Re: [PATCH 07/16] drm/radeon: use DMA_RESV_USAGE_KERNEL

On Wed, Apr 06, 2022 at 09:51:23AM +0200, Christian König wrote:

Always wait for kernel fences before kmap and not only for UVD kmaps.

Signed-off-by: Christian König christian.koenig@amd.com

drivers/gpu/drm/radeon/radeon_object.c | 7 ++++++- drivers/gpu/drm/radeon/radeon_uvd.c | 12 ++---------- 2 files changed, 8 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index cb5c4aa45cef..6c4a6802ca96 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -219,7 +219,12 @@ int radeon_bo_create(struct radeon_device *rdev, int radeon_bo_kmap(struct radeon_bo *bo, void **ptr) { bool is_iomem;

• int r;

• long r;
• r = dma_resv_wait_timeout(bo->tbo.base.resv, DMA_RESV_USAGE_KERNEL,

• 		  false, MAX_SCHEDULE_TIMEOUT);
  

Maybe another reason why we should push this wait into ttm kmap helpers?

Acked-by: Daniel Vetter daniel.vetter@ffwll.ch

• if (r < 0)

• return r;
  
  

  if (bo->kptr) { if (ptr) {

diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c b/drivers/gpu/drm/radeon/radeon_uvd.c index a50750740ab0..a2cda184b2b2 100644 --- a/drivers/gpu/drm/radeon/radeon_uvd.c +++ b/drivers/gpu/drm/radeon/radeon_uvd.c @@ -470,24 +470,16 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, int32_t *msg, msg_type, handle; unsigned img_size = 0; void *ptr;

• long r;
• int i;

• int i, r;

  if (offset & 0x3F) { DRM_ERROR("UVD messages must be 64 byte aligned!\n"); return -EINVAL; }

• r = dma_resv_wait_timeout(bo->tbo.base.resv, DMA_RESV_USAGE_WRITE,

• 		  false, MAX_SCHEDULE_TIMEOUT);
  

• if (r <= 0) {

• DRM_ERROR("Failed waiting for UVD message (%ld)!\n", r);
  

• return r ? r : -ETIME;
  

• }
• r = radeon_bo_kmap(bo, &ptr); if (r) {

• DRM_ERROR("Failed mapping the UVD message (%ld)!\n", r);
  

• DRM_ERROR("Failed mapping the UVD message (%d)!\n", r);
  

  return r; }

-- 2.25.1