu_events_attr_group.attrs) { - ret = -ENOMEM; + spin_lock_init(&pmu->lock); + hrtimer_init(&pmu->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + pmu->timer.function = i915_sample; + + if (!is_igp(i915)) + pmu->name = kasprintf(GFP_KERNEL, + "i915-%s", + dev_name(i915->drm.dev)); + else + pmu->name = "i915"; + if (!pmu->name) goto err; - } + + i915_pmu_events_attr_group.attrs = create_event_attributes(pmu); + if (!i915_pmu_events_attr_group.attrs) + goto err_name; pmu->base.attr_groups = i915_pmu_attr_groups; pmu->base.task_ctx_nr = perf_invalid_context; @@ -1073,13 +1130,9 @@ void i915_pmu_register(struct drm_i915_private *i915) pmu->base.read = i915_pmu_event_read; pmu->base.event_idx = i915_pmu_event_event_idx; - spin_lock_init(&pmu->lock); - hrtimer_init(&pmu->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); - pmu->timer.function = i915_sample; - - ret = perf_pmu_register(&pmu->base, "i915", -1); + ret = perf_pmu_register(&pmu->base, pmu->name, -1); if (ret) - goto err; + goto err_attr; ret = i915_pmu_register_cpuhp_state(pmu); if (ret) @@ -1089,10 +1142,14 @@ void i915_pmu_register(struct drm_i915_private *i915) err_unreg: perf_pmu_unregister(&pmu->base); -err: +err_attr: pmu->base.event_init = NULL; free_event_attributes(pmu); - DRM_NOTE("Failed to register PMU! (err=%d)\n", ret); +err_name: + if (!is_igp(i915)) + kfree(pmu->name); +err: + dev_notice(i915->drm.dev, "Failed to register PMU!\n"); } void i915_pmu_unregister(struct drm_i915_private *i915) @@ -1110,5 +1167,7 @@ void i915_pmu_unregister(struct drm_i915_private *i915) perf_pmu_unregister(&pmu->base); pmu->base.event_init = NULL; + if (!is_igp(i915)) + kfree(pmu->name); free_event_attributes(pmu); } diff --git a/drivers/gpu/drm/i915/i915_pmu.h b/drivers/gpu/drm/i915/i915_pmu.h index 4fc4f2478301..bf52e3983631 100644 --- a/drivers/gpu/drm/i915/i915_pmu.h +++ b/drivers/gpu/drm/i915/i915_pmu.h @@ -47,6 +47,10 @@ struct i915_pmu { */ struct pmu base; /** + * @name: Name as registered with perf core. + */ + const char *name; + /** * @lock: Lock protecting enable mask and ref count handling. */ spinlock_t lock; @@ -97,9 +101,9 @@ struct i915_pmu { */ struct i915_pmu_sample sample[__I915_NUM_PMU_SAMPLERS]; /** - * @suspended_time_last: Cached suspend time from PM core. + * @sleep_last: Last time GT parked for RC6 estimation. */ - u64 suspended_time_last; + ktime_t sleep_last; /** * @i915_attr: Memory block holding device attributes. */ diff --git a/drivers/gpu/drm/i915/i915_priolist_types.h b/drivers/gpu/drm/i915/i915_priolist_types.h index 21037a2e2038..732aad148881 100644 --- a/drivers/gpu/drm/i915/i915_priolist_types.h +++ b/drivers/gpu/drm/i915/i915_priolist_types.h @@ -16,6 +16,12 @@ enum { I915_PRIORITY_MIN = I915_CONTEXT_MIN_USER_PRIORITY - 1, I915_PRIORITY_NORMAL = I915_CONTEXT_DEFAULT_PRIORITY, I915_PRIORITY_MAX = I915_CONTEXT_MAX_USER_PRIORITY + 1, + + /* A preemptive pulse used to monitor the health of each engine */ + I915_PRIORITY_HEARTBEAT, + + /* Interactive workload, scheduled for immediate pageflipping */ + I915_PRIORITY_DISPLAY, }; #define I915_USER_PRIORITY_SHIFT 2 @@ -39,6 +45,7 @@ enum { * active request. */ #define I915_PRIORITY_UNPREEMPTABLE INT_MAX +#define I915_PRIORITY_BARRIER INT_MAX #define __NO_PREEMPTION (I915_PRIORITY_WAIT) diff --git a/drivers/gpu/drm/i915/i915_query.c b/drivers/gpu/drm/i915/i915_query.c index ad9240a0817a..ef25ce6e395e 100644 --- a/drivers/gpu/drm/i915/i915_query.c +++ b/drivers/gpu/drm/i915/i915_query.c @@ -7,6 +7,7 @@ #include <linux/nospec.h> #include "i915_drv.h" +#include "i915_perf.h" #include "i915_query.h" #include <uapi/drm/i915_drm.h> @@ -37,8 +38,6 @@ static int query_topology_info(struct drm_i915_private *dev_priv, const struct sseu_dev_info *sseu = &RUNTIME_INFO(dev_priv)->sseu; struct drm_i915_query_topology_info topo; u32 slice_length, subslice_length, eu_length, total_length; - u8 subslice_stride = GEN_SSEU_STRIDE(sseu->max_subslices); - u8 eu_stride = GEN_SSEU_STRIDE(sseu->max_eus_per_subslice); int ret; if (query_item->flags != 0) @@ -50,8 +49,8 @@ static int query_topology_info(struct drm_i915_private *dev_priv, BUILD_BUG_ON(sizeof(u8) != sizeof(sseu->slice_mask)); slice_length = sizeof(sseu->slice_mask); - subslice_length = sseu->max_slices * subslice_stride; - eu_length = sseu->max_slices * sseu->max_subslices * eu_stride; + subslice_length = sseu->max_slices * sseu->ss_stride; + eu_length = sseu->max_slices * sseu->max_subslices * sseu->eu_stride; total_length = sizeof(topo) + slice_length + subslice_length + eu_length; @@ -69,9 +68,9 @@ static int query_topology_info(struct drm_i915_private *dev_priv, topo.max_eus_per_subslice = sseu->max_eus_per_subslice; topo.subslice_offset = slice_length; - topo.subslice_stride = subslice_stride; + topo.subslice_stride = sseu->ss_stride; topo.eu_offset = slice_length + subslice_length; - topo.eu_stride = eu_stride; + topo.eu_stride = sseu->eu_stride; if (__copy_to_user(u64_to_user_ptr(query_item->data_ptr), &topo, sizeof(topo))) @@ -104,15 +103,18 @@ query_engine_info(struct drm_i915_private *i915, struct drm_i915_engine_info __user *info_ptr; struct drm_i915_query_engine_info query; struct drm_i915_engine_info info = { }; + unsigned int num_uabi_engines = 0; struct intel_engine_cs *engine; int len, ret; if (query_item->flags) return -EINVAL; + for_each_uabi_engine(engine, i915) + num_uabi_engines++; + len = sizeof(struct drm_i915_query_engine_info) + - RUNTIME_INFO(i915)->num_engines * - sizeof(struct drm_i915_engine_info); + num_uabi_engines * sizeof(struct drm_i915_engine_info); ret = copy_query_item(&query, sizeof(query), len, query_item); if (ret != 0) @@ -142,10 +144,305 @@ query_engine_info(struct drm_i915_private *i915, return len; } +static int can_copy_perf_config_registers_or_number(u32 user_n_regs, + u64 user_regs_ptr, + u32 kernel_n_regs) +{ + /* + * We'll just put the number of registers, and won't copy the + * register. + */ + if (user_n_regs == 0) + return 0; + + if (user_n_regs < kernel_n_regs) + return -EINVAL; + + if (!access_ok(u64_to_user_ptr(user_regs_ptr), + 2 * sizeof(u32) * kernel_n_regs)) + return -EFAULT; + + return 0; +} + +static int copy_perf_config_registers_or_number(const struct i915_oa_reg *kernel_regs, + u32 kernel_n_regs, + u64 user_regs_ptr, + u32 *user_n_regs) +{ + u32 r; + + if (*user_n_regs == 0) { + *user_n_regs = kernel_n_regs; + return 0; + } + + *user_n_regs = kernel_n_regs; + + for (r = 0; r < kernel_n_regs; r++) { + u32 __user *user_reg_ptr = + u64_to_user_ptr(user_regs_ptr + sizeof(u32) * r * 2); + u32 __user *user_val_ptr = + u64_to_user_ptr(user_regs_ptr + sizeof(u32) * r * 2 + + sizeof(u32)); + int ret; + + ret = __put_user(i915_mmio_reg_offset(kernel_regs[r].addr), + user_reg_ptr); + if (ret) + return -EFAULT; + + ret = __put_user(kernel_regs[r].value, user_val_ptr); + if (ret) + return -EFAULT; + } + + return 0; +} + +static int query_perf_config_data(struct drm_i915_private *i915, + struct drm_i915_query_item *query_item, + bool use_uuid) +{ + struct drm_i915_query_perf_config __user *user_query_config_ptr = + u64_to_user_ptr(query_item->data_ptr); + struct drm_i915_perf_oa_config __user *user_config_ptr = + u64_to_user_ptr(query_item->data_ptr + + sizeof(struct drm_i915_query_perf_config)); + struct drm_i915_perf_oa_config user_config; + struct i915_perf *perf = &i915->perf; + struct i915_oa_config *oa_config; + char uuid[UUID_STRING_LEN + 1]; + u64 config_id; + u32 flags, total_size; + int ret; + + if (!perf->i915) + return -ENODEV; + + total_size = + sizeof(struct drm_i915_query_perf_config) + + sizeof(struct drm_i915_perf_oa_config); + + if (query_item->length == 0) + return total_size; + + if (query_item->length < total_size) { + DRM_DEBUG("Invalid query config data item size=%u expected=%u\n", + query_item->length, total_size); + return -EINVAL; + } + + if (!access_ok(user_query_config_ptr, total_size)) + return -EFAULT; + + if (__get_user(flags, &user_query_config_ptr->flags)) + return -EFAULT; + + if (flags != 0) + return -EINVAL; + + if (use_uuid) { + struct i915_oa_config *tmp; + int id; + + BUILD_BUG_ON(sizeof(user_query_config_ptr->uuid) >= sizeof(uuid)); + + memset(&uuid, 0, sizeof(uuid)); + if (__copy_from_user(uuid, user_query_config_ptr->uuid, + sizeof(user_query_config_ptr->uuid))) + return -EFAULT; + + oa_config = NULL; + rcu_read_lock(); + idr_for_each_entry(&perf->metrics_idr, tmp, id) { + if (!strcmp(tmp->uuid, uuid)) { + oa_config = i915_oa_config_get(tmp); + break; + } + } + rcu_read_unlock(); + } else { + if (__get_user(config_id, &user_query_config_ptr->config)) + return -EFAULT; + + oa_config = i915_perf_get_oa_config(perf, config_id); + } + if (!oa_config) + return -ENOENT; + + if (__copy_from_user(&user_config, user_config_ptr, + sizeof(user_config))) { + ret = -EFAULT; + goto out; + } + + ret = can_copy_perf_config_registers_or_number(user_config.n_boolean_regs, + user_config.boolean_regs_ptr, + oa_config->b_counter_regs_len); + if (ret) + goto out; + + ret = can_copy_perf_config_registers_or_number(user_config.n_flex_regs, + user_config.flex_regs_ptr, + oa_config->flex_regs_len); + if (ret) + goto out; + + ret = can_copy_perf_config_registers_or_number(user_config.n_mux_regs, + user_config.mux_regs_ptr, + oa_config->mux_regs_len); + if (ret) + goto out; + + ret = copy_perf_config_registers_or_number(oa_config->b_counter_regs, + oa_config->b_counter_regs_len, + user_config.boolean_regs_ptr, + &user_config.n_boolean_regs); + if (ret) + goto out; + + ret = copy_perf_config_registers_or_number(oa_config->flex_regs, + oa_config->flex_regs_len, + user_config.flex_regs_ptr, + &user_config.n_flex_regs); + if (ret) + goto out; + + ret = copy_perf_config_registers_or_number(oa_config->mux_regs, + oa_config->mux_regs_len, + user_config.mux_regs_ptr, + &user_config.n_mux_regs); + if (ret) + goto out; + + memcpy(user_config.uuid, oa_config->uuid, sizeof(user_config.uuid)); + + if (__copy_to_user(user_config_ptr, &user_config, + sizeof(user_config))) { + ret = -EFAULT; + goto out; + } + + ret = total_size; + +out: + i915_oa_config_put(oa_config); + return ret; +} + +static size_t sizeof_perf_config_list(size_t count) +{ + return sizeof(struct drm_i915_query_perf_config) + sizeof(u64) * count; +} + +static size_t sizeof_perf_metrics(struct i915_perf *perf) +{ + struct i915_oa_config *tmp; + size_t i; + int id; + + i = 1; + rcu_read_lock(); + idr_for_each_entry(&perf->metrics_idr, tmp, id) + i++; + rcu_read_unlock(); + + return sizeof_perf_config_list(i); +} + +static int query_perf_config_list(struct drm_i915_private *i915, + struct drm_i915_query_item *query_item) +{ + struct drm_i915_query_perf_config __user *user_query_config_ptr = + u64_to_user_ptr(query_item->data_ptr); + struct i915_perf *perf = &i915->perf; + u64 *oa_config_ids = NULL; + int alloc, n_configs; + u32 flags; + int ret; + + if (!perf->i915) + return -ENODEV; + + if (query_item->length == 0) + return sizeof_perf_metrics(perf); + + if (get_user(flags, &user_query_config_ptr->flags)) + return -EFAULT; + + if (flags != 0) + return -EINVAL; + + n_configs = 1; + do { + struct i915_oa_config *tmp; + u64 *ids; + int id; + + ids = krealloc(oa_config_ids, + n_configs * sizeof(*oa_config_ids), + GFP_KERNEL); + if (!ids) + return -ENOMEM; + + alloc = fetch_and_zero(&n_configs); + + ids[n_configs++] = 1ull; /* reserved for test_config */ + rcu_read_lock(); + idr_for_each_entry(&perf->metrics_idr, tmp, id) { + if (n_configs < alloc) + ids[n_configs] = id; + n_configs++; + } + rcu_read_unlock(); + + oa_config_ids = ids; + } while (n_configs > alloc); + + if (query_item->length < sizeof_perf_config_list(n_configs)) { + DRM_DEBUG("Invalid query config list item size=%u expected=%zu\n", + query_item->length, + sizeof_perf_config_list(n_configs)); + kfree(oa_config_ids); + return -EINVAL; + } + + if (put_user(n_configs, &user_query_config_ptr->config)) { + kfree(oa_config_ids); + return -EFAULT; + } + + ret = copy_to_user(user_query_config_ptr + 1, + oa_config_ids, + n_configs * sizeof(*oa_config_ids)); + kfree(oa_config_ids); + if (ret) + return -EFAULT; + + return sizeof_perf_config_list(n_configs); +} + +static int query_perf_config(struct drm_i915_private *i915, + struct drm_i915_query_item *query_item) +{ + switch (query_item->flags) { + case DRM_I915_QUERY_PERF_CONFIG_LIST: + return query_perf_config_list(i915, query_item); + case DRM_I915_QUERY_PERF_CONFIG_DATA_FOR_UUID: + return query_perf_config_data(i915, query_item, true); + case DRM_I915_QUERY_PERF_CONFIG_DATA_FOR_ID: + return query_perf_config_data(i915, query_item, false); + default: + return -EINVAL; + } +} + static int (* const i915_query_funcs[])(struct drm_i915_private *dev_priv, struct drm_i915_query_item *query_item) = { query_topology_info, query_engine_info, + query_perf_config, }; int i915_query_ioctl(struct drm_device *dev, void *data, struct drm_file *file) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 2abd199093c5..73079b503724 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -413,6 +413,9 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define GEN11_VECS_SFC_USAGE(engine) _MMIO((engine)->mmio_base + 0x2014) #define GEN11_VECS_SFC_USAGE_BIT (1 << 0) +#define GEN12_SFC_DONE(n) _MMIO(0x1cc00 + (n) * 0x100) +#define GEN12_SFC_DONE_MAX 4 + #define RING_PP_DIR_BASE(base) _MMIO((base) + 0x228) #define RING_PP_DIR_BASE_READ(base) _MMIO((base) + 0x518) #define RING_PP_DIR_DCLV(base) _MMIO((base) + 0x220) @@ -471,6 +474,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define ECOCHK_PPGTT_WT_HSW (0x2 << 3) #define ECOCHK_PPGTT_WB_HSW (0x3 << 3) +#define GEN8_RC6_CTX_INFO _MMIO(0x8504) + #define GAC_ECO_BITS _MMIO(0x14090) #define ECOBITS_SNB_BIT (1 << 13) #define ECOBITS_PPGTT_CACHE64B (3 << 8) @@ -545,7 +550,9 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MI_PREDICATE_SRC0_UDW _MMIO(0x2400 + 4) #define MI_PREDICATE_SRC1 _MMIO(0x2408) #define MI_PREDICATE_SRC1_UDW _MMIO(0x2408 + 4) - +#define MI_PREDICATE_DATA _MMIO(0x2410) +#define MI_PREDICATE_RESULT _MMIO(0x2418) +#define MI_PREDICATE_RESULT_1 _MMIO(0x241c) #define MI_PREDICATE_RESULT_2 _MMIO(0x2214) #define LOWER_SLICE_ENABLED (1 << 0) #define LOWER_SLICE_DISABLED (0 << 0) @@ -555,6 +562,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) */ #define BCS_SWCTRL _MMIO(0x22200) +/* There are 16 GPR registers */ +#define BCS_GPR(n) _MMIO(0x22600 + (n) * 8) +#define BCS_GPR_UDW(n) _MMIO(0x22600 + (n) * 8 + 4) + #define GPGPU_THREADS_DISPATCHED _MMIO(0x2290) #define GPGPU_THREADS_DISPATCHED_UDW _MMIO(0x2290 + 4) #define HS_INVOCATION_COUNT _MMIO(0x2300) @@ -682,6 +693,45 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define OABUFFER_SIZE_8M (6 << 3) #define OABUFFER_SIZE_16M (7 << 3) +/* Gen12 OAR unit */ +#define GEN12_OAR_OACONTROL _MMIO(0x2960) +#define GEN12_OAR_OACONTROL_COUNTER_FORMAT_SHIFT 1 +#define GEN12_OAR_OACONTROL_COUNTER_ENABLE (1 << 0) + +#define GEN12_OACTXCONTROL _MMIO(0x2360) +#define GEN12_OAR_OASTATUS _MMIO(0x2968) + +/* Gen12 OAG unit */ +#define GEN12_OAG_OAHEADPTR _MMIO(0xdb00) +#define GEN12_OAG_OAHEADPTR_MASK 0xffffffc0 +#define GEN12_OAG_OATAILPTR _MMIO(0xdb04) +#define GEN12_OAG_OATAILPTR_MASK 0xffffffc0 + +#define GEN12_OAG_OABUFFER _MMIO(0xdb08) +#define GEN12_OAG_OABUFFER_BUFFER_SIZE_MASK (0x7) +#define GEN12_OAG_OABUFFER_BUFFER_SIZE_SHIFT (3) +#define GEN12_OAG_OABUFFER_MEMORY_SELECT (1 << 0) /* 0: PPGTT, 1: GGTT */ + +#define GEN12_OAG_OAGLBCTXCTRL _MMIO(0x2b28) +#define GEN12_OAG_OAGLBCTXCTRL_TIMER_PERIOD_SHIFT 2 +#define GEN12_OAG_OAGLBCTXCTRL_TIMER_ENABLE (1 << 1) +#define GEN12_OAG_OAGLBCTXCTRL_COUNTER_RESUME (1 << 0) + +#define GEN12_OAG_OACONTROL _MMIO(0xdaf4) +#define GEN12_OAG_OACONTROL_OA_COUNTER_FORMAT_SHIFT 2 +#define GEN12_OAG_OACONTROL_OA_COUNTER_ENABLE (1 << 0) + +#define GEN12_OAG_OA_DEBUG _MMIO(0xdaf8) +#define GEN12_OAG_OA_DEBUG_INCLUDE_CLK_RATIO (1 << 6) +#define GEN12_OAG_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS (1 << 5) +#define GEN12_OAG_OA_DEBUG_DISABLE_GO_1_0_REPORTS (1 << 2) +#define GEN12_OAG_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS (1 << 1) + +#define GEN12_OAG_OASTATUS _MMIO(0xdafc) +#define GEN12_OAG_OASTATUS_COUNTER_OVERFLOW (1 << 2) +#define GEN12_OAG_OASTATUS_BUFFER_OVERFLOW (1 << 1) +#define GEN12_OAG_OASTATUS_REPORT_LOST (1 << 0) + /* * Flexible, Aggregate EU Counter Registers. * Note: these aren't contiguous @@ -918,6 +968,26 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define OAREPORTTRIG8_NOA_SELECT_6_SHIFT 24 #define OAREPORTTRIG8_NOA_SELECT_7_SHIFT 28 +/* Same layout as OASTARTTRIGX */ +#define GEN12_OAG_OASTARTTRIG1 _MMIO(0xd900) +#define GEN12_OAG_OASTARTTRIG2 _MMIO(0xd904) +#define GEN12_OAG_OASTARTTRIG3 _MMIO(0xd908) +#define GEN12_OAG_OASTARTTRIG4 _MMIO(0xd90c) +#define GEN12_OAG_OASTARTTRIG5 _MMIO(0xd910) +#define GEN12_OAG_OASTARTTRIG6 _MMIO(0xd914) +#define GEN12_OAG_OASTARTTRIG7 _MMIO(0xd918) +#define GEN12_OAG_OASTARTTRIG8 _MMIO(0xd91c) + +/* Same layout as OAREPORTTRIGX */ +#define GEN12_OAG_OAREPORTTRIG1 _MMIO(0xd920) +#define GEN12_OAG_OAREPORTTRIG2 _MMIO(0xd924) +#define GEN12_OAG_OAREPORTTRIG3 _MMIO(0xd928) +#define GEN12_OAG_OAREPORTTRIG4 _MMIO(0xd92c) +#define GEN12_OAG_OAREPORTTRIG5 _MMIO(0xd930) +#define GEN12_OAG_OAREPORTTRIG6 _MMIO(0xd934) +#define GEN12_OAG_OAREPORTTRIG7 _MMIO(0xd938) +#define GEN12_OAG_OAREPORTTRIG8 _MMIO(0xd93c) + /* CECX_0 */ #define OACEC_COMPARE_LESS_OR_EQUAL 6 #define OACEC_COMPARE_NOT_EQUAL 5 @@ -934,6 +1004,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define OACEC_SELECT_PREV (1 << 19) #define OACEC_SELECT_BOOLEAN (2 << 19) +/* 11-bit array 0: pass-through, 1: negated */ +#define GEN12_OASCEC_NEGATE_MASK 0x7ff +#define GEN12_OASCEC_NEGATE_SHIFT 21 + /* CECX_1 */ #define OACEC_MASK_MASK 0xffff #define OACEC_CONSIDERATIONS_MASK 0xffff @@ -956,6 +1030,42 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define OACEC7_0 _MMIO(0x27a8) #define OACEC7_1 _MMIO(0x27ac) +/* Same layout as CECX_Y */ +#define GEN12_OAG_CEC0_0 _MMIO(0xd940) +#define GEN12_OAG_CEC0_1 _MMIO(0xd944) +#define GEN12_OAG_CEC1_0 _MMIO(0xd948) +#define GEN12_OAG_CEC1_1 _MMIO(0xd94c) +#define GEN12_OAG_CEC2_0 _MMIO(0xd950) +#define GEN12_OAG_CEC2_1 _MMIO(0xd954) +#define GEN12_OAG_CEC3_0 _MMIO(0xd958) +#define GEN12_OAG_CEC3_1 _MMIO(0xd95c) +#define GEN12_OAG_CEC4_0 _MMIO(0xd960) +#define GEN12_OAG_CEC4_1 _MMIO(0xd964) +#define GEN12_OAG_CEC5_0 _MMIO(0xd968) +#define GEN12_OAG_CEC5_1 _MMIO(0xd96c) +#define GEN12_OAG_CEC6_0 _MMIO(0xd970) +#define GEN12_OAG_CEC6_1 _MMIO(0xd974) +#define GEN12_OAG_CEC7_0 _MMIO(0xd978) +#define GEN12_OAG_CEC7_1 _MMIO(0xd97c) + +/* Same layout as CECX_Y + negate 11-bit array */ +#define GEN12_OAG_SCEC0_0 _MMIO(0xdc00) +#define GEN12_OAG_SCEC0_1 _MMIO(0xdc04) +#define GEN12_OAG_SCEC1_0 _MMIO(0xdc08) +#define GEN12_OAG_SCEC1_1 _MMIO(0xdc0c) +#define GEN12_OAG_SCEC2_0 _MMIO(0xdc10) +#define GEN12_OAG_SCEC2_1 _MMIO(0xdc14) +#define GEN12_OAG_SCEC3_0 _MMIO(0xdc18) +#define GEN12_OAG_SCEC3_1 _MMIO(0xdc1c) +#define GEN12_OAG_SCEC4_0 _MMIO(0xdc20) +#define GEN12_OAG_SCEC4_1 _MMIO(0xdc24) +#define GEN12_OAG_SCEC5_0 _MMIO(0xdc28) +#define GEN12_OAG_SCEC5_1 _MMIO(0xdc2c) +#define GEN12_OAG_SCEC6_0 _MMIO(0xdc30) +#define GEN12_OAG_SCEC6_1 _MMIO(0xdc34) +#define GEN12_OAG_SCEC7_0 _MMIO(0xdc38) +#define GEN12_OAG_SCEC7_1 _MMIO(0xdc3c) + /* OA perf counters */ #define OA_PERFCNT1_LO _MMIO(0x91B8) #define OA_PERFCNT1_HI _MMIO(0x91BC) @@ -1036,6 +1146,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MICRO_BP3_COUNT_STATUS23 _MMIO(0x9838) #define MICRO_BP_FIRED_ARMED _MMIO(0x983C) +#define GEN12_OAA_DBG_REG _MMIO(0xdc44) +#define GEN12_OAG_OA_PESS _MMIO(0x2b2c) +#define GEN12_OAG_SPCTR_CNF _MMIO(0xdc40) + #define GDT_CHICKEN_BITS _MMIO(0x9840) #define GT_NOA_ENABLE 0x00000080 @@ -1956,8 +2070,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define ICL_DPHY_CHKN(port) _MMIO(_ICL_COMBOPHY(port) + _ICL_DPHY_CHKN_REG) #define ICL_DPHY_CHKN_AFE_OVER_PPI_STRAP REG_BIT(7) -#define MG_PHY_PORT_LN(ln, port, ln0p1, ln0p2, ln1p1) \ - _MMIO(_PORT((port) - PORT_C, ln0p1, ln0p2) + (ln) * ((ln1p1) - (ln0p1))) +#define MG_PHY_PORT_LN(ln, tc_port, ln0p1, ln0p2, ln1p1) \ + _MMIO(_PORT(tc_port, ln0p1, ln0p2) + (ln) * ((ln1p1) - (ln0p1))) #define MG_TX_LINK_PARAMS_TX1LN0_PORT1 0x16812C #define MG_TX_LINK_PARAMS_TX1LN1_PORT1 0x16852C @@ -1967,10 +2081,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_LINK_PARAMS_TX1LN1_PORT3 0x16A52C #define MG_TX_LINK_PARAMS_TX1LN0_PORT4 0x16B12C #define MG_TX_LINK_PARAMS_TX1LN1_PORT4 0x16B52C -#define MG_TX1_LINK_PARAMS(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_LINK_PARAMS_TX1LN0_PORT1, \ - MG_TX_LINK_PARAMS_TX1LN0_PORT2, \ - MG_TX_LINK_PARAMS_TX1LN1_PORT1) +#define MG_TX1_LINK_PARAMS(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_LINK_PARAMS_TX1LN0_PORT1, \ + MG_TX_LINK_PARAMS_TX1LN0_PORT2, \ + MG_TX_LINK_PARAMS_TX1LN1_PORT1) #define MG_TX_LINK_PARAMS_TX2LN0_PORT1 0x1680AC #define MG_TX_LINK_PARAMS_TX2LN1_PORT1 0x1684AC @@ -1980,10 +2094,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_LINK_PARAMS_TX2LN1_PORT3 0x16A4AC #define MG_TX_LINK_PARAMS_TX2LN0_PORT4 0x16B0AC #define MG_TX_LINK_PARAMS_TX2LN1_PORT4 0x16B4AC -#define MG_TX2_LINK_PARAMS(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_LINK_PARAMS_TX2LN0_PORT1, \ - MG_TX_LINK_PARAMS_TX2LN0_PORT2, \ - MG_TX_LINK_PARAMS_TX2LN1_PORT1) +#define MG_TX2_LINK_PARAMS(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_LINK_PARAMS_TX2LN0_PORT1, \ + MG_TX_LINK_PARAMS_TX2LN0_PORT2, \ + MG_TX_LINK_PARAMS_TX2LN1_PORT1) #define CRI_USE_FS32 (1 << 5) #define MG_TX_PISO_READLOAD_TX1LN0_PORT1 0x16814C @@ -1994,10 +2108,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_PISO_READLOAD_TX1LN1_PORT3 0x16A54C #define MG_TX_PISO_READLOAD_TX1LN0_PORT4 0x16B14C #define MG_TX_PISO_READLOAD_TX1LN1_PORT4 0x16B54C -#define MG_TX1_PISO_READLOAD(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_PISO_READLOAD_TX1LN0_PORT1, \ - MG_TX_PISO_READLOAD_TX1LN0_PORT2, \ - MG_TX_PISO_READLOAD_TX1LN1_PORT1) +#define MG_TX1_PISO_READLOAD(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_PISO_READLOAD_TX1LN0_PORT1, \ + MG_TX_PISO_READLOAD_TX1LN0_PORT2, \ + MG_TX_PISO_READLOAD_TX1LN1_PORT1) #define MG_TX_PISO_READLOAD_TX2LN0_PORT1 0x1680CC #define MG_TX_PISO_READLOAD_TX2LN1_PORT1 0x1684CC @@ -2007,10 +2121,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_PISO_READLOAD_TX2LN1_PORT3 0x16A4CC #define MG_TX_PISO_READLOAD_TX2LN0_PORT4 0x16B0CC #define MG_TX_PISO_READLOAD_TX2LN1_PORT4 0x16B4CC -#define MG_TX2_PISO_READLOAD(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_PISO_READLOAD_TX2LN0_PORT1, \ - MG_TX_PISO_READLOAD_TX2LN0_PORT2, \ - MG_TX_PISO_READLOAD_TX2LN1_PORT1) +#define MG_TX2_PISO_READLOAD(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_PISO_READLOAD_TX2LN0_PORT1, \ + MG_TX_PISO_READLOAD_TX2LN0_PORT2, \ + MG_TX_PISO_READLOAD_TX2LN1_PORT1) #define CRI_CALCINIT (1 << 1) #define MG_TX_SWINGCTRL_TX1LN0_PORT1 0x168148 @@ -2021,10 +2135,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_SWINGCTRL_TX1LN1_PORT3 0x16A548 #define MG_TX_SWINGCTRL_TX1LN0_PORT4 0x16B148 #define MG_TX_SWINGCTRL_TX1LN1_PORT4 0x16B548 -#define MG_TX1_SWINGCTRL(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_SWINGCTRL_TX1LN0_PORT1, \ - MG_TX_SWINGCTRL_TX1LN0_PORT2, \ - MG_TX_SWINGCTRL_TX1LN1_PORT1) +#define MG_TX1_SWINGCTRL(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_SWINGCTRL_TX1LN0_PORT1, \ + MG_TX_SWINGCTRL_TX1LN0_PORT2, \ + MG_TX_SWINGCTRL_TX1LN1_PORT1) #define MG_TX_SWINGCTRL_TX2LN0_PORT1 0x1680C8 #define MG_TX_SWINGCTRL_TX2LN1_PORT1 0x1684C8 @@ -2034,10 +2148,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_SWINGCTRL_TX2LN1_PORT3 0x16A4C8 #define MG_TX_SWINGCTRL_TX2LN0_PORT4 0x16B0C8 #define MG_TX_SWINGCTRL_TX2LN1_PORT4 0x16B4C8 -#define MG_TX2_SWINGCTRL(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_SWINGCTRL_TX2LN0_PORT1, \ - MG_TX_SWINGCTRL_TX2LN0_PORT2, \ - MG_TX_SWINGCTRL_TX2LN1_PORT1) +#define MG_TX2_SWINGCTRL(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_SWINGCTRL_TX2LN0_PORT1, \ + MG_TX_SWINGCTRL_TX2LN0_PORT2, \ + MG_TX_SWINGCTRL_TX2LN1_PORT1) #define CRI_TXDEEMPH_OVERRIDE_17_12(x) ((x) << 0) #define CRI_TXDEEMPH_OVERRIDE_17_12_MASK (0x3F << 0) @@ -2049,10 +2163,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_DRVCTRL_TX1LN1_TXPORT3 0x16A544 #define MG_TX_DRVCTRL_TX1LN0_TXPORT4 0x16B144 #define MG_TX_DRVCTRL_TX1LN1_TXPORT4 0x16B544 -#define MG_TX1_DRVCTRL(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_DRVCTRL_TX1LN0_TXPORT1, \ - MG_TX_DRVCTRL_TX1LN0_TXPORT2, \ - MG_TX_DRVCTRL_TX1LN1_TXPORT1) +#define MG_TX1_DRVCTRL(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_DRVCTRL_TX1LN0_TXPORT1, \ + MG_TX_DRVCTRL_TX1LN0_TXPORT2, \ + MG_TX_DRVCTRL_TX1LN1_TXPORT1) #define MG_TX_DRVCTRL_TX2LN0_PORT1 0x1680C4 #define MG_TX_DRVCTRL_TX2LN1_PORT1 0x1684C4 @@ -2062,10 +2176,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_DRVCTRL_TX2LN1_PORT3 0x16A4C4 #define MG_TX_DRVCTRL_TX2LN0_PORT4 0x16B0C4 #define MG_TX_DRVCTRL_TX2LN1_PORT4 0x16B4C4 -#define MG_TX2_DRVCTRL(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_DRVCTRL_TX2LN0_PORT1, \ - MG_TX_DRVCTRL_TX2LN0_PORT2, \ - MG_TX_DRVCTRL_TX2LN1_PORT1) +#define MG_TX2_DRVCTRL(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_DRVCTRL_TX2LN0_PORT1, \ + MG_TX_DRVCTRL_TX2LN0_PORT2, \ + MG_TX_DRVCTRL_TX2LN1_PORT1) #define CRI_TXDEEMPH_OVERRIDE_11_6(x) ((x) << 24) #define CRI_TXDEEMPH_OVERRIDE_11_6_MASK (0x3F << 24) #define CRI_TXDEEMPH_OVERRIDE_EN (1 << 22) @@ -2082,10 +2196,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_CLKHUB_LN1_PORT3 0x16A79C #define MG_CLKHUB_LN0_PORT4 0x16B39C #define MG_CLKHUB_LN1_PORT4 0x16B79C -#define MG_CLKHUB(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_CLKHUB_LN0_PORT1, \ - MG_CLKHUB_LN0_PORT2, \ - MG_CLKHUB_LN1_PORT1) +#define MG_CLKHUB(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_CLKHUB_LN0_PORT1, \ + MG_CLKHUB_LN0_PORT2, \ + MG_CLKHUB_LN1_PORT1) #define CFG_LOW_RATE_LKREN_EN (1 << 11) #define MG_TX_DCC_TX1LN0_PORT1 0x168110 @@ -2096,10 +2210,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_DCC_TX1LN1_PORT3 0x16A510 #define MG_TX_DCC_TX1LN0_PORT4 0x16B110 #define MG_TX_DCC_TX1LN1_PORT4 0x16B510 -#define MG_TX1_DCC(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_DCC_TX1LN0_PORT1, \ - MG_TX_DCC_TX1LN0_PORT2, \ - MG_TX_DCC_TX1LN1_PORT1) +#define MG_TX1_DCC(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_DCC_TX1LN0_PORT1, \ + MG_TX_DCC_TX1LN0_PORT2, \ + MG_TX_DCC_TX1LN1_PORT1) #define MG_TX_DCC_TX2LN0_PORT1 0x168090 #define MG_TX_DCC_TX2LN1_PORT1 0x168490 #define MG_TX_DCC_TX2LN0_PORT2 0x169090 @@ -2108,10 +2222,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_TX_DCC_TX2LN1_PORT3 0x16A490 #define MG_TX_DCC_TX2LN0_PORT4 0x16B090 #define MG_TX_DCC_TX2LN1_PORT4 0x16B490 -#define MG_TX2_DCC(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_TX_DCC_TX2LN0_PORT1, \ - MG_TX_DCC_TX2LN0_PORT2, \ - MG_TX_DCC_TX2LN1_PORT1) +#define MG_TX2_DCC(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_TX_DCC_TX2LN0_PORT1, \ + MG_TX_DCC_TX2LN0_PORT2, \ + MG_TX_DCC_TX2LN1_PORT1) #define CFG_AMI_CK_DIV_OVERRIDE_VAL(x) ((x) << 25) #define CFG_AMI_CK_DIV_OVERRIDE_VAL_MASK (0x3 << 25) #define CFG_AMI_CK_DIV_OVERRIDE_EN (1 << 24) @@ -2124,10 +2238,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define MG_DP_MODE_LN1_ACU_PORT3 0x16A7A0 #define MG_DP_MODE_LN0_ACU_PORT4 0x16B3A0 #define MG_DP_MODE_LN1_ACU_PORT4 0x16B7A0 -#define MG_DP_MODE(ln, port) \ - MG_PHY_PORT_LN(ln, port, MG_DP_MODE_LN0_ACU_PORT1, \ - MG_DP_MODE_LN0_ACU_PORT2, \ - MG_DP_MODE_LN1_ACU_PORT1) +#define MG_DP_MODE(ln, tc_port) \ + MG_PHY_PORT_LN(ln, tc_port, MG_DP_MODE_LN0_ACU_PORT1, \ + MG_DP_MODE_LN0_ACU_PORT2, \ + MG_DP_MODE_LN1_ACU_PORT1) #define MG_DP_MODE_CFG_DP_X2_MODE (1 << 7) #define MG_DP_MODE_CFG_DP_X1_MODE (1 << 6) #define MG_DP_MODE_CFG_TR2PWR_GATING (1 << 5) @@ -2166,13 +2280,13 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define _MMIO_FIA(fia, off) _MMIO(_FIA(fia) + (off)) /* ICL PHY DFLEX registers */ -#define PORT_TX_DFLEXDPMLE1(fia) _MMIO_FIA((fia), 0x008C0) -#define DFLEXDPMLE1_DPMLETC_MASK(tc_port) (0xf << (4 * (tc_port))) -#define DFLEXDPMLE1_DPMLETC_ML0(tc_port) (1 << (4 * (tc_port))) -#define DFLEXDPMLE1_DPMLETC_ML1_0(tc_port) (3 << (4 * (tc_port))) -#define DFLEXDPMLE1_DPMLETC_ML3(tc_port) (8 << (4 * (tc_port))) -#define DFLEXDPMLE1_DPMLETC_ML3_2(tc_port) (12 << (4 * (tc_port))) -#define DFLEXDPMLE1_DPMLETC_ML3_0(tc_port) (15 << (4 * (tc_port))) +#define PORT_TX_DFLEXDPMLE1(fia) _MMIO_FIA((fia), 0x008C0) +#define DFLEXDPMLE1_DPMLETC_MASK(idx) (0xf << (4 * (idx))) +#define DFLEXDPMLE1_DPMLETC_ML0(idx) (1 << (4 * (idx))) +#define DFLEXDPMLE1_DPMLETC_ML1_0(idx) (3 << (4 * (idx))) +#define DFLEXDPMLE1_DPMLETC_ML3(idx) (8 << (4 * (idx))) +#define DFLEXDPMLE1_DPMLETC_ML3_2(idx) (12 << (4 * (idx))) +#define DFLEXDPMLE1_DPMLETC_ML3_0(idx) (15 << (4 * (idx))) /* BXT PHY Ref registers */ #define _PORT_REF_DW3_A 0x16218C @@ -2453,6 +2567,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define RING_FAULT_FAULT_TYPE(x) (((x) >> 1) & 0x3) #define RING_FAULT_VALID (1 << 0) #define DONE_REG _MMIO(0x40b0) +#define GEN12_GAM_DONE _MMIO(0xcf68) #define GEN8_PRIVATE_PAT_LO _MMIO(0x40e0) #define GEN8_PRIVATE_PAT_HI _MMIO(0x40e0 + 4) #define GEN10_PAT_INDEX(index) _MMIO(0x40e0 + (index) * 4) @@ -2483,7 +2598,12 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define RING_WAIT (1 << 11) /* gen3+, PRBx_CTL */ #define RING_WAIT_SEMAPHORE (1 << 10) /* gen6+ */ +/* There are 16 64-bit CS General Purpose Registers per-engine on Gen8+ */ +#define GEN8_RING_CS_GPR(base, n) _MMIO((base) + 0x600 + (n) * 8) +#define GEN8_RING_CS_GPR_UDW(base, n) _MMIO((base) + 0x600 + (n) * 8 + 4) + #define RING_FORCE_TO_NONPRIV(base, i) _MMIO(((base) + 0x4D0) + (i) * 4) +#define RING_FORCE_TO_NONPRIV_ADDRESS_MASK REG_GENMASK(25, 2) #define RING_FORCE_TO_NONPRIV_ACCESS_RW (0 << 28) /* CFL+ & Gen11+ */ #define RING_FORCE_TO_NONPRIV_ACCESS_RD (1 << 28) #define RING_FORCE_TO_NONPRIV_ACCESS_WR (2 << 28) @@ -2596,6 +2716,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define FAULT_VA_HIGH_BITS (0xf << 0) #define FAULT_GTT_SEL (1 << 4) +#define GEN12_AUX_ERR_DBG _MMIO(0x43f4) + #define FPGA_DBG _MMIO(0x42300) #define FPGA_DBG_RM_NOCLAIM (1 << 31) @@ -2705,6 +2827,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define VLV_GU_CTL0 _MMIO(VLV_DISPLAY_BASE + 0x2030) #define VLV_GU_CTL1 _MMIO(VLV_DISPLAY_BASE + 0x2034) #define SCPD0 _MMIO(0x209c) /* 915+ only */ +#define CSTATE_RENDER_CLOCK_GATE_DISABLE (1 << 5) #define GEN2_IER _MMIO(0x20a0) #define GEN2_IIR _MMIO(0x20a4) #define GEN2_IMR _MMIO(0x20a8) @@ -2878,6 +3001,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define GEN6_RC_SLEEP_PSMI_CONTROL _MMIO(0x2050) #define GEN6_PSMI_SLEEP_MSG_DISABLE (1 << 0) +#define GEN12_WAIT_FOR_EVENT_POWER_DOWN_DISABLE REG_BIT(7) #define GEN8_RC_SEMA_IDLE_MSG_DISABLE (1 << 12) #define GEN8_FF_DOP_CLOCK_GATE_DISABLE (1 << 10) @@ -2956,6 +3080,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define GEN11_GT_SUBSLICE_DISABLE _MMIO(0x913C) +#define GEN12_GT_DSS_ENABLE _MMIO(0x913C) + #define GEN6_BSD_SLEEP_PSMI_CONTROL _MMIO(0x12050) #define GEN6_BSD_SLEEP_MSG_DISABLE (1 << 0) #define GEN6_BSD_SLEEP_FLUSH_DISABLE (1 << 2) @@ -3558,6 +3684,9 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define _PALETTE_A 0xa000 #define _PALETTE_B 0xa800 #define _CHV_PALETTE_C 0xc000 +#define PALETTE_RED_MASK REG_GENMASK(23, 16) +#define PALETTE_GREEN_MASK REG_GENMASK(15, 8) +#define PALETTE_BLUE_MASK REG_GENMASK(7, 0) #define PALETTE(pipe, i) _MMIO(DISPLAY_MMIO_BASE(dev_priv) + \ _PICK((pipe), _PALETTE_A, \ _PALETTE_B, _CHV_PALETTE_C) + \ @@ -4038,10 +4167,15 @@ enum { #define SARBUNIT_CLKGATE_DIS (1 << 5) #define RCCUNIT_CLKGATE_DIS (1 << 7) #define MSCUNIT_CLKGATE_DIS (1 << 10) +#define L3_CLKGATE_DIS REG_BIT(16) +#define L3_CR2X_CLKGATE_DIS REG_BIT(17) #define SUBSLICE_UNIT_LEVEL_CLKGATE _MMIO(0x9524) #define GWUNIT_CLKGATE_DIS (1 << 16) +#define SUBSLICE_UNIT_LEVEL_CLKGATE2 _MMIO(0x9528) +#define CPSSUNIT_CLKGATE_DIS REG_BIT(9) + #define UNSLICE_UNIT_LEVEL_CLKGATE _MMIO(0x9434) #define VFUNIT_CLKGATE_DIS (1 << 20) @@ -4135,6 +4269,7 @@ enum { #define _VTOTAL_A 0x6000c #define _VBLANK_A 0x60010 #define _VSYNC_A 0x60014 +#define _EXITLINE_A 0x60018 #define _PIPEASRC 0x6001c #define _BCLRPAT_A 0x60020 #define _VSYNCSHIFT_A 0x60028 @@ -4186,10 +4321,22 @@ enum { #define PIPESRC(trans) _MMIO_TRANS2(trans, _PIPEASRC) #define PIPE_MULT(trans) _MMIO_TRANS2(trans, _PIPE_MULT_A) -/* HSW+ eDP PSR registers */ -#define HSW_EDP_PSR_BASE 0x64800 -#define BDW_EDP_PSR_BASE 0x6f800 -#define EDP_PSR_CTL _MMIO(dev_priv->psr_mmio_base + 0) +#define EXITLINE(trans) _MMIO_TRANS2(trans, _EXITLINE_A) +#define EXITLINE_ENABLE REG_BIT(31) +#define EXITLINE_MASK REG_GENMASK(12, 0) +#define EXITLINE_SHIFT 0 + +/* + * HSW+ eDP PSR registers + * + * HSW PSR registers are relative to DDIA(_DDI_BUF_CTL_A + 0x800) with just one + * instance of it + */ +#define _HSW_EDP_PSR_BASE 0x64800 +#define _SRD_CTL_A 0x60800 +#define _SRD_CTL_EDP 0x6f800 +#define _PSR_ADJ(tran, reg) (_TRANS2(tran, reg) - dev_priv->hsw_psr_mmio_adjust) +#define EDP_PSR_CTL(tran) _MMIO(_PSR_ADJ(tran, _SRD_CTL_A)) #define EDP_PSR_ENABLE (1 << 31) #define BDW_PSR_SINGLE_FRAME (1 << 30) #define EDP_PSR_RESTORE_PSR_ACTIVE_CTX_MASK (1 << 29) /* SW can't modify */ @@ -4215,27 +4362,40 @@ enum { #define EDP_PSR_TP1_TIME_0us (3 << 4) #define EDP_PSR_IDLE_FRAME_SHIFT 0 -/* Bspec claims those aren't shifted but stay at 0x64800 */ +/* + * Until TGL, IMR/IIR are fixed at 0x648xx. On TGL+ those registers are relative + * to transcoder and bits defined for each one as if using no shift (i.e. as if + * it was for TRANSCODER_EDP) + */ #define EDP_PSR_IMR _MMIO(0x64834) #define EDP_PSR_IIR _MMIO(0x64838) -#define EDP_PSR_ERROR(shift) (1 << ((shift) + 2)) -#define EDP_PSR_POST_EXIT(shift) (1 << ((shift) + 1)) -#define EDP_PSR_PRE_ENTRY(shift) (1 << (shift)) -#define EDP_PSR_TRANSCODER_C_SHIFT 24 -#define EDP_PSR_TRANSCODER_B_SHIFT 16 -#define EDP_PSR_TRANSCODER_A_SHIFT 8 -#define EDP_PSR_TRANSCODER_EDP_SHIFT 0 - -#define EDP_PSR_AUX_CTL _MMIO(dev_priv->psr_mmio_base + 0x10) +#define _PSR_IMR_A 0x60814 +#define _PSR_IIR_A 0x60818 +#define TRANS_PSR_IMR(tran) _MMIO_TRANS2(tran, _PSR_IMR_A) +#define TRANS_PSR_IIR(tran) _MMIO_TRANS2(tran, _PSR_IIR_A) +#define _EDP_PSR_TRANS_SHIFT(trans) ((trans) == TRANSCODER_EDP ? \ + 0 : ((trans) - TRANSCODER_A + 1) * 8) +#define EDP_PSR_TRANS_MASK(trans) (0x7 << _EDP_PSR_TRANS_SHIFT(trans)) +#define EDP_PSR_ERROR(trans) (0x4 << _EDP_PSR_TRANS_SHIFT(trans)) +#define EDP_PSR_POST_EXIT(trans) (0x2 << _EDP_PSR_TRANS_SHIFT(trans)) +#define EDP_PSR_PRE_ENTRY(trans) (0x1 << _EDP_PSR_TRANS_SHIFT(trans)) + +#define _SRD_AUX_CTL_A 0x60810 +#define _SRD_AUX_CTL_EDP 0x6f810 +#define EDP_PSR_AUX_CTL(tran) _MMIO(_PSR_ADJ(tran, _SRD_AUX_CTL_A)) #define EDP_PSR_AUX_CTL_TIME_OUT_MASK (3 << 26) #define EDP_PSR_AUX_CTL_MESSAGE_SIZE_MASK (0x1f << 20) #define EDP_PSR_AUX_CTL_PRECHARGE_2US_MASK (0xf << 16) #define EDP_PSR_AUX_CTL_ERROR_INTERRUPT (1 << 11) #define EDP_PSR_AUX_CTL_BIT_CLOCK_2X_MASK (0x7ff) -#define EDP_PSR_AUX_DATA(i) _MMIO(dev_priv->psr_mmio_base + 0x14 + (i) * 4) /* 5 registers */ +#define _SRD_AUX_DATA_A 0x60814 +#define _SRD_AUX_DATA_EDP 0x6f814 +#define EDP_PSR_AUX_DATA(tran, i) _MMIO(_PSR_ADJ(tran, _SRD_AUX_DATA_A) + (i) + 4) /* 5 registers */ -#define EDP_PSR_STATUS _MMIO(dev_priv->psr_mmio_base + 0x40) +#define _SRD_STATUS_A 0x60840 +#define _SRD_STATUS_EDP 0x6f840 +#define EDP_PSR_STATUS(tran) _MMIO(_PSR_ADJ(tran, _SRD_STATUS_A)) #define EDP_PSR_STATUS_STATE_MASK (7 << 29) #define EDP_PSR_STATUS_STATE_SHIFT 29 #define EDP_PSR_STATUS_STATE_IDLE (0 << 29) @@ -4260,10 +4420,15 @@ enum { #define EDP_PSR_STATUS_SENDING_TP1 (1 << 4) #define EDP_PSR_STATUS_IDLE_MASK 0xf -#define EDP_PSR_PERF_CNT _MMIO(dev_priv->psr_mmio_base + 0x44) +#define _SRD_PERF_CNT_A 0x60844 +#define _SRD_PERF_CNT_EDP 0x6f844 +#define EDP_PSR_PERF_CNT(tran) _MMIO(_PSR_ADJ(tran, _SRD_PERF_CNT_A)) #define EDP_PSR_PERF_CNT_MASK 0xffffff -#define EDP_PSR_DEBUG _MMIO(dev_priv->psr_mmio_base + 0x60) /* PSR_MASK on SKL+ */ +/* PSR_MASK on SKL+ */ +#define _SRD_DEBUG_A 0x60860 +#define _SRD_DEBUG_EDP 0x6f860 +#define EDP_PSR_DEBUG(tran) _MMIO(_PSR_ADJ(tran, _SRD_DEBUG_A)) #define EDP_PSR_DEBUG_MASK_MAX_SLEEP (1 << 28) #define EDP_PSR_DEBUG_MASK_LPSP (1 << 27) #define EDP_PSR_DEBUG_MASK_MEMUP (1 << 26) @@ -4271,7 +4436,9 @@ enum { #define EDP_PSR_DEBUG_MASK_DISP_REG_WRITE (1 << 16) /* Reserved in ICL+ */ #define EDP_PSR_DEBUG_EXIT_ON_PIXEL_UNDERRUN (1 << 15) /* SKL+ */ -#define EDP_PSR2_CTL _MMIO(0x6f900) +#define _PSR2_CTL_A 0x60900 +#define _PSR2_CTL_EDP 0x6f900 +#define EDP_PSR2_CTL(tran) _MMIO_TRANS2(tran, _PSR2_CTL_A) #define EDP_PSR2_ENABLE (1 << 31) #define EDP_SU_TRACK_ENABLE (1 << 30) #define EDP_Y_COORDINATE_VALID (1 << 26) /* GLK and CNL+ */ @@ -4293,8 +4460,8 @@ enum { #define _PSR_EVENT_TRANS_B 0x61848 #define _PSR_EVENT_TRANS_C 0x62848 #define _PSR_EVENT_TRANS_D 0x63848 -#define _PSR_EVENT_TRANS_EDP 0x6F848 -#define PSR_EVENT(trans) _MMIO_TRANS2(trans, _PSR_EVENT_TRANS_A) +#define _PSR_EVENT_TRANS_EDP 0x6f848 +#define PSR_EVENT(tran) _MMIO_TRANS2(tran, _PSR_EVENT_TRANS_A) #define PSR_EVENT_PSR2_WD_TIMER_EXPIRE (1 << 17) #define PSR_EVENT_PSR2_DISABLED (1 << 16) #define PSR_EVENT_SU_DIRTY_FIFO_UNDERRUN (1 << 15) @@ -4312,15 +4479,16 @@ enum { #define PSR_EVENT_LPSP_MODE_EXIT (1 << 1) #define PSR_EVENT_PSR_DISABLE (1 << 0) -#define EDP_PSR2_STATUS _MMIO(0x6f940) +#define _PSR2_STATUS_A 0x60940 +#define _PSR2_STATUS_EDP 0x6f940 +#define EDP_PSR2_STATUS(tran) _MMIO_TRANS2(tran, _PSR2_STATUS_A) #define EDP_PSR2_STATUS_STATE_MASK (0xf << 28) #define EDP_PSR2_STATUS_STATE_SHIFT 28 -#define _PSR2_SU_STATUS_0 0x6F914 -#define _PSR2_SU_STATUS_1 0x6F918 -#define _PSR2_SU_STATUS_2 0x6F91C -#define _PSR2_SU_STATUS(index) _MMIO(_PICK_EVEN((index), _PSR2_SU_STATUS_0, _PSR2_SU_STATUS_1)) -#define PSR2_SU_STATUS(frame) (_PSR2_SU_STATUS((frame) / 3)) +#define _PSR2_SU_STATUS_A 0x60914 +#define _PSR2_SU_STATUS_EDP 0x6f914 +#define _PSR2_SU_STATUS(tran, index) _MMIO(_TRANS2(tran, _PSR2_SU_STATUS_A) + (index) * 4) +#define PSR2_SU_STATUS(tran, frame) (_PSR2_SU_STATUS(tran, (frame) / 3)) #define PSR2_SU_STATUS_SHIFT(frame) (((frame) % 3) * 10) #define PSR2_SU_STATUS_MASK(frame) (0x3ff << PSR2_SU_STATUS_SHIFT(frame)) #define PSR2_SU_STATUS_FRAMES 8 @@ -4646,6 +4814,7 @@ enum { * (Haswell and newer) to see which VIDEO_DIP_DATA byte corresponds to each byte * of the infoframe structure specified by CEA-861. */ #define VIDEO_DIP_DATA_SIZE 32 +#define VIDEO_DIP_GMP_DATA_SIZE 36 #define VIDEO_DIP_VSC_DATA_SIZE 36 #define VIDEO_DIP_PPS_DATA_SIZE 132 #define VIDEO_DIP_CTL _MMIO(0x61170) @@ -5482,45 +5651,9 @@ enum { */ #define _DPA_AUX_CH_CTL (DISPLAY_MMIO_BASE(dev_priv) + 0x64010) #define _DPA_AUX_CH_DATA1 (DISPLAY_MMIO_BASE(dev_priv) + 0x64014) -#define _DPA_AUX_CH_DATA2 (DISPLAY_MMIO_BASE(dev_priv) + 0x64018) -#define _DPA_AUX_CH_DATA3 (DISPLAY_MMIO_BASE(dev_priv) + 0x6401c) -#define _DPA_AUX_CH_DATA4 (DISPLAY_MMIO_BASE(dev_priv) + 0x64020) -#define _DPA_AUX_CH_DATA5 (DISPLAY_MMIO_BASE(dev_priv) + 0x64024) #define _DPB_AUX_CH_CTL (DISPLAY_MMIO_BASE(dev_priv) + 0x64110) #define _DPB_AUX_CH_DATA1 (DISPLAY_MMIO_BASE(dev_priv) + 0x64114) -#define _DPB_AUX_CH_DATA2 (DISPLAY_MMIO_BASE(dev_priv) + 0x64118) -#define _DPB_AUX_CH_DATA3 (DISPLAY_MMIO_BASE(dev_priv) + 0x6411c) -#define _DPB_AUX_CH_DATA4 (DISPLAY_MMIO_BASE(dev_priv) + 0x64120) -#define _DPB_AUX_CH_DATA5 (DISPLAY_MMIO_BASE(dev_priv) + 0x64124) - -#define _DPC_AUX_CH_CTL (DISPLAY_MMIO_BASE(dev_priv) + 0x64210) -#define _DPC_AUX_CH_DATA1 (DISPLAY_MMIO_BASE(dev_priv) + 0x64214) -#define _DPC_AUX_CH_DATA2 (DISPLAY_MMIO_BASE(dev_priv) + 0x64218) -#define _DPC_AUX_CH_DATA3 (DISPLAY_MMIO_BASE(dev_priv) + 0x6421c) -#define _DPC_AUX_CH_DATA4 (DISPLAY_MMIO_BASE(dev_priv) + 0x64220) -#define _DPC_AUX_CH_DATA5 (DISPLAY_MMIO_BASE(dev_priv) + 0x64224) - -#define _DPD_AUX_CH_CTL (DISPLAY_MMIO_BASE(dev_priv) + 0x64310) -#define _DPD_AUX_CH_DATA1 (DISPLAY_MMIO_BASE(dev_priv) + 0x64314) -#define _DPD_AUX_CH_DATA2 (DISPLAY_MMIO_BASE(dev_priv) + 0x64318) -#define _DPD_AUX_CH_DATA3 (DISPLAY_MMIO_BASE(dev_priv) + 0x6431c) -#define _DPD_AUX_CH_DATA4 (DISPLAY_MMIO_BASE(dev_priv) + 0x64320) -#define _DPD_AUX_CH_DATA5 (DISPLAY_MMIO_BASE(dev_priv) + 0x64324) - -#define _DPE_AUX_CH_CTL (DISPLAY_MMIO_BASE(dev_priv) + 0x64410) -#define _DPE_AUX_CH_DATA1 (DISPLAY_MMIO_BASE(dev_priv) + 0x64414) -#define _DPE_AUX_CH_DATA2 (DISPLAY_MMIO_BASE(dev_priv) + 0x64418) -#define _DPE_AUX_CH_DATA3 (DISPLAY_MMIO_BASE(dev_priv) + 0x6441c) -#define _DPE_AUX_CH_DATA4 (DISPLAY_MMIO_BASE(dev_priv) + 0x64420) -#define _DPE_AUX_CH_DATA5 (DISPLAY_MMIO_BASE(dev_priv) + 0x64424) - -#define _DPF_AUX_CH_CTL (DISPLAY_MMIO_BASE(dev_priv) + 0x64510) -#define _DPF_AUX_CH_DATA1 (DISPLAY_MMIO_BASE(dev_priv) + 0x64514) -#define _DPF_AUX_CH_DATA2 (DISPLAY_MMIO_BASE(dev_priv) + 0x64518) -#define _DPF_AUX_CH_DATA3 (DISPLAY_MMIO_BASE(dev_priv) + 0x6451c) -#define _DPF_AUX_CH_DATA4 (DISPLAY_MMIO_BASE(dev_priv) + 0x64520) -#define _DPF_AUX_CH_DATA5 (DISPLAY_MMIO_BASE(dev_priv) + 0x64524) #define DP_AUX_CH_CTL(aux_ch) _MMIO_PORT(aux_ch, _DPA_AUX_CH_CTL, _DPB_AUX_CH_CTL) #define DP_AUX_CH_DATA(aux_ch, i) _MMIO(_PORT(aux_ch, _DPA_AUX_CH_DATA1, _DPB_AUX_CH_DATA1) + (i) * 4) /* 5 registers */ @@ -5652,6 +5785,11 @@ enum { #define PIPECONF_CXSR_DOWNCLOCK (1 << 16) #define PIPECONF_EDP_RR_MODE_SWITCH_VLV (1 << 14) #define PIPECONF_COLOR_RANGE_SELECT (1 << 13) +#define PIPECONF_OUTPUT_COLORSPACE_MASK (3 << 11) /* ilk-ivb */ +#define PIPECONF_OUTPUT_COLORSPACE_RGB (0 << 11) /* ilk-ivb */ +#define PIPECONF_OUTPUT_COLORSPACE_YUV601 (1 << 11) /* ilk-ivb */ +#define PIPECONF_OUTPUT_COLORSPACE_YUV709 (2 << 11) /* ilk-ivb */ +#define PIPECONF_OUTPUT_COLORSPACE_YUV_HSW (1 << 11) /* hsw only */ #define PIPECONF_BPC_MASK (0x7 << 5) #define PIPECONF_8BPC (0 << 5) #define PIPECONF_10BPC (1 << 5) @@ -5739,12 +5877,13 @@ enum { #define _PIPEAGCMAX 0x70010 #define _PIPEBGCMAX 0x71010 +#define PIPEGCMAX_RGB_MASK REG_GENMASK(15, 0) #define PIPEGCMAX(pipe, i) _MMIO_PIPE2(pipe, _PIPEAGCMAX + (i) * 4) #define _PIPE_MISC_A 0x70030 #define _PIPE_MISC_B 0x71030 -#define PIPEMISC_YUV420_ENABLE (1 << 27) -#define PIPEMISC_YUV420_MODE_FULL_BLEND (1 << 26) +#define PIPEMISC_YUV420_ENABLE (1 << 27) /* glk+ */ +#define PIPEMISC_YUV420_MODE_FULL_BLEND (1 << 26) /* glk+ */ #define PIPEMISC_HDR_MODE_PRECISION (1 << 23) /* icl+ */ #define PIPEMISC_OUTPUT_COLORSPACE_YUV (1 << 11) #define PIPEMISC_DITHER_BPC_MASK (7 << 5) @@ -6201,6 +6340,7 @@ enum { #define CHV_CURSOR_C_OFFSET 0x700e0 #define IVB_CURSOR_B_OFFSET 0x71080 #define IVB_CURSOR_C_OFFSET 0x72080 +#define TGL_CURSOR_D_OFFSET 0x73080 /* Display A control */ #define _DSPACNTR 0x70180 @@ -7171,11 +7311,17 @@ enum { /* legacy palette */ #define _LGC_PALETTE_A 0x4a000 #define _LGC_PALETTE_B 0x4a800 +#define LGC_PALETTE_RED_MASK REG_GENMASK(23, 16) +#define LGC_PALETTE_GREEN_MASK REG_GENMASK(15, 8) +#define LGC_PALETTE_BLUE_MASK REG_GENMASK(7, 0) #define LGC_PALETTE(pipe, i) _MMIO(_PIPE(pipe, _LGC_PALETTE_A, _LGC_PALETTE_B) + (i) * 4) /* ilk/snb precision palette */ #define _PREC_PALETTE_A 0x4b000 #define _PREC_PALETTE_B 0x4c000 +#define PREC_PALETTE_RED_MASK REG_GENMASK(29, 20) +#define PREC_PALETTE_GREEN_MASK REG_GENMASK(19, 10) +#define PREC_PALETTE_BLUE_MASK REG_GENMASK(9, 0) #define PREC_PALETTE(pipe, i) _MMIO(_PIPE(pipe, _PREC_PALETTE_A, _PREC_PALETTE_B) + (i) * 4) #define _PREC_PIPEAGCMAX 0x4d000 @@ -7211,6 +7357,12 @@ enum { #define TGL_DMC_DEBUG_DC5_COUNT _MMIO(0x101084) #define TGL_DMC_DEBUG_DC6_COUNT _MMIO(0x101088) +#define DMC_DEBUG3 _MMIO(0x101090) + +/* Display Internal Timeout Register */ +#define RM_TIMEOUT _MMIO(0x42060) +#define MMIO_TIMEOUT_US(us) ((us) << 0) + /* interrupts */ #define DE_MASTER_IRQ_CONTROL (1 << 31) #define DE_SPRITEB_FLIP_DONE (1 << 29) @@ -7322,6 +7474,9 @@ enum { #define GEN8_PIPE_VSYNC (1 << 1) #define GEN8_PIPE_VBLANK (1 << 0) #define GEN9_PIPE_CURSOR_FAULT (1 << 11) +#define GEN11_PIPE_PLANE7_FAULT (1 << 22) +#define GEN11_PIPE_PLANE6_FAULT (1 << 21) +#define GEN11_PIPE_PLANE5_FAULT (1 << 20) #define GEN9_PIPE_PLANE4_FAULT (1 << 10) #define GEN9_PIPE_PLANE3_FAULT (1 << 9) #define GEN9_PIPE_PLANE2_FAULT (1 << 8) @@ -7341,6 +7496,11 @@ enum { GEN9_PIPE_PLANE3_FAULT | \ GEN9_PIPE_PLANE2_FAULT | \ GEN9_PIPE_PLANE1_FAULT) +#define GEN11_DE_PIPE_IRQ_FAULT_ERRORS \ + (GEN9_DE_PIPE_IRQ_FAULT_ERRORS | \ + GEN11_PIPE_PLANE7_FAULT | \ + GEN11_PIPE_PLANE6_FAULT | \ + GEN11_PIPE_PLANE5_FAULT) #define GEN8_DE_PORT_ISR _MMIO(0x44440) #define GEN8_DE_PORT_IMR _MMIO(0x44444) @@ -7360,6 +7520,12 @@ enum { #define GEN8_PORT_DP_A_HOTPLUG (1 << 3) #define BXT_DE_PORT_GMBUS (1 << 1) #define GEN8_AUX_CHANNEL_A (1 << 0) +#define TGL_DE_PORT_AUX_USBC6 (1 << 13) +#define TGL_DE_PORT_AUX_USBC5 (1 << 12) +#define TGL_DE_PORT_AUX_USBC4 (1 << 11) +#define TGL_DE_PORT_AUX_USBC3 (1 << 10) +#define TGL_DE_PORT_AUX_USBC2 (1 << 9) +#define TGL_DE_PORT_AUX_USBC1 (1 << 8) #define TGL_DE_PORT_AUX_DDIC (1 << 2) #define TGL_DE_PORT_AUX_DDIB (1 << 1) #define TGL_DE_PORT_AUX_DDIA (1 << 0) @@ -7548,10 +7714,17 @@ enum { #define BDW_DPRS_MASK_VBLANK_SRD (1 << 0) #define CHICKEN_PIPESL_1(pipe) _MMIO_PIPE(pipe, _CHICKEN_PIPESL_1_A, _CHICKEN_PIPESL_1_B) -#define CHICKEN_TRANS_A _MMIO(0x420c0) -#define CHICKEN_TRANS_B _MMIO(0x420c4) -#define CHICKEN_TRANS_C _MMIO(0x420c8) -#define CHICKEN_TRANS_EDP _MMIO(0x420cc) +#define _CHICKEN_TRANS_A 0x420c0 +#define _CHICKEN_TRANS_B 0x420c4 +#define _CHICKEN_TRANS_C 0x420c8 +#define _CHICKEN_TRANS_EDP 0x420cc +#define _CHICKEN_TRANS_D 0x420d8 +#define CHICKEN_TRANS(trans) _MMIO(_PICK((trans), \ + [TRANSCODER_EDP] = _CHICKEN_TRANS_EDP, \ + [TRANSCODER_A] = _CHICKEN_TRANS_A, \ + [TRANSCODER_B] = _CHICKEN_TRANS_B, \ + [TRANSCODER_C] = _CHICKEN_TRANS_C, \ + [TRANSCODER_D] = _CHICKEN_TRANS_D)) #define VSC_DATA_SEL_SOFTWARE_CONTROL (1 << 25) /* GLK and CNL+ */ #define DDI_TRAINING_OVERRIDE_ENABLE (1 << 19) #define DDI_TRAINING_OVERRIDE_VALUE (1 << 18) @@ -7584,15 +7757,19 @@ enum { #define CNL_DDI_CLOCK_REG_ACCESS_ON (1 << 7) #define SKL_DFSM _MMIO(0x51000) -#define SKL_DFSM_CDCLK_LIMIT_MASK (3 << 23) -#define SKL_DFSM_CDCLK_LIMIT_675 (0 << 23) -#define SKL_DFSM_CDCLK_LIMIT_540 (1 << 23) -#define SKL_DFSM_CDCLK_LIMIT_450 (2 << 23) -#define SKL_DFSM_CDCLK_LIMIT_337_5 (3 << 23) -#define SKL_DFSM_PIPE_A_DISABLE (1 << 30) -#define SKL_DFSM_PIPE_B_DISABLE (1 << 21) -#define SKL_DFSM_PIPE_C_DISABLE (1 << 28) -#define TGL_DFSM_PIPE_D_DISABLE (1 << 22) +#define SKL_DFSM_DISPLAY_PM_DISABLE (1 << 27) +#define SKL_DFSM_DISPLAY_HDCP_DISABLE (1 << 25) +#define SKL_DFSM_CDCLK_LIMIT_MASK (3 << 23) +#define SKL_DFSM_CDCLK_LIMIT_675 (0 << 23) +#define SKL_DFSM_CDCLK_LIMIT_540 (1 << 23) +#define SKL_DFSM_CDCLK_LIMIT_450 (2 << 23) +#define SKL_DFSM_CDCLK_LIMIT_337_5 (3 << 23) +#define ICL_DFSM_DMC_DISABLE (1 << 23) +#define SKL_DFSM_PIPE_A_DISABLE (1 << 30) +#define SKL_DFSM_PIPE_B_DISABLE (1 << 21) +#define SKL_DFSM_PIPE_C_DISABLE (1 << 28) +#define TGL_DFSM_PIPE_D_DISABLE (1 << 22) +#define CNL_DFSM_DISPLAY_DSC_DISABLE (1 << 7) #define SKL_DSSM _MMIO(0x51004) #define CNL_DSSM_CDCLK_PLL_REFCLK_24MHz (1 << 31) @@ -7609,7 +7786,10 @@ enum { #define GEN9_POOLED_EU_LOAD_BALANCING_FIX_DISABLE (1 << 10) #define GEN9_CS_DEBUG_MODE1 _MMIO(0x20ec) +#define FF_DOP_CLOCK_GATE_DISABLE REG_BIT(1) #define GEN9_CTX_PREEMPT_REG _MMIO(0x2248) +#define GEN12_DISABLE_POSH_BUSY_FF_DOP_CG REG_BIT(11) + #define GEN8_CS_CHICKEN1 _MMIO(0x2580) #define GEN9_PREEMPT_3D_OBJECT_LEVEL (1 << 0) #define GEN9_PREEMPT_GPGPU_LEVEL(hi, lo) (((hi) << 2) | ((lo) << 1)) @@ -7634,6 +7814,7 @@ enum { #define GEN11_COMMON_SLICE_CHICKEN3 _MMIO(0x7304) #define GEN11_BLEND_EMB_FIX_DISABLE_IN_RCC (1 << 11) + #define GEN12_DISABLE_CPS_AWARE_COLOR_PIPE (1 << 9) #define HIZ_CHICKEN _MMIO(0x7018) # define CHV_HZ_8X8_MODE_IN_1X (1 << 15) @@ -7818,29 +7999,24 @@ enum { SDE_FDI_RXA_CPT) /* south display engine interrupt: ICP/TGP */ -#define SDE_TC6_HOTPLUG_TGP (1 << 29) -#define SDE_TC5_HOTPLUG_TGP (1 << 28) -#define SDE_TC4_HOTPLUG_ICP (1 << 27) -#define SDE_TC3_HOTPLUG_ICP (1 << 26) -#define SDE_TC2_HOTPLUG_ICP (1 << 25) -#define SDE_TC1_HOTPLUG_ICP (1 << 24) #define SDE_GMBUS_ICP (1 << 23) -#define SDE_DDIC_HOTPLUG_TGP (1 << 18) -#define SDE_DDIB_HOTPLUG_ICP (1 << 17) -#define SDE_DDIA_HOTPLUG_ICP (1 << 16) #define SDE_TC_HOTPLUG_ICP(tc_port) (1 << ((tc_port) + 24)) #define SDE_DDI_HOTPLUG_ICP(port) (1 << ((port) + 16)) -#define SDE_DDI_MASK_ICP (SDE_DDIB_HOTPLUG_ICP | \ - SDE_DDIA_HOTPLUG_ICP) -#define SDE_TC_MASK_ICP (SDE_TC4_HOTPLUG_ICP | \ - SDE_TC3_HOTPLUG_ICP | \ - SDE_TC2_HOTPLUG_ICP | \ - SDE_TC1_HOTPLUG_ICP) -#define SDE_DDI_MASK_TGP (SDE_DDIC_HOTPLUG_TGP | \ - SDE_DDI_MASK_ICP) -#define SDE_TC_MASK_TGP (SDE_TC6_HOTPLUG_TGP | \ - SDE_TC5_HOTPLUG_TGP | \ - SDE_TC_MASK_ICP) +#define SDE_DDI_MASK_ICP (SDE_DDI_HOTPLUG_ICP(PORT_B) | \ + SDE_DDI_HOTPLUG_ICP(PORT_A)) +#define SDE_TC_MASK_ICP (SDE_TC_HOTPLUG_ICP(PORT_TC4) | \ + SDE_TC_HOTPLUG_ICP(PORT_TC3) | \ + SDE_TC_HOTPLUG_ICP(PORT_TC2) | \ + SDE_TC_HOTPLUG_ICP(PORT_TC1)) +#define SDE_DDI_MASK_TGP (SDE_DDI_HOTPLUG_ICP(PORT_C) | \ + SDE_DDI_HOTPLUG_ICP(PORT_B) | \ + SDE_DDI_HOTPLUG_ICP(PORT_A)) +#define SDE_TC_MASK_TGP (SDE_TC_HOTPLUG_ICP(PORT_TC6) | \ + SDE_TC_HOTPLUG_ICP(PORT_TC5) | \ + SDE_TC_HOTPLUG_ICP(PORT_TC4) | \ + SDE_TC_HOTPLUG_ICP(PORT_TC3) | \ + SDE_TC_HOTPLUG_ICP(PORT_TC2) | \ + SDE_TC_HOTPLUG_ICP(PORT_TC1)) #define SDEISR _MMIO(0xc4000) #define SDEIMR _MMIO(0xc4004) @@ -7907,26 +8083,13 @@ enum { * SHOTPLUG_CTL_DDI and SHOTPLUG_CTL_TC. */ -#define SHOTPLUG_CTL_DDI _MMIO(0xc4030) -#define TGP_DDIC_HPD_ENABLE (1 << 11) -#define TGP_DDIC_HPD_STATUS_MASK (3 << 8) -#define TGP_DDIC_HPD_NO_DETECT (0 << 8) -#define TGP_DDIC_HPD_SHORT_DETECT (1 << 8) -#define TGP_DDIC_HPD_LONG_DETECT (2 << 8) -#define TGP_DDIC_HPD_SHORT_LONG_DETECT (3 << 8) -#define ICP_DDIB_HPD_ENABLE (1 << 7) -#define ICP_DDIB_HPD_STATUS_MASK (3 << 4) -#define ICP_DDIB_HPD_NO_DETECT (0 << 4) -#define ICP_DDIB_HPD_SHORT_DETECT (1 << 4) -#define ICP_DDIB_HPD_LONG_DETECT (2 << 4) -#define ICP_DDIB_HPD_SHORT_LONG_DETECT (3 << 4) -#define ICP_DDIA_HPD_ENABLE (1 << 3) -#define ICP_DDIA_HPD_OP_DRIVE_1 (1 << 2) -#define ICP_DDIA_HPD_STATUS_MASK (3 << 0) -#define ICP_DDIA_HPD_NO_DETECT (0 << 0) -#define ICP_DDIA_HPD_SHORT_DETECT (1 << 0) -#define ICP_DDIA_HPD_LONG_DETECT (2 << 0) -#define ICP_DDIA_HPD_SHORT_LONG_DETECT (3 << 0) +#define SHOTPLUG_CTL_DDI _MMIO(0xc4030) +#define SHOTPLUG_CTL_DDI_HPD_ENABLE(port) (0x8 << (4 * (port))) +#define SHOTPLUG_CTL_DDI_HPD_STATUS_MASK(port) (0x3 << (4 * (port))) +#define SHOTPLUG_CTL_DDI_HPD_NO_DETECT(port) (0x0 << (4 * (port))) +#define SHOTPLUG_CTL_DDI_HPD_SHORT_DETECT(port) (0x1 << (4 * (port))) +#define SHOTPLUG_CTL_DDI_HPD_LONG_DETECT(port) (0x2 << (4 * (port))) +#define SHOTPLUG_CTL_DDI_HPD_SHORT_LONG_DETECT(port) (0x3 << (4 * (port))) #define SHOTPLUG_CTL_TC _MMIO(0xc4034) #define ICP_TC_HPD_ENABLE(tc_port) (8 << (tc_port) * 4) @@ -8037,14 +8200,15 @@ enum { #define ICP_TC_HPD_LONG_DETECT(tc_port) (2 << (tc_port) * 4) #define ICP_TC_HPD_SHORT_DETECT(tc_port) (1 << (tc_port) * 4) -#define ICP_DDI_HPD_ENABLE_MASK (ICP_DDIB_HPD_ENABLE | \ - ICP_DDIA_HPD_ENABLE) +#define ICP_DDI_HPD_ENABLE_MASK (SHOTPLUG_CTL_DDI_HPD_ENABLE(PORT_B) | \ + SHOTPLUG_CTL_DDI_HPD_ENABLE(PORT_A)) #define ICP_TC_HPD_ENABLE_MASK (ICP_TC_HPD_ENABLE(PORT_TC4) | \ ICP_TC_HPD_ENABLE(PORT_TC3) | \ ICP_TC_HPD_ENABLE(PORT_TC2) | \ ICP_TC_HPD_ENABLE(PORT_TC1)) -#define TGP_DDI_HPD_ENABLE_MASK (TGP_DDIC_HPD_ENABLE | \ - ICP_DDI_HPD_ENABLE_MASK) +#define TGP_DDI_HPD_ENABLE_MASK (SHOTPLUG_CTL_DDI_HPD_ENABLE(PORT_C) | \ + SHOTPLUG_CTL_DDI_HPD_ENABLE(PORT_B) | \ + SHOTPLUG_CTL_DDI_HPD_ENABLE(PORT_A)) #define TGP_TC_HPD_ENABLE_MASK (ICP_TC_HPD_ENABLE(PORT_TC6) | \ ICP_TC_HPD_ENABLE(PORT_TC5) | \ ICP_TC_HPD_ENABLE_MASK) @@ -8594,6 +8758,10 @@ enum { #define GEN9_PWRGT_MEDIA_STATUS_MASK (1 << 0) #define GEN9_PWRGT_RENDER_STATUS_MASK (1 << 1) +#define POWERGATE_ENABLE _MMIO(0xa210) +#define VDN_HCP_POWERGATE_ENABLE(n) BIT(((n) * 2) + 3) +#define VDN_MFX_POWERGATE_ENABLE(n) BIT(((n) * 2) + 4) + #define GTFIFODBG _MMIO(0x120000) #define GT_FIFO_SBDEDICATE_FREE_ENTRY_CHV (0x1f << 20) #define GT_FIFO_FREE_ENTRIES_CHV (0x7f << 13) @@ -8831,6 +8999,7 @@ enum { #define