Am 13.04.23 um 03:01 schrieb Felix Kuehling:
Am 2023-04-12 um 18:25 schrieb Eric Huang:
It is to avoid redundant eviction for KFD's DMAbuf import
bo when dmaunmapping DMAbuf. The DMAbuf import bo has
been set as AMDGPU_PL_PREEMPT in KFD when mapping.
Signed-off-by: Eric Huang <jinhuieric.huang@xxxxxxx>
Reviewed-by: Felix Kuehling <Felix.Kuehling@xxxxxxx>
I'd like to get an Acked-by from Christian as well before submitting
this.
I have to admit that I only partially followed the internal discussion,
but in general you need a *really* good explanation for this.
E.g. add code comment and explain in the commit message extensively why
this is needed and why there are no alternatives.
Regards,
Christian.
Thanks,
Felix
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 2430f3e9f3a7..64795fe9eecb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -526,7 +526,12 @@ static int amdgpu_bo_move(struct
ttm_buffer_object *bo, bool evict,
if ((old_mem->mem_type == TTM_PL_TT ||
old_mem->mem_type == AMDGPU_PL_PREEMPT) &&
new_mem->mem_type == TTM_PL_SYSTEM) {
- r = ttm_bo_wait_ctx(bo, ctx);
+ if (old_mem->mem_type == AMDGPU_PL_PREEMPT)
+ r = amdgpu_bo_sync_wait(abo,
+ AMDGPU_FENCE_OWNER_KFD,
+ ctx->interruptible);
+ else
+ r = ttm_bo_wait_ctx(bo, ctx);
if (r)
return r;
