drm/amdgpu: linear validate first then bind to GART
authorChristian König <christian.koenig@amd.com>
Mon, 16 Oct 2017 08:32:04 +0000 (10:32 +0200)
committerAlex Deucher <alexander.deucher@amd.com>
Thu, 19 Oct 2017 19:27:21 +0000 (15:27 -0400)
For VM emulation for old UVD/VCE we need to validate the BO with linear
VRAM flag set first and then eventually bind it to GART.

Validating with linear VRAM flag set can move the BO to GART making
UVD/VCE read/write from an unbound GART BO.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
CC: stable@vger.kernel.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

index 32cf83e2f2d9fbfb2589f3be415bec814fa06e05..f7fceb63413c9b1c0769a85d38116ba15e43f2e4 100644 (file)
@@ -1582,14 +1582,14 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
        if (READ_ONCE((*bo)->tbo.resv->lock.ctx) != &parser->ticket)
                return -EINVAL;
 
-       r = amdgpu_ttm_bind(&(*bo)->tbo, &(*bo)->tbo.mem);
-       if (unlikely(r))
-               return r;
-
-       if ((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
-               return 0;
+       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);
+               if (r)
+                       return r;
+       }
 
-       (*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
-       amdgpu_ttm_placement_from_domain(*bo, (*bo)->allowed_domains);
-       return ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, false, false);
+       return amdgpu_ttm_bind(&(*bo)->tbo, &(*bo)->tbo.mem);
 }