[PATCH 01/11] drm/amdgpu: Remove call to reservation_object_test_signaled_rcu before wait
Christian König
christian.koenig at amd.com
Mon Aug 29 08:20:14 UTC 2016
Am 29.08.2016 um 09:08 schrieb Chris Wilson:
> Since fence_wait_timeout_reservation_object_wait_timeout_rcu() with a
> timeout of 0 becomes reservation_object_test_signaled_rcu(), we do not
> need to handle such conversion in the caller. The only challenge are
> those callers that wish to differentiate the error code between the
> nonblocking busy check and potentially blocking wait.
>
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Alex Deucher <alexander.deucher at amd.com>
> Cc: Christian König <christian.koenig at amd.com>
Mhm, actually it was one of our developers who added the shortcut into
reservation_object_wait_timeout_rcu().
But it looks like we forgot to clean this up in the amdgpu driver after
doing this. So thanks for taking care of this.
Patch #1 as well as patch #6 are Reviewed-by: Christian König
<christian.koenig at amd.com>.
For patch #6 there is also an use case for this in TTM
ttm_bo_add_move_fence().
We could change the lock into an RCU if we can make sure that we get an
up to date RCU protected fence and not NULL if the fence becomes freed
right at the wrong time.
Regards,
Christian.
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> index 88fbed2389c0..a3e6f883ac2c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> @@ -407,10 +407,8 @@ int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
> return -ENOENT;
> }
> robj = gem_to_amdgpu_bo(gobj);
> - if (timeout == 0)
> - ret = reservation_object_test_signaled_rcu(robj->tbo.resv, true);
> - else
> - ret = reservation_object_wait_timeout_rcu(robj->tbo.resv, true, true, timeout);
> + ret = reservation_object_wait_timeout_rcu(robj->tbo.resv, true, true,
> + timeout);
>
> /* ret == 0 means not signaled,
> * ret > 0 means signaled
More information about the dri-devel
mailing list