Merge tag 'drm-next-2022-01-07' of git://anongit.freedesktop.org/drm/drm

Pull drm updates from Dave Airlie:
 "Highlights are support for privacy screens found in new laptops, a
  bunch of nomodeset refactoring, and i915 enables ADL-P systems by
  default, while starting to add RPL-S support.

  vmwgfx adds GEM and support for OpenGL 4.3 features in userspace.

  Lots of internal refactorings around dma reservations, and lots of
  driver refactoring as well.

  Summary:

  core:
   - add privacy screen support
   - move nomodeset option into drm subsystem
   - clean up nomodeset handling in drivers
   - make drm_irq.c legacy
   - fix stack_depot name conflicts
   - remove DMA_BUF_SET_NAME ioctl restrictions
   - sysfs: send hotplug event
   - replace several DRM_* logging macros with drm_*
   - move hashtable to legacy code
   - add error return from gem_create_object
   - cma-helper: improve interfaces, drop CONFIG_DRM_KMS_CMA_HELPER
   - kernel.h related include cleanups
   - support XRGB2101010 source buffers

  ttm:
   - don't include drm hashtable
   - stop pruning fences after wait
   - documentation updates

  dma-buf:
   - add dma_resv selftest
   - add debugfs helpers
   - remove dma_resv_get_excl_unlocked
   - documentation
   - make fences mandatory in dma_resv_add_excl_fence

  dp:
   - add link training delay helpers

  gem:
   - link shmem/cma helpers into separate modules
   - use dma_resv iteratior
   - import dma-buf namespace into gem helper modules

  scheduler:
   - fence grab fix
   - lockdep fixes

  bridge:
   - switch to managed MIPI DSI helpers
   - register and attach during probe fixes
   - convert to YAML in several places.

  panel:
   - add bunch of new panesl

  simpledrm:
   - support FB_DAMAGE_CLIPS
   - support virtual screen sizes
   - add Apple M1 support

  amdgpu:
   - enable seamless boot for DCN 3.01
   - runtime PM fixes
   - use drm_kms_helper_connector_hotplug_event
   - get all fences at once
   - use generic drm fb helpers
   - PSR/DPCD/LTTPR/DSC/PM/RAS/OLED/SRIOV fixes
   - add smart trace buffer (STB) for supported GPUs
   - display debugfs entries
   - new SMU debug option
   - Documentation update

  amdkfd:
   - IP discovery enumeration refactor
   - interface between driver fixes
   - SVM fixes
   - kfd uapi header to define some sysfs bitfields.

  i915:
   - support VESA panel backlights
   - enable ADL-P by default
   - add eDP privacy screen support
   - add Raptor Lake S (RPL-S) support
   - DG2 page table support
   - lots of GuC/HuC fw refactoring
   - refactored i915->gt interfaces
   - CD clock squashing support
   - enable 10-bit gamma support
   - update ADL-P DMC fw to v2.14
   - enable runtime PM autosuspend by default
   - ADL-P DSI support
   - per-lane DP drive settings for ICL+
   - add support for pipe C/D DMC firmware
   - Atomic gamma LUT updates
   - remove CCS FB stride restrictions on ADL-P
   - VRR platform support for display 11
   - add support for display audio codec keepalive
   - lots of display refactoring
   - fix runtime PM handling during PXP suspend
   - improved eviction performance with async TTM moves
   - async VMA unbinding improvements
   - VMA locking refactoring
   - improved error capture robustness
   - use per device iommu checks
   - drop bits stealing from i915_sw_fence function ptr
   - remove dma_resv_prune
   - add IC cache invalidation on DG2

  nouveau:
   - crc fixes
   - validate LUTs in atomic check
   - set HDMI AVI RGB quant to full

  tegra:
   - buffer objects reworks for dma-buf compat
   - NVDEC driver uAPI support
   - power management improvements

  etnaviv:
   - IOMMU enabled system support
   - fix > 4GB command buffer mapping
   - close a DoS vector
   - fix spurious GPU resets

  ast:
   - fix i2c initialization

  rcar-du:
   - DSI output support

  exynos:
   - replace legacy gpio interface
   - implement generic GEM object mmap

  msm:
   - dpu plane state cleanup in prep for multirect
   - dpu debugfs cleanups
   - dp support for sc7280
   - a506 support
   - removal of struct_mutex
   - remove old eDP sub-driver

  anx7625:
   - support MIPI DSI input
   - support HDMI audio
   - fix reading EDID

  lvds:
   - fix bridge DT bindings

  megachips:
   - probe both bridges before registering

  dw-hdmi:
   - allow interlace on bridge

  ps8640:
   - enable runtime PM
   - support aux-bus

  tx358768:
   - enable reference clock
   - add pulse mode support

  ti-sn65dsi86:
   - use regmap bulk write
   - add PWM support

  etnaviv:
   - get all fences at once

  gma500:
   - gem object cleanups

  kmb:
   - enable fb console

  radeon:
   - use dma_resv_wait_timeout

  rockchip:
   - add DSP hold timeout
   - suspend/resume fixes
   - PLL clock fixes
   - implement mmap in GEM object functions
   - use generic fbdev emulation

