Am 2023-04-12 um 14:49 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>
The headline is a bit inaccurate. The shortest description I can come up
with is "Ignore KFD eviction fences invalidating preemptible DMABuf
imports".
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 5 ++++-
1 file changed, 4 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..a0828f6d9fbe 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -526,7 +526,10 @@ 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, false);
The interruptible flag (3rd parameter) should come from ctx->interruptible.
Regards,
Felix
+ else
+ r = ttm_bo_wait_ctx(bo, ctx);
if (r)
return r;