GEN9_SAGV_DISABLE 0x0 #define GEN9_SAGV_IS_DISABLED 0x1 #define GEN9_SAGV_ENABLE 0x3 +#define GEN12_PCODE_READ_SAGV_BLOCK_TIME_US 0x23 #define GEN6_PCODE_DATA _MMIO(0x138128) #define GEN6_PCODE_FREQ_IA_RATIO_SHIFT 8 #define GEN6_PCODE_FREQ_RING_RATIO_SHIFT 16 @@ -9094,6 +9263,10 @@ enum { #define HSW_AUD_CHICKENBIT _MMIO(0x65f10) #define SKL_AUD_CODEC_WAKE_SIGNAL (1 << 15) +#define AUD_FREQ_CNTRL _MMIO(0x65900) +#define AUD_PIN_BUF_CTL _MMIO(0x48414) +#define AUD_PIN_BUF_ENABLE REG_BIT(31) + /* * HSW - ICL power wells * @@ -9256,12 +9429,20 @@ enum skl_power_gate { /* HDCP Repeater Registers */ #define HDCP_REP_CTL _MMIO(0x66d00) +#define HDCP_TRANSA_REP_PRESENT BIT(31) +#define HDCP_TRANSB_REP_PRESENT BIT(30) +#define HDCP_TRANSC_REP_PRESENT BIT(29) +#define HDCP_TRANSD_REP_PRESENT BIT(28) #define HDCP_DDIB_REP_PRESENT BIT(30) #define HDCP_DDIA_REP_PRESENT BIT(29) #define HDCP_DDIC_REP_PRESENT BIT(28) #define HDCP_DDID_REP_PRESENT BIT(27) #define HDCP_DDIF_REP_PRESENT BIT(26) #define HDCP_DDIE_REP_PRESENT BIT(25) +#define HDCP_TRANSA_SHA1_M0 (1 << 20) +#define HDCP_TRANSB_SHA1_M0 (2 << 20) +#define HDCP_TRANSC_SHA1_M0 (3 << 20) +#define HDCP_TRANSD_SHA1_M0 (4 << 20) #define HDCP_DDIB_SHA1_M0 (1 << 20) #define HDCP_DDIA_SHA1_M0 (2 << 20) #define HDCP_DDIC_SHA1_M0 (3 << 20) @@ -9301,15 +9482,92 @@ enum skl_power_gate { _PORTE_HDCP_AUTHENC, \ _PORTF_HDCP_AUTHENC) + (x)) #define PORT_HDCP_CONF(port) _PORT_HDCP_AUTHENC(port, 0x0) +#define _TRANSA_HDCP_CONF 0x66400 +#define _TRANSB_HDCP_CONF 0x66500 +#define TRANS_HDCP_CONF(trans) _MMIO_TRANS(trans, _TRANSA_HDCP_CONF, \ + _TRANSB_HDCP_CONF) +#define HDCP_CONF(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP_CONF(trans) : \ + PORT_HDCP_CONF(port)) + #define HDCP_CONF_CAPTURE_AN BIT(0) #define HDCP_CONF_AUTH_AND_ENC (BIT(1) | BIT(0)) #define PORT_HDCP_ANINIT(port) _PORT_HDCP_AUTHENC(port, 0x4) +#define _TRANSA_HDCP_ANINIT 0x66404 +#define _TRANSB_HDCP_ANINIT 0x66504 +#define TRANS_HDCP_ANINIT(trans) _MMIO_TRANS(trans, \ + _TRANSA_HDCP_ANINIT, \ + _TRANSB_HDCP_ANINIT) +#define HDCP_ANINIT(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP_ANINIT(trans) : \ + PORT_HDCP_ANINIT(port)) + #define PORT_HDCP_ANLO(port) _PORT_HDCP_AUTHENC(port, 0x8) +#define _TRANSA_HDCP_ANLO 0x66408 +#define _TRANSB_HDCP_ANLO 0x66508 +#define TRANS_HDCP_ANLO(trans) _MMIO_TRANS(trans, _TRANSA_HDCP_ANLO, \ + _TRANSB_HDCP_ANLO) +#define HDCP_ANLO(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP_ANLO(trans) : \ + PORT_HDCP_ANLO(port)) + #define PORT_HDCP_ANHI(port) _PORT_HDCP_AUTHENC(port, 0xC) +#define _TRANSA_HDCP_ANHI 0x6640C +#define _TRANSB_HDCP_ANHI 0x6650C +#define TRANS_HDCP_ANHI(trans) _MMIO_TRANS(trans, _TRANSA_HDCP_ANHI, \ + _TRANSB_HDCP_ANHI) +#define HDCP_ANHI(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP_ANHI(trans) : \ + PORT_HDCP_ANHI(port)) + #define PORT_HDCP_BKSVLO(port) _PORT_HDCP_AUTHENC(port, 0x10) +#define _TRANSA_HDCP_BKSVLO 0x66410 +#define _TRANSB_HDCP_BKSVLO 0x66510 +#define TRANS_HDCP_BKSVLO(trans) _MMIO_TRANS(trans, \ + _TRANSA_HDCP_BKSVLO, \ + _TRANSB_HDCP_BKSVLO) +#define HDCP_BKSVLO(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP_BKSVLO(trans) : \ + PORT_HDCP_BKSVLO(port)) + #define PORT_HDCP_BKSVHI(port) _PORT_HDCP_AUTHENC(port, 0x14) +#define _TRANSA_HDCP_BKSVHI 0x66414 +#define _TRANSB_HDCP_BKSVHI 0x66514 +#define TRANS_HDCP_BKSVHI(trans) _MMIO_TRANS(trans, \ + _TRANSA_HDCP_BKSVHI, \ + _TRANSB_HDCP_BKSVHI) +#define HDCP_BKSVHI(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP_BKSVHI(trans) : \ + PORT_HDCP_BKSVHI(port)) + #define PORT_HDCP_RPRIME(port) _PORT_HDCP_AUTHENC(port, 0x18) +#define _TRANSA_HDCP_RPRIME 0x66418 +#define _TRANSB_HDCP_RPRIME 0x66518 +#define TRANS_HDCP_RPRIME(trans) _MMIO_TRANS(trans, \ + _TRANSA_HDCP_RPRIME, \ + _TRANSB_HDCP_RPRIME) +#define HDCP_RPRIME(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP_RPRIME(trans) : \ + PORT_HDCP_RPRIME(port)) + #define PORT_HDCP_STATUS(port) _PORT_HDCP_AUTHENC(port, 0x1C) +#define _TRANSA_HDCP_STATUS 0x6641C +#define _TRANSB_HDCP_STATUS 0x6651C +#define TRANS_HDCP_STATUS(trans) _MMIO_TRANS(trans, \ + _TRANSA_HDCP_STATUS, \ + _TRANSB_HDCP_STATUS) +#define HDCP_STATUS(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP_STATUS(trans) : \ + PORT_HDCP_STATUS(port)) + #define HDCP_STATUS_STREAM_A_ENC BIT(31) #define HDCP_STATUS_STREAM_B_ENC BIT(30) #define HDCP_STATUS_STREAM_C_ENC BIT(29) @@ -9336,23 +9594,44 @@ enum skl_power_gate { _PORTD_HDCP2_BASE, \ _PORTE_HDCP2_BASE, \ _PORTF_HDCP2_BASE) + (x)) - -#define HDCP2_AUTH_DDI(port) _PORT_HDCP2_BASE(port, 0x98) +#define PORT_HDCP2_AUTH(port) _PORT_HDCP2_BASE(port, 0x98) +#define _TRANSA_HDCP2_AUTH 0x66498 +#define _TRANSB_HDCP2_AUTH 0x66598 +#define TRANS_HDCP2_AUTH(trans) _MMIO_TRANS(trans, _TRANSA_HDCP2_AUTH, \ + _TRANSB_HDCP2_AUTH) #define AUTH_LINK_AUTHENTICATED BIT(31) #define AUTH_LINK_TYPE BIT(30) #define AUTH_FORCE_CLR_INPUTCTR BIT(19) #define AUTH_CLR_KEYS BIT(18) - -#define HDCP2_CTL_DDI(port) _PORT_HDCP2_BASE(port, 0xB0) +#define HDCP2_AUTH(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP2_AUTH(trans) : \ + PORT_HDCP2_AUTH(port)) + +#define PORT_HDCP2_CTL(port) _PORT_HDCP2_BASE(port, 0xB0) +#define _TRANSA_HDCP2_CTL 0x664B0 +#define _TRANSB_HDCP2_CTL 0x665B0 +#define TRANS_HDCP2_CTL(trans) _MMIO_TRANS(trans, _TRANSA_HDCP2_CTL, \ + _TRANSB_HDCP2_CTL) #define CTL_LINK_ENCRYPTION_REQ BIT(31) - -#define HDCP2_STATUS_DDI(port) _PORT_HDCP2_BASE(port, 0xB4) -#define STREAM_ENCRYPTION_STATUS_A BIT(31) -#define STREAM_ENCRYPTION_STATUS_B BIT(30) -#define STREAM_ENCRYPTION_STATUS_C BIT(29) +#define HDCP2_CTL(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP2_CTL(trans) : \ + PORT_HDCP2_CTL(port)) + +#define PORT_HDCP2_STATUS(port) _PORT_HDCP2_BASE(port, 0xB4) +#define _TRANSA_HDCP2_STATUS 0x664B4 +#define _TRANSB_HDCP2_STATUS 0x665B4 +#define TRANS_HDCP2_STATUS(trans) _MMIO_TRANS(trans, \ + _TRANSA_HDCP2_STATUS, \ + _TRANSB_HDCP2_STATUS) #define LINK_TYPE_STATUS BIT(22) #define LINK_AUTH_STATUS BIT(21) #define LINK_ENCRYPTION_STATUS BIT(20) +#define HDCP2_STATUS(dev_priv, trans, port) \ + (INTEL_GEN(dev_priv) >= 12 ? \ + TRANS_HDCP2_STATUS(trans) : \ + PORT_HDCP2_STATUS(port)) /* Per-pipe DDI Function Control */ #define _TRANS_DDI_FUNC_CTL_A 0x60400 @@ -9392,6 +9671,9 @@ enum skl_power_gate { #define TRANS_DDI_EDP_INPUT_A_ONOFF (4 << 12) #define TRANS_DDI_EDP_INPUT_B_ONOFF (5 << 12) #define TRANS_DDI_EDP_INPUT_C_ONOFF (6 << 12) +#define TRANS_DDI_MST_TRANSPORT_SELECT_MASK REG_GENMASK(11, 10) +#define TRANS_DDI_MST_TRANSPORT_SELECT(trans) \ + REG_FIELD_PREP(TRANS_DDI_MST_TRANSPORT_SELECT_MASK, trans) #define TRANS_DDI_HDCP_SIGNALLING (1 << 9) #define TRANS_DDI_DP_VC_PAYLOAD_ALLOC (1 << 8) #define TRANS_DDI_HDMI_SCRAMBLER_CTS_ENABLE (1 << 7) @@ -9419,7 +9701,9 @@ enum skl_power_gate { /* DisplayPort Transport Control */ #define _DP_TP_CTL_A 0x64040 #define _DP_TP_CTL_B 0x64140 +#define _TGL_DP_TP_CTL_A 0x60540 #define DP_TP_CTL(port) _MMIO_PORT(port, _DP_TP_CTL_A, _DP_TP_CTL_B) +#define TGL_DP_TP_CTL(tran) _MMIO_TRANS2((tran), _TGL_DP_TP_CTL_A) #define DP_TP_CTL_ENABLE (1 << 31) #define DP_TP_CTL_FEC_ENABLE (1 << 30) #define DP_TP_CTL_MODE_SST (0 << 27) @@ -9439,7 +9723,9 @@ enum skl_power_gate { /* DisplayPort Transport Status */ #define _DP_TP_STATUS_A 0x64044 #define _DP_TP_STATUS_B 0x64144 +#define _TGL_DP_TP_STATUS_A 0x60544 #define DP_TP_STATUS(port) _MMIO_PORT(port, _DP_TP_STATUS_A, _DP_TP_STATUS_B) +#define TGL_DP_TP_STATUS(tran) _MMIO_TRANS2((tran), _TGL_DP_TP_STATUS_A) #define DP_TP_STATUS_FEC_ENABLE_LIVE (1 << 28) #define DP_TP_STATUS_IDLE_DONE (1 << 25) #define DP_TP_STATUS_ACT_SENT (1 << 24) @@ -9594,17 +9880,7 @@ enum skl_power_gate { #define _TRANSC_MSA_MISC 0x62410 #define _TRANS_EDP_MSA_MISC 0x6f410 #define TRANS_MSA_MISC(tran) _MMIO_TRANS2(tran, _TRANSA_MSA_MISC) - -#define TRANS_MSA_SYNC_CLK (1 << 0) -#define TRANS_MSA_SAMPLING_444 (2 << 1) -#define TRANS_MSA_CLRSP_YCBCR (2 << 3) -#define TRANS_MSA_6_BPC (0 << 5) -#define TRANS_MSA_8_BPC (1 << 5) -#define TRANS_MSA_10_BPC (2 << 5) -#define TRANS_MSA_12_BPC (3 << 5) -#define TRANS_MSA_16_BPC (4 << 5) -#define TRANS_MSA_CEA_RANGE (1 << 3) -#define TRANS_MSA_USE_VSC_SDP (1 << 14) +/* See DP_MSA_MISC_* for the bit definitions */ /* LCPLL Control */ #define LCPLL_CTL _MMIO(0x130040) @@ -9645,7 +9921,10 @@ enum skl_power_gate { #define BXT_CDCLK_CD2X_PIPE(pipe) ((pipe) << 20) #define CDCLK_DIVMUX_CD_OVERRIDE (1 << 19) #define BXT_CDCLK_CD2X_PIPE_NONE BXT_CDCLK_CD2X_PIPE(3) +#define ICL_CDCLK_CD2X_PIPE(pipe) (_PICK(pipe, 0, 2, 6) << 19) #define ICL_CDCLK_CD2X_PIPE_NONE (7 << 19) +#define TGL_CDCLK_CD2X_PIPE(pipe) BXT_CDCLK_CD2X_PIPE(pipe) +#define TGL_CDCLK_CD2X_PIPE_NONE ICL_CDCLK_CD2X_PIPE_NONE #define BXT_CDCLK_SSA_PRECHARGE_ENABLE (1 << 16) #define CDCLK_FREQ_DECIMAL_MASK (0x7ff) @@ -9966,6 +10245,166 @@ enum skl_power_gate { _TGL_DPLL1_CFGCR1, \ _TGL_TBTPLL_CFGCR1) +#define _DKL_PHY1_BASE 0x168000 +#define _DKL_PHY2_BASE 0x169000 +#define _DKL_PHY3_BASE 0x16A000 +#define _DKL_PHY4_BASE 0x16B000 +#define _DKL_PHY5_BASE 0x16C000 +#define _DKL_PHY6_BASE 0x16D000 + +/* DEKEL PHY MMIO Address = Phy base + (internal address & ~index_mask) */ +#define _DKL_PLL_DIV0 0x200 +#define DKL_PLL_DIV0_INTEG_COEFF(x) ((x) << 16) +#define DKL_PLL_DIV0_INTEG_COEFF_MASK (0x1F << 16) +#define DKL_PLL_DIV0_PROP_COEFF(x) ((x) << 12) +#define DKL_PLL_DIV0_PROP_COEFF_MASK (0xF << 12) +#define DKL_PLL_DIV0_FBPREDIV_SHIFT (8) +#define DKL_PLL_DIV0_FBPREDIV(x) ((x) << DKL_PLL_DIV0_FBPREDIV_SHIFT) +#define DKL_PLL_DIV0_FBPREDIV_MASK (0xF << DKL_PLL_DIV0_FBPREDIV_SHIFT) +#define DKL_PLL_DIV0_FBDIV_INT(x) ((x) << 0) +#define DKL_PLL_DIV0_FBDIV_INT_MASK (0xFF << 0) +#define DKL_PLL_DIV0(tc_port) _MMIO(_PORT(tc_port, _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_PLL_DIV0) + +#define _DKL_PLL_DIV1 0x204 +#define DKL_PLL_DIV1_IREF_TRIM(x) ((x) << 16) +#define DKL_PLL_DIV1_IREF_TRIM_MASK (0x1F << 16) +#define DKL_PLL_DIV1_TDC_TARGET_CNT(x) ((x) << 0) +#define DKL_PLL_DIV1_TDC_TARGET_CNT_MASK (0xFF << 0) +#define DKL_PLL_DIV1(tc_port) _MMIO(_PORT(tc_port, _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_PLL_DIV1) + +#define _DKL_PLL_SSC 0x210 +#define DKL_PLL_SSC_IREF_NDIV_RATIO(x) ((x) << 29) +#define DKL_PLL_SSC_IREF_NDIV_RATIO_MASK (0x7 << 29) +#define DKL_PLL_SSC_STEP_LEN(x) ((x) << 16) +#define DKL_PLL_SSC_STEP_LEN_MASK (0xFF << 16) +#define DKL_PLL_SSC_STEP_NUM(x) ((x) << 11) +#define DKL_PLL_SSC_STEP_NUM_MASK (0x7 << 11) +#define DKL_PLL_SSC_EN (1 << 9) +#define DKL_PLL_SSC(tc_port) _MMIO(_PORT(tc_port, _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_PLL_SSC) + +#define _DKL_PLL_BIAS 0x214 +#define DKL_PLL_BIAS_FRAC_EN_H (1 << 30) +#define DKL_PLL_BIAS_FBDIV_SHIFT (8) +#define DKL_PLL_BIAS_FBDIV_FRAC(x) ((x) << DKL_PLL_BIAS_FBDIV_SHIFT) +#define DKL_PLL_BIAS_FBDIV_FRAC_MASK (0x3FFFFF << DKL_PLL_BIAS_FBDIV_SHIFT) +#define DKL_PLL_BIAS(tc_port) _MMIO(_PORT(tc_port, _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_PLL_BIAS) + +#define _DKL_PLL_TDC_COLDST_BIAS 0x218 +#define DKL_PLL_TDC_SSC_STEP_SIZE(x) ((x) << 8) +#define DKL_PLL_TDC_SSC_STEP_SIZE_MASK (0xFF << 8) +#define DKL_PLL_TDC_FEED_FWD_GAIN(x) ((x) << 0) +#define DKL_PLL_TDC_FEED_FWD_GAIN_MASK (0xFF << 0) +#define DKL_PLL_TDC_COLDST_BIAS(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_PLL_TDC_COLDST_BIAS) + +#define _DKL_REFCLKIN_CTL 0x12C +/* Bits are the same as MG_REFCLKIN_CTL */ +#define DKL_REFCLKIN_CTL(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_REFCLKIN_CTL) + +#define _DKL_CLKTOP2_HSCLKCTL 0xD4 +/* Bits are the same as MG_CLKTOP2_HSCLKCTL */ +#define DKL_CLKTOP2_HSCLKCTL(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_CLKTOP2_HSCLKCTL) + +#define _DKL_CLKTOP2_CORECLKCTL1 0xD8 +/* Bits are the same as MG_CLKTOP2_CORECLKCTL1 */ +#define DKL_CLKTOP2_CORECLKCTL1(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_CLKTOP2_CORECLKCTL1) + +#define _DKL_TX_DPCNTL0 0x2C0 +#define DKL_TX_PRESHOOT_COEFF(x) ((x) << 13) +#define DKL_TX_PRESHOOT_COEFF_MASK (0x1f << 13) +#define DKL_TX_DE_EMPHASIS_COEFF(x) ((x) << 8) +#define DKL_TX_DE_EMPAHSIS_COEFF_MASK (0x1f << 8) +#define DKL_TX_VSWING_CONTROL(x) ((x) << 0) +#define DKL_TX_VSWING_CONTROL_MASK (0x7 << 0) +#define DKL_TX_DPCNTL0(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_TX_DPCNTL0) + +#define _DKL_TX_DPCNTL1 0x2C4 +/* Bits are the same as DKL_TX_DPCNTRL0 */ +#define DKL_TX_DPCNTL1(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_TX_DPCNTL1) + +#define _DKL_TX_DPCNTL2 0x2C8 +#define DKL_TX_DP20BITMODE (1 << 2) +#define DKL_TX_DPCNTL2(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_TX_DPCNTL2) + +#define _DKL_TX_FW_CALIB 0x2F8 +#define DKL_TX_CFG_DISABLE_WAIT_INIT (1 << 7) +#define DKL_TX_FW_CALIB(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_TX_FW_CALIB) + +#define _DKL_TX_PMD_LANE_SUS 0xD00 +#define DKL_TX_PMD_LANE_SUS(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_TX_PMD_LANE_SUS) + +#define _DKL_TX_DW17 0xDC4 +#define DKL_TX_DW17(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_TX_DW17) + +#define _DKL_TX_DW18 0xDC8 +#define DKL_TX_DW18(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_TX_DW18) + +#define _DKL_DP_MODE 0xA0 +#define DKL_DP_MODE(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_DP_MODE) + +#define _DKL_CMN_UC_DW27 0x36C +#define DKL_CMN_UC_DW27_UC_HEALTH (0x1 << 15) +#define DKL_CMN_UC_DW_27(tc_port) _MMIO(_PORT(tc_port, \ + _DKL_PHY1_BASE, \ + _DKL_PHY2_BASE) + \ + _DKL_CMN_UC_DW27) + +/* + * Each Dekel PHY is addressed through a 4KB aperture. Each PHY has more than + * 4KB of register space, so a separate index is programmed in HIP_INDEX_REG0 + * or HIP_INDEX_REG1, based on the port number, to set the upper 2 address + * bits that point the 4KB window into the full PHY register space. + */ +#define _HIP_INDEX_REG0 0x1010A0 +#define _HIP_INDEX_REG1 0x1010A4 +#define HIP_INDEX_REG(tc_port) _MMIO((tc_port) < 4 ? _HIP_INDEX_REG0 \ + : _HIP_INDEX_REG1) +#define _HIP_INDEX_SHIFT(tc_port) (8 * ((tc_port) % 4)) +#define HIP_INDEX_VAL(tc_port, val) ((val) << _HIP_INDEX_SHIFT(tc_port)) + /* BXT display engine PLL */ #define BXT_DE_PLL_CTL _MMIO(0x6d000) #define BXT_DE_PLL_RATIO(x) (x) /* {60,65,100} * 19.2MHz */ @@ -9980,6 +10419,8 @@ enum skl_power_gate { /* GEN9 DC */ #define DC_STATE_EN _MMIO(0x45504) #define DC_STATE_DISABLE 0 +#define DC_STATE_EN_DC3CO REG_BIT(30) +#define DC_STATE_DC3CO_STATUS REG_BIT(29) #define DC_STATE_EN_UPTO_DC5 (1 << 0) #define DC_STATE_EN_DC9 (1 << 3) #define DC_STATE_EN_UPTO_DC6 (2 << 0) @@ -10108,11 +10549,11 @@ enum skl_power_gate { #define _PIPE_A_CSC_COEFF_BV 0x49024 #define _PIPE_A_CSC_MODE 0x49028 -#define ICL_CSC_ENABLE (1 << 31) -#define ICL_OUTPUT_CSC_ENABLE (1 << 30) -#define CSC_BLACK_SCREEN_OFFSET (1 << 2) -#define CSC_POSITION_BEFORE_GAMMA (1 << 1) -#define CSC_MODE_YUV_TO_RGB (1 << 0) +#define ICL_CSC_ENABLE (1 << 31) /* icl+ */ +#define ICL_OUTPUT_CSC_ENABLE (1 << 30) /* icl+ */ +#define CSC_BLACK_SCREEN_OFFSET (1 << 2) /* ilk/snb */ +#define CSC_POSITION_BEFORE_GAMMA (1 << 1) /* pre-glk */ +#define CSC_MODE_YUV_TO_RGB (1 << 0) /* ilk/snb */ #define _PIPE_A_CSC_PREOFF_HI 0x49030 #define _PIPE_A_CSC_PREOFF_ME 0x49034 @@ -10228,6 +10669,9 @@ enum skl_power_gate { #define _PAL_PREC_GC_MAX_A 0x4A410 #define _PAL_PREC_GC_MAX_B 0x4AC10 #define _PAL_PREC_GC_MAX_C 0x4B410 +#define PREC_PAL_DATA_RED_MASK REG_GENMASK(29, 20) +#define PREC_PAL_DATA_GREEN_MASK REG_GENMASK(19, 10) +#define PREC_PAL_DATA_BLUE_MASK REG_GENMASK(9, 0) #define _PAL_PREC_EXT_GC_MAX_A 0x4A420 #define _PAL_PREC_EXT_GC_MAX_B 0x4AC20 #define _PAL_PREC_EXT_GC_MAX_C 0x4B420 @@ -10280,6 +10724,9 @@ enum skl_power_gate { #define CGM_PIPE_MODE_GAMMA (1 << 2) #define CGM_PIPE_MODE_CSC (1 << 1) #define CGM_PIPE_MODE_DEGAMMA (1 << 0) +#define CGM_PIPE_GAMMA_RED_MASK REG_GENMASK(9, 0) +#define CGM_PIPE_GAMMA_GREEN_MASK REG_GENMASK(25, 16) +#define CGM_PIPE_GAMMA_BLUE_MASK REG_GENMASK(9, 0) #define _CGM_PIPE_B_CSC_COEFF01 (VLV_DISPLAY_BASE + 0x69900) #define _CGM_PIPE_B_CSC_COEFF23 (VLV_DISPLAY_BASE + 0x69904) @@ -11527,16 +11974,31 @@ enum skl_power_gate { #define PORT_TX_DFLEXDPSP(fia) _MMIO_FIA((fia), 0x008A0) #define MODULAR_FIA_MASK (1 << 4) -#define TC_LIVE_STATE_TBT(tc_port) (1 << ((tc_port) * 8 + 6)) -#define TC_LIVE_STATE_TC(tc_port) (1 << ((tc_port) * 8 + 5)) -#define DP_LANE_ASSIGNMENT_SHIFT(tc_port) ((tc_port) * 8) -#define DP_LANE_ASSIGNMENT_MASK(tc_port) (0xf << ((tc_port) * 8)) -#define DP_LANE_ASSIGNMENT(tc_port, x) ((x) << ((tc_port) * 8)) +#define TC_LIVE_STATE_TBT(idx) (1 << ((idx) * 8 + 6)) +#define TC_LIVE_STATE_TC(idx) (1 << ((idx) * 8 + 5)) +#define DP_LANE_ASSIGNMENT_SHIFT(idx) ((idx) * 8) +#define DP_LANE_ASSIGNMENT_MASK(idx) (0xf << ((idx) * 8)) +#define DP_LANE_ASSIGNMENT(idx, x) ((x) << ((idx) * 8)) #define PORT_TX_DFLEXDPPMS(fia) _MMIO_FIA((fia), 0x00890) -#define DP_PHY_MODE_STATUS_COMPLETED(tc_port) (1 << (tc_port)) +#define DP_PHY_MODE_STATUS_COMPLETED(idx) (1 << (idx)) #define PORT_TX_DFLEXDPCSSS(fia) _MMIO_FIA((fia), 0x00894) -#define DP_PHY_MODE_STATUS_NOT_SAFE(tc_port) (1 << (tc_port)) +#define DP_PHY_MODE_STATUS_NOT_SAFE(idx) (1 << (idx)) + +#define PORT_TX_DFLEXPA1(fia) _MMIO_FIA((fia), 0x00880) +#define DP_PIN_ASSIGNMENT_SHIFT(idx) ((idx) * 4) +#define DP_PIN_ASSIGNMENT_MASK(idx) (0xf << ((idx) * 4)) +#define DP_PIN_ASSIGNMENT(idx, x) ((x) << ((idx) * 4)) + +/* This register controls the Display State Buffer (DSB) engines. */ +#define _DSBSL_INSTANCE_BASE 0x70B00 +#define DSBSL_INSTANCE(pipe, id) (_DSBSL_INSTANCE_BASE + \ + (pipe) * 0x1000 + (id) * 100) +#define DSB_HEAD(pipe, id) _MMIO(DSBSL_INSTANCE(pipe, id) + 0x0) +#define DSB_TAIL(pipe, id) _MMIO(DSBSL_INSTANCE(pipe, id) + 0x4) +#define DSB_CTRL(pipe, id) _MMIO(DSBSL_INSTANCE(pipe, id) + 0x8) +#define DSB_ENABLE (1 << 31) +#define DSB_STATUS (1 << 0) #endif /* _I915_REG_H_ */ diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c index a53777dd371c..bbd71af00a91 100644 --- a/drivers/gpu/drm/i915/i915_request.c +++ b/drivers/gpu/drm/i915/i915_request.c @@ -31,6 +31,8 @@ #include "gem/i915_gem_context.h" #include "gt/intel_context.h" +#include "gt/intel_ring.h" +#include "gt/intel_rps.h" #include "i915_active.h" #include "i915_drv.h" @@ -169,16 +171,17 @@ remove_from_client(struct i915_request *request) { struct drm_i915_file_private *file_priv; - file_priv = READ_ONCE(request->file_priv); - if (!file_priv) + if (!READ_ONCE(request->file_priv)) return; - spin_lock(&file_priv->mm.lock); - if (request->file_priv) { + rcu_read_lock(); + file_priv = xchg(&request->file_priv, NULL); + if (file_priv) { + spin_lock(&file_priv->mm.lock); list_del(&request->client_link); - request->file_priv = NULL; + spin_unlock(&file_priv->mm.lock); } - spin_unlock(&file_priv->mm.lock); + rcu_read_unlock(); } static void free_capture_list(struct i915_request *request) @@ -194,11 +197,29 @@ static void free_capture_list(struct i915_request *request) } } -static bool i915_request_retire(struct i915_request *rq) +static void remove_from_engine(struct i915_request *rq) { - struct i915_active_request *active, *next; + struct intel_engine_cs *engine, *locked; - lockdep_assert_held(&rq->timeline->mutex); + /* + * Virtual engines complicate acquiring the engine timeline lock, + * as their rq->engine pointer is not stable until under that + * engine lock. The simple ploy we use is to take the lock then + * check that the rq still belongs to the newly locked engine. + */ + locked = READ_ONCE(rq->engine); + spin_lock_irq(&locked->active.lock); + while (unlikely(locked != (engine = READ_ONCE(rq->engine)))) { + spin_unlock(&locked->active.lock); + spin_lock(&engine->active.lock); + locked = engine; + } + list_del(&rq->sched.link); + spin_unlock_irq(&locked->active.lock); +} + +bool i915_request_retire(struct i915_request *rq) +{ if (!i915_request_completed(rq)) return false; @@ -219,68 +240,34 @@ static bool i915_request_retire(struct i915_request *rq) * Note this requires that we are always called in request * completion order. */ - GEM_BUG_ON(!list_is_first(&rq->link, &rq->timeline->requests)); + GEM_BUG_ON(!list_is_first(&rq->link, + &i915_request_timeline(rq)->requests)); rq->ring->head = rq->postfix; /* - * Walk through the active list, calling retire on each. This allows - * objects to track their GPU activity and mark themselves as idle - * when their *last* active request is completed (updating state - * tracking lists for eviction, active references for GEM, etc). - * - * As the ->retire() may free the node, we decouple it first and - * pass along the auxiliary information (to avoid dereferencing - * the node after the callback). - */ - list_for_each_entry_safe(active, next, &rq->active_list, link) { - /* - * In microbenchmarks or focusing upon time inside the kernel, - * we may spend an inordinate amount of time simply handling - * the retirement of requests and processing their callbacks. - * Of which, this loop itself is particularly hot due to the - * cache misses when jumping around the list of - * i915_active_request. So we try to keep this loop as - * streamlined as possible and also prefetch the next - * i915_active_request to try and hide the likely cache miss. - */ - prefetchw(next); - - INIT_LIST_HEAD(&active->link); - RCU_INIT_POINTER(active->request, NULL); - - active->retire(active, rq); - } - - local_irq_disable(); - - /* * We only loosely track inflight requests across preemption, * and so we may find ourselves attempting to retire a _completed_ * request that we have removed from the HW and put back on a run * queue. */ - spin_lock(&rq->engine->active.lock); - list_del(&rq->sched.link); - spin_unlock(&rq->engine->active.lock); + remove_from_engine(rq); - spin_lock(&rq->lock); + spin_lock_irq(&rq->lock); i915_request_mark_complete(rq); if (!i915_request_signaled(rq)) dma_fence_signal_locked(&rq->fence); if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &rq->fence.flags)) i915_request_cancel_breadcrumb(rq); if (i915_request_has_waitboost(rq)) { - GEM_BUG_ON(!atomic_read(&rq->i915->gt_pm.rps.num_waiters)); - atomic_dec(&rq->i915->gt_pm.rps.num_waiters); + GEM_BUG_ON(!atomic_read(&rq->engine->gt->rps.num_waiters)); + atomic_dec(&rq->engine->gt->rps.num_waiters); } if (!test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags)) { set_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags); __notify_execute_cb(rq); } GEM_BUG_ON(!list_empty(&rq->execute_cb)); - spin_unlock(&rq->lock); - - local_irq_enable(); + spin_unlock_irq(&rq->lock); remove_from_client(rq); list_del(&rq->link); @@ -297,7 +284,7 @@ static bool i915_request_retire(struct i915_request *rq) void i915_request_retire_upto(struct i915_request *rq) { - struct intel_timeline * const tl = rq->timeline; + struct intel_timeline * const tl = i915_request_timeline(rq); struct i915_request *tmp; GEM_TRACE("%s fence %llx:%lld, current %d\n", @@ -305,7 +292,6 @@ void i915_request_retire_upto(struct i915_request *rq) rq->fence.context, rq->fence.seqno, hwsp_seqno(rq)); - lockdep_assert_held(&tl->mutex); GEM_BUG_ON(!i915_request_completed(rq)); do { @@ -358,9 +344,10 @@ __i915_request_await_execution(struct i915_request *rq, return 0; } -void __i915_request_submit(struct i915_request *request) +bool __i915_request_submit(struct i915_request *request) { struct intel_engine_cs *engine = request->engine; + bool result = false; GEM_TRACE("%s fence %llx:%lld, current %d\n", engine->name, @@ -370,6 +357,25 @@ void __i915_request_submit(struct i915_request *request) GEM_BUG_ON(!irqs_disabled()); lockdep_assert_held(&engine->active.lock); + /* + * With the advent of preempt-to-busy, we frequently encounter + * requests that we have unsubmitted from HW, but left running + * until the next ack and so have completed in the meantime. On + * resubmission of that completed request, we can skip + * updating the payload, and execlists can even skip submitting + * the request. + * + * We must remove the request from the caller's priority queue, + * and the caller must only call us when the request is in their + * priority queue, under the active.lock. This ensures that the + * request has *not* yet been retired and we can safely move + * the request into the engine->active.list where it will be + * dropped upon retiring. (Otherwise if resubmit a *retired* + * request, this would be a horrible use-after-free.) + */ + if (i915_request_completed(request)) + goto xfer; + if (i915_gem_context_is_banned(request->gem_context)) i915_request_skip(request, -EIO); @@ -393,13 +399,18 @@ void __i915_request_submit(struct i915_request *request) i915_sw_fence_signaled(&request->semaphore)) engine->saturated |= request->sched.semaphores; - /* We may be recursing from the signal callback of another i915 fence */ - spin_lock_nested(&request->lock, SINGLE_DEPTH_NESTING); + engine->emit_fini_breadcrumb(request, + request->ring->vaddr + request->postfix); - list_move_tail(&request->sched.link, &engine->active.requests); + trace_i915_request_execute(request); + engine->serial++; + result = true; - GEM_BUG_ON(test_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags)); - set_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags); +xfer: /* We may be recursing from the signal callback of another i915 fence */ + spin_lock_nested(&request->lock, SINGLE_DEPTH_NESTING); + + if (!test_and_set_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags)) + list_move_tail(&request->sched.link, &engine->active.requests); if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags) && !test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &request->fence.flags) && @@ -410,12 +421,7 @@ void __i915_request_submit(struct i915_request *request) spin_unlock(&request->lock); - engine->emit_fini_breadcrumb(request, - request->ring->vaddr + request->postfix); - - engine->serial++; - - trace_i915_request_execute(request); + return result; } void i915_request_submit(struct i915_request *request) @@ -641,9 +647,12 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp) rq->gem_context = ce->gem_context; rq->engine = ce->engine; rq->ring = ce->ring; - rq->timeline = tl; + rq->execution_mask = ce->engine->mask; + + rcu_assign_pointer(rq->timeline, tl); rq->hwsp_seqno = tl->hwsp_seqno; rq->hwsp_cacheline = tl->hwsp_cacheline; + rq->rcustate = get_state_synchronize_rcu(); /* acts as smp_mb() */ spin_lock_init(&rq->lock); @@ -661,9 +670,7 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp) rq->batch = NULL; rq->capture_list = NULL; rq->flags = 0; - rq->execution_mask = ALL_ENGINES; - INIT_LIST_HEAD(&rq->active_list); INIT_LIST_HEAD(&rq->execute_cb); /* @@ -702,7 +709,6 @@ err_unwind: ce->ring->emit = rq->head; /* Make sure we didn't add ourselves to external state before freeing */ - GEM_BUG_ON(!list_empty(&rq->active_list)); GEM_BUG_ON(!list_empty(&rq->sched.signalers_list)); GEM_BUG_ON(!list_empty(&rq->sched.waiters_list)); @@ -747,16 +753,43 @@ err_unlock: static int i915_request_await_start(struct i915_request *rq, struct i915_request *signal) { - if (list_is_first(&signal->link, &signal->timeline->requests)) - return 0; + struct intel_timeline *tl; + struct dma_fence *fence; + int err; + + GEM_BUG_ON(i915_request_timeline(rq) == + rcu_access_pointer(signal->timeline)); - signal = list_prev_entry(signal, link); - if (intel_timeline_sync_is_later(rq->timeline, &signal->fence)) + rcu_read_lock(); + tl = rcu_dereference(signal->timeline); + if (i915_request_started(signal) || !kref_get_unless_zero(&tl->kref)) + tl = NULL; + rcu_read_unlock(); + if (!tl) /* already started or maybe even completed */ return 0; - return i915_sw_fence_await_dma_fence(&rq->submit, - &signal->fence, 0, - I915_FENCE_GFP); + fence = ERR_PTR(-EBUSY); + if (mutex_trylock(&tl->mutex)) { + fence = NULL; + if (!i915_request_started(signal) && + !list_is_first(&signal->link, &tl->requests)) { + signal = list_prev_entry(signal, link); + fence = dma_fence_get(&signal->fence); + } + mutex_unlock(&tl->mutex); + } + intel_timeline_put(tl); + if (IS_ERR_OR_NULL(fence)) + return PTR_ERR_OR_ZERO(fence); + + err = 0; + if (intel_timeline_sync_is_later(i915_request_timeline(rq), fence)) + err = i915_sw_fence_await_dma_fence(&rq->submit, + fence, 0, + I915_FENCE_GFP); + dma_fence_put(fence); + + return err; } static intel_engine_mask_t @@ -782,34 +815,33 @@ emit_semaphore_wait(struct i915_request *to, struct i915_request *from, gfp_t gfp) { + const int has_token = INTEL_GEN(to->i915) >= 12; u32 hwsp_offset; + int len; u32 *cs; - int err; - GEM_BUG_ON(!from->timeline->has_initial_breadcrumb); GEM_BUG_ON(INTEL_GEN(to->i915) < 8); /* Just emit the first semaphore we see as request space is limited. */ if (already_busywaiting(to) & from->engine->mask) - return i915_sw_fence_await_dma_fence(&to->submit, - &from->fence, 0, - I915_FENCE_GFP); + goto await_fence; - err = i915_request_await_start(to, from); - if (err < 0) - return err; + if (i915_request_await_start(to, from) < 0) + goto await_fence; /* Only submit our spinner after the signaler is running! */ - err = __i915_request_await_execution(to, from, NULL, gfp); - if (err) - return err; + if (__i915_request_await_execution(to, from, NULL, gfp)) + goto await_fence; /* We need to pin the signaler's HWSP until we are finished reading. */ - err = intel_timeline_read_hwsp(from, to, &hwsp_offset); - if (err) - return err; + if (intel_timeline_read_hwsp(from, to, &hwsp_offset)) + goto await_fence; - cs = intel_ring_begin(to, 4); + len = 4; + if (has_token) + len += 2; + + cs = intel_ring_begin(to, len); if (IS_ERR(cs)) return PTR_ERR(cs); @@ -821,18 +853,28 @@ emit_semaphore_wait(struct i915_request *to, * (post-wrap) values than they were expecting (and so wait * forever). */ - *cs++ = MI_SEMAPHORE_WAIT | - MI_SEMAPHORE_GLOBAL_GTT | - MI_SEMAPHORE_POLL | - MI_SEMAPHORE_SAD_GTE_SDD; + *cs++ = (MI_SEMAPHORE_WAIT | + MI_SEMAPHORE_GLOBAL_GTT | + MI_SEMAPHORE_POLL | + MI_SEMAPHORE_SAD_GTE_SDD) + + has_token; *cs++ = from->fence.seqno; *cs++ = hwsp_offset; *cs++ = 0; + if (has_token) { + *cs++ = 0; + *cs++ = MI_NOOP; + } intel_ring_advance(to, cs); to->sched.semaphores |= from->engine->mask; to->sched.flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN; return 0; + +await_fence: + return i915_sw_fence_await_dma_fence(&to->submit, + &from->fence, 0, + I915_FENCE_GFP); } static int @@ -916,21 +958,23 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence) /* Squash repeated waits to the same timelines */ if (fence->context && - intel_timeline_sync_is_later(rq->timeline, fence)) + intel_timeline_sync_is_later(i915_request_timeline(rq), + fence)) continue; if (dma_fence_is_i915(fence)) ret = i915_request_await_request(rq, to_request(fence)); else ret = i915_sw_fence_await_dma_fence(&rq->submit, fence, - I915_FENCE_TIMEOUT, + fence->context ? I915_FENCE_TIMEOUT : 0, I915_FENCE_GFP); if (ret < 0) return ret; /* Record the latest fence used against each timeline */ if (fence->context) - intel_timeline_sync_set(rq->timeline, fence); + intel_timeline_sync_set(i915_request_timeline(rq), + fence); } while (--nchild); return 0; @@ -1072,7 +1116,7 @@ void i915_request_skip(struct i915_request *rq, int error) static struct i915_request * __i915_request_add_to_timeline(struct i915_request *rq) { - struct intel_timeline *timeline = rq->timeline; + struct intel_timeline *timeline = i915_request_timeline(rq); struct i915_request *prev; /* @@ -1095,8 +1139,8 @@ __i915_request_add_to_timeline(struct i915_request *rq) * precludes optimising to use semaphores serialisation of a single * timeline across engines. */ - prev = rcu_dereference_protected(timeline->last_request.request, - lockdep_is_held(&timeline->mutex)); + prev = to_request(__i915_active_fence_set(&timeline->last_request, + &rq->fence)); if (prev && !i915_request_completed(prev)) { if (is_power_of_2(prev->engine->mask | rq->engine->mask)) i915_sw_fence_await_sw_fence(&rq->submit, @@ -1121,7 +1165,6 @@ __i915_request_add_to_timeline(struct i915_request *rq) * us, the timeline will hold its seqno which is later than ours. */ GEM_BUG_ON(timeline->seqno != rq->fence.seqno); - __i915_active_request_set(&timeline->last_request, rq); return prev; } @@ -1185,7 +1228,7 @@ void __i915_request_queue(struct i915_request *rq, void i915_request_add(struct i915_request *rq) { struct i915_sched_attr attr = rq->gem_context->sched; - struct intel_timeline * const tl = rq->timeline; + struct intel_timeline * const tl = i915_request_timeline(rq); struct i915_request *prev; lockdep_assert_held(&tl->mutex); @@ -1240,7 +1283,9 @@ void i915_request_add(struct i915_request *rq) * work on behalf of others -- but instead we should benefit from * improved resource management. (Well, that's the theory at least.) */ - if (prev && i915_request_completed(prev) && prev->timeline == tl) + if (prev && + i915_request_completed(prev) && + rcu_access_pointer(prev->timeline) == tl) i915_request_retire_upto(prev); mutex_unlock(&tl->mutex); @@ -1403,7 +1448,7 @@ long i915_request_wait(struct i915_request *rq, * completion. That requires having a good predictor for the request * duration, which we currently lack. */ - if (CONFIG_DRM_I915_SPIN_REQUEST && + if (IS_ACTIVE(CONFIG_DRM_I915_SPIN_REQUEST) && __i915_spin_request(rq, state, CONFIG_DRM_I915_SPIN_REQUEST)) { dma_fence_signal(&rq->fence); goto out; @@ -1423,7 +1468,7 @@ long i915_request_wait(struct i915_request *rq, */ if (flags & I915_WAIT_PRIORITY) { if (!i915_request_started(rq) && INTEL_GEN(rq->i915) >= 6) - gen6_rps_boost(rq); + intel_rps_boost(rq); i915_schedule_bump_priority(rq, I915_PRIORITY_WAIT); } @@ -1449,6 +1494,7 @@ long i915_request_wait(struct i915_request *rq, break; } + intel_engine_flush_submission(rq->engine); timeout = io_schedule_timeout(timeout); } __set_current_state(TASK_RUNNING); @@ -1456,53 +1502,11 @@ long i915_request_wait(struct i915_request *rq, dma_fence_remove_callback(&rq->fence, &wait.cb); out: - mutex_release(&rq->engine->gt->reset.mutex.dep_map, 0, _THIS_IP_); + mutex_release(&rq->engine->gt->reset.mutex.dep_map, _THIS_IP_); trace_i915_request_wait_end(rq); return timeout; } -bool i915_retire_requests(struct drm_i915_private *i915) -{ - struct intel_gt_timelines *timelines = &i915->gt.timelines; - struct intel_timeline *tl, *tn; - unsigned long flags; - LIST_HEAD(free); - - spin_lock_irqsave(&timelines->lock, flags); - list_for_each_entry_safe(tl, tn, &timelines->active_list, link) { - if (!mutex_trylock(&tl->mutex)) - continue; - - intel_timeline_get(tl); - GEM_BUG_ON(!tl->active_count); - tl->active_count++; /* pin the list element */ - spin_unlock_irqrestore(&timelines->lock, flags); - - retire_requests(tl); - - spin_lock_irqsave(&timelines->lock, flags); - - /* Resume iteration after dropping lock */ - list_safe_reset_next(tl, tn, link); - if (!--tl->active_count) - list_del(&tl->link); - - mutex_unlock(&tl->mutex); - - /* Defer the final release to after the spinlock */ - if (refcount_dec_and_test(&tl->kref.refcount)) { - GEM_BUG_ON(tl->active_count); - list_add(&tl->link, &free); - } - } - spin_unlock_irqrestore(&timelines->lock, flags); - - list_for_each_entry_safe(tl, tn, &free, link) - __intel_timeline_free(&tl->kref); - - return !list_empty(&timelines->active_list); -} - #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) #include "selftests/mock_request.c" #include "selftests/i915_request.c" diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h index 8ac6e1226a56..96991d64759c 100644 --- a/drivers/gpu/drm/i915/i915_request.h +++ b/drivers/gpu/drm/i915/i915_request.h @@ -113,7 +113,7 @@ struct i915_request { struct intel_engine_cs *engine; struct intel_context *hw_context; struct intel_ring *ring; - struct intel_timeline *timeline; + struct intel_timeline __rcu *timeline; struct list_head signal_link; /* @@ -211,14 +211,14 @@ struct i915_request { * on the active_list (of their final request). */ struct i915_capture_list *capture_list; - struct list_head active_list; /** Time at which this request was emitted, in jiffies. */ unsigned long emitted_jiffies; unsigned long flags; -#define I915_REQUEST_WAITBOOST BIT(0) -#define I915_REQUEST_NOPREEMPT BIT(1) +#define I915_REQUEST_WAITBOOST BIT(0) +#define I915_REQUEST_NOPREEMPT BIT(1) +#define I915_REQUEST_SENTINEL BIT(2) /** timeline->request entry for this request */ struct list_head link; @@ -251,6 +251,7 @@ struct i915_request *__i915_request_commit(struct i915_request *request); void __i915_request_queue(struct i915_request *rq, const struct i915_sched_attr *attr); +bool i915_request_retire(struct i915_request *rq); void i915_request_retire_upto(struct i915_request *rq); static inline struct i915_request * @@ -292,7 +293,7 @@ int i915_request_await_execution(struct i915_request *rq, void i915_request_add(struct i915_request *rq); -void __i915_request_submit(struct i915_request *request); +bool __i915_request_submit(struct i915_request *request); void i915_request_submit(struct i915_request *request); void i915_request_skip(struct i915_request *request, int error); @@ -309,10 +310,8 @@ long i915_request_wait(struct i915_request *rq, long timeout) __attribute__((nonnull(1))); #define I915_WAIT_INTERRUPTIBLE BIT(0) -#define I915_WAIT_LOCKED BIT(1) /* struct_mutex held, handle GPU reset */ -#define I915_WAIT_PRIORITY BIT(2) /* small priority bump for the request */ -#define I915_WAIT_ALL BIT(3) /* used by i915_gem_object_wait() */ -#define I915_WAIT_FOR_IDLE_BOOST BIT(4) +#define I915_WAIT_PRIORITY BIT(1) /* small priority bump for the request */ +#define I915_WAIT_ALL BIT(2) /* used by i915_gem_object_wait() */ static inline bool i915_request_signaled(const struct i915_request *rq) { @@ -442,6 +441,29 @@ static inline bool i915_request_has_nopreempt(const struct i915_request *rq) return unlikely(rq->flags & I915_REQUEST_NOPREEMPT); } -bool i915_retire_requests(struct drm_i915_private *i915); +static inline bool i915_request_has_sentinel(const struct i915_request *rq) +{ + return unlikely(rq->flags & I915_REQUEST_SENTINEL); +} + +static inline struct intel_timeline * +i915_request_timeline(struct i915_request *rq) +{ + /* Valid only while the request is being constructed (or retired). */ + return rcu_dereference_protected(rq->timeline, + lockdep_is_held(&rcu_access_pointer(rq->timeline)->mutex)); +} + +static inline struct intel_timeline * +i915_request_active_timeline(struct i915_request *rq) +{ + /* + * When in use during submission, we are protected by a guarantee that + * the context/timeline is pinned and must remain pinned until after + * this submission. + */ + return rcu_dereference_protected(rq->timeline, + lockdep_is_held(&rq->engine->active.lock)); +} #endif /* I915_REQUEST_H */ diff --git a/drivers/gpu/drm/i915/i915_scatterlist.h b/drivers/gpu/drm/i915/i915_scatterlist.h index 6617963df9ed..b7b59328cb76 100644 --- a/drivers/gpu/drm/i915/i915_scatterlist.h +++ b/drivers/gpu/drm/i915/i915_scatterlist.h @@ -67,15 +67,15 @@ static inline struct scatterlist *__sg_next(struct scatterlist *sg) } /** - * __for_each_sgt_dma - iterate over the DMA addresses of the given sg_table - * @__dmap: DMA address (output) + * __for_each_sgt_daddr - iterate over the device addresses of the given sg_table + * @__dp: Device address (output) * @__iter: 'struct sgt_iter' (iterator state, internal) * @__sgt: sg_table to iterate over (input) * @__step: step size */ -#define __for_each_sgt_dma(__dmap, __iter, __sgt, __step) \ +#define __for_each_sgt_daddr(__dp, __iter, __sgt, __step) \ for ((__iter) = __sgt_iter((__sgt)->sgl, true); \ - ((__dmap) = (__iter).dma + (__iter).curr); \ + ((__dp) = (__iter).dma + (__iter).curr), (__iter).sgp; \ (((__iter).curr += (__step)) >= (__iter).max) ? \ (__iter) = __sgt_iter(__sg_next((__iter).sgp), true), 0 : 0) diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c index 7b84ebca2901..010d67f48ad9 100644 --- a/drivers/gpu/drm/i915/i915_scheduler.c +++ b/drivers/gpu/drm/i915/i915_scheduler.c @@ -177,21 +177,51 @@ static inline int rq_prio(const struct i915_request *rq) return rq->sched.attr.priority | __NO_PREEMPTION; } -static void kick_submission(struct intel_engine_cs *engine, int prio) +static inline bool need_preempt(int prio, int active) { - const struct i915_request *inflight = *engine->execlists.active; + /* + * Allow preemption of low -> normal -> high, but we do + * not allow low priority tasks to preempt other low priority + * tasks under the impression that latency for low priority + * tasks does not matter (as much as background throughput), + * so kiss. + */ + return prio >= max(I915_PRIORITY_NORMAL, active); +} + +static void kick_submission(struct intel_engine_cs *engine, + const struct i915_request *rq, + int prio) +{ + const struct i915_request *inflight; /* - * If we are already the currently executing context, don't - * bother evaluating if we should preempt ourselves, or if - * we expect nothing to change as a result of running the - * tasklet, i.e. we have not change the priority queue - * sufficiently to oust the running context. + * We only need to kick the tasklet once for the high priority + * new context we add into the queue. */ - if (!inflight || !i915_scheduler_need_preempt(prio, rq_prio(inflight))) + if (prio <= engine->execlists.queue_priority_hint) return; - tasklet_hi_schedule(&engine->execlists.tasklet); + rcu_read_lock(); + + /* Nothing currently active? We're overdue for a submission! */ + inflight = execlists_active(&engine->execlists); + if (!inflight) + goto unlock; + + /* + * If we are already the currently executing context, don't + * bother evaluating if we should preempt ourselves. + */ + if (inflight->hw_context == rq->hw_context) + goto unlock; + + engine->execlists.queue_priority_hint = prio; + if (need_preempt(prio, rq_prio(inflight))) + tasklet_hi_schedule(&engine->execlists.tasklet); + +unlock: + rcu_read_unlock(); } static void __i915_schedule(struct i915_sched_node *node, @@ -317,13 +347,8 @@ static void __i915_schedule(struct i915_sched_node *node, list_move_tail(&node->link, cache.priolist); } - if (prio <= engine->execlists.queue_priority_hint) - continue; - - engine->execlists.queue_priority_hint = prio; - /* Defer (tasklet) submission until after all of our updates. */ - kick_submission(engine, prio); + kick_submission(engine, node_to_request(node), prio); } spin_unlock(&engine->active.lock); diff --git a/drivers/gpu/drm/i915/i915_scheduler.h b/drivers/gpu/drm/i915/i915_scheduler.h index 7eefccff39bf..07d243acf553 100644 --- a/drivers/gpu/drm/i915/i915_scheduler.h +++ b/drivers/gpu/drm/i915/i915_scheduler.h @@ -52,22 +52,4 @@ static inline void i915_priolist_free(struct i915_priolist *p) __i915_priolist_free(p); } -static inline bool i915_scheduler_need_preempt(int prio, int active) -{ - /* - * Allow preemption of low -> normal -> high, but we do - * not allow low priority tasks to preempt other low priority - * tasks under the impression that latency for low priority - * tasks does not matter (as much as background throughput), - * so kiss. - * - * More naturally we would write - * prio >= max(0, last); - * except that we wish to prevent triggering preemption at the same - * priority level: the task that is running should remain running - * to preserve FIFO ordering of dependencies. - */ - return prio > max(I915_PRIORITY_NORMAL - 1, active); -} - #endif /* _I915_SCHEDULER_H_ */ diff --git a/drivers/gpu/drm/i915/i915_scheduler_types.h b/drivers/gpu/drm/i915/i915_scheduler_types.h index aad81acba9dc..d18e70550054 100644 --- a/drivers/gpu/drm/i915/i915_scheduler_types.h +++ b/drivers/gpu/drm/i915/i915_scheduler_types.h @@ -49,6 +49,15 @@ struct i915_sched_attr { * DAG of each request, we are able to insert it into a sorted queue when it * is ready, and are able to reorder its portion of the graph to accommodate * dynamic priority changes. + * + * Ok, there is now one active element to the "scheduler" in the backends. + * We let a new context run for a small amount of time before re-evaluating + * the run order. As we re-evaluate, we maintain the strict ordering of + * dependencies, but attempt to rotate the active contexts (the current context + * is put to the back of its priority queue, then reshuffling its dependents). + * This provides minimal timeslicing and prevents a userspace hog (e.g. + * something waiting on a user semaphore [VkEvent]) from denying service to + * others. */ struct i915_sched_node { struct list_head signalers_list; /* those before us, we depend upon */ diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c index 8508a01ad8b9..8812cdd9007f 100644 --- a/drivers/gpu/drm/i915/i915_suspend.c +++ b/drivers/gpu/drm/i915/i915_suspend.c @@ -28,6 +28,7 @@ #include "display/intel_fbc.h" #include "display/intel_gmbus.h" +#include "display/intel_vga.h" #include "i915_drv.h" #include "i915_reg.h" @@ -57,7 +58,7 @@ static void i915_restore_display(struct drm_i915_private *dev_priv) if (HAS_FBC(dev_priv) && INTEL_GEN(dev_priv) <= 4 && !IS_G4X(dev_priv)) I915_WRITE(FBC_CONTROL, dev_priv->regfile.saveFBC_CONTROL); - i915_redisable_vga(dev_priv); + intel_vga_redisable(dev_priv); } int i915_save_state(struct drm_i915_private *dev_priv) @@ -65,8 +66,6 @@ int i915_save_state(struct drm_i915_private *dev_priv) struct pci_dev *pdev = dev_priv->drm.pdev; int i; - mutex_lock(&dev_priv->drm.struct_mutex); - i915_save_display(dev_priv); if (IS_GEN(dev_priv, 4)) @@ -100,8 +99,6 @@ int i915_save_state(struct drm_i915_private *dev_priv) dev_priv->regfile.saveSWF3[i] = I915_READ(SWF3(i)); } - mutex_unlock(&dev_priv->drm.struct_mutex); - return 0; } @@ -110,8 +107,6 @@ int i915_restore_state(struct drm_i915_private *dev_priv) struct pci_dev *pdev = dev_priv->drm.pdev; int i; - mutex_lock(&dev_priv->drm.struct_mutex); - if (IS_GEN(dev_priv, 4)) pci_write_config_word(pdev, GCDGMBUS, dev_priv->regfile.saveGCDGMBUS); @@ -145,8 +140,6 @@ int i915_restore_state(struct drm_i915_private *dev_priv) I915_WRITE(SWF3(i), dev_priv->regfile.saveSWF3[i]); } - mutex_unlock(&dev_priv->drm.struct_mutex); - intel_gmbus_reset(dev_priv); return 0; diff --git a/drivers/gpu/drm/i915/i915_switcheroo.c b/drivers/gpu/drm/i915/i915_switcheroo.c new file mode 100644 index 000000000000..39c79e1c5b52 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_switcheroo.c @@ -0,0 +1,67 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include <linux/vga_switcheroo.h> + +#include "i915_drv.h" +#include "i915_switcheroo.h" + +static void i915_switcheroo_set_state(struct pci_dev *pdev, + enum vga_switcheroo_state state) +{ + struct drm_i915_private *i915 = pdev_to_i915(pdev); + pm_message_t pmm = { .event = PM_EVENT_SUSPEND }; + + if (!i915) { + dev_err(&pdev->dev, "DRM not initialized, aborting switch.\n"); + return; + } + + if (state == VGA_SWITCHEROO_ON) { + pr_info("switched on\n"); + i915->drm.switch_power_state = DRM_SWITCH_POWER_CHANGING; + /* i915 resume handler doesn't set to D0 */ + pci_set_power_state(pdev, PCI_D0); + i915_resume_switcheroo(i915); + i915->drm.switch_power_state = DRM_SWITCH_POWER_ON; + } else { + pr_info("switched off\n"); + i915->drm.switch_power_state = DRM_SWITCH_POWER_CHANGING; + i915_suspend_switcheroo(i915, pmm); + i915->drm.switch_power_state = DRM_SWITCH_POWER_OFF; + } +} + +static bool i915_switcheroo_can_switch(struct pci_dev *pdev) +{ + struct drm_i915_private *i915 = pdev_to_i915(pdev); + + /* + * FIXME: open_count is protected by drm_global_mutex but that would lead to + * locking inversion with the driver load path. And the access here is + * completely racy anyway. So don't bother with locking for now. + */ + return i915 && i915->drm.open_count == 0; +} + +static const struct vga_switcheroo_client_ops i915_switcheroo_ops = { + .set_gpu_state = i915_switcheroo_set_state, + .reprobe = NULL, + .can_switch = i915_switcheroo_can_switch, +}; + +int i915_switcheroo_register(struct drm_i915_private *i915) +{ + struct pci_dev *pdev = i915->drm.pdev; + + return vga_switcheroo_register_client(pdev, &i915_switcheroo_ops, false); +} + +void i915_switcheroo_unregister(struct drm_i915_private *i915) +{ + struct pci_dev *pdev = i915->drm.pdev; + + vga_switcheroo_unregister_client(pdev); +} diff --git a/drivers/gpu/drm/i915/i915_switcheroo.h b/drivers/gpu/drm/i915/i915_switcheroo.h new file mode 100644 index 000000000000..59b6c1e07d75 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_switcheroo.h @@ -0,0 +1,14 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __I915_SWITCHEROO__ +#define __I915_SWITCHEROO__ + +struct drm_i915_private; + +int i915_switcheroo_register(struct drm_i915_private *i915); +void i915_switcheroo_unregister(struct drm_i915_private *i915); + +#endif /* __I915_SWITCHEROO__ */ diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c index d8a3b180c084..65476909d1bf 100644 --- a/drivers/gpu/drm/i915/i915_sysfs.c +++ b/drivers/gpu/drm/i915/i915_sysfs.c @@ -30,6 +30,9 @@ #include <linux/stat.h> #include <linux/sysfs.h> +#include "gt/intel_rc6.h" +#include "gt/intel_rps.h" + #include "i915_drv.h" #include "i915_sysfs.h" #include "intel_pm.h" @@ -49,7 +52,7 @@ static u32 calc_residency(struct drm_i915_private *dev_priv, u64 res = 0; with_intel_runtime_pm(&dev_priv->runtime_pm, wakeref) - res = intel_rc6_residency_us(dev_priv, reg); + res = intel_rc6_residency_us(&dev_priv->gt.rc6, reg); return DIV_ROUND_CLOSEST_ULL(res, 1000); } @@ -142,12 +145,12 @@ static const struct attribute_group media_rc6_attr_group = { }; #endif -static int l3_access_valid(struct drm_i915_private *dev_priv, loff_t offset) +static int l3_access_valid(struct drm_i915_private *i915, loff_t offset) { - if (!HAS_L3_DPF(dev_priv)) + if (!HAS_L3_DPF(i915)) return -EPERM; - if (offset % 4 != 0) + if (!IS_ALIGNED(offset, sizeof(u32))) return -EINVAL; if (offset >= GEN7_L3LOG_SIZE) @@ -162,31 +165,24 @@ i915_l3_read(struct file *filp, struct kobject *kobj, loff_t offset, size_t count) { struct device *kdev = kobj_to_dev(kobj); - struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); - struct drm_device *dev = &dev_priv->drm; + struct drm_i915_private *i915 = kdev_minor_to_i915(kdev); int slice = (int)(uintptr_t)attr->private; int ret; - count = round_down(count, 4); - - ret = l3_access_valid(dev_priv, offset); + ret = l3_access_valid(i915, offset); if (ret) return ret; + count = round_down(count, sizeof(u32)); count = min_t(size_t, GEN7_L3LOG_SIZE - offset, count); + memset(buf, 0, count); - ret = i915_mutex_lock_interruptible(dev); - if (ret) - return ret; - - if (dev_priv->l3_parity.remap_info[slice]) + spin_lock(&i915->gem.contexts.lock); + if (i915->l3_parity.remap_info[slice]) memcpy(buf, - dev_priv->l3_parity.remap_info[slice] + (offset/4), + i915->l3_parity.remap_info[slice] + offset / sizeof(u32), count); - else - memset(buf, 0, count); - - mutex_unlock(&dev->struct_mutex); + spin_unlock(&i915->gem.contexts.lock); return count; } @@ -197,46 +193,49 @@ i915_l3_write(struct file *filp, struct kobject *kobj, loff_t offset, size_t count) { struct device *kdev = kobj_to_dev(kobj); - struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); - struct drm_device *dev = &dev_priv->drm; - struct i915_gem_context *ctx; + struct drm_i915_private *i915 = kdev_minor_to_i915(kdev); int slice = (int)(uintptr_t)attr->private; - u32 **remap_info; + u32 *remap_info, *freeme = NULL; + struct i915_gem_context *ctx; int ret; - ret = l3_access_valid(dev_priv, offset); + ret = l3_access_valid(i915, offset); if (ret) return ret; - ret = i915_mutex_lock_interruptible(dev); - if (ret) - return ret; + if (count < sizeof(u32)) + return -EINVAL; - remap_info = &dev_priv->l3_parity.remap_info[slice]; - if (!*remap_info) { - *remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL); - if (!