On 21/05/31 08:49PM, David Sterba wrote:

On Sun, May 30, 2021 at 08:24:05PM +0530, Ritesh Harjani wrote:

quoted

We always return 0 even in case of an error in btrfs_mark_extent_written().
Fix it to return proper error value in case of a failure.

Oh right, this looks like it got forgotten, the whole function does the
right thing regarding errors and the callers also handle it.

quoted

Signed-off-by: Ritesh Harjani <redacted>
---
Tested fstests with "-g quick" group. No new failure seen.

That won't be probably enough to trigger the error paths but the patch

Yes. However, I found the bug while stress testing btrfs/062 failure
with bs < ps patch series from Qu.
On analyzing the kernel panic [1], it looked like it was caused by not
properly returning an error from btrfs_mark_extent_written().
So, I thought of just running "-g quick" test to avoid any obvious issue.

[1]: https://www.spinics.net/lists/linux-btrfs/msg113280.html

otherwise looks correct to me. Added to misc-next, thanks.

Thanks