On 2017-11-29 08:10 AM, Christian König wrote: > Hi Andres, > > just a gentle ping to see if you have noticed this. > > Thanks, > Christian. > > Am 24.11.2017 um 13:49 schrieb Christian König: >> Fixes a bug introduced by AMDGPU_GEM_CREATE_EXPLICIT_SYNC. We still need >> to wait for pipelined moves in the shared fences list. >> >> v2: fix typo >> >> Signed-off-by: Christian König <christian.koenig at amd.com> Hi Christian, Sorry, last few weeks have been a little hectic. This patch looks good to me. You can add: Reviewed-by: Andres Rodriguez <andresx7 at gmail.com> The steamvr explicit sync use cases are untouched by this patch, so we should be good on that front as well. Kind Regards, Andres >> --- >>  drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c | 7 ++----- >>  1 file changed, 2 insertions(+), 5 deletions(-) >> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c >> index a4bf21f8f1c1..bbbc40d630a0 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c >> @@ -191,9 +191,6 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, >>      f = reservation_object_get_excl(resv); >>      r = amdgpu_sync_fence(adev, sync, f); >> -   if (explicit_sync) >> -       return r; >> - >>      flist = reservation_object_get_list(resv); >>      if (!flist || r) >>          return r; >> @@ -212,11 +209,11 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, >>                   (fence_owner == AMDGPU_FENCE_OWNER_VM))) >>                  continue; >> -           /* Ignore fence from the same owner as >> +           /* Ignore fence from the same owner and explicit one as >>               * long as it isn't undefined. >>               */ >>              if (owner != AMDGPU_FENCE_OWNER_UNDEFINED && >> -               fence_owner == owner) >> +               (fence_owner == owner || explicit_sync)) >>                  continue; >>          } >