*remap_info) { - ret = -ENOMEM; - goto out; - } + remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL); + if (!remap_info) + return -ENOMEM; + + spin_lock(&i915->gem.contexts.lock); + + if (i915->l3_parity.remap_info[slice]) { + freeme = remap_info; + remap_info = i915->l3_parity.remap_info[slice]; + } else { + i915->l3_parity.remap_info[slice] = remap_info; } - /* TODO: Ideally we really want a GPU reset here to make sure errors - * aren't propagated. Since I cannot find a stable way to reset the GPU - * at this point it is left as a TODO. - */ - memcpy(*remap_info + (offset/4), buf, count); + count = round_down(count, sizeof(u32)); + memcpy(remap_info + offset / sizeof(u32), buf, count); /* NB: We defer the remapping until we switch to the context */ - list_for_each_entry(ctx, &dev_priv->contexts.list, link) - ctx->remap_slice |= (1<<slice); + list_for_each_entry(ctx, &i915->gem.contexts.list, link) + ctx->remap_slice |= BIT(slice); - ret = count; + spin_unlock(&i915->gem.contexts.lock); + kfree(freeme); -out: - mutex_unlock(&dev->struct_mutex); + /* + * TODO: Ideally we really want a GPU reset here to make sure errors + * aren't propagated. Since I cannot find a stable way to reset the GPU + * at this point it is left as a TODO. + */ - return ret; + return count; } static const struct bin_attribute dpf_attrs = { @@ -261,6 +260,7 @@ static ssize_t gt_act_freq_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); + struct intel_rps *rps = &dev_priv->gt.rps; intel_wakeref_t wakeref; u32 freq; @@ -273,31 +273,31 @@ static ssize_t gt_act_freq_mhz_show(struct device *kdev, freq = (freq >> 8) & 0xff; } else { - freq = intel_get_cagf(dev_priv, I915_READ(GEN6_RPSTAT1)); + freq = intel_get_cagf(rps, I915_READ(GEN6_RPSTAT1)); } intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); - return snprintf(buf, PAGE_SIZE, "%d\n", intel_gpu_freq(dev_priv, freq)); + return snprintf(buf, PAGE_SIZE, "%d\n", intel_gpu_freq(rps, freq)); } static ssize_t gt_cur_freq_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); + struct intel_rps *rps = &dev_priv->gt.rps; return snprintf(buf, PAGE_SIZE, "%d\n", - intel_gpu_freq(dev_priv, - dev_priv->gt_pm.rps.cur_freq)); + intel_gpu_freq(rps, rps->cur_freq)); } static ssize_t gt_boost_freq_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); + struct intel_rps *rps = &dev_priv->gt.rps; return snprintf(buf, PAGE_SIZE, "%d\n", - intel_gpu_freq(dev_priv, - dev_priv->gt_pm.rps.boost_freq)); + intel_gpu_freq(rps, rps->boost_freq)); } static ssize_t gt_boost_freq_mhz_store(struct device *kdev, @@ -305,7 +305,7 @@ static ssize_t gt_boost_freq_mhz_store(struct device *kdev, const char *buf, size_t count) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); - struct intel_rps *rps = &dev_priv->gt_pm.rps; + struct intel_rps *rps = &dev_priv->gt.rps; bool boost = false; ssize_t ret; u32 val; @@ -315,7 +315,7 @@ static ssize_t gt_boost_freq_mhz_store(struct device *kdev, return ret; /* Validate against (static) hardware limits */ - val = intel_freq_opcode(dev_priv, val); + val = intel_freq_opcode(rps, val); if (val < rps->min_freq || val > rps->max_freq) return -EINVAL; @@ -335,19 +335,19 @@ static ssize_t vlv_rpe_freq_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); + struct intel_rps *rps = &dev_priv->gt.rps; return snprintf(buf, PAGE_SIZE, "%d\n", - intel_gpu_freq(dev_priv, - dev_priv->gt_pm.rps.efficient_freq)); + intel_gpu_freq(rps, rps->efficient_freq)); } static ssize_t gt_max_freq_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); + struct intel_rps *rps = &dev_priv->gt.rps; return snprintf(buf, PAGE_SIZE, "%d\n", - intel_gpu_freq(dev_priv, - dev_priv->gt_pm.rps.max_freq_softlimit)); + intel_gpu_freq(rps, rps->max_freq_softlimit)); } static ssize_t gt_max_freq_mhz_store(struct device *kdev, @@ -355,19 +355,17 @@ static ssize_t gt_max_freq_mhz_store(struct device *kdev, const char *buf, size_t count) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); - struct intel_rps *rps = &dev_priv->gt_pm.rps; - intel_wakeref_t wakeref; - u32 val; + struct intel_rps *rps = &dev_priv->gt.rps; ssize_t ret; + u32 val; ret = kstrtou32(buf, 0, &val); if (ret) return ret; - wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); mutex_lock(&rps->lock); - val = intel_freq_opcode(dev_priv, val); + val = intel_freq_opcode(rps, val); if (val < rps->min_freq || val > rps->max_freq || val < rps->min_freq_softlimit) { @@ -377,7 +375,7 @@ static ssize_t gt_max_freq_mhz_store(struct device *kdev, if (val > rps->rp0_freq) DRM_DEBUG("User requested overclocking to %d\n", - intel_gpu_freq(dev_priv, val)); + intel_gpu_freq(rps, val)); rps->max_freq_softlimit = val; @@ -385,14 +383,15 @@ static ssize_t gt_max_freq_mhz_store(struct device *kdev, rps->min_freq_softlimit, rps->max_freq_softlimit); - /* We still need *_set_rps to process the new max_delay and + /* + * We still need *_set_rps to process the new max_delay and * update the interrupt limits and PMINTRMSK even though - * frequency request may be unchanged. */ - ret = intel_set_rps(dev_priv, val); + * frequency request may be unchanged. + */ + intel_rps_set(rps, val); unlock: mutex_unlock(&rps->lock); - intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); return ret ?: count; } @@ -400,10 +399,10 @@ unlock: static ssize_t gt_min_freq_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); + struct intel_rps *rps = &dev_priv->gt.rps; return snprintf(buf, PAGE_SIZE, "%d\n", - intel_gpu_freq(dev_priv, - dev_priv->gt_pm.rps.min_freq_softlimit)); + intel_gpu_freq(rps, rps->min_freq_softlimit)); } static ssize_t gt_min_freq_mhz_store(struct device *kdev, @@ -411,19 +410,17 @@ static ssize_t gt_min_freq_mhz_store(struct device *kdev, const char *buf, size_t count) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); - struct intel_rps *rps = &dev_priv->gt_pm.rps; - intel_wakeref_t wakeref; - u32 val; + struct intel_rps *rps = &dev_priv->gt.rps; ssize_t ret; + u32 val; ret = kstrtou32(buf, 0, &val); if (ret) return ret; - wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); mutex_lock(&rps->lock); - val = intel_freq_opcode(dev_priv, val); + val = intel_freq_opcode(rps, val); if (val < rps->min_freq || val > rps->max_freq || val > rps->max_freq_softlimit) { @@ -437,14 +434,15 @@ static ssize_t gt_min_freq_mhz_store(struct device *kdev, rps->min_freq_softlimit, rps->max_freq_softlimit); - /* We still need *_set_rps to process the new min_delay and + /* + * We still need *_set_rps to process the new min_delay and * update the interrupt limits and PMINTRMSK even though - * frequency request may be unchanged. */ - ret = intel_set_rps(dev_priv, val); + * frequency request may be unchanged. + */ + intel_rps_set(rps, val); unlock: mutex_unlock(&rps->lock); - intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); return ret ?: count; } @@ -466,15 +464,15 @@ static DEVICE_ATTR(gt_RPn_freq_mhz, S_IRUGO, gt_rp_mhz_show, NULL); static ssize_t gt_rp_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf) { struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev); - struct intel_rps *rps = &dev_priv->gt_pm.rps; + struct intel_rps *rps = &dev_priv->gt.rps; u32 val; if (attr == &dev_attr_gt_RP0_freq_mhz) - val = intel_gpu_freq(dev_priv, rps->rp0_freq); + val = intel_gpu_freq(rps, rps->rp0_freq); else if (attr == &dev_attr_gt_RP1_freq_mhz) - val = intel_gpu_freq(dev_priv, rps->rp1_freq); + val = intel_gpu_freq(rps, rps->rp1_freq); else if (attr == &dev_attr_gt_RPn_freq_mhz) - val = intel_gpu_freq(dev_priv, rps->min_freq); + val = intel_gpu_freq(rps, rps->min_freq); else BUG(); diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h index 24f2944da09d..7ef7a1e1664c 100644 --- a/drivers/gpu/drm/i915/i915_trace.h +++ b/drivers/gpu/drm/i915/i915_trace.h @@ -665,7 +665,6 @@ TRACE_EVENT(i915_request_queue, TP_STRUCT__entry( __field(u32, dev) - __field(u32, hw_id) __field(u64, ctx) __field(u16, class) __field(u16, instance) @@ -675,7 +674,6 @@ TRACE_EVENT(i915_request_queue, TP_fast_assign( __entry->dev = rq->i915->drm.primary->index; - __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; @@ -683,10 +681,9 @@ TRACE_EVENT(i915_request_queue, __entry->flags = flags; ), - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, flags=0x%x", + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, flags=0x%x", __entry->dev, __entry->class, __entry->instance, - __entry->hw_id, __entry->ctx, __entry->seqno, - __entry->flags) + __entry->ctx, __entry->seqno, __entry->flags) ); DECLARE_EVENT_CLASS(i915_request, @@ -695,7 +692,6 @@ DECLARE_EVENT_CLASS(i915_request, TP_STRUCT__entry( __field(u32, dev) - __field(u32, hw_id) __field(u64, ctx) __field(u16, class) __field(u16, instance) @@ -704,16 +700,15 @@ DECLARE_EVENT_CLASS(i915_request, TP_fast_assign( __entry->dev = rq->i915->drm.primary->index; - __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; __entry->seqno = rq->fence.seqno; ), - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u", + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u", __entry->dev, __entry->class, __entry->instance, - __entry->hw_id, __entry->ctx, __entry->seqno) + __entry->ctx, __entry->seqno) ); DEFINE_EVENT(i915_request, i915_request_add, @@ -738,7 +733,6 @@ TRACE_EVENT(i915_request_in, TP_STRUCT__entry( __field(u32, dev) - __field(u32, hw_id) __field(u64, ctx) __field(u16, class) __field(u16, instance) @@ -749,7 +743,6 @@ TRACE_EVENT(i915_request_in, TP_fast_assign( __entry->dev = rq->i915->drm.primary->index; - __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; @@ -758,9 +751,9 @@ TRACE_EVENT(i915_request_in, __entry->port = port; ), - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, prio=%u, port=%u", + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, prio=%u, port=%u", __entry->dev, __entry->class, __entry->instance, - __entry->hw_id, __entry->ctx, __entry->seqno, + __entry->ctx, __entry->seqno, __entry->prio, __entry->port) ); @@ -770,7 +763,6 @@ TRACE_EVENT(i915_request_out, TP_STRUCT__entry( __field(u32, dev) - __field(u32, hw_id) __field(u64, ctx) __field(u16, class) __field(u16, instance) @@ -780,7 +772,6 @@ TRACE_EVENT(i915_request_out, TP_fast_assign( __entry->dev = rq->i915->drm.primary->index; - __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; @@ -788,10 +779,9 @@ TRACE_EVENT(i915_request_out, __entry->completed = i915_request_completed(rq); ), - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, completed?=%u", + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, completed?=%u", __entry->dev, __entry->class, __entry->instance, - __entry->hw_id, __entry->ctx, __entry->seqno, - __entry->completed) + __entry->ctx, __entry->seqno, __entry->completed) ); #else @@ -829,7 +819,6 @@ TRACE_EVENT(i915_request_wait_begin, TP_STRUCT__entry( __field(u32, dev) - __field(u32, hw_id) __field(u64, ctx) __field(u16, class) __field(u16, instance) @@ -845,7 +834,6 @@ TRACE_EVENT(i915_request_wait_begin, */ TP_fast_assign( __entry->dev = rq->i915->drm.primary->index; - __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; @@ -853,9 +841,9 @@ TRACE_EVENT(i915_request_wait_begin, __entry->flags = flags; ), - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, flags=0x%x", + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, flags=0x%x", __entry->dev, __entry->class, __entry->instance, - __entry->hw_id, __entry->ctx, __entry->seqno, + __entry->ctx, __entry->seqno, __entry->flags) ); @@ -958,19 +946,17 @@ DECLARE_EVENT_CLASS(i915_context, TP_STRUCT__entry( __field(u32, dev) __field(struct i915_gem_context *, ctx) - __field(u32, hw_id) __field(struct i915_address_space *, vm) ), TP_fast_assign( __entry->dev = ctx->i915->drm.primary->index; __entry->ctx = ctx; - __entry->hw_id = ctx->hw_id; - __entry->vm = ctx->vm; + __entry->vm = rcu_access_pointer(ctx->vm); ), - TP_printk("dev=%u, ctx=%p, ctx_vm=%p, hw_id=%u", - __entry->dev, __entry->ctx, __entry->vm, __entry->hw_id) + TP_printk("dev=%u, ctx=%p, ctx_vm=%p", + __entry->dev, __entry->ctx, __entry->vm) ) DEFINE_EVENT(i915_context, i915_context_create, diff --git a/drivers/gpu/drm/i915/i915_utils.c b/drivers/gpu/drm/i915/i915_utils.c index 16acdf7bdbe6..0348c6d0ef5f 100644 --- a/drivers/gpu/drm/i915/i915_utils.c +++ b/drivers/gpu/drm/i915/i915_utils.c @@ -54,25 +54,54 @@ __i915_printk(struct drm_i915_private *dev_priv, const char *level, #if IS_ENABLED(CONFIG_DRM_I915_DEBUG) static unsigned int i915_probe_fail_count; -int __i915_inject_load_error(struct drm_i915_private *i915, int err, - const char *func, int line) +int __i915_inject_probe_error(struct drm_i915_private *i915, int err, + const char *func, int line) { - if (i915_probe_fail_count >= i915_modparams.inject_load_failure) + if (i915_probe_fail_count >= i915_modparams.inject_probe_failure) return 0; - if (++i915_probe_fail_count < i915_modparams.inject_load_failure) + if (++i915_probe_fail_count < i915_modparams.inject_probe_failure) return 0; __i915_printk(i915, KERN_INFO, "Injecting failure %d at checkpoint %u [%s:%d]\n", - err, i915_modparams.inject_load_failure, func, line); - i915_modparams.inject_load_failure = 0; + err, i915_modparams.inject_probe_failure, func, line); + i915_modparams.inject_probe_failure = 0; return err; } bool i915_error_injected(void) { - return i915_probe_fail_count && !i915_modparams.inject_load_failure; + return i915_probe_fail_count && !i915_modparams.inject_probe_failure; } #endif + +void cancel_timer(struct timer_list *t) +{ + if (!READ_ONCE(t->expires)) + return; + + del_timer(t); + WRITE_ONCE(t->expires, 0); +} + +void set_timer_ms(struct timer_list *t, unsigned long timeout) +{ + if (!timeout) { + cancel_timer(t); + return; + } + + timeout = msecs_to_jiffies_timeout(timeout); + + /* + * Paranoia to make sure the compiler computes the timeout before + * loading 'jiffies' as jiffies is volatile and may be updated in + * the background by a timer tick. All to reduce the complexity + * of the addition and reduce the risk of losing a jiffie. + */ + barrier(); + + mod_timer(t, jiffies + timeout); +} diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h index 562f756da421..04139ba1191e 100644 --- a/drivers/gpu/drm/i915/i915_utils.h +++ b/drivers/gpu/drm/i915/i915_utils.h @@ -32,6 +32,7 @@ #include <linux/workqueue.h> struct drm_i915_private; +struct timer_list; #undef WARN_ON /* Many gcc seem to no see through this and fall over :( */ @@ -60,20 +61,20 @@ __i915_printk(struct drm_i915_private *dev_priv, const char *level, #if IS_ENABLED(CONFIG_DRM_I915_DEBUG) -int __i915_inject_load_error(struct drm_i915_private *i915, int err, - const char *func, int line); -#define i915_inject_load_error(_i915, _err) \ - __i915_inject_load_error((_i915), (_err), __func__, __LINE__) +int __i915_inject_probe_error(struct drm_i915_private *i915, int err, + const char *func, int line); +#define i915_inject_probe_error(_i915, _err) \ + __i915_inject_probe_error((_i915), (_err), __func__, __LINE__) bool i915_error_injected(void); #else -#define i915_inject_load_error(_i915, _err) 0 +#define i915_inject_probe_error(_i915, _err) 0 #define i915_error_injected() false #endif -#define i915_inject_probe_failure(i915) i915_inject_load_error((i915), -ENODEV) +#define i915_inject_probe_failure(i915) i915_inject_probe_error((i915), -ENODEV) #define i915_probe_error(i915, fmt, ...) \ __i915_printk(i915, i915_error_injected() ? KERN_DEBUG : KERN_ERR, \ @@ -421,4 +422,25 @@ static inline void add_taint_for_CI(unsigned int taint) add_taint(taint, LOCKDEP_STILL_OK); } +void cancel_timer(struct timer_list *t); +void set_timer_ms(struct timer_list *t, unsigned long timeout); + +static inline bool timer_expired(const struct timer_list *t) +{ + return READ_ONCE(t->expires) && !timer_pending(t); +} + +/* + * This is a lookalike for IS_ENABLED() that takes a kconfig value, + * e.g. CONFIG_DRM_I915_SPIN_REQUEST, and evaluates whether it is non-zero + * i.e. whether the configuration is active. Wrapping up the config inside + * a boolean context prevents clang and smatch from complaining about potential + * issues in confusing logical-&& with bitwise-& for constants. + * + * Sadly IS_ENABLED() itself does not work with kconfig values. + * + * Returns 0 if @config is 0, 1 if set to any value. + */ +#define IS_ACTIVE(config) ((config) != 0) + #endif /* !__I915_UTILS_H */ diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c index e0e677b2a3a9..e5512f26e20a 100644 --- a/drivers/gpu/drm/i915/i915_vma.c +++ b/drivers/gpu/drm/i915/i915_vma.c @@ -32,6 +32,7 @@ #include "i915_drv.h" #include "i915_globals.h" +#include "i915_sw_fence_work.h" #include "i915_trace.h" #include "i915_vma.h" @@ -90,6 +91,7 @@ static int __i915_vma_active(struct i915_active *ref) return i915_vma_tryget(active_to_vma(ref)) ? 0 : -ENOENT; } +__i915_active_call static void __i915_vma_retire(struct i915_active *ref) { i915_vma_put(active_to_vma(ref)); @@ -104,21 +106,21 @@ vma_create(struct drm_i915_gem_object *obj, struct rb_node *rb, **p; /* The aliasing_ppgtt should never be used directly! */ - GEM_BUG_ON(vm == &vm->i915->ggtt.alias->vm); + GEM_BUG_ON(vm == &vm->gt->ggtt->alias->vm); vma = i915_vma_alloc(); if (vma == NULL) return ERR_PTR(-ENOMEM); - vma->vm = vm; + mutex_init(&vma->pages_mutex); + vma->vm = i915_vm_get(vm); vma->ops = &vm->vma_ops; vma->obj = obj; vma->resv = obj->base.resv; vma->size = obj->base.size; vma->display_alignment = I915_GTT_MIN_ALIGNMENT; - i915_active_init(vm->i915, &vma->active, - __i915_vma_active, __i915_vma_retire); + i915_active_init(&vma->active, __i915_vma_active, __i915_vma_retire); /* Declare ourselves safe for use inside shrinkers */ if (IS_ENABLED(CONFIG_LOCKDEP)) { @@ -171,7 +173,7 @@ vma_create(struct drm_i915_gem_object *obj, i915_gem_object_get_stride(obj)); GEM_BUG_ON(!is_power_of_2(vma->fence_alignment)); - vma->flags |= I915_VMA_GGTT; + __set_bit(I915_VMA_GGTT_BIT, __i915_vma_flags(vma)); } spin_lock(&obj->vma.lock); @@ -218,10 +220,6 @@ vma_create(struct drm_i915_gem_object *obj, spin_unlock(&obj->vma.lock); - mutex_lock(&vm->mutex); - list_add(&vma->vm_link, &vm->unbound_list); - mutex_unlock(&vm->mutex); - return vma; err_vma: @@ -265,8 +263,6 @@ vma_lookup(struct drm_i915_gem_object *obj, * Once created, the VMA is kept until either the object is freed, or the * address space is closed. * - * Must be called with struct_mutex held. - * * Returns the vma, or an error pointer. */ struct i915_vma * @@ -277,7 +273,7 @@ i915_vma_instance(struct drm_i915_gem_object *obj, struct i915_vma *vma; GEM_BUG_ON(view && !i915_is_ggtt(vm)); - GEM_BUG_ON(vm->closed); + GEM_BUG_ON(!atomic_read(&vm->open)); spin_lock(&obj->vma.lock); vma = vma_lookup(obj, vm, view); @@ -291,18 +287,63 @@ i915_vma_instance(struct drm_i915_gem_object *obj, return vma; } +struct i915_vma_work { + struct dma_fence_work base; + struct i915_vma *vma; + enum i915_cache_level cache_level; + unsigned int flags; +}; + +static int __vma_bind(struct dma_fence_work *work) +{ + struct i915_vma_work *vw = container_of(work, typeof(*vw), base); + struct i915_vma *vma = vw->vma; + int err; + + err = vma->ops->bind_vma(vma, vw->cache_level, vw->flags); + if (err) + atomic_or(I915_VMA_ERROR, &vma->flags); + + if (vma->obj) + __i915_gem_object_unpin_pages(vma->obj); + + return err; +} + +static const struct dma_fence_work_ops bind_ops = { + .name = "bind", + .work = __vma_bind, +}; + +struct i915_vma_work *i915_vma_work(void) +{ + struct i915_vma_work *vw; + + vw = kzalloc(sizeof(*vw), GFP_KERNEL); + if (!vw) + return NULL; + + dma_fence_work_init(&vw->base, &bind_ops); + vw->base.dma.error = -EAGAIN; /* disable the worker by default */ + + return vw; +} + /** * i915_vma_bind - Sets up PTEs for an VMA in it's corresponding address space. * @vma: VMA to map * @cache_level: mapping cache level * @flags: flags like global or local mapping + * @work: preallocated worker for allocating and binding the PTE * * DMA addresses are taken from the scatter-gather table of this object (or of * this VMA in case of non-default GGTT views) and PTE entries set up. * Note that DMA addresses are also the only part of the SG table we care about. */ -int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level, - u32 flags) +int i915_vma_bind(struct i915_vma *vma, + enum i915_cache_level cache_level, + u32 flags, + struct i915_vma_work *work) { u32 bind_flags; u32 vma_flags; @@ -319,13 +360,11 @@ int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level, if (GEM_DEBUG_WARN_ON(!flags)) return -EINVAL; - bind_flags = 0; - if (flags & PIN_GLOBAL) - bind_flags |= I915_VMA_GLOBAL_BIND; - if (flags & PIN_USER) - bind_flags |= I915_VMA_LOCAL_BIND; + bind_flags = flags; + bind_flags &= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND; - vma_flags = vma->flags & (I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND); + vma_flags = atomic_read(&vma->flags); + vma_flags &= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND; if (flags & PIN_UPDATE) bind_flags |= vma_flags; else @@ -336,11 +375,34 @@ int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level, GEM_BUG_ON(!vma->pages); trace_i915_vma_bind(vma, bind_flags); - ret = vma->ops->bind_vma(vma, cache_level, bind_flags); - if (ret) - return ret; + if (work && (bind_flags & ~vma_flags) & vma->vm->bind_async_flags) { + work->vma = vma; + work->cache_level = cache_level; + work->flags = bind_flags | I915_VMA_ALLOC; - vma->flags |= bind_flags; + /* + * Note we only want to chain up to the migration fence on + * the pages (not the object itself). As we don't track that, + * yet, we have to use the exclusive fence instead. + * + * Also note that we do not want to track the async vma as + * part of the obj->resv->excl_fence as it only affects + * execution and not content or object's backing store lifetime. + */ + GEM_BUG_ON(i915_active_has_exclusive(&vma->active)); + i915_active_set_exclusive(&vma->active, &work->base.dma); + work->base.dma.error = 0; /* enable the queue_work() */ + + if (vma->obj) + __i915_gem_object_pin_pages(vma->obj); + } else { + GEM_BUG_ON((bind_flags & ~vma_flags) & vma->vm->bind_async_flags); + ret = vma->ops->bind_vma(vma, cache_level, bind_flags); + if (ret) + return ret; + } + + atomic_or(bind_flags, &vma->flags); return 0; } @@ -350,18 +412,16 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma) int err; /* Access through the GTT requires the device to be awake. */ - assert_rpm_wakelock_held(&vma->vm->i915->runtime_pm); - - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); - if (WARN_ON(!i915_vma_is_map_and_fenceable(vma))) { + assert_rpm_wakelock_held(vma->vm->gt->uncore->rpm); + if (GEM_WARN_ON(!i915_vma_is_map_and_fenceable(vma))) { err = -ENODEV; goto err; } GEM_BUG_ON(!i915_vma_is_ggtt(vma)); - GEM_BUG_ON((vma->flags & I915_VMA_GLOBAL_BIND) == 0); + GEM_BUG_ON(!i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND)); - ptr = vma->iomap; + ptr = READ_ONCE(vma->iomap); if (ptr == NULL) { ptr = io_mapping_map_wc(&i915_vm_to_ggtt(vma->vm)->iomap, vma->node.start, @@ -371,7 +431,10 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma) goto err; } - vma->iomap = ptr; + if (unlikely(cmpxchg(&vma->iomap, NULL, ptr))) { + io_mapping_unmap(ptr); + ptr = vma->iomap; + } } __i915_vma_pin(vma); @@ -391,18 +454,12 @@ err: void i915_vma_flush_writes(struct i915_vma *vma) { - if (!i915_vma_has_ggtt_write(vma)) - return; - - intel_gt_flush_ggtt_writes(vma->vm->gt); - - i915_vma_unset_ggtt_write(vma); + if (i915_vma_unset_ggtt_write(vma)) + intel_gt_flush_ggtt_writes(vma->vm->gt); } void i915_vma_unpin_iomap(struct i915_vma *vma) { - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); - GEM_BUG_ON(vma->iomap == NULL); i915_vma_flush_writes(vma); @@ -438,6 +495,9 @@ bool i915_vma_misplaced(const struct i915_vma *vma, if (!drm_mm_node_allocated(&vma->node)) return false; + if (test_bit(I915_VMA_ERROR_BIT, __i915_vma_flags(vma))) + return true; + if (vma->node.size < size) return true; @@ -472,17 +532,12 @@ void __i915_vma_set_map_and_fenceable(struct i915_vma *vma) mappable = vma->node.start + vma->fence_size <= i915_vm_to_ggtt(vma->vm)->mappable_end; if (mappable && fenceable) - vma->flags |= I915_VMA_CAN_FENCE; + set_bit(I915_VMA_CAN_FENCE_BIT, __i915_vma_flags(vma)); else - vma->flags &= ~I915_VMA_CAN_FENCE; -} - -static bool color_differs(struct drm_mm_node *node, unsigned long color) -{ - return node->allocated && node->color != color; + clear_bit(I915_VMA_CAN_FENCE_BIT, __i915_vma_flags(vma)); } -bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long cache_level) +bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long color) { struct drm_mm_node *node = &vma->node; struct drm_mm_node *other; @@ -494,7 +549,7 @@ bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long cache_level) * these constraints apply and set the drm_mm.color_adjust * appropriately. */ - if (vma->vm->mm.color_adjust == NULL) + if (!i915_vm_has_cache_coloring(vma->vm)) return true; /* Only valid to be called on an already inserted vma */ @@ -502,11 +557,13 @@ bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long cache_level) GEM_BUG_ON(list_empty(&node->node_list)); other = list_prev_entry(node, node_list); - if (color_differs(other, cache_level) && !drm_mm_hole_follows(other)) + if (i915_node_color_differs(other, color) && + !drm_mm_hole_follows(other)) return false; other = list_next_entry(node, node_list); - if (color_differs(other, cache_level) && !drm_mm_hole_follows(node)) + if (i915_node_color_differs(other, color) && + !drm_mm_hole_follows(node)) return false; return true; @@ -541,13 +598,12 @@ static void assert_bind_count(const struct drm_i915_gem_object *obj) static int i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags) { - struct drm_i915_private *dev_priv = vma->vm->i915; - unsigned int cache_level; + unsigned long color; u64 start, end; int ret; GEM_BUG_ON(i915_vma_is_closed(vma)); - GEM_BUG_ON(vma->flags & (I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND)); + GEM_BUG_ON(i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND)); GEM_BUG_ON(drm_mm_node_allocated(&vma->node)); size = max(size, vma->size); @@ -567,7 +623,7 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags) end = vma->vm->total; if (flags & PIN_MAPPABLE) - end = min_t(u64, end, dev_priv->ggtt.mappable_end); + end = min_t(u64, end, i915_vm_to_ggtt(vma->vm)->mappable_end); if (flags & PIN_ZONE_4G) end = min_t(u64, end, (1ULL << 32) - I915_GTT_PAGE_SIZE); GEM_BUG_ON(!IS_ALIGNED(end, I915_GTT_PAGE_SIZE)); @@ -583,35 +639,21 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags) return -ENOSPC; } - if (vma->obj) { - ret = i915_gem_object_pin_pages(vma->obj); - if (ret) - return ret; - - cache_level = vma->obj->cache_level; - } else { - cache_level = 0; - } - - GEM_BUG_ON(vma->pages); - - ret = vma->ops->set_pages(vma); - if (ret) - goto err_unpin; + color = 0; + if (vma->obj && i915_vm_has_cache_coloring(vma->vm)) + color = vma->obj->cache_level; if (flags & PIN_OFFSET_FIXED) { u64 offset = flags & PIN_OFFSET_MASK; if (!IS_ALIGNED(offset, alignment) || - range_overflows(offset, size, end)) { - ret = -EINVAL; - goto err_clear; - } + range_overflows(offset, size, end)) + return -EINVAL; ret = i915_gem_gtt_reserve(vma->vm, &vma->node, - size, offset, cache_level, + size, offset, color, flags); if (ret) - goto err_clear; + return ret; } else { /* * We only support huge gtt pages through the 48b PPGTT, @@ -647,116 +689,259 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags) } ret = i915_gem_gtt_insert(vma->vm, &vma->node, - size, alignment, cache_level, + size, alignment, color, start, end, flags); if (ret) - goto err_clear; + return ret; GEM_BUG_ON(vma->node.start < start); GEM_BUG_ON(vma->node.start + vma->node.size > end); } GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); - GEM_BUG_ON(!i915_gem_valid_gtt_space(vma, cache_level)); - - mutex_lock(&vma->vm->mutex); - list_move_tail(&vma->vm_link, &vma->vm->bound_list); - mutex_unlock(&vma->vm->mutex); + GEM_BUG_ON(!i915_gem_valid_gtt_space(vma, color)); if (vma->obj) { - atomic_inc(&vma->obj->bind_count); - assert_bind_count(vma->obj); + struct drm_i915_gem_object *obj = vma->obj; + + atomic_inc(&obj->bind_count); + assert_bind_count(obj); } + list_add_tail(&vma->vm_link, &vma->vm->bound_list); return 0; - -err_clear: - vma->ops->clear_pages(vma); -err_unpin: - if (vma->obj) - i915_gem_object_unpin_pages(vma->obj); - return ret; } static void -i915_vma_remove(struct i915_vma *vma) +i915_vma_detach(struct i915_vma *vma) { GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); - GEM_BUG_ON(vma->flags & (I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND)); - - vma->ops->clear_pages(vma); - - mutex_lock(&vma->vm->mutex); - drm_mm_remove_node(&vma->node); - list_move_tail(&vma->vm_link, &vma->vm->unbound_list); - mutex_unlock(&vma->vm->mutex); + GEM_BUG_ON(i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND)); /* - * Since the unbound list is global, only move to that list if - * no more VMAs exist. + * And finally now the object is completely decoupled from this + * vma, we can drop its hold on the backing storage and allow + * it to be reaped by the shrinker. */ + list_del(&vma->vm_link); if (vma->obj) { struct drm_i915_gem_object *obj = vma->obj; + assert_bind_count(obj); atomic_dec(&obj->bind_count); + } +} - /* - * And finally now the object is completely decoupled from this - * vma, we can drop its hold on the backing storage and allow - * it to be reaped by the shrinker. - */ - i915_gem_object_unpin_pages(obj); - assert_bind_count(obj); +static bool try_qad_pin(struct i915_vma *vma, unsigned int flags) +{ + unsigned int bound; + bool pinned = true; + + bound = atomic_read(&vma->flags); + do { + if (unlikely(flags & ~bound)) + return false; + + if (unlikely(bound & (I915_VMA_OVERFLOW | I915_VMA_ERROR))) + return false; + + if (!(bound & I915_VMA_PIN_MASK)) + goto unpinned; + + GEM_BUG_ON(((bound + 1) & I915_VMA_PIN_MASK) == 0); + } while (!atomic_try_cmpxchg(&vma->flags, &bound, bound + 1)); + + return true; + +unpinned: + /* + * If pin_count==0, but we are bound, check under the lock to avoid + * racing with a concurrent i915_vma_unbind(). + */ + mutex_lock(&vma->vm->mutex); + do { + if (unlikely(bound & (I915_VMA_OVERFLOW | I915_VMA_ERROR))) { + pinned = false; + break; + } + + if (unlikely(flags & ~bound)) { + pinned = false; + break; + } + } while (!atomic_try_cmpxchg(&vma->flags, &bound, bound + 1)); + mutex_unlock(&vma->vm->mutex); + + return pinned; +} + +static int vma_get_pages(struct i915_vma *vma) +{ + int err = 0; + + if (atomic_add_unless(&vma->pages_count, 1, 0)) + return 0; + + /* Allocations ahoy! */ + if (mutex_lock_interruptible(&vma->pages_mutex)) + return -EINTR; + + if (!atomic_read(&vma->pages_count)) { + if (vma->obj) { + err = i915_gem_object_pin_pages(vma->obj); + if (err) + goto unlock; + } + + err = vma->ops->set_pages(vma); + if (err) { + if (vma->obj) + i915_gem_object_unpin_pages(vma->obj); + goto unlock; + } } + atomic_inc(&vma->pages_count); + +unlock: + mutex_unlock(&vma->pages_mutex); + + return err; } -int __i915_vma_do_pin(struct i915_vma *vma, - u64 size, u64 alignment, u64 flags) +static void __vma_put_pages(struct i915_vma *vma, unsigned int count) { - const unsigned int bound = vma->flags; - int ret; + /* We allocate under vma_get_pages, so beware the shrinker */ + mutex_lock_nested(&vma->pages_mutex, SINGLE_DEPTH_NESTING); + GEM_BUG_ON(atomic_read(&vma->pages_count) < count); + if (atomic_sub_return(count, &vma->pages_count) == 0) { + vma->ops->clear_pages(vma); + GEM_BUG_ON(vma->pages); + if (vma->obj) + i915_gem_object_unpin_pages(vma->obj); + } + mutex_unlock(&vma->pages_mutex); +} - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); - GEM_BUG_ON((flags & (PIN_GLOBAL | PIN_USER)) == 0); - GEM_BUG_ON((flags & PIN_GLOBAL) && !i915_vma_is_ggtt(vma)); +static void vma_put_pages(struct i915_vma *vma) +{ + if (atomic_add_unless(&vma->pages_count, -1, 1)) + return; - if (WARN_ON(bound & I915_VMA_PIN_OVERFLOW)) { - ret = -EBUSY; - goto err_unpin; + __vma_put_pages(vma, 1); +} + +static void vma_unbind_pages(struct i915_vma *vma) +{ + unsigned int count; + + lockdep_assert_held(&vma->vm->mutex); + + /* The upper portion of pages_count is the number of bindings */ + count = atomic_read(&vma->pages_count); + count >>= I915_VMA_PAGES_BIAS; + GEM_BUG_ON(!count); + + __vma_put_pages(vma, count | count << I915_VMA_PAGES_BIAS); +} + +int i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags) +{ + struct i915_vma_work *work = NULL; + unsigned int bound; + int err; + + BUILD_BUG_ON(PIN_GLOBAL != I915_VMA_GLOBAL_BIND); + BUILD_BUG_ON(PIN_USER != I915_VMA_LOCAL_BIND); + + GEM_BUG_ON(flags & PIN_UPDATE); + GEM_BUG_ON(!(flags & (PIN_USER | PIN_GLOBAL))); + + /* First try and grab the pin without rebinding the vma */ + if (try_qad_pin(vma, flags & I915_VMA_BIND_MASK)) + return 0; + + err = vma_get_pages(vma); + if (err) + return err; + + if (flags & vma->vm->bind_async_flags) { + work = i915_vma_work(); + if (!work) { + err = -ENOMEM; + goto err_pages; + } } - if ((bound & I915_VMA_BIND_MASK) == 0) { - ret = i915_vma_insert(vma, size, alignment, flags); - if (ret) - goto err_unpin; + /* No more allocations allowed once we hold vm->mutex */ + err = mutex_lock_interruptible(&vma->vm->mutex); + if (err) + goto err_fence; + + bound = atomic_read(&vma->flags); + if (unlikely(bound & I915_VMA_ERROR)) { + err = -ENOMEM; + goto err_unlock; } - GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); - ret = i915_vma_bind(vma, vma->obj ? vma->obj->cache_level : 0, flags); - if (ret) - goto err_remove; + if (unlikely(!((bound + 1) & I915_VMA_PIN_MASK))) { + err = -EAGAIN; /* pins are meant to be fairly temporary */ + goto err_unlock; + } - GEM_BUG_ON((vma->flags & I915_VMA_BIND_MASK) == 0); + if (unlikely(!(flags & ~bound & I915_VMA_BIND_MASK))) { + __i915_vma_pin(vma); + goto err_unlock; + } - if ((bound ^ vma->flags) & I915_VMA_GLOBAL_BIND) - __i915_vma_set_map_and_fenceable(vma); + err = i915_active_acquire(&vma->active); + if (err) + goto err_unlock; + + if (!(bound & I915_VMA_BIND_MASK)) { + err = i915_vma_insert(vma, size, alignment, flags); + if (err) + goto err_active; + + if (i915_is_ggtt(vma->vm)) + __i915_vma_set_map_and_fenceable(vma); + } + + GEM_BUG_ON(!vma->pages); + err = i915_vma_bind(vma, + vma->obj ? vma->obj->cache_level : 0, + flags, work); + if (err) + goto err_remove; + + /* There should only be at most 2 active bindings (user, global) */ + GEM_BUG_ON(bound + I915_VMA_PAGES_ACTIVE < bound); + atomic_add(I915_VMA_PAGES_ACTIVE, &vma->pages_count); + list_move_tail(&vma->vm_link, &vma->vm->bound_list); + __i915_vma_pin(vma); + GEM_BUG_ON(!i915_vma_is_pinned(vma)); + GEM_BUG_ON(!i915_vma_is_bound(vma, flags)); GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags)); - return 0; err_remove: - if ((bound & I915_VMA_BIND_MASK) == 0) { - i915_vma_remove(vma); - GEM_BUG_ON(vma->pages); - GEM_BUG_ON(vma->flags & I915_VMA_BIND_MASK); + if (!i915_vma_is_bound(vma, I915_VMA_BIND_MASK)) { + i915_vma_detach(vma); + drm_mm_remove_node(&vma->node); } -err_unpin: - __i915_vma_unpin(vma); - return ret; +err_active: + i915_active_release(&vma->active); +err_unlock: + mutex_unlock(&vma->vm->mutex); +err_fence: + if (work) + dma_fence_work_commit(&work->base); +err_pages: + vma_put_pages(vma); + return err; } void i915_vma_close(struct i915_vma *vma) { - struct drm_i915_private *i915 = vma->vm->i915; + struct intel_gt *gt = vma->vm->gt; unsigned long flags; GEM_BUG_ON(i915_vma_is_closed(vma)); @@ -773,79 +958,87 @@ void i915_vma_close(struct i915_vma *vma) * causing us to rebind the VMA once more. This ends up being a lot * of wasted work for the steady state. */ - spin_lock_irqsave(&i915->gt.closed_lock, flags); - list_add(&vma->closed_link, &i915->gt.closed_vma); - spin_unlock_irqrestore(&i915->gt.closed_lock, flags); + spin_lock_irqsave(>->closed_lock, flags); + list_add(&vma->closed_link, >->closed_vma); + spin_unlock_irqrestore(>->closed_lock, flags); } static void __i915_vma_remove_closed(struct i915_vma *vma) { - struct drm_i915_private *i915 = vma->vm->i915; - - if (!i915_vma_is_closed(vma)) - return; + struct intel_gt *gt = vma->vm->gt; - spin_lock_irq(&i915->gt.closed_lock); + spin_lock_irq(>->closed_lock); list_del_init(&vma->closed_link); - spin_unlock_irq(&i915->gt.closed_lock); + spin_unlock_irq(>->closed_lock); } void i915_vma_reopen(struct i915_vma *vma) { - __i915_vma_remove_closed(vma); + if (i915_vma_is_closed(vma)) + __i915_vma_remove_closed(vma); } -static void __i915_vma_destroy(struct i915_vma *vma) +void i915_vma_destroy(struct i915_vma *vma) { - GEM_BUG_ON(vma->node.allocated); - GEM_BUG_ON(vma->fence); - - mutex_lock(&vma->vm->mutex); - list_del(&vma->vm_link); - mutex_unlock(&vma->vm->mutex); + if (drm_mm_node_allocated(&vma->node)) { + mutex_lock(&vma->vm->mutex); + atomic_and(~I915_VMA_PIN_MASK, &vma->flags); + WARN_ON(__i915_vma_unbind(vma)); + mutex_unlock(&vma->vm->mutex); + GEM_BUG_ON(drm_mm_node_allocated(&vma->node)); + } + GEM_BUG_ON(i915_vma_is_active(vma)); if (vma->obj) { struct drm_i915_gem_object *obj = vma->obj; spin_lock(&obj->vma.lock); list_del(&vma->obj_link); - rb_erase(&vma->obj_node, &vma->obj->vma.tree); + rb_erase(&vma->obj_node, &obj->vma.tree); spin_unlock(&obj->vma.lock); } - i915_active_fini(&vma->active); + __i915_vma_remove_closed(vma); + i915_vm_put(vma->vm); + i915_active_fini(&vma->active); i915_vma_free(vma); } -void i915_vma_destroy(struct i915_vma *vma) +void i915_vma_parked(struct intel_gt *gt) { - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); + struct i915_vma *vma, *next; - GEM_BUG_ON(i915_vma_is_pinned(vma)); + spin_lock_irq(>->closed_lock); + list_for_each_entry_safe(vma, next, >->closed_vma, closed_link) { + struct drm_i915_gem_object *obj = vma->obj; + struct i915_address_space *vm = vma->vm; - __i915_vma_remove_closed(vma); + /* XXX All to avoid keeping a reference on i915_vma itself */ - WARN_ON(i915_vma_unbind(vma)); - GEM_BUG_ON(i915_vma_is_active(vma)); + if (!kref_get_unless_zero(&obj->base.refcount)) + continue; - __i915_vma_destroy(vma); -} + if (!i915_vm_tryopen(vm)) { + i915_gem_object_put(obj); + obj = NULL; + } -void i915_vma_parked(struct drm_i915_private *i915) -{ - struct i915_vma *vma, *next; + spin_unlock_irq(>->closed_lock); - spin_lock_irq(&i915->gt.closed_lock); - list_for_each_entry_safe(vma, next, &i915->gt.closed_vma, closed_link) { - list_del_init(&vma->closed_link); - spin_unlock_irq(&i915->gt.closed_lock); + if (obj) { + i915_vma_destroy(vma); + i915_gem_object_put(obj); + } - i915_vma_destroy(vma); + i915_vm_close(vm); - spin_lock_irq(&i915->gt.closed_lock); + /* Restart after dropping lock */ + spin_lock_irq(>->closed_lock); + next = list_first_entry(>->closed_vma, + typeof(*next), closed_link); } - spin_unlock_irq(&i915->gt.closed_lock); + spin_unlock_irq(>->closed_lock); } static void __i915_vma_iounmap(struct i915_vma *vma) @@ -883,6 +1076,20 @@ void i915_vma_revoke_mmap(struct i915_vma *vma) list_del(&vma->obj->userfault_link); } +int __i915_vma_move_to_active(struct i915_vma *vma, struct i915_request *rq) +{ + int err; + + GEM_BUG_ON(!i915_vma_is_pinned(vma)); + + /* Wait for the vma to be bound before we start! */ + err = i915_request_await_active(rq, &vma->active); + if (err) + return err; + + return i915_active_add_request(&vma->active, rq); +} + int i915_vma_move_to_active(struct i915_vma *vma, struct i915_request *rq, unsigned int flags) @@ -890,27 +1097,15 @@ int i915_vma_move_to_active(struct i915_vma *vma, struct drm_i915_gem_object *obj = vma->obj; int err; - assert_vma_held(vma); assert_object_held(obj); - GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); - /* - * Add a reference if we're newly entering the active list. - * The order in which we add operations to the retirement queue is - * vital here: mark_active adds to the start of the callback list, - * such that subsequent callbacks are called first. Therefore we - * add the active reference first and queue for it to be dropped - * *last*. - */ - err = i915_active_ref(&vma->active, rq->timeline, rq); + err = __i915_vma_move_to_active(vma, rq); if (unlikely(err)) return err; if (flags & EXEC_OBJECT_WRITE) { if (intel_frontbuffer_invalidate(obj->frontbuffer, ORIGIN_CS)) - i915_active_ref(&obj->frontbuffer->write, - rq->timeline, - rq); + i915_active_add_request(&obj->frontbuffer->write, rq); dma_resv_add_excl_fence(vma->resv, &rq->fence); obj->write_domain = I915_GEM_DOMAIN_RENDER; @@ -930,44 +1125,31 @@ int i915_vma_move_to_active(struct i915_vma *vma, return 0; } -int i915_vma_unbind(struct i915_vma *vma) +int __i915_vma_unbind(struct i915_vma *vma) { int ret; - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); + lockdep_assert_held(&vma->vm->mutex); /* * First wait upon any activity as retiring the request may * have side-effects such as unpinning or even unbinding this vma. + * + * XXX Actually waiting under the vm->mutex is a hinderance and + * should be pipelined wherever possible. In cases where that is + * unavoidable, we should lift the wait to before the mutex. */ - might_sleep(); - if (i915_vma_is_active(vma)) { - /* - * When a closed VMA is retired, it is unbound - eek. - * In order to prevent it from being recursively closed, - * take a pin on the vma so that the second unbind is - * aborted. - * - * Even more scary is that the retire callback may free - * the object (last active vma). To prevent the explosion - * we defer the actual object free to a worker that can - * only proceed once it acquires the struct_mutex (which - * we currently hold, therefore it cannot free this object - * before we are finished). - */ - __i915_vma_pin(vma); - ret = i915_active_wait(&vma->active); - __i915_vma_unpin(vma); - if (ret) - return ret; - } - GEM_BUG_ON(i915_vma_is_active(vma)); + ret = i915_vma_sync(vma); + if (ret) + return ret; + GEM_BUG_ON(i915_vma_is_active(vma)); if (i915_vma_is_pinned(vma)) { vma_print_allocator(vma, "is pinned"); return -EBUSY; } + GEM_BUG_ON(i915_vma_is_active(vma)); if (!drm_mm_node_allocated(&vma->node)) return 0; @@ -982,34 +1164,47 @@ int i915_vma_unbind(struct i915_vma *vma) GEM_BUG_ON(i915_vma_has_ggtt_write(vma)); /* release the fence reg _after_ flushing */ - mutex_lock(&vma->vm->mutex); ret = i915_vma_revoke_fence(vma); - mutex_unlock(&vma->vm->mutex); if (ret) return ret; /* Force a pagefault for domain tracking on next user access */ - mutex_lock(&vma->vm->mutex); i915_vma_revoke_mmap(vma); - mutex_unlock(&vma->vm->mutex); __i915_vma_iounmap(vma); - vma->flags &= ~I915_VMA_CAN_FENCE; + clear_bit(I915_VMA_CAN_FENCE_BIT, __i915_vma_flags(vma)); } GEM_BUG_ON(vma->fence); GEM_BUG_ON(i915_vma_has_userfault(vma)); - if (likely(!vma->vm->closed)) { + if (likely(atomic_read(&vma->vm->open))) { trace_i915_vma_unbind(vma); vma->ops->unbind_vma(vma); } - vma->flags &= ~(I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND); + atomic_and(~(I915_VMA_BIND_MASK | I915_VMA_ERROR), &vma->flags); - i915_vma_remove(vma); + i915_vma_detach(vma); + vma_unbind_pages(vma); + drm_mm_remove_node(&vma->node); /* pairs with i915_vma_destroy() */ return 0; } +int i915_vma_unbind(struct i915_vma *vma) +{ + struct i915_address_space *vm = vma->vm; + int err; + + err = mutex_lock_interruptible(&vm->mutex); + if (err) + return err; + + err = __i915_vma_unbind(vma); + mutex_unlock(&vm->mutex); + + return err; +} + struct i915_vma *i915_vma_make_unshrinkable(struct i915_vma *vma) { i915_gem_object_make_unshrinkable(vma->obj); diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h index 889fc7cb910a..465932813bc5 100644 --- a/drivers/gpu/drm/i915/i915_vma.h +++ b/drivers/gpu/drm/i915/i915_vma.h @@ -72,7 +72,7 @@ struct i915_vma { * that exist in the ctx->handle_vmas LUT for this vma. */ atomic_t open_count; - unsigned long flags; + atomic_t flags; /** * How many users have pinned this object in GTT space. * @@ -96,22 +96,41 @@ struct i915_vma { * exclusive cachelines of a single page, so a maximum of 64 possible * users. */ -#define I915_VMA_PIN_MASK 0xff -#define I915_VMA_PIN_OVERFLOW BIT(8) +#define I915_VMA_PIN_MASK 0x3ff +#define I915_VMA_OVERFLOW 0x200 /** Flags and address space this VMA is bound to */ -#define I915_VMA_GLOBAL_BIND BIT(9) -#define I915_VMA_LOCAL_BIND BIT(10) -#define I915_VMA_BIND_MASK (I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND | I915_VMA_PIN_OVERFLOW) +#define I915_VMA_GLOBAL_BIND_BIT 10 +#define I915_VMA_LOCAL_BIND_BIT 11 -#define I915_VMA_GGTT BIT(11) -#define I915_VMA_CAN_FENCE BIT(12) -#define I915_VMA_USERFAULT_BIT 13 -#define I915_VMA_USERFAULT BIT(I915_VMA_USERFAULT_BIT) -#define I915_VMA_GGTT_WRITE BIT(14) +#define I915_VMA_GLOBAL_BIND ((int)BIT(I915_VMA_GLOBAL_BIND_BIT)) +#define I915_VMA_LOCAL_BIND ((int)BIT(I915_VMA_LOCAL_BIND_BIT)) + +#define I915_VMA_BIND_MASK (I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND) + +#define I915_VMA_ALLOC_BIT 12 +#define I915_VMA_ALLOC ((int)BIT(I915_VMA_ALLOC_BIT)) + +#define I915_VMA_ERROR_BIT 13 +#define I915_VMA_ERROR ((int)BIT(I915_VMA_ERROR_BIT)) + +#define I915_VMA_GGTT_BIT 14 +#define I915_VMA_CAN_FENCE_BIT 15 +#define I915_VMA_USERFAULT_BIT 16 +#define I915_VMA_GGTT_WRITE_BIT 17 + +#define I915_VMA_GGTT ((int)BIT(I915_VMA_GGTT_BIT)) +#define I915_VMA_CAN_FENCE ((int)BIT(I915_VMA_CAN_FENCE_BIT)) +#define I915_VMA_USERFAULT ((int)BIT(I915_VMA_USERFAULT_BIT)) +#define I915_VMA_GGTT_WRITE ((int)BIT(I915_VMA_GGTT_WRITE_BIT)) struct i915_active active; +#define I915_VMA_PAGES_BIAS 24 +#define I915_VMA_PAGES_ACTIVE (BIT(24) | 1) + atomic_t pages_count; /* number of active binds to the pages */ + struct mutex pages_mutex; /* protect acquire/release of backing pages */ + /** * Support different GGTT views into the same object. * This means there can be multiple VMA mappings per object and per VM. @@ -158,52 +177,57 @@ static inline bool i915_vma_is_active(const struct i915_vma *vma) return !i915_active_is_idle(&vma->active); } +int __must_check __i915_vma_move_to_active(struct i915_vma *vma, + struct i915_request *rq); int __must_check i915_vma_move_to_active(struct i915_vma *vma, struct i915_request *rq, unsigned int flags); +#define __i915_vma_flags(v) ((unsigned long *)&(v)->flags.counter) + static inline bool i915_vma_is_ggtt(const struct i915_vma *vma) { - return vma->flags & I915_VMA_GGTT; + return test_bit(I915_VMA_GGTT_BIT, __i915_vma_flags(vma)); } static inline bool i915_vma_has_ggtt_write(const struct i915_vma *vma) { - return vma->flags & I915_VMA_GGTT_WRITE; + return test_bit(I915_VMA_GGTT_WRITE_BIT, __i915_vma_flags(vma)); } static inline void i915_vma_set_ggtt_write(struct i915_vma *vma) { GEM_BUG_ON(!i915_vma_is_ggtt(vma)); - vma->flags |= I915_VMA_GGTT_WRITE; + set_bit(I915_VMA_GGTT_WRITE_BIT, __i915_vma_flags(vma)); } -static inline void i915_vma_unset_ggtt_write(struct i915_vma *vma) +static inline bool i915_vma_unset_ggtt_write(struct i915_vma *vma) { - vma->flags &= ~I915_VMA_GGTT_WRITE; + return test_and_clear_bit(I915_VMA_GGTT_WRITE_BIT, + __i915_vma_flags(vma)); } void i915_vma_flush_writes(struct i915_vma *vma); static inline bool i915_vma_is_map_and_fenceable(const struct i915_vma *vma) { - return vma->flags & I915_VMA_CAN_FENCE; + return test_bit(I915_VMA_CAN_FENCE_BIT, __i915_vma_flags(vma)); } static inline bool i915_vma_set_userfault(struct i915_vma *vma) { GEM_BUG_ON(!i915_vma_is_map_and_fenceable(vma)); - return __test_and_set_bit(I915_VMA_USERFAULT_BIT, &vma->flags); + return test_and_set_bit(I915_VMA_USERFAULT_BIT, __i915_vma_flags(vma)); } static inline void i915_vma_unset_userfault(struct i915_vma *vma) { - return __clear_bit(I915_VMA_USERFAULT_BIT, &vma->flags); + return clear_bit(I915_VMA_USERFAULT_BIT, __i915_vma_flags(vma)); } static inline bool i915_vma_has_userfault(const struct i915_vma *vma) { - return test_bit(I915_VMA_USERFAULT_BIT, &vma->flags); + return test_bit(I915_VMA_USERFAULT_BIT, __i915_vma_flags(vma)); } static inline bool i915_vma_is_closed(const struct i915_vma *vma) @@ -214,7 +238,7 @@ static inline bool i915_vma_is_closed(const struct i915_vma *vma) static inline u32 i915_ggtt_offset(const struct i915_vma *vma) { GEM_BUG_ON(!i915_vma_is_ggtt(vma)); - GEM_BUG_ON(!vma->node.allocated); + GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); GEM_BUG_ON(upper_32_bits(vma->node.start)); GEM_BUG_ON(upper_32_bits(vma->node.start + vma->node.size - 1)); return lower_32_bits(vma->node.start); @@ -293,13 +317,18 @@ i915_vma_compare(struct i915_vma *vma, return memcmp(&vma->ggtt_view.partial, &view->partial, view->type); } -int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level, - u32 flags); -bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long cache_level); +struct i915_vma_work *i915_vma_work(void); +int i915_vma_bind(struct i915_vma *vma, + enum i915_cache_level cache_level, + u32 flags, + struct i915_vma_work *work); + +bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long color); bool i915_vma_misplaced(const struct i915_vma *vma, u64 size, u64 alignment, u64 flags); void __i915_vma_set_map_and_fenceable(struct i915_vma *vma); void i915_vma_revoke_mmap(struct i915_vma *vma); +int __i915_vma_unbind(struct i915_vma *vma); int __must_check i915_vma_unbind(struct i915_vma *vma); void i915_vma_unlink_ctx(struct i915_vma *vma); void i915_vma_close(struct i915_vma *vma); @@ -318,30 +347,12 @@ static inline void i915_vma_unlock(struct i915_vma *vma) dma_resv_unlock(vma->resv); } -int __i915_vma_do_pin(struct i915_vma *vma, - u64 size, u64 alignment, u64 flags); -static inline int __must_check -i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags) -{ - BUILD_BUG_ON(PIN_MBZ != I915_VMA_PIN_OVERFLOW); - BUILD_BUG_ON(PIN_GLOBAL != I915_VMA_GLOBAL_BIND); - BUILD_BUG_ON(PIN_USER != I915_VMA_LOCAL_BIND); - - /* Pin early to prevent the shrinker/eviction logic from destroying - * our vma as we insert and bind. - */ - if (likely(((++vma->flags ^ flags) & I915_VMA_BIND_MASK) == 0)) { - GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); - GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags)); - return 0; - } - - return __i915_vma_do_pin(vma, size, alignment, flags); -} +int __must_check +i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags); static inline int i915_vma_pin_count(const struct i915_vma *vma) { - return vma->flags & I915_VMA_PIN_MASK; + return atomic_read(&vma->flags) & I915_VMA_PIN_MASK; } static inline bool i915_vma_is_pinned(const struct i915_vma *vma) @@ -351,18 +362,18 @@ static inline bool i915_vma_is_pinned(const struct i915_vma *vma) static inline void __i915_vma_pin(struct i915_vma *vma) { - vma->flags++; - GEM_BUG_ON(vma->flags & I915_VMA_PIN_OVERFLOW); + atomic_inc(&vma->flags); + GEM_BUG_ON(!i915_vma_is_pinned(vma)); } static inline void __i915_vma_unpin(struct i915_vma *vma) { - vma->flags--; + GEM_BUG_ON(!i915_vma_is_pinned(vma)); + atomic_dec(&vma->flags); } static inline void i915_vma_unpin(struct i915_vma *vma) { - GEM_BUG_ON(!i915_vma_is_pinned(vma)); GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); __i915_vma_unpin(vma); } @@ -370,7 +381,13 @@ static inline void i915_vma_unpin(struct i915_vma *vma) static inline bool i915_vma_is_bound(const struct i915_vma *vma, unsigned int where) { - return vma->flags & where; + return atomic_read(&vma->flags) & where; +} + +static inline bool i915_node_color_differs(const struct drm_mm_node *node, + unsigned long color) +{ + return drm_mm_node_allocated(node) && node->color != color; } /** @@ -382,8 +399,6 @@ static inline bool i915_vma_is_bound(const struct i915_vma *vma, * the caller must call i915_vma_unpin_iomap to relinquish the pinning * after the iomapping is no longer required. * - * Callers must hold the struct_mutex. - * * Returns a valid iomapped pointer or ERR_PTR. */ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma); @@ -395,8 +410,8 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma); * * Unpins the previously iomapped VMA from i915_vma_pin_iomap(). * - * Callers must hold the struct_mutex. This function is only valid to be - * called on a VMA previously iomapped by the caller with i915_vma_pin_iomap(). + * This function is only valid to be called on a VMA previously + * iomapped by the caller with i915_vma_pin_iomap(). */ void i915_vma_unpin_iomap(struct i915_vma *vma); @@ -424,6 +439,8 @@ static inline struct page *i915_vma_first_page(struct i915_vma *vma) int __must_check i915_vma_pin_fence(struct i915_vma *vma); int __must_check i915_vma_revoke_fence(struct i915_vma *vma); +int __i915_vma_pin_fence(struct i915_vma *vma); + static inline void __i915_vma_unpin_fence(struct i915_vma *vma) { GEM_BUG_ON(atomic_read(&vma->fence->pin_count) <= 0); @@ -441,12 +458,11 @@ static inline void __i915_vma_unpin_fence(struct i915_vma *vma) static inline void i915_vma_unpin_fence(struct i915_vma *vma) { - /* lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); */ if (vma->fence) __i915_vma_unpin_fence(vma); } -void i915_vma_parked(struct drm_i915_private *i915); +void i915_vma_parked(struct intel_gt *gt); #define for_each_until(cond) if (cond) break; else @@ -470,4 +486,10 @@ struct i915_vma *i915_vma_make_unshrinkable(struct i915_vma *vma); void