The fbdev dirty callback (drm_atomic_helper_dirtyfb()) can fail with
-ENOMEM when the atomic commit allocation fails under memory pressure.
That is a transient runtime condition, not a driver bug: the damage is
only cosmetic and the frame is redrawn on the next flush.

drm_WARN_ONCE() turns this recoverable failure into a WARN splat (and a
panic with panic_on_warn), which syzbot keeps reporting:

  virtio-pci 0000:00:01.0: [drm] Dirty helper failed: ret=-12
  WARNING: drivers/gpu/drm/drm_fbdev_shmem.c:118 at ...

Keep the warning for other failures, but do not warn about -ENOMEM.  The
error is still propagated to the caller.

Fixes: 150f431a0831 ("drm/fbdev: Add fbdev-shmem")
Reported-by: [email protected]
Closes: https://syzkaller.appspot.com/bug?extid=369ee6a6e9d0bbf14b37
Signed-off-by: Jiahao Xu <[email protected]>
---
v2:
 - Add the Fixes: tag reported by sashiko-bot.  No code change.
---
 drivers/gpu/drm/drm_fbdev_shmem.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/drm_fbdev_shmem.c 
b/drivers/gpu/drm/drm_fbdev_shmem.c
index 8eb05fe9b..c3fa5a663 100644
--- a/drivers/gpu/drm/drm_fbdev_shmem.c
+++ b/drivers/gpu/drm/drm_fbdev_shmem.c
@@ -115,7 +115,14 @@ static int drm_fbdev_shmem_helper_fb_dirty(struct 
drm_fb_helper *helper,
 
        if (helper->fb->funcs->dirty) {
                ret = helper->fb->funcs->dirty(helper->fb, NULL, 0, 0, clip, 1);
-               if (drm_WARN_ONCE(dev, ret, "Dirty helper failed: ret=%d\n", 
ret))
+               /*
+                * An allocation failure is a transient condition under memory
+                * pressure, not a driver bug, so don't warn about it.  The
+                * error itself is still propagated to the caller.
+                */
+               drm_WARN_ONCE(dev, ret && ret != -ENOMEM,
+                             "Dirty helper failed: ret=%d\n", ret);
+               if (ret)
                        return ret;
        }
 
-- 
2.43.0


Reply via email to