[lvc-project] [PATCH v2] fs: minix: Fix handling of corrupted directories

[Kitotavrik]

>  guess the easiest is to fudge i_nlinks count in memory to 2 to
> avoid issues...

But if a subdirectory was in the corrupted directory(nlinks= 3), it
will be replaced with nlinks 2. And after deleting subdirectory,
nlinks was 1 and the problem will remain. Maybe should return EUCLEAN,
regardless of the mounting mode.


ср, 14 мая 2025 г. в 18:54, Jan Kara <jack at suse.cz>:
>
> On Wed 14-05-25 13:38:35, Andrey Kriulin wrote:
> > If the directory is corrupted and the number of nlinks is less than 2
> > (valid nlinks have at least 2), then when the directory is deleted, the
> > minix_rmdir will try to reduce the nlinks(unsigned int) to a negative
> > value.
> >
> > Make nlinks validity check for directory.
> >
> > Found by Linux Verification Center (linuxtesting.org) with Syzkaller.
> >
> > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> > Cc: stable at vger.kernel.org
> > Signed-off-by: Andrey Kriulin <kitotavrik.media at gmail.com>
> > Signed-off-by: Andrey Kriulin <kitotavrik.s at gmail.com>
> > ---
> > v2: Move nlinks validaty check to V[12]_minix_iget() per Jan Kara
> > <jack at suse.cz> request. Change return error code to EUCLEAN. Don't block
> > directory in r/o mode per Al Viro <viro at zeniv.linux.org.uk> request.
> >
> >  fs/minix/inode.c | 16 ++++++++++++++++
> >  1 file changed, 16 insertions(+)
> >
> > diff --git a/fs/minix/inode.c b/fs/minix/inode.c
> > index f007e389d5d2..d815397b8b0d 100644
> > --- a/fs/minix/inode.c
> > +++ b/fs/minix/inode.c
> > @@ -517,6 +517,14 @@ static struct inode *V1_minix_iget(struct inode *inode)
> >               iget_failed(inode);
> >               return ERR_PTR(-ESTALE);
> >       }
> > +     if (S_ISDIR(raw_inode->i_mode) && raw_inode->i_nlinks < 2) {
> > +             printk("MINIX-fs: inode directory with corrupted number of links");
>
> A message like this is rather useless because it shows nothing either about
> the inode or the link count or the filesystem where this happened. I'd
> either improve or delete it.
>
> > +             if (!sb_rdonly(inode->i_sb)) {
> > +                     brelse(bh);
> > +                     iget_failed(inode);
> > +                     return ERR_PTR(-EUCLEAN);
> > +             }
>
> OK, but when the inode is cached in memory with the wrong link count and
> then the filesystem is remounted read-write, you will get the same problem
> as before? I guess the easiest is to fudge i_nlinks count in memory to 2 to
> avoid issues...
>
>
>                                                                 Honza
> --
> Jan Kara <jack at suse.com>
> SUSE Labs, CR