From d1f6dc1a9a106a73510181cfad9b4a7a0b140990 Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Thu, 19 Oct 2017 14:29:46 -0400 Subject: drm/amdgpu: Avoid accessing job->entity after the job is scheduled. Bug: amdgpu_job_free_cb was accessing s_job->s_entity when the allocated amdgpu_ctx (and the entity inside it) were already deallocated from amdgpu_cs_parser_fini. Fix: Save job's priority on it's creation instead of accessing it from s_entity later on. Signed-off-by: Andrey Grodzovsky Reviewed-by: Andres Rodriguez Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 57abf7abd7a9..b4b3ecf9f96d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1194,8 +1194,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, job->uf_sequence = seq; amdgpu_job_free_resources(job); - amdgpu_ring_priority_get(job->ring, - amd_sched_get_job_priority(&job->base)); + amdgpu_ring_priority_get(job->ring, job->base.s_priority); trace_amdgpu_cs_ioctl(job); amd_sched_entity_push_job(&job->base); -- cgit From 7716ea564f6538236c34749cdf3a6351c2844c1a Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Tue, 17 Oct 2017 12:08:02 +0800 Subject: drm/amdgpu:skip job for guilty ctx in parser_init MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Monk Liu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index b4b3ecf9f96d..16947bad5b49 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -90,6 +90,12 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) goto free_chunk; } + /* skip guilty context job */ + if (atomic_read(&p->ctx->guilty) == 1) { + ret = -ECANCELED; + goto free_chunk; + } + mutex_lock(&p->ctx->lock); /* get chunks */ -- cgit From a4176cb484ac457a08b44c93da06fce09c6e281c Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Tue, 24 Oct 2017 13:30:16 -0400 Subject: drm/amdgpu: Remove job->s_entity to avoid keeping reference to stale pointer. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Andrey Grodzovsky Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 7 ++++--- drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h | 9 ++++----- drivers/gpu/drm/amd/scheduler/gpu_scheduler.c | 19 +++++++++---------- drivers/gpu/drm/amd/scheduler/gpu_scheduler.h | 7 ++++--- 5 files changed, 22 insertions(+), 22 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 16947bad5b49..bf1aad00bb8d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1203,7 +1203,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, amdgpu_ring_priority_get(job->ring, job->base.s_priority); trace_amdgpu_cs_ioctl(job); - amd_sched_entity_push_job(&job->base); + amd_sched_entity_push_job(&job->base, entity); ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence); amdgpu_mn_unlock(p->mn); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index a58e3c5dd84b..f60662e03761 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -142,12 +142,13 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring, *f = dma_fence_get(&job->base.s_fence->finished); amdgpu_job_free_resources(job); amdgpu_ring_priority_get(job->ring, job->base.s_priority); - amd_sched_entity_push_job(&job->base); + amd_sched_entity_push_job(&job->base, entity); return 0; } -static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job) +static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job, + struct amd_sched_entity *s_entity) { struct amdgpu_job *job = to_amdgpu_job(sched_job); struct amdgpu_vm *vm = job->vm; @@ -155,7 +156,7 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job) struct dma_fence *fence = amdgpu_sync_get_fence(&job->dep_sync); int r; - if (amd_sched_dependency_optimized(fence, sched_job->s_entity)) { + if (amd_sched_dependency_optimized(fence, s_entity)) { r = amdgpu_sync_fence(job->adev, &job->sched_sync, fence); if (r) DRM_ERROR("Error adding fence to sync (%d)\n", r); diff --git a/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h b/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h index 705380eb693c..eebe323c7159 100644 --- a/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h +++ b/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h @@ -13,8 +13,8 @@ #define TRACE_INCLUDE_FILE gpu_sched_trace TRACE_EVENT(amd_sched_job, - TP_PROTO(struct amd_sched_job *sched_job), - TP_ARGS(sched_job), + TP_PROTO(struct amd_sched_job *sched_job, struct amd_sched_entity *entity), + TP_ARGS(sched_job, entity), TP_STRUCT__entry( __field(struct amd_sched_entity *, entity) __field(struct dma_fence *, fence) @@ -25,12 +25,11 @@ TRACE_EVENT(amd_sched_job, ), TP_fast_assign( - __entry->entity = sched_job->s_entity; + __entry->entity = entity; __entry->id = sched_job->id; __entry->fence = &sched_job->s_fence->finished; __entry->name = sched_job->sched->name; - __entry->job_count = spsc_queue_count( - &sched_job->s_entity->job_queue); + __entry->job_count = spsc_queue_count(&entity->job_queue); __entry->hw_job_count = atomic_read( &sched_job->sched->hw_rq_count); ), diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c index 1a2267ce62a8..f116de798204 100644 --- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c +++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c @@ -341,11 +341,10 @@ amd_sched_entity_pop_job(struct amd_sched_entity *entity) if (!sched_job) return NULL; - while ((entity->dependency = sched->ops->dependency(sched_job))) + while ((entity->dependency = sched->ops->dependency(sched_job, entity))) if (amd_sched_entity_add_dependency_cb(entity)) return NULL; - sched_job->s_entity = NULL; spsc_queue_pop(&entity->job_queue); return sched_job; } @@ -357,13 +356,13 @@ amd_sched_entity_pop_job(struct amd_sched_entity *entity) * * Returns 0 for success, negative error code otherwise. */ -void amd_sched_entity_push_job(struct amd_sched_job *sched_job) +void amd_sched_entity_push_job(struct amd_sched_job *sched_job, + struct amd_sched_entity *entity) { struct amd_gpu_scheduler *sched = sched_job->sched; - struct amd_sched_entity *entity = sched_job->s_entity; bool first = false; - trace_amd_sched_job(sched_job); + trace_amd_sched_job(sched_job, entity); spin_lock(&entity->queue_lock); first = spsc_queue_push(&entity->job_queue, &sched_job->queue_node); @@ -442,11 +441,12 @@ static void amd_sched_job_timedout(struct work_struct *work) job->sched->ops->timedout_job(job); } -static void amd_sched_set_guilty(struct amd_sched_job *s_job) +static void amd_sched_set_guilty(struct amd_sched_job *s_job, + struct amd_sched_entity *s_entity) { if (atomic_inc_return(&s_job->karma) > s_job->sched->hang_limit) - if (s_job->s_entity->guilty) - atomic_set(s_job->s_entity->guilty, 1); + if (s_entity->guilty) + atomic_set(s_entity->guilty, 1); } void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched, struct amd_sched_job *bad) @@ -477,7 +477,7 @@ void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched, struct amd_sched_jo list_for_each_entry_safe(entity, tmp, &rq->entities, list) { if (bad->s_fence->scheduled.context == entity->fence_context) { found = true; - amd_sched_set_guilty(bad); + amd_sched_set_guilty(bad, entity); break; } } @@ -541,7 +541,6 @@ int amd_sched_job_init(struct amd_sched_job *job, void *owner) { job->sched = sched; - job->s_entity = entity; job->s_priority = entity->rq - sched->sched_rq; job->s_fence = amd_sched_fence_create(entity, owner); if (!job->s_fence) diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h index f9e3a83cddc6..b590fcc2786a 100644 --- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h +++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h @@ -91,7 +91,6 @@ struct amd_sched_fence { struct amd_sched_job { struct spsc_node queue_node; struct amd_gpu_scheduler *sched; - struct amd_sched_entity *s_entity; struct amd_sched_fence *s_fence; struct dma_fence_cb finish_cb; struct work_struct finish_work; @@ -125,7 +124,8 @@ static inline bool amd_sched_invalidate_job(struct amd_sched_job *s_job, int thr * these functions should be implemented in driver side */ struct amd_sched_backend_ops { - struct dma_fence *(*dependency)(struct amd_sched_job *sched_job); + struct dma_fence *(*dependency)(struct amd_sched_job *sched_job, + struct amd_sched_entity *s_entity); struct dma_fence *(*run_job)(struct amd_sched_job *sched_job); void (*timedout_job)(struct amd_sched_job *sched_job); void (*free_job)(struct amd_sched_job *sched_job); @@ -161,7 +161,8 @@ int amd_sched_entity_init(struct amd_gpu_scheduler *sched, uint32_t jobs, atomic_t* guilty); void amd_sched_entity_fini(struct