amdgpu: Use drmIoctl in amdgpu_ioctl_wait_cs

This is safe now because the ioctl uses an absolute timeout.

This prevents amdgpu_cs_query_fence_status from returning early e.g.
when a signal is delivered, which in turn caused Mesa winsys code to
assume a BO was idle when it actually wasn't yet.

Reviewed-by: Christian König <christian.koenig@amd.com>

1 files changed, 2 insertions, 6 deletions

diff --git a/amdgpu/amdgpu_cs.c b/amdgpu/amdgpu_cs.c
index 91a7b059..d9aa22d6 100644
--- a/amdgpu/amdgpu_cs.c
+++ b/amdgpu/amdgpu_cs.c
@@ -367,12 +367,8 @@ static int amdgpu_ioctl_wait_cs(amdgpu_context_handle context,
 	else
 		args.in.timeout = amdgpu_cs_calculate_timeout(timeout_ns);
 
-	/* Handle errors manually here because of timeout */
-	r = ioctl(dev->fd, DRM_IOCTL_AMDGPU_WAIT_CS, &args);
-	if (r == -1 && (errno == EINTR || errno == EAGAIN)) {
-		*busy = true;
-		return 0;
-	} else if (r)
+	r = drmIoctl(dev->fd, DRM_IOCTL_AMDGPU_WAIT_CS, &args);
+	if (r)
 		return -errno;
 
 	*busy = args.out.status;