Re: [PATCH] nilfs2: handle corrupted checkpoint count gracefully during deletion

Ryusuke Konishi <[email protected]> Tue, 7 Jul 2026 01:39:12 +0900
Newsgroups org.kernel.vger.linux-nilfs,org.kernel.vger.linux-kernel
Message-ID <CAKFNMo=XyMb_Tmm4F1awoyB0f=ndmp9vmRn44+fAQBmABtmTEg@mail.gmail.com>
On Mon, Jul 6, 2026 at 8:08=E2=80=AFPM Igor Putko wrote:
>
> Syzkaller reported a kernel warning in nilfs_cpfile_delete_checkpoints()
> due to a corrupted checkpoint count on the storage medium where
> le32_to_cpu(cp->cp_checkpoints_count) is less than the number of
> checkpoints being deleted.
> Triggering a WARN_ON() for disk image corruption is suboptimal. Fix
> this by returning -EIO and reporting a filesystem error via
> nilfs_error() instead of interrupting execution with a kernel warning.
>
> Reported-by: [email protected]
> Closes: https://syzkaller.appspot.com/bug?extid=3D79b815da3aec0a6a4d02
> Signed-off-by: Igor Putko <[email protected]>
> ---
>  fs/nilfs2/cpfile.c | 20 ++++++++++++++++----
>  1 file changed, 16 insertions(+), 4 deletions(-)

Thanks for the patch proposal!

A filesystem inconsistency is indeed the root cause, and WARN_ON is
undesirably catching it.

The proposed fix looks appropriate to me, so I plan to apply it after
running some tests on my end.

Thanks,
Ryusuke Konishi

>
> diff --git a/fs/nilfs2/cpfile.c b/fs/nilfs2/cpfile.c
> index 4bbdc832d..d3349fa58 100644
> --- a/fs/nilfs2/cpfile.c
> +++ b/fs/nilfs2/cpfile.c
> @@ -81,18 +81,26 @@ nilfs_cpfile_block_add_valid_checkpoints(const struct=
 inode *cpfile,
>         return count;
>  }
>
> -static unsigned int
> +static int
>  nilfs_cpfile_block_sub_valid_checkpoints(const struct inode *cpfile,
>                                          struct buffer_head *bh,
>                                          unsigned int n)
>  {
>         struct nilfs_checkpoint *cp;
> -       unsigned int count;
> +       unsigned int checkpoints_count;
> +       int count;
>
>         cp =3D kmap_local_folio(bh->b_folio,
>                               offset_in_folio(bh->b_folio, bh->b_data));
> -       WARN_ON(le32_to_cpu(cp->cp_checkpoints_count) < n);
> -       count =3D le32_to_cpu(cp->cp_checkpoints_count) - n;
> +       checkpoints_count =3D le32_to_cpu(cp->cp_checkpoints_count);
> +       if (unlikely(checkpoints_count < n)) {
> +               nilfs_error(cpfile->i_sb,
> +                           "deleted checkpoints count %u exceeds block c=
ount %u",
> +                           n, checkpoints_count);
> +               kunmap_local(cp);
> +               return -EIO;
> +       }
> +       count =3D checkpoints_count - n;
>         cp->cp_checkpoints_count =3D cpu_to_le32(count);
>         kunmap_local(cp);
>         return count;
> @@ -522,6 +530,10 @@ int nilfs_cpfile_delete_checkpoints(struct inode *cp=
file,
>                 count =3D nilfs_cpfile_block_sub_valid_checkpoints(cpfile=
, cp_bh,
>                                                                  nicps);
>                 brelse(cp_bh);
> +               if (unlikely(count < 0)) {
> +                       ret =3D count;
> +                       break;
> +               }
>                 if (count)
>                         continue;
>
> --
> 2.47.3
>