Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <202307132107.2ce4ea2f-oliver.sang@intel.com> <20230713150923.GA28246@lst.de>
 <ZLAZn_SBmoIFG5F5@yuki>
In-Reply-To: <ZLAZn_SBmoIFG5F5@yuki>
From:   Jan Stancek <jstancek@redhat.com>
Date:   Fri, 14 Jul 2023 13:50:58 +0200
Message-ID: <CAASaF6xbgSf+X+SF8wLjFrsMA4=XxHti0SXDZQP1ZqdGYp4aUQ@mail.gmail.com>
Subject: Re: [LTP] [linus:master] [iomap] 219580eea1: ltp.writev07.fail
To:     Cyril Hrubis <chrubis@suse.cz>
Cc:     Christoph Hellwig <hch@lst.de>,
        "Darrick J. Wong" <djwong@kernel.org>,
        Andreas Gruenbacher <agruenba@redhat.com>,
        Miklos Szeredi <miklos@szeredi.hu>,
        Matthew Wilcox <willy@infradead.org>, cluster-devel@redhat.com,
        Ilya Dryomov <idryomov@gmail.com>,
        Miklos Szeredi <mszeredi@redhat.com>,
        Chao Yu <chao@kernel.org>, oe-lkp@lists.linux.dev,
        Al Viro <viro@zeniv.linux.org.uk>,
        Jaegeuk Kim <jaegeuk@kernel.org>, Xiubo Li <xiubli@redhat.com>,
        Trond Myklebust <trond.myklebust@hammerspace.com>,
        ltp@lists.linux.it, lkp@intel.com, Jens Axboe <axboe@kernel.dk>,
        Christian Brauner <brauner@kernel.org>,
        "Theodore Ts'o" <tytso@mit.edu>,
        Johannes Thumshirn <johannes.thumshirn@wdc.com>,
        linux-kernel@vger.kernel.org, linux-xfs@vger.kernel.org,
        kernel test robot <oliver.sang@intel.com>,
        Anna Schumaker <anna@kernel.org>,
        linux-fsdevel@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>,
        Hannes Reinecke <hare@suse.de>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Jul 13, 2023 at 5:38=E2=80=AFPM Cyril Hrubis <chrubis@suse.cz> wrot=
e:
>
> Hi!
> > I can't reproduce this on current mainline.  Is this a robust failure
> > or flapping test?  Especiall as the FAIL conditions look rather
> > unrelated.

It's consistently reproducible for me on xfs with HEAD at:
eb26cbb1a754 ("Merge tag 'platform-drivers-x86-v6.5-2' of
git://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86")

>
> Actually the test is spot on, the difference is that previously the
> error was returned form the iomap_file_buffered_write() only if we
> failed with the first buffer from the iov, now we always return the
> error and we do not advance the offset.
>
> The change that broke it:
>
> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> index 063133ec77f4..550525a525c4 100644
> --- a/fs/iomap/buffered-io.c
> +++ b/fs/iomap/buffered-io.c
> @@ -864,16 +864,19 @@ iomap_file_buffered_write(struct kiocb *iocb, struc=
t iov_iter *i,
>                 .len            =3D iov_iter_count(i),
>                 .flags          =3D IOMAP_WRITE,
>         };
> -       int ret;
> +       ssize_t ret;
>
>         if (iocb->ki_flags & IOCB_NOWAIT)
>                 iter.flags |=3D IOMAP_NOWAIT;
>
>         while ((ret =3D iomap_iter(&iter, ops)) > 0)
>                 iter.processed =3D iomap_write_iter(&iter, i);
> -       if (iter.pos =3D=3D iocb->ki_pos)
> +
> +       if (unlikely(ret < 0))
>                 return ret;
> -       return iter.pos - iocb->ki_pos;
> +       ret =3D iter.pos - iocb->ki_pos;
> +       iocb->ki_pos +=3D ret;
> +       return ret;
>  }
>
> I suppose that we shoudl fix is with something as:
>
> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> index adb92cdb24b0..bfb39f7bc303 100644
> --- a/fs/iomap/buffered-io.c
> +++ b/fs/iomap/buffered-io.c
> @@ -872,11 +872,12 @@ iomap_file_buffered_write(struct kiocb *iocb, struc=
t iov_iter *i,
>         while ((ret =3D iomap_iter(&iter, ops)) > 0)
>                 iter.processed =3D iomap_write_iter(&iter, i);
>
> +       iocb->ki_pos +=3D iter.pos - iocb->ki_pos;
> +
>         if (unlikely(ret < 0))
>                 return ret;
> -       ret =3D iter.pos - iocb->ki_pos;
> -       iocb->ki_pos +=3D ret;
> -       return ret;
> +
> +       return iter.pos - iocb->ki_pos;

Replacing "ret" with "iter.pos - iocb->ki_pos" here doesn't look
equivalent to original,
because you already updated "iocb->ki_pos" few lines above.

Wouldn't it be enough to bring the old condition back?

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index adb92cdb24b0..7cc9f7274883 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -872,7 +872,7 @@ iomap_file_buffered_write(struct kiocb *iocb,
struct iov_iter *i,
        while ((ret =3D iomap_iter(&iter, ops)) > 0)
                iter.processed =3D iomap_write_iter(&iter, i);

-       if (unlikely(ret < 0))
+       if (unlikely(iter.pos =3D=3D iocb->ki_pos))
                return ret;
        ret =3D iter.pos - iocb->ki_pos;
        iocb->ki_pos +=3D ret;

(with hunk above applied)
# ./writev07
tst_test.c:1526: TINFO: Timeout per run is 0h 00m 30s
writev07.c:50: TINFO: starting test with initial file offset: 0
writev07.c:94: TINFO: writev() has written 64 bytes
writev07.c:105: TPASS: file has expected content
writev07.c:116: TPASS: offset at 64 as expected
writev07.c:50: TINFO: starting test with initial file offset: 65
writev07.c:94: TINFO: writev() has written 64 bytes
writev07.c:105: TPASS: file has expected content
writev07.c:116: TPASS: offset at 129 as expected
writev07.c:50: TINFO: starting test with initial file offset: 4096
writev07.c:94: TINFO: writev() has written 64 bytes
writev07.c:105: TPASS: file has expected content
writev07.c:116: TPASS: offset at 4160 as expected
writev07.c:50: TINFO: starting test with initial file offset: 4097
writev07.c:94: TINFO: writev() has written 64 bytes
writev07.c:105: TPASS: file has expected content
writev07.c:116: TPASS: offset at 4161 as expected


>  }
>  EXPORT_SYMBOL_GPL(iomap_file_buffered_write);
>
>
> --
> Cyril Hrubis
> chrubis@suse.cz
>
> --
> Mailing list info: https://lists.linux.it/listinfo/ltp
>