Re: [PATCH 5/6] xfs: fix handling of synchronous errors in xfs_buf_submit

["Darrick J. Wong" <djwong@kernel.org>]

On Thu, Jun 25, 2026 at 03:58:35PM +0200, Christoph Hellwig wrote:
> Synchronous readers and writers already run __xfs_buf_ioend from
> xfs_buf_iowait after being woken through bp->b_iowait, so we
> should not call it here, which can lead to double completions.

How hard is it to trip the double completion?  I guess all you need is
an IO failure (or a verifier failure) on a synchronous read?

> Fixes: 4b90de5bc0f5 ("xfs: reduce context switches for synchronous buffered I/O")
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com>

If so, then

Cc: <stable@vger.kernel.org> # v6.14
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>

--D

> ---
>  fs/xfs/xfs_buf.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
> index 3eea9d9ea829..83549573e2cc 100644
> --- a/fs/xfs/xfs_buf.c
> +++ b/fs/xfs/xfs_buf.c
> @@ -1403,12 +1403,12 @@ xfs_buf_submit(
>  	bp->b_flags &= ~XBF_DONE;
>  	xfs_buf_stale(bp);
>  end_io:
> -	if (!__xfs_buf_ioend(bp))
> -		return;
> -	if (bp->b_flags & XBF_ASYNC)
> -		xfs_buf_relse(bp);
> -	else
> +	if (bp->b_flags & XBF_ASYNC) {
> +		if (__xfs_buf_ioend(bp))
> +			xfs_buf_relse(bp);
> +	} else {
>  		complete(&bp->b_iowait);
> +	}
>  }
>  
>  /*
> -- 
> 2.53.0
> 
>