i915_vma_make_shrinkable(struct i915_vma *vma); void i915_vma_make_purgeable(struct i915_vma *vma); +static inline int i915_vma_sync(struct i915_vma *vma) +{ + /* Wait for the asynchronous bindings and pending GPU reads */ + return i915_active_wait(&vma->active); +} + #endif diff --git a/drivers/gpu/drm/i915/intel_csr.c b/drivers/gpu/drm/i915/intel_csr.c index 546577e39b4e..09870a31b4f0 100644 --- a/drivers/gpu/drm/i915/intel_csr.c +++ b/drivers/gpu/drm/i915/intel_csr.c @@ -44,8 +44,8 @@ #define TGL_CSR_MAX_FW_SIZE 0x6000 MODULE_FIRMWARE(TGL_CSR_PATH); -#define ICL_CSR_PATH "i915/icl_dmc_ver1_07.bin" -#define ICL_CSR_VERSION_REQUIRED CSR_VERSION(1, 7) +#define ICL_CSR_PATH "i915/icl_dmc_ver1_09.bin" +#define ICL_CSR_VERSION_REQUIRED CSR_VERSION(1, 9) #define ICL_CSR_MAX_FW_SIZE 0x6000 MODULE_FIRMWARE(ICL_CSR_PATH); diff --git a/drivers/gpu/drm/i915/intel_device_info.c b/drivers/gpu/drm/i915/intel_device_info.c index d0ed44d33484..a5b571364cf6 100644 --- a/drivers/gpu/drm/i915/intel_device_info.c +++ b/drivers/gpu/drm/i915/intel_device_info.c @@ -93,9 +93,9 @@ static void sseu_dump(const struct sseu_dev_info *sseu, struct drm_printer *p) hweight8(sseu->slice_mask), sseu->slice_mask); drm_printf(p, "subslice total: %u\n", intel_sseu_subslice_total(sseu)); for (s = 0; s < sseu->max_slices; s++) { - drm_printf(p, "slice%d: %u subslices, mask=%04x\n", + drm_printf(p, "slice%d: %u subslices, mask=%08x\n", s, intel_sseu_subslices_per_slice(sseu, s), - sseu->subslice_mask[s]); + intel_sseu_get_subslices(sseu, s)); } drm_printf(p, "EU total: %u\n", sseu->eu_total); drm_printf(p, "EU per subslice: %u\n", sseu->eu_per_subslice); @@ -118,10 +118,9 @@ void intel_device_info_dump_runtime(const struct intel_runtime_info *info, static int sseu_eu_idx(const struct sseu_dev_info *sseu, int slice, int subslice) { - int subslice_stride = GEN_SSEU_STRIDE(sseu->max_eus_per_subslice); - int slice_stride = sseu->max_subslices * subslice_stride; + int slice_stride = sseu->max_subslices * sseu->eu_stride; - return slice * slice_stride + subslice * subslice_stride; + return slice * slice_stride + subslice * sseu->eu_stride; } static u16 sseu_get_eus(const struct sseu_dev_info *sseu, int slice, @@ -130,7 +129,7 @@ static u16 sseu_get_eus(const struct sseu_dev_info *sseu, int slice, int i, offset = sseu_eu_idx(sseu, slice, subslice); u16 eu_mask = 0; - for (i = 0; i < GEN_SSEU_STRIDE(sseu->max_eus_per_subslice); i++) { + for (i = 0; i < sseu->eu_stride; i++) { eu_mask |= ((u16)sseu->eu_mask[offset + i]) << (i * BITS_PER_BYTE); } @@ -143,7 +142,7 @@ static void sseu_set_eus(struct sseu_dev_info *sseu, int slice, int subslice, { int i, offset = sseu_eu_idx(sseu, slice, subslice); - for (i = 0; i < GEN_SSEU_STRIDE(sseu->max_eus_per_subslice); i++) { + for (i = 0; i < sseu->eu_stride; i++) { sseu->eu_mask[offset + i] = (eu_mask >> (BITS_PER_BYTE * i)) & 0xff; } @@ -160,9 +159,9 @@ void intel_device_info_dump_topology(const struct sseu_dev_info *sseu, } for (s = 0; s < sseu->max_slices; s++) { - drm_printf(p, "slice%d: %u subslice(s) (0x%hhx):\n", + drm_printf(p, "slice%d: %u subslice(s) (0x%08x):\n", s, intel_sseu_subslices_per_slice(sseu, s), - sseu->subslice_mask[s]); + intel_sseu_get_subslices(sseu, s)); for (ss = 0; ss < sseu->max_subslices; ss++) { u16 enabled_eus = sseu_get_eus(sseu, s, ss); @@ -183,44 +182,80 @@ static u16 compute_eu_total(const struct sseu_dev_info *sseu) return total; } +static void gen11_compute_sseu_info(struct sseu_dev_info *sseu, + u8 s_en, u32 ss_en, u16 eu_en) +{ + int s, ss; + + /* ss_en represents entire subslice mask across all slices */ + GEM_BUG_ON(sseu->max_slices * sseu->max_subslices > + sizeof(ss_en) * BITS_PER_BYTE); + + for (s = 0; s < sseu->max_slices; s++) { + if ((s_en & BIT(s)) == 0) + continue; + + sseu->slice_mask |= BIT(s); + + intel_sseu_set_subslices(sseu, s, ss_en); + + for (ss = 0; ss < sseu->max_subslices; ss++) + if (intel_sseu_has_subslice(sseu, s, ss)) + sseu_set_eus(sseu, s, ss, eu_en); + } + sseu->eu_per_subslice = hweight16(eu_en); + sseu->eu_total = compute_eu_total(sseu); +} + +static void gen12_sseu_info_init(struct drm_i915_private *dev_priv) +{ + struct sseu_dev_info *sseu = &RUNTIME_INFO(dev_priv)->sseu; + u8 s_en; + u32 dss_en; + u16 eu_en = 0; + u8 eu_en_fuse; + int eu; + + /* + * Gen12 has Dual-Subslices, which behave similarly to 2 gen11 SS. + * Instead of splitting these, provide userspace with an array + * of DSS to more closely represent the hardware resource. + */ + intel_sseu_set_info(sseu, 1, 6, 16); + + s_en = I915_READ(GEN11_GT_SLICE_ENABLE) & GEN11_GT_S_ENA_MASK; + + dss_en = I915_READ(GEN12_GT_DSS_ENABLE); + + /* one bit per pair of EUs */ + eu_en_fuse = ~(I915_READ(GEN11_EU_DISABLE) & GEN11_EU_DIS_MASK); + for (eu = 0; eu < sseu->max_eus_per_subslice / 2; eu++) + if (eu_en_fuse & BIT(eu)) + eu_en |= BIT(eu * 2) | BIT(eu * 2 + 1); + + gen11_compute_sseu_info(sseu, s_en, dss_en, eu_en); + + /* TGL only supports slice-level power gating */ + sseu->has_slice_pg = 1; +} + static void gen11_sseu_info_init(struct drm_i915_private *dev_priv) { struct sseu_dev_info *sseu = &RUNTIME_INFO(dev_priv)->sseu; u8 s_en; - u32 ss_en, ss_en_mask; + u32 ss_en; u8 eu_en; - int s; - if (IS_ELKHARTLAKE(dev_priv)) { - sseu->max_slices = 1; - sseu->max_subslices = 4; - sseu->max_eus_per_subslice = 8; - } else { - sseu->max_slices = 1; - sseu->max_subslices = 8; - sseu->max_eus_per_subslice = 8; - } + if (IS_ELKHARTLAKE(dev_priv)) + intel_sseu_set_info(sseu, 1, 4, 8); + else + intel_sseu_set_info(sseu, 1, 8, 8); s_en = I915_READ(GEN11_GT_SLICE_ENABLE) & GEN11_GT_S_ENA_MASK; ss_en = ~I915_READ(GEN11_GT_SUBSLICE_DISABLE); - ss_en_mask = BIT(sseu->max_subslices) - 1; eu_en = ~(I915_READ(GEN11_EU_DISABLE) & GEN11_EU_DIS_MASK); - for (s = 0; s < sseu->max_slices; s++) { - if (s_en & BIT(s)) { - int ss_idx = sseu->max_subslices * s; - int ss; - - sseu->slice_mask |= BIT(s); - sseu->subslice_mask[s] = (ss_en >> ss_idx) & ss_en_mask; - for (ss = 0; ss < sseu->max_subslices; ss++) { - if (sseu->subslice_mask[s] & BIT(ss)) - sseu_set_eus(sseu, s, ss, eu_en); - } - } - } - sseu->eu_per_subslice = hweight8(eu_en); - sseu->eu_total = compute_eu_total(sseu); + gen11_compute_sseu_info(sseu, s_en, ss_en, eu_en); /* ICL has no power gating restrictions. */ sseu->has_slice_pg = 1; @@ -236,23 +271,10 @@ static void gen10_sseu_info_init(struct drm_i915_private *dev_priv) const int eu_mask = 0xff; u32 subslice_mask, eu_en; + intel_sseu_set_info(sseu, 6, 4, 8); + sseu->slice_mask = (fuse2 & GEN10_F2_S_ENA_MASK) >> GEN10_F2_S_ENA_SHIFT; - sseu->max_slices = 6; - sseu->max_subslices = 4; - sseu->max_eus_per_subslice = 8; - - subslice_mask = (1 << 4) - 1; - subslice_mask &= ~((fuse2 & GEN10_F2_SS_DIS_MASK) >> - GEN10_F2_SS_DIS_SHIFT); - - /* - * Slice0 can have up to 3 subslices, but there are only 2 in - * slice1/2. - */ - sseu->subslice_mask[0] = subslice_mask; - for (s = 1; s < sseu->max_slices; s++) - sseu->subslice_mask[s] = subslice_mask & 0x3; /* Slice0 */ eu_en = ~I915_READ(GEN8_EU_DISABLE0); @@ -277,14 +299,25 @@ static void gen10_sseu_info_init(struct drm_i915_private *dev_priv) eu_en = ~I915_READ(GEN10_EU_DISABLE3); sseu_set_eus(sseu, 5, 1, eu_en & eu_mask); - /* Do a second pass where we mark the subslices disabled if all their - * eus are off. - */ + subslice_mask = (1 << 4) - 1; + subslice_mask &= ~((fuse2 & GEN10_F2_SS_DIS_MASK) >> + GEN10_F2_SS_DIS_SHIFT); + for (s = 0; s < sseu->max_slices; s++) { + u32 subslice_mask_with_eus = subslice_mask; + for (ss = 0; ss < sseu->max_subslices; ss++) { if (sseu_get_eus(sseu, s, ss) == 0) - sseu->subslice_mask[s] &= ~BIT(ss); + subslice_mask_with_eus &= ~BIT(ss); } + + /* + * Slice0 can have up to 3 subslices, but there are only 2 in + * slice1/2. + */ + intel_sseu_set_subslices(sseu, s, s == 0 ? + subslice_mask_with_eus : + subslice_mask_with_eus & 0x3); } sseu->eu_total = compute_eu_total(sseu); @@ -310,13 +343,12 @@ static void cherryview_sseu_info_init(struct drm_i915_private *dev_priv) { struct sseu_dev_info *sseu = &RUNTIME_INFO(dev_priv)->sseu; u32 fuse; + u8 subslice_mask = 0; fuse = I915_READ(CHV_FUSE_GT); sseu->slice_mask = BIT(0); - sseu->max_slices = 1; - sseu->max_subslices = 2; - sseu->max_eus_per_subslice = 8; + intel_sseu_set_info(sseu, 1, 2, 8); if (!(fuse & CHV_FGT_DISABLE_SS0)) { u8 disabled_mask = @@ -325,7 +357,7 @@ static void cherryview_sseu_info_init(struct drm_i915_private *dev_priv) (((fuse & CHV_FGT_EU_DIS_SS0_R1_MASK) >> CHV_FGT_EU_DIS_SS0_R1_SHIFT) << 4); - sseu->subslice_mask[0] |= BIT(0); + subslice_mask |= BIT(0); sseu_set_eus(sseu, 0, 0, ~disabled_mask); } @@ -336,10 +368,12 @@ static void cherryview_sseu_info_init(struct drm_i915_private *dev_priv) (((fuse & CHV_FGT_EU_DIS_SS1_R1_MASK) >> CHV_FGT_EU_DIS_SS1_R1_SHIFT) << 4); - sseu->subslice_mask[0] |= BIT(1); + subslice_mask |= BIT(1); sseu_set_eus(sseu, 0, 1, ~disabled_mask); } + intel_sseu_set_subslices(sseu, 0, subslice_mask); + sseu->eu_total = compute_eu_total(sseu); /* @@ -372,9 +406,8 @@ static void gen9_sseu_info_init(struct drm_i915_private *dev_priv) sseu->slice_mask = (fuse2 & GEN8_F2_S_ENA_MASK) >> GEN8_F2_S_ENA_SHIFT; /* BXT has a single slice and at most 3 subslices. */ - sseu->max_slices = IS_GEN9_LP(dev_priv) ? 1 : 3; - sseu->max_subslices = IS_GEN9_LP(dev_priv) ? 3 : 4; - sseu->max_eus_per_subslice = 8; + intel_sseu_set_info(sseu, IS_GEN9_LP(dev_priv) ? 1 : 3, + IS_GEN9_LP(dev_priv) ? 3 : 4, 8); /* * The subslice disable field is global, i.e. it applies @@ -393,14 +426,14 @@ static void gen9_sseu_info_init(struct drm_i915_private *dev_priv) /* skip disabled slice */ continue; - sseu->subslice_mask[s] = subslice_mask; + intel_sseu_set_subslices(sseu, s, subslice_mask); eu_disable = I915_READ(GEN9_EU_DISABLE(s)); for (ss = 0; ss < sseu->max_subslices; ss++) { int eu_per_ss; u8 eu_disabled_mask; - if (!(sseu->subslice_mask[s] & BIT(ss))) + if (!intel_sseu_has_subslice(sseu, s, ss)) /* skip disabled subslice */ continue; @@ -473,9 +506,7 @@ static void broadwell_sseu_info_init(struct drm_i915_private *dev_priv) fuse2 = I915_READ(GEN8_FUSE2); sseu->slice_mask = (fuse2 & GEN8_F2_S_ENA_MASK) >> GEN8_F2_S_ENA_SHIFT; - sseu->max_slices = 3; - sseu->max_subslices = 3; - sseu->max_eus_per_subslice = 8; + intel_sseu_set_info(sseu, 3, 3, 8); /* * The subslice disable field is global, i.e. it applies @@ -502,13 +533,13 @@ static void broadwell_sseu_info_init(struct drm_i915_private *dev_priv) /* skip disabled slice */ continue; - sseu->subslice_mask[s] = subslice_mask; + intel_sseu_set_subslices(sseu, s, subslice_mask); for (ss = 0; ss < sseu->max_subslices; ss++) { u8 eu_disabled_mask; u32 n_disabled; - if (!(sseu->subslice_mask[s] & BIT(ss))) + if (!intel_sseu_has_subslice(sseu, s, ss)) /* skip disabled subslice */ continue; @@ -552,6 +583,7 @@ static void haswell_sseu_info_init(struct drm_i915_private *dev_priv) { struct sseu_dev_info *sseu = &RUNTIME_INFO(dev_priv)->sseu; u32 fuse1; + u8 subslice_mask = 0; int s, ss; /* @@ -564,22 +596,18 @@ static void haswell_sseu_info_init(struct drm_i915_private *dev_priv) /* fall through */ case 1: sseu->slice_mask = BIT(0); - sseu->subslice_mask[0] = BIT(0); + subslice_mask = BIT(0); break; case 2: sseu->slice_mask = BIT(0); - sseu->subslice_mask[0] = BIT(0) | BIT(1); + subslice_mask = BIT(0) | BIT(1); break; case 3: sseu->slice_mask = BIT(0) | BIT(1); - sseu->subslice_mask[0] = BIT(0) | BIT(1); - sseu->subslice_mask[1] = BIT(0) | BIT(1); + subslice_mask = BIT(0) | BIT(1); break; } - sseu->max_slices = hweight8(sseu->slice_mask); - sseu->max_subslices = hweight8(sseu->subslice_mask[0]); - fuse1 = I915_READ(HSW_PAVP_FUSE1); switch ((fuse1 & HSW_F1_EU_DIS_MASK) >> HSW_F1_EU_DIS_SHIFT) { default: @@ -596,9 +624,14 @@ static void haswell_sseu_info_init(struct drm_i915_private *dev_priv) sseu->eu_per_subslice = 6; break; } - sseu->max_eus_per_subslice = sseu->eu_per_subslice; + + intel_sseu_set_info(sseu, hweight8(sseu->slice_mask), + hweight8(subslice_mask), + sseu->eu_per_subslice); for (s = 0; s < sseu->max_slices; s++) { + intel_sseu_set_subslices(sseu, s, subslice_mask); + for (ss = 0; ss < sseu->max_subslices; ss++) { sseu_set_eus(sseu, s, ss, (1UL << sseu->eu_per_subslice) - 1); @@ -900,12 +933,8 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) runtime->num_sprites[pipe] = 1; } - if (i915_modparams.disable_display) { - DRM_INFO("Display disabled (module parameter)\n"); - info->num_pipes = 0; - } else if (HAS_DISPLAY(dev_priv) && - (IS_GEN_RANGE(dev_priv, 7, 8)) && - HAS_PCH_SPLIT(dev_priv)) { + if (HAS_DISPLAY(dev_priv) && IS_GEN_RANGE(dev_priv, 7, 8) && + HAS_PCH_SPLIT(dev_priv)) { u32 fuse_strap = I915_READ(FUSE_STRAP); u32 sfuse_strap = I915_READ(SFUSE_STRAP); @@ -923,14 +952,14 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) (HAS_PCH_CPT(dev_priv) && !(sfuse_strap & SFUSE_STRAP_FUSE_LOCK))) { DRM_INFO("Display fused off, disabling\n"); - info->num_pipes = 0; + info->pipe_mask = 0; } else if (fuse_strap & IVB_PIPE_C_DISABLE) { DRM_INFO("PipeC fused off\n"); - info->num_pipes -= 1; + info->pipe_mask &= ~BIT(PIPE_C); } } else if (HAS_DISPLAY(dev_priv) && INTEL_GEN(dev_priv) >= 9) { u32 dfsm = I915_READ(SKL_DFSM); - u8 enabled_mask = BIT(info->num_pipes) - 1; + u8 enabled_mask = info->pipe_mask; if (dfsm & SKL_DFSM_PIPE_A_DISABLE) enabled_mask &= ~BIT(PIPE_A); @@ -951,7 +980,20 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) DRM_ERROR("invalid pipe fuse configuration: enabled_mask=0x%x\n", enabled_mask); else - info->num_pipes = hweight8(enabled_mask); + info->pipe_mask = enabled_mask; + + if (dfsm & SKL_DFSM_DISPLAY_HDCP_DISABLE) + info->display.has_hdcp = 0; + + if (dfsm & SKL_DFSM_DISPLAY_PM_DISABLE) + info->display.has_fbc = 0; + + if (INTEL_GEN(dev_priv) >= 11 && (dfsm & ICL_DFSM_DMC_DISABLE)) + info->display.has_csr = 0; + + if (INTEL_GEN(dev_priv) >= 10 && + (dfsm & CNL_DFSM_DISPLAY_DSC_DISABLE)) + info->display.has_dsc = 0; } /* Initialize slice/subslice/EU info */ @@ -965,8 +1007,10 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) gen9_sseu_info_init(dev_priv); else if (IS_GEN(dev_priv, 10)) gen10_sseu_info_init(dev_priv); - else if (INTEL_GEN(dev_priv) >= 11) + else if (IS_GEN(dev_priv, 11)) gen11_sseu_info_init(dev_priv); + else if (INTEL_GEN(dev_priv) >= 12) + gen12_sseu_info_init(dev_priv); if (IS_GEN(dev_priv, 6) && intel_vtd_active()) { DRM_INFO("Disabling ppGTT for VT-d support\n"); @@ -1010,8 +1054,10 @@ void intel_device_info_init_mmio(struct drm_i915_private *dev_priv) GEN11_GT_VEBOX_DISABLE_SHIFT; for (i = 0; i < I915_MAX_VCS; i++) { - if (!HAS_ENGINE(dev_priv, _VCS(i))) + if (!HAS_ENGINE(dev_priv, _VCS(i))) { + vdbox_mask &= ~BIT(i); continue; + } if (!(BIT(i) & vdbox_mask)) { info->engine_mask &= ~BIT(_VCS(i)); @@ -1032,8 +1078,10 @@ void intel_device_info_init_mmio(struct drm_i915_private *dev_priv) GEM_BUG_ON(vdbox_mask != VDBOX_MASK(dev_priv)); for (i = 0; i < I915_MAX_VECS; i++) { - if (!HAS_ENGINE(dev_priv, _VECS(i))) + if (!HAS_ENGINE(dev_priv, _VECS(i))) { + vebox_mask &= ~BIT(i); continue; + } if (!(BIT(i) & vebox_mask)) { info->engine_mask &= ~BIT(_VECS(i)); diff --git a/drivers/gpu/drm/i915/intel_device_info.h b/drivers/gpu/drm/i915/intel_device_info.h index 92e0c2e0954c..4bdf8a6cfb47 100644 --- a/drivers/gpu/drm/i915/intel_device_info.h +++ b/drivers/gpu/drm/i915/intel_device_info.h @@ -107,6 +107,7 @@ enum intel_ppgtt_type { func(is_mobile); \ func(is_lp); \ func(require_force_probe); \ + func(is_dgfx); \ /* Keep has_* in alphabetical order */ \ func(has_64bit_reloc); \ func(gpu_reset_clobbers_display); \ @@ -135,8 +136,11 @@ enum intel_ppgtt_type { func(has_csr); \ func(has_ddi); \ func(has_dp_mst); \ + func(has_dsb); \ + func(has_dsc); \ func(has_fbc); \ func(has_gmch); \ + func(has_hdcp); \ func(has_hotplug); \ func(has_ipc); \ func(has_modular_fia); \ @@ -159,9 +163,11 @@ struct intel_device_info { unsigned int page_sizes; /* page sizes supported by the HW */ + u32 memory_regions; /* regions supported by the HW */ + u32 display_mmio_offset; - u8 num_pipes; + u8 pipe_mask; #define DEFINE_FLAG(name) u8 name:1 DEV_INFO_FOR_EACH_FLAG(DEFINE_FLAG); diff --git a/drivers/gpu/drm/i915/intel_memory_region.c b/drivers/gpu/drm/i915/intel_memory_region.c new file mode 100644 index 000000000000..baaeaecc64af --- /dev/null +++ b/drivers/gpu/drm/i915/intel_memory_region.c @@ -0,0 +1,272 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include "intel_memory_region.h" +#include "i915_drv.h" + +/* XXX: Hysterical raisins. BIT(inst) needs to just be (inst) at some point. */ +#define REGION_MAP(type, inst) \ + BIT((type) + INTEL_MEMORY_TYPE_SHIFT) | BIT(inst) + +const u32 intel_region_map[] = { + [INTEL_REGION_SMEM] = REGION_MAP(INTEL_MEMORY_SYSTEM, 0), + [INTEL_REGION_LMEM] = REGION_MAP(INTEL_MEMORY_LOCAL, 0), + [INTEL_REGION_STOLEN] = REGION_MAP(INTEL_MEMORY_STOLEN, 0), +}; + +static u64 +intel_memory_region_free_pages(struct intel_memory_region *mem, + struct list_head *blocks) +{ + struct i915_buddy_block *block, *on; + u64 size = 0; + + list_for_each_entry_safe(block, on, blocks, link) { + size += i915_buddy_block_size(&mem->mm, block); + i915_buddy_free(&mem->mm, block); + } + INIT_LIST_HEAD(blocks); + + return size; +} + +void +__intel_memory_region_put_pages_buddy(struct intel_memory_region *mem, + struct list_head *blocks) +{ + mutex_lock(&mem->mm_lock); + intel_memory_region_free_pages(mem, blocks); + mutex_unlock(&mem->mm_lock); +} + +void +__intel_memory_region_put_block_buddy(struct i915_buddy_block *block) +{ + struct list_head blocks; + + INIT_LIST_HEAD(&blocks); + list_add(&block->link, &blocks); + __intel_memory_region_put_pages_buddy(block->private, &blocks); +} + +int +__intel_memory_region_get_pages_buddy(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags, + struct list_head *blocks) +{ + unsigned int min_order = 0; + unsigned long n_pages; + + GEM_BUG_ON(!IS_ALIGNED(size, mem->mm.chunk_size)); + GEM_BUG_ON(!list_empty(blocks)); + + if (flags & I915_ALLOC_MIN_PAGE_SIZE) { + min_order = ilog2(mem->min_page_size) - + ilog2(mem->mm.chunk_size); + } + + if (flags & I915_ALLOC_CONTIGUOUS) { + size = roundup_pow_of_two(size); + min_order = ilog2(size) - ilog2(mem->mm.chunk_size); + } + + n_pages = size >> ilog2(mem->mm.chunk_size); + + mutex_lock(&mem->mm_lock); + + do { + struct i915_buddy_block *block; + unsigned int order; + + order = fls(n_pages) - 1; + GEM_BUG_ON(order > mem->mm.max_order); + GEM_BUG_ON(order < min_order); + + do { + block = i915_buddy_alloc(&mem->mm, order); + if (!IS_ERR(block)) + break; + + if (order-- == min_order) + goto err_free_blocks; + } while (1); + + n_pages -= BIT(order); + + block->private = mem; + list_add(&block->link, blocks); + + if (!n_pages) + break; + } while (1); + + mutex_unlock(&mem->mm_lock); + return 0; + +err_free_blocks: + intel_memory_region_free_pages(mem, blocks); + mutex_unlock(&mem->mm_lock); + return -ENXIO; +} + +struct i915_buddy_block * +__intel_memory_region_get_block_buddy(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags) +{ + struct i915_buddy_block *block; + LIST_HEAD(blocks); + int ret; + + ret = __intel_memory_region_get_pages_buddy(mem, size, flags, &blocks); + if (ret) + return ERR_PTR(ret); + + block = list_first_entry(&blocks, typeof(*block), link); + list_del_init(&block->link); + return block; +} + +int intel_memory_region_init_buddy(struct intel_memory_region *mem) +{ + return i915_buddy_init(&mem->mm, resource_size(&mem->region), + PAGE_SIZE); +} + +void intel_memory_region_release_buddy(struct intel_memory_region *mem) +{ + i915_buddy_fini(&mem->mm); +} + +struct intel_memory_region * +intel_memory_region_create(struct drm_i915_private *i915, + resource_size_t start, + resource_size_t size, + resource_size_t min_page_size, + resource_size_t io_start, + const struct intel_memory_region_ops *ops) +{ + struct intel_memory_region *mem; + int err; + + mem = kzalloc(sizeof(*mem), GFP_KERNEL); + if (!mem) + return ERR_PTR(-ENOMEM); + + mem->i915 = i915; + mem->region = (struct resource)DEFINE_RES_MEM(start, size); + mem->io_start = io_start; + mem->min_page_size = min_page_size; + mem->ops = ops; + + mutex_init(&mem->objects.lock); + INIT_LIST_HEAD(&mem->objects.list); + INIT_LIST_HEAD(&mem->objects.purgeable); + + mutex_init(&mem->mm_lock); + + if (ops->init) { + err = ops->init(mem); + if (err) + goto err_free; + } + + kref_init(&mem->kref); + return mem; + +err_free: + kfree(mem); + return ERR_PTR(err); +} + +static void __intel_memory_region_destroy(struct kref *kref) +{ + struct intel_memory_region *mem = + container_of(kref, typeof(*mem), kref); + + if (mem->ops->release) + mem->ops->release(mem); + + mutex_destroy(&mem->mm_lock); + mutex_destroy(&mem->objects.lock); + kfree(mem); +} + +struct intel_memory_region * +intel_memory_region_get(struct intel_memory_region *mem) +{ + kref_get(&mem->kref); + return mem; +} + +void intel_memory_region_put(struct intel_memory_region *mem) +{ + kref_put(&mem->kref, __intel_memory_region_destroy); +} + +/* Global memory region registration -- only slight layer inversions! */ + +int intel_memory_regions_hw_probe(struct drm_i915_private *i915) +{ + int err, i; + + for (i = 0; i < ARRAY_SIZE(i915->mm.regions); i++) { + struct intel_memory_region *mem = ERR_PTR(-ENODEV); + u32 type; + + if (!HAS_REGION(i915, BIT(i))) + continue; + + type = MEMORY_TYPE_FROM_REGION(intel_region_map[i]); + switch (type) { + case INTEL_MEMORY_SYSTEM: + mem = i915_gem_shmem_setup(i915); + break; + case INTEL_MEMORY_STOLEN: + mem = i915_gem_stolen_setup(i915); + break; + case INTEL_MEMORY_LOCAL: + mem = intel_setup_fake_lmem(i915); + break; + } + + if (IS_ERR(mem)) { + err = PTR_ERR(mem); + DRM_ERROR("Failed to setup region(%d) type=%d\n", err, type); + goto out_cleanup; + } + + mem->id = intel_region_map[i]; + mem->type = type; + mem->instance = MEMORY_INSTANCE_FROM_REGION(intel_region_map[i]); + + i915->mm.regions[i] = mem; + } + + return 0; + +out_cleanup: + intel_memory_regions_driver_release(i915); + return err; +} + +void intel_memory_regions_driver_release(struct drm_i915_private *i915) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(i915->mm.regions); i++) { + struct intel_memory_region *region = + fetch_and_zero(&i915->mm.regions[i]); + + if (region) + intel_memory_region_put(region); + } +} + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "selftests/intel_memory_region.c" +#include "selftests/mock_region.c" +#endif diff --git a/drivers/gpu/drm/i915/intel_memory_region.h b/drivers/gpu/drm/i915/intel_memory_region.h new file mode 100644 index 000000000000..238722009677 --- /dev/null +++ b/drivers/gpu/drm/i915/intel_memory_region.h @@ -0,0 +1,129 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __INTEL_MEMORY_REGION_H__ +#define __INTEL_MEMORY_REGION_H__ + +#include <linux/kref.h> +#include <linux/ioport.h> +#include <linux/mutex.h> +#include <linux/io-mapping.h> +#include <drm/drm_mm.h> + +#include "i915_buddy.h" + +struct drm_i915_private; +struct drm_i915_gem_object; +struct intel_memory_region; +struct sg_table; + +/** + * Base memory type + */ +enum intel_memory_type { + INTEL_MEMORY_SYSTEM = 0, + INTEL_MEMORY_LOCAL, + INTEL_MEMORY_STOLEN, +}; + +enum intel_region_id { + INTEL_REGION_SMEM = 0, + INTEL_REGION_LMEM, + INTEL_REGION_STOLEN, + INTEL_REGION_UNKNOWN, /* Should be last */ +}; + +#define REGION_SMEM BIT(INTEL_REGION_SMEM) +#define REGION_LMEM BIT(INTEL_REGION_LMEM) +#define REGION_STOLEN BIT(INTEL_REGION_STOLEN) + +#define INTEL_MEMORY_TYPE_SHIFT 16 + +#define MEMORY_TYPE_FROM_REGION(r) (ilog2((r) >> INTEL_MEMORY_TYPE_SHIFT)) +#define MEMORY_INSTANCE_FROM_REGION(r) (ilog2((r) & 0xffff)) + +#define I915_ALLOC_MIN_PAGE_SIZE BIT(0) +#define I915_ALLOC_CONTIGUOUS BIT(1) + +/** + * Memory regions encoded as type | instance + */ +extern const u32 intel_region_map[]; + +struct intel_memory_region_ops { + unsigned int flags; + + int (*init)(struct intel_memory_region *mem); + void (*release)(struct intel_memory_region *mem); + + struct drm_i915_gem_object * + (*create_object)(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags); +}; + +struct intel_memory_region { + struct drm_i915_private *i915; + + const struct intel_memory_region_ops *ops; + + struct io_mapping iomap; + struct resource region; + + /* For fake LMEM */ + struct drm_mm_node fake_mappable; + + struct i915_buddy_mm mm; + struct mutex mm_lock; + + struct kref kref; + + resource_size_t io_start; + resource_size_t min_page_size; + + unsigned int type; + unsigned int instance; + unsigned int id; + + dma_addr_t remap_addr; + + struct { + struct mutex lock; /* Protects access to objects */ + struct list_head list; + struct list_head purgeable; + } objects; +}; + +int intel_memory_region_init_buddy(struct intel_memory_region *mem); +void intel_memory_region_release_buddy(struct intel_memory_region *mem); + +int __intel_memory_region_get_pages_buddy(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags, + struct list_head *blocks); +struct i915_buddy_block * +__intel_memory_region_get_block_buddy(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags); +void __intel_memory_region_put_pages_buddy(struct intel_memory_region *mem, + struct list_head *blocks); +void __intel_memory_region_put_block_buddy(struct i915_buddy_block *block); + +struct intel_memory_region * +intel_memory_region_create(struct drm_i915_private *i915, + resource_size_t start, + resource_size_t size, + resource_size_t min_page_size, + resource_size_t io_start, + const struct intel_memory_region_ops *ops); + +struct intel_memory_region * +intel_memory_region_get(struct intel_memory_region *mem); +void intel_memory_region_put(struct intel_memory_region *mem); + +int intel_memory_regions_hw_probe(struct drm_i915_private *i915); +void intel_memory_regions_driver_release(struct drm_i915_private *i915); + +#endif diff --git a/drivers/gpu/drm/i915/intel_pch.c b/drivers/gpu/drm/i915/intel_pch.c index fa864d8f2b73..8fd92b9130a7 100644 --- a/drivers/gpu/drm/i915/intel_pch.c +++ b/drivers/gpu/drm/i915/intel_pch.c @@ -52,7 +52,8 @@ intel_pch_type(const struct drm_i915_private *dev_priv, unsigned short id) return PCH_SPT; case INTEL_PCH_SPT_LP_DEVICE_ID_TYPE: DRM_DEBUG_KMS("Found SunrisePoint LP PCH\n"); - WARN_ON(!IS_SKYLAKE(dev_priv) && !IS_KABYLAKE(dev_priv)); + WARN_ON(!IS_SKYLAKE(dev_priv) && !IS_KABYLAKE(dev_priv) && + !IS_COFFEELAKE(dev_priv)); return PCH_SPT; case INTEL_PCH_KBP_DEVICE_ID_TYPE: DRM_DEBUG_KMS("Found Kaby Lake PCH (KBP)\n"); @@ -69,16 +70,21 @@ intel_pch_type(const struct drm_i915_private *dev_priv, unsigned short id) WARN_ON(!IS_CANNONLAKE(dev_priv) && !IS_COFFEELAKE(dev_priv)); return PCH_CNP; case INTEL_PCH_CMP_DEVICE_ID_TYPE: + case INTEL_PCH_CMP2_DEVICE_ID_TYPE: DRM_DEBUG_KMS("Found Comet Lake PCH (CMP)\n"); WARN_ON(!IS_COFFEELAKE(dev_priv)); /* CometPoint is CNP Compatible */ return PCH_CNP; + case INTEL_PCH_CMP_V_DEVICE_ID_TYPE: + DRM_DEBUG_KMS("Found Comet Lake V PCH (CMP-V)\n"); + WARN_ON(!IS_COFFEELAKE(dev_priv)); + /* Comet Lake V PCH is based on KBP, which is SPT compatible */ + return PCH_SPT; case INTEL_PCH_ICP_DEVICE_ID_TYPE: DRM_DEBUG_KMS("Found Ice Lake PCH\n"); WARN_ON(!IS_ICELAKE(dev_priv)); return PCH_ICP; case INTEL_PCH_MCC_DEVICE_ID_TYPE: - case INTEL_PCH_MCC2_DEVICE_ID_TYPE: DRM_DEBUG_KMS("Found Mule Creek Canyon PCH\n"); WARN_ON(!IS_ELKHARTLAKE(dev_priv)); return PCH_MCC; @@ -86,6 +92,11 @@ intel_pch_type(const struct drm_i915_private *dev_priv, unsigned short id) DRM_DEBUG_KMS("Found Tiger Lake LP PCH\n"); WARN_ON(!IS_TIGERLAKE(dev_priv)); return PCH_TGP; + case INTEL_PCH_JSP_DEVICE_ID_TYPE: + case INTEL_PCH_JSP2_DEVICE_ID_TYPE: + DRM_DEBUG_KMS("Found Jasper Lake PCH\n"); + WARN_ON(!IS_ELKHARTLAKE(dev_priv)); + return PCH_JSP; default: return PCH_NONE; } diff --git a/drivers/gpu/drm/i915/intel_pch.h b/drivers/gpu/drm/i915/intel_pch.h index e6a2d65f19c6..d26c25dd8d54 100644 --- a/drivers/gpu/drm/i915/intel_pch.h +++ b/drivers/gpu/drm/i915/intel_pch.h @@ -23,6 +23,7 @@ enum intel_pch { PCH_SPT, /* Sunrisepoint/Kaby Lake PCH */ PCH_CNP, /* Cannon/Comet Lake PCH */ PCH_ICP, /* Ice Lake PCH */ + PCH_JSP, /* Jasper Lake PCH */ PCH_MCC, /* Mule Creek Canyon PCH */ PCH_TGP, /* Tiger Lake PCH */ }; @@ -41,16 +42,20 @@ enum intel_pch { #define INTEL_PCH_CNP_DEVICE_ID_TYPE 0xA300 #define INTEL_PCH_CNP_LP_DEVICE_ID_TYPE 0x9D80 #define INTEL_PCH_CMP_DEVICE_ID_TYPE 0x0280 +#define INTEL_PCH_CMP2_DEVICE_ID_TYPE 0x0680 +#define INTEL_PCH_CMP_V_DEVICE_ID_TYPE 0xA380 #define INTEL_PCH_ICP_DEVICE_ID_TYPE 0x3480 #define INTEL_PCH_MCC_DEVICE_ID_TYPE 0x4B00 -#define INTEL_PCH_MCC2_DEVICE_ID_TYPE 0x3880 #define INTEL_PCH_TGP_DEVICE_ID_TYPE 0xA080 +#define INTEL_PCH_JSP_DEVICE_ID_TYPE 0x4D80 +#define INTEL_PCH_JSP2_DEVICE_ID_TYPE 0x3880 #define INTEL_PCH_P2X_DEVICE_ID_TYPE 0x7100 #define INTEL_PCH_P3X_DEVICE_ID_TYPE 0x7000 #define INTEL_PCH_QEMU_DEVICE_ID_TYPE 0x2900 /* qemu q35 has 2918 */ #define INTEL_PCH_TYPE(dev_priv) ((dev_priv)->pch_type) #define INTEL_PCH_ID(dev_priv) ((dev_priv)->pch_id) +#define HAS_PCH_JSP(dev_priv) (INTEL_PCH_TYPE(dev_priv) == PCH_JSP) #define HAS_PCH_MCC(dev_priv) (INTEL_PCH_TYPE(dev_priv) == PCH_MCC) #define HAS_PCH_TGP(dev_priv) (INTEL_PCH_TYPE(dev_priv) == PCH_TGP) #define HAS_PCH_ICP(dev_priv) (INTEL_PCH_TYPE(dev_priv) == PCH_ICP) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 75ee027abb80..809bff955b5a 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -25,7 +25,6 @@ * */ -#include <linux/cpufreq.h> #include <linux/module.h> #include <linux/pm_runtime.h> @@ -38,6 +37,8 @@ #include "display/intel_fbc.h" #include "display/intel_sprite.h" +#include "gt/intel_llc.h" + #include "i915_drv.h" #include "i915_irq.h" #include "i915_trace.h" @@ -45,26 +46,6 @@ #include "intel_sideband.h" #include "../../../platform/x86/intel_ips.h" -/** - * DOC: RC6 - * - * RC6 is a special power stage which allows the GPU to enter an very - * low-voltage mode when idle, using down to 0V while at this stage. This - * stage is entered automatically when the GPU is idle when RC6 support is - * enabled, and as soon as new workload arises GPU wakes up automatically as well. - * - * There are different RC6 modes available in Intel GPU, which differentiate - * among each other with the latency required to enter and leave RC6 and - * voltage consumed by the GPU in different states. - * - * The combination of the following flags define which states GPU is allowed - * to enter, while RC6 is the normal RC6 state, RC6p is the deep RC6, and - * RC6pp is deepest RC6. Their support by hardware varies according to the - * GPU, BIOS, chipset and platform. RC6 is usually the safest one and the one - * which brings the most power savings; deeper states save more power, but - * require higher latency to switch to and wake up. - */ - static void gen9_init_clock_gating(struct drm_i915_private *dev_priv) { if (HAS_LLC(dev_priv)) { @@ -126,6 +107,14 @@ static void bxt_init_clock_gating(struct drm_i915_private *dev_priv) */ I915_WRITE(GEN9_CLKGATE_DIS_0, I915_READ(GEN9_CLKGATE_DIS_0) | PWM1_GATING_DIS | PWM2_GATING_DIS); + + /* + * Lower the display internal timeout. + * This is needed to avoid any hard hangs when DSI port PLL + * is off and a MMIO access is attempted by any privilege + * application, using batch buffers or any other means. + */ + I915_WRITE(RM_TIMEOUT, MMIO_TIMEOUT_US(950)); } static void glk_init_clock_gating(struct drm_i915_private *dev_priv) @@ -216,8 +205,6 @@ static void i915_ironlake_get_mem_freq(struct drm_i915_private *dev_priv) break; } - dev_priv->ips.r_t = dev_priv->mem_freq; - switch (csipll & 0x3ff) { case 0x00c: dev_priv->fsb_freq = 3200; @@ -246,14 +233,6 @@ static void i915_ironlake_get_mem_freq(struct drm_i915_private *dev_priv) dev_priv->fsb_freq = 0; break; } - - if (dev_priv->fsb_freq == 3200) { - dev_priv->ips.c_m = 0; - } else if (dev_priv->fsb_freq > 3200 && dev_priv->fsb_freq <= 4800) { - dev_priv->ips.c_m = 1; - } else { - dev_priv->ips.c_m = 2; - } } static const struct cxsr_latency cxsr_latency_table[] = { @@ -1137,10 +1116,7 @@ static u16 g4x_compute_wm(const struct intel_crtc_state *crtc_state, clock = adjusted_mode->crtc_clock; htotal = adjusted_mode->crtc_htotal; - if (plane->id == PLANE_CURSOR) - width = plane_state->base.crtc_w; - else - width = drm_rect_width(&plane_state->base.dst); + width = drm_rect_width(&plane_state->base.dst); if (plane->id == PLANE_CURSOR) { wm = intel_wm_method2(clock, htotal, width, cpp, latency); @@ -1327,8 +1303,8 @@ static int g4x_compute_pipe_wm(struct intel_crtc_state *crtc_state) struct intel_atomic_state *state = to_intel_atomic_state(crtc_state->base.state); struct g4x_wm_state *wm_state = &crtc_state->wm.g4x.optimal; - int num_active_planes = hweight32(crtc_state->active_planes & - ~BIT(PLANE_CURSOR)); + int num_active_planes = hweight8(crtc_state->active_planes & + ~BIT(PLANE_CURSOR)); const struct g4x_pipe_wm *raw; const struct intel_plane_state *old_plane_state; const struct intel_plane_state *new_plane_state; @@ -1490,7 +1466,7 @@ static void g4x_merge_wm(struct drm_i915_private *dev_priv, struct g4x_wm_values *wm) { struct intel_crtc *crtc; - int num_active_crtcs = 0; + int num_active_pipes = 0; wm->cxsr = true; wm->hpll_en = true; @@ -1509,10 +1485,10 @@ static void g4x_merge_wm(struct drm_i915_private *dev_priv, if (!wm_state->fbc_en) wm->fbc_en = false; - num_active_crtcs++; + num_active_pipes++; } - if (num_active_crtcs != 1) { + if (num_active_pipes != 1) { wm->cxsr = false; wm->hpll_en = false; wm->fbc_en = false; @@ -1659,7 +1635,7 @@ static int vlv_compute_fifo(struct intel_crtc_state *crtc_state) &crtc_state->wm.vlv.raw[VLV_WM_LEVEL_PM2]; struct vlv_fifo_state *fifo_state = &crtc_state->wm.vlv.fifo_state; unsigned int active_planes = crtc_state->active_planes & ~BIT(PLANE_CURSOR); - int num_active_planes = hweight32(active_planes); + int num_active_planes = hweight8(active_planes); const int fifo_size = 511; int fifo_extra, fifo_left = fifo_size; int sprite0_fifo_extra = 0; @@ -1848,8 +1824,8 @@ static int vlv_compute_pipe_wm(struct intel_crtc_state *crtc_state) struct vlv_wm_state *wm_state = &crtc_state->wm.vlv.optimal; const struct vlv_fifo_state *fifo_state = &crtc_state->wm.vlv.fifo_state; - int num_active_planes = hweight32(crtc_state->active_planes & - ~BIT(PLANE_CURSOR)); + int num_active_planes = hweight8(crtc_state->active_planes & + ~BIT(PLANE_CURSOR)); bool needs_modeset = drm_atomic_crtc_needs_modeset(&crtc_state->base); const struct intel_plane_state *old_plane_state; const struct intel_plane_state *new_plane_state; @@ -1909,7 +1885,7 @@ static int vlv_compute_pipe_wm(struct intel_crtc_state *crtc_state) for (level = 0; level < wm_state->num_levels; level++) { const struct g4x_pipe_wm *raw = &crtc_state->wm.vlv.raw[level]; - const int sr_fifo_size = INTEL_INFO(dev_priv)->num_pipes * 512 - 1; + const int sr_fifo_size = INTEL_NUM_PIPES(dev_priv) * 512 - 1; if (!vlv_raw_crtc_wm_is_valid(crtc_state, level)) break; @@ -2098,7 +2074,7 @@ static void vlv_merge_wm(struct drm_i915_private *dev_priv, struct vlv_wm_values *wm) { struct intel_crtc *crtc; - int num_active_crtcs = 0; + int num_active_pipes = 0; wm->level = dev_priv->wm.max_level; wm->cxsr = true; @@ -2112,14 +2088,14 @@ static void vlv_merge_wm(struct drm_i915_private *dev_priv, if (!wm_state->cxsr) wm->cxsr = false; - num_active_crtcs++; + num_active_pipes++; wm->level = min_t(int, wm->level, wm_state->num_levels - 1); } - if (num_active_crtcs != 1) + if (num_active_pipes != 1) wm->cxsr = false; - if (num_active_crtcs > 1) + if (num_active_pipes > 1) wm->level = VLV_WM_LEVEL_PM2; for_each_intel_crtc(&dev_priv->drm, crtc) { @@ -2569,7 +2545,8 @@ static u32 ilk_compute_cur_wm(const struct intel_crtc_state *crtc_state, return ilk_wm_method2(crtc_state->pixel_rate, crtc_state->base.adjusted_mode.crtc_htotal, - plane_state->base.crtc_w, cpp, mem_value); + drm_rect_width(&plane_state->base.dst), + cpp, mem_value); } /* Only for WM_LP. */ @@ -2648,7 +2625,7 @@ static unsigned int ilk_plane_wm_max(const struct drm_i915_private *dev_priv, /* HSW allows LP1+ watermarks even with multiple pipes */ if (level == 0 || config->num_pipes_active > 1) { - fifo_size /= INTEL_INFO(dev_priv)->num_pipes; + fifo_size /= INTEL_NUM_PIPES(dev_priv); /* * For some reason the non self refresh @@ -3109,8 +3086,8 @@ static int ilk_compute_pipe_wm(struct intel_crtc_state *crtc_state) struct intel_pipe_wm *pipe_wm; struct drm_device *dev = state->dev; const struct drm_i915_private *dev_priv = to_i915(dev); - struct drm_plane *plane; - const struct drm_plane_state *plane_state; + struct intel_plane *plane; + const struct intel_plane_state *plane_state; const struct intel_plane_state *pristate = NULL; const struct intel_plane_state *sprstate = NULL; const struct intel_plane_state *curstate = NULL; @@ -3119,15 +3096,13 @@ static int ilk_compute_pipe_wm(struct intel_crtc_state *crtc_state) pipe_wm = &crtc_state->wm.ilk.optimal; - drm_atomic_crtc_state_for_each_plane_state(plane, plane_state, &crtc_state->base) { - const struct intel_plane_state *ps = to_intel_plane_state(plane_state); - - if (plane->type == DRM_PLANE_TYPE_PRIMARY) - pristate = ps; - else if (plane->type == DRM_PLANE_TYPE_OVERLAY) - sprstate = ps; - else if (plane->type == DRM_PLANE_TYPE_CURSOR) - curstate = ps; + intel_atomic_crtc_state_for_each_plane_state(plane, plane_state, crtc_state) { + if (plane->base.type == DRM_PLANE_TYPE_PRIMARY) + pristate = plane_state; + else if (plane->base.type == DRM_PLANE_TYPE_OVERLAY) + sprstate = plane_state; + else if (plane->base.type == DRM_PLANE_TYPE_CURSOR) + curstate = plane_state; } pipe_wm->pipe_enabled = crtc_state->base.active; @@ -3654,10 +3629,47 @@ static bool skl_needs_memory_bw_wa(struct drm_i915_private *dev_priv) static bool intel_has_sagv(struct drm_i915_private *dev_priv) { + /* HACK! */ + if (IS_GEN(dev_priv, 12)) + return false; + return (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10) && dev_priv->sagv_status != I915_SAGV_NOT_CONTROLLED; } +static void +skl_setup_sagv_block_time(struct drm_i915_private *dev_priv) +{ + if (INTEL_GEN(dev_priv) >= 12) { + u32 val = 0; + int ret; + + ret = sandybridge_pcode_read(dev_priv, + GEN12_PCODE_READ_SAGV_BLOCK_TIME_US, + &val, NULL); + if (!ret) { + dev_priv->sagv_block_time_us = val; + return; + } + + DRM_DEBUG_DRIVER("Couldn't read SAGV block time!\n"); + } else if (IS_GEN(dev_priv, 11)) { + dev_priv->sagv_block_time_us = 10; + return; + } else if (IS_GEN(dev_priv, 10)) { + dev_priv->sagv_block_time_us = 20; + return; + } else if (IS_GEN(dev_priv, 9)) { + dev_priv->sagv_block_time_us = 30; + return; + } else { + MISSING_CASE(INTEL_GEN(dev_priv)); + } + + /* Default to an unusable block time */ + dev_priv->sagv_block_time_us = -1; +} + /* * SAGV dynamically adjusts the system agent voltage and clock frequencies * depending on power and performance requirements. The display engine access @@ -3746,33 +3758,25 @@ bool intel_can_enable_sagv(struct intel_atomic_state *state) struct intel_crtc_state *crtc_state; enum pipe pipe; int level, latency; - int sagv_block_time_us; if (!intel_has_sagv(dev_priv)) return false; - if (IS_GEN(dev_priv, 9)) - sagv_block_time_us = 30; - else if (IS_GEN(dev_priv, 10)) - sagv_block_time_us = 20; - else - sagv_block_time_us = 10; - /* * If there are no active CRTCs, no additional checks need be performed */ - if (hweight32(state->active_crtcs) == 0) + if (hweight8(state->active_pipes) == 0) return true; /* * SKL+ workaround: bspec recommends we disable SAGV when we have * more then one pipe enabled */ - if (hweight32(state->active_crtcs) > 1) + if (hweight8(state->active_pipes) > 1) return false; /* Since we're now guaranteed to only have one active CRTC... */ - pipe = ffs(state->active_crtcs) - 1; + pipe = ffs(state->active_pipes) - 1; crtc = intel_get_crtc_for_pipe(dev_priv, pipe); crtc_state = to_intel_crtc_state(crtc->base.state); @@ -3804,7 +3808,7 @@ bool intel_can_enable_sagv(struct intel_atomic_state *state) * incur memory latencies higher than sagv_block_time_us we * can't enable SAGV. */ - if (latency < sagv_block_time_us) + if (latency < dev_priv->sagv_block_time_us) return false; } @@ -3867,14 +3871,14 @@ skl_ddb_get_pipe_allocation_limits(struct drm_i915_private *dev_priv, if (WARN_ON(!state) || !crtc_state->base.active) { alloc->start = 0; alloc->end = 0; - *num_active = hweight32(dev_priv->active_crtcs); + *num_active = hweight8(dev_priv->active_pipes); return; } if (intel_state->active_pipe_changes) - *num_active = hweight32(intel_state->active_crtcs); + *num_active = hweight8(intel_state->active_pipes); else - *num_active = hweight32(dev_priv->active_crtcs); + *num_active = hweight8(dev_priv->active_pipes); ddb_size = intel_get_ddb_size(dev_priv, crtc_state, total_data_rate, *num_active, ddb); @@ -4005,7 +4009,8 @@ skl_ddb_get_hw_plane_state(struct drm_i915_private *dev_priv, val = I915_READ(PLANE_BUF_CFG(pipe, plane_id)); val2 = I915_READ(PLANE_NV12_BUF_CFG(pipe, plane_id)); - if (is_planar_yuv_format(fourcc)) + if (fourcc && + drm_format_info_is_yuv_semiplanar(drm_format_info(fourcc))) swap(val, val2); skl_ddb_entry_init_from_hw(dev_priv, ddb_y, val); @@ -4063,7 +4068,6 @@ static uint_fixed_16_16_t skl_plane_downscale_amount(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state) { - struct intel_plane *plane = to_intel_plane(plane_state->base.plane); u32 src_w, src_h, dst_w, dst_h; uint_fixed_16_16_t fp_w_ratio, fp_h_ratio; uint_fixed_16_16_t downscale_h, downscale_w; @@ -4071,27 +4075,17 @@ skl_plane_downscale_amount(const struct intel_crtc_state *crtc_state, if (WARN_ON(!intel_wm_plane_visible(crtc_state, plane_state))) return u32_to_fixed16(0); - /* n.b., src is 16.16 fixed point, dst is whole integer */ - if (plane->id == PLANE_CURSOR) { - /* - * Cursors only support 0/180 degree rotation, - * hence no need to account for rotation here. - */ - src_w = plane_state->base.src_w >> 16; - src_h = plane_state->base.src_h >> 16; - dst_w = plane_state->base.crtc_w; - dst_h = plane_state->base.crtc_h; - } else { - /* - * Src coordinates are already rotated by 270 degrees for - * the 90/270 degree plane rotation cases (to match the - * GTT mapping), hence no need to account for rotation here. - */ - src_w = drm_rect_width(&plane_state->base.src) >> 16; - src_h = drm_rect_height(&plane_state->base.src) >> 16; - dst_w = drm_rect_width(&plane_state->base.dst); - dst_h = drm_rect_height(&plane_state->base.dst); - } + /* + * Src coordinates are already rotated by 270 degrees for + * the 90/270 degree plane rotation cases (to match the + * GTT mapping), hence no need to account for rotation here. + * + * n.b., src is 16.16 fixed point, dst is whole integer. + */ + src_w = drm_rect_width(&plane_state->base.src) >> 16; + src_h = drm_rect_height(&plane_state->base.src) >> 16; + dst_w = drm_rect_width(&plane_state->base.dst); + dst_h = drm_rect_height(&plane_state->base.dst); fp_w_ratio = div_fixed16(src_w, dst_w); fp_h_ratio = div_fixed16(src_h, dst_h); @@ -4101,117 +4095,26 @@ skl_plane_downscale_amount(const struct intel_crtc_state *crtc_state, return mul_fixed16(downscale_w, downscale_h); } -static uint_fixed_16_16_t -skl_pipe_downscale_amount(const struct intel_crtc_state *crtc_state) -{ - uint_fixed_16_16_t pipe_downscale = u32_to_fixed16(1); - - if (!crtc_state->base.enable) - return pipe_downscale; - - if (crtc_state->pch_pfit.enabled) { - u32 src_w, src_h, dst_w, dst_h; - u32 pfit_size = crtc_state->pch_pfit.size; - uint_fixed_16_16_t fp_w_ratio, fp_h_ratio; - uint_fixed_16_16_t downscale_h, downscale_w; - - src_w = crtc_state->pipe_src_w; - src_h = crtc_state->pipe_src_h; - dst_w = pfit_size >> 16; - dst_h = pfit_size & 0xffff; - - if (!dst_w || !dst_h) - return pipe_downscale; - - fp_w_ratio = div_fixed16(src_w, dst_w); - fp_h_ratio = div_fixed16(src_h, dst_h); - downscale_w = max_fixed16(fp_w_ratio, u32_to_fixed16(1)); - downscale_h = max_fixed16(fp_h_ratio, u32_to_fixed16(1)); - - pipe_downscale = mul_fixed16(downscale_w, downscale_h); - } - - return pipe_downscale; -} - -int skl_check_pipe_max_pixel_rate(struct intel_crtc *intel_crtc, - struct intel_crtc_state *crtc_state) -{ - struct drm_i915_private *dev_priv = to_i915(intel_crtc->base.dev); - struct drm_atomic_state *state = crtc_state->base.state; - struct drm_plane *plane; - const struct drm_plane_state *drm_plane_state; - int crtc_clock, dotclk; - u32 pipe_max_pixel_rate; - uint_fixed_16_16_t pipe_downscale; - uint_fixed_16_16_t max_downscale = u32_to_fixed16(1); - - if (!crtc_state->base.enable) - return 0; - - drm_atomic_crtc_state_for_each_plane_state(plane, drm_plane_state, &crtc_state->base) { - uint_fixed_16_16_t plane_downscale; - uint_fixed_16_16_t fp_9_div_8 = div_fixed16(9, 8); - int bpp; - const struct intel_plane_state *plane_state = - to_intel_plane_state(drm_plane_state); - - if (!intel_wm_plane_visible(crtc_state, plane_state)) - continue; - - if (WARN_ON(!plane_state->base.fb)) - return -EINVAL; - - plane_downscale = skl_plane_downscale_amount(crtc_state, plane_state); - bpp = plane_state->base.fb->format->cpp[0] * 8; - if (bpp == 64) - plane_downscale = mul_fixed16(plane_downscale, - fp_9_div_8); - - max_downscale = max_fixed16(plane_downscale, max_downscale); - } - pipe_downscale = skl_pipe_downscale_amount(crtc_state); - - pipe_downscale = mul_fixed16(pipe_downscale, max_downscale); - - crtc_clock = crtc_state->base.adjusted_mode.crtc_clock; - dotclk = to_intel_atomic_state(state)->cdclk.logical.cdclk; - - if (IS_GEMINILAKE(dev_priv) || INTEL_GEN(dev_priv) >= 10) - dotclk *= 2; - - pipe_max_pixel_rate = div_round_up_u32_fixed16(dotclk, pipe_downscale); - - if (pipe_max_pixel_rate < crtc_clock) { - DRM_DEBUG_KMS("Max supported pixel clock with scaling exceeded\n"); - return -EINVAL; - } - - return 0; -} - static u64 skl_plane_relative_data_rate(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state, - const int plane) + int color_plane) { - struct intel_plane *intel_plane = to_intel_plane(plane_state->base.plane); + struct intel_plane *plane = to_intel_plane(plane_state->base.plane); + const struct drm_framebuffer *fb = plane_state->base.fb; u32 data_rate; u32 width = 0, height = 0; - struct drm_framebuffer *fb; - u32 format; uint_fixed_16_16_t down_scale_amount; u64 rate; if (!plane_state->base.visible) return 0; - fb = plane_state->base.fb; - format = fb->format->format; - - if (intel_plane->id == PLANE_CURSOR) + if (plane->id == PLANE_CURSOR) return 0; - if (plane == 1 && !is_planar_yuv_format(format)) + + if (color_plane == 1 && + !drm_format_info_is_yuv_semiplanar(fb->format)) return 0; /* @@ -4223,7 +4126,7 @@ skl_plane_relative_data_rate(const struct intel_crtc_state *crtc_state, height = drm_rect_height(&plane_state->base.src) >> 16; /* UV plane does 1/2 pixel sub-sampling */ - if (plane == 1 && is_planar_yuv_format(format)) { + if (color_plane == 1) { width /= 2; height /= 2; } @@ -4234,7 +4137,7 @@ skl_plane_relative_data_rate(const struct intel_crtc_state *crtc_state, rate = mul_round_up_u32_fixed16(data_rate, down_scale_amount); - rate *= fb->format->cpp[plane]; + rate *= fb->format->cpp[color_plane]; return rate; } @@ -4244,18 +4147,16 @@ skl_get_total_relative_data_rate(struct intel_crtc_state *crtc_state, u64 *uv_plane_data_rate) { struct drm_atomic_state *state = crtc_state->base.state; - struct drm_plane *plane; - const struct drm_plane_state *drm_plane_state; + struct intel_plane *plane; + const struct intel_plane_state *plane_state; u64 total_data_rate = 0; if (WARN_ON(!state)) return 0; /* Calculate and cache data rate for each plane */ - drm_atomic_crtc_state_for_each_plane_state(plane, drm_plane_state, &crtc_state->base) { - enum plane_id plane_id = to_intel_plane(plane)->id; - const struct intel_plane_state *plane_state = - to_intel_plane_state(drm_plane_state); + intel_atomic_crtc_state_for_each_plane_state(plane, plane_state, crtc_state) { + enum plane_id plane_id = plane->id; u64 rate; /* packed/y */ @@ -4276,21 +4177,19 @@ static u64 icl_get_total_relative_data_rate(struct intel_crtc_state *crtc_state, u64 *plane_data_rate) { - struct drm_plane *plane; - const struct drm_plane_state *drm_plane_state; + struct intel_plane *plane; + const struct intel_plane_state *plane_state; u64 total_data_rate = 0; if (WARN_ON(!crtc_state->base.state)) return 0; /* Calculate and cache data rate for each plane */ - drm_atomic_crtc_state_for_each_plane_state(plane, drm_plane_state, &crtc_state->base) { - const struct intel_plane_state *plane_state = - to_intel_plane_state(drm_plane_state); - enum plane_id plane_id = to_intel_plane(plane)->id; + intel_atomic_crtc_state_for_each_plane_state(plane, plane_state, crtc_state) { + enum plane_id plane_id = plane->id; u64 rate; - if (!plane_state->linked_plane) { + if (!plane_state->planar_linked_plane) { rate = skl_plane_relative_data_rate(crtc_state, plane_state, 0); plane_data_rate[plane_id] = rate; total_data_rate += rate; @@ -4299,17 +4198,17 @@ icl_get_total_relative_data_rate(struct intel_crtc_state *crtc_state, /* * The slave plane might not iterate in - * drm_atomic_crtc_state_for_each_plane_state(), + * intel_atomic_crtc_state_for_each_plane_state(), * and needs the master plane state which may be * NULL if we try get_new_plane_state(), so we * always calculate from the master. */ - if (plane_state->slave) + if (plane_state->planar_slave) continue; /* Y plane rate is calculated on the slave */ rate = skl_plane_relative_data_rate(crtc_state, plane_state, 0); - y_plane_id = plane_state->linked_plane->id; + y_plane_id = plane_state->planar_linked_plane->id; plane_data_rate[y_plane_id] = rate; total_data_rate += rate; @@ -4639,7 +4538,7 @@ skl_compute_wm_params(const struct intel_crtc_state *crtc_state, u32 interm_pbpl; /* only planar format has two planes */ - if (color_plane == 1 && !is_planar_yuv_format(format->format)) { + if (color_plane == 1 && !drm_format_info_is_yuv_semiplanar(format)) { DRM_DEBUG_KMS("Non planar format have single plane\n"); return -EINVAL; } @@ -4651,7 +4550,7 @@ skl_compute_wm_params(const struct intel_crtc_state *crtc_state, wp->x_tiled = modifier == I915_FORMAT_MOD_X_TILED; wp->rc_surface = modifier == I915_FORMAT_MOD_Y_TILED_CCS || modifier == I915_FORMAT_MOD_Yf_TILED_CCS; - wp->is_planar = is_planar_yuv_format(format->format); + wp->is_planar = drm_format_info_is_yuv_semiplanar(format); wp->width = width; if (color_plane == 1 && wp->is_planar) @@ -4723,20 +4622,15 @@ skl_compute_plane_wm_params(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state, struct skl_wm_params *wp, int color_plane) { - struct intel_plane *plane = to_intel_plane(plane_state->base.plane); const struct drm_framebuffer *fb = plane_state->base.fb; int width; - if (plane->id == PLANE_CURSOR) { - width = plane_state->base.crtc_w; - } else { - /* - * Src coordinates are already rotated by 270 degrees for - * the 90/270 degree plane rotation cases (to match the - * GTT mapping), hence no need to account for rotation here. - */ - width = drm_rect_width(&plane_state->base.src) >> 16; - } + /* + * Src coordinates are already rotated by 270 degrees for + * the 90/270 degree plane rotation cases (to match the + * GTT mapping), hence no need to account for rotation here. + */ + width = drm_rect_width(&plane_state->base.src) >> 16; return skl_compute_wm_params(crtc_state, width, fb->format, fb->modifier, @@ -5048,12 +4942,12 @@ static int icl_build_plane_wm(struct intel_crtc_state *crtc_state, int ret; /* Watermarks calculated in master */ - if (plane_state->slave) + if (plane_state->planar_slave) return 0; - if (plane_state->linked_plane) { + if (plane_state->planar_linked_plane) { const struct drm_framebuffer *fb = plane_state->base.fb; - enum plane_id y_plane_id = plane_state->linked_plane->id; + enum plane_id y_plane_id = plane_state->planar_linked_plane->id; WARN_ON(!intel_wm_plane_visible(crtc_state, plane_state)); WARN_ON(!fb->format->is_yuv || @@ -5082,8 +4976,8 @@ static int skl_build_pipe_wm(struct intel_crtc_state *crtc_state) { struct drm_i915_private *dev_priv = to_i915(crtc_state->base.crtc->dev); struct skl_pipe_wm *pipe_wm = &crtc_state->wm.skl.optimal; - struct drm_plane *plane; - const struct drm_plane_state *drm_plane_state; + struct intel_plane *plane; + const struct intel_plane_state *plane_state; int ret; /* @@ -5092,10 +4986,8 @@ static int skl_build_pipe_wm(struct intel_crtc_state *crtc_state) */ memset(pipe_wm->planes, 0, sizeof(pipe_wm->planes)); - drm_atomic_crtc_state_for_each_plane_state(plane, drm_plane_state, - &crtc_state->base) { - const struct intel_plane_state *plane_state = - to_intel_plane_state(drm_plane_state); + intel_atomic_crtc_state_for_each_plane_state(plane, plane_state, + crtc_state) { if (INTEL_GEN(dev_priv) >= 11) ret = icl_build_plane_wm(crtc_state, plane_state); @@ -5255,19 +5147,6 @@ bool skl_ddb_allocation_overlaps(const struct skl_ddb_entry *ddb, return false; } -static u32 -pipes_modified(struct intel_atomic_state *state) -{ - struct intel_crtc *crtc; - struct intel_crtc_state *crtc_state; - u32 i, ret = 0; - - for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) - ret |= drm_crtc_mask(&crtc->base); - - return ret; -} - static int skl_ddb_add_affected_planes(const struct intel_crtc_state *old_crtc_state, struct intel_crtc_state *new_crtc_state) @@ -5443,36 +5322,27 @@ skl_print_wm_changes(struct intel_atomic_state *state) } } -static int -skl_ddb_add_affected_pipes(struct intel_atomic_state *state, bool *changed) +static int intel_add_all_pipes(struct intel_atomic_state *state) { - struct drm_device *dev = state->base.dev; - const struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(state->base.dev); struct intel_crtc *crtc; - struct intel_crtc_state *crtc_state; - u32 realloc_pipes = pipes_modified(state); - int ret, i; - /* - * When we distrust bios wm we always need to recompute to set the - * expected DDB allocations for each CRTC. - */ - if (dev_priv->wm.distrust_bios_wm) - (*changed) = true; + for_each_intel_crtc(&dev_priv->drm, crtc) { + struct intel_crtc_state *crtc_state; - /* - * If this transaction isn't actually touching any CRTC's, don't - * bother with watermark calculation. Note that if we pass this - * test, we're guaranteed to hold at least one CRTC state mutex, - * which means we can safely use values like dev_priv->active_crtcs - * since any racing commits that want to update them would need to - * hold _all_ CRTC state mutexes. - */ - for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) - (*changed) = true; + crtc_state = intel_atomic_get_crtc_state(&state->base, crtc); + if (IS_ERR(crtc_state)) + return PTR_ERR(crtc_state); + } - if (!