diff options
Diffstat (limited to 'drivers/gpu/drm/i915/gvt')
| -rw-r--r-- | drivers/gpu/drm/i915/gvt/aperture_gm.c | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gvt/fb_decoder.c | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gvt/gtt.c | 116 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gvt/gtt.h | 10 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gvt/handlers.c | 8 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gvt/mmio_context.c | 4 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gvt/scheduler.c | 28 |
7 files changed, 104 insertions, 66 deletions
diff --git a/drivers/gpu/drm/i915/gvt/aperture_gm.c b/drivers/gpu/drm/i915/gvt/aperture_gm.c index fe754022e356..359d37d5c958 100644 --- a/drivers/gpu/drm/i915/gvt/aperture_gm.c +++ b/drivers/gpu/drm/i915/gvt/aperture_gm.c @@ -61,10 +61,12 @@ static int alloc_gm(struct intel_vgpu *vgpu, bool high_gm) } mutex_lock(&dev_priv->drm.struct_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); if (ret) gvt_err("fail to alloc %s gm space from host\n", diff --git a/drivers/gpu/drm/i915/gvt/fb_decoder.c b/drivers/gpu/drm/i915/gvt/fb_decoder.c index 481896fb712a..85e6736f0a32 100644 --- a/drivers/gpu/drm/i915/gvt/fb_decoder.c +++ b/drivers/gpu/drm/i915/gvt/fb_decoder.c @@ -235,7 +235,7 @@ int intel_vgpu_decode_primary_plane(struct intel_vgpu *vgpu, plane->bpp = skl_pixel_formats[fmt].bpp; plane->drm_format = skl_pixel_formats[fmt].drm_format; } else { - plane->tiled = !!(val & DISPPLANE_TILED); + plane->tiled = val & DISPPLANE_TILED; fmt = bdw_format_to_drm(val & DISPPLANE_PIXFORMAT_MASK); plane->bpp = bdw_pixel_formats[fmt].bpp; plane->drm_format = bdw_pixel_formats[fmt].drm_format; diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 2402395a068d..c7103dd2d8d5 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -1905,7 +1905,6 @@ static struct intel_vgpu_mm *intel_vgpu_create_ggtt_mm(struct intel_vgpu *vgpu) vgpu_free_mm(mm); return ERR_PTR(-ENOMEM); } - mm->ggtt_mm.last_partial_off = -1UL; return mm; } @@ -1930,7 +1929,6 @@ void _intel_vgpu_mm_release(struct kref *mm_ref) invalidate_ppgtt_mm(mm); } else { vfree(mm->ggtt_mm.virtual_ggtt); - mm->ggtt_mm.last_partial_off = -1UL; } vgpu_free_mm(mm); @@ -2168,6 +2166,8 @@ static int emulate_ggtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off, struct intel_gvt_gtt_entry e, m; dma_addr_t dma_addr; int ret; + struct intel_gvt_partial_pte *partial_pte, *pos, *n; + bool partial_update = false; if (bytes != 4 && bytes != 8) return -EINVAL; @@ -2178,68 +2178,57 @@ static int emulate_ggtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off, if (!vgpu_gmadr_is_valid(vgpu, gma)) return 0; - ggtt_get_guest_entry(ggtt_mm, &e, g_gtt_index); - + e.type = GTT_TYPE_GGTT_PTE; memcpy((void *)&e.val64 + (off & (info->gtt_entry_size - 1)), p_data, bytes); /* If ggtt entry size is 8 bytes, and it's split into two 4 bytes - * write, we assume the two 4 bytes writes are consecutive. - * Otherwise, we abort and report error + * write, save the first 4 bytes in a list and update virtual + * PTE. Only update shadow PTE when the second 4 bytes comes. */ if (bytes < info->gtt_entry_size) { - if (ggtt_mm->ggtt_mm.last_partial_off == -1UL) { - /* the first partial part*/ - ggtt_mm->ggtt_mm.last_partial_off = off; - ggtt_mm->ggtt_mm.last_partial_data = e.val64; - return 0; - } else if ((g_gtt_index == - (ggtt_mm->ggtt_mm.last_partial_off >> - info->gtt_entry_size_shift)) && - (off != ggtt_mm->ggtt_mm.last_partial_off)) { - /* the second partial part */ - - int last_off = ggtt_mm->ggtt_mm.last_partial_off & - (info->gtt_entry_size - 1); - - memcpy((void *)&e.val64 + last_off, - (void *)&ggtt_mm->ggtt_mm.last_partial_data + - last_off, bytes); - - ggtt_mm->ggtt_mm.last_partial_off = -1UL; - } else { - int last_offset; - - gvt_vgpu_err("failed to populate guest ggtt entry: abnormal ggtt entry write sequence, last_partial_off=%lx, offset=%x, bytes=%d, ggtt entry size=%d\n", - ggtt_mm->ggtt_mm.last_partial_off, off, - bytes, info->gtt_entry_size); - - /* set host ggtt entry to scratch page and clear - * virtual ggtt entry as not present for last - * partially write offset - */ - last_offset = ggtt_mm->ggtt_mm.last_partial_off & - (~(info->gtt_entry_size - 1)); - - ggtt_get_host_entry(ggtt_mm, &m, last_offset); - ggtt_invalidate_pte(vgpu, &m); - ops->set_pfn(&m, gvt->gtt.scratch_mfn); - ops->clear_present(&m); - ggtt_set_host_entry(ggtt_mm, &m, last_offset); - ggtt_invalidate(gvt->dev_priv); - - ggtt_get_guest_entry(ggtt_mm, &e, last_offset); - ops->clear_present(&e); - ggtt_set_guest_entry(ggtt_mm, &e, last_offset); - - ggtt_mm->ggtt_mm.last_partial_off = off; - ggtt_mm->ggtt_mm.last_partial_data = e.val64; + bool found = false; + + list_for_each_entry_safe(pos, n, + &ggtt_mm->ggtt_mm.partial_pte_list, list) { + if (g_gtt_index == pos->offset >> + info->gtt_entry_size_shift) { + if (off != pos->offset) { + /* the second partial part*/ + int last_off = pos->offset & + (info->gtt_entry_size - 1); + + memcpy((void *)&e.val64 + last_off, + (void *)&pos->data + last_off, + bytes); + + list_del(&pos->list); + kfree(pos); + found = true; + break; + } + + /* update of the first partial part */ + pos->data = e.val64; + ggtt_set_guest_entry(ggtt_mm, &e, g_gtt_index); + return 0; + } + } - return 0; + if (!found) { + /* the first partial part */ + partial_pte = kzalloc(sizeof(*partial_pte), GFP_KERNEL); + if (!partial_pte) + return -ENOMEM; + partial_pte->offset = off; + partial_pte->data = e.val64; + list_add_tail(&partial_pte->list, + &ggtt_mm->ggtt_mm.partial_pte_list); + partial_update = true; } } - if (ops->test_present(&e)) { + if (!partial_update && (ops->test_present(&e))) { gfn = ops->get_pfn(&e); m = e; @@ -2263,16 +2252,18 @@ static int emulate_ggtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off, } else ops->set_pfn(&m, dma_addr >> PAGE_SHIFT); } else { - ggtt_get_host_entry(ggtt_mm, &m, g_gtt_index); - ggtt_invalidate_pte(vgpu, &m); ops->set_pfn(&m, gvt->gtt.scratch_mfn); ops->clear_present(&m); } out: + ggtt_set_guest_entry(ggtt_mm, &e, g_gtt_index); + + ggtt_get_host_entry(ggtt_mm, &e, g_gtt_index); + ggtt_invalidate_pte(vgpu, &e); + ggtt_set_host_entry(ggtt_mm, &m, g_gtt_index); ggtt_invalidate(gvt->dev_priv); - ggtt_set_guest_entry(ggtt_mm, &e, g_gtt_index); return 0; } @@ -2430,6 +2421,8 @@ int intel_vgpu_init_gtt(struct intel_vgpu *vgpu) intel_vgpu_reset_ggtt(vgpu, false); + INIT_LIST_HEAD(>t->ggtt_mm->ggtt_mm.partial_pte_list); + return create_scratch_page_tree(vgpu); } @@ -2454,6 +2447,15 @@ static void intel_vgpu_destroy_all_ppgtt_mm(struct intel_vgpu *vgpu) static void intel_vgpu_destroy_ggtt_mm(struct intel_vgpu *vgpu) { + struct intel_gvt_partial_pte *pos, *next; + + list_for_each_entry_safe(pos, next, + &vgpu->gtt.ggtt_mm->ggtt_mm.partial_pte_list, + list) { + gvt_dbg_mm("partial PTE update on hold 0x%lx : 0x%llx\n", + pos->offset, pos->data); + kfree(pos); + } intel_vgpu_destroy_mm(vgpu->gtt.ggtt_mm); vgpu->gtt.ggtt_mm = NULL; } diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h index 7a9b36176efb..d8cb04cc946d 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.h +++ b/drivers/gpu/drm/i915/gvt/gtt.h @@ -35,7 +35,6 @@ #define _GVT_GTT_H_ #define I915_GTT_PAGE_SHIFT 12 -#define I915_GTT_PAGE_MASK (~(I915_GTT_PAGE_SIZE - 1)) struct intel_vgpu_mm; @@ -133,6 +132,12 @@ enum intel_gvt_mm_type { #define GVT_RING_CTX_NR_PDPS GEN8_3LVL_PDPES +struct intel_gvt_partial_pte { + unsigned long offset; + u64 data; + struct list_head list; +}; + struct