Am 06.04.22 um 14:42 schrieb Daniel Vetter:
On Wed, Apr 06, 2022 at 09:51:22AM +0200, Christian König wrote:
Wait only for kernel fences before kmap or UVD direct submission.
This also makes sure that we always wait in amdgpu_bo_kmap() even when
returning a cached pointer.
Signed-off-by: Christian König <christian.koenig@xxxxxxx>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++-----
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 2 +-
2 files changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index a3cdf8a24377..5832c05ab10d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -761,6 +761,11 @@ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr)
if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)
return -EPERM;
+ r = dma_resv_wait_timeout(bo->tbo.base.resv, DMA_RESV_USAGE_KERNEL,
+ false, MAX_SCHEDULE_TIMEOUT);
+ if (r < 0)
+ return r;
+
kptr = amdgpu_bo_kptr(bo);
if (kptr) {
if (ptr)
@@ -768,11 +773,6 @@ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr)
return 0;
}
- r = dma_resv_wait_timeout(bo->tbo.base.resv, DMA_RESV_USAGE_WRITE,
- false, MAX_SCHEDULE_TIMEOUT);
- if (r < 0)
- return r;
-
r = ttm_bo_kmap(&bo->tbo, 0, bo->tbo.resource->num_pages, &bo->kmap);
I wonder whether waiting for kernel fences shouldn't be ttm's duty here.
Anyway patch makes some sense to me.
I was thinking the same and already had it halve implemented until I
realized that this won't work because of the ptr caching.
Need to move that around as well and rework the whole handling.
Christian.
Acked-by: Daniel Vetter <daniel.vetter@xxxxxxxx>
if (r)
return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
index 3654326219e0..6eac649499d3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
@@ -1164,7 +1164,7 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
if (direct) {
r = dma_resv_wait_timeout(bo->tbo.base.resv,
- DMA_RESV_USAGE_WRITE, false,
+ DMA_RESV_USAGE_KERNEL, false,
msecs_to_jiffies(10));
if (r == 0)
r = -ETIMEDOUT;
--
2.25.1
