From 89351925a477441ae5fdd0136aec80b49ea1c53c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 29 Jul 2020 17:42:18 +0100 Subject: drm/i915/gt: Switch to object allocations for page directories The GEM object is grossly overweight for the practicality of tracking large numbers of individual pages, yet it is currently our only abstraction for tracking DMA allocations. Since those allocations need to be reserved upfront before an operation, and that we need to break away from simple system memory, we need to ditch using plain struct page wrappers. In the process, we drop the WC mapping as we ended up clflushing everything anyway due to various issues across a wider range of platforms. Though in a future step, we need to drop the kmap_atomic approach which suggests we need to pre-map all the pages and keep them mapped. v2: Verify our large scratch page is suitably DMA aligned; and manually clear the scratch since we are allocating plain struct pages full of prior content. Signed-off-by: Chris Wilson Cc: Matthew Auld Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-2-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/gvt/scheduler.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 3c3b9842bbbd..1570eb8aa978 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -403,6 +403,14 @@ static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) wa_ctx->indirect_ctx.shadow_va = NULL; } +static void set_dma_address(struct i915_page_directory *pd, dma_addr_t addr) +{ + struct scatterlist *sg = pd->pt.base->mm.pages->sgl; + + /* This is not a good idea */ + sg->dma_address = addr; +} + static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload, struct intel_context *ce) { @@ -411,7 +419,7 @@ static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload, int i = 0; if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) { - px_dma(ppgtt->pd) = mm->ppgtt_mm.shadow_pdps[0]; + set_dma_address(ppgtt->pd, mm->ppgtt_mm.shadow_pdps[0]); } else { for (i = 0; i < GVT_RING_CTX_NR_PDPS; i++) { struct i915_page_directory * const pd = @@ -421,7 +429,8 @@ static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload, shadow ppgtt. */ if (!pd) break; - px_dma(pd) = mm->ppgtt_mm.shadow_pdps[i]; + + set_dma_address(pd, mm->ppgtt_mm.shadow_pdps[i]); } } } @@ -1240,13 +1249,13 @@ i915_context_ppgtt_root_restore(struct intel_vgpu_submission *s, int i; if (i915_vm_is_4lvl(&ppgtt->vm)) { - px_dma(ppgtt->pd) = s->i915_context_pml4; + set_dma_address(ppgtt->pd, s->i915_context_pml4); } else { for (i = 0; i < GEN8_3LVL_PDPES; i++) { struct i915_page_directory * const pd = i915_pd_entry(ppgtt->pd, i); - px_dma(pd) = s->i915_context_pdps[i]; + set_dma_address(pd, s->i915_context_pdps[i]); } } } -- cgit v1.2.3 From 80f0b679d6f0683f23cf98a511af3e44dd509472 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Wed, 19 Aug 2020 16:08:45 +0200 Subject: drm/i915: Add an implementation for i915_gem_ww_ctx locking, v2. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit i915_gem_ww_ctx is used to lock all gem bo's for pinning and memory eviction. We don't use it yet, but lets start adding the definition first. To use it, we have to pass a non-NULL ww to gem_object_lock, and don't unlock directly. It is done in i915_gem_ww_ctx_fini. Changes since v1: - Change ww_ctx and obj order in locking functions (Jonas Lahtinen) Signed-off-by: Maarten Lankhorst Reviewed-by: Thomas Hellström Link: https://patchwork.freedesktop.org/patch/msgid/20200819140904.1708856-6-maarten.lankhorst@linux.intel.com Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/display/intel_display.c | 4 +- drivers/gpu/drm/i915/gem/i915_gem_client_blt.c | 2 +- drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c | 4 +- drivers/gpu/drm/i915/gem/i915_gem_domain.c | 10 ++--- drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/gem/i915_gem_object.h | 38 +++++++++++++--- drivers/gpu/drm/i915/gem/i915_gem_object_types.h | 9 ++++ drivers/gpu/drm/i915/gem/i915_gem_pm.c | 2 +- drivers/gpu/drm/i915/gem/i915_gem_tiling.c | 2 +- drivers/gpu/drm/i915/gem/selftests/huge_pages.c | 2 +- .../drm/i915/gem/selftests/i915_gem_client_blt.c | 2 +- .../drm/i915/gem/selftests/i915_gem_coherency.c | 10 ++--- .../gpu/drm/i915/gem/selftests/i915_gem_context.c | 4 +- drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c | 4 +- drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c | 2 +- drivers/gpu/drm/i915/gt/selftest_workarounds.c | 2 +- drivers/gpu/drm/i915/gvt/cmd_parser.c | 2 +- drivers/gpu/drm/i915/i915_gem.c | 52 ++++++++++++++++++++-- drivers/gpu/drm/i915/i915_gem.h | 11 +++++ drivers/gpu/drm/i915/selftests/i915_gem.c | 41 +++++++++++++++++ drivers/gpu/drm/i915/selftests/i915_vma.c | 2 +- .../gpu/drm/i915/selftests/intel_memory_region.c | 2 +- 22 files changed, 170 insertions(+), 39 deletions(-) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c index 7d50b7177d40..27c0a2140441 100644 --- a/drivers/gpu/drm/i915/display/intel_display.c +++ b/drivers/gpu/drm/i915/display/intel_display.c @@ -2311,7 +2311,7 @@ err: void intel_unpin_fb_vma(struct i915_vma *vma, unsigned long flags) { - i915_gem_object_lock(vma->obj); + i915_gem_object_lock(vma->obj, NULL); if (flags & PLANE_HAS_FENCE) i915_vma_unpin_fence(vma); i915_gem_object_unpin_from_display_plane(vma); @@ -17194,7 +17194,7 @@ static int intel_framebuffer_init(struct intel_framebuffer *intel_fb, if (!intel_fb->frontbuffer) return -ENOMEM; - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); tiling = i915_gem_object_get_tiling(obj); stride = i915_gem_object_get_stride(obj); i915_gem_object_unlock(obj); 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 947c8aa8e13e..a85907f29c53 