  sun4i:
   - use CMA helpers without vmap support

  vc4:
   - fix HDMI-CEC hang with display is off
   - power on HDMI controller while disabling
   - support 4K@60Hz modes
   - support 10-bit YUV 4:2:0 output

  vmwgfx:
   - fix leak on probe errors
   - fail probing on broken hosts
   - new placement for MOB page tables
   - hide internal BOs from userspace
   - implement GEM support
   - implement GL 4.3 support

  virtio:
   - overflow fixes

  xen:
   - implement mmap as GEM object function

  omapdrm:
   - fix scatterlist export
   - support virtual planes

  mediatek:
   - MT8192 support
   - CMDQ refinement"

* tag 'drm-next-2022-01-07' of git://anongit.freedesktop.org/drm/drm: (1241 commits)
  drm/amdgpu: no DC support for headless chips
  drm/amd/display: fix dereference before NULL check
  drm/amdgpu: always reset the asic in suspend (v2)
  drm/amdgpu: put SMU into proper state on runpm suspending for BOCO capable platform
  drm/amd/display: Fix the uninitialized variable in enable_stream_features()
  drm/amdgpu: fix runpm documentation
  amdgpu/pm: Make sysfs pm attributes as read-only for VFs
  drm/amdgpu: save error count in RAS poison handler
  drm/amdgpu: drop redundant semicolon
  drm/amd/display: get and restore link res map
  drm/amd/display: support dynamic HPO DP link encoder allocation
  drm/amd/display: access hpo dp link encoder only through link resource
  drm/amd/display: populate link res in both detection and validation
  drm/amd/display: define link res and make it accessible to all link interfaces
  drm/amd/display: 3.2.167
  drm/amd/display: [FW Promotion] Release 0.0.98
  drm/amd/display: Undo ODM combine
  drm/amd/display: Add reg defs for DCN303
  drm/amd/display: Changed pipe split policy to allow for multi-display pipe split
  drm/amd/display: Set optimize_pwr_state for DCN31
  ...

1 files changed, 103 insertions, 1 deletions

diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
index 6e90209cd543..55f443328d8e 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
@@ -42,7 +42,15 @@ struct a6xx_gpu_state {
 	struct a6xx_gpu_state_obj *cx_debugbus;
 	int nr_cx_debugbus;
 
+	struct msm_gpu_state_bo *gmu_log;
+	struct msm_gpu_state_bo *gmu_hfi;
+	struct msm_gpu_state_bo *gmu_debug;
+
+	s32 hfi_queue_history[2][HFI_HISTORY_SZ];
+
 	struct list_head objs;
+
+	bool gpu_initialized;
 };
 
 static inline int CRASHDUMP_WRITE(u64 *in, u32 reg, u32 val)
@@ -800,6 +808,45 @@ static void a6xx_get_gmu_registers(struct msm_gpu *gpu,
 		&a6xx_state->gmu_registers[2], false);
 }
 
+static struct msm_gpu_state_bo *a6xx_snapshot_gmu_bo(
+		struct a6xx_gpu_state *a6xx_state, struct a6xx_gmu_bo *bo)
+{
+	struct msm_gpu_state_bo *snapshot;
+
+	snapshot = state_kcalloc(a6xx_state, 1, sizeof(*snapshot));
+	if (!snapshot)
+		return NULL;
+
+	snapshot->iova = bo->iova;
+	snapshot->size = bo->size;
+	snapshot->data = kvzalloc(snapshot->size, GFP_KERNEL);
+	if (!snapshot->data)
+		return NULL;
+
+	memcpy(snapshot->data, bo->virt, bo->size);
+
+	return snapshot;
+}
+
+static void a6xx_snapshot_gmu_hfi_history(struct msm_gpu *gpu,
+					  struct a6xx_gpu_state *a6xx_state)
+{
+	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
+	struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
+	struct a6xx_gmu *gmu = &a6xx_gpu->gmu;
+	unsigned i, j;
+
+	BUILD_BUG_ON(ARRAY_SIZE(gmu->queues) != ARRAY_SIZE(a6xx_state->hfi_queue_history));
+
+	for (i = 0; i < ARRAY_SIZE(gmu->queues); i++) {
+		struct a6xx_hfi_queue *queue = &gmu->queues[i];
+		for (j = 0; j < HFI_HISTORY_SZ; j++) {
+			unsigned idx = (j + queue->history_idx) % HFI_HISTORY_SZ;
+			a6xx_state->hfi_queue_history[i][j] = queue->history[idx];
+		}
+	}
+}
+
 #define A6XX_GBIF_REGLIST_SIZE   1
 static void a6xx_get_registers(struct msm_gpu *gpu,
 		struct a6xx_gpu_state *a6xx_state,
@@ -937,6 +984,12 @@ struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
 
 	a6xx_get_gmu_registers(gpu, a6xx_state);
 