intel_vgpu_mm { enum intel_gvt_mm_type type; struct intel_vgpu *vgpu; @@ -157,8 +162,7 @@ struct intel_vgpu_mm { } ppgtt_mm; struct { void *virtual_ggtt; - unsigned long last_partial_off; - u64 last_partial_data; + struct list_head partial_pte_list; } ggtt_mm; }; }; diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 90f50f67909a..aa280bb07125 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1609,7 +1609,7 @@ static int bxt_gt_disp_pwron_write(struct intel_vgpu *vgpu, return 0; } -static int bxt_edp_psr_imr_iir_write(struct intel_vgpu *vgpu, +static int edp_psr_imr_iir_write(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { vgpu_vreg(vgpu, offset) = 0; @@ -2607,6 +2607,9 @@ static int init_generic_mmio_info(struct intel_gvt *gvt) MMIO_DFH(_MMIO(0x1a178), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL); MMIO_DFH(_MMIO(0x1a17c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL); MMIO_DFH(_MMIO(0x2217c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL); + + MMIO_DH(EDP_PSR_IMR, D_BDW_PLUS, NULL, edp_psr_imr_iir_write); + MMIO_DH(EDP_PSR_IIR, D_BDW_PLUS, NULL, edp_psr_imr_iir_write); return 0; } @@ -3205,9 +3208,6 @@ static int init_bxt_mmio_info(struct intel_gvt *gvt) MMIO_D(HSW_TVIDEO_DIP_GCP(TRANSCODER_B), D_BXT); MMIO_D(HSW_TVIDEO_DIP_GCP(TRANSCODER_C), D_BXT); - MMIO_DH(EDP_PSR_IMR, D_BXT, NULL, bxt_edp_psr_imr_iir_write); - MMIO_DH(EDP_PSR_IIR, D_BXT, NULL, bxt_edp_psr_imr_iir_write); - MMIO_D(RC6_CTX_BASE, D_BXT); MMIO_D(GEN8_PUSHBUS_CONTROL, D_BXT); diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c index 10e63eea5492..d6e02c15ef97 100644 --- a/drivers/gpu/drm/i915/gvt/mmio_context.c +++ b/drivers/gpu/drm/i915/gvt/mmio_context.c @@ -131,7 +131,7 @@ static struct engine_mmio gen9_engine_mmio_list[] __cacheline_aligned = { {RCS, GAMT_CHKN_BIT_REG, 0x0, false}, /* 0x4ab8 */ {RCS, GEN9_GAMT_ECO_REG_RW_IA, 0x0, false}, /* 0x4ab0 */ - {RCS, GEN9_CSFE_CHICKEN1_RCS, 0x0, false}, /* 0x20d4 */ + {RCS, GEN9_CSFE_CHICKEN1_RCS, 0xffff, false}, /* 0x20d4 */ {RCS, GEN8_GARBCNTL, 0x0, false}, /* 0xb004 */ {RCS, GEN7_FF_THREAD_MODE, 0x0, false}, /* 0x20a0 */ @@ -158,6 +158,8 @@ static void load_render_mocs(struct drm_i915_private *dev_priv) int ring_id, i; for (ring_id = 0; ring_id < ARRAY_SIZE(regs); ring_id++) { + if (!HAS_ENGINE(dev_priv, ring_id)) + continue; offset.reg = regs[ring_id]; for (i = 0; i < GEN9_MOCS_SIZE; i++) { gen9_render_mocs.control_table[ring_id][i] = diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index ea34003d6dd2..b8fbe3fabea3 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -334,6 +334,28 @@ static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) i915_gem_object_put(wa_ctx->indirect_ctx.obj); } +static int 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_hw_ppgtt *ppgtt = ctx->ppgtt; + int i = 0; + + if (mm->type != INTEL_GVT_MM_PPGTT || !mm->ppgtt_mm.shadowed) + return -1; + + if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) { + px_dma(&ppgtt->pml4) = mm->ppgtt_mm.shadow_pdps[0]; + } else { + for (i = 0; i < GVT_RING_CTX_NR_PDPS; i++) { + px_dma(ppgtt->pdp.page_directory[i]) = + mm->ppgtt_mm.shadow_pdps[i]; + } + } + + return 0; +} + /** * intel_gvt_scan_and_shadow_workload - audit the workload by scanning and * shadow it as well, include ringbuffer,wa_ctx and ctx. @@ -358,6 +380,12 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) if (workload->req) return 0; + ret = set_context_ppgtt_from_shadow(workload, shadow_ctx); + if (ret < 0) { + gvt_vgpu_err("workload shadow ppgtt isn't ready\n"); + return ret; + } + /* pin shadow context by gvt even the shadow context will be pinned * when i915 alloc request. That is because gvt will update the guest * context from shadow context when workload is completed, and at that |