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c @@ -288,7 +288,7 @@ int i915_gem_schedule_fill_pages_blt(struct drm_i915_gem_object *obj, dma_fence_init(&work->dma, &clear_pages_work_ops, &fence_lock, 0, 0); i915_sw_fence_init(&work->wait, clear_pages_work_notify); - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); err = i915_sw_fence_await_reservation(&work->wait, obj->base.resv, NULL, true, 0, I915_FENCE_GFP); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c index 2679380159fc..27fddc22a7c6 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c @@ -128,7 +128,7 @@ static int i915_gem_begin_cpu_access(struct dma_buf *dma_buf, enum dma_data_dire if (err) return err; - err = i915_gem_object_lock_interruptible(obj); + err = i915_gem_object_lock_interruptible(obj, NULL); if (err) goto out; @@ -149,7 +149,7 @@ static int i915_gem_end_cpu_access(struct dma_buf *dma_buf, enum dma_data_direct if (err) return err; - err = i915_gem_object_lock_interruptible(obj); + err = i915_gem_object_lock_interruptible(obj, NULL); if (err) goto out; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c index 7f76fc68f498..c0acfc97fae3 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c @@ -32,7 +32,7 @@ void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj) if (!i915_gem_object_is_framebuffer(obj)) return; - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); __i915_gem_object_flush_for_display(obj); i915_gem_object_unlock(obj); } @@ -197,7 +197,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj, if (ret) return ret; - ret = i915_gem_object_lock_interruptible(obj); + ret = i915_gem_object_lock_interruptible(obj, NULL); if (ret) return ret; @@ -536,7 +536,7 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data, if (err) goto out; - err = i915_gem_object_lock_interruptible(obj); + err = i915_gem_object_lock_interruptible(obj, NULL); if (err) goto out_unpin; @@ -576,7 +576,7 @@ int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj, if (!i915_gem_object_has_struct_page(obj)) return -ENODEV; - ret = i915_gem_object_lock_interruptible(obj); + ret = i915_gem_object_lock_interruptible(obj, NULL); if (ret) return ret; @@ -630,7 +630,7 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj, if (!i915_gem_object_has_struct_page(obj)) return -ENODEV; - ret = i915_gem_object_lock_interruptible(obj); + ret = i915_gem_object_lock_interruptible(obj, NULL); if (ret) return ret; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index ccaaf9e599a0..51395a0ceb15 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -1076,7 +1076,7 @@ static void *reloc_iomap(struct drm_i915_gem_object *obj, if (use_cpu_reloc(cache, obj)) return NULL; - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_gtt_domain(obj, true); i915_gem_object_unlock(obj); if (err) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h index e5b9276d254c..488459143366 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h @@ -110,20 +110,44 @@ i915_gem_object_put(struct drm_i915_gem_object *obj) #define assert_object_held(obj) dma_resv_assert_held((obj)->base.resv) -static inline void i915_gem_object_lock(struct drm_i915_gem_object *obj) +static inline int __i915_gem_object_lock(struct drm_i915_gem_object *obj, + struct i915_gem_ww_ctx *ww, + bool intr) { - dma_resv_lock(obj->base.resv, NULL); + int ret; + + if (intr) + ret = dma_resv_lock_interruptible(obj->base.resv, ww ? &ww->ctx : NULL); + else + ret = dma_resv_lock(obj->base.resv, ww ? &ww->ctx : NULL); + + if (!ret && ww) + list_add_tail(&obj->obj_link, &ww->obj_list); + if (ret == -EALREADY) + ret = 0; + + if (ret == -EDEADLK) + ww->contended = obj; + + return ret; } -static inline bool i915_gem_object_trylock(struct drm_i915_gem_object *obj) +static inline int i915_gem_object_lock(struct drm_i915_gem_object *obj, + struct i915_gem_ww_ctx *ww) { - return dma_resv_trylock(obj->base.resv); + return __i915_gem_object_lock(obj, ww, ww && ww->intr); } -static inline int -i915_gem_object_lock_interruptible(struct drm_i915_gem_object *obj) +static inline int i915_gem_object_lock_interruptible(struct drm_i915_gem_object *obj, + struct i915_gem_ww_ctx *ww) { - return dma_resv_lock_interruptible(obj->base.resv, NULL); + WARN_ON(ww && !ww->intr); + return __i915_gem_object_lock(obj, ww, true); +} + +static inline bool i915_gem_object_trylock(struct drm_i915_gem_object *obj) +{ + return dma_resv_trylock(obj->base.resv); } static inline void i915_gem_object_unlock(struct drm_i915_gem_object *obj) 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 d0847d7896f9..b5c15557cc87 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h @@ -123,6 +123,15 @@ struct drm_i915_gem_object { struct list_head lut_list; spinlock_t lut_lock; /* guards lut_list */ + /** + * @obj_link: Link into @i915_gem_ww_ctx.obj_list + * + * When we lock this object through i915_gem_object_lock() with a + * context, we add it to the list to ensure we can unlock everything + * when i915_gem_ww_ctx_backoff() or i915_gem_ww_ctx_fini() are called. + */ + struct list_head obj_link; + /** Stolen memory for this object, instead of being backed by shmem. */ struct drm_mm_node *stolen; union { diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c index 3d215164dd5a..40d3e40500fa 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c @@ -84,7 +84,7 @@ void i915_gem_suspend_late(struct drm_i915_private *i915) spin_unlock_irqrestore(&i915->mm.obj_lock, flags); - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); drm_WARN_ON(&i915->drm, i915_gem_object_set_to_gtt_domain(obj, false)); i915_gem_object_unlock(obj); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c index ff72ee2fd9cd..ffcaee74a249 