*changed) - return 0; + return 0; +} + +static int +skl_ddb_add_affected_pipes(struct intel_atomic_state *state) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + int ret; /* * If this is our first atomic update following hardware readout, @@ -5481,7 +5351,7 @@ skl_ddb_add_affected_pipes(struct intel_atomic_state *state, bool *changed) * ensure a full DDB recompute. */ if (dev_priv->wm.distrust_bios_wm) { - ret = drm_modeset_lock(&dev->mode_config.connection_mutex, + ret = drm_modeset_lock(&dev_priv->drm.mode_config.connection_mutex, state->base.acquire_ctx); if (ret) return ret; @@ -5489,13 +5359,13 @@ skl_ddb_add_affected_pipes(struct intel_atomic_state *state, bool *changed) state->active_pipe_changes = ~0; /* - * We usually only initialize state->active_crtcs if we + * We usually only initialize state->active_pipes if we * we're doing a modeset; make sure this field is always * initialized during the sanitization process that happens * on the first commit too. */ if (!state->modeset) - state->active_crtcs = dev_priv->active_crtcs; + state->active_pipes = dev_priv->active_pipes; } /* @@ -5512,18 +5382,11 @@ skl_ddb_add_affected_pipes(struct intel_atomic_state *state, bool *changed) * to grab the lock on *all* CRTC's. */ if (state->active_pipe_changes || state->modeset) { - realloc_pipes = ~0; state->wm_results.dirty_pipes = ~0; - } - /* - * We're not recomputing for the pipes not included in the commit, so - * make sure we start with the current state. - */ - for_each_intel_crtc_mask(dev, crtc, realloc_pipes) { - crtc_state = intel_atomic_get_crtc_state(&state->base, crtc); - if (IS_ERR(crtc_state)) - return PTR_ERR(crtc_state); + ret = intel_add_all_pipes(state); + if (ret) + return ret; } return 0; @@ -5596,14 +5459,13 @@ skl_compute_wm(struct intel_atomic_state *state) struct intel_crtc_state *new_crtc_state; struct intel_crtc_state *old_crtc_state; struct skl_ddb_values *results = &state->wm_results; - bool changed = false; int ret, i; /* Clear all dirty flags */ results->dirty_pipes = 0; - ret = skl_ddb_add_affected_pipes(state, &changed); - if (ret || !changed) + ret = skl_ddb_add_affected_pipes(state); + if (ret) return ret; /* @@ -5625,7 +5487,7 @@ skl_compute_wm(struct intel_atomic_state *state) if (!skl_pipe_wm_equals(crtc, &old_crtc_state->wm.skl.optimal, &new_crtc_state->wm.skl.optimal)) - results->dirty_pipes |= drm_crtc_mask(&crtc->base); + results->dirty_pipes |= BIT(crtc->pipe); } ret = skl_compute_ddb(state); @@ -5645,7 +5507,7 @@ static void skl_atomic_update_crtc_wm(struct intel_atomic_state *state, struct skl_pipe_wm *pipe_wm = &crtc_state->wm.skl.optimal; enum pipe pipe = crtc->pipe; - if (!(state->wm_results.dirty_pipes & drm_crtc_mask(&crtc->base))) + if ((state->wm_results.dirty_pipes & BIT(crtc->pipe)) == 0) return; I915_WRITE(PIPE_WM_LINETIME(pipe), pipe_wm->linetime); @@ -5654,12 +5516,11 @@ static void skl_atomic_update_crtc_wm(struct intel_atomic_state *state, static void skl_initial_wm(struct intel_atomic_state *state, struct intel_crtc_state *crtc_state) { - struct intel_crtc *intel_crtc = to_intel_crtc(crtc_state->base.crtc); - struct drm_device *dev = intel_crtc->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); struct skl_ddb_values *results = &state->wm_results; - if ((results->dirty_pipes & drm_crtc_mask(&intel_crtc->base)) == 0) + if ((results->dirty_pipes & BIT(crtc->pipe)) == 0) return; mutex_lock(&dev_priv->wm.wm_mutex); @@ -5808,10 +5669,10 @@ void skl_wm_get_hw_state(struct drm_i915_private *dev_priv) skl_pipe_wm_get_hw_state(crtc, &crtc_state->wm.skl.optimal); if (crtc->active) - hw->dirty_pipes |= drm_crtc_mask(&crtc->base); + hw->dirty_pipes |= BIT(crtc->pipe); } - if (dev_priv->active_crtcs) { + if (dev_priv->active_pipes) { /* Fully recompute DDB on first atomic commit */ dev_priv->wm.distrust_bios_wm = true; } @@ -6389,2378 +6250,6 @@ void intel_init_ipc(struct drm_i915_private *dev_priv) intel_enable_ipc(dev_priv); } -/* - * Lock protecting IPS related data structures - */ -DEFINE_SPINLOCK(mchdev_lock); - -bool ironlake_set_drps(struct drm_i915_private *i915, u8 val) -{ - struct intel_uncore *uncore = &i915->uncore; - u16 rgvswctl; - - lockdep_assert_held(&mchdev_lock); - - rgvswctl = intel_uncore_read16(uncore, MEMSWCTL); - if (rgvswctl & MEMCTL_CMD_STS) { - DRM_DEBUG("gpu busy, RCS change rejected\n"); - return false; /* still busy with another command */ - } - - rgvswctl = (MEMCTL_CMD_CHFREQ << MEMCTL_CMD_SHIFT) | - (val << MEMCTL_FREQ_SHIFT) | MEMCTL_SFCAVM; - intel_uncore_write16(uncore, MEMSWCTL, rgvswctl); - intel_uncore_posting_read16(uncore, MEMSWCTL); - - rgvswctl |= MEMCTL_CMD_STS; - intel_uncore_write16(uncore, MEMSWCTL, rgvswctl); - - return true; -} - -static void ironlake_enable_drps(struct drm_i915_private *dev_priv) -{ - struct intel_uncore *uncore = &dev_priv->uncore; - u32 rgvmodectl; - u8 fmax, fmin, fstart, vstart; - - spin_lock_irq(&mchdev_lock); - - rgvmodectl = intel_uncore_read(uncore, MEMMODECTL); - - /* Enable temp reporting */ - intel_uncore_write16(uncore, PMMISC, I915_READ(PMMISC) | MCPPCE_EN); - intel_uncore_write16(uncore, TSC1, I915_READ(TSC1) | TSE); - - /* 100ms RC evaluation intervals */ - intel_uncore_write(uncore, RCUPEI, 100000); - intel_uncore_write(uncore, RCDNEI, 100000); - - /* Set max/min thresholds to 90ms and 80ms respectively */ - intel_uncore_write(uncore, RCBMAXAVG, 90000); - intel_uncore_write(uncore, RCBMINAVG, 80000); - - intel_uncore_write(uncore, MEMIHYST, 1); - - /* Set up min, max, and cur for interrupt handling */ - fmax = (rgvmodectl & MEMMODE_FMAX_MASK) >> MEMMODE_FMAX_SHIFT; - fmin = (rgvmodectl & MEMMODE_FMIN_MASK); - fstart = (rgvmodectl & MEMMODE_FSTART_MASK) >> - MEMMODE_FSTART_SHIFT; - - vstart = (intel_uncore_read(uncore, PXVFREQ(fstart)) & - PXVFREQ_PX_MASK) >> PXVFREQ_PX_SHIFT; - - dev_priv->ips.fmax = fmax; /* IPS callback will increase this */ - dev_priv->ips.fstart = fstart; - - dev_priv->ips.max_delay = fstart; - dev_priv->ips.min_delay = fmin; - dev_priv->ips.cur_delay = fstart; - - DRM_DEBUG_DRIVER("fmax: %d, fmin: %d, fstart: %d\n", - fmax, fmin, fstart); - - intel_uncore_write(uncore, - MEMINTREN, - MEMINT_CX_SUPR_EN | MEMINT_EVAL_CHG_EN); - - /* - * Interrupts will be enabled in ironlake_irq_postinstall - */ - - intel_uncore_write(uncore, VIDSTART, vstart); - intel_uncore_posting_read(uncore, VIDSTART); - - rgvmodectl |= MEMMODE_SWMODE_EN; - intel_uncore_write(uncore, MEMMODECTL, rgvmodectl); - - if (wait_for_atomic((intel_uncore_read(uncore, MEMSWCTL) & - MEMCTL_CMD_STS) == 0, 10)) - DRM_ERROR("stuck trying to change perf mode\n"); - mdelay(1); - - ironlake_set_drps(dev_priv, fstart); - - dev_priv->ips.last_count1 = - intel_uncore_read(uncore, DMIEC) + - intel_uncore_read(uncore, DDREC) + - intel_uncore_read(uncore, CSIEC); - dev_priv->ips.last_time1 = jiffies_to_msecs(jiffies); - dev_priv->ips.last_count2 = intel_uncore_read(uncore, GFXEC); - dev_priv->ips.last_time2 = ktime_get_raw_ns(); - - spin_unlock_irq(&mchdev_lock); -} - -static void ironlake_disable_drps(struct drm_i915_private *i915) -{ - struct intel_uncore *uncore = &i915->uncore; - u16 rgvswctl; - - spin_lock_irq(&mchdev_lock); - - rgvswctl = intel_uncore_read16(uncore, MEMSWCTL); - - /* Ack interrupts, disable EFC interrupt */ - intel_uncore_write(uncore, - MEMINTREN, - intel_uncore_read(uncore, MEMINTREN) & - ~MEMINT_EVAL_CHG_EN); - intel_uncore_write(uncore, MEMINTRSTS, MEMINT_EVAL_CHG); - intel_uncore_write(uncore, - DEIER, - intel_uncore_read(uncore, DEIER) & ~DE_PCU_EVENT); - intel_uncore_write(uncore, DEIIR, DE_PCU_EVENT); - intel_uncore_write(uncore, - DEIMR, - intel_uncore_read(uncore, DEIMR) | DE_PCU_EVENT); - - /* Go back to the starting frequency */ - ironlake_set_drps(i915, i915->ips.fstart); - mdelay(1); - rgvswctl |= MEMCTL_CMD_STS; - intel_uncore_write(uncore, MEMSWCTL, rgvswctl); - mdelay(1); - - spin_unlock_irq(&mchdev_lock); -} - -/* There's a funny hw issue where the hw returns all 0 when reading from - * GEN6_RP_INTERRUPT_LIMITS. Hence we always need to compute the desired value - * ourselves, instead of doing a rmw cycle (which might result in us clearing - * all limits and the gpu stuck at whatever frequency it is at atm). - */ -static u32 intel_rps_limits(struct drm_i915_private *dev_priv, u8 val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - u32 limits; - - /* Only set the down limit when we've reached the lowest level to avoid - * getting more interrupts, otherwise leave this clear. This prevents a - * race in the hw when coming out of rc6: There's a tiny window where - * the hw runs at the minimal clock before selecting the desired - * frequency, if the down threshold expires in that window we will not - * receive a down interrupt. */ - if (INTEL_GEN(dev_priv) >= 9) { - limits = (rps->max_freq_softlimit) << 23; - if (val <= rps->min_freq_softlimit) - limits |= (rps->min_freq_softlimit) << 14; - } else { - limits = rps->max_freq_softlimit << 24; - if (val <= rps->min_freq_softlimit) - limits |= rps->min_freq_softlimit << 16; - } - - return limits; -} - -static void rps_set_power(struct drm_i915_private *dev_priv, int new_power) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - u32 threshold_up = 0, threshold_down = 0; /* in % */ - u32 ei_up = 0, ei_down = 0; - - lockdep_assert_held(&rps->power.mutex); - - if (new_power == rps->power.mode) - return; - - /* Note the units here are not exactly 1us, but 1280ns. */ - switch (new_power) { - case LOW_POWER: - /* Upclock if more than 95% busy over 16ms */ - ei_up = 16000; - threshold_up = 95; - - /* Downclock if less than 85% busy over 32ms */ - ei_down = 32000; - threshold_down = 85; - break; - - case BETWEEN: - /* Upclock if more than 90% busy over 13ms */ - ei_up = 13000; - threshold_up = 90; - - /* Downclock if less than 75% busy over 32ms */ - ei_down = 32000; - threshold_down = 75; - break; - - case HIGH_POWER: - /* Upclock if more than 85% busy over 10ms */ - ei_up = 10000; - threshold_up = 85; - - /* Downclock if less than 60% busy over 32ms */ - ei_down = 32000; - threshold_down = 60; - break; - } - - /* When byt can survive without system hang with dynamic - * sw freq adjustments, this restriction can be lifted. - */ - if (IS_VALLEYVIEW(dev_priv)) - goto skip_hw_write; - - I915_WRITE(GEN6_RP_UP_EI, - GT_INTERVAL_FROM_US(dev_priv, ei_up)); - I915_WRITE(GEN6_RP_UP_THRESHOLD, - GT_INTERVAL_FROM_US(dev_priv, - ei_up * threshold_up / 100)); - - I915_WRITE(GEN6_RP_DOWN_EI, - GT_INTERVAL_FROM_US(dev_priv, ei_down)); - I915_WRITE(GEN6_RP_DOWN_THRESHOLD, - GT_INTERVAL_FROM_US(dev_priv, - ei_down * threshold_down / 100)); - - I915_WRITE(GEN6_RP_CONTROL, - (INTEL_GEN(dev_priv) > 9 ? 0 : GEN6_RP_MEDIA_TURBO) | - GEN6_RP_MEDIA_HW_NORMAL_MODE | - GEN6_RP_MEDIA_IS_GFX | - GEN6_RP_ENABLE | - GEN6_RP_UP_BUSY_AVG | - GEN6_RP_DOWN_IDLE_AVG); - -skip_hw_write: - rps->power.mode = new_power; - rps->power.up_threshold = threshold_up; - rps->power.down_threshold = threshold_down; -} - -static void gen6_set_rps_thresholds(struct drm_i915_private *dev_priv, u8 val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - int new_power; - - new_power = rps->power.mode; - switch (rps->power.mode) { - case LOW_POWER: - if (val > rps->efficient_freq + 1 && - val > rps->cur_freq) - new_power = BETWEEN; - break; - - case BETWEEN: - if (val <= rps->efficient_freq && - val < rps->cur_freq) - new_power = LOW_POWER; - else if (val >= rps->rp0_freq && - val > rps->cur_freq) - new_power = HIGH_POWER; - break; - - case HIGH_POWER: - if (val < (rps->rp1_freq + rps->rp0_freq) >> 1 && - val < rps->cur_freq) - new_power = BETWEEN; - break; - } - /* Max/min bins are special */ - if (val <= rps->min_freq_softlimit) - new_power = LOW_POWER; - if (val >= rps->max_freq_softlimit) - new_power = HIGH_POWER; - - mutex_lock(&rps->power.mutex); - if (rps->power.interactive) - new_power = HIGH_POWER; - rps_set_power(dev_priv, new_power); - mutex_unlock(&rps->power.mutex); -} - -void intel_rps_mark_interactive(struct drm_i915_private *i915, bool interactive) -{ - struct intel_rps *rps = &i915->gt_pm.rps; - - if (INTEL_GEN(i915) < 6) - return; - - mutex_lock(&rps->power.mutex); - if (interactive) { - if (!rps->power.interactive++ && READ_ONCE(i915->gt.awake)) - rps_set_power(i915, HIGH_POWER); - } else { - GEM_BUG_ON(!rps->power.interactive); - rps->power.interactive--; - } - mutex_unlock(&rps->power.mutex); -} - -static u32 gen6_rps_pm_mask(struct drm_i915_private *dev_priv, u8 val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - u32 mask = 0; - - /* We use UP_EI_EXPIRED interupts for both up/down in manual mode */ - if (val > rps->min_freq_softlimit) - mask |= GEN6_PM_RP_UP_EI_EXPIRED | GEN6_PM_RP_DOWN_THRESHOLD | GEN6_PM_RP_DOWN_TIMEOUT; - if (val < rps->max_freq_softlimit) - mask |= GEN6_PM_RP_UP_EI_EXPIRED | GEN6_PM_RP_UP_THRESHOLD; - - mask &= dev_priv->pm_rps_events; - - return gen6_sanitize_rps_pm_mask(dev_priv, ~mask); -} - -/* gen6_set_rps is called to update the frequency request, but should also be - * called when the range (min_delay and max_delay) is modified so that we can - * update the GEN6_RP_INTERRUPT_LIMITS register accordingly. */ -static int gen6_set_rps(struct drm_i915_private *dev_priv, u8 val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - /* min/max delay may still have been modified so be sure to - * write the limits value. - */ - if (val != rps->cur_freq) { - gen6_set_rps_thresholds(dev_priv, val); - - if (INTEL_GEN(dev_priv) >= 9) - I915_WRITE(GEN6_RPNSWREQ, - GEN9_FREQUENCY(val)); - else if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) - I915_WRITE(GEN6_RPNSWREQ, - HSW_FREQUENCY(val)); - else - I915_WRITE(GEN6_RPNSWREQ, - GEN6_FREQUENCY(val) | - GEN6_OFFSET(0) | - GEN6_AGGRESSIVE_TURBO); - } - - /* Make sure we continue to get interrupts - * until we hit the minimum or maximum frequencies. - */ - I915_WRITE(GEN6_RP_INTERRUPT_LIMITS, intel_rps_limits(dev_priv, val)); - I915_WRITE(GEN6_PMINTRMSK, gen6_rps_pm_mask(dev_priv, val)); - - rps->cur_freq = val; - trace_intel_gpu_freq_change(intel_gpu_freq(dev_priv, val)); - - return 0; -} - -static int valleyview_set_rps(struct drm_i915_private *dev_priv, u8 val) -{ - int err; - - if (WARN_ONCE(IS_CHERRYVIEW(dev_priv) && (val & 1), - "Odd GPU freq value\n")) - val &= ~1; - - I915_WRITE(GEN6_PMINTRMSK, gen6_rps_pm_mask(dev_priv, val)); - - if (val != dev_priv->gt_pm.rps.cur_freq) { - vlv_punit_get(dev_priv); - err = vlv_punit_write(dev_priv, PUNIT_REG_GPU_FREQ_REQ, val); - vlv_punit_put(dev_priv); - if (err) - return err; - - gen6_set_rps_thresholds(dev_priv, val); - } - - dev_priv->gt_pm.rps.cur_freq = val; - trace_intel_gpu_freq_change(intel_gpu_freq(dev_priv, val)); - - return 0; -} - -/* vlv_set_rps_idle: Set the frequency to idle, if Gfx clocks are down - * - * * If Gfx is Idle, then - * 1. Forcewake Media well. - * 2. Request idle freq. - * 3. Release Forcewake of Media well. -*/ -static void vlv_set_rps_idle(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - u32 val = rps->idle_freq; - int err; - - if (rps->cur_freq <= val) - return; - - /* The punit delays the write of the frequency and voltage until it - * determines the GPU is awake. During normal usage we don't want to - * waste power changing the frequency if the GPU is sleeping (rc6). - * However, the GPU and driver is now idle and we do not want to delay - * switching to minimum voltage (reducing power whilst idle) as we do - * not expect to be woken in the near future and so must flush the - * change by waking the device. - * - * We choose to take the media powerwell (either would do to trick the - * punit into committing the voltage change) as that takes a lot less - * power than the render powerwell. - */ - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_MEDIA); - err = valleyview_set_rps(dev_priv, val); - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_MEDIA); - - if (err) - DRM_ERROR("Failed to set RPS for idle\n"); -} - -void gen6_rps_busy(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - mutex_lock(&rps->lock); - if (rps->enabled) { - u8 freq; - - if (dev_priv->pm_rps_events & GEN6_PM_RP_UP_EI_EXPIRED) - gen6_rps_reset_ei(dev_priv); - I915_WRITE(GEN6_PMINTRMSK, - gen6_rps_pm_mask(dev_priv, rps->cur_freq)); - - gen6_enable_rps_interrupts(dev_priv); - - /* Use the user's desired frequency as a guide, but for better - * performance, jump directly to RPe as our starting frequency. - */ - freq = max(rps->cur_freq, - rps->efficient_freq); - - if (intel_set_rps(dev_priv, - clamp(freq, - rps->min_freq_softlimit, - rps->max_freq_softlimit))) - DRM_DEBUG_DRIVER("Failed to set idle frequency\n"); - } - mutex_unlock(&rps->lock); -} - -void gen6_rps_idle(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - /* Flush our bottom-half so that it does not race with us - * setting the idle frequency and so that it is bounded by - * our rpm wakeref. And then disable the interrupts to stop any - * futher RPS reclocking whilst we are asleep. - */ - gen6_disable_rps_interrupts(dev_priv); - - mutex_lock(&rps->lock); - if (rps->enabled) { - if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) - vlv_set_rps_idle(dev_priv); - else - gen6_set_rps(dev_priv, rps->idle_freq); - rps->last_adj = 0; - I915_WRITE(GEN6_PMINTRMSK, - gen6_sanitize_rps_pm_mask(dev_priv, ~0)); - } - mutex_unlock(&rps->lock); -} - -void gen6_rps_boost(struct i915_request *rq) -{ - struct intel_rps *rps = &rq->i915->gt_pm.rps; - unsigned long flags; - bool boost; - - /* This is intentionally racy! We peek at the state here, then - * validate inside the RPS worker. - */ - if (!rps->enabled) - return; - - if (i915_request_signaled(rq)) - return; - - /* Serializes with i915_request_retire() */ - boost = false; - spin_lock_irqsave(&rq->lock, flags); - if (!i915_request_has_waitboost(rq) && - !dma_fence_is_signaled_locked(&rq->fence)) { - boost = !atomic_fetch_inc(&rps->num_waiters); - rq->flags |= I915_REQUEST_WAITBOOST; - } - spin_unlock_irqrestore(&rq->lock, flags); - if (!boost) - return; - - if (READ_ONCE(rps->cur_freq) < rps->boost_freq) - schedule_work(&rps->work); - - atomic_inc(&rps->boosts); -} - -int intel_set_rps(struct drm_i915_private *dev_priv, u8 val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - int err; - - lockdep_assert_held(&rps->lock); - GEM_BUG_ON(val > rps->max_freq); - GEM_BUG_ON(val < rps->min_freq); - - if (!rps->enabled) { - rps->cur_freq = val; - return 0; - } - - if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) - err = valleyview_set_rps(dev_priv, val); - else - err = gen6_set_rps(dev_priv, val); - - return err; -} - -static void gen9_disable_rc6(struct drm_i915_private *dev_priv) -{ - I915_WRITE(GEN6_RC_CONTROL, 0); - I915_WRITE(GEN9_PG_ENABLE, 0); -} - -static void gen9_disable_rps(struct drm_i915_private *dev_priv) -{ - I915_WRITE(GEN6_RP_CONTROL, 0); -} - -static void gen6_disable_rc6(struct drm_i915_private *dev_priv) -{ - I915_WRITE(GEN6_RC_CONTROL, 0); -} - -static void gen6_disable_rps(struct drm_i915_private *dev_priv) -{ - I915_WRITE(GEN6_RPNSWREQ, 1 << 31); - I915_WRITE(GEN6_RP_CONTROL, 0); -} - -static void cherryview_disable_rc6(struct drm_i915_private *dev_priv) -{ - I915_WRITE(GEN6_RC_CONTROL, 0); -} - -static void cherryview_disable_rps(struct drm_i915_private *dev_priv) -{ - I915_WRITE(GEN6_RP_CONTROL, 0); -} - -static void valleyview_disable_rc6(struct drm_i915_private *dev_priv) -{ - /* We're doing forcewake before Disabling RC6, - * This what the BIOS expects when going into suspend */ - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - I915_WRITE(GEN6_RC_CONTROL, 0); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void valleyview_disable_rps(struct drm_i915_private *dev_priv) -{ - I915_WRITE(GEN6_RP_CONTROL, 0); -} - -static bool bxt_check_bios_rc6_setup(struct drm_i915_private *dev_priv) -{ - bool enable_rc6 = true; - unsigned long rc6_ctx_base; - u32 rc_ctl; - int rc_sw_target; - - rc_ctl = I915_READ(GEN6_RC_CONTROL); - rc_sw_target = (I915_READ(GEN6_RC_STATE) & RC_SW_TARGET_STATE_MASK) >> - RC_SW_TARGET_STATE_SHIFT; - DRM_DEBUG_DRIVER("BIOS enabled RC states: " - "HW_CTRL %s HW_RC6 %s SW_TARGET_STATE %x\n", - onoff(rc_ctl & GEN6_RC_CTL_HW_ENABLE), - onoff(rc_ctl & GEN6_RC_CTL_RC6_ENABLE), - rc_sw_target); - - if (!(I915_READ(RC6_LOCATION) & RC6_CTX_IN_DRAM)) { - DRM_DEBUG_DRIVER("RC6 Base location not set properly.\n"); - enable_rc6 = false; - } - - /* - * The exact context size is not known for BXT, so assume a page size - * for this check. - */ - rc6_ctx_base = I915_READ(RC6_CTX_BASE) & RC6_CTX_BASE_MASK; - if (!((rc6_ctx_base >= dev_priv->dsm_reserved.start) && - (rc6_ctx_base + PAGE_SIZE < dev_priv->dsm_reserved.end))) { - DRM_DEBUG_DRIVER("RC6 Base address not as expected.\n"); - enable_rc6 = false; - } - - if (!(((I915_READ(PWRCTX_MAXCNT_RCSUNIT) & IDLE_TIME_MASK) > 1) && - ((I915_READ(PWRCTX_MAXCNT_VCSUNIT0) & IDLE_TIME_MASK) > 1) && - ((I915_READ(PWRCTX_MAXCNT_BCSUNIT) & IDLE_TIME_MASK) > 1) && - ((I915_READ(PWRCTX_MAXCNT_VECSUNIT) & IDLE_TIME_MASK) > 1))) { - DRM_DEBUG_DRIVER("Engine Idle wait time not set properly.\n"); - enable_rc6 = false; - } - - if (!I915_READ(GEN8_PUSHBUS_CONTROL) || - !I915_READ(GEN8_PUSHBUS_ENABLE) || - !I915_READ(GEN8_PUSHBUS_SHIFT)) { - DRM_DEBUG_DRIVER("Pushbus not setup properly.\n"); - enable_rc6 = false; - } - - if (!I915_READ(GEN6_GFXPAUSE)) { - DRM_DEBUG_DRIVER("GFX pause not setup properly.\n"); - enable_rc6 = false; - } - - if (!I915_READ(GEN8_MISC_CTRL0)) { - DRM_DEBUG_DRIVER("GPM control not setup properly.\n"); - enable_rc6 = false; - } - - return enable_rc6; -} - -static bool sanitize_rc6(struct drm_i915_private *i915) -{ - struct intel_device_info *info = mkwrite_device_info(i915); - - /* Powersaving is controlled by the host when inside a VM */ - if (intel_vgpu_active(i915)) { - info->has_rc6 = 0; - info->has_rps = false; - } - - if (info->has_rc6 && - IS_GEN9_LP(i915) && !bxt_check_bios_rc6_setup(i915)) { - DRM_INFO("RC6 disabled by BIOS\n"); - info->has_rc6 = 0; - } - - /* - * We assume that we do not have any deep rc6 levels if we don't have - * have the previous rc6 level supported, i.e. we use HAS_RC6() - * as the initial coarse check for rc6 in general, moving on to - * progressively finer/deeper levels. - */ - if (!info->has_rc6 && info->has_rc6p) - info->has_rc6p = 0; - - return info->has_rc6; -} - -static void gen6_init_rps_frequencies(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - /* All of these values are in units of 50MHz */ - - /* static values from HW: RP0 > RP1 > RPn (min_freq) */ - if (IS_GEN9_LP(dev_priv)) { - u32 rp_state_cap = I915_READ(BXT_RP_STATE_CAP); - rps->rp0_freq = (rp_state_cap >> 16) & 0xff; - rps->rp1_freq = (rp_state_cap >> 8) & 0xff; - rps->min_freq = (rp_state_cap >> 0) & 0xff; - } else { - u32 rp_state_cap = I915_READ(GEN6_RP_STATE_CAP); - rps->rp0_freq = (rp_state_cap >> 0) & 0xff; - rps->rp1_freq = (rp_state_cap >> 8) & 0xff; - rps->min_freq = (rp_state_cap >> 16) & 0xff; - } - /* hw_max = RP0 until we check for overclocking */ - rps->max_freq = rps->rp0_freq; - - rps->efficient_freq = rps->rp1_freq; - if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv) || - IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10) { - u32 ddcc_status = 0; - - if (sandybridge_pcode_read(dev_priv, - HSW_PCODE_DYNAMIC_DUTY_CYCLE_CONTROL, - &ddcc_status, NULL) == 0) - rps->efficient_freq = - clamp_t(u8, - ((ddcc_status >> 8) & 0xff), - rps->min_freq, - rps->max_freq); - } - - if (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10) { - /* Store the frequency values in 16.66 MHZ units, which is - * the natural hardware unit for SKL - */ - rps->rp0_freq *= GEN9_FREQ_SCALER; - rps->rp1_freq *= GEN9_FREQ_SCALER; - rps->min_freq *= GEN9_FREQ_SCALER; - rps->max_freq *= GEN9_FREQ_SCALER; - rps->efficient_freq *= GEN9_FREQ_SCALER; - } -} - -static void reset_rps(struct drm_i915_private *dev_priv, - int (*set)(struct drm_i915_private *, u8)) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - u8 freq = rps->cur_freq; - - /* force a reset */ - rps->power.mode = -1; - rps->cur_freq = -1; - - if (set(dev_priv, freq)) - DRM_ERROR("Failed to reset RPS to initial values\n"); -} - -/* See the Gen9_GT_PM_Programming_Guide doc for the below */ -static void gen9_enable_rps(struct drm_i915_private *dev_priv) -{ - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* Program defaults and thresholds for RPS */ - if (IS_GEN(dev_priv, 9)) - I915_WRITE(GEN6_RC_VIDEO_FREQ, - GEN9_FREQUENCY(dev_priv->gt_pm.rps.rp1_freq)); - - /* 1 second timeout*/ - I915_WRITE(GEN6_RP_DOWN_TIMEOUT, - GT_INTERVAL_FROM_US(dev_priv, 1000000)); - - I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 0xa); - - /* Leaning on the below call to gen6_set_rps to program/setup the - * Up/Down EI & threshold registers, as well as the RP_CONTROL, - * RP_INTERRUPT_LIMITS & RPNSWREQ registers */ - reset_rps(dev_priv, gen6_set_rps); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void gen11_enable_rc6(struct drm_i915_private *dev_priv) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - - /* 1a: Software RC state - RC0 */ - I915_WRITE(GEN6_RC_STATE, 0); - - /* - * 1b: Get forcewake during program sequence. Although the driver - * hasn't enabled a state yet where we need forcewake, BIOS may have. - */ - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* 2a: Disable RC states. */ - I915_WRITE(GEN6_RC_CONTROL, 0); - - /* 2b: Program RC6 thresholds.*/ - I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85); - I915_WRITE(GEN10_MEDIA_WAKE_RATE_LIMIT, 150); - - I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ - I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ - for_each_engine(engine, dev_priv, id) - I915_WRITE(RING_MAX_IDLE(engine->mmio_base), 10); - - if (HAS_GT_UC(dev_priv)) - I915_WRITE(GUC_MAX_IDLE_COUNT, 0xA); - - I915_WRITE(GEN6_RC_SLEEP, 0); - - I915_WRITE(GEN6_RC6_THRESHOLD, 50000); /* 50/125ms per EI */ - - /* - * 2c: Program Coarse Power Gating Policies. - * - * Bspec's guidance is to use 25us (really 25 * 1280ns) here. What we - * use instead is a more conservative estimate for the maximum time - * it takes us to service a CS interrupt and submit a new ELSP - that - * is the time which the GPU is idle waiting for the CPU to select the - * next request to execute. If the idle hysteresis is less than that - * interrupt service latency, the hardware will automatically gate - * the power well and we will then incur the wake up cost on top of - * the service latency. A similar guide from plane_state is that we - * do not want the enable hysteresis to less than the wakeup latency. - * - * igt/gem_exec_nop/sequential provides a rough estimate for the - * service latency, and puts it around 10us for Broadwell (and other - * big core) and around 40us for Broxton (and other low power cores). - * [Note that for legacy ringbuffer submission, this is less than 1us!] - * However, the wakeup latency on Broxton is closer to 100us. To be - * conservative, we have to factor in a context switch on top (due - * to ksoftirqd). - */ - I915_WRITE(GEN9_MEDIA_PG_IDLE_HYSTERESIS, 250); - I915_WRITE(GEN9_RENDER_PG_IDLE_HYSTERESIS, 250); - - /* 3a: Enable RC6 */ - I915_WRITE(GEN6_RC_CONTROL, - GEN6_RC_CTL_HW_ENABLE | - GEN6_RC_CTL_RC6_ENABLE | - GEN6_RC_CTL_EI_MODE(1)); - - /* 3b: Enable Coarse Power Gating only when RC6 is enabled. */ - I915_WRITE(GEN9_PG_ENABLE, - GEN9_RENDER_PG_ENABLE | - GEN9_MEDIA_PG_ENABLE | - GEN11_MEDIA_SAMPLER_PG_ENABLE); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void gen9_enable_rc6(struct drm_i915_private *dev_priv) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - u32 rc6_mode; - - /* 1a: Software RC state - RC0 */ - I915_WRITE(GEN6_RC_STATE, 0); - - /* 1b: Get forcewake during program sequence. Although the driver - * hasn't enabled a state yet where we need forcewake, BIOS may have.*/ - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* 2a: Disable RC states. */ - I915_WRITE(GEN6_RC_CONTROL, 0); - - /* 2b: Program RC6 thresholds.*/ - if (INTEL_GEN(dev_priv) >= 10) { - I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85); - I915_WRITE(GEN10_MEDIA_WAKE_RATE_LIMIT, 150); - } else if (IS_SKYLAKE(dev_priv)) { - /* - * WaRsDoubleRc6WrlWithCoarsePowerGating:skl Doubling WRL only - * when CPG is enabled - */ - I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 108 << 16); - } else { - I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16); - } - - I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ - I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ - for_each_engine(engine, dev_priv, id) - I915_WRITE(RING_MAX_IDLE(engine->mmio_base), 10); - - if (HAS_GT_UC(dev_priv)) - I915_WRITE(GUC_MAX_IDLE_COUNT, 0xA); - - I915_WRITE(GEN6_RC_SLEEP, 0); - - /* - * 2c: Program Coarse Power Gating Policies. - * - * Bspec's guidance is to use 25us (really 25 * 1280ns) here. What we - * use instead is a more conservative estimate for the maximum time - * it takes us to service a CS interrupt and submit a new ELSP - that - * is the time which the GPU is idle waiting for the CPU to select the - * next request to execute. If the idle hysteresis is less than that - * interrupt service latency, the hardware will automatically gate - * the power well and we will then incur the wake up cost on top of - * the service latency. A similar guide from plane_state is that we - * do not want the enable hysteresis to less than the wakeup latency. - * - * igt/gem_exec_nop/sequential provides a rough estimate for the - * service latency, and puts it around 10us for Broadwell (and other - * big core) and around 40us for Broxton (and other low power cores). - * [Note that for legacy ringbuffer submission, this is less than 1us!] - * However, the wakeup latency on Broxton is closer to 100us. To be - * conservative, we have to factor in a context switch on top (due - * to ksoftirqd). - */ - I915_WRITE(GEN9_MEDIA_PG_IDLE_HYSTERESIS, 250); - I915_WRITE(GEN9_RENDER_PG_IDLE_HYSTERESIS, 250); - - /* 3a: Enable RC6 */ - I915_WRITE(GEN6_RC6_THRESHOLD, 37500); /* 37.5/125ms per EI */ - - /* WaRsUseTimeoutMode:cnl (pre-prod) */ - if (IS_CNL_REVID(dev_priv, CNL_REVID_A0, CNL_REVID_C0)) - rc6_mode = GEN7_RC_CTL_TO_MODE; - else - rc6_mode = GEN6_RC_CTL_EI_MODE(1); - - I915_WRITE(GEN6_RC_CONTROL, - GEN6_RC_CTL_HW_ENABLE | - GEN6_RC_CTL_RC6_ENABLE | - rc6_mode); - - /* - * 3b: Enable Coarse Power Gating only when RC6 is enabled. - * WaRsDisableCoarsePowerGating:skl,cnl - Render/Media PG need to be disabled with RC6. - */ - if (NEEDS_WaRsDisableCoarsePowerGating(dev_priv)) - I915_WRITE(GEN9_PG_ENABLE, 0); - else - I915_WRITE(GEN9_PG_ENABLE, - GEN9_RENDER_PG_ENABLE | GEN9_MEDIA_PG_ENABLE); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void gen8_enable_rc6(struct drm_i915_private *dev_priv) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - - /* 1a: Software RC state - RC0 */ - I915_WRITE(GEN6_RC_STATE, 0); - - /* 1b: Get forcewake during program sequence. Although the driver - * hasn't enabled a state yet where we need forcewake, BIOS may have.*/ - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* 2a: Disable RC states. */ - I915_WRITE(GEN6_RC_CONTROL, 0); - - /* 2b: Program RC6 thresholds.*/ - I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16); - I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ - I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ - for_each_engine(engine, dev_priv, id) - I915_WRITE(RING_MAX_IDLE(engine->mmio_base), 10); - I915_WRITE(GEN6_RC_SLEEP, 0); - I915_WRITE(GEN6_RC6_THRESHOLD, 625); /* 800us/1.28 for TO */ - - /* 3: Enable RC6 */ - - I915_WRITE(GEN6_RC_CONTROL, - GEN6_RC_CTL_HW_ENABLE | - GEN7_RC_CTL_TO_MODE | - GEN6_RC_CTL_RC6_ENABLE); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void gen8_enable_rps(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* 1 Program defaults and thresholds for RPS*/ - I915_WRITE(GEN6_RPNSWREQ, - HSW_FREQUENCY(rps->rp1_freq)); - I915_WRITE(GEN6_RC_VIDEO_FREQ, - HSW_FREQUENCY(rps->rp1_freq)); - /* NB: Docs say 1s, and 1000000 - which aren't equivalent */ - I915_WRITE(GEN6_RP_DOWN_TIMEOUT, 100000000 / 128); /* 1 second timeout */ - - /* Docs recommend 900MHz, and 300 MHz respectively */ - I915_WRITE(GEN6_RP_INTERRUPT_LIMITS, - rps->max_freq_softlimit << 24 | - rps->min_freq_softlimit << 16); - - I915_WRITE(GEN6_RP_UP_THRESHOLD, 7600000 / 128); /* 76ms busyness per EI, 90% */ - I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 31300000 / 128); /* 313ms busyness per EI, 70%*/ - I915_WRITE(GEN6_RP_UP_EI, 66000); /* 84.48ms, XXX: random? */ - I915_WRITE(GEN6_RP_DOWN_EI, 350000); /* 448ms, XXX: random? */ - - I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10); - - /* 2: Enable RPS */ - I915_WRITE(GEN6_RP_CONTROL, - GEN6_RP_MEDIA_TURBO | - GEN6_RP_MEDIA_HW_NORMAL_MODE | - GEN6_RP_MEDIA_IS_GFX | - GEN6_RP_ENABLE | - GEN6_RP_UP_BUSY_AVG | - GEN6_RP_DOWN_IDLE_AVG); - - reset_rps(dev_priv, gen6_set_rps); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void gen6_enable_rc6(struct drm_i915_private *dev_priv) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - u32 rc6vids, rc6_mask; - u32 gtfifodbg; - int ret; - - I915_WRITE(GEN6_RC_STATE, 0); - - /* Clear the DBG now so we don't confuse earlier errors */ - gtfifodbg = I915_READ(GTFIFODBG); - if (gtfifodbg) { - DRM_ERROR("GT fifo had a previous error %x\n", gtfifodbg); - I915_WRITE(GTFIFODBG, gtfifodbg); - } - - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* disable the counters and set deterministic thresholds */ - I915_WRITE(GEN6_RC_CONTROL, 0); - - I915_WRITE(GEN6_RC1_WAKE_RATE_LIMIT, 1000 << 16); - I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16 | 30); - I915_WRITE(GEN6_RC6pp_WAKE_RATE_LIMIT, 30); - I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000); - I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25); - - for_each_engine(engine, dev_priv, id) - I915_WRITE(RING_MAX_IDLE(engine->mmio_base), 10); - - I915_WRITE(GEN6_RC_SLEEP, 0); - I915_WRITE(GEN6_RC1e_THRESHOLD, 1000); - if (IS_IVYBRIDGE(dev_priv)) - I915_WRITE(GEN6_RC6_THRESHOLD, 125000); - else - I915_WRITE(GEN6_RC6_THRESHOLD, 50000); - I915_WRITE(GEN6_RC6p_THRESHOLD, 150000); - I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */ - - /* We don't use those on Haswell */ - rc6_mask = GEN6_RC_CTL_RC6_ENABLE; - if (HAS_RC6p(dev_priv)) - rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE; - if (HAS_RC6pp(dev_priv)) - rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE; - I915_WRITE(GEN6_RC_CONTROL, - rc6_mask | - GEN6_RC_CTL_EI_MODE(1) | - GEN6_RC_CTL_HW_ENABLE); - - rc6vids = 0; - ret = sandybridge_pcode_read(dev_priv, GEN6_PCODE_READ_RC6VIDS, - &rc6vids, NULL); - if (IS_GEN(dev_priv, 6) && ret) { - DRM_DEBUG_DRIVER("Couldn't check for BIOS workaround\n"); - } else if (IS_GEN(dev_priv, 6) && (GEN6_DECODE_RC6_VID(rc6vids & 0xff) < 450)) { - DRM_DEBUG_DRIVER("You should update your BIOS. Correcting minimum rc6 voltage (%dmV->%dmV)\n", - GEN6_DECODE_RC6_VID(rc6vids & 0xff), 450); - rc6vids &= 0xffff00; - rc6vids |= GEN6_ENCODE_RC6_VID(450); - ret = sandybridge_pcode_write(dev_priv, GEN6_PCODE_WRITE_RC6VIDS, rc6vids); - if (ret) - DRM_ERROR("Couldn't fix incorrect rc6 voltage\n"); - } - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void gen6_enable_rps(struct drm_i915_private *dev_priv) -{ - /* Here begins a magic sequence of register writes to enable - * auto-downclocking. - * - * Perhaps there might be some value in exposing these to - * userspace... - */ - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* Power down if completely idle for over 50ms */ - I915_WRITE(GEN6_RP_DOWN_TIMEOUT, 50000); - I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10); - - reset_rps(dev_priv, gen6_set_rps); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void gen6_update_ring_freq(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - const int min_freq = 15; - const int scaling_factor = 180; - unsigned int gpu_freq; - unsigned int max_ia_freq, min_ring_freq; - unsigned int max_gpu_freq, min_gpu_freq; - struct cpufreq_policy *policy; - - lockdep_assert_held(&rps->lock); - - if (rps->max_freq <= rps->min_freq) - return; - - policy = cpufreq_cpu_get(0); - if (policy) { - max_ia_freq = policy->cpuinfo.max_freq; - cpufreq_cpu_put(policy); - } else { - /* - * Default to measured freq if none found, PCU will ensure we - * don't go over - */ - max_ia_freq = tsc_khz; - } - - /* Convert from kHz to MHz */ - max_ia_freq /= 1000; - - min_ring_freq = I915_READ(DCLK) & 0xf; - /* convert DDR frequency from units of 266.6MHz to bandwidth */ - min_ring_freq = mult_frac(min_ring_freq, 8, 3); - - min_gpu_freq = rps->min_freq; - max_gpu_freq = rps->max_freq; - if (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10) { - /* Convert GT frequency to 50 HZ units */ - min_gpu_freq /= GEN9_FREQ_SCALER; - max_gpu_freq /= GEN9_FREQ_SCALER; - } - - /* - * For each potential GPU frequency, load a ring frequency we'd like - * to use for memory access. We do this by specifying the IA frequency - * the PCU should use as a reference to determine the ring frequency. - */ - for (gpu_freq = max_gpu_freq; gpu_freq >= min_gpu_freq; gpu_freq--) { - const int diff = max_gpu_freq - gpu_freq; - unsigned int ia_freq = 0, ring_freq = 0; - - if (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10) { - /* - * ring_freq = 2 * GT. ring_freq is in 100MHz units - * No floor required for ring frequency on SKL. - */ - ring_freq = gpu_freq; - } else if (INTEL_GEN(dev_priv) >= 8) { - /* max(2 * GT, DDR). NB: GT is 50MHz units */ - ring_freq = max(min_ring_freq, gpu_freq); - } else if (IS_HASWELL(dev_priv)) { - ring_freq = mult_frac(gpu_freq, 5, 4); - ring_freq = max(min_ring_freq, ring_freq); - /* leave ia_freq as the default, chosen by cpufreq */ - } else { - /* On older processors, there is no separate ring - * clock domain, so in order to boost the bandwidth - * of the ring, we need to upclock the CPU (ia_freq). - * - * For GPU frequencies less than 750MHz, - * just use the lowest ring freq. - */ - if (gpu_freq < min_freq) - ia_freq = 800; - else - ia_freq = max_ia_freq - ((diff * scaling_factor) / 2); - ia_freq = DIV_ROUND_CLOSEST(ia_freq, 100); - } - - sandybridge_pcode_write(dev_priv, - GEN6_PCODE_WRITE_MIN_FREQ_TABLE, - ia_freq << GEN6_PCODE_FREQ_IA_RATIO_SHIFT | - ring_freq << GEN6_PCODE_FREQ_RING_RATIO_SHIFT | - gpu_freq); - } -} - -static int cherryview_rps_max_freq(struct drm_i915_private *dev_priv) -{ - u32 val, rp0; - - val = vlv_punit_read(dev_priv, FB_GFX_FMAX_AT_VMAX_FUSE); - - switch (RUNTIME_INFO(dev_priv)->sseu.eu_total) { - case 8: - /* (2 * 4) config */ - rp0 = (val >> FB_GFX_FMAX_AT_VMAX_2SS4EU_FUSE_SHIFT); - break; - case 12: - /* (2 * 6) config */ - rp0 = (val >> FB_GFX_FMAX_AT_VMAX_2SS6EU_FUSE_SHIFT); - break; - case 16: - /* (2 * 8) config */ - default: - /* Setting (2 * 8) Min RP0 for any other combination */ - rp0 = (val >> FB_GFX_FMAX_AT_VMAX_2SS8EU_FUSE_SHIFT); - break; - } - - rp0 = (rp0 & FB_GFX_FREQ_FUSE_MASK); - - return rp0; -} - -static int cherryview_rps_rpe_freq(struct drm_i915_private *dev_priv) -{ - u32 val, rpe; - - val = vlv_punit_read(dev_priv, PUNIT_GPU_DUTYCYCLE_REG); - rpe = (val >> PUNIT_GPU_DUTYCYCLE_RPE_FREQ_SHIFT) & PUNIT_GPU_DUTYCYCLE_RPE_FREQ_MASK; - - return rpe; -} - -static int cherryview_rps_guar_freq(struct drm_i915_private *dev_priv) -{ - u32 val, rp1; - - val = vlv_punit_read(dev_priv, FB_GFX_FMAX_AT_VMAX_FUSE); - rp1 = (val & FB_GFX_FREQ_FUSE_MASK); - - return rp1; -} - -static u32 cherryview_rps_min_freq(struct drm_i915_private *dev_priv) -{ - u32 val, rpn; - - val = vlv_punit_read(dev_priv, FB_GFX_FMIN_AT_VMIN_FUSE); - rpn = ((val >> FB_GFX_FMIN_AT_VMIN_FUSE_SHIFT) & - FB_GFX_FREQ_FUSE_MASK); - - return rpn; -} - -static int valleyview_rps_guar_freq(struct drm_i915_private *dev_priv) -{ - u32 val, rp1; - - val = vlv_nc_read(dev_priv, IOSF_NC_FB_GFX_FREQ_FUSE); - - rp1 = (val & FB_GFX_FGUARANTEED_FREQ_FUSE_MASK) >> FB_GFX_FGUARANTEED_FREQ_FUSE_SHIFT; - - return rp1; -} - -static int valleyview_rps_max_freq(struct drm_i915_private *dev_priv) -{ - u32 val, rp0; - - val = vlv_nc_read(dev_priv, IOSF_NC_FB_GFX_FREQ_FUSE); - - rp0 = (val & FB_GFX_MAX_FREQ_FUSE_MASK) >> FB_GFX_MAX_FREQ_FUSE_SHIFT; - /* Clamp to max */ - rp0 = min_t(u32, rp0, 0xea); - - return rp0; -} - -static int valleyview_rps_rpe_freq(struct drm_i915_private *dev_priv) -{ - u32 val, rpe; - - val = vlv_nc_read(dev_priv, IOSF_NC_FB_GFX_FMAX_FUSE_LO); - rpe = (val & FB_FMAX_VMIN_FREQ_LO_MASK) >> FB_FMAX_VMIN_FREQ_LO_SHIFT; - val = vlv_nc_read(dev_priv, IOSF_NC_FB_GFX_FMAX_FUSE_HI); - rpe |= (val & FB_FMAX_VMIN_FREQ_HI_MASK) << 5; - - return rpe; -} - -static int valleyview_rps_min_freq(struct drm_i915_private *dev_priv) -{ - u32 val; - - val = vlv_punit_read(dev_priv, PUNIT_REG_GPU_LFM) & 0xff; - /* - * According to the BYT Punit GPU turbo HAS 1.1.6.3 the minimum value - * for the minimum frequency in GPLL mode is 0xc1. Contrary to this on - * a BYT-M B0 the above register contains 0xbf. Moreover when setting - * a frequency Punit will not allow values below 0xc0. Clamp it 0xc0 - * to make sure it matches what Punit accepts. - */ - return max_t(u32, val, 0xc0); -} - -/* Check that the pctx buffer wasn't move under us. */ -static void valleyview_check_pctx(struct drm_i915_private *dev_priv) -{ - unsigned long pctx_addr = I915_READ(VLV_PCBR) & ~4095; - - WARN_ON(pctx_addr != dev_priv->dsm.start + - dev_priv->vlv_pctx->stolen->start); -} - - -/* Check that the pcbr address is not empty. */ -static void cherryview_check_pctx(struct drm_i915_private *dev_priv) -{ - unsigned long pctx_addr = I915_READ(VLV_PCBR) & ~4095; - - WARN_ON((pctx_addr >> VLV_PCBR_ADDR_SHIFT) == 0); -} - -static void cherryview_setup_pctx(struct drm_i915_private *dev_priv) -{ - resource_size_t pctx_paddr, paddr; - resource_size_t pctx_size = 32*1024; - u32 pcbr; - - pcbr = I915_READ(VLV_PCBR); - if ((pcbr >> VLV_PCBR_ADDR_SHIFT) == 0) { - DRM_DEBUG_DRIVER("BIOS didn't set up PCBR, fixing up\n"); - paddr = dev_priv->dsm.end + 1 - pctx_size; - GEM_BUG_ON(paddr > U32_MAX); - - pctx_paddr = (paddr & (~4095)); - I915_WRITE(VLV_PCBR, pctx_paddr); - } - - DRM_DEBUG_DRIVER("PCBR: 0x%08x\n", I915_READ(VLV_PCBR)); -} - -static void valleyview_setup_pctx(struct drm_i915_private *dev_priv) -{ - struct drm_i915_gem_object *pctx; - resource_size_t pctx_paddr; - resource_size_t pctx_size = 24*1024; - u32 pcbr; - - pcbr = I915_READ(VLV_PCBR); - if (pcbr) { - /* BIOS set it up already, grab the pre-alloc'd space */ - resource_size_t pcbr_offset; - - pcbr_offset = (pcbr & (~4095)) - dev_priv->dsm.start; - pctx = i915_gem_object_create_stolen_for_preallocated(dev_priv, - pcbr_offset, - I915_GTT_OFFSET_NONE, - pctx_size); - goto out; - } - - DRM_DEBUG_DRIVER("BIOS didn't set up PCBR, fixing up\n"); - - /* - * From the Gunit register HAS: - * The Gfx driver is expected to program this register and ensure - * proper allocation within Gfx stolen memory. For example, this - * register should be programmed such than the PCBR range does not - * overlap with other ranges, such as the frame buffer, protected - * memory, or any other relevant ranges. - */ - pctx = i915_gem_object_create_stolen(dev_priv, pctx_size); - if (!pctx) { - DRM_DEBUG("not enough stolen space for PCTX, disabling\n"); - goto out; - } - - GEM_BUG_ON(range_overflows_t(u64, - dev_priv->dsm.start, - pctx->stolen->start, - U32_MAX)); - pctx_paddr = dev_priv->dsm.start + pctx->stolen->start; - I915_WRITE(VLV_PCBR, pctx_paddr); - -out: - DRM_DEBUG_DRIVER("PCBR: 0x%08x\n", I915_READ(VLV_PCBR)); - dev_priv->vlv_pctx = pctx; -} - -static void valleyview_cleanup_pctx(struct drm_i915_private *dev_priv) -{ - struct drm_i915_gem_object *pctx; - - pctx = fetch_and_zero(&dev_priv->vlv_pctx); - if (pctx) - i915_gem_object_put(pctx); -} - -static void vlv_init_gpll_ref_freq(struct drm_i915_private *dev_priv) -{ - dev_priv->gt_pm.rps.gpll_ref_freq = - vlv_get_cck_clock(dev_priv, "GPLL ref", - CCK_GPLL_CLOCK_CONTROL, - dev_priv->czclk_freq); - - DRM_DEBUG_DRIVER("GPLL reference freq: %d kHz\n", - dev_priv->gt_pm.rps.gpll_ref_freq); -} - -static void valleyview_init_gt_powersave(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - u32 val; - - valleyview_setup_pctx(dev_priv); - - vlv_iosf_sb_get(dev_priv, - BIT(VLV_IOSF_SB_PUNIT) | - BIT(VLV_IOSF_SB_NC) | - BIT(VLV_IOSF_SB_CCK)); - - vlv_init_gpll_ref_freq(dev_priv); - - val = vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS); - switch ((val >> 6) & 3) { - case 0: - case 1: - dev_priv->mem_freq = 800; - break; - case 2: - dev_priv->mem_freq = 1066; - break; - case 3: - dev_priv->mem_freq = 1333; - break; - } - DRM_DEBUG_DRIVER("DDR speed: %d MHz\n", dev_priv->mem_freq); - - rps->max_freq = valleyview_rps_max_freq(dev_priv); - rps->rp0_freq = rps->max_freq; - DRM_DEBUG_DRIVER("max GPU freq: %d MHz (%u)\n", - intel_gpu_freq(dev_priv, rps->max_freq), - rps->max_freq); - - rps->efficient_freq = valleyview_rps_rpe_freq(dev_priv); - DRM_DEBUG_DRIVER("RPe GPU freq: %d MHz (%u)\n", - intel_gpu_freq(dev_priv, rps->efficient_freq), - rps->efficient_freq); - - rps->rp1_freq = valleyview_rps_guar_freq(dev_priv); - DRM_DEBUG_DRIVER("RP1(Guar Freq) GPU freq: %d MHz (%u)\n", - intel_gpu_freq(dev_priv, rps->rp1_freq), - rps->rp1_freq); - - rps->min_freq = valleyview_rps_min_freq(dev_priv); - DRM_DEBUG_DRIVER("min GPU freq: %d MHz (%u)\n", - intel_gpu_freq(dev_priv, rps->min_freq), - rps->min_freq); - - vlv_iosf_sb_put(dev_priv, - BIT(VLV_IOSF_SB_PUNIT) | - BIT(VLV_IOSF_SB_NC) | - BIT(VLV_IOSF_SB_CCK)); -} - -static void cherryview_init_gt_powersave(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - u32 val; - - cherryview_setup_pctx(dev_priv); - - vlv_iosf_sb_get(dev_priv, - BIT(VLV_IOSF_SB_PUNIT) | - BIT(VLV_IOSF_SB_NC) | - BIT(VLV_IOSF_SB_CCK)); - - vlv_init_gpll_ref_freq(dev_priv); - - val = vlv_cck_read(dev_priv, CCK_FUSE_REG); - - switch ((val >> 2) & 0x7) { - case 3: - dev_priv->mem_freq = 2000; - break; - default: - dev_priv->mem_freq = 1600; - break; - } - DRM_DEBUG_DRIVER("DDR speed: %d MHz\n", dev_priv->mem_freq); - - rps->max_freq = cherryview_rps_max_freq(dev_priv); - rps->rp0_freq = rps->max_freq; - DRM_DEBUG_DRIVER("max GPU freq: %d MHz (%u)\n", - intel_gpu_freq(dev_priv, rps->max_freq), - rps->max_freq); - - rps->efficient_freq = cherryview_rps_rpe_freq(dev_priv); - DRM_DEBUG_DRIVER("RPe GPU freq: %d MHz (%u)\n", - intel_gpu_freq(dev_priv, rps->efficient_freq), - rps->efficient_freq); - - rps->rp1_freq = cherryview_rps_guar_freq(dev_priv); - DRM_DEBUG_DRIVER("RP1(Guar) GPU freq: %d MHz (%u)\n", - intel_gpu_freq(dev_priv, rps->rp1_freq), - rps->rp1_freq); - - rps->min_freq = cherryview_rps_min_freq(dev_priv); - DRM_DEBUG_DRIVER("min GPU freq: %d MHz (%u)\n", - intel_gpu_freq(dev_priv, rps->min_freq), - rps->min_freq); - - vlv_iosf_sb_put(dev_priv, - BIT(VLV_IOSF_SB_PUNIT) | - BIT(VLV_IOSF_SB_NC) | - BIT(VLV_IOSF_SB_CCK)); - - WARN_ONCE((rps->max_freq | rps->efficient_freq | rps->rp1_freq | - rps->min_freq) & 1, - "Odd GPU freq values\n"); -} - -static void valleyview_cleanup_gt_powersave(struct drm_i915_private *dev_priv) -{ - valleyview_cleanup_pctx(dev_priv); -} - -static void cherryview_enable_rc6(struct drm_i915_private *dev_priv) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - u32 gtfifodbg, rc6_mode, pcbr; - - gtfifodbg = I915_READ(GTFIFODBG) & ~(GT_FIFO_SBDEDICATE_FREE_ENTRY_CHV | - GT_FIFO_FREE_ENTRIES_CHV); - if (gtfifodbg) { - DRM_DEBUG_DRIVER("GT fifo had a previous error %x\n", - gtfifodbg); - I915_WRITE(GTFIFODBG, gtfifodbg); - } - - cherryview_check_pctx(dev_priv); - - /* 1a & 1b: Get forcewake during program sequence. Although the driver - * hasn't enabled a state yet where we need forcewake, BIOS may have.*/ - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* Disable RC states. */ - I915_WRITE(GEN6_RC_CONTROL, 0); - - /* 2a: Program RC6 thresholds.