[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAHc6FU6vAokT9ugX1DA8iQLbeu7=Eixr9bq6z0o77_Nq+PyXvw@mail.gmail.com>
Date: Sun, 21 Dec 2025 16:39:47 +0100
From: Andreas Gruenbacher <agruenba@...hat.com>
To: Christoph Hellwig <hch@...radead.org>, Jens Axboe <axboe@...nel.dk>, Chris Mason <clm@...com>,
David Sterba <dsterba@...e.com>
Cc: linux-block@...r.kernel.org, linux-btrfs@...r.kernel.org,
linux-raid@...r.kernel.org, dm-devel@...ts.linux.dev,
linux-kernel@...r.kernel.org
Subject: Re: [RFC v2 01/17] xfs: don't clobber bi_status in xfs_zone_alloc_and_submit
On Sun, Dec 21, 2025 at 3:52 AM Andreas Gruenbacher <agruenba@...hat.com> wrote:
> Function xfs_zone_alloc_and_submit() sets bio->bi_status and then it
> calls bio_io_error(), which overwrites that value again. Fix that by
> completing the bio separately after setting bio->bi_status.
By the way, this bug makes me wonder if we shouldn't just get rid of
bio_io_error() in favour of bio_endio_status(bio, BLK_STS_IOERR). The
latter would be a lot more obvious.
Andreas
> Signed-off-by: Andreas Gruenbacher <agruenba@...hat.com>
> ---
> fs/xfs/xfs_zone_alloc.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/fs/xfs/xfs_zone_alloc.c b/fs/xfs/xfs_zone_alloc.c
> index ef7a931ebde5..bd6f3ef095cb 100644
> --- a/fs/xfs/xfs_zone_alloc.c
> +++ b/fs/xfs/xfs_zone_alloc.c
> @@ -897,6 +897,9 @@ xfs_zone_alloc_and_submit(
>
> out_split_error:
> ioend->io_bio.bi_status = errno_to_blk_status(PTR_ERR(split));
> + bio_endio(&ioend->io_bio);
> + return;
> +
> out_error:
> bio_io_error(&ioend->io_bio);
> }
> --
> 2.52.0
>
Powered by blists - more mailing lists