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c @@ -249,7 +249,7 @@ i915_gem_object_set_tiling(struct drm_i915_gem_object *obj, * whilst executing a fenced command for an untiled object. */ - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); if (i915_gem_object_is_framebuffer(obj)) { i915_gem_object_unlock(obj); return -EBUSY; diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c index e2f3d014acb2..68c3631f3746 100644 --- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c @@ -947,7 +947,7 @@ static int gpu_write(struct intel_context *ce, { int err; - i915_gem_object_lock(vma->obj); + i915_gem_object_lock(vma->obj, NULL); err = i915_gem_object_set_to_gtt_domain(vma->obj, true); i915_gem_object_unlock(vma->obj); if (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 299c29e9ad86..4e36d4897ea6 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 @@ -75,7 +75,7 @@ static int __igt_client_fill(struct intel_engine_cs *engine) if (err) goto err_unpin; - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_cpu_domain(obj, false); i915_gem_object_unlock(obj); if (err) 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 87d7d8aa080f..1de2959b153c 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c @@ -82,7 +82,7 @@ static int gtt_set(struct context *ctx, unsigned long offset, u32 v) u32 __iomem *map; int err = 0; - i915_gem_object_lock(ctx->obj); + i915_gem_object_lock(ctx->obj, NULL); err = i915_gem_object_set_to_gtt_domain(ctx->obj, true); i915_gem_object_unlock(ctx->obj); if (err) @@ -115,7 +115,7 @@ static int gtt_get(struct context *ctx, unsigned long offset, u32 *v) u32 __iomem *map; int err = 0; - i915_gem_object_lock(ctx->obj); + i915_gem_object_lock(ctx->obj, NULL); err = i915_gem_object_set_to_gtt_domain(ctx->obj, false); i915_gem_object_unlock(ctx->obj); if (err) @@ -147,7 +147,7 @@ static int wc_set(struct context *ctx, unsigned long offset, u32 v) u32 *map; int err; - i915_gem_object_lock(ctx->obj); + i915_gem_object_lock(ctx->obj, NULL); err = i915_gem_object_set_to_wc_domain(ctx->obj, true); i915_gem_object_unlock(ctx->obj); if (err) @@ -170,7 +170,7 @@ static int wc_get(struct context *ctx, unsigned long offset, u32 *v) u32 *map; int err; - i915_gem_object_lock(ctx->obj); + i915_gem_object_lock(ctx->obj, NULL); err = i915_gem_object_set_to_wc_domain(ctx->obj, false); i915_gem_object_unlock(ctx->obj); if (err) @@ -193,7 +193,7 @@ static int gpu_set(struct context *ctx, unsigned long offset, u32 v) u32 *cs; int err; - i915_gem_object_lock(ctx->obj); + i915_gem_object_lock(ctx->obj, NULL); err = i915_gem_object_set_to_gtt_domain(ctx->obj, true); i915_gem_object_unlock(ctx->obj); if (err) 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 d176b015353f..52af9a52b743 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c @@ -950,7 +950,7 @@ emit_rpcs_query(struct drm_i915_gem_object *obj, if (IS_ERR(vma)) return PTR_ERR(vma); - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_gtt_domain(obj, false); i915_gem_object_unlock(obj); if (err) @@ -1709,7 +1709,7 @@ static int read_from_scratch(struct i915_gem_context *ctx, i915_request_add(rq); - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_cpu_domain(obj, false); i915_gem_object_unlock(obj); if (err) 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 9c7402ce5bf9..9fb95a45bcad 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c @@ -103,7 +103,7 @@ static int check_partial_mapping(struct drm_i915_gem_object *obj, 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); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_gtt_domain(obj, true); i915_gem_object_unlock(obj); if (err) { @@ -188,7 +188,7 @@ static int check_partial_mappings(struct drm_i915_gem_object *obj, 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); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_gtt_domain(obj, true); i915_gem_object_unlock(obj); if (err) { 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 34932871b3a5..a94243dc4c5c 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c @@ -44,7 +44,7 @@ static int mock_phys_object(void *arg) } /* Make the object dirty so that put_pages must do copy back the data */ - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_gtt_domain(obj, true); i915_gem_object_unlock(obj); if (err) { diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c index febc9e6692ba..61a0532d0f3d 100644 --- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c +++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c @@ -214,7 +214,7 @@ static int check_whitelist(struct i915_gem_context *ctx, return PTR_ERR(results); err = 0; - i915_gem_object_lock(results); + i915_gem_object_lock(results, NULL); 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); diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index f1940939260a..943c8d232703 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -2982,7 +2982,7 @@ static int shadow_indirect_ctx(struct intel_shadow_wa_ctx *wa_ctx) goto put_obj; } - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); ret = i915_gem_object_set_to_cpu_domain(obj, false); i915_gem_object_unlock(obj); if (ret) { diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index e1de50780ed5..4bc8fdb94019 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -420,7 +420,7 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj, GEM_BUG_ON(!drm_mm_node_allocated(&node)); } - ret = i915_gem_object_lock_interruptible(obj); + ret = i915_gem_object_lock_interruptible(obj, NULL); if (ret) goto out_unpin; @@ -619,7 +619,7 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj, GEM_BUG_ON(!drm_mm_node_allocated(&node)); } - ret = i915_gem_object_lock_interruptible(obj); + ret = i915_gem_object_lock_interruptible(obj, NULL); if (ret) goto out_unpin; @@ -1290,7 +1290,7 @@ int i915_gem_freeze_late(struct drm_i915_private *i915) i915_gem_drain_freed_objects(i915); list_for_each_entry(obj, &i915->mm.shrink_list, mm.link) { - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); drm_WARN_ON(&i915->drm, i915_gem_object_set_to_cpu_domain(obj, true)); i915_gem_object_unlock(obj); @@ -1326,6 +1326,52 @@ int i915_gem_open(struct drm_i915_private *i915, struct drm_file *file) return ret; } +void i915_gem_ww_ctx_init(struct i915_gem_ww_ctx *ww, bool intr) +{ + ww_acquire_init(&ww->ctx, &reservation_ww_class); + INIT_LIST_HEAD(&ww->obj_list); + ww->intr = intr; + ww->contended = NULL; +} + +static void i915_gem_ww_ctx_unlock_all(struct i915_gem_ww_ctx *ww) +{ + struct drm_i915_gem_object *obj; + + while ((obj = list_first_entry_or_null(&ww->obj_list, struct drm_i915_gem_object, obj_link))) { + list_del(&obj->obj_link); + i915_gem_object_unlock(obj); + } +} + +void i915_gem_ww_ctx_fini(struct i915_gem_ww_ctx *ww) +{ + i915_gem_ww_ctx_unlock_all(ww); + WARN_ON(ww->contended); + ww_acquire_fini(&ww->ctx); +} + +int __must_check i915_gem_ww_ctx_backoff(struct i915_gem_ww_ctx *ww) +{ + int ret = 0; + + if (WARN_ON(!ww->contended)) + return -EINVAL; + + i915_gem_ww_ctx_unlock_all(ww); + if (ww->intr) + ret = dma_resv_lock_slow_interruptible(ww->contended->base.resv, &ww->ctx); + else + dma_resv_lock_slow(ww->contended->base.resv, &ww->ctx); + + if (!ret) + list_add_tail(&ww->contended->obj_link, &ww->obj_list); + + ww->contended = NULL; + + return ret; +} + #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) #include "selftests/mock_gem_device.c" #include "selftests/i915_gem.c" diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h index f333e88a2b6e..427ebd02fceb 100644 --- a/drivers/gpu/drm/i915/i915_gem.h +++ b/drivers/gpu/drm/i915/i915_gem.h @@ -116,4 +116,15 @@ static inline bool __tasklet_is_scheduled(struct tasklet_struct *t) return test_bit(TASKLET_STATE_SCHED, &t->state); } +struct i915_gem_ww_ctx { + struct ww_acquire_ctx ctx; + struct list_head obj_list; + bool intr; + struct drm_i915_gem_object *contended; +}; + +void i915_gem_ww_ctx_init(struct i915_gem_ww_ctx *ctx, bool intr); +void i915_gem_ww_ctx_fini(struct i915_gem_ww_ctx *ctx); +int __must_check i915_gem_ww_ctx_backoff(struct i915_gem_ww_ctx *ctx); + #endif /* __I915_GEM_H__ */ diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c index 88d400b9df88..23a6132c5f4e 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem.c @@ -199,11 +199,52 @@ out: return err; } +static int igt_gem_ww_ctx(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct drm_i915_gem_object *obj, *obj2; + struct i915_gem_ww_ctx ww; + int err = 0; + + obj = i915_gem_object_create_internal(i915, PAGE_SIZE); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + obj2 = i915_gem_object_create_internal(i915, PAGE_SIZE); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + goto put1; + } + + i915_gem_ww_ctx_init(&ww, true); +retry: + /* Lock the objects, twice for good measure (-EALREADY handling) */ + err = i915_gem_object_lock(obj, &ww); + if (!err) + err = i915_gem_object_lock_interruptible(obj, &ww); + if (!err) + err = i915_gem_object_lock_interruptible(obj2, &ww); + if (!err) + err = i915_gem_object_lock(obj2, &ww); + + if (err == -EDEADLK) { + err = i915_gem_ww_ctx_backoff(&ww); + if (!err) + goto retry; + } + i915_gem_ww_ctx_fini(&ww); + i915_gem_object_put(obj2); +put1: + i915_gem_object_put(obj); + return err; +} + int i915_gem_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(igt_gem_suspend), SUBTEST(igt_gem_hibernate), + SUBTEST(igt_gem_ww_ctx), }; if (intel_gt_is_wedged(&i915->gt)) diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c index af89c7fc8f59..88c5e9acb84c 100644 --- a/drivers/gpu/drm/i915/selftests/i915_vma.c +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c @@ -892,7 +892,7 @@ static int igt_vma_remapped_gtt(void *arg) unsigned int x, y; int err; - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_gtt_domain(obj, true); i915_gem_object_unlock(obj); if (err) diff --git a/drivers/gpu/drm/i915/selftests/intel_memory_region.c b/drivers/gpu/drm/i915/selftests/intel_memory_region.c index 741270fcb6b9..93a38a323584 100644 --- a/drivers/gpu/drm/i915/selftests/intel_memory_region.c +++ b/drivers/gpu/drm/i915/selftests/intel_memory_region.c @@ -509,7 +509,7 @@ static int igt_lmem_write_cpu(void *arg) if (err) goto out_unpin; - i915_gem_object_lock(obj); + i915_gem_object_lock(obj, NULL); err = i915_gem_object_set_to_wc_domain(obj, true); i915_gem_object_unlock(obj); if (err) -- cgit v1.2.3 From 1af343cdc1971ff953f08f9dcf0efad464cf057f Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Wed, 19 Aug 2020 16:08:46 +0200 Subject: drm/i915: Remove locking from i915_gem_object_prepare_read/write Execbuffer submission will perform its own WW locking, and we cannot rely on the implicit lock there. This also makes it clear that the GVT code will get a lockdep splat when multiple batchbuffer shadows need to be performed in the same instance, fix that up. Signed-off-by: Maarten Lankhorst Reviewed-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20200819140904.1708856-7-maarten.lankhorst@linux.intel.com Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/gem/i915_gem_domain.c | 20 ++++++-------------- drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 13 +++++++++++-- drivers/gpu/drm/i915/gem/i915_gem_object.h | 1 - drivers/gpu/drm/i915/gem/selftests/huge_pages.c | 5 ++++- .../gpu/drm/i915/gem/selftests/i915_gem_coherency.c | 14 ++++++++++---- .../gpu/drm/i915/gem/selftests/i915_gem_context.c | 12 +++++++++--- drivers/gpu/drm/i915/gvt/cmd_parser.c | 1 + drivers/gpu/drm/i915/i915_gem.c | 20 ++++++++++++++++++-- 8 files changed, 59 insertions(+), 27 deletions(-) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c index c0acfc97fae3..8ebceebd11b0 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c @@ -576,19 +576,17 @@ int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj, if (!i915_gem_object_has_struct_page(obj)) return -ENODEV; - ret = i915_gem_object_lock_interruptible(obj, NULL); - if (ret) - return ret; + assert_object_held(obj); ret = i915_gem_object_wait(obj, I915_WAIT_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT); if (ret) - goto err_unlock; + return ret; ret = i915_gem_object_pin_pages(obj); if (ret) - goto err_unlock; + return ret; if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ || !static_cpu_has(X86_FEATURE_CLFLUSH)) { @@ -616,8 +614,6 @@ out: err_unpin: i915_gem_object_unpin_pages(obj); -err_unlock: - i915_gem_object_unlock(obj); return ret; } @@ -630,20 +626,18 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj, if (!i915_gem_object_has_struct_page(obj)) return -ENODEV; - ret = i915_gem_object_lock_interruptible(obj, NULL); - if (ret) - return ret; + assert_object_held(obj); ret = i915_gem_object_wait(obj, I915_WAIT_INTERRUPTIBLE | I915_WAIT_ALL, MAX_SCHEDULE_TIMEOUT); if (ret) - goto err_unlock; + return ret; ret = i915_gem_object_pin_pages(obj); if (ret) - goto err_unlock; + return ret; if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE || !static_cpu_has(X86_FEATURE_CLFLUSH)) { @@ -680,7 +674,5 @@ out: err_unpin: i915_gem_object_unpin_pages(obj); -err_unlock: - i915_gem_object_unlock(obj); return ret; } diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index 51395a0ceb15..ac2001f10b61 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -991,11 +991,14 @@ static void reloc_cache_reset(struct reloc_cache *cache) vaddr = unmask_page(cache->vaddr); if (cache->vaddr & KMAP) { + struct drm_i915_gem_object *obj = + (struct drm_i915_gem_object *)cache->node.mm; if (cache->vaddr & CLFLUSH_AFTER) mb(); kunmap_atomic(vaddr); - i915_gem_object_finish_access((struct drm_i915_gem_object *)cache->node.mm); + i915_gem_object_finish_access(obj); + i915_gem_object_unlock(obj); } else { struct i915_ggtt *ggtt = cache_to_ggtt(cache); @@ -1031,10 +1034,16 @@ static void *reloc_kmap(struct drm_i915_gem_object *obj, unsigned int flushes; int err; - err = i915_gem_object_prepare_write(obj, &flushes); + err = i915_gem_object_lock_interruptible(obj, NULL); if (err) return ERR_PTR(err); + err = i915_gem_object_prepare_write(obj, &flushes); + if (err) { + i915_gem_object_unlock(obj); + return ERR_PTR(err); + } + BUILD_BUG_ON(KMAP & CLFLUSH_FLAGS); BUILD_BUG_ON((KMAP | CLFLUSH_FLAGS) & PAGE_MASK); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h index 488459143366..45d79d75e73a 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h @@ -432,7 +432,6 @@ static inline void i915_gem_object_finish_access(struct drm_i915_gem_object *obj) { i915_gem_object_unpin_pages(obj); - i915_gem_object_unlock(obj); } static inline struct intel_engine_cs * diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c index 68c3631f3746..5daf4a2be422 100644 --- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c @@ -964,9 +964,10 @@ __cpu_check_shmem(struct drm_i915_gem_object *obj, u32 dword, u32 val) unsigned long n; int err; + i915_gem_object_lock(obj, NULL); err = i915_gem_object_prepare_read(obj, &needs_flush); if (err) - return err; + goto err_unlock; for (n = 0; n < obj->base.size >> PAGE_SHIFT; ++n) { u32 *ptr = kmap_atomic(i915_gem_object_get_page(obj, n)); @@ -986,6 +987,8 @@ __cpu_check_shmem(struct drm_i915_gem_object *obj, u32 dword, u32 val) } i915_gem_object_finish_access(obj); +err_unlock: + i915_gem_object_unlock(obj); return err; } 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 1de2959b153c..dcdfc396f2f8 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c @@ -27,9 +27,10 @@ static int cpu_set(struct context *ctx, unsigned long offset, u32 v) u32 *cpu; int err; + i915_gem_object_lock(ctx->obj, NULL); err = i915_gem_object_prepare_write(ctx->obj, &needs_clflush); if (err) - return err; + goto out; page = i915_gem_object_get_page(ctx->obj, offset >> PAGE_SHIFT); map = kmap_atomic(page); @@ -46,7 +47,9 @@ static int cpu_set(struct context *ctx, unsigned long offset, u32 v) kunmap_atomic(map); i915_gem_object_finish_access(ctx->obj); - return 0; +out: + i915_gem_object_unlock(ctx->obj); + return err; } static int cpu_get(struct context *ctx, unsigned long offset, u32 *v) @@ -57,9 +60,10 @@ static int cpu_get(struct context *ctx, unsigned long offset, u32 *v) u32 *cpu; int err; + i915_gem_object_lock(ctx->obj, NULL); err = i915_gem_object_prepare_read(ctx->obj, &needs_clflush); if (err) - return err; + goto out; page = i915_gem_object_get_page(ctx->obj, offset >> PAGE_SHIFT); map = kmap_atomic(page); @@ -73,7 +77,9 @@ static int cpu_get(struct context *ctx, unsigned long offset, u32 *v) kunmap_atomic(map); i915_gem_object_finish_access(ctx->obj); - return 0; +out: + i915_gem_object_unlock(ctx->obj); + return err; } static int gtt_set(struct context *ctx, unsigned long offset, u32 v) 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 52af9a52b743..9160a335c684 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c @@ -461,9 +461,10 @@ static int cpu_fill(struct drm_i915_gem_object *obj, u32 value) unsigned int n, m, need_flush; int err; + i915_gem_object_lock(obj, NULL); err = i915_gem_object_prepare_write(obj, &need_flush); if (err) - return err; + goto out; for (n = 0; n < real_page_count(obj); n++) { u32 *map; @@ -479,7 +480,9 @@ static int cpu_fill(struct