*/ - I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16); - I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ - I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ - - for_each_engine(engine, dev_priv, id) - I915_WRITE(RING_MAX_IDLE(engine->mmio_base), 10); - I915_WRITE(GEN6_RC_SLEEP, 0); - - /* TO threshold set to 500 us ( 0x186 * 1.28 us) */ - I915_WRITE(GEN6_RC6_THRESHOLD, 0x186); - - /* Allows RC6 residency counter to work */ - I915_WRITE(VLV_COUNTER_CONTROL, - _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH | - VLV_MEDIA_RC6_COUNT_EN | - VLV_RENDER_RC6_COUNT_EN)); - - /* For now we assume BIOS is allocating and populating the PCBR */ - pcbr = I915_READ(VLV_PCBR); - - /* 3: Enable RC6 */ - rc6_mode = 0; - if (pcbr >> VLV_PCBR_ADDR_SHIFT) - rc6_mode = GEN7_RC_CTL_TO_MODE; - I915_WRITE(GEN6_RC_CONTROL, rc6_mode); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void cherryview_enable_rps(struct drm_i915_private *dev_priv) -{ - u32 val; - - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* 1: Program defaults and thresholds for RPS*/ - I915_WRITE(GEN6_RP_DOWN_TIMEOUT, 1000000); - I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400); - I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000); - I915_WRITE(GEN6_RP_UP_EI, 66000); - I915_WRITE(GEN6_RP_DOWN_EI, 350000); - - I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10); - - /* 2: Enable RPS */ - I915_WRITE(GEN6_RP_CONTROL, - GEN6_RP_MEDIA_HW_NORMAL_MODE | - GEN6_RP_MEDIA_IS_GFX | - GEN6_RP_ENABLE | - GEN6_RP_UP_BUSY_AVG | - GEN6_RP_DOWN_IDLE_AVG); - - /* Setting Fixed Bias */ - vlv_punit_get(dev_priv); - - val = VLV_OVERRIDE_EN | VLV_SOC_TDP_EN | CHV_BIAS_CPU_50_SOC_50; - vlv_punit_write(dev_priv, VLV_TURBO_SOC_OVERRIDE, val); - - val = vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS); - - vlv_punit_put(dev_priv); - - /* RPS code assumes GPLL is used */ - WARN_ONCE((val & GPLLENABLE) == 0, "GPLL not enabled\n"); - - DRM_DEBUG_DRIVER("GPLL enabled? %s\n", yesno(val & GPLLENABLE)); - DRM_DEBUG_DRIVER("GPU status: 0x%08x\n", val); - - reset_rps(dev_priv, valleyview_set_rps); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void valleyview_enable_rc6(struct drm_i915_private *dev_priv) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - u32 gtfifodbg; - - valleyview_check_pctx(dev_priv); - - gtfifodbg = I915_READ(GTFIFODBG); - if (gtfifodbg) { - DRM_DEBUG_DRIVER("GT fifo had a previous error %x\n", - gtfifodbg); - I915_WRITE(GTFIFODBG, gtfifodbg); - } - - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - /* Disable RC states. */ - I915_WRITE(GEN6_RC_CONTROL, 0); - - I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 0x00280000); - I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000); - I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25); - - for_each_engine(engine, dev_priv, id) - I915_WRITE(RING_MAX_IDLE(engine->mmio_base), 10); - - I915_WRITE(GEN6_RC6_THRESHOLD, 0x557); - - /* Allows RC6 residency counter to work */ - I915_WRITE(VLV_COUNTER_CONTROL, - _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH | - VLV_MEDIA_RC0_COUNT_EN | - VLV_RENDER_RC0_COUNT_EN | - VLV_MEDIA_RC6_COUNT_EN | - VLV_RENDER_RC6_COUNT_EN)); - - I915_WRITE(GEN6_RC_CONTROL, - GEN7_RC_CTL_TO_MODE | VLV_RC_CTL_CTX_RST_PARALLEL); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static void valleyview_enable_rps(struct drm_i915_private *dev_priv) -{ - u32 val; - - intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL); - - I915_WRITE(GEN6_RP_DOWN_TIMEOUT, 1000000); - I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400); - I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000); - I915_WRITE(GEN6_RP_UP_EI, 66000); - I915_WRITE(GEN6_RP_DOWN_EI, 350000); - - I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10); - - I915_WRITE(GEN6_RP_CONTROL, - GEN6_RP_MEDIA_TURBO | - GEN6_RP_MEDIA_HW_NORMAL_MODE | - GEN6_RP_MEDIA_IS_GFX | - GEN6_RP_ENABLE | - GEN6_RP_UP_BUSY_AVG | - GEN6_RP_DOWN_IDLE_CONT); - - vlv_punit_get(dev_priv); - - /* Setting Fixed Bias */ - val = VLV_OVERRIDE_EN | VLV_SOC_TDP_EN | VLV_BIAS_CPU_125_SOC_875; - vlv_punit_write(dev_priv, VLV_TURBO_SOC_OVERRIDE, val); - - val = vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS); - - vlv_punit_put(dev_priv); - - /* RPS code assumes GPLL is used */ - WARN_ONCE((val & GPLLENABLE) == 0, "GPLL not enabled\n"); - - DRM_DEBUG_DRIVER("GPLL enabled? %s\n", yesno(val & GPLLENABLE)); - DRM_DEBUG_DRIVER("GPU status: 0x%08x\n", val); - - reset_rps(dev_priv, valleyview_set_rps); - - intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL); -} - -static unsigned long intel_pxfreq(u32 vidfreq) -{ - unsigned long freq; - int div = (vidfreq & 0x3f0000) >> 16; - int post = (vidfreq & 0x3000) >> 12; - int pre = (vidfreq & 0x7); - - if (!pre) - return 0; - - freq = ((div * 133333) / ((1<<post) * pre)); - - return freq; -} - -static const struct cparams { - u16 i; - u16 t; - u16 m; - u16 c; -} cparams[] = { - { 1, 1333, 301, 28664 }, - { 1, 1066, 294, 24460 }, - { 1, 800, 294, 25192 }, - { 0, 1333, 276, 27605 }, - { 0, 1066, 276, 27605 }, - { 0, 800, 231, 23784 }, -}; - -static unsigned long __i915_chipset_val(struct drm_i915_private *dev_priv) -{ - u64 total_count, diff, ret; - u32 count1, count2, count3, m = 0, c = 0; - unsigned long now = jiffies_to_msecs(jiffies), diff1; - int i; - - lockdep_assert_held(&mchdev_lock); - - diff1 = now - dev_priv->ips.last_time1; - - /* Prevent division-by-zero if we are asking too fast. - * Also, we don't get interesting results if we are polling - * faster than once in 10ms, so just return the saved value - * in such cases. - */ - if (diff1 <= 10) - return dev_priv->ips.chipset_power; - - count1 = I915_READ(DMIEC); - count2 = I915_READ(DDREC); - count3 = I915_READ(CSIEC); - - total_count = count1 + count2 + count3; - - /* FIXME: handle per-counter overflow */ - if (total_count < dev_priv->ips.last_count1) { - diff = ~0UL - dev_priv->ips.last_count1; - diff += total_count; - } else { - diff = total_count - dev_priv->ips.last_count1; - } - - for (i = 0; i < ARRAY_SIZE(cparams); i++) { - if (cparams[i].i == dev_priv->ips.c_m && - cparams[i].t == dev_priv->ips.r_t) { - m = cparams[i].m; - c = cparams[i].c; - break; - } - } - - diff = div_u64(diff, diff1); - ret = ((m * diff) + c); - ret = div_u64(ret, 10); - - dev_priv->ips.last_count1 = total_count; - dev_priv->ips.last_time1 = now; - - dev_priv->ips.chipset_power = ret; - - return ret; -} - -unsigned long i915_chipset_val(struct drm_i915_private *dev_priv) -{ - intel_wakeref_t wakeref; - unsigned long val = 0; - - if (!IS_GEN(dev_priv, 5)) - return 0; - - with_intel_runtime_pm(&dev_priv->runtime_pm, wakeref) { - spin_lock_irq(&mchdev_lock); - val = __i915_chipset_val(dev_priv); - spin_unlock_irq(&mchdev_lock); - } - - return val; -} - -unsigned long i915_mch_val(struct drm_i915_private *i915) -{ - unsigned long m, x, b; - u32 tsfs; - - tsfs = intel_uncore_read(&i915->uncore, TSFS); - - m = ((tsfs & TSFS_SLOPE_MASK) >> TSFS_SLOPE_SHIFT); - x = intel_uncore_read8(&i915->uncore, TR1); - - b = tsfs & TSFS_INTR_MASK; - - return ((m * x) / 127) - b; -} - -static int _pxvid_to_vd(u8 pxvid) -{ - if (pxvid == 0) - return 0; - - if (pxvid >= 8 && pxvid < 31) - pxvid = 31; - - return (pxvid + 2) * 125; -} - -static u32 pvid_to_extvid(struct drm_i915_private *dev_priv, u8 pxvid) -{ - const int vd = _pxvid_to_vd(pxvid); - const int vm = vd - 1125; - - if (INTEL_INFO(dev_priv)->is_mobile) - return vm > 0 ? vm : 0; - - return vd; -} - -static void __i915_update_gfx_val(struct drm_i915_private *dev_priv) -{ - u64 now, diff, diffms; - u32 count; - - lockdep_assert_held(&mchdev_lock); - - now = ktime_get_raw_ns(); - diffms = now - dev_priv->ips.last_time2; - do_div(diffms, NSEC_PER_MSEC); - - /* Don't divide by 0 */ - if (!diffms) - return; - - count = I915_READ(GFXEC); - - if (count < dev_priv->ips.last_count2) { - diff = ~0UL - dev_priv->ips.last_count2; - diff += count; - } else { - diff = count - dev_priv->ips.last_count2; - } - - dev_priv->ips.last_count2 = count; - dev_priv->ips.last_time2 = now; - - /* More magic constants... */ - diff = diff * 1181; - diff = div_u64(diff, diffms * 10); - dev_priv->ips.gfx_power = diff; -} - -void i915_update_gfx_val(struct drm_i915_private *dev_priv) -{ - intel_wakeref_t wakeref; - - if (!IS_GEN(dev_priv, 5)) - return; - - with_intel_runtime_pm(&dev_priv->runtime_pm, wakeref) { - spin_lock_irq(&mchdev_lock); - __i915_update_gfx_val(dev_priv); - spin_unlock_irq(&mchdev_lock); - } -} - -static unsigned long __i915_gfx_val(struct drm_i915_private *dev_priv) -{ - unsigned long t, corr, state1, corr2, state2; - u32 pxvid, ext_v; - - lockdep_assert_held(&mchdev_lock); - - pxvid = I915_READ(PXVFREQ(dev_priv->gt_pm.rps.cur_freq)); - pxvid = (pxvid >> 24) & 0x7f; - ext_v = pvid_to_extvid(dev_priv, pxvid); - - state1 = ext_v; - - t = i915_mch_val(dev_priv); - - /* Revel in the empirically derived constants */ - - /* Correction factor in 1/100000 units */ - if (t > 80) - corr = ((t * 2349) + 135940); - else if (t >= 50) - corr = ((t * 964) + 29317); - else /* < 50 */ - corr = ((t * 301) + 1004); - - corr = corr * ((150142 * state1) / 10000 - 78642); - corr /= 100000; - corr2 = (corr * dev_priv->ips.corr); - - state2 = (corr2 * state1) / 10000; - state2 /= 100; /* convert to mW */ - - __i915_update_gfx_val(dev_priv); - - return dev_priv->ips.gfx_power + state2; -} - -unsigned long i915_gfx_val(struct drm_i915_private *dev_priv) -{ - intel_wakeref_t wakeref; - unsigned long val = 0; - - if (!IS_GEN(dev_priv, 5)) - return 0; - - with_intel_runtime_pm(&dev_priv->runtime_pm, wakeref) { - spin_lock_irq(&mchdev_lock); - val = __i915_gfx_val(dev_priv); - spin_unlock_irq(&mchdev_lock); - } - - return val; -} - -static struct drm_i915_private __rcu *i915_mch_dev; - -static struct drm_i915_private *mchdev_get(void) -{ - struct drm_i915_private *i915; - - rcu_read_lock(); - i915 = rcu_dereference(i915_mch_dev); - if (!kref_get_unless_zero(&i915->drm.ref)) - i915 = NULL; - rcu_read_unlock(); - - return i915; -} - -/** - * i915_read_mch_val - return value for IPS use - * - * Calculate and return a value for the IPS driver to use when deciding whether - * we have thermal and power headroom to increase CPU or GPU power budget. - */ -unsigned long i915_read_mch_val(void) -{ - struct drm_i915_private *i915; - unsigned long chipset_val = 0; - unsigned long graphics_val = 0; - intel_wakeref_t wakeref; - - i915 = mchdev_get(); - if (!i915) - return 0; - - with_intel_runtime_pm(&i915->runtime_pm, wakeref) { - spin_lock_irq(&mchdev_lock); - chipset_val = __i915_chipset_val(i915); - graphics_val = __i915_gfx_val(i915); - spin_unlock_irq(&mchdev_lock); - } - - drm_dev_put(&i915->drm); - return chipset_val + graphics_val; -} -EXPORT_SYMBOL_GPL(i915_read_mch_val); - -/** - * i915_gpu_raise - raise GPU frequency limit - * - * Raise the limit; IPS indicates we have thermal headroom. - */ -bool i915_gpu_raise(void) -{ - struct drm_i915_private *i915; - - i915 = mchdev_get(); - if (!i915) - return false; - - spin_lock_irq(&mchdev_lock); - if (i915->ips.max_delay > i915->ips.fmax) - i915->ips.max_delay--; - spin_unlock_irq(&mchdev_lock); - - drm_dev_put(&i915->drm); - return true; -} -EXPORT_SYMBOL_GPL(i915_gpu_raise); - -/** - * i915_gpu_lower - lower GPU frequency limit - * - * IPS indicates we're close to a thermal limit, so throttle back the GPU - * frequency maximum. - */ -bool i915_gpu_lower(void) -{ - struct drm_i915_private *i915; - - i915 = mchdev_get(); - if (!i915) - return false; - - spin_lock_irq(&mchdev_lock); - if (i915->ips.max_delay < i915->ips.min_delay) - i915->ips.max_delay++; - spin_unlock_irq(&mchdev_lock); - - drm_dev_put(&i915->drm); - return true; -} -EXPORT_SYMBOL_GPL(i915_gpu_lower); - -/** - * i915_gpu_busy - indicate GPU business to IPS - * - * Tell the IPS driver whether or not the GPU is busy. - */ -bool i915_gpu_busy(void) -{ - struct drm_i915_private *i915; - bool ret; - - i915 = mchdev_get(); - if (!i915) - return false; - - ret = i915->gt.awake; - - drm_dev_put(&i915->drm); - return ret; -} -EXPORT_SYMBOL_GPL(i915_gpu_busy); - -/** - * i915_gpu_turbo_disable - disable graphics turbo - * - * Disable graphics turbo by resetting the max frequency and setting the - * current frequency to the default. - */ -bool i915_gpu_turbo_disable(void) -{ - struct drm_i915_private *i915; - bool ret; - - i915 = mchdev_get(); - if (!i915) - return false; - - spin_lock_irq(&mchdev_lock); - i915->ips.max_delay = i915->ips.fstart; - ret = ironlake_set_drps(i915, i915->ips.fstart); - spin_unlock_irq(&mchdev_lock); - - drm_dev_put(&i915->drm); - return ret; -} -EXPORT_SYMBOL_GPL(i915_gpu_turbo_disable); - -/** - * Tells the intel_ips driver that the i915 driver is now loaded, if - * IPS got loaded first. - * - * This awkward dance is so that neither module has to depend on the - * other in order for IPS to do the appropriate communication of - * GPU turbo limits to i915. - */ -static void -ips_ping_for_i915_load(void) -{ - void (*link)(void); - - link = symbol_get(ips_link_to_i915_driver); - if (link) { - link(); - symbol_put(ips_link_to_i915_driver); - } -} - -void intel_gpu_ips_init(struct drm_i915_private *dev_priv) -{ - /* We only register the i915 ips part with intel-ips once everything is - * set up, to avoid intel-ips sneaking in and reading bogus values. */ - rcu_assign_pointer(i915_mch_dev, dev_priv); - - ips_ping_for_i915_load(); -} - -void intel_gpu_ips_teardown(void) -{ - rcu_assign_pointer(i915_mch_dev, NULL); -} - -static void intel_init_emon(struct drm_i915_private *dev_priv) -{ - u32 lcfuse; - u8 pxw[16]; - int i; - - /* Disable to program */ - I915_WRITE(ECR, 0); - POSTING_READ(ECR); - - /* Program energy weights for various events */ - I915_WRITE(SDEW, 0x15040d00); - I915_WRITE(CSIEW0, 0x007f0000); - I915_WRITE(CSIEW1, 0x1e220004); - I915_WRITE(CSIEW2, 0x04000004); - - for (i = 0; i < 5; i++) - I915_WRITE(PEW(i), 0); - for (i = 0; i < 3; i++) - I915_WRITE(DEW(i), 0); - - /* Program P-state weights to account for frequency power adjustment */ - for (i = 0; i < 16; i++) { - u32 pxvidfreq = I915_READ(PXVFREQ(i)); - unsigned long freq = intel_pxfreq(pxvidfreq); - unsigned long vid = (pxvidfreq & PXVFREQ_PX_MASK) >> - PXVFREQ_PX_SHIFT; - unsigned long val; - - val = vid * vid; - val *= (freq / 1000); - val *= 255; - val /= (127*127*900); - if (val > 0xff) - DRM_ERROR("bad pxval: %ld\n", val); - pxw[i] = val; - } - /* Render standby states get 0 weight */ - pxw[14] = 0; - pxw[15] = 0; - - for (i = 0; i < 4; i++) { - u32 val = (pxw[i*4] << 24) | (pxw[(i*4)+1] << 16) | - (pxw[(i*4)+2] << 8) | (pxw[(i*4)+3]); - I915_WRITE(PXW(i), val); - } - - /* Adjust magic regs to magic values (more experimental results) */ - I915_WRITE(OGW0, 0); - I915_WRITE(OGW1, 0); - I915_WRITE(EG0, 0x00007f00); - I915_WRITE(EG1, 0x0000000e); - I915_WRITE(EG2, 0x000e0000); - I915_WRITE(EG3, 0x68000300); - I915_WRITE(EG4, 0x42000000); - I915_WRITE(EG5, 0x00140031); - I915_WRITE(EG6, 0); - I915_WRITE(EG7, 0); - - for (i = 0; i < 8; i++) - I915_WRITE(PXWL(i), 0); - - /* Enable PMON + select events */ - I915_WRITE(ECR, 0x80000019); - - lcfuse = I915_READ(LCFUSE02); - - dev_priv->ips.corr = (lcfuse & LCFUSE_HIV_MASK); -} - -void intel_init_gt_powersave(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - /* - * RPM depends on RC6 to save restore the GT HW context, so make RC6 a - * requirement. - */ - if (!sanitize_rc6(dev_priv)) { - DRM_INFO("RC6 disabled, disabling runtime PM support\n"); - pm_runtime_get(&dev_priv->drm.pdev->dev); - } - - /* Initialize RPS limits (for userspace) */ - if (IS_CHERRYVIEW(dev_priv)) - cherryview_init_gt_powersave(dev_priv); - else if (IS_VALLEYVIEW(dev_priv)) - valleyview_init_gt_powersave(dev_priv); - else if (INTEL_GEN(dev_priv) >= 6) - gen6_init_rps_frequencies(dev_priv); - - /* Derive initial user preferences/limits from the hardware limits */ - rps->max_freq_softlimit = rps->max_freq; - rps->min_freq_softlimit = rps->min_freq; - - /* After setting max-softlimit, find the overclock max freq */ - if (IS_GEN(dev_priv, 6) || - IS_IVYBRIDGE(dev_priv) || IS_HASWELL(dev_priv)) { - u32 params = 0; - - sandybridge_pcode_read(dev_priv, GEN6_READ_OC_PARAMS, - ¶ms, NULL); - if (params & BIT(31)) { /* OC supported */ - DRM_DEBUG_DRIVER("Overclocking supported, max: %dMHz, overclock: %dMHz\n", - (rps->max_freq & 0xff) * 50, - (params & 0xff) * 50); - rps->max_freq = params & 0xff; - } - } - - /* Finally allow us to boost to max by default */ - rps->boost_freq = rps->max_freq; - rps->idle_freq = rps->min_freq; - rps->cur_freq = rps->idle_freq; -} - -void intel_cleanup_gt_powersave(struct drm_i915_private *dev_priv) -{ - if (IS_VALLEYVIEW(dev_priv)) - valleyview_cleanup_gt_powersave(dev_priv); - - if (!HAS_RC6(dev_priv)) - pm_runtime_put(&dev_priv->drm.pdev->dev); -} - -void intel_sanitize_gt_powersave(struct drm_i915_private *dev_priv) -{ - dev_priv->gt_pm.rps.enabled = true; /* force RPS disabling */ - dev_priv->gt_pm.rc6.enabled = true; /* force RC6 disabling */ - intel_disable_gt_powersave(dev_priv); - - if (INTEL_GEN(dev_priv) >= 11) - gen11_reset_rps_interrupts(dev_priv); - else if (INTEL_GEN(dev_priv) >= 6) - gen6_reset_rps_interrupts(dev_priv); -} - -static inline void intel_disable_llc_pstate(struct drm_i915_private *i915) -{ - lockdep_assert_held(&i915->gt_pm.rps.lock); - - if (!i915->gt_pm.llc_pstate.enabled) - return; - - /* Currently there is no HW configuration to be done to disable. */ - - i915->gt_pm.llc_pstate.enabled = false; -} - -static void intel_disable_rc6(struct drm_i915_private *dev_priv) -{ - lockdep_assert_held(&dev_priv->gt_pm.rps.lock); - - if (!dev_priv->gt_pm.rc6.enabled) - return; - - if (INTEL_GEN(dev_priv) >= 9) - gen9_disable_rc6(dev_priv); - else if (IS_CHERRYVIEW(dev_priv)) - cherryview_disable_rc6(dev_priv); - else if (IS_VALLEYVIEW(dev_priv)) - valleyview_disable_rc6(dev_priv); - else if (INTEL_GEN(dev_priv) >= 6) - gen6_disable_rc6(dev_priv); - - dev_priv->gt_pm.rc6.enabled = false; -} - -static void intel_disable_rps(struct drm_i915_private *dev_priv) -{ - lockdep_assert_held(&dev_priv->gt_pm.rps.lock); - - if (!dev_priv->gt_pm.rps.enabled) - return; - - if (INTEL_GEN(dev_priv) >= 9) - gen9_disable_rps(dev_priv); - else if (IS_CHERRYVIEW(dev_priv)) - cherryview_disable_rps(dev_priv); - else if (IS_VALLEYVIEW(dev_priv)) - valleyview_disable_rps(dev_priv); - else if (INTEL_GEN(dev_priv) >= 6) - gen6_disable_rps(dev_priv); - else if (IS_IRONLAKE_M(dev_priv)) - ironlake_disable_drps(dev_priv); - - dev_priv->gt_pm.rps.enabled = false; -} - -void intel_disable_gt_powersave(struct drm_i915_private *dev_priv) -{ - mutex_lock(&dev_priv->gt_pm.rps.lock); - - intel_disable_rc6(dev_priv); - intel_disable_rps(dev_priv); - if (HAS_LLC(dev_priv)) - intel_disable_llc_pstate(dev_priv); - - mutex_unlock(&dev_priv->gt_pm.rps.lock); -} - -static inline void intel_enable_llc_pstate(struct drm_i915_private *i915) -{ - lockdep_assert_held(&i915->gt_pm.rps.lock); - - if (i915->gt_pm.llc_pstate.enabled) - return; - - gen6_update_ring_freq(i915); - - i915->gt_pm.llc_pstate.enabled = true; -} - -static void intel_enable_rc6(struct drm_i915_private *dev_priv) -{ - lockdep_assert_held(&dev_priv->gt_pm.rps.lock); - - if (dev_priv->gt_pm.rc6.enabled) - return; - - if (IS_CHERRYVIEW(dev_priv)) - cherryview_enable_rc6(dev_priv); - else if (IS_VALLEYVIEW(dev_priv)) - valleyview_enable_rc6(dev_priv); - else if (INTEL_GEN(dev_priv) >= 11) - gen11_enable_rc6(dev_priv); - else if (INTEL_GEN(dev_priv) >= 9) - gen9_enable_rc6(dev_priv); - else if (IS_BROADWELL(dev_priv)) - gen8_enable_rc6(dev_priv); - else if (INTEL_GEN(dev_priv) >= 6) - gen6_enable_rc6(dev_priv); - - dev_priv->gt_pm.rc6.enabled = true; -} - -static void intel_enable_rps(struct drm_i915_private *dev_priv) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - lockdep_assert_held(&rps->lock); - - if (rps->enabled) - return; - - if (IS_CHERRYVIEW(dev_priv)) { - cherryview_enable_rps(dev_priv); - } else if (IS_VALLEYVIEW(dev_priv)) { - valleyview_enable_rps(dev_priv); - } else if (INTEL_GEN(dev_priv) >= 9) { - gen9_enable_rps(dev_priv); - } else if (IS_BROADWELL(dev_priv)) { - gen8_enable_rps(dev_priv); - } else if (INTEL_GEN(dev_priv) >= 6) { - gen6_enable_rps(dev_priv); - } else if (IS_IRONLAKE_M(dev_priv)) { - ironlake_enable_drps(dev_priv); - intel_init_emon(dev_priv); - } - - WARN_ON(rps->max_freq < rps->min_freq); - WARN_ON(rps->idle_freq > rps->max_freq); - - WARN_ON(rps->efficient_freq < rps->min_freq); - WARN_ON(rps->efficient_freq > rps->max_freq); - - rps->enabled = true; -} - -void intel_enable_gt_powersave(struct drm_i915_private *dev_priv) -{ - /* Powersaving is controlled by the host when inside a VM */ - if (intel_vgpu_active(dev_priv)) - return; - - mutex_lock(&dev_priv->gt_pm.rps.lock); - - if (HAS_RC6(dev_priv)) - intel_enable_rc6(dev_priv); - if (HAS_RPS(dev_priv)) - intel_enable_rps(dev_priv); - if (HAS_LLC(dev_priv)) - intel_enable_llc_pstate(dev_priv); - - mutex_unlock(&dev_priv->gt_pm.rps.lock); -} - static void ibx_init_clock_gating(struct drm_i915_private *dev_priv) { /* @@ -8858,7 +6347,7 @@ static void ilk_init_clock_gating(struct drm_i915_private *dev_priv) static void cpt_init_clock_gating(struct drm_i915_private *dev_priv) { - int pipe; + enum pipe pipe; u32 val; /* @@ -9078,6 +6567,22 @@ static void icl_init_clock_gating(struct drm_i915_private *dev_priv) _MASKED_BIT_ENABLE(GEN11_ENABLE_32_PLANE_MODE)); } +static void tgl_init_clock_gating(struct drm_i915_private *dev_priv) +{ + u32 vd_pg_enable = 0; + unsigned int i; + + /* This is not a WA. Enable VD HCP & MFX_ENC powergate */ + for (i = 0; i < I915_MAX_VCS; i++) { + if (HAS_ENGINE(dev_priv, _VCS(i))) + vd_pg_enable |= VDN_HCP_POWERGATE_ENABLE(i) | + VDN_MFX_POWERGATE_ENABLE(i); + } + + I915_WRITE(POWERGATE_ENABLE, + I915_READ(POWERGATE_ENABLE) | vd_pg_enable); +} + static void cnp_init_clock_gating(struct drm_i915_private *dev_priv) { if (!HAS_PCH_CNP(dev_priv)) @@ -9598,7 +7103,7 @@ static void nop_init_clock_gating(struct drm_i915_private *dev_priv) void intel_init_clock_gating_hooks(struct drm_i915_private *dev_priv) { if (IS_GEN(dev_priv, 12)) - dev_priv->display.init_clock_gating = nop_init_clock_gating; + dev_priv->display.init_clock_gating = tgl_init_clock_gating; else if (IS_GEN(dev_priv, 11)) dev_priv->display.init_clock_gating = icl_init_clock_gating; else if (IS_CANNONLAKE(dev_priv)) @@ -9654,6 +7159,9 @@ void intel_init_pm(struct drm_i915_private *dev_priv) else if (IS_GEN(dev_priv, 5)) i915_ironlake_get_mem_freq(dev_priv); + if (intel_has_sagv(dev_priv)) + skl_setup_sagv_block_time(dev_priv); + /* For FIFO watermark updates */ if (INTEL_GEN(dev_priv) >= 9) { skl_setup_wm_latency(dev_priv); @@ -9712,7 +7220,7 @@ void intel_init_pm(struct drm_i915_private *dev_priv) dev_priv->display.update_wm = i9xx_update_wm; dev_priv->display.get_fifo_size = i9xx_get_fifo_size; } else if (IS_GEN(dev_priv, 2)) { - if (INTEL_INFO(dev_priv)->num_pipes == 1) { + if (INTEL_NUM_PIPES(dev_priv) == 1) { dev_priv->display.update_wm = i845_update_wm; dev_priv->display.get_fifo_size = i845_get_fifo_size; } else { @@ -9724,217 +7232,8 @@ void intel_init_pm(struct drm_i915_private *dev_priv) } } -static int byt_gpu_freq(struct drm_i915_private *dev_priv, int val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - /* - * N = val - 0xb7 - * Slow = Fast = GPLL ref * N - */ - return DIV_ROUND_CLOSEST(rps->gpll_ref_freq * (val - 0xb7), 1000); -} - -static int byt_freq_opcode(struct drm_i915_private *dev_priv, int val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - return DIV_ROUND_CLOSEST(1000 * val, rps->gpll_ref_freq) + 0xb7; -} - -static int chv_gpu_freq(struct drm_i915_private *dev_priv, int val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - /* - * N = val / 2 - * CU (slow) = CU2x (fast) / 2 = GPLL ref * N / 2 - */ - return DIV_ROUND_CLOSEST(rps->gpll_ref_freq * val, 2 * 2 * 1000); -} - -static int chv_freq_opcode(struct drm_i915_private *dev_priv, int val) -{ - struct intel_rps *rps = &dev_priv->gt_pm.rps; - - /* CHV needs even values */ - return DIV_ROUND_CLOSEST(2 * 1000 * val, rps->gpll_ref_freq) * 2; -} - -int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) -{ - if (INTEL_GEN(dev_priv) >= 9) - return DIV_ROUND_CLOSEST(val * GT_FREQUENCY_MULTIPLIER, - GEN9_FREQ_SCALER); - else if (IS_CHERRYVIEW(dev_priv)) - return chv_gpu_freq(dev_priv, val); - else if (IS_VALLEYVIEW(dev_priv)) - return byt_gpu_freq(dev_priv, val); - else - return val * GT_FREQUENCY_MULTIPLIER; -} - -int intel_freq_opcode(struct drm_i915_private *dev_priv, int val) -{ - if (INTEL_GEN(dev_priv) >= 9) - return DIV_ROUND_CLOSEST(val * GEN9_FREQ_SCALER, - GT_FREQUENCY_MULTIPLIER); - else if (IS_CHERRYVIEW(dev_priv)) - return chv_freq_opcode(dev_priv, val); - else if (IS_VALLEYVIEW(dev_priv)) - return byt_freq_opcode(dev_priv, val); - else - return DIV_ROUND_CLOSEST(val, GT_FREQUENCY_MULTIPLIER); -} - void intel_pm_setup(struct drm_i915_private *dev_priv) { - mutex_init(&dev_priv->gt_pm.rps.lock); - mutex_init(&dev_priv->gt_pm.rps.power.mutex); - - atomic_set(&dev_priv->gt_pm.rps.num_waiters, 0); - dev_priv->runtime_pm.suspended = false; atomic_set(&dev_priv->runtime_pm.wakeref_count, 0); } - -static u64 vlv_residency_raw(struct drm_i915_private *dev_priv, - const i915_reg_t reg) -{ - u32 lower, upper, tmp; - int loop = 2; - - /* - * The register accessed do not need forcewake. We borrow - * uncore lock to prevent concurrent access to range reg. - */ - lockdep_assert_held(&dev_priv->uncore.lock); - - /* - * vlv and chv residency counters are 40 bits in width. - * With a control bit, we can choose between upper or lower - * 32bit window into this counter. - * - * Although we always use the counter in high-range mode elsewhere, - * userspace may attempt to read the value before rc6 is initialised, - * before we have set the default VLV_COUNTER_CONTROL value. So always - * set the high bit to be safe. - */ - I915_WRITE_FW(VLV_COUNTER_CONTROL, - _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH)); - upper = I915_READ_FW(reg); - do { - tmp = upper; - - I915_WRITE_FW(VLV_COUNTER_CONTROL, - _MASKED_BIT_DISABLE(VLV_COUNT_RANGE_HIGH)); - lower = I915_READ_FW(reg); - - I915_WRITE_FW(VLV_COUNTER_CONTROL, - _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH)); - upper = I915_READ_FW(reg); - } while (upper != tmp && --loop); - - /* - * Everywhere else we always use VLV_COUNTER_CONTROL with the - * VLV_COUNT_RANGE_HIGH bit set - so it is safe to leave it set - * now. - */ - - return lower | (u64)upper << 8; -} - -u64 intel_rc6_residency_ns(struct drm_i915_private *dev_priv, - const i915_reg_t reg) -{ - struct intel_uncore *uncore = &dev_priv->uncore; - u64 time_hw, prev_hw, overflow_hw; - unsigned int fw_domains; - unsigned long flags; - unsigned int i; - u32 mul, div; - - if (!HAS_RC6(dev_priv)) - return 0; - - /* - * Store previous hw counter values for counter wrap-around handling. - * - * There are only four interesting registers and they live next to each - * other so we can use the relative address, compared to the smallest - * one as the index into driver storage. - */ - i = (i915_mmio_reg_offset(reg) - - i915_mmio_reg_offset(GEN6_GT_GFX_RC6_LOCKED)) / sizeof(u32); - if (WARN_ON_ONCE(i >= ARRAY_SIZE(dev_priv->gt_pm.rc6.cur_residency))) - return 0; - - fw_domains = intel_uncore_forcewake_for_reg(uncore, reg, FW_REG_READ); - - spin_lock_irqsave(&uncore->lock, flags); - intel_uncore_forcewake_get__locked(uncore, fw_domains); - - /* On VLV and CHV, residency time is in CZ units rather than 1.28us */ - if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) { - mul = 1000000; - div = dev_priv->czclk_freq; - overflow_hw = BIT_ULL(40); - time_hw = vlv_residency_raw(dev_priv, reg); - } else { - /* 833.33ns units on Gen9LP, 1.28us elsewhere. */ - if (IS_GEN9_LP(dev_priv)) { - mul = 10000; - div = 12; - } else { - mul = 1280; - div = 1; - } - - overflow_hw = BIT_ULL(32); - time_hw = intel_uncore_read_fw(uncore, reg); - } - - /* - * Counter wrap handling. - * - * But relying on a sufficient frequency of queries otherwise counters - * can still wrap. - */ - prev_hw = dev_priv->gt_pm.rc6.prev_hw_residency[i]; - dev_priv->gt_pm.rc6.prev_hw_residency[i] = time_hw; - - /* RC6 delta from last sample. */ - if (time_hw >= prev_hw) - time_hw -= prev_hw; - else - time_hw += overflow_hw - prev_hw; - - /* Add delta to RC6 extended raw driver copy. */ - time_hw += dev_priv->gt_pm.rc6.cur_residency[i]; - dev_priv->gt_pm.rc6.cur_residency[i] = time_hw; - - intel_uncore_forcewake_put__locked(uncore, fw_domains); - spin_unlock_irqrestore(&uncore->lock, flags); - - return mul_u64_u32_div(time_hw, mul, div); -} - -u64 intel_rc6_residency_us(struct drm_i915_private *dev_priv, - i915_reg_t reg) -{ - return DIV_ROUND_UP_ULL(intel_rc6_residency_ns(dev_priv, reg), 1000); -} - -u32 intel_get_cagf(struct drm_i915_private *dev_priv, u32 rpstat) -{ - u32 cagf; - - if (INTEL_GEN(dev_priv) >= 9) - cagf = (rpstat & GEN9_CAGF_MASK) >> GEN9_CAGF_SHIFT; - else if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) - cagf = (rpstat & HSW_CAGF_MASK) >> HSW_CAGF_SHIFT; - else - cagf = (rpstat & GEN6_CAGF_MASK) >> GEN6_CAGF_SHIFT; - - return cagf; -} diff --git a/drivers/gpu/drm/i915/intel_pm.h b/drivers/gpu/drm/i915/intel_pm.h index e3573e1e16e3..b579c724b915 100644 --- a/drivers/gpu/drm/i915/intel_pm.h +++ b/drivers/gpu/drm/i915/intel_pm.h @@ -29,16 +29,6 @@ void intel_update_watermarks(struct intel_crtc *crtc); void intel_init_pm(struct drm_i915_private *dev_priv); void intel_init_clock_gating_hooks(struct drm_i915_private *dev_priv); void intel_pm_setup(struct drm_i915_private *dev_priv); -void intel_gpu_ips_init(struct drm_i915_private *dev_priv); -void intel_gpu_ips_teardown(void); -void intel_init_gt_powersave(struct drm_i915_private *dev_priv); -void intel_cleanup_gt_powersave(struct drm_i915_private *dev_priv); -void intel_sanitize_gt_powersave(struct drm_i915_private *dev_priv); -void intel_enable_gt_powersave(struct drm_i915_private *dev_priv); -void intel_disable_gt_powersave(struct drm_i915_private *dev_priv); -void gen6_rps_busy(struct drm_i915_private *dev_priv); -void gen6_rps_idle(struct drm_i915_private *dev_priv); -void gen6_rps_boost(struct i915_request *rq); void g4x_wm_get_hw_state(struct drm_i915_private *dev_priv); void vlv_wm_get_hw_state(struct drm_i915_private *dev_priv); void ilk_wm_get_hw_state(struct drm_i915_private *dev_priv); @@ -65,26 +55,9 @@ void skl_write_plane_wm(struct intel_plane *plane, void skl_write_cursor_wm(struct intel_plane *plane, const struct intel_crtc_state *crtc_state); bool ilk_disable_lp_wm(struct drm_device *dev); -int skl_check_pipe_max_pixel_rate(struct intel_crtc *intel_crtc, - struct intel_crtc_state *cstate); void intel_init_ipc(struct drm_i915_private *dev_priv); void intel_enable_ipc(struct drm_i915_private *dev_priv); -int intel_gpu_freq(struct drm_i915_private *dev_priv, int val); -int intel_freq_opcode(struct drm_i915_private *dev_priv, int val); -u64 intel_rc6_residency_ns(struct drm_i915_private *dev_priv, i915_reg_t reg); -u64 intel_rc6_residency_us(struct drm_i915_private *dev_priv, i915_reg_t reg); - -u32 intel_get_cagf(struct drm_i915_private *dev_priv, u32 rpstat1); - -unsigned long i915_chipset_val(struct drm_i915_private *dev_priv); -unsigned long i915_mch_val(struct drm_i915_private *dev_priv); -unsigned long i915_gfx_val(struct drm_i915_private *dev_priv); -void i915_update_gfx_val(struct drm_i915_private *dev_priv); - -bool ironlake_set_drps(struct drm_i915_private *dev_priv, u8 val); -int intel_set_rps(struct drm_i915_private *dev_priv, u8 val); -void intel_rps_mark_interactive(struct drm_i915_private *i915, bool interactive); bool intel_set_memory_cxsr(struct drm_i915_private *dev_priv, bool enable); #endif /* __INTEL_PM_H__ */ diff --git a/drivers/gpu/drm/i915/intel_region_lmem.c b/drivers/gpu/drm/i915/intel_region_lmem.c new file mode 100644 index 000000000000..583118095635 --- /dev/null +++ b/drivers/gpu/drm/i915/intel_region_lmem.c @@ -0,0 +1,132 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include "i915_drv.h" +#include "intel_memory_region.h" +#include "gem/i915_gem_lmem.h" +#include "gem/i915_gem_region.h" +#include "intel_region_lmem.h" + +static int init_fake_lmem_bar(struct intel_memory_region *mem) +{ + struct drm_i915_private *i915 = mem->i915; + struct i915_ggtt *ggtt = &i915->ggtt; + unsigned long n; + int ret; + + /* We want to 1:1 map the mappable aperture to our reserved region */ + + mem->fake_mappable.start = 0; + mem->fake_mappable.size = resource_size(&mem->region); + mem->fake_mappable.color = I915_COLOR_UNEVICTABLE; + + ret = drm_mm_reserve_node(&ggtt->vm.mm, &mem->fake_mappable); + if (ret) + return ret; + + mem->remap_addr = dma_map_resource(&i915->drm.pdev->dev, + mem->region.start, + mem->fake_mappable.size, + PCI_DMA_BIDIRECTIONAL, + DMA_ATTR_FORCE_CONTIGUOUS); + if (dma_mapping_error(&i915->drm.pdev->dev, mem->remap_addr)) { + drm_mm_remove_node(&mem->fake_mappable); + return -EINVAL; + } + + for (n = 0; n < mem->fake_mappable.size >> PAGE_SHIFT; ++n) { + ggtt->vm.insert_page(&ggtt->vm, + mem->remap_addr + (n << PAGE_SHIFT), + n << PAGE_SHIFT, + I915_CACHE_NONE, 0); + } + + mem->region = (struct resource)DEFINE_RES_MEM(mem->remap_addr, + mem->fake_mappable.size); + + return 0; +} + +static void release_fake_lmem_bar(struct intel_memory_region *mem) +{ + if (drm_mm_node_allocated(&mem->fake_mappable)) + drm_mm_remove_node(&mem->fake_mappable); + + dma_unmap_resource(&mem->i915->drm.pdev->dev, + mem->remap_addr, + mem->fake_mappable.size, + PCI_DMA_BIDIRECTIONAL, + DMA_ATTR_FORCE_CONTIGUOUS); +} + +static void +region_lmem_release(struct intel_memory_region *mem) +{ + release_fake_lmem_bar(mem); + io_mapping_fini(&mem->iomap); + intel_memory_region_release_buddy(mem); +} + +static int +region_lmem_init(struct intel_memory_region *mem) +{ + int ret; + + if (i915_modparams.fake_lmem_start) { + ret = init_fake_lmem_bar(mem); + GEM_BUG_ON(ret); + } + + if (!io_mapping_init_wc(&mem->iomap, + mem->io_start, + resource_size(&mem->region))) + return -EIO; + + ret = intel_memory_region_init_buddy(mem); + if (ret) + io_mapping_fini(&mem->iomap); + + return ret; +} + +const struct intel_memory_region_ops intel_region_lmem_ops = { + .init = region_lmem_init, + .release = region_lmem_release, + .create_object = __i915_gem_lmem_object_create, +}; + +struct intel_memory_region * +intel_setup_fake_lmem(struct drm_i915_private *i915) +{ + struct pci_dev *pdev = i915->drm.pdev; + struct intel_memory_region *mem; + resource_size_t mappable_end; + resource_size_t io_start; + resource_size_t start; + + GEM_BUG_ON(i915_ggtt_has_aperture(&i915->ggtt)); + GEM_BUG_ON(!i915_modparams.fake_lmem_start); + + /* Your mappable aperture belongs to me now! */ + mappable_end = pci_resource_len(pdev, 2); + io_start = pci_resource_start(pdev, 2), + start = i915_modparams.fake_lmem_start; + + mem = intel_memory_region_create(i915, + start, + mappable_end, + PAGE_SIZE, + io_start, + &intel_region_lmem_ops); + if (!IS_ERR(mem)) { + DRM_INFO("Intel graphics fake LMEM: %pR\n", &mem->region); + DRM_INFO("Intel graphics fake LMEM IO start: %llx\n", + (u64)mem->io_start); + DRM_INFO("Intel graphics fake LMEM size: %llx\n", + (u64)resource_size(&mem->region)); + } + + return mem; +} diff --git a/drivers/gpu/drm/i915/intel_region_lmem.h b/drivers/gpu/drm/i915/intel_region_lmem.h new file mode 100644 index 000000000000..213def7c7b8a --- /dev/null +++ b/drivers/gpu/drm/i915/intel_region_lmem.h @@ -0,0 +1,16 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __INTEL_REGION_LMEM_H +#define __INTEL_REGION_LMEM_H + +struct drm_i915_private; + +extern const struct intel_memory_region_ops intel_region_lmem_ops; + +struct intel_memory_region * +intel_setup_fake_lmem(struct drm_i915_private *i915); + +#endif /* !__INTEL_REGION_LMEM_H */ diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index 2fd3c097e1f5..ad719c9602af 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -27,7 +27,6 @@ */ #include <linux/pm_runtime.h> -#include <linux/vgaarb.h> #include <drm/drm_print.h> diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 9e583f13a9e4..94a97bf8c021 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -805,9 +805,6 @@ void assert_forcewakes_active(struct intel_uncore *uncore, /* We give fast paths for the really cool registers */ #define NEEDS_FORCE_WAKE(reg) ((reg) < 0x40000) -#define GEN11_NEEDS_FORCE_WAKE(reg) \ - ((reg) < 0x40000 || ((reg) >= 0x1c0000 && (reg) < 0x1dc000)) - #define __gen6_reg_read_fw_domains(uncore, offset) \ ({ \ enum forcewake_domains __fwd; \ @@ -903,12 +900,10 @@ static const struct intel_forcewake_range __vlv_fw_ranges[] = { }) #define __gen11_fwtable_reg_read_fw_domains(uncore, offset) \ -({ \ - enum forcewake_domains __fwd = 0; \ - if (GEN11_NEEDS_FORCE_WAKE((offset))) \ - __fwd = find_fw_domain(uncore, offset); \ - __fwd; \ -}) + find_fw_domain(uncore, offset) + +#define __gen12_fwtable_reg_read_fw_domains(uncore, offset) \ + find_fw_domain(uncore, offset) /* *Must* be sorted by offset! See intel_shadow_table_check(). */ static const i915_reg_t gen8_shadowed_regs[] = { @@ -935,6 +930,20 @@ static const i915_reg_t gen11_shadowed_regs[] = { /* TODO: Other registers are not yet used */ }; +static const i915_reg_t gen12_shadowed_regs[] = { + RING_TAIL(RENDER_RING_BASE), /* 0x2000 (base) */ + GEN6_RPNSWREQ, /* 0xA008 */ + GEN6_RC_VIDEO_FREQ, /* 0xA00C */ + RING_TAIL(BLT_RING_BASE), /* 0x22000 (base) */ + RING_TAIL(GEN11_BSD_RING_BASE), /* 0x1C0000 (base) */ + RING_TAIL(GEN11_BSD2_RING_BASE), /* 0x1C4000 (base) */ + RING_TAIL(GEN11_VEBOX_RING_BASE), /* 0x1C8000 (base) */ + RING_TAIL(GEN11_BSD3_RING_BASE), /* 0x1D0000 (base) */ + RING_TAIL(GEN11_BSD4_RING_BASE), /* 0x1D4000 (base) */ + RING_TAIL(GEN11_VEBOX2_RING_BASE), /* 0x1D8000 (base) */ + /* TODO: Other registers are not yet used */ +}; + static int mmio_reg_cmp(u32 key, const i915_reg_t *reg) { u32 offset = i915_mmio_reg_offset(*reg); @@ -957,6 +966,7 @@ static bool is_gen##x##_shadowed(u32 offset) \ __is_genX_shadowed(8) __is_genX_shadowed(11) +__is_genX_shadowed(12) static enum forcewake_domains gen6_reg_write_fw_domains(struct intel_uncore *uncore, i915_reg_t reg) @@ -1005,8 +1015,18 @@ static const struct intel_forcewake_range __chv_fw_ranges[] = { #define __gen11_fwtable_reg_write_fw_domains(uncore, offset) \ ({ \ enum forcewake_domains __fwd = 0; \ - if (GEN11_NEEDS_FORCE_WAKE((offset)) && !is_gen11_shadowed(offset)) \ - __fwd = find_fw_domain(uncore, offset); \ + const u32 __offset = (offset); \ + if (!is_gen11_shadowed(__offset)) \ + __fwd = find_fw_domain(uncore, __offset); \ + __fwd; \ +}) + +#define __gen12_fwtable_reg_write_fw_domains(uncore, offset) \ +({ \ + enum forcewake_domains __fwd = 0; \ + const u32 __offset = (offset); \ + if (!is_gen12_shadowed(__offset)) \ + __fwd = find_fw_domain(uncore, __offset); \ __fwd; \ }) @@ -1065,9 +1085,51 @@ static const struct intel_forcewake_range __gen11_fw_ranges[] = { GEN_FW_RANGE(0x9400, 0x97ff, FORCEWAKE_ALL), GEN_FW_RANGE(0x9800, 0xafff, FORCEWAKE_BLITTER), GEN_FW_RANGE(0xb000, 0xb47f, FORCEWAKE_RENDER), + GEN_FW_RANGE(0xb480, 0xdeff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0xdf00, 0xe8ff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0xe900, 0x16dff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x16e00, 0x19fff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x1a000, 0x243ff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x24400, 0x247ff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x24800, 0x3ffff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x40000, 0x1bffff, 0), + GEN_FW_RANGE(0x1c0000, 0x1c3fff, FORCEWAKE_MEDIA_VDBOX0), + GEN_FW_RANGE(0x1c4000, 0x1c7fff, FORCEWAKE_MEDIA_VDBOX1), + GEN_FW_RANGE(0x1c8000, 0x1cbfff, FORCEWAKE_MEDIA_VEBOX0), + GEN_FW_RANGE(0x1cc000, 0x1cffff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x1d0000, 0x1d3fff, FORCEWAKE_MEDIA_VDBOX2), + GEN_FW_RANGE(0x1d4000, 0x1d7fff, FORCEWAKE_MEDIA_VDBOX3), + GEN_FW_RANGE(0x1d8000, 0x1dbfff, FORCEWAKE_MEDIA_VEBOX1) +}; + +/* *Must* be sorted by offset ranges! See intel_fw_table_check(). */ +static const struct intel_forcewake_range __gen12_fw_ranges[] = { + GEN_FW_RANGE(0x0, 0xaff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0xb00, 0x1fff, 0), /* uncore range */ + GEN_FW_RANGE(0x2000, 0x26ff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x2700, 0x2fff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x3000, 0x3fff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x4000, 0x51ff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x5200, 0x7fff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x8000, 0x813f, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x8140, 0x815f, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x8160, 0x82ff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x8300, 0x84ff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x8500, 0x8bff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x8c00, 0x8cff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x8d00, 0x93ff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x9400, 0x97ff, FORCEWAKE_ALL), + GEN_FW_RANGE(0x9800, 0xafff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0xb000, 0xb47f, FORCEWAKE_RENDER), GEN_FW_RANGE(0xb480, 0xdfff, FORCEWAKE_BLITTER), GEN_FW_RANGE(0xe000, 0xe8ff, FORCEWAKE_RENDER), - GEN_FW_RANGE(0xe900, 0x243ff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0xe900, 0x147ff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x14800, 0x148ff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x14900, 0x19fff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x1a000, 0x1a7ff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x1a800, 0x1afff, FORCEWAKE_BLITTER), + GEN_FW_RANGE(0x1b000, 0x1bfff, FORCEWAKE_RENDER), + GEN_FW_RANGE(0x1c000, 0x243ff, FORCEWAKE_BLITTER), GEN_FW_RANGE(0x24400, 0x247ff, FORCEWAKE_RENDER), GEN_FW_RANGE(0x24800, 0x3ffff, FORCEWAKE_BLITTER), GEN_FW_RANGE(0x40000, 0x1bffff, 0), @@ -1228,6 +1290,7 @@ __gen_read(func, 16) \ __gen_read(func, 32) \ __gen_read(func, 64) +__gen_reg_read_funcs(gen12_fwtable); __gen_reg_read_funcs(gen11_fwtable); __gen_reg_read_funcs(fwtable); __gen_reg_read_funcs(gen6); @@ -1319,6 +1382,7 @@ __gen_write(func, 8) \ __gen_write(func, 16) \ __gen_write(func, 32) +__gen_reg_write_funcs(gen12_fwtable); __gen_reg_write_funcs(gen11_fwtable); __gen_reg_write_funcs(fwtable); __gen_reg_write_funcs(gen8); @@ -1690,10 +1754,14 @@ static int uncore_forcewake_init(struct intel_uncore *uncore) ASSIGN_FW_DOMAINS_TABLE(uncore, __gen9_fw_ranges); ASSIGN_WRITE_MMIO_VFUNCS(uncore, fwtable); ASSIGN_READ_MMIO_VFUNCS(uncore, fwtable); - } else { + } else if (IS_GEN(i915, 11)) { ASSIGN_FW_DOMAINS_TABLE(uncore, __gen11_fw_ranges); ASSIGN_WRITE_MMIO_VFUNCS(uncore, gen11_fwtable); ASSIGN_READ_MMIO_VFUNCS(uncore, gen11_fwtable); + } else { + ASSIGN_FW_DOMAINS_TABLE(uncore, __gen12_fw_ranges); + ASSIGN_WRITE_MMIO_VFUNCS(uncore, gen12_fwtable); + ASSIGN_READ_MMIO_VFUNCS(uncore, gen12_fwtable); } uncore->pmic_bus_access_nb.notifier_call = i915_pmic_bus_access_notifier; diff --git a/drivers/gpu/drm/i915/intel_uncore.h b/drivers/gpu/drm/i915/intel_uncore.h index 414fc2cb0459..dcfa243892c6 100644 --- a/drivers/gpu/drm/i915/intel_uncore.h +++ b/drivers/gpu/drm/i915/intel_uncore.h @@ -378,23 +378,23 @@ intel_uncore_read64_2x32(struct intel_uncore *uncore, static inline void intel_uncore_rmw(struct intel_uncore *uncore, i915_reg_t reg, u32 clear, u32 set) { - u32 val; + u32 old, val; - val = intel_uncore_read(uncore, reg); - val &= ~clear; - val |= set; - intel_uncore_write(uncore, reg, val); + old = intel_uncore_read(uncore, reg); + val = (old & ~clear) | set; + if (val != old) + intel_uncore_write(uncore, reg, val); } static inline void intel_uncore_rmw_fw(struct intel_uncore *uncore, i915_reg_t reg, u32 clear, u32 set) { - u32 val; + u32 old, val; - val = intel_uncore_read_fw(uncore, reg); - val &= ~clear; - val |= set; - intel_uncore_write_fw(uncore, reg, val); + old = intel_uncore_read_fw(uncore, reg); + val = (old & ~clear) | set; + if (val != old) + intel_uncore_write_fw(uncore, reg, val); } static inline int intel_uncore_write_and_verify(struct intel_uncore *uncore, diff --git a/drivers/gpu/drm/i915/intel_wakeref.c b/drivers/gpu/drm/i915/intel_wakeref.c index 868cc78048d0..59aa1b6f1827 100644 --- a/drivers/gpu/drm/i915/intel_wakeref.c +++ b/drivers/gpu/drm/i915/intel_wakeref.c @@ -54,7 +54,8 @@ int __intel_wakeref_get_first(struct intel_wakeref *wf) static void ____intel_wakeref_put_last(struct intel_wakeref *wf) { - if (!atomic_dec_and_test(&wf->count)) + INTEL_WAKEREF_BUG_ON(atomic_read(&wf->count) <= 0); + if (unlikely(!atomic_dec_and_test(&wf->count))) goto unlock; /* ops->put() must reschedule its own release on error/deferral */ @@ -67,13 +68,12 @@ unlock: mutex_unlock(&wf->mutex); } -void __intel_wakeref_put_last(struct intel_wakeref *wf) +void __intel_wakeref_put_last(struct intel_wakeref *wf, unsigned long flags) { INTEL_WAKEREF_BUG_ON(work_pending(&wf->work)); /* Assume we are not in process context and so cannot sleep. */ - if (wf->ops->flags & INTEL_WAKEREF_PUT_ASYNC || - !mutex_trylock(&wf->mutex)) { + if (flags & INTEL_WAKEREF_PUT_ASYNC || !mutex_trylock(&wf->mutex)) { schedule_work(&wf->work); return; } @@ -109,8 +109,17 @@ void __intel_wakeref_init(struct intel_wakeref *wf, int intel_wakeref_wait_for_idle(struct intel_wakeref *wf) { - return wait_var_event_killable(&wf->wakeref, - !intel_wakeref_is_active(wf)); + int err; + + might_sleep(); + + err = wait_var_event_killable(&wf->wakeref, + !intel_wakeref_is_active(wf)); + if (err) + return err; + + intel_wakeref_unlock_wait(wf); + return 0; } static void wakeref_auto_timeout(struct timer_list *t) diff --git a/drivers/gpu/drm/i915/intel_wakeref.h b/drivers/gpu/drm/i915/intel_wakeref.h index 5f0c972a80fb..da6e8fd506e6 100644 --- a/drivers/gpu/drm/i915/intel_wakeref.h +++ b/drivers/gpu/drm/i915/intel_wakeref.h @@ -9,6 +9,7 @@ #include <linux/atomic.h> #include <linux/bits.h> +#include <linux/lockdep.h> #include <linux/mutex.h> #include <linux/refcount.h> #include <linux/stackdepot.h> @@ -29,9 +30,6 @@ typedef depot_stack_handle_t intel_wakeref_t; struct intel_wakeref_ops { int (*get)(struct intel_wakeref *wf); int (*put)(struct intel_wakeref *wf); - - unsigned long flags; -#define INTEL_WAKEREF_PUT_ASYNC BIT(0) }; struct intel_wakeref { @@ -57,7 +55,7 @@ void __intel_wakeref_init(struct intel_wakeref *wf, } while (0) int __intel_wakeref_get_first(struct intel_wakeref *wf); -void __intel_wakeref_put_last(struct intel_wakeref *wf); +void __intel_wakeref_put_last(struct intel_wakeref *wf, unsigned long flags); /** * intel_wakeref_get: Acquire the wakeref @@ -100,10 +98,9 @@ intel_wakeref_get_if_active(struct intel_wakeref *wf) } /** - * intel_wakeref_put: Release the wakeref - * @i915: the drm_i915_private device + * intel_wakeref_put_flags: Release the wakeref * @wf: the wakeref - * @fn: callback for releasing the wakeref, called only on final release. + * @flags: control flags * * Release our hold on the wakeref. When there are no more users, * the runtime pm wakeref will be released after the @fn callback is called @@ -116,11 +113,25 @@ intel_wakeref_get_if_active(struct intel_wakeref *wf) * code otherwise. */ static inline void -intel_wakeref_put(struct intel_wakeref *wf) +__intel_wakeref_put(struct intel_wakeref *wf, unsigned long flags) +#define INTEL_WAKEREF_PUT_ASYNC BIT(0) { INTEL_WAKEREF_BUG_ON(atomic_read(&wf->count) <= 0); if (unlikely(!atomic_add_unless(&wf->count, -1, 1))) - __intel_wakeref_put_last(wf); + __intel_wakeref_put_last(wf, flags); +} + +static inline void +intel_wakeref_put(struct intel_wakeref *wf) +{ + might_sleep(); + __intel_wakeref_put(wf, 0); +} + +static inline void +intel_wakeref_put_async(struct intel_wakeref *wf) +{ + __intel_wakeref_put(wf, INTEL_WAKEREF_PUT_ASYNC); } /** @@ -152,6 +163,21 @@ intel_wakeref_unlock(struct intel_wakeref *wf) } /** + * intel_wakeref_unlock_wait: Wait until the active callback is complete + * @wf: the wakeref + * + * Waits for the active callback (under the @wf->mutex or another CPU) is + * complete. + */ +static inline void +intel_wakeref_unlock_wait(struct intel_wakeref *wf) +{ + mutex_lock(&wf->mutex); + mutex_unlock(&wf->mutex); + flush_work(&wf->work); +} + +/** * intel_wakeref_is_active: Query whether the wakeref is currently held * @wf: the wakeref * @@ -170,6 +196,7 @@ intel_wakeref_is_active(const struct intel_wakeref *wf) static inline void __intel_wakeref_defer_park(struct intel_wakeref *wf) { + lockdep_assert_held(&wf->mutex); INTEL_WAKEREF_BUG_ON(atomic_read(&wf->count)); atomic_set_release(&wf->count, 1); } diff --git a/drivers/gpu/drm/i915/oa/i915_oa_tgl.c b/drivers/gpu/drm/i915/oa/i915_oa_tgl.c new file mode 100644 index 000000000000..a29d93707345 --- /dev/null +++ b/drivers/gpu/drm/i915/oa/i915_oa_tgl.c @@ -0,0 +1,121 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2018 Intel Corporation + * + * Autogenerated file by GPU Top : https://github.com/rib/gputop + * DO NOT EDIT manually! + */ + +#include <linux/sysfs.h> + +#include "i915_drv.h" +#include "i915_oa_tgl.h" + +static const struct i915_oa_reg b_counter_config_test_oa[] = { + { _MMIO(0xD920), 0x00000000 }, + { _MMIO(0xD900), 0x00000000 }, + { _MMIO(0xD904), 0xF0800000 }, + { _MMIO(0xD910), 0x00000000 }, + { _MMIO(0xD914), 0xF0800000 }, + { _MMIO(0xDC40), 0x00FF0000 }, + { _MMIO(0xD940), 0x00000004 }, + { _MMIO(0xD944), 0x0000FFFF }, + { _MMIO(0xDC00), 0x00000004 }, + { _MMIO(0xDC04), 0x0000FFFF }, + { _MMIO(0xD948), 0x00000003 }, + { _MMIO(0xD94C), 0x0000FFFF }, + { _MMIO(0xDC08), 0x00000003 }, + { _MMIO(0xDC0C), 0x0000FFFF }, + { _MMIO(0xD950), 0x00000007 }, + { _MMIO(0xD954), 0x0000FFFF }, + { _MMIO(0xDC10), 0x00000007 }, + { _MMIO(0xDC14), 0x0000FFFF }, + { _MMIO(0xD958), 0x00100002 }, + { _MMIO(0xD95C), 0x0000FFF7 }, + { _MMIO(0xDC18), 0x00100002 }, + { _MMIO(0xDC1C), 0x0000FFF7 }, + { _MMIO(0xD960), 0x00100002 }, + { _MMIO(0xD964), 0x0000FFCF }, + { _MMIO(0xDC20), 0x00100002 }, + { _MMIO(0xDC24), 0x0000FFCF }, + { _MMIO(0xD968), 0x00100082 }, + { _MMIO(0xD96C), 0x0000FFEF }, + { _MMIO(0xDC28), 0x00100082 }, + { _MMIO(0xDC2C), 0x0000FFEF }, + { _MMIO(0xD970), 0x001000C2 }, + { _MMIO(0xD974), 0x0000FFE7 }, + { _MMIO(0xDC30), 0x001000C2 }, + { _MMIO(0xDC34), 0x0000FFE7 }, + { _MMIO(0xD978), 0x00100001 }, + { _MMIO(0xD97C), 0x0000FFE7 }, + { _MMIO(0xDC38), 0x00100001 }, + { _MMIO(0xDC3C), 0x0000FFE7 }, +}; + +static const struct i915_oa_reg flex_eu_config_test_oa[] = { +}; + +static const struct i915_oa_reg mux_config_test_oa[] = { + { _MMIO(0x0D04), 0x00000200 }, + { _MMIO(0x9840), 0x00000000 }, + { _MMIO(0x9884), 0x00000000 }, + { _MMIO(0x9888), 0x280E0000 }, + { _MMIO(0x9888), 0x1E0E0147 }, + { _MMIO(0x9888), 0x180E0000 }, + { _MMIO(0x9888), 0x160E0000 }, + { _MMIO(0x9888), 0x1E0F1000 }, + { _MMIO(0x9888), 0x1E104000 }, + { _MMIO(0x9888), 0x2E020100 }, + { _MMIO(0x9888), 0x2C030004 }, + { _MMIO(0x9888), 0x38003000 }, + { _MMIO(0x9888), 0x1E0A8000 }, + { _MMIO(0x9884), 0x00000003 }, + { _MMIO(0x9888), 0x49110000 }, + { _MMIO(0x9888), 0x5D101400 }, + { _MMIO(0x9888), 0x1D140020 }, + { _MMIO(0x9888), 0x1D1103A3 }, + { _MMIO(0x9888), 0x01110000 }, + { _MMIO(0x9888), 0x61111000 }, + { _MMIO(0x9888), 0x1F128000 }, + { _MMIO(0x9888), 0x17100000 }, + { _MMIO(0x9888), 0x55100630 }, + { _MMIO(0x9888), 0x57100000 }, + { _MMIO(0x9888), 0x31100000 }, + { _MMIO(0x9884), 0x00000003 }, + { _MMIO(0x9888), 0x65100002 }, + { _MMIO(0x9884), 0x00000000 }, + { _MMIO(0x9888), 0x42000001 }, +}; + +static ssize_t +show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) +{ + return sprintf(buf, "1\n"); +} + +void +i915_perf_load_test_config_tgl(struct drm_i915_private *dev_priv) +{ + strlcpy(dev_priv->perf.test_config.uuid, + "80a833f0-2504-4321-8894-e9277844ce7b", + sizeof(dev_priv->perf.test_config.uuid)); + dev_priv->perf.test_config.id = 1; + + dev_priv->perf.test_config.mux_regs = mux_config_test_oa; + dev_priv->perf.test_config.mux_regs_len = ARRAY_SIZE(mux_config_test_oa); + + dev_priv->perf.test_config.b_counter_regs = b_counter_config_test_oa; + dev_priv->perf.test_config.b_counter_regs_len = ARRAY_SIZE(b_counter_config_test_oa); + + dev_priv->perf.test_config.flex_regs = flex_eu_config_test_oa; + dev_priv->perf.test_config.flex_regs_len = ARRAY_SIZE(flex_eu_config_test_oa); + + dev_priv->perf.test_config.sysfs_metric.name = "80a833f0-2504-4321-8894-e9277844ce7b"; + dev_priv->perf.test_config.sysfs_metric.attrs = dev_priv->perf.test_config.attrs; + + dev_priv->perf.test_config.attrs[0] = &dev_priv->perf.test_config.sysfs_metric_id.attr; + + dev_priv->perf.test_config.sysfs_metric_id.attr.name = "id"; + dev_priv->perf.test_config.sysfs_metric_id.attr.mode = 0444; + dev_priv->perf.test_config.sysfs_metric_id.show = show_test_oa_id; +} diff --git a/drivers/gpu/drm/i915/oa/i915_oa_tgl.h b/drivers/gpu/drm/i915/oa/i915_oa_tgl.h new file mode 100644 index 000000000000..4c25f0be825c --- /dev/null +++ b/drivers/gpu/drm/i915/oa/i915_oa_tgl.h @@ -0,0 +1,16 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2018 Intel Corporation + * + * Autogenerated file by GPU Top : https://github.com/rib/gputop + * DO NOT EDIT manually! + */ + +#ifndef __I915_OA_TGL_H__ +#define __I915_OA_TGL_H__ + +struct drm_i915_private; + +void i915_perf_load_test_config_tgl(struct drm_i915_private *dev_priv); + +#endif diff --git a/drivers/gpu/drm/i915/selftests/i915_active.c b/drivers/gpu/drm/i915/selftests/i915_active.c index 77d844ac8b71..260b0ee5d1e3 100644 --- a/drivers/gpu/drm/i915/selftests/i915_active.c +++ b/drivers/gpu/drm/i915/selftests/i915_active.c @@ -68,7 +68,7 @@ static struct live_active *__live_alloc(struct drm_i915_private *i915) return NULL; kref_init(&active->ref); - i915_active_init(i915, &active->base, __live_active, __live_retire); + i915_active_init(&active->base, __live_active, __live_retire); return active; } @@ -79,7 +79,6 @@ __live_active_setup(struct drm_i915_private *i915) struct intel_engine_cs *engine; struct i915_sw_fence *submit; struct live_active *active; - enum intel_engine_id id; unsigned int count = 0; int err = 0; @@ -97,7 +96,7 @@ __live_active_setup(struct drm_i915_private *i915) if (err) goto out; - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { struct i915_request *rq; rq = i915_request_create(engine->kernel_context); @@ -110,7 +109,7 @@ __live_active_setup(struct drm_i915_private *i915) submit, GFP_KERNEL); if (err >= 0) - err = i915_active_ref(&active->base, rq->timeline, rq); + err = i915_active_add_request(&active->base, rq); i915_request_add(rq); if (err) { pr_err("Failed to track active ref!