amd_gpu_scheduler *sched, struct amd_sched_entity *entity); -void amd_sched_entity_push_job(struct amd_sched_job *sched_job); +void amd_sched_entity_push_job(struct amd_sched_job *sched_job, + struct amd_sched_entity *entity); void amd_sched_entity_set_rq(struct amd_sched_entity *entity, struct amd_sched_rq *rq); -- cgit From 4ff23be3d50a142ed0787fa6f276afc02ac2577d Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 16 Oct 2017 17:24:21 +0200 Subject: drm/amdgpu: remove extra parameter from amdgpu_ttm_bind() v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We always use the BO mem now. v2: minor rebase Signed-off-by: Christian König Reviewed-by: Michel Dänzer Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h | 2 +- 4 files changed, 6 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index bf1aad00bb8d..43df84c13f21 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -684,7 +684,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, if (!r && p->uf_entry.robj) { struct amdgpu_bo *uf = p->uf_entry.robj; - r = amdgpu_ttm_bind(&uf->tbo, &uf->tbo.mem); + r = amdgpu_ttm_bind(&uf->tbo); p->job->uf_addr += amdgpu_bo_gpu_offset(uf); } @@ -1601,5 +1601,5 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, return r; } - return amdgpu_ttm_bind(&(*bo)->tbo, &(*bo)->tbo.mem); + return amdgpu_ttm_bind(&(*bo)->tbo); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 6d4b22fdc1b4..589aa5099881 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -688,7 +688,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, goto error; } - r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem); + r = amdgpu_ttm_bind(&bo->tbo); if (unlikely(r)) { dev_err(adev->dev, "%p bind failed\n", bo); goto error; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 470b911200ae..6a724f9f1b86 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -889,7 +889,7 @@ bool amdgpu_ttm_is_bound(struct ttm_tt *ttm) return gtt && !list_empty(>t->list); } -int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem) +int amdgpu_ttm_bind(struct ttm_buffer_object *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct ttm_tt *ttm = bo->ttm; @@ -1628,7 +1628,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, } if (bo->tbo.mem.mem_type == TTM_PL_TT) { - r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem); + r = amdgpu_ttm_bind(&bo->tbo); if (r) return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index abd4084982a3..0108a59559cd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -91,7 +91,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma); bool amdgpu_ttm_is_bound(struct ttm_tt *ttm); -int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem); +int amdgpu_ttm_bind(struct ttm_buffer_object *bo); int amdgpu_ttm_recover_gart(struct amdgpu_device *adev); int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages); -- cgit From c5835bbb1195996c6c5bf2bf2cdbae08c936fc93 Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 27 Oct 2017 15:43:14 +0200 Subject: drm/amdgpu: rename amdgpu_ttm_bind to amdgpu_ttm_alloc_gart MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We actually don't bind here, but rather allocate GART space if necessary. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h | 2 +- 4 files changed, 6 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 43df84c13f21..bc3a7500fbbd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -684,7 +684,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, if (!r && p->uf_entry.robj) { struct amdgpu_bo *uf = p->uf_entry.robj; - r = amdgpu_ttm_bind(&uf->tbo); + r = amdgpu_ttm_alloc_gart(&uf->tbo); p->job->uf_addr += amdgpu_bo_gpu_offset(uf); } @@ -1601,5 +1601,5 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, return r; } - return amdgpu_ttm_bind(&(*bo)->tbo); + return amdgpu_ttm_alloc_gart(&(*bo)->tbo); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index ff6f842655d1..a937c49590a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -688,7 +688,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, goto error; } - r = amdgpu_ttm_bind(&bo->tbo); + r = amdgpu_ttm_alloc_gart(&bo->tbo); if (unlikely(r)) { dev_err(adev->dev, "%p bind failed\n", bo); goto error; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 34dbe7afb600..793f41dd4ee4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -875,7 +875,7 @@ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm, return r; } -int amdgpu_ttm_bind(struct ttm_buffer_object *bo) +int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct ttm_mem_reg tmp; @@ -1601,7 +1601,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, } if (bo->tbo.mem.mem_type == TTM_PL_TT) { - r = amdgpu_ttm_bind(&bo->tbo); + r = amdgpu_ttm_alloc_gart(&bo->tbo); if (r) return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index d2985def4168..4f9433e61406 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -91,7 +91,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, struct dma_fence **fence); int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma); -int amdgpu_ttm_bind(struct ttm_buffer_object *bo); +int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo); int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo); int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages); -- cgit From 6f16b4fb60011cbc7d4530e112739ea4416c6ea6 Mon Sep 17 00:00:00 2001 From: Chunming Zhou Date: Mon, 13 Nov 2017 10:51:57 +0800 Subject: drm/amdgpu: use dep_sync for CS dependency/syncobj MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Otherwise, they could be optimized by scheduled fence. Signed-off-by: Chunming Zhou Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index bc3a7500fbbd..743875ad4404 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1039,7 +1039,7 @@ static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p, amdgpu_ctx_put(ctx); return r; } else if (fence) { - r = amdgpu_sync_fence(p->adev, &p->job->sync, + r = amdgpu_sync_fence(p->adev, &p->job->dep_sync, fence); dma_fence_put(fence); amdgpu_ctx_put(ctx); @@ -1059,7 +1059,7 @@ static int amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p, if (r) return r; - r = amdgpu_sync_fence(p->adev, &p->job->sync, fence); + r = amdgpu_sync_fence(p->adev, &p->job->dep_sync, fence); dma_fence_put(fence); return r; -- cgit From 19be5570107108fba772bc2e3a1eb22ec32fb021 Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 12 Apr 2017 14:24:39 +0200 Subject: drm/ttm: add operation ctx to ttm_bo_validate v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Give moving a BO into place an operation context to work with. v2: rebased Signed-off-by: Christian König Reviewed-by: Michel Dänzer Reviewed-by: Chunming Zhou Tested-by: Dieter Nützel Tested-by: Michel Dänzer Acked-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 14 ++++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 3 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 12 ++++++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 6 ++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 3 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 3 ++- drivers/gpu/drm/ast/ast_ttm.c | 9 ++++++--- drivers/gpu/drm/bochs/bochs_mm.c | 6 ++++-- drivers/gpu/drm/cirrus/cirrus_ttm.c | 6 ++++-- drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c | 6 ++++-- drivers/gpu/drm/mgag200/mgag200_ttm.c | 9 ++++++--- drivers/gpu/drm/nouveau/nouveau_bo.c | 4 ++-- drivers/gpu/drm/qxl/qxl_ioctl.c | 4 ++-- drivers/gpu/drm/qxl/qxl_object.c | 6 ++++-- drivers/gpu/drm/qxl/qxl_release.c | 4 ++-- drivers/gpu/drm/radeon/radeon_gem.c | 3 ++- drivers/gpu/drm/radeon/radeon_mn.c | 3 ++- drivers/gpu/drm/radeon/radeon_object.c | 14 +++++++++----- drivers/gpu/drm/radeon/radeon_vm.c | 3 ++- drivers/gpu/drm/ttm/ttm_bo.c | 16 +++++++++------- drivers/gpu/drm/virtio/virtgpu_ioctl.c | 11 ++++++----- drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c | 3 ++- drivers/gpu/drm/vmwgfx/vmwgfx_dmabuf.c | 21 +++++++++++++-------- drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c | 9 ++++----- drivers/gpu/drm/vmwgfx/vmwgfx_resource.c | 6 ++++-- drivers/gpu/drm/vmwgfx/vmwgfx_shader.c | 3 ++- include/drm/ttm/ttm_bo_api.h | 20 ++++++++++++++++---- 27 files changed, 131 insertions(+), 76 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 743875ad4404..faab662ce680 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -343,6 +343,7 @@ static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + struct ttm_operation_ctx ctx = { true, false }; u64 initial_bytes_moved, bytes_moved; uint32_t domain; int r; @@ -374,7 +375,7 @@ static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, retry: amdgpu_ttm_placement_from_domain(bo, domain); initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); bytes_moved = atomic64_read(&adev->num_bytes_moved) - initial_bytes_moved; p->bytes_moved += bytes_moved; @@ -396,6 +397,7 @@ static bool amdgpu_cs_try_evict(struct amdgpu_cs_parser *p, struct amdgpu_bo *validated) { uint32_t domain = validated->allowed_domains; + struct ttm_operation_ctx ctx = { true, false }; int r; if (!p->evictable) @@ -437,7 +439,7 @@ static bool amdgpu_cs_try_evict(struct amdgpu_cs_parser *p, bo->tbo.mem.mem_type == TTM_PL_VRAM && bo->tbo.mem.start < adev->mc.visible_vram_size >> PAGE_SHIFT; initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); bytes_moved = atomic64_read(&adev->num_bytes_moved) - initial_bytes_moved; p->bytes_moved += bytes_moved; @@ -476,6 +478,7 @@ static int amdgpu_cs_validate(void *param, struct amdgpu_bo *bo) static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, struct list_head *validated) { + struct ttm_operation_ctx ctx = { true, false }; struct amdgpu_bo_list_entry *lobj; int r; @@ -493,8 +496,7 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, lobj->user_pages) { amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, - false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) return r; amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, @@ -1575,6 +1577,7 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, struct amdgpu_bo_va_mapping **map) { struct amdgpu_fpriv *fpriv = parser->filp->driver_priv; + struct ttm_operation_ctx ctx = { false, false }; struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_va_mapping *mapping; int r; @@ -1595,8 +1598,7 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, if (!((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)) { (*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; amdgpu_ttm_placement_from_domain(*bo, (*bo)->allowed_domains); - r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, false, - false); + r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, &ctx); if (r) return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 3ad4cf0f22f8..c16579287aee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -282,6 +282,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { + struct ttm_operation_ctx ctx = { true, false }; struct amdgpu_device *adev = dev->dev_private; struct drm_amdgpu_gem_userptr *args = data; struct drm_gem_object *gobj; @@ -335,7 +336,7 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, goto free_pages; amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT); - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); amdgpu_bo_unreserve(bo); if (r) goto free_pages; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 3233d5988f66..c2419bc6b3df 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -552,6 +552,7 @@ err: int amdgpu_bo_validate(struct amdgpu_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; uint32_t domain; int r; @@ -562,7 +563,7 @@ int amdgpu_bo_validate(struct amdgpu_bo *bo) retry: amdgpu_ttm_placement_from_domain(bo, domain); - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) { domain = bo->allowed_domains; goto retry; @@ -673,6 +674,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, u64 *gpu_addr) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + struct ttm_operation_ctx ctx = { false, false }; int r, i; if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) @@ -723,7 +725,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; } - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (unlikely(r)) { dev_err(adev->dev, "%p pin failed\n", bo); goto error; @@ -760,6 +762,7 @@ int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain, u64 *gpu_addr) int amdgpu_bo_unpin(struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + struct ttm_operation_ctx ctx = { false, false }; int r, i; if (!bo->pin_count) { @@ -773,7 +776,7 @@ int amdgpu_bo_unpin(struct amdgpu_bo *bo) bo->placements[i].lpfn = 0; bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; } - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (unlikely(r)) { dev_err(adev->dev, "%p validate failed for unpin\n", bo); goto error; @@ -945,6 +948,7 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); + struct ttm_operation_ctx ctx = { false, false }; struct amdgpu_bo *abo; unsigned long offset, size; int r; @@ -978,7 +982,7 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) abo->placement.num_busy_placement = 1; abo->placement.busy_placement = &abo->placements[1]; - r = ttm_bo_validate(bo, &abo->placement, false, false); + r = ttm_bo_validate(bo, &abo->placement, &ctx); if (unlikely(r != 0)) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index be607b2be4e9..2f2a9e17fdb4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -408,6 +408,7 @@ static u64 amdgpu_uvd_get_addr_from_ctx(struct amdgpu_uvd_cs_ctx *ctx) */ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx) { + struct ttm_operation_ctx tctx = { false, false }; struct amdgpu_bo_va_mapping *mapping; struct amdgpu_bo *bo; uint32_t cmd; @@ -430,7 +431,7 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx) } amdgpu_uvd_force_into_uvd_segment(bo); - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &tctx); } return r; @@ -949,6 +950,7 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx) static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, bool direct, struct dma_fence **fence) { + struct ttm_operation_ctx ctx = { true, false }; struct ttm_validate_buffer tv; struct ww_acquire_ctx ticket; struct list_head head; @@ -975,7 +977,7 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, amdgpu_uvd_force_into_uvd_segment(bo); } - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) goto err; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index a91abfb32746..ba6d846b08ff 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -558,6 +558,7 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p, uint32_t ib_idx, int lo, int hi, unsigned size, int32_t index) { int64_t offset = ((uint64_t)size) * ((int64_t)index); + struct ttm_operation_ctx ctx = { false, false }; struct amdgpu_bo_va_mapping *mapping; unsigned i, fpfn, lpfn; struct amdgpu_bo *bo; @@ -587,7 +588,7 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p, uint32_t ib_idx, bo->placements[i].lpfn = bo->placements[i].fpfn ? min(bo->placements[i].fpfn, lpfn) : lpfn; } - return ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + return ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index dabaca4da7f2..df218df332b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -274,6 +274,7 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring) static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, bool direct, struct dma_fence **fence) { + struct ttm_operation_ctx ctx = { true, false }; struct ttm_validate_buffer tv; struct ww_acquire_ctx ticket; struct list_head head; @@ -294,7 +295,7 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *b if (r) return r; - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) goto err; diff --git a/drivers/gpu/drm/ast/ast_ttm.c b/drivers/gpu/drm/ast/ast_ttm.c index 696a15dc2f3f..28da7c2b7ed9 100644 --- a/drivers/gpu/drm/ast/ast_ttm.c +++ b/drivers/gpu/drm/ast/ast_ttm.c @@ -354,6 +354,7 @@ static inline u64 ast_bo_gpu_offset(struct ast_bo *bo) int ast_bo_pin(struct ast_bo *bo, u32 pl_flag, u64 *gpu_addr) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (bo->pin_count) { @@ -365,7 +366,7 @@ int ast_bo_pin(struct ast_bo *bo, u32 pl_flag, u64 *gpu_addr) ast_ttm_placement(bo, pl_flag); for (i = 0; i < bo->placement.num_placement; i++) bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) return ret; @@ -377,6 +378,7 @@ int ast_bo_pin(struct ast_bo *bo, u32 pl_flag, u64 *gpu_addr) int ast_bo_unpin(struct ast_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; int i; if (!bo->pin_count) { DRM_ERROR("unpin bad %p\n", bo); @@ -388,11 +390,12 @@ int ast_bo_unpin(struct ast_bo *bo) for (i = 0; i < bo->placement.num_placement ; i++) bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; - return ttm_bo_validate(&bo->bo, &bo->placement, false, false); + return ttm_bo_validate(&bo->bo, &bo->placement, &ctx); } int ast_bo_push_sysram(struct