drm_i915_gem_object *obj, u32 value) i915_gem_object_finish_access(obj); obj->read_domains = I915_GEM_DOMAIN_GTT | I915_GEM_DOMAIN_CPU; obj->write_domain = 0; - return 0; +out: + i915_gem_object_unlock(obj); + return err; } static noinline int cpu_check(struct drm_i915_gem_object *obj, @@ -488,9 +491,10 @@ static noinline int cpu_check(struct drm_i915_gem_object *obj, unsigned int n, m, needs_flush; int err; + i915_gem_object_lock(obj, NULL); err = i915_gem_object_prepare_read(obj, &needs_flush); if (err) - return err; + goto out_unlock; for (n = 0; n < real_page_count(obj); n++) { u32 *map; @@ -527,6 +531,8 @@ out_unmap: } i915_gem_object_finish_access(obj); +out_unlock: + i915_gem_object_unlock(obj); return err; } diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 943c8d232703..d0a599b51bfe 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -1923,6 +1923,7 @@ static int perform_bb_shadow(struct parser_exec_state *s) if (ret) goto err_unmap; + i915_gem_object_unlock(bb->obj); INIT_LIST_HEAD(&bb->list); list_add(&bb->list, &s->workload->shadow_bb); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 4bc8fdb94019..55158e487a7f 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -335,12 +335,20 @@ i915_gem_shmem_pread(struct drm_i915_gem_object *obj, u64 remain; int ret; - ret = i915_gem_object_prepare_read(obj, &needs_clflush); + ret = i915_gem_object_lock_interruptible(obj, NULL); if (ret) return ret; + ret = i915_gem_object_prepare_read(obj, &needs_clflush); + if (ret) { + i915_gem_object_unlock(obj); + return ret; + } + fence = i915_gem_object_lock_fence(obj); i915_gem_object_finish_access(obj); + i915_gem_object_unlock(obj); + if (!fence) return -ENOMEM; @@ -734,12 +742,20 @@ i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj, u64 remain; int ret; - ret = i915_gem_object_prepare_write(obj, &needs_clflush); + ret = i915_gem_object_lock_interruptible(obj, NULL); if (ret) return ret; + ret = i915_gem_object_prepare_write(obj, &needs_clflush); + if (ret) { + i915_gem_object_unlock(obj); + return ret; + } + fence = i915_gem_object_lock_fence(obj); i915_gem_object_finish_access(obj); + i915_gem_object_unlock(obj); + if (!fence) return -ENOMEM; -- cgit v1.2.3 From 56d44649a7c207752be8907b801dd698b3451cad Mon Sep 17 00:00:00 2001 From: Yan Zhao Date: Tue, 11 Aug 2020 14:09:44 +0800 Subject: drm/i915/gvt: rename F_IN_CTX flag to F_SR_IN_CTX F_IN_CTX is an inaccurate flag name, because people may wrongly think all MMIOs in context image are with this flag. But actually, this flag is only for MMIOs both in GVT's save-restore list and in hardare logical context's image. Reviewed-by: Zhenyu Wang Signed-off-by: Yan Zhao Signed-off-by: Zhenyu Wang Link: http://patchwork.freedesktop.org/patch/msgid/20200811060944.3039-1-yan.y.zhao@intel.com --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 2 +- drivers/gpu/drm/i915/gvt/gvt.h | 23 ++++++++++++++--------- drivers/gpu/drm/i915/gvt/mmio_context.c | 2 +- 3 files changed, 16 insertions(+), 11 deletions(-) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index f1940939260a..018509eeb0fe 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -976,7 +976,7 @@ static int cmd_reg_handler(struct parser_exec_state *s, * inhibit context will restore with correct values */ if (IS_GEN(s->engine->i915, 9) && - intel_gvt_mmio_is_in_ctx(gvt, offset) && + intel_gvt_mmio_is_sr_in_ctx(gvt, offset) && !strncmp(cmd, "lri", 3)) { intel_gvt_hypervisor_read_gpa(s->vgpu, s->workload->ring_context_gpa + 12, &ctx_sr_ctl, 4); diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index a4a6db6b7f90..899945b91cc6 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -256,8 +256,10 @@ struct intel_gvt_mmio { #define F_CMD_ACCESSED (1 << 5) /* This reg could be accessed by unaligned address */ #define F_UNALIGN (1 << 6) -/* This reg is saved/restored in context */ -#define F_IN_CTX (1 << 7) +/* This reg is in GVT's mmio save-restor list and in hardware + * logical context image + */ +#define F_SR_IN_CTX (1 << 7) struct gvt_mmio_block *mmio_block; unsigned int num_mmio_block; @@ -645,30 +647,33 @@ static inline bool intel_gvt_mmio_has_mode_mask( } /** - * intel_gvt_mmio_is_in_ctx - check if a MMIO has in-ctx mask + * intel_gvt_mmio_is_sr_in_ctx - + * check if an MMIO has F_SR_IN_CTX mask * @gvt: a GVT device * @offset: register offset * * Returns: - * True if a MMIO has a in-context mask, false if it isn't. + * True if an MMIO has an F_SR_IN_CTX mask, false if it isn't. * */ -static inline bool intel_gvt_mmio_is_in_ctx( +static inline bool intel_gvt_mmio_is_sr_in_ctx( struct intel_gvt *gvt, unsigned int offset) { - return gvt->mmio.mmio_attribute[offset >> 2] & F_IN_CTX; + return gvt->mmio.mmio_attribute[offset >> 2] & F_SR_IN_CTX; } /** - * intel_gvt_mmio_set_in_ctx - mask a MMIO in logical context + * intel_gvt_mmio_set_sr_in_ctx - + * mask an MMIO in GVT's mmio save-restore list and also + * in hardware logical context image * @gvt: a GVT device * @offset: register offset * */ -static inline void intel_gvt_mmio_set_in_ctx( +static inline void intel_gvt_mmio_set_sr_in_ctx( struct intel_gvt *gvt, unsigned int offset) { - gvt->mmio.mmio_attribute[offset >> 2] |= F_IN_CTX; + gvt->mmio.mmio_attribute[offset >> 2] |= F_SR_IN_CTX; } void intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu); diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c index 86a60bdf0818..afe574d6b3b5 100644 --- a/drivers/gpu/drm/i915/gvt/mmio_context.c +++ b/drivers/gpu/drm/i915/gvt/mmio_context.c @@ -595,7 +595,7 @@ void intel_gvt_init_engine_mmio_context(struct intel_gvt *gvt) i915_mmio_reg_valid(mmio->reg); mmio++) { if (mmio->in_context) { gvt->engine_mmio_list.ctx_mmio_count[mmio->id]++; - intel_gvt_mmio_set_in_ctx(gvt, mmio->reg.reg); + intel_gvt_mmio_set_sr_in_ctx(gvt, mmio->reg.reg); } } } -- cgit v1.2.3 From a6c5817a38cfaddb3b5b4b3f2bc4b3750f264c57 Mon Sep 17 00:00:00 2001 From: Yan Zhao Date: Tue, 11 Aug 2020 14:37:44 +0800 Subject: drm/i915/gvt: remove flag F_CMD_ACCESSED Flag F_CMD_ACCESSED is not used. just remove it. Reviewed-by: Zhenyu Wang Signed-off-by: Yan Zhao Signed-off-by: Zhenyu Wang Link: http://patchwork.freedesktop.org/patch/msgid/20200811063744.3272-1-yan.y.zhao@intel.com --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 2 -- drivers/gpu/drm/i915/gvt/gvt.h | 14 -------------- 2 files changed, 16 deletions(-) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 018509eeb0fe..0292e3c1deb4 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -992,8 +992,6 @@ static int cmd_reg_handler(struct parser_exec_state *s, } } - /* TODO: Update the global mask if this MMIO is a masked-MMIO */ - intel_gvt_mmio_set_cmd_accessed(gvt, offset); return 0; } diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 899945b91cc6..8fbe65471c01 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -253,8 +253,6 @@ struct intel_gvt_mmio { /* This reg has been accessed by a VM */ #define F_ACCESSED (1 << 4) /* This reg has been accessed through GPU commands */ -#define F_CMD_ACCESSED (1 << 5) -/* This reg could be accessed by unaligned address */ #define F_UNALIGN (1 << 6) /* This reg is in GVT's mmio save-restor list and in hardware * logical context image @@ -619,18 +617,6 @@ static inline bool intel_gvt_mmio_is_unalign( return gvt->mmio.mmio_attribute[offset >> 2] & F_UNALIGN; } -/** - * intel_gvt_mmio_set_cmd_accessed - mark a MMIO has been accessed by command - * @gvt: a GVT device - * @offset: register offset - * - */ -static inline void intel_gvt_mmio_set_cmd_accessed( - struct intel_gvt *gvt, unsigned int offset) -{ - gvt->mmio.mmio_attribute[offset >> 2] |= F_CMD_ACCESSED; -} - /** * intel_gvt_mmio_has_mode_mask - if a MMIO has a mode mask * @gvt: a GVT device -- cgit v1.2.3 From 7e93a0806f5a8b42c0a845a02816b6f80d406f79 Mon Sep 17 00:00:00 2001 From: Yan Zhao Date: Tue, 11 Aug 2020 15:02:33 +0800 Subject: drm/i915/gvt: add/modify interfaces for flag F_CMD_ACCESS flag F_CMD_ACCESS represents whether an MMIO is able to be accessed by GPU commands. In this patch, 1. add interface to set this flag 2. rename intel_gvt_mmio_is_cmd_access() to intel_gvt_mmio_is_cmd_accessible() and update its description message. Reviewed-by: Zhenyu Wang Signed-off-by: Yan Zhao Signed-off-by: Zhenyu Wang Link: http://patchwork.freedesktop.org/patch/msgid/20200811070233.3387-1-yan.y.zhao@intel.com --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 2 +- drivers/gpu/drm/i915/gvt/gvt.h | 19 +++++++++++++++++-- 2 files changed, 18 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 0292e3c1deb4..affd05d2c201 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -936,7 +936,7 @@ static int cmd_reg_handler(struct parser_exec_state *s, return -EFAULT; } - if (!intel_gvt_mmio_is_cmd_access(gvt, offset)) { + if (!intel_gvt_mmio_is_cmd_accessible(gvt, offset)) { gvt_vgpu_err("%s access to non-render register (%x)\n", cmd, offset); return -EBADRQC; diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 8fbe65471c01..d7279cd654ca 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -594,17 +594,32 @@ static inline void intel_gvt_mmio_set_accessed( } /** - * intel_gvt_mmio_is_cmd_accessed - mark a MMIO could be accessed by command + * intel_gvt_mmio_is_cmd_accessible - if a MMIO could be accessed by command * @gvt: a GVT device * @offset: register offset * + * Returns: + * True if an MMIO is able to be accessed by GPU commands */ -static inline bool intel_gvt_mmio_is_cmd_access( +static inline bool intel_gvt_mmio_is_cmd_accessible( struct intel_gvt *gvt, unsigned int offset) { return gvt->mmio.mmio_attribute[offset >> 2] & F_CMD_ACCESS; } +/** + * intel_gvt_mmio_set_cmd_accessible - + * mark a MMIO could be accessible by command + * @gvt: a GVT device + * @offset: register offset + * + */ +static inline void intel_gvt_mmio_set_cmd_accessible( + struct intel_gvt *gvt, unsigned int offset) +{ + gvt->mmio.mmio_attribute[offset >> 2] |= F_CMD_ACCESS; +} + /** * intel_gvt_mmio_is_unalign - mark a MMIO could be accessed unaligned * @gvt: a GVT device -- cgit v1.2.3 From b2feabc6ebd39976b0085afd8205734dc3a9793b Mon Sep 17 00:00:00 2001 From: Yan Zhao Date: Tue, 11 Aug 2020 15:27:20 +0800 Subject: drm/i915/gvt: remove F_CMD_ACCESS flag for some registers some registers cannot be cmd accessible. remove them from the list Reviewed-by: Zhenyu Wang Signed-off-by: Wang Zhi Signed-off-by: Yan Zhao Signed-off-by: Zhenyu Wang Link: http://patchwork.freedesktop.org/patch/msgid/20200811072720.3525-1-yan.y.zhao@intel.com --- drivers/gpu/drm/i915/gvt/handlers.c | 28 ++++++++++++++++------------ 1 file changed, 16 insertions(+), 12 deletions(-) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 63bba7b4bb2f..d55cb43c9bd4 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1892,7 +1892,7 @@ static int init_generic_mmio_info(struct intel_gvt *gvt) struct drm_i915_private *dev_priv = gvt->gt->i915; int ret; - MMIO_RING_DFH(RING_IMR, D_ALL, F_CMD_ACCESS, NULL, + MMIO_RING_DFH(RING_IMR, D_ALL, 0, NULL, intel_vgpu_reg_imr_handler); MMIO_DFH(SDEIMR, D_ALL, 0, NULL, intel_vgpu_reg_imr_handler); @@ -1900,7 +1900,8 @@ static int init_generic_mmio_info(struct intel_gvt *gvt) MMIO_DFH(SDEIIR, D_ALL, 0, NULL, intel_vgpu_reg_iir_handler); MMIO_D(SDEISR, D_ALL); - MMIO_RING_DFH(RING_HWSTAM, D_ALL, F_CMD_ACCESS, NULL, NULL); + MMIO_RING_DFH(RING_HWSTAM, D_ALL, 