drm/amdgpu: refuse to reserve io mem for split VRAM buffers
authorNicolai Hähnle <nicolai.haehnle@amd.com>
Thu, 16 Feb 2017 22:49:12 +0000 (23:49 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 21 Oct 2017 15:21:36 +0000 (17:21 +0200)
[ Upstream commit 4694335dad7357e9b3d7822ab13049014d74d8b0 ]

When the fast blit path fails while attempting to move a buffer from RAM
to VRAM, we fall back to a CPU-based memcpy that cannot handle split VRAM
buffers. Instead of crashing, simply fail the buffer move.

Ideally, we would teach TTM about split buffers so that the fallback still
works in this case, but that is quite involved. So for now, apply the
simplest possible fix.

Fixes: 40361bb1704b ("drm/amdgpu: add VRAM manager v2")
Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

index 264899df9bfc3523c7ee2a9848ee9efa15b0c302..05ff98b43c50f3677dcd5ef971c9fd5bc0548203 100644 (file)
@@ -491,6 +491,9 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_
        case TTM_PL_TT:
                break;
        case TTM_PL_VRAM:
+               if (mem->start == AMDGPU_BO_INVALID_OFFSET)
+                       return -EINVAL;
+
                mem->bus.offset = mem->start << PAGE_SHIFT;
                /* check if it's visible */
                if ((mem->bus.offset + mem->bus.size) > adev->mc.visible_vram_size)