\n"); @@ -121,7 +120,7 @@ __live_active_setup(struct drm_i915_private *i915) } i915_active_release(&active->base); - if (active->retired && count) { + if (READ_ONCE(active->retired) && count) { pr_err("i915_active retired before submission!\n"); err = -EINVAL; } @@ -146,35 +145,25 @@ static int live_active_wait(void *arg) { struct drm_i915_private *i915 = arg; struct live_active *active; - intel_wakeref_t wakeref; int err = 0; /* Check that we get a callback when requests retire upon waiting */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - active = __live_active_setup(i915); - if (IS_ERR(active)) { - err = PTR_ERR(active); - goto err; - } + if (IS_ERR(active)) + return PTR_ERR(active); i915_active_wait(&active->base); - if (!active->retired) { + if (!READ_ONCE(active->retired)) { pr_err("i915_active not retired after waiting!\n"); err = -EINVAL; } __live_put(active); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; -err: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - return err; } @@ -182,35 +171,25 @@ static int live_active_retire(void *arg) { struct drm_i915_private *i915 = arg; struct live_active *active; - intel_wakeref_t wakeref; int err = 0; /* Check that we get a callback when requests are indirectly retired */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - active = __live_active_setup(i915); - if (IS_ERR(active)) { - err = PTR_ERR(active); - goto err; - } + if (IS_ERR(active)) + return PTR_ERR(active); /* waits for & retires all requests */ - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; - if (!active->retired) { + if (!READ_ONCE(active->retired)) { pr_err("i915_active not retired after flushing!\n"); err = -EINVAL; } __live_put(active); -err: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - return err; } @@ -226,3 +205,48 @@ int i915_active_live_selftests(struct drm_i915_private *i915) return i915_subtests(tests, i915); } + +static struct intel_engine_cs *node_to_barrier(struct active_node *it) +{ + struct intel_engine_cs *engine; + + if (!is_barrier(&it->base)) + return NULL; + + engine = __barrier_to_engine(it); + smp_rmb(); /* serialise with add_active_barriers */ + if (!is_barrier(&it->base)) + return NULL; + + return engine; +} + +void i915_active_print(struct i915_active *ref, struct drm_printer *m) +{ + drm_printf(m, "active %pS:%pS\n", ref->active, ref->retire); + drm_printf(m, "\tcount: %d\n", atomic_read(&ref->count)); + drm_printf(m, "\tpreallocated barriers? %s\n", + yesno(!llist_empty(&ref->preallocated_barriers))); + + if (i915_active_acquire_if_busy(ref)) { + struct active_node *it, *n; + + rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) { + struct intel_engine_cs *engine; + + engine = node_to_barrier(it); + if (engine) { + drm_printf(m, "\tbarrier: %s\n", engine->name); + continue; + } + + if (i915_active_fence_isset(&it->base)) { + drm_printf(m, + "\ttimeline: %llx\n", it->timeline); + continue; + } + } + + i915_active_release(ref); + } +} diff --git a/drivers/gpu/drm/i915/selftests/i915_buddy.c b/drivers/gpu/drm/i915/selftests/i915_buddy.c index 23f784eae1e7..1b856bae67b5 100644 --- a/drivers/gpu/drm/i915/selftests/i915_buddy.c +++ b/drivers/gpu/drm/i915/selftests/i915_buddy.c @@ -375,6 +375,8 @@ retry: if (err) break; + + cond_resched(); } if (err == -ENOMEM) @@ -687,6 +689,8 @@ static int igt_buddy_alloc_range(void *arg) rem -= size; if (!rem) break; + + cond_resched(); } if (err == -ENOMEM) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c index bb6dd54a6ff3..d83f6bf6d9d4 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem.c @@ -15,23 +15,26 @@ #include "igt_flush_test.h" #include "mock_drm.h" -static int switch_to_context(struct drm_i915_private *i915, - struct i915_gem_context *ctx) +static int switch_to_context(struct i915_gem_context *ctx) { - struct intel_engine_cs *engine; - enum intel_engine_id id; + struct i915_gem_engines_iter it; + struct intel_context *ce; + int err = 0; - for_each_engine(engine, i915, id) { + for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { struct i915_request *rq; - rq = igt_request_alloc(ctx, engine); - if (IS_ERR(rq)) - return PTR_ERR(rq); + rq = intel_context_create_request(ce); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + break; + } i915_request_add(rq); } + i915_gem_context_unlock_engines(ctx); - return 0; + return err; } static void trash_stolen(struct drm_i915_private *i915) @@ -42,6 +45,10 @@ static void trash_stolen(struct drm_i915_private *i915) unsigned long page; u32 prng = 0x12345678; + /* XXX: fsck. needs some more thought... */ + if (!i915_ggtt_has_aperture(ggtt)) + return; + for (page = 0; page < size; page += PAGE_SIZE) { const dma_addr_t dma = i915->dsm.start + page; u32 __iomem *s; @@ -117,7 +124,10 @@ static void pm_resume(struct drm_i915_private *i915) */ with_intel_runtime_pm(&i915->runtime_pm, wakeref) { intel_gt_sanitize(&i915->gt, false); - i915_gem_sanitize(i915); + + i915_gem_restore_gtt_mappings(i915); + i915_gem_restore_fences(&i915->ggtt); + i915_gem_resume(i915); } } @@ -134,11 +144,9 @@ static int igt_gem_suspend(void *arg) return PTR_ERR(file); err = -ENOMEM; - mutex_lock(&i915->drm.struct_mutex); ctx = live_context(i915, file); if (!IS_ERR(ctx)) - err = switch_to_context(i915, ctx); - mutex_unlock(&i915->drm.struct_mutex); + err = switch_to_context(ctx); if (err) goto out; @@ -153,9 +161,7 @@ static int igt_gem_suspend(void *arg) pm_resume(i915); - mutex_lock(&i915->drm.struct_mutex); - err = switch_to_context(i915, ctx); - mutex_unlock(&i915->drm.struct_mutex); + err = switch_to_context(ctx); out: mock_file_free(i915, file); return err; @@ -173,11 +179,9 @@ static int igt_gem_hibernate(void *arg) return PTR_ERR(file); err = -ENOMEM; - mutex_lock(&i915->drm.struct_mutex); ctx = live_context(i915, file); if (!IS_ERR(ctx)) - err = switch_to_context(i915, ctx); - mutex_unlock(&i915->drm.struct_mutex); + err = switch_to_context(ctx); if (err) goto out; @@ -192,9 +196,7 @@ static int igt_gem_hibernate(void *arg) pm_resume(i915); - mutex_lock(&i915->drm.struct_mutex); - err = switch_to_context(i915, ctx); - mutex_unlock(&i915->drm.struct_mutex); + err = switch_to_context(ctx); out: mock_file_free(i915, file); return err; diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c index cb30c669b1b7..42e948144f1b 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c @@ -43,8 +43,7 @@ static void quirk_add(struct drm_i915_gem_object *obj, list_add(&obj->st_link, objects); } -static int populate_ggtt(struct drm_i915_private *i915, - struct list_head *objects) +static int populate_ggtt(struct i915_ggtt *ggtt, struct list_head *objects) { unsigned long unbound, bound, count; struct drm_i915_gem_object *obj; @@ -53,7 +52,8 @@ static int populate_ggtt(struct drm_i915_private *i915, do { struct i915_vma *vma; - obj = i915_gem_object_create_internal(i915, I915_GTT_PAGE_SIZE); + obj = i915_gem_object_create_internal(ggtt->vm.i915, + I915_GTT_PAGE_SIZE); if (IS_ERR(obj)) return PTR_ERR(obj); @@ -70,7 +70,7 @@ static int populate_ggtt(struct drm_i915_private *i915, count++; } while (1); pr_debug("Filled GGTT with %lu pages [%llu total]\n", - count, i915->ggtt.vm.total / PAGE_SIZE); + count, ggtt->vm.total / PAGE_SIZE); bound = 0; unbound = 0; @@ -96,7 +96,7 @@ static int populate_ggtt(struct drm_i915_private *i915, return -EINVAL; } - if (list_empty(&i915->ggtt.vm.bound_list)) { + if (list_empty(&ggtt->vm.bound_list)) { pr_err("No objects on the GGTT inactive list!\n"); return -EINVAL; } @@ -104,20 +104,16 @@ static int populate_ggtt(struct drm_i915_private *i915, return 0; } -static void unpin_ggtt(struct drm_i915_private *i915) +static void unpin_ggtt(struct i915_ggtt *ggtt) { - struct i915_ggtt *ggtt = &i915->ggtt; struct i915_vma *vma; - mutex_lock(&ggtt->vm.mutex); - list_for_each_entry(vma, &i915->ggtt.vm.bound_list, vm_link) + list_for_each_entry(vma, &ggtt->vm.bound_list, vm_link) if (vma->obj->mm.quirked) i915_vma_unpin(vma); - mutex_unlock(&ggtt->vm.mutex); } -static void cleanup_objects(struct drm_i915_private *i915, - struct list_head *list) +static void cleanup_objects(struct i915_ggtt *ggtt, struct list_head *list) { struct drm_i915_gem_object *obj, *on; @@ -127,44 +123,44 @@ static void cleanup_objects(struct drm_i915_private *i915, i915_gem_object_put(obj); } - mutex_unlock(&i915->drm.struct_mutex); - - i915_gem_drain_freed_objects(i915); - - mutex_lock(&i915->drm.struct_mutex); + i915_gem_drain_freed_objects(ggtt->vm.i915); } static int igt_evict_something(void *arg) { - struct drm_i915_private *i915 = arg; - struct i915_ggtt *ggtt = &i915->ggtt; + struct intel_gt *gt = arg; + struct i915_ggtt *ggtt = gt->ggtt; LIST_HEAD(objects); int err; /* Fill the GGTT with pinned objects and try to evict one. */ - err = populate_ggtt(i915, &objects); + err = populate_ggtt(ggtt, &objects); if (err) goto cleanup; /* Everything is pinned, nothing should happen */ + mutex_lock(&ggtt->vm.mutex); err = i915_gem_evict_something(&ggtt->vm, I915_GTT_PAGE_SIZE, 0, 0, 0, U64_MAX, 0); + mutex_unlock(&ggtt->vm.mutex); if (err != -ENOSPC) { pr_err("i915_gem_evict_something failed on a full GGTT with err=%d\n", err); goto cleanup; } - unpin_ggtt(i915); + unpin_ggtt(ggtt); /* Everything is unpinned, we should be able to evict something */ + mutex_lock(&ggtt->vm.mutex); err = i915_gem_evict_something(&ggtt->vm, I915_GTT_PAGE_SIZE, 0, 0, 0, U64_MAX, 0); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_evict_something failed on a full GGTT with err=%d\n", err); @@ -172,13 +168,14 @@ static int igt_evict_something(void *arg) } cleanup: - cleanup_objects(i915, &objects); + cleanup_objects(ggtt, &objects); return err; } static int igt_overcommit(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; + struct i915_ggtt *ggtt = gt->ggtt; struct drm_i915_gem_object *obj; struct i915_vma *vma; LIST_HEAD(objects); @@ -188,11 +185,11 @@ static int igt_overcommit(void *arg) * We expect it to fail. */ - err = populate_ggtt(i915, &objects); + err = populate_ggtt(ggtt, &objects); if (err) goto cleanup; - obj = i915_gem_object_create_internal(i915, I915_GTT_PAGE_SIZE); + obj = i915_gem_object_create_internal(gt->i915, I915_GTT_PAGE_SIZE); if (IS_ERR(obj)) { err = PTR_ERR(obj); goto cleanup; @@ -208,14 +205,14 @@ static int igt_overcommit(void *arg) } cleanup: - cleanup_objects(i915, &objects); + cleanup_objects(ggtt, &objects); return err; } static int igt_evict_for_vma(void *arg) { - struct drm_i915_private *i915 = arg; - struct i915_ggtt *ggtt = &i915->ggtt; + struct intel_gt *gt = arg; + struct i915_ggtt *ggtt = gt->ggtt; struct drm_mm_node target = { .start = 0, .size = 4096, @@ -225,22 +222,26 @@ static int igt_evict_for_vma(void *arg) /* Fill the GGTT with pinned objects and try to evict a range. */ - err = populate_ggtt(i915, &objects); + err = populate_ggtt(ggtt, &objects); if (err) goto cleanup; /* Everything is pinned, nothing should happen */ + mutex_lock(&ggtt->vm.mutex); err = i915_gem_evict_for_node(&ggtt->vm, &target, 0); + mutex_unlock(&ggtt->vm.mutex); if (err != -ENOSPC) { pr_err("i915_gem_evict_for_node on a full GGTT returned err=%d\n", err); goto cleanup; } - unpin_ggtt(i915); + unpin_ggtt(ggtt); /* Everything is unpinned, we should be able to evict the node */ + mutex_lock(&ggtt->vm.mutex); err = i915_gem_evict_for_node(&ggtt->vm, &target, 0); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_evict_for_node returned err=%d\n", err); @@ -248,7 +249,7 @@ static int igt_evict_for_vma(void *arg) } cleanup: - cleanup_objects(i915, &objects); + cleanup_objects(ggtt, &objects); return err; } @@ -261,8 +262,8 @@ static void mock_color_adjust(const struct drm_mm_node *node, static int igt_evict_for_cache_color(void *arg) { - struct drm_i915_private *i915 = arg; - struct i915_ggtt *ggtt = &i915->ggtt; + struct intel_gt *gt = arg; + struct i915_ggtt *ggtt = gt->ggtt; const unsigned long flags = PIN_OFFSET_FIXED; struct drm_mm_node target = { .start = I915_GTT_PAGE_SIZE * 2, @@ -274,14 +275,16 @@ static int igt_evict_for_cache_color(void *arg) LIST_HEAD(objects); int err; - /* Currently the use of color_adjust is limited to cache domains within - * the ggtt, and so the presence of mm.color_adjust is assumed to be - * i915_gtt_color_adjust throughout our driver, so using a mock color - * adjust will work just fine for our purposes. + /* + * Currently the use of color_adjust for the GGTT is limited to cache + * coloring and guard pages, and so the presence of mm.color_adjust for + * the GGTT is assumed to be i915_ggtt_color_adjust, hence using a mock + * color adjust will work just fine for our purposes. */ ggtt->vm.mm.color_adjust = mock_color_adjust; + GEM_BUG_ON(!i915_vm_has_cache_coloring(&ggtt->vm)); - obj = i915_gem_object_create_internal(i915, I915_GTT_PAGE_SIZE); + obj = i915_gem_object_create_internal(gt->i915, I915_GTT_PAGE_SIZE); if (IS_ERR(obj)) { err = PTR_ERR(obj); goto cleanup; @@ -297,7 +300,7 @@ static int igt_evict_for_cache_color(void *arg) goto cleanup; } - obj = i915_gem_object_create_internal(i915, I915_GTT_PAGE_SIZE); + obj = i915_gem_object_create_internal(gt->i915, I915_GTT_PAGE_SIZE); if (IS_ERR(obj)) { err = PTR_ERR(obj); goto cleanup; @@ -317,7 +320,9 @@ static int igt_evict_for_cache_color(void *arg) i915_vma_unpin(vma); /* Remove just the second vma */ + mutex_lock(&ggtt->vm.mutex); err = i915_gem_evict_for_node(&ggtt->vm, &target, 0); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("[0]i915_gem_evict_for_node returned err=%d\n", err); goto cleanup; @@ -328,7 +333,9 @@ static int igt_evict_for_cache_color(void *arg) */ target.color = I915_CACHE_L3_LLC; + mutex_lock(&ggtt->vm.mutex); err = i915_gem_evict_for_node(&ggtt->vm, &target, 0); + mutex_unlock(&ggtt->vm.mutex); if (!err) { pr_err("[1]i915_gem_evict_for_node returned err=%d\n", err); err = -EINVAL; @@ -338,36 +345,40 @@ static int igt_evict_for_cache_color(void *arg) err = 0; cleanup: - unpin_ggtt(i915); - cleanup_objects(i915, &objects); + unpin_ggtt(ggtt); + cleanup_objects(ggtt, &objects); ggtt->vm.mm.color_adjust = NULL; return err; } static int igt_evict_vm(void *arg) { - struct drm_i915_private *i915 = arg; - struct i915_ggtt *ggtt = &i915->ggtt; + struct intel_gt *gt = arg; + struct i915_ggtt *ggtt = gt->ggtt; LIST_HEAD(objects); int err; /* Fill the GGTT with pinned objects and try to evict everything. */ - err = populate_ggtt(i915, &objects); + err = populate_ggtt(ggtt, &objects); if (err) goto cleanup; /* Everything is pinned, nothing should happen */ + mutex_lock(&ggtt->vm.mutex); err = i915_gem_evict_vm(&ggtt->vm); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_evict_vm on a full GGTT returned err=%d]\n", err); goto cleanup; } - unpin_ggtt(i915); + unpin_ggtt(ggtt); + mutex_lock(&ggtt->vm.mutex); err = i915_gem_evict_vm(&ggtt->vm); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_evict_vm on a full GGTT returned err=%d]\n", err); @@ -375,14 +386,16 @@ static int igt_evict_vm(void *arg) } cleanup: - cleanup_objects(i915, &objects); + cleanup_objects(ggtt, &objects); return err; } static int igt_evict_contexts(void *arg) { const u64 PRETEND_GGTT_SIZE = 16ull << 20; - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; + struct i915_ggtt *ggtt = gt->ggtt; + struct drm_i915_private *i915 = gt->i915; struct intel_engine_cs *engine; enum intel_engine_id id; struct reserved { @@ -408,14 +421,14 @@ static int igt_evict_contexts(void *arg) if (!HAS_FULL_PPGTT(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); wakeref = intel_runtime_pm_get(&i915->runtime_pm); /* Reserve a block so that we know we have enough to fit a few rq */ memset(&hole, 0, sizeof(hole)); - err = i915_gem_gtt_insert(&i915->ggtt.vm, &hole, + mutex_lock(&ggtt->vm.mutex); + err = i915_gem_gtt_insert(&ggtt->vm, &hole, PRETEND_GGTT_SIZE, 0, I915_COLOR_UNEVICTABLE, - 0, i915->ggtt.vm.total, + 0, ggtt->vm.total, PIN_NOEVICT); if (err) goto out_locked; @@ -425,15 +438,17 @@ static int igt_evict_contexts(void *arg) do { struct reserved *r; + mutex_unlock(&ggtt->vm.mutex); r = kcalloc(1, sizeof(*r), GFP_KERNEL); + mutex_lock(&ggtt->vm.mutex); if (!r) { err = -ENOMEM; goto out_locked; } - if (i915_gem_gtt_insert(&i915->ggtt.vm, &r->node, + if (i915_gem_gtt_insert(&ggtt->vm, &r->node, 1ul << 20, 0, I915_COLOR_UNEVICTABLE, - 0, i915->ggtt.vm.total, + 0, ggtt->vm.total, PIN_NOEVICT)) { kfree(r); break; @@ -445,11 +460,11 @@ static int igt_evict_contexts(void *arg) count++; } while (1); drm_mm_remove_node(&hole); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(&ggtt->vm.mutex); pr_info("Filled GGTT with %lu 1MiB nodes\n", count); /* Overfill the GGTT with context objects and so try to evict one. */ - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { struct i915_sw_fence fence; struct drm_file *file; @@ -460,7 +475,6 @@ static int igt_evict_contexts(void *arg) } count = 0; - mutex_lock(&i915->drm.struct_mutex); onstack_fence_init(&fence); do { struct i915_request *rq; @@ -478,8 +492,8 @@ static int igt_evict_contexts(void *arg) if (IS_ERR(rq)) { /* When full, fail_if_busy will trigger EBUSY */ if (PTR_ERR(rq) != -EBUSY) { - pr_err("Unexpected error from request alloc (ctx hw id %u, on %s): %d\n", - ctx->hw_id, engine->name, + pr_err("Unexpected error from request alloc (on %s): %d\n", + engine->name, (int)PTR_ERR(rq)); err = PTR_ERR(rq); } @@ -497,8 +511,6 @@ static int igt_evict_contexts(void *arg) count++; err = 0; } while(1); - mutex_unlock(&i915->drm.struct_mutex); - onstack_fence_fini(&fence); pr_info("Submitted %lu contexts/requests on %s\n", count, engine->name); @@ -508,9 +520,9 @@ static int igt_evict_contexts(void *arg) break; } - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(&ggtt->vm.mutex); out_locked: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; while (reserved) { struct reserved *next = reserved->next; @@ -522,8 +534,8 @@ out_locked: } if (drm_mm_node_allocated(&hole)) drm_mm_remove_node(&hole); + mutex_unlock(&ggtt->vm.mutex); intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -545,11 +557,8 @@ int i915_gem_evict_mock_selftests(void) if (!i915) return -ENOMEM; - mutex_lock(&i915->drm.struct_mutex); with_intel_runtime_pm(&i915->runtime_pm, wakeref) - err = i915_subtests(tests, i915); - - mutex_unlock(&i915->drm.struct_mutex); + err = i915_subtests(tests, &i915->gt); drm_dev_put(&i915->drm); return err; @@ -564,5 +573,5 @@ int i915_gem_evict_live_selftests(struct drm_i915_private *i915) if (intel_gt_is_wedged(&i915->gt)) return 0; - return i915_subtests(tests, i915); + return intel_gt_live_subtests(tests, &i915->gt); } diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c index 31a51ca1ddcb..3f7e80fb3bbd 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c @@ -25,26 +25,20 @@ #include <linux/list_sort.h> #include <linux/prime_numbers.h> +#include "gem/i915_gem_context.h" #include "gem/selftests/mock_context.h" +#include "gt/intel_context.h" #include "i915_random.h" #include "i915_selftest.h" #include "mock_drm.h" #include "mock_gem_device.h" +#include "igt_flush_test.h" static void cleanup_freed_objects(struct drm_i915_private *i915) { - /* - * As we may hold onto the struct_mutex for inordinate lengths of - * time, the NMI khungtaskd detector may fire for the free objects - * worker. - */ - mutex_unlock(&i915->drm.struct_mutex); - i915_gem_drain_freed_objects(i915); - - mutex_lock(&i915->drm.struct_mutex); } static void fake_free_pages(struct drm_i915_gem_object *obj, @@ -88,8 +82,6 @@ static int fake_get_pages(struct drm_i915_gem_object *obj) } GEM_BUG_ON(rem); - obj->mm.madv = I915_MADV_DONTNEED; - __i915_gem_object_set_pages(obj, pages, sg_page_sizes); return 0; @@ -101,7 +93,6 @@ static void fake_put_pages(struct drm_i915_gem_object *obj, { fake_free_pages(obj, pages); obj->mm.dirty = false; - obj->mm.madv = I915_MADV_WILLNEED; } static const struct drm_i915_gem_object_ops fake_ops = { @@ -113,6 +104,7 @@ static const struct drm_i915_gem_object_ops fake_ops = { static struct drm_i915_gem_object * fake_dma_object(struct drm_i915_private *i915, u64 size) { + static struct lock_class_key lock_class; struct drm_i915_gem_object *obj; GEM_BUG_ON(!size); @@ -126,7 +118,9 @@ fake_dma_object(struct drm_i915_private *i915, u64 size) goto err; drm_gem_private_object_init(&i915->drm, &obj->base, size); - i915_gem_object_init(obj, &fake_ops); + i915_gem_object_init(obj, &fake_ops, &lock_class); + + i915_gem_object_set_volatile(obj); obj->write_domain = I915_GEM_DOMAIN_CPU; obj->read_domains = I915_GEM_DOMAIN_CPU; @@ -293,18 +287,20 @@ static int lowlevel_hole(struct drm_i915_private *i915, mock_vma.node.size = BIT_ULL(size); mock_vma.node.start = addr; - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - vm->insert_entries(vm, &mock_vma, I915_CACHE_NONE, 0); - intel_runtime_pm_put(&i915->runtime_pm, wakeref); + with_intel_runtime_pm(&i915->runtime_pm, wakeref) + vm->insert_entries(vm, &mock_vma, + I915_CACHE_NONE, 0); } count = n; i915_random_reorder(order, count, &prng); for (n = 0; n < count; n++) { u64 addr = hole_start + order[n] * BIT_ULL(size); + intel_wakeref_t wakeref; GEM_BUG_ON(addr + BIT_ULL(size) > vm->total); - vm->clear_range(vm, addr, BIT_ULL(size)); + with_intel_runtime_pm(&i915->runtime_pm, wakeref) + vm->clear_range(vm, addr, BIT_ULL(size)); } i915_gem_object_unpin_pages(obj); @@ -875,6 +871,15 @@ static int __shrink_hole(struct drm_i915_private *i915, i915_vma_unpin(vma); addr += size; + /* + * Since we are injecting allocation faults at random intervals, + * wait for this allocation to complete before we change the + * faultinjection. + */ + err = i915_vma_sync(vma); + if (err) + break; + if (igt_timeout(end_time, "%s timed out at ofset %llx [%llx - %llx]\n", __func__, addr, hole_start, hole_end)) { @@ -1008,21 +1013,19 @@ static int exercise_ppgtt(struct drm_i915_private *dev_priv, if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&dev_priv->drm.struct_mutex); ppgtt = i915_ppgtt_create(dev_priv); if (IS_ERR(ppgtt)) { err = PTR_ERR(ppgtt); - goto out_unlock; + goto out_free; } GEM_BUG_ON(offset_in_page(ppgtt->vm.total)); - GEM_BUG_ON(ppgtt->vm.closed); + GEM_BUG_ON(!atomic_read(&ppgtt->vm.open)); err = func(dev_priv, &ppgtt->vm, 0, ppgtt->vm.total, end_time); i915_vm_put(&ppgtt->vm); -out_unlock: - mutex_unlock(&dev_priv->drm.struct_mutex); +out_free: mock_file_free(dev_priv, file); return err; } @@ -1085,7 +1088,6 @@ static int exercise_ggtt(struct drm_i915_private *i915, IGT_TIMEOUT(end_time); int err = 0; - mutex_lock(&i915->drm.struct_mutex); restart: list_sort(NULL, &ggtt->vm.mm.hole_stack, sort_holes); drm_mm_for_each_hole(node, &ggtt->vm.mm, hole_start, hole_end) { @@ -1106,7 +1108,6 @@ restart: last = hole_end; goto restart; } - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -1148,13 +1149,12 @@ static int igt_ggtt_page(void *arg) unsigned int *order, n; int err; - mutex_lock(&i915->drm.struct_mutex); + if (!i915_ggtt_has_aperture(ggtt)) + return 0; obj = i915_gem_object_create_internal(i915, PAGE_SIZE); - if (IS_ERR(obj)) { - err = PTR_ERR(obj); - goto out_unlock; - } + if (IS_ERR(obj)) + return PTR_ERR(obj); err = i915_gem_object_pin_pages(obj); if (err) @@ -1222,8 +1222,6 @@ out_unpin: i915_gem_object_unpin_pages(obj); out_free: i915_gem_object_put(obj); -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -1234,10 +1232,13 @@ static void track_vma_bind(struct i915_vma *vma) atomic_inc(&obj->bind_count); /* track for eviction later */ __i915_gem_object_pin_pages(obj); + GEM_BUG_ON(vma->pages); + atomic_set(&vma->pages_count, I915_VMA_PAGES_ACTIVE); + __i915_gem_object_pin_pages(obj); vma->pages = obj->mm.pages; mutex_lock(&vma->vm->mutex); - list_move_tail(&vma->vm_link, &vma->vm->bound_list); + list_add_tail(&vma->vm_link, &vma->vm->bound_list); mutex_unlock(&vma->vm->mutex); } @@ -1248,6 +1249,7 @@ static int exercise_mock(struct drm_i915_private *i915, unsigned long end_time)) { const u64 limit = totalram_pages() << PAGE_SHIFT; + struct i915_address_space *vm; struct i915_gem_context *ctx; IGT_TIMEOUT(end_time); int err; @@ -1256,7 +1258,9 @@ static int exercise_mock(struct drm_i915_private *i915, if (!ctx) return -ENOMEM; - err = func(i915, ctx->vm, 0, min(ctx->vm->total, limit), end_time); + vm = i915_gem_context_get_vm_rcu(ctx); + err = func(i915, vm, 0, min(vm->total, limit), end_time); + i915_vm_put(vm); mock_context_close(ctx); return err; @@ -1294,6 +1298,7 @@ static int igt_gtt_reserve(void *arg) { struct i915_ggtt *ggtt = arg; struct drm_i915_gem_object *obj, *on; + I915_RND_STATE(prng); LIST_HEAD(objects); u64 total; int err = -ENODEV; @@ -1330,11 +1335,13 @@ static int igt_gtt_reserve(void *arg) goto out; } + mutex_lock(&ggtt->vm.mutex); err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node, obj->base.size, total, obj->cache_level, 0); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_gtt_reserve (pass 1) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1380,11 +1387,13 @@ static int igt_gtt_reserve(void *arg) goto out; } + mutex_lock(&ggtt->vm.mutex); err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node, obj->base.size, total, obj->cache_level, 0); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_gtt_reserve (pass 2) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1420,15 +1429,18 @@ static int igt_gtt_reserve(void *arg) goto out; } - offset = random_offset(0, ggtt->vm.total, - 2*I915_GTT_PAGE_SIZE, - I915_GTT_MIN_ALIGNMENT); + offset = igt_random_offset(&prng, + 0, ggtt->vm.total, + 2 * I915_GTT_PAGE_SIZE, + I915_GTT_MIN_ALIGNMENT); + mutex_lock(&ggtt->vm.mutex); err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node, obj->base.size, offset, obj->cache_level, 0); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_gtt_reserve (pass 3) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1497,11 +1509,13 @@ static int igt_gtt_insert(void *arg) /* Check a couple of obviously invalid requests */ for (ii = invalid_insert; ii->size; ii++) { + mutex_lock(&ggtt->vm.mutex); err = i915_gem_gtt_insert(&ggtt->vm, &tmp, ii->size, ii->alignment, I915_COLOR_UNEVICTABLE, ii->start, ii->end, 0); + mutex_unlock(&ggtt->vm.mutex); if (err != -ENOSPC) { pr_err("Invalid i915_gem_gtt_insert(.size=%llx, .alignment=%llx, .start=%llx, .end=%llx) succeeded (err=%d)\n", ii->size, ii->alignment, ii->start, ii->end, @@ -1537,10 +1551,12 @@ static int igt_gtt_insert(void *arg) goto out; } + mutex_lock(&ggtt->vm.mutex); err = i915_gem_gtt_insert(&ggtt->vm, &vma->node, obj->base.size, 0, obj->cache_level, 0, ggtt->vm.total, 0); + mutex_unlock(&ggtt->vm.mutex); if (err == -ENOSPC) { /* maxed out the GGTT space */ i915_gem_object_put(obj); @@ -1595,10 +1611,12 @@ static int igt_gtt_insert(void *arg) goto out; } + mutex_lock(&ggtt->vm.mutex); err = i915_gem_gtt_insert(&ggtt->vm, &vma->node, obj->base.size, 0, obj->cache_level, 0, ggtt->vm.total, 0); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_gtt_insert (pass 2) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1642,10 +1660,12 @@ static int igt_gtt_insert(void *arg) goto out; } + mutex_lock(&ggtt->vm.mutex); err = i915_gem_gtt_insert(&ggtt->vm, &vma->node, obj->base.size, 0, obj->cache_level, 0, ggtt->vm.total, 0); + mutex_unlock(&ggtt->vm.mutex); if (err) { pr_err("i915_gem_gtt_insert (pass 3) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1689,13 +1709,10 @@ int i915_gem_gtt_mock_selftests(void) } mock_init_ggtt(i915, ggtt); - mutex_lock(&i915->drm.struct_mutex); err = i915_subtests(tests, ggtt); - mock_device_flush(i915); - mutex_unlock(&i915->drm.struct_mutex); + mock_device_flush(i915); i915_gem_drain_freed_objects(i915); - mock_fini_ggtt(ggtt); kfree(ggtt); out_put: @@ -1703,6 +1720,312 @@ out_put: return err; } +static int context_sync(struct intel_context *ce) +{ + struct i915_request *rq; + long timeout; + + rq = intel_context_create_request(ce); + if (IS_ERR(rq)) + return PTR_ERR(rq); + + i915_request_get(rq); + i915_request_add(rq); + + timeout = i915_request_wait(rq, 0, HZ / 5); + i915_request_put(rq); + + return timeout < 0 ? -EIO : 0; +} + +static struct i915_request * +submit_batch(struct intel_context *ce, u64 addr) +{ + struct i915_request *rq; + int err; + + rq = intel_context_create_request(ce); + if (IS_ERR(rq)) + return rq; + + err = 0; + if (rq->engine->emit_init_breadcrumb) /* detect a hang */ + err = rq->engine->emit_init_breadcrumb(rq); + if (err == 0) + err = rq->engine->emit_bb_start(rq, addr, 0, 0); + + if (err == 0) + i915_request_get(rq); + i915_request_add(rq); + + return err ? ERR_PTR(err) : rq; +} + +static u32 *spinner(u32 *batch, int i) +{ + return batch + i * 64 / sizeof(*batch) + 4; +} + +static void end_spin(u32 *batch, int i) +{ + *spinner(batch, i) = MI_BATCH_BUFFER_END; + wmb(); +} + +static int igt_cs_tlb(void *arg) +{ + const unsigned int count = PAGE_SIZE / 64; + const unsigned int chunk_size = count * PAGE_SIZE; + struct drm_i915_private *i915 = arg; + struct drm_i915_gem_object *bbe, *act, *out; + struct i915_gem_engines_iter it; + struct i915_address_space *vm; + struct i915_gem_context *ctx; + struct intel_context *ce; + struct drm_file *file; + struct i915_vma *vma; + I915_RND_STATE(prng); + unsigned int i; + u32 *result; + u32 *batch; + int err = 0; + + /* + * Our mission here is to fool the hardware to execute something + * from scratch as it has not seen the batch move (due to missing + * the TLB invalidate). + */ + + file = mock_file(i915); + if (IS_ERR(file)) + return PTR_ERR(file); + + ctx = live_context(i915, file); + if (IS_ERR(ctx)) { + err = PTR_ERR(ctx); + goto out_unlock; + } + + vm = i915_gem_context_get_vm_rcu(ctx); + if (i915_is_ggtt(vm)) + goto out_vm; + + /* Create two pages; dummy we prefill the TLB, and intended */ + bbe = i915_gem_object_create_internal(i915, PAGE_SIZE); + if (IS_ERR(bbe)) { + err = PTR_ERR(bbe); + goto out_vm; + } + + batch = i915_gem_object_pin_map(bbe, I915_MAP_WC); + if (IS_ERR(batch)) { + err = PTR_ERR(batch); + goto out_put_bbe; + } + memset32(batch, MI_BATCH_BUFFER_END, PAGE_SIZE / sizeof(u32)); + i915_gem_object_flush_map(bbe); + i915_gem_object_unpin_map(bbe); + + act = i915_gem_object_create_internal(i915, PAGE_SIZE); + if (IS_ERR(act)) { + err = PTR_ERR(act); + goto out_put_bbe; + } + + /* Track the execution of each request by writing into different slot */ + batch = i915_gem_object_pin_map(act, I915_MAP_WC); + if (IS_ERR(batch)) { + err = PTR_ERR(batch); + goto out_put_act; + } + for (i = 0; i < count; i++) { + u32 *cs = batch + i * 64 / sizeof(*cs); + u64 addr = (vm->total - PAGE_SIZE) + i * sizeof(u32); + + GEM_BUG_ON(INTEL_GEN(i915) < 6); + cs[0] = MI_STORE_DWORD_IMM_GEN4; + if (INTEL_GEN(i915) >= 8) { + cs[1] = lower_32_bits(addr); + cs[2] = upper_32_bits(addr); + cs[3] = i; + cs[4] = MI_NOOP; + cs[5] = MI_BATCH_BUFFER_START_GEN8; + } else { + cs[1] = 0; + cs[2] = lower_32_bits(addr); + cs[3] = i; + cs[4] = MI_NOOP; + cs[5] = MI_BATCH_BUFFER_START; + } + } + + out = i915_gem_object_create_internal(i915, PAGE_SIZE); + if (IS_ERR(out)) { + err = PTR_ERR(out); + goto out_put_batch; + } + i915_gem_object_set_cache_coherency(out, I915_CACHING_CACHED); + + vma = i915_vma_instance(out, vm, NULL); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto out_put_batch; + } + + err = i915_vma_pin(vma, 0, 0, + PIN_USER | + PIN_OFFSET_FIXED | + (vm->total - PAGE_SIZE)); + if (err) + goto out_put_out; + GEM_BUG_ON(vma->node.start != vm->total - PAGE_SIZE); + + result = i915_gem_object_pin_map(out, I915_MAP_WB); + if (IS_ERR(result)) { + err = PTR_ERR(result); + goto out_put_out; + } + + for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { + IGT_TIMEOUT(end_time); + unsigned long pass = 0; + + if (!intel_engine_can_store_dword(ce->engine)) + continue; + + while (!__igt_timeout(end_time, NULL)) { + struct i915_request *rq; + u64 offset; + + offset = igt_random_offset(&prng, + 0, vm->total - PAGE_SIZE, + chunk_size, PAGE_SIZE); + + err = vm->allocate_va_range(vm, offset, chunk_size); + if (err) + goto end; + + memset32(result, STACK_MAGIC, PAGE_SIZE / sizeof(u32)); + + vma = i915_vma_instance(bbe, vm, NULL); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto end; + } + + err = vma->ops->set_pages(vma); + if (err) + goto end; + + /* Prime the TLB with the dummy pages */ + for (i = 0; i < count; i++) { + vma->node.start = offset + i * PAGE_SIZE; + vm->insert_entries(vm, vma, I915_CACHE_NONE, 0); + + rq = submit_batch(ce, vma->node.start); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto end; + } + i915_request_put(rq); + } + + vma->ops->clear_pages(vma); + + err = context_sync(ce); + if (err) { + pr_err("%s: dummy setup timed out\n", + ce->engine->name); + goto end; + } + + vma = i915_vma_instance(act, vm, NULL); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto end; + } + + err = vma->ops->set_pages(vma); + if (err) + goto end; + + /* Replace the TLB with target batches */ + for (i = 0; i < count; i++) { + struct i915_request *rq; + u32 *cs = batch + i * 64 / sizeof(*cs); + u64 addr; + + vma->node.start = offset + i * PAGE_SIZE; + vm->insert_entries(vm, vma, I915_CACHE_NONE, 0); + + addr = vma->node.start + i * 64; + cs[4] = MI_NOOP; + cs[6] = lower_32_bits(addr); + cs[7] = upper_32_bits(addr); + wmb(); + + rq = submit_batch(ce, addr); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto end; + } + + /* Wait until the context chain has started */ + if (i == 0) { + while (READ_ONCE(result[i]) && + !i915_request_completed(rq)) + cond_resched(); + } else { + end_spin(batch, i - 1); + } + + i915_request_put(rq); + } + end_spin(batch, count - 1); + + vma->ops->clear_pages(vma); + + err = context_sync(ce); + if (err) { + pr_err("%s: writes timed out\n", + ce->engine->name); + goto end; + } + + for (i = 0; i < count; i++) { + if (result[i] != i) { + pr_err("%s: Write lost on pass %lu, at offset %llx, index %d, found %x, expected %x\n", + ce->engine->name, pass, + offset, i, result[i], i); + err = -EINVAL; + goto end; + } + } + + vm->clear_range(vm, offset, chunk_size); + pass++; + } + } +end: + if (igt_flush_test(i915)) + err = -EIO; + i915_gem_context_unlock_engines(ctx); + i915_gem_object_unpin_map(out); +out_put_out: + i915_gem_object_put(out); +out_put_batch: + i915_gem_object_unpin_map(act); +out_put_act: + i915_gem_object_put(act); +out_put_bbe: + i915_gem_object_put(bbe); +out_vm: + i915_vm_put(vm); +out_unlock: + mock_file_free(i915, file); + return err; +} + int i915_gem_gtt_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { @@ -1720,6 +2043,7 @@ int i915_gem_gtt_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_ggtt_pot), SUBTEST(igt_ggtt_fill), SUBTEST(igt_ggtt_page), + SUBTEST(igt_cs_tlb), }; GEM_BUG_ON(offset_in_page(i915->ggtt.vm.total)); diff --git a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h index 1ccf0f731ac0..4b3cac73e291 100644 --- a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h +++ b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h @@ -15,6 +15,9 @@ selftest(workarounds, intel_workarounds_live_selftests) selftest(gt_engines, intel_engine_live_selftests) selftest(gt_timelines, intel_timeline_live_selftests) selftest(gt_contexts, intel_context_live_selftests) +selftest(gt_lrc, intel_lrc_live_selftests) +selftest(gt_pm, intel_gt_pm_live_selftests) +selftest(gt_heartbeat, intel_heartbeat_live_selftests) selftest(requests, i915_request_live_selftests) selftest(active, i915_active_live_selftests) selftest(objects, i915_gem_object_live_selftests) @@ -30,6 +33,8 @@ selftest(gem_contexts, i915_gem_context_live_selftests) selftest(blt, i915_gem_object_blt_live_selftests) selftest(client, i915_gem_client_blt_live_selftests) selftest(reset, intel_reset_live_selftests) +selftest(memory_region, intel_memory_region_live_selftests) selftest(hangcheck, intel_hangcheck_live_selftests) selftest(execlists, intel_execlists_live_selftests) selftest(guc, intel_guc_live_selftest) +selftest(perf, i915_perf_live_selftests) diff --git a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h index b88084fe3269..aa5a0e7f5d9e 100644 --- a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h +++ b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h @@ -26,3 +26,4 @@ selftest(gtt, i915_gem_gtt_mock_selftests) selftest(hugepages, i915_gem_huge_page_mock_selftests) selftest(contexts, i915_gem_context_mock_selftests) selftest(buddy, i915_buddy_mock_selftests) +selftest(memory_region, intel_memory_region_mock_selftests) diff --git a/drivers/gpu/drm/i915/selftests/i915_perf.c b/drivers/gpu/drm/i915/selftests/i915_perf.c new file mode 100644 index 000000000000..aabd07f67e49 --- /dev/null +++ b/drivers/gpu/drm/i915/selftests/i915_perf.c @@ -0,0 +1,217 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include <linux/kref.h> + +#include "gem/i915_gem_pm.h" +#include "gt/intel_gt.h" + +#include "i915_selftest.h" + +#include "igt_flush_test.h" +#include "lib_sw_fence.h" + +static struct i915_perf_stream * +test_stream(struct i915_perf *perf) +{ + struct drm_i915_perf_open_param param = {}; + struct perf_open_properties props = { + .engine = intel_engine_lookup_user(perf->i915, + I915_ENGINE_CLASS_RENDER, + 0), + .sample_flags = SAMPLE_OA_REPORT, + .oa_format = IS_GEN(perf->i915, 12) ? + I915_OA_FORMAT_A32u40_A4u32_B8_C8 : I915_OA_FORMAT_C4_B8, + .metrics_set = 1, + }; + struct i915_perf_stream *stream; + + stream = kzalloc(sizeof(*stream), GFP_KERNEL); + if (!stream) + return NULL; + + stream->perf = perf; + + mutex_lock(&perf->lock); + if (i915_oa_stream_init(stream, ¶m, &props)) { + kfree(stream); + stream = NULL; + } + mutex_unlock(&perf->lock); + + return stream; +} + +static void stream_destroy(struct i915_perf_stream *stream) +{ + struct i915_perf *perf = stream->perf; + + mutex_lock(&perf->lock); + i915_perf_destroy_locked(stream); + mutex_unlock(&perf->lock); +} + +static int live_sanitycheck(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct i915_perf_stream *stream; + + /* Quick check we can create a perf stream */ + + stream = test_stream(&i915->perf); + if (!stream) + return -EINVAL; + + stream_destroy(stream); + return 0; +} + +static int write_timestamp(struct i915_request *rq, int slot) +{ + u32 *cs; + int len; + + cs = intel_ring_begin(rq, 6); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + len = 5; + if (INTEL_GEN(rq->i915) >= 8) + len++; + + *cs++ = GFX_OP_PIPE_CONTROL(len); + *cs++ = PIPE_CONTROL_GLOBAL_GTT_IVB | + PIPE_CONTROL_STORE_DATA_INDEX | + PIPE_CONTROL_WRITE_TIMESTAMP; + *cs++ = slot * sizeof(u32); + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + + intel_ring_advance(rq, cs); + + return 0; +} + +static ktime_t poll_status(struct i915_request *rq, int slot) +{ + while (!intel_read_status_page(rq->engine, slot) && + !i915_request_completed(rq)) + cpu_relax(); + + return ktime_get(); +} + +static int live_noa_delay(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct i915_perf_stream *stream; + struct i915_request *rq; + ktime_t t0, t1; + u64 expected; + u32 delay; + int err; + int i; + + /* Check that the GPU delays matches expectations */ + + stream = test_stream(&i915->perf); + if (!stream) + return -ENOMEM; + + expected = atomic64_read(&stream->perf->noa_programming_delay); + + if (stream->engine->class != RENDER_CLASS) { + err = -ENODEV; + goto out; + } + + for (i = 0; i < 4; i++) + intel_write_status_page(stream->engine, 0x100 + i, 0); + + rq = i915_request_create(stream->engine->kernel_context); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto out; + } + + if (rq->engine->emit_init_breadcrumb && + i915_request_timeline(rq)->has_initial_breadcrumb) { + err = rq->engine->emit_init_breadcrumb(rq); + if (err) { + i915_request_add(rq); + goto out; + } + } + + err = write_timestamp(rq, 0x100); + if (err) { + i915_request_add(rq); + goto out; + } + + err = rq->engine->emit_bb_start(rq, + i915_ggtt_offset(stream->noa_wait), 0, + I915_DISPATCH_SECURE); + if (err) { + i915_request_add(rq); + goto out; + } + + err = write_timestamp(rq, 0x102); + if (err) { + i915_request_add(rq); + goto out; + } + + i915_request_get(rq); + i915_request_add(rq); + + preempt_disable(); + t0 = poll_status(rq, 0x100); + t1 = poll_status(rq, 0x102); + preempt_enable(); + + pr_info("CPU delay: %lluns, expected %lluns\n", + ktime_sub(t1, t0), expected); + + delay = intel_read_status_page(stream->engine, 0x102); + delay -= intel_read_status_page(stream->engine, 0x100); + delay = div_u64(mul_u32_u32(delay, 1000 * 1000), + RUNTIME_INFO(i915)->cs_timestamp_frequency_khz); + pr_info("GPU delay: %uns, expected %lluns\n", + delay, expected); + + if (4 * delay < 3 * expected || 2 * delay > 3 * expected) { + pr_err("GPU delay [%uus] outside of expected threshold! [%lluus, %lluus]\n", + delay / 1000, + div_u64(3 * expected, 4000), + div_u64(3 * expected, 2000)); + err = -EINVAL; + } + + i915_request_put(rq); +out: + stream_destroy(stream); + return err; +} + +int i915_perf_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(live_sanitycheck), + SUBTEST(live_noa_delay), + }; + struct i915_perf *perf = &i915->perf; + + if (!perf->metrics_kobj || !perf->ops.enable_metric_set) + return 0; + + if (intel_gt_is_wedged(&i915->gt)) + return 0; + + return i915_subtests(tests, i915); +} diff --git a/drivers/gpu/drm/i915/selftests/i915_random.c b/drivers/gpu/drm/i915/selftests/i915_random.c index 716a3f19f030..abdfadcf626b 100644 --- a/drivers/gpu/drm/i915/selftests/i915_random.c +++ b/drivers/gpu/drm/i915/selftests/i915_random.c @@ -29,6 +29,7 @@ #include <linux/types.h> #include "i915_random.h" +#include "i915_utils.h" u64 i915_prandom_u64_state(struct rnd_state *rnd) { @@ -87,3 +88,22 @@ unsigned int *i915_random_order(unsigned int count, struct rnd_state *state) i915_random_reorder(order, count, state); return order; } + +u64 igt_random_offset(struct rnd_state *state, + u64 start, u64 end, + u64 len, u64 align) +{ + u64 range, addr; + + BUG_ON(range_overflows(start, len, end)); + BUG_ON(round_up(start, align) > round_down(end - len, align)); + + range = round_down(end - len, align) - round_up(start, align); + if (range) { + addr = i915_prandom_u64_state(state); + div64_u64_rem(addr, range, &addr); + start += addr; + } + + return round_up(start, align); +} diff --git a/drivers/gpu/drm/i915/selftests/i915_random.h b/drivers/gpu/drm/i915/selftests/i915_random.h index 8e1ff9c105b6..35cc69a3a1b9 100644 --- a/drivers/gpu/drm/i915/selftests/i915_random.h +++ b/drivers/gpu/drm/i915/selftests/i915_random.h @@ -57,4 +57,8 @@ void i915_random_reorder(unsigned int *order, void i915_prandom_shuffle(void *arr, size_t elsz, size_t count, struct rnd_state *state); +u64 igt_random_offset(struct rnd_state *state, + u64 start, u64 end, + u64 len, u64 align); + #endif /* !__I915_SELFTESTS_RANDOM_H__ */ diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c index b3688543ed7d..8618a4dc0701 100644 --- a/drivers/gpu/drm/i915/selftests/i915_request.c +++ b/drivers/gpu/drm/i915/selftests/i915_request.c @@ -37,25 +37,32 @@ #include "mock_drm.h" #include "mock_gem_device.h" +static unsigned int num_uabi_engines(struct drm_i915_private *i915) +{ + struct intel_engine_cs *engine; + unsigned int count; + + count = 0; + for_each_uabi_engine(engine, i915) + count++; + + return count; +} + static int igt_add_request(void *arg) { struct drm_i915_private *i915 = arg; struct i915_request *request; - int err = -ENOMEM; /* Basic preliminary test to create a request and let it loose! */ - mutex_lock(&i915->drm.struct_mutex); request = mock_request(i915->engine[RCS0]->kernel_context, HZ / 10); if (!request) - goto out_unlock; + return -ENOMEM; i915_request_add(request); - err = 0; -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - return err; + return 0; } static int igt_wait_request(void *arg) @@ -67,12 +74,10 @@ static int igt_wait_request(void *arg) /* Submit a request, then wait upon it */ - mutex_lock(&i915->drm.struct_mutex); request = mock_request(i915->engine[RCS0]->kernel_context, T); - if (!request) { - err = -ENOMEM; - goto out_unlock; - } + if (!request) + return -ENOMEM; + i915_request_get(request); if (i915_request_wait(request, 0, 0) != -ETIME) { @@ -125,9 +130,7 @@ static int igt_wait_request(void *arg) err = 0; out_request: i915_request_put(request); -out_unlock: mock_device_flush(i915); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -140,52 +143,45 @@ static int igt_fence_wait(void *arg) /* Submit a request, treat it as a fence and wait upon it */ - mutex_lock(&i915->drm.struct_mutex); request = mock_request(i915->engine[RCS0]->kernel_context, T); - if (!request) { - err = -ENOMEM; - goto out_locked; - } + if (!request) + return -ENOMEM; if (dma_fence_wait_timeout(&request->fence, false, T) != -ETIME) { pr_err("fence wait success before submit (expected timeout)!\n"); - goto out_locked; + goto out; } i915_request_add(request); - mutex_unlock(&i915->drm.struct_mutex); if (dma_fence_is_signaled(&request->fence)) { pr_err("fence signaled immediately!\n"); - goto out_device; + goto out; } if (dma_fence_wait_timeout(&request->fence, false, T / 2) != -ETIME) { pr_err("fence wait success after submit (expected timeout)!\n"); - goto out_device; + goto out; } if (dma_fence_wait_timeout(&request->fence, false, T) <= 0) { pr_err("fence wait timed out (expected success)!\n"); - goto out_device; + goto out; } if (!dma_fence_is_signaled(&request->fence)) { pr_err("fence unsignaled after waiting!\n"); - goto out_device; + goto out; } if (dma_fence_wait_timeout(&request->fence, false, T) <= 0) { pr_err("fence wait timed out when complete (expected success)!\n"); - goto out_device; + goto out; } err = 0; -out_device: - mutex_lock(&i915->drm.struct_mutex); -out_locked: +out: mock_device_flush(i915); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -197,8 +193,8 @@ static int igt_request_rewind(void *arg) struct intel_context *ce; int err = -EINVAL; - mutex_lock(&i915->drm.struct_mutex); ctx[0] = mock_context(i915, "A"); + ce = i915_gem_context_get_engine(ctx[0], RCS0); GEM_BUG_ON(IS_ERR(ce)); request = mock_request(ce, 2 * HZ); @@ -212,6 +208,7 @@ static int igt_request_rewind(void *arg) i915_request_add(request); ctx[1] = mock_context(i915, "B"); + ce = i915_gem_context_get_engine(ctx[1], RCS0); GEM_BUG_ON(IS_ERR(ce)); vip = mock_request(ce, 0); @@ -233,7 +230,6 @@ static int igt_request_rewind(void *arg) request->engine->submit_request(request); rcu_read_unlock(); - mutex_unlock(&i915->drm.struct_mutex); if (i915_request_wait(vip, 0, HZ) == -ETIME) { pr_err("timed out waiting for high priority request\n"); @@ -248,14 +244,12 @@ static int igt_request_rewind(void *arg) err = 0; err: i915_request_put(vip); - mutex_lock(&i915->drm.struct_mutex); err_context_1: mock_context_close(ctx[1]); i915_request_put(request); err_context_0: mock_context_close(ctx[0]); mock_device_flush(i915); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -282,7 +276,6 @@ __live_request_alloc(struct intel_context *ce) static int __igt_breadcrumbs_smoketest(void *arg) { struct smoketest *t = arg; - struct mutex * const BKL = &t->engine->i915->drm.struct_mutex; const unsigned int max_batch = min(t->ncontexts, t->max_batch) - 1; const unsigned int total = 4 * t->ncontexts + 1; unsigned int num_waits = 0, num_fences = 0; @@ -300,7 +293,7 @@ static int __igt_breadcrumbs_smoketest(void *arg) * that the fences were marked as signaled. */ - requests = kmalloc_array(total, sizeof(*requests), GFP_KERNEL); + requests = kcalloc(total, sizeof(*requests), GFP_KERNEL); if (!requests) return -ENOMEM; @@ -337,14 +330,11 @@ static int __igt_breadcrumbs_smoketest(void *arg) struct i915_request *rq; struct intel_context *ce; - mutex_lock(BKL); - ce = i915_gem_context_get_engine(ctx, t->engine->legacy_idx); GEM_BUG_ON(IS_ERR(ce)); rq = t->request_alloc(ce); intel_context_put(ce); if (IS_ERR(rq)) { - mutex_unlock(BKL); err = PTR_ERR(rq); count = n; break; @@ -357,8 +347,6 @@ static int __igt_breadcrumbs_smoketest(void *arg) requests[n] = i915_request_get(rq); i915_request_add(rq); - mutex_unlock(BKL); - if (err >= 0) err = i915_sw_fence_await_dma_fence(wait, &rq->fence, @@ -446,18 +434,16 @@ static int mock_breadcrumbs_smoketest(void *arg) * See __igt_breadcrumbs_smoketest(); */ - threads = kmalloc_array(ncpus, sizeof(*threads), GFP_KERNEL); + threads = kcalloc(ncpus, sizeof(*threads), GFP_KERNEL); if (!threads) return -ENOMEM; - t.contexts = - kmalloc_array(t.ncontexts, sizeof(*t.contexts), GFP_KERNEL); + t.contexts = kcalloc(t.ncontexts, sizeof(*t.contexts), GFP_KERNEL); if (!t.contexts) { ret = -ENOMEM; goto out_threads; } - mutex_lock(&t.engine->i915->drm.struct_mutex); for (n = 0; n < t.ncontexts; n++) { t.contexts[n] = mock_context(t.engine->i915, "mock"); if (!t.contexts[n]) { @@ -465,7 +451,6 @@ static int mock_breadcrumbs_smoketest(void *arg) goto out_contexts; } } - mutex_unlock(&t.engine->i915->drm.struct_mutex); for (n = 0; n < ncpus; n++) { threads[n] = kthread_run(__igt_breadcrumbs_smoketest, @@ -479,6 +464,7 @@ static int mock_breadcrumbs_smoketest(void *arg) get_task_struct(threads[n]); } + yield(); /* start all threads before we begin */ msleep(jiffies_to_msecs(i915_selftest.timeout_jiffies)); for (n = 0; n < ncpus; n++) { @@ -495,18 +481,15 @@ static int mock_breadcrumbs_smoketest(void *arg) atomic_long_read(&t.num_fences), ncpus); - mutex_lock(&t.engine->i915->drm.struct_mutex); out_contexts: for (n = 0; n < t.ncontexts; n++) { if (!t.contexts[n]) break; mock_context_close(t.contexts[n]); } - mutex_unlock(&t.engine->i915->drm.struct_mutex); kfree(t.contexts); out_threads: kfree(threads); - return ret; } @@ -539,40 +522,37 @@ static int live_nop_request(void *arg) { struct drm_i915_private *i915 = arg; struct intel_engine_cs *engine; - intel_wakeref_t wakeref; struct igt_live_test t; - unsigned int id; int err = -ENODEV; - /* Submit various sized batches of empty requests, to each engine + /* + * Submit various sized batches of empty requests, to each engine * (individually), and wait for the batch to complete. We can check * the overhead of submitting requests to the hardware. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - for_each_engine(engine, i915, id) { - struct i915_request *request = NULL; + for_each_uabi_engine(engine, i915) { unsigned long n, prime; IGT_TIMEOUT(end_time); ktime_t times[2] = {}; err = igt_live_test_begin(&t, i915, __func__, engine->name); if (err) - goto out_unlock; + return err; for_each_prime_number_from(prime, 1, 8192) { + struct i915_request *request = NULL; + times[1] = ktime_get_raw(); for (n = 0; n < prime; n++) { + i915_request_put(request); request = i915_request_create(engine->kernel_context); - if (IS_ERR(request)) { - err = PTR_ERR(request); - goto out_unlock; - } + if (IS_ERR(request)) + return PTR_ERR(request); - /* This space is left intentionally blank. + /* + * This space is left intentionally blank. * * We do not actually want to perform any * action with this request, we just want @@ -585,9 +565,11 @@ static int live_nop_request(void *arg) * for latency. */ + i915_request_get(request); i915_request_add(request); } i915_request_wait(request, 0, MAX_SCHEDULE_TIMEOUT); + i915_request_put(request); times[1] = ktime_sub(ktime_get_raw(), times[1]); if (prime == 1) @@ -599,7 +581,7 @@ static int live_nop_request(void *arg) err = igt_live_test_end(&t); if (err) - goto out_unlock; + return err; pr_info("Request latencies on %s: 1 = %lluns, %lu = %lluns\n", engine->name, @@ -607,9 +589,6 @@ static int live_nop_request(void *arg) prime, div64_u64(ktime_to_ns(times[1]), prime)); } -out_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -647,8 +626,15 @@ static struct i915_vma *empty_batch(struct drm_i915_private *i915) if (err) goto err; + /* Force the wait wait now to avoid including it in the benchmark */ + err = i915_vma_sync(vma); + if (err) + goto err_pin; + return vma; +err_pin: + i915_vma_unpin(vma); err: i915_gem_object_put(obj); return ERR_PTR(err); @@ -672,6 +658,7 @@ empty_request(struct intel_engine_cs *engine, if (err) goto out_request; + i915_request_get(request); out_request: i915_request_add(request); return err ? ERR_PTR(err) : request; @@ -681,27 +668,21 @@ static int live_empty_request(void *arg) { struct drm_i915_private *i915 = arg; struct intel_engine_cs *engine; - intel_wakeref_t wakeref; struct igt_live_test t; struct i915_vma *batch; - unsigned int id; int err = 0; - /* Submit various sized batches of empty requests, to each engine + /* + * Submit various sized batches of empty requests, to each engine * (individually), and wait for the batch to complete. We can check * the overhead of submitting requests to the hardware. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - batch = empty_batch(i915); - if (IS_ERR(batch)) { - err = PTR_ERR(batch); - goto out_unlock; - } + if (IS_ERR(batch)) + return PTR_ERR(batch); - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { IGT_TIMEOUT(end_time); struct i915_request *request; unsigned long n, prime; @@ -723,6 +704,7 @@ static int live_empty_request(void *arg) times[1] = ktime_get_raw(); for (n = 0; n < prime; n++) { + i915_request_put(request); request = empty_request(engine, batch); if (IS_ERR(request)) { err = PTR_ERR(request); @@ -738,6 +720,7 @@ static int live_empty_request(void *arg) if (__igt_timeout(end_time, NULL)) break; } + i915_request_put(request); err = igt_live_test_end(&t); if (err) @@ -752,18 +735,15 @@ static int live_empty_request(void *arg) out_batch: i915_vma_unpin(batch); i915_vma_put(batch); -out_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } static struct i915_vma *recursive_batch(struct drm_i915_private *i915) { struct i915_gem_context *ctx = i915->kernel_context; - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm; struct drm_i915_gem_object *obj; const int gen = INTEL_GEN(i915); + struct i915_address_space *vm; struct i915_vma *vma; u32 *cmd; int err; @@ -772,7 +752,9 @@ static struct i915_vma *recursive_batch(struct drm_i915_private *i915) if (IS_ERR(obj)) return ERR_CAST(obj); + vm = i915_gem_context_get_vm_rcu(ctx); vma = i915_vma_instance(obj, vm, NULL); + i915_vm_put(vm); if (IS_ERR(vma)) { err = PTR_ERR(vma); goto err; @@ -832,67 +814,73 @@ static int recursive_batch_resolve(struct i915_vma *batch) static int live_all_engines(void *arg) { struct drm_i915_private *i915 = arg; + const unsigned int nengines = num_uabi_engines(i915); struct intel_engine_cs *engine; - struct i915_request *request[I915_NUM_ENGINES]; - intel_wakeref_t wakeref; + struct i915_request **request; struct igt_live_test t; struct i915_vma *batch; - unsigned int id; + unsigned int idx; int err; - /* Check we can submit requests to all engines simultaneously. We + /* + * Check we can submit requests to all engines simultaneously. We * send a recursive batch to each engine - checking that we don't * block doing so, and that they don't complete too soon. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + request = kcalloc(nengines, sizeof(*request), GFP_KERNEL); + if (!request) + return -ENOMEM; err = igt_live_test_begin(&t, i915, __func__, ""); if (err) - goto out_unlock; + goto out_free; batch = recursive_batch(i915); if (IS_ERR(batch)) { err = PTR_ERR(batch); pr_err("%s: Unable to create batch, err=%d\n", __func__, err); - goto out_unlock; + goto out_free; } - for_each_engine(engine, i915, id) { - request[id] = i915_request_create(engine->kernel_context); - if (IS_ERR(request[id])) { - err = PTR_ERR(request[id]); + idx = 0; + for_each_uabi_engine(engine, i915) { + request[idx] = i915_request_create(engine->kernel_context); + if (IS_ERR(request[idx])) { + err = PTR_ERR(request[idx]); pr_err("%s: Request allocation failed with err=%d\n", __func__, err); goto out_request; } - err = engine->emit_bb_start(request[id], + err = engine->emit_bb_start(request[idx], batch->node.start, batch->node.size, 0); GEM_BUG_ON(err); - request[id]->batch = batch; + request[idx]->batch = batch; i915_vma_lock(batch); - err = i915_request_await_object(request[id], batch->obj, 0); + err = i915_request_await_object(request[idx], batch->obj, 0); if (err == 0) - err = i915_vma_move_to_active(batch, request[id], 0); + err = i915_vma_move_to_active(batch, request[idx], 0); i915_vma_unlock(batch); GEM_BUG_ON(err); - i915_request_get(request[id]); - i915_request_add(request[id]); + i915_request_get(request[idx]); + i915_request_add(request[idx]); + idx++; } - for_each_engine(engine, i915, id) { - if (i915_request_completed(request[id])) { + idx = 0; + for_each_uabi_engine(engine, i915) { + if (i915_request_completed(request[idx])) { pr_err("%s(%s): request completed too early!\n", __func__, engine->name); err = -EINVAL; goto out_request; } + idx++; } err = recursive_batch_resolve(batch); @@ -901,10 +889,11 @@ static int live_all_engines(void *arg) goto out_request; } - for_each_engine(engine, i915, id) { + idx = 0; + for_each_uabi_engine(engine, i915) { long timeout; - timeout = i915_request_wait(request[id], 0, + timeout = i915_request_wait(request[idx], 0, MAX_SCHEDULE_TIMEOUT); if (timeout < 0) { err = timeout; @@ -913,50 +902,56 @@ static int live_all_engines(void *arg) goto out_request; } - GEM_BUG_ON(!i915_request_completed(request[id])); - i915_request_put(request[id]); - request[id] = NULL; + GEM_BUG_ON(!i915_request_completed(request[idx])); + i915_request_put(request[idx]); + request[idx] = NULL; + idx++; } err = igt_live_test_end(&t); out_request: - for_each_engine(engine, i915, id) - if (request[id]) - i915_request_put(request[id]); + idx = 0; + for_each_uabi_engine(engine, i915) { + if (request[idx]) + i915_request_put(request[idx]); + idx++; + } i915_vma_unpin(batch); i915_vma_put(batch); -out_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); +out_free: + kfree(request); return err; } static int live_sequential_engines(void *arg) { struct drm_i915_private *i915 = arg; - struct i915_request *request[I915_NUM_ENGINES] = {}; + const unsigned int nengines = num_uabi_engines(i915); + struct i915_request **request; struct i915_request *prev = NULL; struct intel_engine_cs *engine; - intel_wakeref_t wakeref; struct igt_live_test t; - unsigned int id; + unsigned int idx; int err; - /* Check we can submit requests to all engines sequentially, such + /* + * Check we can submit requests to all engines sequentially, such * that each successive request waits for the earlier ones. This * tests that we don't execute requests out of order, even though * they are running on independent engines. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + request = kcalloc(nengines, sizeof(*request), GFP_KERNEL); + if (!request) + return -ENOMEM; err = igt_live_test_begin(&t, i915, __func__, ""); if (err) - goto out_unlock; + goto out_free; - for_each_engine(engine, i915, id) { + idx = 0; + for_each_uabi_engine(engine, i915) { struct i915_vma *batch; batch = recursive_batch(i915); @@ -964,66 +959,69 @@ static int live_sequential_engines(void *arg) err = PTR_ERR(batch); pr_err("%s: Unable to create batch for %s, err=%d\n", __func__, engine->name, err); - goto out_unlock; + goto out_free; } - request[id] = i915_request_create(engine->kernel_context); - if (IS_ERR(request[id])) { - err = PTR_ERR(request[id]); + request[idx] = i915_request_create(engine->kernel_context); + if (IS_ERR(request[idx])) { + err = PTR_ERR(request[idx]); pr_err("%s: Request allocation failed for %s with err=%d\n", __func__, engine->name, err); goto out_request; } if (prev) { - err = i915_request_await_dma_fence(request[id], + err = i915_request_await_dma_fence(request[idx], &prev->fence); if (err) { - i915_request_add(request[id]); + i915_request_add(request[idx]); pr_err("%s: Request await failed for %s with err=%d\n", __func__, engine->name, err); goto out_request; } } - err = engine->emit_bb_start(request[id], + err = engine->emit_bb_start(request[idx], batch->node.start, batch->node.size, 0); GEM_BUG_ON(err); - request[id]->batch = batch; + request[idx]->batch = batch; i915_vma_lock(batch); - err = i915_request_await_object(request[id], batch->obj, false); + err = i915_request_await_object(request[idx], + batch->obj, false); if (err == 0) - err = i915_vma_move_to_active(batch, request[id], 0); + err = i915_vma_move_to_active(batch, request[idx], 0); i915_vma_unlock(batch); GEM_BUG_ON(err); - i915_request_get(request[id]); - i915_request_add(request[id]); + i915_request_get(request[idx]); + i915_request_add(request[idx]); - prev = request[id]; + prev = request[idx]; + idx++; } - for_each_engine(engine, i915, id) { + idx = 0; + for_each_uabi_engine(engine, i915) { long timeout; - if (i915_request_completed(request[id])) { + if (i915_request_completed(request[idx])) { pr_err("%s(%s): request completed too early!\n", __func__, engine->name); err = -EINVAL; goto out_request; } - err = recursive_batch_resolve(request[id]->batch); + err = recursive_batch_resolve(request[idx]->batch); if (err) { pr_err("%s: failed to resolve batch, err=%d\n", __func__, err); goto out_request; } - timeout = i915_request_wait(request[id], 0, + timeout = i915_request_wait(request[idx], 0, MAX_SCHEDULE_TIMEOUT); if (timeout < 0) { err = timeout; @@ -1032,33 +1030,156 @@ static int live_sequential_engines(void *arg) goto out_request; } - GEM_BUG_ON(!i915_request_completed(request[id])); + GEM_BUG_ON(!i915_request_completed(request[idx])); + idx++; } err = igt_live_test_end(&t); out_request: - for_each_engine(engine, i915, id) { + idx = 0; + for_each_uabi_engine(engine, i915) { u32 *cmd; - if (!request[id]) + if (!request[idx]) break; - cmd = i915_gem_object_pin_map(request[id]->batch->obj, + cmd = i915_gem_object_pin_map(request[idx]->batch->obj, I915_MAP_WC); if (!IS_ERR(cmd)) { *cmd = MI_BATCH_BUFFER_END; intel_gt_chipset_flush(engine->gt); - i915_gem_object_unpin_map(request[id]->batch->obj); + i915_gem_object_unpin_map(request[idx]->batch->obj); } - i915_vma_put(request[id]->batch); - i915_request_put(request[id]); + i915_vma_put(request[idx]->batch); + i915_request_put(request[idx]); + idx++; } -out_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); +out_free: + kfree(request); + return err; +} + +static int __live_parallel_engine1(void *arg) +{ + struct intel_engine_cs *engine = arg; + IGT_TIMEOUT(end_time); + unsigned long count; + + count = 0; + do { + struct i915_request *rq; + int err; + + rq = i915_request_create(engine->kernel_context); + if (IS_ERR(rq)) + return PTR_ERR(rq); + + i915_request_get(rq); + i915_request_add(rq); + + err = 0; + if (i915_request_wait(rq, 0, HZ / 5) < 0) + err = -ETIME; + i915_request_put(rq); + if (err) + return err; + + count++; + } while (!__igt_timeout(end_time, NULL)); + + pr_info("%s: %lu request + sync\n", engine->name, count); + return 0; +} + +static int __live_parallel_engineN(void *arg) +{ + struct intel_engine_cs *engine = arg; + IGT_TIMEOUT(end_time); + unsigned long count; + + count = 0; + do { + struct i915_request *rq; + + rq = i915_request_create(engine->kernel_context); + if (IS_ERR(rq)) + return PTR_ERR(rq); + + i915_request_add(rq); + count++; + } while (!__igt_timeout(end_time, NULL)); + + pr_info("%s: %lu requests\n", engine->name, count); + return 0; +} + +static int live_parallel_engines(void *arg) +{ + struct drm_i915_private *i915 = arg; + static int (* const func[])(void *arg) = { + __live_parallel_engine1, + __live_parallel_engineN, + NULL, + }; + const unsigned int nengines = num_uabi_engines(i915); + struct intel_engine_cs *engine; + int (* const *fn)(void *arg); + struct task_struct **tsk; + int err = 0; + + /* + * Check we can submit requests to all engines concurrently. This + * tests that we load up the system maximally. + */ + + tsk = kcalloc(nengines, sizeof(*tsk), GFP_KERNEL); + if (!tsk) + return -ENOMEM; + + for (fn = func; !err && *fn; fn++) { + struct igt_live_test t; + unsigned int idx; + + err = igt_live_test_begin(&t, i915, __func__, ""); + if (err) + break; + + idx = 0; + for_each_uabi_engine(engine, i915) { + tsk[idx] = kthread_run(*fn, engine, + "igt/parallel:%s", + engine->name); + if (IS_ERR(tsk[idx])) { + err = PTR_ERR(tsk[idx]); + break; + } + get_task_struct(tsk[idx++]); + } + + yield(); /* start all threads before we kthread_stop() */ + + idx = 0; + for_each_uabi_engine(engine, i915) { + int status; + + if (IS_ERR(tsk[idx])) + break; + + status = kthread_stop(tsk[idx]); + if (status && !err) + err = status; + + put_task_struct(tsk[idx++]); + } + + if (igt_live_test_end(&t)) + err = -EIO; + } + + kfree(tsk); return err; } @@ -1102,16 +1223,16 @@ max_batches(struct i915_gem_context *ctx, struct intel_engine_cs *engine) static int live_breadcrumbs_smoketest(void *arg) { struct drm_i915_private *i915 = arg; - struct smoketest t[I915_NUM_ENGINES]; - unsigned int ncpus = num_online_cpus(); + const unsigned int nengines = num_uabi_engines(i915); + const unsigned int ncpus = num_online_cpus(); unsigned long num_waits, num_fences; struct intel_engine_cs *engine; struct task_struct **threads; struct igt_live_test live; - enum intel_engine_id id; intel_wakeref_t wakeref; struct drm_file *file; - unsigned int n; + struct smoketest *smoke; + unsigned int n, idx; int ret = 0; /* @@ -1130,29 +1251,31 @@ static int live_breadcrumbs_smoketest(void *arg) goto out_rpm; } - threads = kcalloc(ncpus * I915_NUM_ENGINES, - sizeof(*threads), - GFP_KERNEL); - if (!threads) { + smoke = kcalloc(nengines, sizeof(*smoke), GFP_KERNEL); + if (!smoke) { ret = -ENOMEM; goto out_file; } - memset(&t[0], 0, sizeof(t[0])); - t[0].request_alloc = __live_request_alloc; - t[0].ncontexts = 64; - t[0].contexts = kmalloc_array(t[0].ncontexts, - sizeof(*t[0].contexts), - GFP_KERNEL); - if (!t[0].contexts) { + threads = kcalloc(ncpus * nengines, sizeof(*threads), GFP_KERNEL); + if (!threads) { + ret = -ENOMEM; + goto out_smoke; + } + + smoke[0].request_alloc = __live_request_alloc; + smoke[0].ncontexts = 64; + smoke[0].contexts = kcalloc(smoke[0].ncontexts, + sizeof(*smoke[0].contexts), + GFP_KERNEL); + if (!smoke[0].contexts) { ret = -ENOMEM; goto out_threads; } - mutex_lock(&i915->drm.struct_mutex); - for (n = 0; n < t[0].ncontexts; n++) { - t[0].contexts[n] = live_context(i915, file); - if (!t[0].contexts[n]) { + for (n = 0; n < smoke[0].ncontexts; n++) { + smoke[0].contexts[n] = live_context(i915, file); + if (!smoke[0].contexts[n]) { ret = -ENOMEM; goto out_contexts; } @@ -1162,45 +1285,48 @@ static int live_breadcrumbs_smoketest(void *arg) if (ret) goto out_contexts; - for_each_engine(engine, i915, id) { - t[id] = t[0]; - t[id].engine = engine; - t[id].max_batch = max_batches(t[0].contexts[0], engine); - if (t[id].max_batch < 0) { - ret = t[id].max_batch; - mutex_unlock(&i915->drm.struct_mutex); + idx = 0; + for_each_uabi_engine(engine, i915) { + smoke[idx] = smoke[0]; + smoke[idx].engine = engine; + smoke[idx].max_batch = + max_batches(smoke[0].contexts[0], engine); + if (smoke[idx].max_batch < 0) { + ret = smoke[idx].max_batch; goto out_flush; } /* One ring interleaved between requests from all cpus */ - t[id].max_batch /= num_online_cpus() + 1; + smoke[idx].max_batch /= num_online_cpus() + 1; pr_debug("Limiting batches to %d requests on %s\n", - t[id].max_batch, engine->name); + smoke[idx].max_batch, engine->name); for (n = 0; n < ncpus; n++) { struct task_struct *tsk; tsk = kthread_run(__igt_breadcrumbs_smoketest, - &t[id], "igt/%d.%d", id, n); + &smoke[idx], "igt/%d.%d", idx, n); if (IS_ERR(tsk)) { ret = PTR_ERR(tsk); - mutex_unlock(&i915->drm.struct_mutex); goto out_flush; } get_task_struct(tsk); - threads[id * ncpus + n] = tsk; + threads[idx * ncpus + n] = tsk; } + + idx++; } - mutex_unlock(&i915->drm.struct_mutex); + yield(); /* start all threads before we begin */ msleep(jiffies_to_msecs(i915_selftest.timeout_jiffies)); out_flush: + idx = 0; num_waits = 0; num_fences = 0; - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { for (n = 0; n < ncpus; n++) { - struct task_struct *tsk = threads[id * ncpus + n]; + struct task_struct *tsk = threads[idx * ncpus + n]; int err; if (!tsk) @@ -1213,19 +1339,20 @@ out_flush: put_task_struct(tsk); } - num_waits += atomic_long_read(&t[id].num_waits); - num_fences += atomic_long_read(&t[id].num_fences); + num_waits += atomic_long_read(&smoke[idx].num_waits); + num_fences += atomic_long_read(&smoke[idx].num_fences); + idx++; } pr_info("Completed %lu waits for %lu fences across %d engines and %d cpus\n", num_waits, num_fences, RUNTIME_INFO(i915)->num_engines, ncpus); - mutex_lock(&i915->drm.struct_mutex); ret = igt_live_test_end(&live) ?: ret; out_contexts: - mutex_unlock(&i915->drm.struct_mutex); - kfree(t[0].contexts); + kfree(smoke[0].contexts); out_threads: kfree(threads); +out_smoke: + kfree(smoke); out_file: mock_file_free(i915, file); out_rpm: @@ -1240,6 +1367,7 @@ int i915_request_live_selftests(struct drm_i915_private *i915) SUBTEST(live_nop_request), SUBTEST(live_all_engines), SUBTEST(live_sequential_engines), + SUBTEST(live_parallel_engines), SUBTEST(live_empty_request), SUBTEST(live_breadcrumbs_smoketest), }; diff --git a/drivers/gpu/drm/i915/selftests/i915_selftest.c b/drivers/gpu/drm/i915/selftests/i915_selftest.c index 438ea0eaa416..a6cca4ad96f6 100644 --- a/drivers/gpu/drm/i915/selftests/i915_selftest.c +++ b/drivers/gpu/drm/i915/selftests/i915_selftest.c @@ -23,13 +23,14 @@ #include <linux/random.h> -#include "../i915_drv.h" -#include "../i915_selftest.h" +#include "gt/intel_gt_pm.h" +#include "i915_drv.h" +#include "i915_selftest.h" #include "igt_flush_test.h" struct i915_selftest i915_selftest __read_mostly = { - .timeout_ms = 1000, + .timeout_ms = 500, }; int i915_mock_sanitycheck(void) @@ -256,6 +257,10 @@ int __i915_live_setup(void *data) { struct drm_i915_private *i915 = data; + /* The selftests expect an idle system */ + if (intel_gt_pm_wait_for_idle(&i915->gt)) + return -EIO; + return intel_gt_terminally_wedged(&i915->gt); } @@ -263,10 +268,8 @@ int __i915_live_teardown(int err, void *data) { struct drm_i915_private *i915 = data; - mutex_lock(&i915->drm.struct_mutex); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; - mutex_unlock(&i915->drm.struct_mutex); i915_gem_drain_freed_objects(i915); @@ -277,6 +280,10 @@ int __intel_gt_live_setup(void *data) { struct intel_gt *gt = data; + /* The selftests expect an idle system */ + if (intel_gt_pm_wait_for_idle(gt)) + return -EIO; + return intel_gt_terminally_wedged(gt); } @@ -284,10 +291,8 @@ int __intel_gt_live_teardown(int err, void *data) { struct intel_gt *gt = data; - mutex_lock(>->i915->drm.struct_mutex); - if (igt_flush_test(gt->i915, I915_WAIT_LOCKED)) + if (igt_flush_test(gt->i915)) err = -EIO; - mutex_unlock(>->i915->drm.struct_mutex); i915_gem_drain_freed_objects(gt->i915); diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c index a5bec0a4cdcc..58b5f40a07dd 100644 --- a/drivers/gpu/drm/i915/selftests/i915_vma.c +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c @@ -24,6 +24,7 @@ #include <linux/prime_numbers.h> +#include "gem/i915_gem_context.h" #include "gem/selftests/mock_context.h" #include "i915_scatterlist.h" @@ -38,7 +39,7 @@ static bool assert_vma(struct i915_vma *vma, { bool ok = true; - if (vma->vm != ctx->vm) { + if (vma->vm != rcu_access_pointer(ctx->vm)) { pr_err("VMA created with wrong VM\n"); ok = false; } @@ -113,11 +114,13 @@ static int create_vmas(struct drm_i915_private *i915, list_for_each_entry(obj, objects, st_link) { for (pinned = 0; pinned <= 1; pinned++) { list_for_each_entry(ctx, contexts, link) { - struct i915_address_space *vm = ctx->vm; + struct i915_address_space *vm; struct i915_vma *vma; int err; + vm = i915_gem_context_get_vm_rcu(ctx); vma = checked_vma_instance(obj, vm, NULL); + i915_vm_put(vm); if (IS_ERR(vma)) return PTR_ERR(vma); @@ -170,7 +173,7 @@ static int igt_vma_create(void *arg) } nc = 0; - for_each_prime_number(num_ctx, MAX_CONTEXT_HW_ID) { + for_each_prime_number(num_ctx, 2 * NUM_CONTEXT_TAG) { for (; nc < num_ctx; nc++) { ctx = mock_context(i915, "mock"); if (!ctx) @@ -623,7 +626,7 @@ static bool assert_partial(struct drm_i915_gem_object *obj, struct sgt_iter sgt; dma_addr_t dma; - for_each_sgt_dma(dma, sgt, vma->pages) { + for_each_sgt_daddr(dma, sgt, vma->pages) { dma_addr_t src; if (!size) { @@ -831,13 +834,10 @@ int i915_vma_mock_selftests(void) } mock_init_ggtt(i915, ggtt); - mutex_lock(&i915->drm.struct_mutex); err = i915_subtests(tests, ggtt); - mock_device_flush(i915); - mutex_unlock(&i915->drm.struct_mutex); + mock_device_flush(i915); i915_gem_drain_freed_objects(i915); - mock_fini_ggtt(ggtt); kfree(ggtt); out_put: @@ -879,8 +879,6 @@ static int igt_vma_remapped_gtt(void *arg) if (IS_ERR(obj)) return PTR_ERR(obj); - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); for (t = types; *t; t++) { @@ -976,7 +974,6 @@ static int igt_vma_remapped_gtt(void *arg) out: intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); i915_gem_object_put(obj); return err; diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.c b/drivers/gpu/drm/i915/selftests/igt_flush_test.c index d3b5eb402d33..7b0939e3f007 100644 --- a/drivers/gpu/drm/i915/selftests/igt_flush_test.c +++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.c @@ -4,39 +4,32 @@ * Copyright © 2018 Intel Corporation */ -#include "gem/i915_gem_context.h" #include "gt/intel_gt.h" +#include "gt/intel_gt_requests.h" #include "i915_drv.h" #include "i915_selftest.h" #include "igt_flush_test.h" -int igt_flush_test(struct drm_i915_private *i915, unsigned int flags) +int igt_flush_test(struct drm_i915_private *i915) { - int ret = intel_gt_is_wedged(&i915->gt) ? -EIO : 0; - int repeat = !!(flags & I915_WAIT_LOCKED); + struct intel_gt *gt = &i915->gt; + int ret = intel_gt_is_wedged(gt) ? -EIO : 0; cond_resched(); - do { - if (i915_gem_wait_for_idle(i915, flags, HZ / 5) == -ETIME) { - pr_err("%pS timed out, cancelling all further testing.\n", - __builtin_return_address(0)); + if (intel_gt_wait_for_idle(gt, HZ / 5) == -ETIME) { + pr_err("%pS timed out, cancelling all further testing.\n", + __builtin_return_address(0)); - GEM_TRACE("%pS timed out.\n", - __builtin_return_address(0)); - GEM_TRACE_DUMP(); + GEM_TRACE("%pS timed out.\n", + __builtin_return_address(0)); + GEM_TRACE_DUMP(); - intel_gt_set_wedged(&i915->gt); - repeat = 0; - ret = -EIO; - } - - /* Ensure we also flush after wedging. */ - if (flags & I915_WAIT_LOCKED) - i915_retire_requests(i915); - } while (repeat--); + intel_gt_set_wedged(gt); + ret = -EIO; + } return ret; } diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.h b/drivers/gpu/drm/i915/selftests/igt_flush_test.h index 63e009927c43..7541fa74e641 100644 --- a/drivers/gpu/drm/i915/selftests/igt_flush_test.h +++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.h @@ -9,6 +9,6 @@ struct drm_i915_private; -int igt_flush_test(struct drm_i915_private *i915, unsigned int flags); +int igt_flush_test(struct drm_i915_private *i915); #endif /* IGT_FLUSH_TEST_H */ diff --git a/drivers/gpu/drm/i915/selftests/igt_live_test.c b/drivers/gpu/drm/i915/selftests/igt_live_test.c index 3e902761cd16..c130010a7033 100644 --- a/drivers/gpu/drm/i915/selftests/igt_live_test.c +++ b/drivers/gpu/drm/i915/selftests/igt_live_test.c @@ -4,7 +4,8 @@ * Copyright © 2018 Intel Corporation */ -#include "../i915_drv.h" +#include "i915_drv.h" +#include "gt/intel_gt_requests.h" #include "../i915_selftest.h" #include "igt_flush_test.h" @@ -15,20 +16,16 @@ int igt_live_test_begin(struct igt_live_test *t, const char *func, const char *name) { + struct intel_gt *gt = &i915->gt; struct intel_engine_cs *engine; enum intel_engine_id id; int err; - lockdep_assert_held(&i915->drm.struct_mutex); - t->i915 = i915; t->func = func; t->name = name; - err = i915_gem_wait_for_idle(i915, - I915_WAIT_INTERRUPTIBLE | - I915_WAIT_LOCKED, - MAX_SCHEDULE_TIMEOUT); + err = intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT); if (err) { pr_err("%s(%s): failed to idle before, with err=%d!", func, name, err); @@ -37,7 +34,7 @@ int igt_live_test_begin(struct igt_live_test *t, t->reset_global = i915_reset_count(&i915->gpu_error); - for_each_engine(engine, i915, id) + for_each_engine(engine, gt, id) t->reset_engine[id] = i915_reset_engine_count(&i915->gpu_error, engine); @@ -50,9 +47,7 @@ int igt_live_test_end(struct igt_live_test *t) struct intel_engine_cs *engine; enum intel_engine_id id; - lockdep_assert_held(&i915->drm.struct_mutex); - - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) return -EIO; if (t->reset_global != i915_reset_count(&i915->gpu_error)) { @@ -62,7 +57,7 @@ int igt_live_test_end(struct igt_live_test *t) return -EIO; } - for_each_engine(engine, i915, id) { + for_each_engine(engine, &i915->gt, id) { if (t->reset_engine[id] == i915_reset_engine_count(&i915->gpu_error, engine)) continue; diff --git a/drivers/gpu/drm/i915/selftests/igt_reset.c b/drivers/gpu/drm/i915/selftests/igt_reset.c index 7ec8f8b049c6..9f8590b868a9 100644 --- a/drivers/gpu/drm/i915/selftests/igt_reset.c +++ b/drivers/gpu/drm/i915/selftests/igt_reset.c @@ -22,7 +22,7 @@ void igt_global_reset_lock(struct intel_gt *gt) wait_event(gt->reset.queue, !test_bit(I915_RESET_BACKOFF, >->reset.flags)); - for_each_engine(engine, gt->i915, id) { + for_each_engine(engine, gt, id) { while (test_and_set_bit(I915_RESET_ENGINE + id, >->reset.flags)) wait_on_bit(>->reset.flags, I915_RESET_ENGINE + id, @@ -35,7 +35,7 @@ void igt_global_reset_unlock(struct intel_gt *gt) struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, gt->i915, id) + for_each_engine(engine, gt, id) clear_bit(I915_RESET_ENGINE + id, >->reset.flags); clear_bit(I915_RESET_BACKOFF, >->reset.flags); diff --git a/drivers/gpu/drm/i915/selftests/igt_spinner.c b/drivers/gpu/drm/i915/selftests/igt_spinner.c index 11f04ad48e68..ee8450b871da 100644 --- a/drivers/gpu/drm/i915/selftests/igt_spinner.c +++ b/drivers/gpu/drm/i915/selftests/igt_spinner.c @@ -147,7 +147,7 @@ igt_spinner_create_request(struct igt_spinner *spin, intel_gt_chipset_flush(engine->gt); if (engine->emit_init_breadcrumb && - rq->timeline->has_initial_breadcrumb) { + i915_request_timeline(rq)->has_initial_breadcrumb) { err = engine->emit_init_breadcrumb(rq); if (err) goto cancel_rq; diff --git a/drivers/gpu/drm/i915/selftests/intel_memory_region.c b/drivers/gpu/drm/i915/selftests/intel_memory_region.c new file mode 100644 index 000000000000..19e1cca8f143 --- /dev/null +++ b/drivers/gpu/drm/i915/selftests/intel_memory_region.c @@ -0,0 +1,624 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include <linux/prime_numbers.h> + +#include "../i915_selftest.h" + +#include "mock_drm.h" +#include "mock_gem_device.h" +#include "mock_region.h" + +#include "gem/i915_gem_context.h" +#include "gem/i915_gem_lmem.h" +#include "gem/i915_gem_region.h" +#include "gem/i915_gem_object_blt.h" +#include "gem/selftests/igt_gem_utils.h" +#include "gem/selftests/mock_context.h" +#include "gt/intel_engine_user.h" +#include "gt/intel_gt.h" +#include "selftests/igt_flush_test.h" +#include "selftests/i915_random.h" + +static void close_objects(struct intel_memory_region *mem, + struct list_head *objects) +{ + struct drm_i915_private *i915 = mem->i915; + struct drm_i915_gem_object *obj, *on; + + list_for_each_entry_safe(obj, on, objects, st_link) { + if (i915_gem_object_has_pinned_pages(obj)) + i915_gem_object_unpin_pages(obj); + /* No polluting the memory region between tests */ + __i915_gem_object_put_pages(obj, I915_MM_NORMAL); + list_del(&obj->st_link); + i915_gem_object_put(obj); + } + + cond_resched(); + + i915_gem_drain_freed_objects(i915); +} + +static int igt_mock_fill(void *arg) +{ + struct intel_memory_region *mem = arg; + resource_size_t total = resource_size(&mem->region); + resource_size_t page_size; + resource_size_t rem; + unsigned long max_pages; + unsigned long page_num; + LIST_HEAD(objects); + int err = 0; + + page_size = mem->mm.chunk_size; + max_pages = div64_u64(total, page_size); + rem = total; + + for_each_prime_number_from(page_num, 1, max_pages) { + resource_size_t size = page_num * page_size; + struct drm_i915_gem_object *obj; + + obj = i915_gem_object_create_region(mem, size, 0); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + break; + } + + err = i915_gem_object_pin_pages(obj); + if (err) { + i915_gem_object_put(obj); + break; + } + + list_add(&obj->st_link, &objects); + rem -= size; + } + + if (err == -ENOMEM) + err = 0; + if (err == -ENXIO) { + if (page_num * page_size <= rem) { + pr_err("%s failed, space still left in region\n", + __func__); + err = -EINVAL; + } else { + err = 0; + } + } + + close_objects(mem, &objects); + + return err; +} + +static struct drm_i915_gem_object * +igt_object_create(struct intel_memory_region *mem, + struct list_head *objects, + u64 size, + unsigned int flags) +{ + struct drm_i915_gem_object *obj; + int err; + + obj = i915_gem_object_create_region(mem, size, flags); + if (IS_ERR(obj)) + return obj; + + err = i915_gem_object_pin_pages(obj); + if (err) + goto put; + + list_add(&obj->st_link, objects); + return obj; + +put: + i915_gem_object_put(obj); + return ERR_PTR(err); +} + +static void igt_object_release(struct drm_i915_gem_object *obj) +{ + i915_gem_object_unpin_pages(obj); + __i915_gem_object_put_pages(obj, I915_MM_NORMAL); + list_del(&obj->st_link); + i915_gem_object_put(obj); +} + +static int igt_mock_contiguous(void *arg) +{ + struct intel_memory_region *mem = arg; + struct drm_i915_gem_object *obj; + unsigned long n_objects; + LIST_HEAD(objects); + LIST_HEAD(holes); + I915_RND_STATE(prng); + resource_size_t total; + resource_size_t min; + u64 target; + int err = 0; + + total = resource_size(&mem->region); + + /* Min size */ + obj = igt_object_create(mem, &objects, mem->mm.chunk_size, + I915_BO_ALLOC_CONTIGUOUS); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + if (obj->mm.pages->nents != 1) { + pr_err("%s min object spans multiple sg entries\n", __func__); + err = -EINVAL; + goto err_close_objects; + } + + igt_object_release(obj); + + /* Max size */ + obj = igt_object_create(mem, &objects, total, I915_BO_ALLOC_CONTIGUOUS); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + if (obj->mm.pages->nents != 1) { + pr_err("%s max object spans multiple sg entries\n", __func__); + err = -EINVAL; + goto err_close_objects; + } + + igt_object_release(obj); + + /* Internal fragmentation should not bleed into the object size */ + target = i915_prandom_u64_state(&prng); + div64_u64_rem(target, total, &target); + target = round_up(target, PAGE_SIZE); + target = max_t(u64, PAGE_SIZE, target); + + obj = igt_object_create(mem, &objects, target, + I915_BO_ALLOC_CONTIGUOUS); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + if (obj->base.size != target) { + pr_err("%s obj->base.size(%zx) != target(%llx)\n", __func__, + obj->base.size, target); + err = -EINVAL; + goto err_close_objects; + } + + if (obj->mm.pages->nents != 1) { + pr_err("%s object spans multiple sg entries\n", __func__); + err = -EINVAL; + goto err_close_objects; + } + + igt_object_release(obj); + + /* + * Try to fragment the address space, such that half of it is free, but + * the max contiguous block size is SZ_64K. + */ + + target = SZ_64K; + n_objects = div64_u64(total, target); + + while (n_objects--) { + struct list_head *list; + + if (n_objects % 2) + list = &holes; + else + list = &objects; + + obj = igt_object_create(mem, list, target, + I915_BO_ALLOC_CONTIGUOUS); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + goto err_close_objects; + } + } + + close_objects(mem, &holes); + + min = target; + target = total >> 1; + + /* Make sure we can still allocate all the fragmented space */ + obj = igt_object_create(mem, &objects, target, 0); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + goto err_close_objects; + } + + igt_object_release(obj); + + /* + * Even though we have enough free space, we don't have a big enough + * contiguous block. Make sure that holds true. + */ + + do { + bool should_fail = target > min; + + obj = igt_object_create(mem, &objects, target, + I915_BO_ALLOC_CONTIGUOUS); + if (should_fail != IS_ERR(obj)) { + pr_err("%s target allocation(%llx) mismatch\n", + __func__, target); + err = -EINVAL; + goto err_close_objects; + } + + target >>= 1; + } while (target >= mem->mm.chunk_size); + +err_close_objects: + list_splice_tail(&holes, &objects); + close_objects(mem, &objects); + return err; +} + +static int igt_gpu_write_dw(struct intel_context *ce, + struct i915_vma *vma, + u32 dword, + u32 value) +{ + return igt_gpu_fill_dw(ce, vma, dword * sizeof(u32), + vma->size >> PAGE_SHIFT, value); +} + +static int igt_cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val) +{ + unsigned long n; + int err; + + i915_gem_object_lock(obj); + err = i915_gem_object_set_to_wc_domain(obj, false); + i915_gem_object_unlock(obj); + if (err) + return err; + + err = i915_gem_object_pin_pages(obj); + if (err) + return err; + + for (n = 0; n < obj->base.size >> PAGE_SHIFT; ++n) { + u32 __iomem *base; + u32 read_val; + + base = i915_gem_object_lmem_io_map_page_atomic(obj, n); + + read_val = ioread32(base + dword); + io_mapping_unmap_atomic(base); + if (read_val != val) { + pr_err("n=%lu base[%u]=%u, val=%u\n", + n, dword, read_val, val); + err = -EINVAL; + break; + } + } + + i915_gem_object_unpin_pages(obj); + return err; +} + +static int igt_gpu_write(struct i915_gem_context *ctx, + struct drm_i915_gem_object *obj) +{ + struct i915_gem_engines *engines; + struct i915_gem_engines_iter it; + struct i915_address_space *vm; + struct intel_context *ce; + I915_RND_STATE(prng); + IGT_TIMEOUT(end_time); + unsigned int count; + struct i915_vma *vma; + int *order; + int i, n; + int err = 0; + + GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj)); + + n = 0; + count = 0; + for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) { + count++; + if (!intel_engine_can_store_dword(ce->engine)) + continue; + + vm = ce->vm; + n++; + } + i915_gem_context_unlock_engines(ctx); + if (!n) + return 0; + + order = i915_random_order(count * count, &prng); + if (!order) + return -ENOMEM; + + vma = i915_vma_instance(obj, vm, NULL); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto out_free; + } + + err = i915_vma_pin(vma, 0, 0, PIN_USER); + if (err) + goto out_free; + + i = 0; + engines = i915_gem_context_lock_engines(ctx); + do { + u32 rng = prandom_u32_state(&prng); + u32 dword = offset_in_page(rng) / 4; + + ce = engines->engines[order[i] % engines->num_engines]; + i = (i + 1) % (count * count); + if (!ce || !intel_engine_can_store_dword(ce->engine)) + continue; + + err = igt_gpu_write_dw(ce, vma, dword, rng); + if (err) + break; + + err = igt_cpu_check(obj, dword, rng); + if (err) + break; + } while (!__igt_timeout(end_time, NULL)); + i915_gem_context_unlock_engines(ctx); + +out_free: + kfree(order); + + if (err == -ENOMEM) + err = 0; + + return err; +} + +static int igt_lmem_create(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct drm_i915_gem_object *obj; + int err = 0; + + obj = i915_gem_object_create_lmem(i915, PAGE_SIZE, 0); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + err = i915_gem_object_pin_pages(obj); + if (err) + goto out_put; + + i915_gem_object_unpin_pages(obj); +out_put: + i915_gem_object_put(obj); + + return err; +} + +static int igt_lmem_write_gpu(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct drm_i915_gem_object *obj; + struct i915_gem_context *ctx; + struct drm_file *file; + I915_RND_STATE(prng); + u32 sz; + int err; + + file = mock_file(i915); + if (IS_ERR(file)) + return PTR_ERR(file); + + ctx = live_context(i915, file); + if (IS_ERR(ctx)) { + err = PTR_ERR(ctx); + goto out_file; + } + + sz = round_up(prandom_u32_state(&prng) % SZ_32M, PAGE_SIZE); + + obj = i915_gem_object_create_lmem(i915, sz, 0); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + goto out_file; + } + + err = i915_gem_object_pin_pages(obj); + if (err) + goto out_put; + + err = igt_gpu_write(ctx, obj); + if (err) + pr_err("igt_gpu_write failed(%d)\n", err); + + i915_gem_object_unpin_pages(obj); +out_put: + i915_gem_object_put(obj); +out_file: + mock_file_free(i915, file); + return err; +} + +static struct intel_engine_cs * +random_engine_class(struct drm_i915_private *i915, + unsigned int class, + struct rnd_state *prng) +{ + struct intel_engine_cs *engine; + unsigned int count; + + count = 0; + for (engine = intel_engine_lookup_user(i915, class, 0); + engine && engine->uabi_class == class; + engine = rb_entry_safe(rb_next(&engine->uabi_node), + typeof(*engine), uabi_node)) + count++; + + count = i915_prandom_u32_max_state(count, prng); + return intel_engine_lookup_user(i915, class, count); +} + +static int igt_lmem_write_cpu(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct drm_i915_gem_object *obj; + I915_RND_STATE(prng); + IGT_TIMEOUT(end_time); + u32 bytes[] = { + 0, /* rng placeholder */ + sizeof(u32), + sizeof(u64), + 64, /* cl */ + PAGE_SIZE, + PAGE_SIZE - sizeof(u32), + PAGE_SIZE - sizeof(u64), + PAGE_SIZE - 64, + }; + struct intel_engine_cs *engine; + u32 *vaddr; + u32 sz; + u32 i; + int *order; + int count; + int err; + + engine = random_engine_class(i915, I915_ENGINE_CLASS_COPY, &prng); + if (!engine) + return 0; + + pr_info("%s: using %s\n", __func__, engine->name); + + sz = round_up(prandom_u32_state(&prng) % SZ_32M, PAGE_SIZE); + sz = max_t(u32, 2 * PAGE_SIZE, sz); + + obj = i915_gem_object_create_lmem(i915, sz, I915_BO_ALLOC_CONTIGUOUS); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + vaddr = i915_gem_object_pin_map(obj, I915_MAP_WC); + if (IS_ERR(vaddr)) { + err = PTR_ERR(vaddr); + goto out_put; + } + + /* Put the pages into a known state -- from the gpu for added fun */ + err = i915_gem_object_fill_blt(obj, engine->kernel_context, 0xdeadbeaf); + if (err) + goto out_unpin; + + i915_gem_object_lock(obj); + err = i915_gem_object_set_to_wc_domain(obj, true); + i915_gem_object_unlock(obj); + if (err) + goto out_unpin; + + count = ARRAY_SIZE(bytes); + order = i915_random_order(count * count, &prng); + if (!order) { + err = -ENOMEM; + goto out_unpin; + } + + /* We want to throw in a random width/align */ + bytes[0] = igt_random_offset(&prng, 0, PAGE_SIZE, sizeof(u32), + sizeof(u32)); + + i = 0; + do { + u32 offset; + u32 align; + u32 dword; + u32 size; + u32 val; + + size = bytes[order[i] % count]; + i = (i + 1) % (count * count); + + align = bytes[order[i] % count]; + i = (i + 1) % (count * count); + + align = max_t(u32, sizeof(u32), rounddown_pow_of_two(align)); + + offset = igt_random_offset(&prng, 0, obj->base.size, + size, align); + + val = prandom_u32_state(&prng); + memset32(vaddr + offset / sizeof(u32), val ^ 0xdeadbeaf, + size / sizeof(u32)); + + /* + * Sample random dw -- don't waste precious time reading every + * single dw. + */ + dword = igt_random_offset(&prng, offset, + offset + size, + sizeof(u32), sizeof(u32)); + dword /= sizeof(u32); + if (vaddr[dword] != (val ^ 0xdeadbeaf)) { + pr_err("%s vaddr[%u]=%u, val=%u, size=%u, align=%u, offset=%u\n", + __func__, dword, vaddr[dword], val ^ 0xdeadbeaf, + size, align, offset); + err = -EINVAL; + break; + } + } while (!__igt_timeout(end_time, NULL)); + +out_unpin: + i915_gem_object_unpin_map(obj); +out_put: + i915_gem_object_put(obj); + + return err; +} + +int intel_memory_region_mock_selftests(void) +{ + static const struct i915_subtest tests[] = { + SUBTEST(igt_mock_fill), + SUBTEST(igt_mock_contiguous), + }; + struct intel_memory_region *mem; + struct drm_i915_private *i915; + int err; + + i915 = mock_gem_device(); + if (!i915) + return -ENOMEM; + + mem = mock_region_create(i915, 0, SZ_2G, I915_GTT_PAGE_SIZE_4K, 0); + if (IS_ERR(mem)) { + pr_err("failed to create memory region\n"); + err = PTR_ERR(mem); + goto out_unref; + } + + err = i915_subtests(tests, mem); + + intel_memory_region_put(mem); +out_unref: + drm_dev_put(&i915->drm); + return err; +} + +int intel_memory_region_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(igt_lmem_create), + SUBTEST(igt_lmem_write_cpu), + SUBTEST(igt_lmem_write_gpu), + }; + + if (!HAS_LMEM(i915)) { + pr_info("device lacks LMEM support, skipping\n"); + return 0; + } + + if (intel_gt_is_wedged(&i915->gt)) + return 0; + + return i915_live_subtests(tests, i915); +} diff --git a/drivers/gpu/drm/i915/selftests/intel_uncore.c b/drivers/gpu/drm/i915/selftests/intel_uncore.c index 86815c6072a1..0e4e6be0101d 100644 --- a/drivers/gpu/drm/i915/selftests/intel_uncore.c +++ b/drivers/gpu/drm/i915/selftests/intel_uncore.c @@ -67,6 +67,7 @@ static int intel_shadow_table_check(void) } reg_lists[] = { { gen8_shadowed_regs, ARRAY_SIZE(gen8_shadowed_regs) }, { gen11_shadowed_regs, ARRAY_SIZE(gen11_shadowed_regs) }, + { gen12_shadowed_regs, ARRAY_SIZE(gen12_shadowed_regs) }, }; const i915_reg_t *reg; unsigned int i, j; @@ -101,6 +102,7 @@ int intel_uncore_mock_selftests(void) { __chv_fw_ranges, ARRAY_SIZE(__chv_fw_ranges), false }, { __gen9_fw_ranges, ARRAY_SIZE(__gen9_fw_ranges), true }, { __gen11_fw_ranges, ARRAY_SIZE(__gen11_fw_ranges), true }, + { __gen12_fw_ranges, ARRAY_SIZE(__gen12_fw_ranges), true }, }; int err, i; @@ -138,19 +140,19 @@ static int live_forcewake_ops(void *arg) } }; const struct reg *r; - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct intel_uncore_forcewake_domain *domain; - struct intel_uncore *uncore = &i915->uncore; + struct intel_uncore *uncore = gt->uncore; struct intel_engine_cs *engine; enum intel_engine_id id; intel_wakeref_t wakeref; unsigned int tmp; int err = 0; - GEM_BUG_ON(i915->gt.awake); + GEM_BUG_ON(gt->awake); /* vlv/chv with their pcu behave differently wrt reads */ - if (IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915)) { + if (IS_VALLEYVIEW(gt->i915) || IS_CHERRYVIEW(gt->i915)) { pr_debug("PCU fakes forcewake badly; skipping\n"); return 0; } @@ -168,15 +170,15 @@ static int live_forcewake_ops(void *arg) /* We have to pick carefully to get the exact behaviour we need */ for (r = registers; r->name; r++) - if (r->platforms & INTEL_INFO(i915)->gen_mask) + if (r->platforms & INTEL_INFO(gt->i915)->gen_mask) break; if (!r->name) { pr_debug("Forcewaked register not known for %s; skipping\n", - intel_platform_name(INTEL_INFO(i915)->platform)); + intel_platform_name(INTEL_INFO(gt->i915)->platform)); return 0; } - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + wakeref = intel_runtime_pm_get(uncore->rpm); for_each_fw_domain(domain, uncore, tmp) { smp_store_mb(domain->active, false); @@ -186,7 +188,7 @@ static int live_forcewake_ops(void *arg) intel_uncore_fw_release_timer(&domain->timer); } - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt, id) { i915_reg_t mmio = _MMIO(engine->mmio_base + r->offset); u32 __iomem *reg = uncore->regs + engine->mmio_base + r->offset; enum forcewake_domains fw_domains; @@ -247,22 +249,22 @@ static int live_forcewake_ops(void *arg) } out_rpm: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); + intel_runtime_pm_put(uncore->rpm, wakeref); return err; } static int live_forcewake_domains(void *arg) { #define FW_RANGE 0x40000 - struct drm_i915_private *dev_priv = arg; - struct intel_uncore *uncore = &dev_priv->uncore; + struct intel_gt *gt = arg; + struct intel_uncore *uncore = gt->uncore; unsigned long *valid; u32 offset; int err; - if (!HAS_FPGA_DBG_UNCLAIMED(dev_priv) && - !IS_VALLEYVIEW(dev_priv) && - !IS_CHERRYVIEW(dev_priv)) + if (!HAS_FPGA_DBG_UNCLAIMED(gt->i915) && + !IS_VALLEYVIEW(gt->i915) && + !IS_CHERRYVIEW(gt->i915)) return 0; /* @@ -281,7 +283,7 @@ static int live_forcewake_domains(void *arg) for (offset = 0; offset < FW_RANGE; offset += 4) { i915_reg_t reg = { offset }; - (void)I915_READ_FW(reg); + intel_uncore_posting_read_fw(uncore, reg); if (!check_for_unclaimed_mmio(uncore)) set_bit(offset, valid); } @@ -298,7 +300,7 @@ static int live_forcewake_domains(void *arg) check_for_unclaimed_mmio(uncore); - (void)I915_READ(reg); + intel_uncore_posting_read_fw(uncore, reg); if (check_for_unclaimed_mmio(uncore)) { pr_err("Unclaimed mmio read to register 0x%04x\n", offset); @@ -310,21 +312,23 @@ static int live_forcewake_domains(void *arg) return err; } +static int live_fw_table(void *arg) +{ + struct intel_gt *gt = arg; + + /* Confirm the table we load is still valid */ + return intel_fw_table_check(gt->uncore->fw_domains_table, + gt->uncore->fw_domains_table_entries, + INTEL_GEN(gt->i915) >= 9); +} + int intel_uncore_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { + SUBTEST(live_fw_table), SUBTEST(live_forcewake_ops), SUBTEST(live_forcewake_domains), }; - int err; - - /* Confirm the table we load is still valid */ - err = intel_fw_table_check(i915->uncore.fw_domains_table, - i915->uncore.fw_domains_table_entries, - INTEL_GEN(i915) >= 9); - if (err) - return err; - - return i915_subtests(tests, i915); + return intel_gt_live_subtests(tests, &i915->gt); } diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c index 01a89c071bf5..27ed3cee6a9b 100644 --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c @@ -26,27 +26,29 @@ #include <linux/pm_runtime.h> #include "gt/intel_gt.h" +#include "gt/intel_gt_requests.h" #include "gt/mock_engine.h" +#include "intel_memory_region.h" #include "mock_request.h" #include "mock_gem_device.h" #include "mock_gtt.h" #include "mock_uncore.h" +#include "mock_region.h" #include "gem/selftests/mock_context.h" #include "gem/selftests/mock_gem_object.h" void mock_device_flush(struct drm_i915_private *i915) { + struct intel_gt *gt = &i915->gt; struct intel_engine_cs *engine; enum intel_engine_id id; - lockdep_assert_held(&i915->drm.struct_mutex); - do { - for_each_engine(engine, i915, id) + for_each_engine(engine, gt, id) mock_engine_flush(engine); - } while (i915_retire_requests(i915)); + } while (intel_gt_retire_requests_timeout(gt, MAX_SCHEDULE_TIMEOUT)); } static void mock_device_release(struct drm_device *dev) @@ -55,31 +57,23 @@ static void mock_device_release(struct drm_device *dev) struct intel_engine_cs *engine; enum intel_engine_id id; - mutex_lock(&i915->drm.struct_mutex); mock_device_flush(i915); - mutex_unlock(&i915->drm.struct_mutex); - flush_work(&i915->gem.idle_work); i915_gem_drain_workqueue(i915); - mutex_lock(&i915->drm.struct_mutex); - for_each_engine(engine, i915, id) + for_each_engine(engine, &i915->gt, id) mock_engine_free(engine); - i915_gem_contexts_fini(i915); - mutex_unlock(&i915->drm.struct_mutex); + i915_gem_driver_release__contexts(i915); intel_timelines_fini(i915); drain_workqueue(i915->wq); i915_gem_drain_freed_objects(i915); - mutex_lock(&i915->drm.struct_mutex); mock_fini_ggtt(&i915->ggtt); - mutex_unlock(&i915->drm.struct_mutex); - destroy_workqueue(i915->wq); - i915_gemfs_fini(i915); + intel_memory_regions_driver_release(i915); drm_mode_config_cleanup(&i915->drm); @@ -103,14 +97,6 @@ static void release_dev(struct device *dev) kfree(pdev); } -static void mock_retire_work_handler(struct work_struct *work) -{ -} - -static void mock_idle_work_handler(struct work_struct *work) -{ -} - static int pm_domain_resume(struct device *dev) { return pm_generic_runtime_resume(dev); @@ -178,10 +164,15 @@ struct drm_i915_private *mock_gem_device(void) I915_GTT_PAGE_SIZE_64K | I915_GTT_PAGE_SIZE_2M; - mock_uncore_init(&i915->uncore); + mkwrite_device_info(i915)->memory_regions = REGION_SMEM; + intel_memory_regions_hw_probe(i915); + + mock_uncore_init(&i915->uncore, i915); + i915_gem_init__mm(i915); intel_gt_init_early(&i915->gt, i915); atomic_inc(&i915->gt.wakeref.count); /* disable; no hw support */ + i915->gt.awake = -ENODEV; i915->wq = alloc_ordered_workqueue("mock", 0); if (!i915->wq) @@ -189,15 +180,8 @@ struct drm_i915_private *mock_gem_device(void) mock_init_contexts(i915); - INIT_DELAYED_WORK(&i915->gem.retire_work, mock_retire_work_handler); - INIT_WORK(&i915->gem.idle_work, mock_idle_work_handler); - - i915->gt.awake = true; - intel_timelines_init(i915); - mutex_lock(&i915->drm.struct_mutex); - mock_init_ggtt(i915, &i915->ggtt); mkwrite_device_info(i915)->engine_mask = BIT(0); @@ -214,21 +198,18 @@ struct drm_i915_private *mock_gem_device(void) goto err_context; intel_engines_driver_register(i915); - mutex_unlock(&i915->drm.struct_mutex); - - WARN_ON(i915_gemfs_init(i915)); return i915; err_context: - i915_gem_contexts_fini(i915); + i915_gem_driver_release__contexts(i915); err_engine: mock_engine_free(i915->engine[RCS0]); err_unlock: - mutex_unlock(&i915->drm.struct_mutex); intel_timelines_fini(i915); destroy_workqueue(i915->wq); err_drv: + intel_memory_regions_driver_release(i915); drm_mode_config_cleanup(&i915->drm); drm_dev_fini(&i915->drm); put_device: diff --git a/drivers/gpu/drm/i915/selftests/mock_gtt.c b/drivers/gpu/drm/i915/selftests/mock_gtt.c index e62a67e0f79c..20ac3844edec 100644 --- a/drivers/gpu/drm/i915/selftests/mock_gtt.c +++ b/drivers/gpu/drm/i915/selftests/mock_gtt.c @@ -43,7 +43,7 @@ static int mock_bind_ppgtt(struct i915_vma *vma, u32 flags) { GEM_BUG_ON(flags & I915_VMA_GLOBAL_BIND); - vma->flags |= I915_VMA_LOCAL_BIND; + set_bit(I915_VMA_LOCAL_BIND_BIT, __i915_vma_flags(vma)); return 0; } @@ -63,6 +63,7 @@ struct i915_ppgtt *mock_ppgtt(struct drm_i915_private *i915, const char *name) if (!ppgtt) return NULL; + ppgtt->vm.gt = &i915->gt; ppgtt->vm.i915 = i915; ppgtt->vm.total = round_down(U64_MAX, PAGE_SIZE); ppgtt->vm.file = ERR_PTR(-ENODEV); @@ -86,7 +87,7 @@ static int mock_bind_ggtt(struct i915_vma *vma, enum i915_cache_level cache_level, u32 flags) { - vma->flags |= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND; + atomic_or(I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND, &vma->flags); return 0; } @@ -117,8 +118,7 @@ void mock_init_ggtt(struct drm_i915_private *i915, struct i915_ggtt *ggtt) ggtt->vm.vma_ops.clear_pages = clear_pages; i915_address_space_init(&ggtt->vm, VM_CLASS_GGTT); - - intel_gt_init_hw(i915); + i915->gt.ggtt = ggtt; } void mock_fini_ggtt(struct i915_ggtt *ggtt) diff --git a/drivers/gpu/drm/i915/selftests/mock_region.c b/drivers/gpu/drm/i915/selftests/mock_region.c new file mode 100644 index 000000000000..b2ad41c27e67 --- /dev/null +++ b/drivers/gpu/drm/i915/selftests/mock_region.c @@ -0,0 +1,60 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include "gem/i915_gem_region.h" +#include "intel_memory_region.h" + +#include "mock_region.h" + +static const struct drm_i915_gem_object_ops mock_region_obj_ops = { + .get_pages = i915_gem_object_get_pages_buddy, + .put_pages = i915_gem_object_put_pages_buddy, + .release = i915_gem_object_release_memory_region, +}; + +static struct drm_i915_gem_object * +mock_object_create(struct intel_memory_region *mem, + resource_size_t size, + unsigned int flags) +{ + static struct lock_class_key lock_class; + struct drm_i915_private *i915 = mem->i915; + struct drm_i915_gem_object *obj; + + if (size > BIT(mem->mm.max_order) * mem->mm.chunk_size) + return ERR_PTR(-E2BIG); + + obj = i915_gem_object_alloc(); + if (!obj) + return ERR_PTR(-ENOMEM); + + drm_gem_private_object_init(&i915->drm, &obj->base, size); + i915_gem_object_init(obj, &mock_region_obj_ops, &lock_class); + + obj->read_domains = I915_GEM_DOMAIN_CPU | I915_GEM_DOMAIN_GTT; + + i915_gem_object_set_cache_coherency(obj, I915_CACHE_NONE); + + i915_gem_object_init_memory_region(obj, mem, flags); + + return obj; +} + +static const struct intel_memory_region_ops mock_region_ops = { + .init = intel_memory_region_init_buddy, + .release = intel_memory_region_release_buddy, + .create_object = mock_object_create, +}; + +struct intel_memory_region * +mock_region_create(struct drm_i915_private *i915, + resource_size_t start, + resource_size_t size, + resource_size_t min_page_size, + resource_size_t io_start) +{ + return intel_memory_region_create(i915, start, size, min_page_size, + io_start, &mock_region_ops); +} diff --git a/drivers/gpu/drm/i915/selftests/mock_region.h b/drivers/gpu/drm/i915/selftests/mock_region.h new file mode 100644 index 000000000000..24608089d833 --- /dev/null +++ b/drivers/gpu/drm/i915/selftests/mock_region.h @@ -0,0 +1,16 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __MOCK_REGION_H +#define __MOCK_REGION_H + +struct intel_memory_region * +mock_region_create(struct drm_i915_private *i915, + resource_size_t start, + resource_size_t size, + resource_size_t min_page_size, + resource_size_t io_start); + +#endif /* !__MOCK_REGION_H */ diff --git a/drivers/gpu/drm/i915/selftests/mock_uncore.c b/drivers/gpu/drm/i915/selftests/mock_uncore.c index 49585f16d4a2..ca57e4008701 100644 --- a/drivers/gpu/drm/i915/selftests/mock_uncore.c +++ b/drivers/gpu/drm/i915/selftests/mock_uncore.c @@ -39,8 +39,11 @@ __nop_read(16) __nop_read(32) __nop_read(64) -void mock_uncore_init(struct intel_uncore *uncore) +void mock_uncore_init(struct intel_uncore *uncore, + struct drm_i915_private *i915) { + intel_uncore_init_early(uncore, i915); + ASSIGN_RAW_WRITE_MMIO_VFUNCS(uncore, nop); ASSIGN_RAW_READ_MMIO_VFUNCS(uncore, nop); } diff --git a/drivers/gpu/drm/i915/selftests/mock_uncore.h b/drivers/gpu/drm/i915/selftests/mock_uncore.h index dacb36b5ffcd..8a2cc553f466 100644 --- a/drivers/gpu/drm/i915/selftests/mock_uncore.h +++ b/drivers/gpu/drm/i915/selftests/mock_uncore.h @@ -25,6 +25,7 @@ #ifndef __MOCK_UNCORE_H #define __MOCK_UNCORE_H -void mock_uncore_init(struct intel_uncore *uncore); +void mock_uncore_init(struct intel_uncore *uncore, + struct drm_i915_private *i915); #endif /* !__MOCK_UNCORE_H */ |