0, NULL, NULL); + MMIO_DH(GEN8_GAMW_ECO_DEV_RW_IA, D_BDW_PLUS, NULL, gamw_echo_dev_rw_ia_write); @@ -1927,11 +1928,11 @@ static int init_generic_mmio_info(struct intel_gvt *gvt) MMIO_GM_RDR(_MMIO(0x12198), D_ALL, NULL, NULL); MMIO_D(GEN7_CXT_SIZE, D_ALL); - MMIO_RING_DFH(RING_TAIL, D_ALL, F_CMD_ACCESS, NULL, NULL); - MMIO_RING_DFH(RING_HEAD, D_ALL, F_CMD_ACCESS, NULL, NULL); - MMIO_RING_DFH(RING_CTL, D_ALL, F_CMD_ACCESS, NULL, NULL); - MMIO_RING_DFH(RING_ACTHD, D_ALL, F_CMD_ACCESS, mmio_read_from_hw, NULL); - MMIO_RING_GM_RDR(RING_START, D_ALL, NULL, NULL); + MMIO_RING_DFH(RING_TAIL, D_ALL, 0, NULL, NULL); + MMIO_RING_DFH(RING_HEAD, D_ALL, 0, NULL, NULL); + MMIO_RING_DFH(RING_CTL, D_ALL, 0, NULL, NULL); + MMIO_RING_DFH(RING_ACTHD, D_ALL, 0, mmio_read_from_hw, NULL); + MMIO_RING_GM(RING_START, D_ALL, NULL, NULL); /* RING MODE */ #define RING_REG(base) _MMIO((base) + 0x29c) @@ -2686,7 +2687,7 @@ static int init_generic_mmio_info(struct intel_gvt *gvt) MMIO_DFH(_MMIO(0x4094), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL); MMIO_DFH(ARB_MODE, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL); - MMIO_RING_GM_RDR(RING_BBADDR, D_ALL, NULL, NULL); + MMIO_RING_GM(RING_BBADDR, D_ALL, NULL, NULL); MMIO_DFH(_MMIO(0x2220), D_ALL, F_CMD_ACCESS, NULL, NULL); MMIO_DFH(_MMIO(0x12220), D_ALL, F_CMD_ACCESS, NULL, NULL); MMIO_DFH(_MMIO(0x22220), D_ALL, F_CMD_ACCESS, NULL, NULL); @@ -2771,7 +2772,7 @@ static int init_bdw_mmio_info(struct intel_gvt *gvt) MMIO_DH(GEN8_MASTER_IRQ, D_BDW_PLUS, NULL, intel_vgpu_reg_master_irq_handler); - MMIO_RING_DFH(RING_ACTHD_UDW, D_BDW_PLUS, F_CMD_ACCESS, + MMIO_RING_DFH(RING_ACTHD_UDW, D_BDW_PLUS, 0, mmio_read_from_hw, NULL); #define RING_REG(base) _MMIO((base) + 0xd0) @@ -2785,7 +2786,7 @@ static int init_bdw_mmio_info(struct intel_gvt *gvt) #undef RING_REG #define RING_REG(base) _MMIO((base) + 0x234) - MMIO_RING_F(RING_REG, 8, F_RO | F_CMD_ACCESS, 0, ~0, D_BDW_PLUS, + MMIO_RING_F(RING_REG, 8, F_RO, 0, ~0, D_BDW_PLUS, NULL, NULL); #undef RING_REG @@ -2820,7 +2821,7 @@ static int init_bdw_mmio_info(struct intel_gvt *gvt) MMIO_RING_F(RING_REG, 32, F_CMD_ACCESS, 0, 0, D_BDW_PLUS, NULL, NULL); #undef RING_REG - MMIO_RING_GM_RDR(RING_HWS_PGA, D_BDW_PLUS, NULL, hws_pga_write); + MMIO_RING_GM(RING_HWS_PGA, D_BDW_PLUS, NULL, hws_pga_write); MMIO_DFH(HDC_CHICKEN0, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL); @@ -3357,7 +3358,10 @@ void intel_gvt_clean_mmio_info(struct intel_gvt *gvt) gvt->mmio.mmio_attribute = NULL; } -/* Special MMIO blocks. */ +/* Special MMIO blocks. registers in MMIO block ranges should not be command + * accessible (should have no F_CMD_ACCESS flag). + * otherwise, need to update cmd_reg_handler in cmd_parser.c + */ static struct gvt_mmio_block mmio_blocks[] = { {D_SKL_PLUS, _MMIO(CSR_MMIO_START_RANGE), 0x3000, NULL, NULL}, {D_ALL, _MMIO(MCHBAR_MIRROR_BASE_SNB), 0x40000, NULL, NULL}, -- cgit v1.2.3 From d0a011094a0c1ba93dbf12a80b69d788b9a5437b Mon Sep 17 00:00:00 2001 From: Colin Xu Date: Wed, 19 Aug 2020 09:08:01 +0800 Subject: drm/i915/gvt: Add F_CMD_ACCESS for some GEN9 SKU WA MMIO access Without F_CMD_ACCESS, guest LRI cmd will fail due to "access to non-render register" when init below WAs: WaDisableDynamicCreditSharing: GAMT_CHKN_BIT_REG WaCompressedResourceSamplerPbeMediaNewHashMode: MMCD_MISC_CTRL So add F_CMD_ACCESS to the two MMIO. Reviewed-by: Zhenyu Wang Signed-off-by: Colin Xu Signed-off-by: Zhenyu Wang Link: http://patchwork.freedesktop.org/patch/msgid/20200819010801.53411-1-colin.xu@intel.com --- drivers/gpu/drm/i915/gvt/handlers.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index d55cb43c9bd4..840572add2d4 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -2922,7 +2922,7 @@ static int init_skl_mmio_info(struct intel_gvt *gvt) MMIO_D(GEN9_MEDIA_PG_IDLE_HYSTERESIS, D_SKL_PLUS); MMIO_D(GEN9_RENDER_PG_IDLE_HYSTERESIS, D_SKL_PLUS); MMIO_DFH(GEN9_GAMT_ECO_REG_RW_IA, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL); - MMIO_DH(MMCD_MISC_CTRL, D_SKL_PLUS, NULL, NULL); + MMIO_DFH(MMCD_MISC_CTRL, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL); MMIO_DH(CHICKEN_PAR1_1, D_SKL_PLUS, NULL, NULL); MMIO_D(DC_STATE_EN, D_SKL_PLUS); MMIO_D(DC_STATE_DEBUG, D_SKL_PLUS); @@ -3138,7 +3138,7 @@ static int init_skl_mmio_info(struct intel_gvt *gvt) MMIO_DFH(GEN9_WM_CHICKEN3, D_SKL_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL); - MMIO_D(GAMT_CHKN_BIT_REG, D_KBL | D_CFL); + MMIO_DFH(GAMT_CHKN_BIT_REG, D_KBL | D_CFL, F_CMD_ACCESS, NULL, NULL); MMIO_D(GEN9_CTX_PREEMPT_REG, D_SKL_PLUS); return 0; -- cgit v1.2.3 From df398e33b8fd3ac28b3c7166de555e38d26e7391 Mon Sep 17 00:00:00 2001 From: Colin Xu Date: Wed, 19 Aug 2020 09:09:00 +0800 Subject: drm/i915/gvt: Init vreg GUC_STATUS to GS_MIA_IN_RESET Although GVT doesn't support guest GuC, MIA core is still expected to be GS_MIA_IN_RESET after uc HW reset. Reviewed-by: Zhenyu Wang Signed-off-by: Colin Xu Signed-off-by: Zhenyu Wang Link: http://patchwork.freedesktop.org/patch/msgid/20200819010900.54598-1-colin.xu@intel.com --- drivers/gpu/drm/i915/gvt/mmio.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/i915/gvt') diff --git a/drivers/gpu/drm/i915/gvt/mmio.c b/drivers/gpu/drm/i915/gvt/mmio.c index 291993615af9..b6811f6a230d 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.c +++ b/drivers/gpu/drm/i915/gvt/mmio.c @@ -251,6 +251,9 @@ void intel_vgpu_reset_mmio(struct intel_vgpu *vgpu, bool dmlr) /* set the bit 0:2(Core C-State ) to C0 */ vgpu_vreg_t(vgpu, GEN6_GT_CORE_STATUS) = 0; + /* uc reset hw expect GS_MIA_IN_RESET */ + vgpu_vreg_t(vgpu, GUC_STATUS) |= GS_MIA_IN_RESET; + if (IS_BROXTON(vgpu->gvt->gt->i915)) { vgpu_vreg_t(vgpu, BXT_P_CR_GT_DISP_PWRON) &= ~(BIT(0) | BIT(1)); -- cgit v1.2.3