ast_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (!bo->pin_count) { DRM_ERROR("unpin bad %p\n", bo); @@ -409,7 +412,7 @@ int ast_bo_push_sysram(struct ast_bo *bo) for (i = 0; i < bo->placement.num_placement ; i++) bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) { DRM_ERROR("pushing to VRAM failed\n"); return ret; diff --git a/drivers/gpu/drm/bochs/bochs_mm.c b/drivers/gpu/drm/bochs/bochs_mm.c index c4cadb638460..8250b5e612d2 100644 --- a/drivers/gpu/drm/bochs/bochs_mm.c +++ b/drivers/gpu/drm/bochs/bochs_mm.c @@ -283,6 +283,7 @@ static inline u64 bochs_bo_gpu_offset(struct bochs_bo *bo) int bochs_bo_pin(struct bochs_bo *bo, u32 pl_flag, u64 *gpu_addr) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (bo->pin_count) { @@ -295,7 +296,7 @@ int bochs_bo_pin(struct bochs_bo *bo, u32 pl_flag, u64 *gpu_addr) bochs_ttm_placement(bo, pl_flag); for (i = 0; i < bo->placement.num_placement; i++) bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) return ret; @@ -307,6 +308,7 @@ int bochs_bo_pin(struct bochs_bo *bo, u32 pl_flag, u64 *gpu_addr) int bochs_bo_unpin(struct bochs_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (!bo->pin_count) { @@ -320,7 +322,7 @@ int bochs_bo_unpin(struct bochs_bo *bo) for (i = 0; i < bo->placement.num_placement; i++) bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) return ret; diff --git a/drivers/gpu/drm/cirrus/cirrus_ttm.c b/drivers/gpu/drm/cirrus/cirrus_ttm.c index 1ff1838c0d44..2a5b54d3a03a 100644 --- a/drivers/gpu/drm/cirrus/cirrus_ttm.c +++ b/drivers/gpu/drm/cirrus/cirrus_ttm.c @@ -358,6 +358,7 @@ static inline u64 cirrus_bo_gpu_offset(struct cirrus_bo *bo) int cirrus_bo_pin(struct cirrus_bo *bo, u32 pl_flag, u64 *gpu_addr) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (bo->pin_count) { @@ -369,7 +370,7 @@ int cirrus_bo_pin(struct cirrus_bo *bo, u32 pl_flag, u64 *gpu_addr) cirrus_ttm_placement(bo, pl_flag); for (i = 0; i < bo->placement.num_placement; i++) bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) return ret; @@ -381,6 +382,7 @@ int cirrus_bo_pin(struct cirrus_bo *bo, u32 pl_flag, u64 *gpu_addr) int cirrus_bo_push_sysram(struct cirrus_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (!bo->pin_count) { DRM_ERROR("unpin bad %p\n", bo); @@ -397,7 +399,7 @@ int cirrus_bo_push_sysram(struct cirrus_bo *bo) for (i = 0; i < bo->placement.num_placement ; i++) bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) { DRM_ERROR("pushing to VRAM failed\n"); return ret; diff --git a/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c b/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c index 3518167a7dc4..ab4ee5953615 100644 --- a/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c +++ b/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c @@ -344,6 +344,7 @@ int hibmc_bo_create(struct drm_device *dev, int size, int align, int hibmc_bo_pin(struct hibmc_bo *bo, u32 pl_flag, u64 *gpu_addr) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (bo->pin_count) { @@ -356,7 +357,7 @@ int hibmc_bo_pin(struct hibmc_bo *bo, u32 pl_flag, u64 *gpu_addr) hibmc_ttm_placement(bo, pl_flag); for (i = 0; i < bo->placement.num_placement; i++) bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) return ret; @@ -368,6 +369,7 @@ int hibmc_bo_pin(struct hibmc_bo *bo, u32 pl_flag, u64 *gpu_addr) int hibmc_bo_unpin(struct hibmc_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (!bo->pin_count) { @@ -380,7 +382,7 @@ int hibmc_bo_unpin(struct hibmc_bo *bo) for (i = 0; i < bo->placement.num_placement ; i++) bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) { DRM_ERROR("validate failed for unpin: %d\n", ret); return ret; diff --git a/drivers/gpu/drm/mgag200/mgag200_ttm.c b/drivers/gpu/drm/mgag200/mgag200_ttm.c index 3e7e1cd31395..f03da63abc7b 100644 --- a/drivers/gpu/drm/mgag200/mgag200_ttm.c +++ b/drivers/gpu/drm/mgag200/mgag200_ttm.c @@ -354,6 +354,7 @@ static inline u64 mgag200_bo_gpu_offset(struct mgag200_bo *bo) int mgag200_bo_pin(struct mgag200_bo *bo, u32 pl_flag, u64 *gpu_addr) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (bo->pin_count) { @@ -366,7 +367,7 @@ int mgag200_bo_pin(struct mgag200_bo *bo, u32 pl_flag, u64 *gpu_addr) mgag200_ttm_placement(bo, pl_flag); for (i = 0; i < bo->placement.num_placement; i++) bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) return ret; @@ -378,6 +379,7 @@ int mgag200_bo_pin(struct mgag200_bo *bo, u32 pl_flag, u64 *gpu_addr) int mgag200_bo_unpin(struct mgag200_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; int i; if (!bo->pin_count) { DRM_ERROR("unpin bad %p\n", bo); @@ -389,11 +391,12 @@ int mgag200_bo_unpin(struct mgag200_bo *bo) for (i = 0; i < bo->placement.num_placement ; i++) bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; - return ttm_bo_validate(&bo->bo, &bo->placement, false, false); + return ttm_bo_validate(&bo->bo, &bo->placement, &ctx); } int mgag200_bo_push_sysram(struct mgag200_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; int i, ret; if (!bo->pin_count) { DRM_ERROR("unpin bad %p\n", bo); @@ -410,7 +413,7 @@ int mgag200_bo_push_sysram(struct mgag200_bo *bo) for (i = 0; i < bo->placement.num_placement ; i++) bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; - ret = ttm_bo_validate(&bo->bo, &bo->placement, false, false); + ret = ttm_bo_validate(&bo->bo, &bo->placement, &ctx); if (ret) { DRM_ERROR("pushing to VRAM failed\n"); return ret; diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c index 2615912430cc..1cf3da3d7bea 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.c +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c @@ -548,10 +548,10 @@ int nouveau_bo_validate(struct nouveau_bo *nvbo, bool interruptible, bool no_wait_gpu) { + struct ttm_operation_ctx ctx = { interruptible, no_wait_gpu }; int ret; - ret = ttm_bo_validate(&nvbo->bo, &nvbo->placement, - interruptible, no_wait_gpu); + ret = ttm_bo_validate(&nvbo->bo, &nvbo->placement, &ctx); if (ret) return ret; diff --git a/drivers/gpu/drm/qxl/qxl_ioctl.c b/drivers/gpu/drm/qxl/qxl_ioctl.c index 31effed4a3c8..e8c0b1037230 100644 --- a/drivers/gpu/drm/qxl/qxl_ioctl.c +++ b/drivers/gpu/drm/qxl/qxl_ioctl.c @@ -309,6 +309,7 @@ static int qxl_update_area_ioctl(struct drm_device *dev, void *data, int ret; struct drm_gem_object *gobj = NULL; struct qxl_bo *qobj = NULL; + struct ttm_operation_ctx ctx = { true, false }; if (update_area->left >= update_area->right || update_area->top >= update_area->bottom) @@ -326,8 +327,7 @@ static int qxl_update_area_ioctl(struct drm_device *dev, void *data, if (!qobj->pin_count) { qxl_ttm_placement_from_domain(qobj, qobj->type, false); - ret = ttm_bo_validate(&qobj->tbo, &qobj->placement, - true, false); + ret = ttm_bo_validate(&qobj->tbo, &qobj->placement, &ctx); if (unlikely(ret)) goto out; } diff --git a/drivers/gpu/drm/qxl/qxl_object.c b/drivers/gpu/drm/qxl/qxl_object.c index 0a67ddf19c3d..f6b80fe47d1f 100644 --- a/drivers/gpu/drm/qxl/qxl_object.c +++ b/drivers/gpu/drm/qxl/qxl_object.c @@ -223,6 +223,7 @@ struct qxl_bo *qxl_bo_ref(struct qxl_bo *bo) static int __qxl_bo_pin(struct qxl_bo *bo, u32 domain, u64 *gpu_addr) { + struct ttm_operation_ctx ctx = { false, false }; struct drm_device *ddev = bo->gem_base.dev; int r; @@ -233,7 +234,7 @@ static int __qxl_bo_pin(struct qxl_bo *bo, u32 domain, u64 *gpu_addr) return 0; } qxl_ttm_placement_from_domain(bo, domain, true); - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (likely(r == 0)) { bo->pin_count = 1; if (gpu_addr != NULL) @@ -246,6 +247,7 @@ static int __qxl_bo_pin(struct qxl_bo *bo, u32 domain, u64 *gpu_addr) static int __qxl_bo_unpin(struct qxl_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; struct drm_device *ddev = bo->gem_base.dev; int r, i; @@ -258,7 +260,7 @@ static int __qxl_bo_unpin(struct qxl_bo *bo) return 0; for (i = 0; i < bo->placement.num_placement; i++) bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (unlikely(r != 0)) dev_err(ddev->dev, "%p validate failed for unpin\n", bo); return r; diff --git a/drivers/gpu/drm/qxl/qxl_release.c b/drivers/gpu/drm/qxl/qxl_release.c index f27777daae63..b223c8d0a491 100644 --- a/drivers/gpu/drm/qxl/qxl_release.c +++ b/drivers/gpu/drm/qxl/qxl_release.c @@ -230,12 +230,12 @@ int qxl_release_list_add(struct qxl_release *release, struct qxl_bo *bo) static int qxl_release_validate_bo(struct qxl_bo *bo) { + struct ttm_operation_ctx ctx = { true, false }; int ret; if (!bo->pin_count) { qxl_ttm_placement_from_domain(bo, bo->type, false); - ret = ttm_bo_validate(&bo->tbo, &bo->placement, - true, false); + ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (ret) return ret; } diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index cf3deb283da5..a9962ffba720 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -285,6 +285,7 @@ int radeon_gem_create_ioctl(struct drm_device *dev, void *data, int radeon_gem_userptr_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { + struct ttm_operation_ctx ctx = { true, false }; struct radeon_device *rdev = dev->dev_private; struct drm_radeon_gem_userptr *args = data; struct drm_gem_object *gobj; @@ -343,7 +344,7 @@ int radeon_gem_userptr_ioctl(struct drm_device *dev, void *data, } radeon_ttm_placement_from_domain(bo, RADEON_GEM_DOMAIN_GTT); - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); radeon_bo_unreserve(bo); up_read(¤t->mm->mmap_sem); if (r) diff --git a/drivers/gpu/drm/radeon/radeon_mn.c b/drivers/gpu/drm/radeon/radeon_mn.c index 1d62288b7ee3..abd24975c9b1 100644 --- a/drivers/gpu/drm/radeon/radeon_mn.c +++ b/drivers/gpu/drm/radeon/radeon_mn.c @@ -124,6 +124,7 @@ static void radeon_mn_invalidate_range_start(struct mmu_notifier *mn, unsigned long end) { struct radeon_mn *rmn = container_of(mn, struct radeon_mn, mn); + struct ttm_operation_ctx ctx = { false, false }; struct interval_tree_node *it; /* notification is exclusive, but interval is inclusive */ @@ -157,7 +158,7 @@ static void radeon_mn_invalidate_range_start(struct mmu_notifier *mn, DRM_ERROR("(%ld) failed to wait for user bo\n", r); radeon_ttm_placement_from_domain(bo, RADEON_GEM_DOMAIN_CPU); - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) DRM_ERROR("(%ld) failed to validate user bo\n", r); diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index 093594976126..15404af9d740 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -329,6 +329,7 @@ void radeon_bo_unref(struct radeon_bo **bo) int radeon_bo_pin_restricted(struct radeon_bo *bo, u32 domain, u64 max_offset, u64 *gpu_addr) { + struct ttm_operation_ctx ctx = { false, false }; int r, i; if (radeon_ttm_tt_has_userptr(bo->tbo.ttm)) @@ -371,7 +372,7 @@ int radeon_bo_pin_restricted(struct radeon_bo *bo, u32 domain, u64 max_offset, bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; } - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (likely(r == 0)) { bo->pin_count = 1; if (gpu_addr != NULL) @@ -393,6 +394,7 @@ int radeon_bo_pin(struct radeon_bo *bo, u32 domain, u64 *gpu_addr) int radeon_bo_unpin(struct radeon_bo *bo) { + struct ttm_operation_ctx ctx = { false, false }; int r, i; if (!bo->pin_count) { @@ -406,7 +408,7 @@ int radeon_bo_unpin(struct radeon_bo *bo) bo->placements[i].lpfn = 0; bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; } - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (likely(r == 0)) { if (bo->tbo.mem.mem_type == TTM_PL_VRAM) bo->rdev->vram_pin_size -= radeon_bo_size(bo); @@ -531,6 +533,7 @@ int radeon_bo_list_validate(struct radeon_device *rdev, struct ww_acquire_ctx *ticket, struct list_head *head, int ring) { + struct ttm_operation_ctx ctx = { true, false }; struct radeon_bo_list *lobj; struct list_head duplicates; int r; @@ -572,7 +575,7 @@ int radeon_bo_list_validate(struct radeon_device *rdev, radeon_uvd_force_into_uvd_segment(bo, allowed); initial_bytes_moved = atomic64_read(&rdev->num_bytes_moved); - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); bytes_moved += atomic64_read(&rdev->num_bytes_moved) - initial_bytes_moved; @@ -792,6 +795,7 @@ void radeon_bo_move_notify(struct ttm_buffer_object *bo, int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo) { + struct ttm_operation_ctx ctx = { false, false }; struct radeon_device *rdev; struct radeon_bo *rbo; unsigned long offset, size, lpfn; @@ -823,10 +827,10 @@ int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo) (!rbo->placements[i].lpfn || rbo->placements[i].lpfn > lpfn)) rbo->placements[i].lpfn = lpfn; } - r = ttm_bo_validate(bo, &rbo->placement, false, false); + r = ttm_bo_validate(bo, &rbo->placement, &ctx); if (unlikely(r == -ENOMEM)) { radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_GTT); - return ttm_bo_validate(bo, &rbo->placement, false, false); + return ttm_bo_validate(bo, &rbo->placement, &ctx); } else if (unlikely(r != 0)) { return r; } diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index e5c0e635e371..7f1a9c787bd1 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -387,6 +387,7 @@ static void radeon_vm_set_pages(struct radeon_device *rdev, static int radeon_vm_clear_bo(struct radeon_device *rdev, struct radeon_bo *bo) { + struct ttm_operation_ctx ctx = { true, false }; struct radeon_ib ib; unsigned entries; uint64_t addr; @@ -396,7 +397,7 @@ static int radeon_vm_clear_bo(struct radeon_device *rdev, if (r) return r; - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) goto error_unreserve; diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index 77a0fd22e5ba..5347c3f3e2f4 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -1091,9 +1091,8 @@ bool ttm_bo_mem_compat(struct ttm_placement *placement, EXPORT_SYMBOL(ttm_bo_mem_compat); int ttm_bo_validate(struct ttm_buffer_object *bo, - struct ttm_placement *placement, - bool interruptible, - bool no_wait_gpu) + struct ttm_placement *placement, + struct ttm_operation_ctx *ctx) { int ret; uint32_t new_flags; @@ -1103,8 +1102,8 @@ int ttm_bo_validate(struct ttm_buffer_object *bo, * Check whether we need to move buffer. */ if (!ttm_bo_mem_compat(placement, &bo->mem, &new_flags)) { - ret = ttm_bo_move_buffer(bo, placement, interruptible, - no_wait_gpu); + ret = ttm_bo_move_buffer(bo, placement, ctx->interruptible, + ctx->no_wait_gpu); if (ret) return ret; } else { @@ -1219,8 +1218,11 @@ int ttm_bo_init_reserved(struct ttm_bo_device *bdev, WARN_ON(!locked); } - if (likely(!ret)) - ret = ttm_bo_validate(bo, placement, interruptible, false); + if (likely(!ret)) { + struct ttm_operation_ctx ctx = { interruptible, false }; + + ret = ttm_bo_validate(bo, placement, &ctx); + } if (unlikely(ret)) { if (!resv) diff --git a/drivers/gpu/drm/virtio/virtgpu_ioctl.c b/drivers/gpu/drm/virtio/virtgpu_ioctl.c index 461f81aa1bbe..5720a0d4ac0a 100644 --- a/drivers/gpu/drm/virtio/virtgpu_ioctl.c +++ b/drivers/gpu/drm/virtio/virtgpu_ioctl.c @@ -56,6 +56,7 @@ static int virtio_gpu_map_ioctl(struct drm_device *dev, void *data, static int virtio_gpu_object_list_validate(struct ww_acquire_ctx *ticket, struct list_head *head) { + struct ttm_operation_ctx ctx = { false, false }; struct ttm_validate_buffer *buf; struct ttm_buffer_object *bo; struct virtio_gpu_object *qobj; @@ -68,7 +69,7 @@ static int virtio_gpu_object_list_validate(struct ww_acquire_ctx *ticket, list_for_each_entry(buf, head, head) { bo = buf->bo; qobj = container_of(bo, struct virtio_gpu_object, tbo); - ret = ttm_bo_validate(bo, &qobj->placement, false, false); + ret = ttm_bo_validate(bo, &qobj->placement, &ctx); if (ret) { ttm_eu_backoff_reservation(ticket, head); return ret; @@ -352,6 +353,7 @@ static int virtio_gpu_transfer_from_host_ioctl(struct drm_device *dev, struct virtio_gpu_device *vgdev = dev->dev_private; struct virtio_gpu_fpriv *vfpriv = file->driver_priv; struct drm_virtgpu_3d_transfer_from_host *args = data; + struct ttm_operation_ctx ctx = { true, false }; struct drm_gem_object *gobj = NULL; struct virtio_gpu_object *qobj = NULL; struct virtio_gpu_fence *fence; @@ -372,8 +374,7 @@ static int virtio_gpu_transfer_from_host_ioctl(struct drm_device *dev, if (ret) goto out; - ret = ttm_bo_validate(&qobj->tbo, &qobj->placement, - true, false); + ret = ttm_bo_validate(&qobj->tbo, &qobj->placement, &ctx); if (unlikely(ret)) goto out_unres; @@ -399,6 +400,7 @@ static int virtio_gpu_transfer_to_host_ioctl(struct drm_device *dev, void *data, struct virtio_gpu_device *vgdev = dev->dev_private; struct virtio_gpu_fpriv *vfpriv = file->driver_priv; struct drm_virtgpu_3d_transfer_to_host *args = data; + struct ttm_operation_ctx ctx = { true, false }; struct drm_gem_object *gobj = NULL; struct virtio_gpu_object *qobj = NULL; struct virtio_gpu_fence *fence; @@ -416,8 +418,7 @@ static int virtio_gpu_transfer_to_host_ioctl(struct drm_device *dev, void *data, if (ret) goto out; - ret = ttm_bo_validate(&qobj->tbo, &qobj->placement, - true, false); + ret = ttm_bo_validate(&qobj->tbo, &qobj->placement, &ctx); if (unlikely(ret)) goto out_unres; diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c b/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c index d87861bbe971..92df0b08c194 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c @@ -387,6 +387,7 @@ static int vmw_cotable_readback(struct vmw_resource *res) */ static int vmw_cotable_resize(struct vmw_resource *res, size_t new_size) { + struct ttm_operation_ctx ctx = { false, false }; struct vmw_private *dev_priv = res->dev_priv; struct vmw_cotable *vcotbl = vmw_cotable(res); struct vmw_dma_buffer *buf, *old_buf = res->backup; @@ -455,7 +456,7 @@ static int vmw_cotable_resize(struct vmw_resource *res, size_t new_size) } /* Unpin new buffer, and switch backup buffers. */ - ret = ttm_bo_validate(bo, &vmw_mob_placement, false, false); + ret = ttm_bo_validate(bo, &vmw_mob_placement, &ctx); if (unlikely(ret != 0)) { DRM_ERROR("Failed validating new COTable backup buffer.