Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit cebb52b7 authored by Andrey Grodzovsky's avatar Andrey Grodzovsky Committed by Alex Deucher
Browse files

drm/amdgpu: Get rid of dep_sync as a seperate object. 



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: default avatarAndrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: default avatarChristian König <christian.koenig@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent 97489129

drivers/gpu/drm/amd/amdgpu/amdgpu.h

+0 −1
Original line number Diff line number Diff line
@@ -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 */

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

+7 −7
Original line number Diff line number Diff line
@@ -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;

drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c

+1 −1
Original line number Diff line number Diff line
@@ -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);

drivers/gpu/drm/amd/amdgpu/amdgpu_job.c

+11 −13
Original line number Diff line number Diff line
@@ -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;

	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);

			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;

drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c

+10 −5
Original line number Diff line number Diff line
@@ -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);