]> git.baikalelectronics.ru Git - kernel.git/commitdiff
drm/amdgpu: refine SR-IOV firmware VRAM reservation to protect data
authorHorace Chen <horace.chen@amd.com>
Wed, 1 Nov 2017 11:32:11 +0000 (19:32 +0800)
committerAlex Deucher <alexander.deucher@amd.com>
Mon, 4 Dec 2017 21:41:42 +0000 (16:41 -0500)
The previous solution will create a zero buffer on the system
domain and then move the zeroes to the VRAM. This will break the
original data on the VRAM.

Refine the code to create bo on VRAM domain directly and then remove
and re-create mem node to the exact position before bo_pin. This can
avoid breaking the data and will not cause eviction.

Signed-off-by: Horace Chen <horace.chen@amd.com>
Reviewed-by: monk liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

index 69e724c96442b21f7d38b0b781d2253a899272cc..83dbd02004b0981c2403e38586d06af575b81ffa 100644 (file)
@@ -680,8 +680,12 @@ void amdgpu_fw_reserve_vram_fini(struct amdgpu_device *adev)
 int amdgpu_fw_reserve_vram_init(struct amdgpu_device *adev)
 {
        int r = 0;
+       int i;
        u64 gpu_addr;
        u64 vram_size = adev->mc.visible_vram_size;
+       u64 offset = adev->fw_vram_usage.start_offset;
+       u64 size = adev->fw_vram_usage.size;
+       struct amdgpu_bo *bo;
 
        adev->fw_vram_usage.va = NULL;
        adev->fw_vram_usage.reserved_bo = NULL;
@@ -690,7 +694,7 @@ int amdgpu_fw_reserve_vram_init(struct amdgpu_device *adev)
                adev->fw_vram_usage.size <= vram_size) {
 
                r = amdgpu_bo_create(adev, adev->fw_vram_usage.size,
-                       PAGE_SIZE, true, 0,
+                       PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,
                        AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
                        AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL, 0,
                        &adev->fw_vram_usage.reserved_bo);
@@ -700,6 +704,23 @@ int amdgpu_fw_reserve_vram_init(struct amdgpu_device *adev)
                r = amdgpu_bo_reserve(adev->fw_vram_usage.reserved_bo, false);
                if (r)
                        goto error_reserve;
+
+               /* remove the original mem node and create a new one at the
+                * request position
+                */
+               bo = adev->fw_vram_usage.reserved_bo;
+               offset = ALIGN(offset, PAGE_SIZE);
+               for (i = 0; i < bo->placement.num_placement; ++i) {
+                       bo->placements[i].fpfn = offset >> PAGE_SHIFT;
+                       bo->placements[i].lpfn = (offset + size) >> PAGE_SHIFT;
+               }
+
+               ttm_bo_mem_put(&bo->tbo, &bo->tbo.mem);
+               r = ttm_bo_mem_space(&bo->tbo, &bo->placement, &bo->tbo.mem,
+                                    false, false);
+               if (r)
+                       goto error_pin;
+
                r = amdgpu_bo_pin_restricted(adev->fw_vram_usage.reserved_bo,
                        AMDGPU_GEM_DOMAIN_VRAM,
                        adev->fw_vram_usage.start_offset,