\n"); goto out_wait; diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_dmabuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_dmabuf.c index 0cd889015dc5..d45d2caffa5a 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_dmabuf.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_dmabuf.c @@ -47,6 +47,7 @@ int vmw_dmabuf_pin_in_placement(struct vmw_private *dev_priv, struct ttm_placement *placement, bool interruptible) { + struct ttm_operation_ctx ctx = {interruptible, false }; struct ttm_buffer_object *bo = &buf->base; int ret; uint32_t new_flags; @@ -65,7 +66,7 @@ int vmw_dmabuf_pin_in_placement(struct vmw_private *dev_priv, ret = ttm_bo_mem_compat(placement, &bo->mem, &new_flags) == true ? 0 : -EINVAL; else - ret = ttm_bo_validate(bo, placement, interruptible, false); + ret = ttm_bo_validate(bo, placement, &ctx); if (!ret) vmw_bo_pin_reserved(buf, true); @@ -95,6 +96,7 @@ int vmw_dmabuf_pin_in_vram_or_gmr(struct vmw_private *dev_priv, struct vmw_dma_buffer *buf, bool interruptible) { + struct ttm_operation_ctx ctx = {interruptible, false }; struct ttm_buffer_object *bo = &buf->base; int ret; uint32_t new_flags; @@ -115,12 +117,11 @@ int vmw_dmabuf_pin_in_vram_or_gmr(struct vmw_private *dev_priv, goto out_unreserve; } - ret = ttm_bo_validate(bo, &vmw_vram_gmr_placement, interruptible, - false); + ret = ttm_bo_validate(bo, &vmw_vram_gmr_placement, &ctx); if (likely(ret == 0) || ret == -ERESTARTSYS) goto out_unreserve; - ret = ttm_bo_validate(bo, &vmw_vram_placement, interruptible, false); + ret = ttm_bo_validate(bo, &vmw_vram_placement, &ctx); out_unreserve: if (!ret) @@ -170,6 +171,7 @@ int vmw_dmabuf_pin_in_start_of_vram(struct vmw_private *dev_priv, struct vmw_dma_buffer *buf, bool interruptible) { + struct ttm_operation_ctx ctx = {interruptible, false }; struct ttm_buffer_object *bo = &buf->base; struct ttm_placement placement; struct ttm_place place; @@ -200,14 +202,16 @@ int vmw_dmabuf_pin_in_start_of_vram(struct vmw_private *dev_priv, if (bo->mem.mem_type == TTM_PL_VRAM && bo->mem.start < bo->num_pages && bo->mem.start > 0 && - buf->pin_count == 0) - (void) ttm_bo_validate(bo, &vmw_sys_placement, false, false); + buf->pin_count == 0) { + ctx.interruptible = false; + (void) ttm_bo_validate(bo, &vmw_sys_placement, &ctx); + } if (buf->pin_count > 0) ret = ttm_bo_mem_compat(&placement, &bo->mem, &new_flags) == true ? 0 : -EINVAL; else - ret = ttm_bo_validate(bo, &placement, interruptible, false); + ret = ttm_bo_validate(bo, &placement, &ctx); /* For some reason we didn't end up at the start of vram */ WARN_ON(ret == 0 && bo->offset != 0); @@ -286,6 +290,7 @@ void vmw_bo_get_guest_ptr(const struct ttm_buffer_object *bo, */ void vmw_bo_pin_reserved(struct vmw_dma_buffer *vbo, bool pin) { + struct ttm_operation_ctx ctx = { false, true }; struct ttm_place pl; struct ttm_placement placement; struct ttm_buffer_object *bo = &vbo->base; @@ -314,7 +319,7 @@ void vmw_bo_pin_reserved(struct vmw_dma_buffer *vbo, bool pin) placement.num_placement = 1; placement.placement = &pl; - ret = ttm_bo_validate(bo, &placement, false, true); + ret = ttm_bo_validate(bo, &placement, &ctx); BUG_ON(ret != 0 || bo->mem.mem_type != old_mem_type); } diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c index 21c62a34e558..b700667f6f0b 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c @@ -3701,14 +3701,14 @@ int vmw_validate_single_buffer(struct vmw_private *dev_priv, { struct vmw_dma_buffer *vbo = container_of(bo, struct vmw_dma_buffer, base); + struct ttm_operation_ctx ctx = { interruptible, true }; int ret; if (vbo->pin_count > 0) return 0; if (validate_as_mob) - return ttm_bo_validate(bo, &vmw_mob_placement, interruptible, - false); + return ttm_bo_validate(bo, &vmw_mob_placement, &ctx); /** * Put BO in VRAM if there is space, otherwise as a GMR. @@ -3717,8 +3717,7 @@ int vmw_validate_single_buffer(struct vmw_private *dev_priv, * used as a GMR, this will return -ENOMEM. */ - ret = ttm_bo_validate(bo, &vmw_vram_gmr_placement, interruptible, - false); + ret = ttm_bo_validate(bo, &vmw_vram_gmr_placement, &ctx); if (likely(ret == 0 || ret == -ERESTARTSYS)) return ret; @@ -3727,7 +3726,7 @@ int vmw_validate_single_buffer(struct vmw_private *dev_priv, * previous contents. */ - ret = ttm_bo_validate(bo, &vmw_vram_placement, interruptible, false); + ret = ttm_bo_validate(bo, &vmw_vram_placement, &ctx); return ret; } diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c index a96f90f017d1..200904ff9a22 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c @@ -968,6 +968,7 @@ vmw_resource_check_buffer(struct vmw_resource *res, bool interruptible, struct ttm_validate_buffer *val_buf) { + struct ttm_operation_ctx ctx = { true, false }; struct list_head val_list; bool backup_dirty = false; int ret; @@ -992,7 +993,7 @@ vmw_resource_check_buffer(struct vmw_resource *res, backup_dirty = res->backup_dirty; ret = ttm_bo_validate(&res->backup->base, res->func->backup_placement, - true, false); + &ctx); if (unlikely(ret != 0)) goto out_no_validate; @@ -1446,6 +1447,7 @@ void vmw_resource_evict_all(struct vmw_private *dev_priv) */ int vmw_resource_pin(struct vmw_resource *res, bool interruptible) { + struct ttm_operation_ctx ctx = { interruptible, false }; struct vmw_private *dev_priv = res->dev_priv; int ret; @@ -1466,7 +1468,7 @@ int vmw_resource_pin(struct vmw_resource *res, bool interruptible) ret = ttm_bo_validate (&vbo->base, res->func->backup_placement, - interruptible, false); + &ctx); if (ret) { ttm_bo_unreserve(&vbo->base); goto out_no_validate; diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c index 9b832f136813..004e18b8832c 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c @@ -970,6 +970,7 @@ int vmw_compat_shader_add(struct vmw_private *dev_priv, size_t size, struct list_head *list) { + struct ttm_operation_ctx ctx = { false, true }; struct vmw_dma_buffer *buf; struct ttm_bo_kmap_obj map; bool is_iomem; @@ -1005,7 +1006,7 @@ int vmw_compat_shader_add(struct vmw_private *dev_priv, WARN_ON(is_iomem); ttm_bo_kunmap(&map); - ret = ttm_bo_validate(&buf->base, &vmw_sys_placement, false, true); + ret = ttm_bo_validate(&buf->base, &vmw_sys_placement, &ctx); WARN_ON(ret != 0); ttm_bo_unreserve(&buf->base); diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h index 833c3ad24091..097951e999bc 100644 --- a/include/drm/ttm/ttm_bo_api.h +++ b/include/drm/ttm/ttm_bo_api.h @@ -258,6 +258,20 @@ struct ttm_bo_kmap_obj { struct ttm_buffer_object *bo; }; +/** + * struct ttm_operation_ctx + * + * @interruptible: Sleep interruptible if sleeping. + * @no_wait_gpu: Return immediately if the GPU is busy. + * + * Context for TTM operations like changing buffer placement or general memory + * allocation. + */ +struct ttm_operation_ctx { + bool interruptible; + bool no_wait_gpu; +}; + /** * ttm_bo_reference - reference a struct ttm_buffer_object * @@ -306,8 +320,7 @@ bool ttm_bo_mem_compat(struct ttm_placement *placement, struct ttm_mem_reg *mem, * * @bo: The buffer object. * @placement: Proposed placement for the buffer object. - * @interruptible: Sleep interruptible if sleeping. - * @no_wait_gpu: Return immediately if the GPU is busy. + * @ctx: validation parameters. * * Changes placement and caching policy of the buffer object * according proposed placement. @@ -319,8 +332,7 @@ bool ttm_bo_mem_compat(struct ttm_placement *placement, struct ttm_mem_reg *mem, */ int ttm_bo_validate(struct ttm_buffer_object *bo, struct ttm_placement *placement, - bool interruptible, - bool no_wait_gpu); + struct ttm_operation_ctx *ctx); /** * ttm_bo_unref -- cgit From 6af046d26f34278eacd6ecddb37170624f6d4251 Mon Sep 17 00:00:00 2001 From: Christian König Date: Thu, 27 Apr 2017 18:20:47 +0200 Subject: drm/amdgpu: use the new TTM bytes moved counter v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Instead of the global statistics use the per context bytes moved counter. v2: rebased Signed-off-by: Christian König Reviewed-by: Michel Dänzer Reviewed-by: Chunming Zhou Tested-by: Dieter Nützel Tested-by: Michel Dänzer Acked-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 9 +++------ drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++------- 2 files changed, 6 insertions(+), 13 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index faab662ce680..5e89d7a7178f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -344,7 +344,6 @@ static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); struct ttm_operation_ctx ctx = { true, false }; - u64 initial_bytes_moved, bytes_moved; uint32_t domain; int r; @@ -374,15 +373,13 @@ static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, retry: amdgpu_ttm_placement_from_domain(bo, domain); - initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); - bytes_moved = atomic64_read(&adev->num_bytes_moved) - - initial_bytes_moved; - p->bytes_moved += bytes_moved; + + p->bytes_moved += ctx.bytes_moved; if (adev->mc.visible_vram_size < adev->mc.real_vram_size && bo->tbo.mem.mem_type == TTM_PL_VRAM && bo->tbo.mem.start < adev->mc.visible_vram_size >> PAGE_SHIFT) - p->bytes_moved_vis += bytes_moved; + p->bytes_moved_vis += ctx.bytes_moved; if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) { domain = bo->allowed_domains; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 15027f751e07..dc0a8be98043 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -331,7 +331,6 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, struct amdgpu_bo *bo; enum ttm_bo_type type; unsigned long page_align; - u64 initial_bytes_moved, bytes_moved; size_t acc_size; int r; @@ -406,22 +405,19 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, bo->tbo.bdev = &adev->mman.bdev; amdgpu_ttm_placement_from_domain(bo, domain); - initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); - /* Kernel allocation are uninterruptible */ r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type, &bo->placement, page_align, &ctx, NULL, acc_size, sg, resv, &amdgpu_ttm_bo_destroy); if (unlikely(r != 0)) return r; - bytes_moved = atomic64_read(&adev->num_bytes_moved) - - initial_bytes_moved; if (adev->mc.visible_vram_size < adev->mc.real_vram_size && bo->tbo.mem.mem_type == TTM_PL_VRAM && bo->tbo.mem.start < adev->mc.visible_vram_size >> PAGE_SHIFT) - amdgpu_cs_report_moved_bytes(adev, bytes_moved, bytes_moved); + amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, + ctx.bytes_moved); else - amdgpu_cs_report_moved_bytes(adev, bytes_moved, 0); + amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, 0); if (kernel) bo->tbo.priority = 1; -- cgit From bb7939b2030ab55acd203c86160c37db22f5796a Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 6 Nov 2017 15:37:01 +0100 Subject: drm/amdgpu: fix VA hole handling on Vega10 v3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Similar to the CPU address space the VA on Vega10 has a hole in it. v2: use dev_dbg instead of dev_err v3: add some more comments to explain how the hw works Signed-off-by: Christian König Reviewed-by: Alex Deucher CC: stable@vger.kernel.org Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 10 +++++----- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 11 +++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 4 +++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 13 +++++++++++++ 4 files changed, 32 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 5e89d7a7178f..93d3cef66503 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -870,8 +870,8 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping *m; struct amdgpu_bo *aobj = NULL; struct amdgpu_cs_chunk *chunk; + uint64_t offset, va_start; struct amdgpu_ib *ib; - uint64_t offset; uint8_t *kptr; chunk = &p->chunks[i]; @@ -881,14 +881,14 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB) continue; - r = amdgpu_cs_find_mapping(p, chunk_ib->va_start, - &aobj, &m); + va_start = chunk_ib->va_start & AMDGPU_VA_HOLE_MASK; + r = amdgpu_cs_find_mapping(p, va_start, &aobj, &m); if (r) { DRM_ERROR("IB va_start is invalid\n"); return r; } - if ((chunk_ib->va_start + chunk_ib->ib_bytes) > + if ((va_start + chunk_ib->ib_bytes) > (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) { DRM_ERROR("IB va_start+ib_bytes is invalid\n"); return -EINVAL; @@ -901,7 +901,7 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, } offset = m->start * AMDGPU_GPU_PAGE_SIZE; - kptr += chunk_ib->va_start - offset; + kptr += va_start - offset; memcpy(ib->ptr, kptr, chunk_ib->ib_bytes); amdgpu_bo_kunmap(aobj); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index c16579287aee..59c9facf9bd7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -564,6 +564,17 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, return -EINVAL; } + if (args->va_address >= AMDGPU_VA_HOLE_START && + args->va_address < AMDGPU_VA_HOLE_END) { + dev_dbg(&dev->pdev->dev, + "va_address 0x%LX is in VA hole 0x%LX-0x%LX\n", + args->va_address, AMDGPU_VA_HOLE_START, + AMDGPU_VA_HOLE_END); + return -EINVAL; + } + + args->va_address &= AMDGPU_VA_HOLE_MASK; + if ((args->flags & ~valid_flags) && (args->flags & ~prt_flags)) { dev_err(&dev->pdev->dev, "invalid flags combination 0x%08X\n", args->flags); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index f55021ae788d..2614269c4d7f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -578,7 +578,9 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file if (amdgpu_sriov_vf(adev)) dev_info.ids_flags |= AMDGPU_IDS_FLAGS_PREEMPTION; dev_info.virtual_address_offset = AMDGPU_VA_RESERVED_SIZE; - dev_info.virtual_address_max = (uint64_t)adev->vm_manager.max_pfn * AMDGPU_GPU_PAGE_SIZE; + dev_info.virtual_address_max = + min(adev->vm_manager.max_pfn * AMDGPU_GPU_PAGE_SIZE, + AMDGPU_VA_HOLE_START); dev_info.virtual_address_alignment = max((int)PAGE_SIZE, AMDGPU_GPU_PAGE_SIZE); dev_info.pte_fragment_size = (1 << adev->vm_manager.fragment_size) * AMDGPU_GPU_PAGE_SIZE; dev_info.gart_page_size = AMDGPU_GPU_PAGE_SIZE; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index e8f8896d18db..c80d45dd2bd3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -96,6 +96,19 @@ struct amdgpu_bo_list_entry; /* hardcode that limit for now */ #define AMDGPU_VA_RESERVED_SIZE (8ULL << 20) +/* VA hole for 48bit addresses on Vega10 */ +#define AMDGPU_VA_HOLE_START 0x0000800000000000ULL +#define AMDGPU_VA_HOLE_END 0xffff800000000000ULL + +/* + * Hardware is programmed as if the hole doesn't exists with start and end + * address values. + * + * This mask is used to remove the upper 16bits of the VA and so come up with + * the linear addr value. + */ +#define AMDGPU_VA_HOLE_MASK 0x0000ffffffffffffULL + /* max vmids dedicated for process */ #define AMDGPU_VM_MAX_RESERVED_VMID 1 -- cgit From cebb52b7bc325863600aff930407bba773010938 Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Mon, 13 Nov 2017 14:47:52 -0500 Subject: drm/amdgpu: Get rid of dep_sync as a seperate object. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Instead mark fence as explicit in it's amdgpu_sync_entry. v2: Fix use after free bug and add new parameter description. Signed-off-by: Andrey Grodzovsky Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 14 +++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 24 +++++++++++------------- drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c | 15 ++++++++++----- drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 12 ++++++------ 7 files changed, 37 insertions(+), 35 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index ecc2e60e5f0c..5e2958a79928 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1121,7 +1121,6 @@ struct amdgpu_job { struct amdgpu_vm *vm; struct amdgpu_ring *ring; struct amdgpu_sync sync; - struct amdgpu_sync dep_sync; struct amdgpu_sync sched_sync; struct amdgpu_ib *ibs; struct dma_fence *fence; /* the hw fence */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 93d3cef66503..4cea9ab237ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -786,7 +786,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) return r; r = amdgpu_sync_fence(adev, &p->job->sync, - fpriv->prt_va->last_pt_update); + fpriv->prt_va->last_pt_update, false); if (r) return r; @@ -800,7 +800,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) return r; f = bo_va->last_pt_update; - r = amdgpu_sync_fence(adev, &p->job->sync, f); + r = amdgpu_sync_fence(adev, &p->job->sync, f, false); if (r) return r; } @@ -823,7 +823,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) return r; f = bo_va->last_pt_update; - r = amdgpu_sync_fence(adev, &p->job->sync, f); + r = amdgpu_sync_fence(adev, &p->job->sync, f, false); if (r) return r; } @@ -834,7 +834,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) if (r) return r; - r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_update); + r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_update, false); if (r) return r; @@ -1038,8 +1038,8 @@ static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p, amdgpu_ctx_put(ctx); return r; } else if (fence) { - r = amdgpu_sync_fence(p->adev, &p->job->dep_sync, - fence); + r = amdgpu_sync_fence(p->adev, &p->job->sync, fence, + true); dma_fence_put(fence); amdgpu_ctx_put(ctx); if (r) @@ -1058,7 +1058,7 @@ static int amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p, if (r) return r; - r = amdgpu_sync_fence(p->adev, &p->job->dep_sync, fence); + r = amdgpu_sync_fence(p->adev, &p->job->sync, fence, true); dma_fence_put(fence); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 659997bfff30..0cf86eb357d6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -164,7 +164,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, } if (ring->funcs->emit_pipeline_sync && job && - ((tmp = amdgpu_sync_get_fence(&job->sched_sync)) || + ((tmp = amdgpu_sync_get_fence(&job->sched_sync, NULL)) || amdgpu_vm_need_pipeline_sync(ring, job))) { need_pipe_sync = true; dma_fence_put(tmp); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 18770a880393..bdc210ac74f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -60,7 +60,6 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, (*job)->num_ibs = num_ibs; amdgpu_sync_create(&(*job)->sync); - amdgpu_sync_create(&(*job)->dep_sync); amdgpu_sync_create(&(*job)->sched_sync); (*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter); @@ -104,7 +103,6 @@ static void amdgpu_job_free_cb(struct amd_sched_job *s_job) amdgpu_ring_priority_put(job->ring, s_job->s_priority); dma_fence_put(job->fence); amdgpu_sync_free(&job->sync); - amdgpu_sync_free(&job->dep_sync); amdgpu_sync_free(&job->sched_sync); kfree(job); } @@ -115,7 +113,6 @@ void amdgpu_job_free(struct amdgpu_job *job) dma_fence_put(job->fence); amdgpu_sync_free(&job->sync); - amdgpu_sync_free(&job->dep_sync); amdgpu_sync_free(&job->sched_sync); kfree(job); } @@ -149,17 +146,18 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job, { struct amdgpu_job *job = to_amdgpu_job(sched_job); struct amdgpu_vm *vm = job->vm; - - struct dma_fence *fence = amdgpu_sync_get_fence(&job->dep_sync); + bool explicit = false; int r; - - if (amd_sched_dependency_optimized(fence, s_entity)) { - r = amdgpu_sync_fence(job->adev, &job->sched_sync, fence); - if (r) - DRM_ERROR("Error adding fence to sync (%d)\n", r); + struct dma_fence *fence = amdgpu_sync_get_fence(&job->sync, &explicit); + + if (fence && explicit) { + if (amd_sched_dependency_optimized(fence, s_entity)) { + r = amdgpu_sync_fence(job->adev, &job->sched_sync, fence, false); + if (r) + DRM_ERROR("Error adding fence to sync (%d)\n", r); + } } - if (!fence) - fence = amdgpu_sync_get_fence(&job->sync); + while (fence == NULL && vm && !job->vm_id) { struct amdgpu_ring *ring = job->ring; @@ -169,7 +167,7 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job, if (r) DRM_ERROR("Error getting VM ID (%d)\n", r); - fence = amdgpu_sync_get_fence(&job->sync); + fence = amdgpu_sync_get_fence(&job->sync, NULL); } return fence; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c index a4bf21f8f1c1..f3d1a25b660f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c @@ -35,6 +35,7 @@ struct amdgpu_sync_entry { struct hlist_node node; struct dma_fence *fence; + bool explicit; }; static struct kmem_cache *amdgpu_sync_slab; @@ -141,7 +142,7 @@ static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f) * */ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync, - struct dma_fence *f) + struct dma_fence *f, bool explicit) { struct amdgpu_sync_entry *e; @@ -159,6 +160,8 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync, if (!e) return -ENOMEM; + e->explicit = explicit; + hash_add(sync->fences, &e->node, f->context); e->fence = dma_fence_get(f); return 0; @@ -189,7 +192,7 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, /* always sync to the exclusive fence */ f = reservation_object_get_excl(resv); - r = amdgpu_sync_fence(adev, sync, f); + r = amdgpu_sync_fence(adev, sync, f, false); if (explicit_sync) return r; @@ -220,7 +223,7 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, continue; } - r = amdgpu_sync_fence(adev, sync, f); + r = amdgpu_sync_fence(adev, sync, f, false); if (r) break; } @@ -275,19 +278,21 @@ struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, * amdgpu_sync_get_fence - get the next fence from the sync object * * @sync: sync object to use + * @explicit: true if the next fence is explicit * * Get and removes the next fence from the sync object not signaled yet. */ -struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync) +struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync, bool *explicit) { struct amdgpu_sync_entry *e; struct hlist_node *tmp; struct dma_fence *f; int i; - hash_for_each_safe(sync->fences, i, tmp, e, node) { f = e->fence; + if (explicit) + *explicit = e->explicit; hash_del(&e->node); kmem_cache_free(amdgpu_sync_slab, e); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h index 70d7e3a279a0..7aba38d5c9df 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h @@ -41,7 +41,7 @@ struct amdgpu_sync { void amdgpu_sync_create(struct amdgpu_sync *sync); int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync, - struct dma_fence *f); + struct dma_fence *f, bool explicit); int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync, struct reservation_object *resv, @@ -49,7 +49,7 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, bool explicit_sync); struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, struct amdgpu_ring *ring); -struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync); +struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync, bool *explicit); int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr); void amdgpu_sync_free(struct amdgpu_sync *sync); int amdgpu_sync_init(void); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 7de519b86b78..3ecdbdfb04dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -488,7 +488,7 @@ static int amdgpu_vm_grab_reserved_vmid_locked(struct amdgpu_vm *vm, id->pd_gpu_addr = 0; tmp = amdgpu_sync_peek_fence(&id->active, ring); if (tmp) { - r = amdgpu_sync_fence(adev, sync, tmp); + r = amdgpu_sync_fence(adev, sync, tmp, false); return r; } } @@ -496,7 +496,7 @@ static int amdgpu_vm_grab_reserved_vmid_locked(struct amdgpu_vm *vm, /* Good we can use this VMID. Remember this submission as * user of the VMID. */ - r = amdgpu_sync_fence(ring->adev, &id->active, fence); + r = amdgpu_sync_fence(ring->adev, &id->active, fence, false); if (r) goto out; @@ -583,7 +583,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, } - r = amdgpu_sync_fence(ring->adev, sync, &array->base); + r = amdgpu_sync_fence(ring->adev, sync, &array->base, false); dma_fence_put(&array->base); if (r) goto error; @@ -626,7 +626,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, /* Good we can use this VMID. Remember this submission as * user of the VMID. */ - r = amdgpu_sync_fence(ring->adev, &id->active, fence); + r = amdgpu_sync_fence(ring->adev, &id->active, fence, false); if (r) goto error; @@ -646,7 +646,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, id = idle; /* Remember this submission as user of the VMID */ - r = amdgpu_sync_fence(ring->adev, &id->active, fence); + r = amdgpu_sync_fence(ring->adev, &id->active, fence, false); if (r) goto error; @@ -1657,7 +1657,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, addr = 0; } - r = amdgpu_sync_fence(adev, &job->sync, exclusive); + r = amdgpu_sync_fence(adev, &job->sync, exclusive, false); if (r) goto error_free; -- cgit