+	a6xx_state->gmu_log = a6xx_snapshot_gmu_bo(a6xx_state, &a6xx_gpu->gmu.log);
+	a6xx_state->gmu_hfi = a6xx_snapshot_gmu_bo(a6xx_state, &a6xx_gpu->gmu.hfi);
+	a6xx_state->gmu_debug = a6xx_snapshot_gmu_bo(a6xx_state, &a6xx_gpu->gmu.debug);
+
+	a6xx_snapshot_gmu_hfi_history(gpu, a6xx_state);
+
 	/* If GX isn't on the rest of the data isn't going to be accessible */
 	if (!a6xx_gmu_gx_is_on(&a6xx_gpu->gmu))
 		return &a6xx_state->base;
@@ -950,7 +1003,8 @@ struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
 	 * write out GPU state, so we need to skip this when the SMMU is
 	 * stalled in response to an iova fault
 	 */
-	if (!stalled && !a6xx_crashdumper_init(gpu, &_dumper)) {
+	if (!stalled && !gpu->needs_hw_init &&
+	    !a6xx_crashdumper_init(gpu, &_dumper)) {
 		dumper = &_dumper;
 	}
 
@@ -967,6 +1021,8 @@ struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
 	if (snapshot_debugbus)
 		a6xx_get_debugbus(gpu, a6xx_state);
 
+	a6xx_state->gpu_initialized = !gpu->needs_hw_init;
+
 	return  &a6xx_state->base;
 }
 
@@ -978,6 +1034,12 @@ static void a6xx_gpu_state_destroy(struct kref *kref)
 	struct a6xx_gpu_state *a6xx_state = container_of(state,
 			struct a6xx_gpu_state, base);
 
+	if (a6xx_state->gmu_log)
+		kvfree(a6xx_state->gmu_log->data);
+
+	if (a6xx_state->gmu_hfi)
+		kvfree(a6xx_state->gmu_hfi->data);
+
 	list_for_each_entry_safe(obj, tmp, &a6xx_state->objs, node)
 		kfree(obj);
 
@@ -1189,8 +1251,48 @@ void a6xx_show(struct msm_gpu *gpu, struct msm_gpu_state *state,
 	if (IS_ERR_OR_NULL(state))
 		return;
 
+	drm_printf(p, "gpu-initialized: %d\n", a6xx_state->gpu_initialized);
+
 	adreno_show(gpu, state, p);
 
+	drm_puts(p, "gmu-log:\n");
+	if (a6xx_state->gmu_log) {
+		struct msm_gpu_state_bo *gmu_log = a6xx_state->gmu_log;
+
+		drm_printf(p, "    iova: 0x%016llx\n", gmu_log->iova);
+		drm_printf(p, "    size: %zu\n", gmu_log->size);
+		adreno_show_object(p, &gmu_log->data, gmu_log->size,
+				&gmu_log->encoded);
+	}
+
+	drm_puts(p, "gmu-hfi:\n");
+	if (a6xx_state->gmu_hfi) {
+		struct msm_gpu_state_bo *gmu_hfi = a6xx_state->gmu_hfi;
+		unsigned i, j;
+
+		drm_printf(p, "    iova: 0x%016llx\n", gmu_hfi->iova);
+		drm_printf(p, "    size: %zu\n", gmu_hfi->size);
+		for (i = 0; i < ARRAY_SIZE(a6xx_state->hfi_queue_history); i++) {
+			drm_printf(p, "    queue-history[%u]:", i);
+			for (j = 0; j < HFI_HISTORY_SZ; j++) {
+				drm_printf(p, " %d", a6xx_state->hfi_queue_history[i][j]);
+			}
+			drm_printf(p, "\n");
+		}
+		adreno_show_object(p, &gmu_hfi->data, gmu_hfi->size,
+				&gmu_hfi->encoded);
+	}
+
+	drm_puts(p, "gmu-debug:\n");
+	if (a6xx_state->gmu_debug) {
+		struct msm_gpu_state_bo *gmu_debug = a6xx_state->gmu_debug;
+
+		drm_printf(p, "    iova: 0x%016llx\n", gmu_debug->iova);
+		drm_printf(p, "    size: %zu\n", gmu_debug->size);
+		adreno_show_object(p, &gmu_debug->data, gmu_debug->size,
+				&gmu_debug->encoded);
+	}
+
 	drm_puts(p, "registers:\n");
 	for (i = 0; i < a6xx_state->nr_registers; i++) {
 		struct a6xx_gpu_state_obj *obj = &a6xx_state->registers[i];