On Wed, Aug 24, 2016 at 02:01:55PM -0400, Brian Foster wrote:
> On Fri, Aug 19, 2016 at 01:30:22PM -0700, Darrick J. Wong wrote:
> > If the caller passes in a cursor to a zero-height btree (which is
> > impossible), we never set block to anything but NULL, which causes the
> > later dereference of it to crash. Instead, just return -EFSCORRUPTED.
> >
> > Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> > ---
>
> Did something actually cause this to happen?
Well, if you took a reflink xfs that was formatted before we added the
rmap/refcount btree block counters to the AGF and then tried to mount
it after will blow up like this, because the refcount_level field got
moved to somewhere that's zero in the old image.
Anyone being malicious with xfs_db can also do this to any other _level field.
(Yay automounting!)
--D
>
> Brian
>
> > fs/xfs/libxfs/xfs_btree.c | 4 ++++
> > 1 file changed, 4 insertions(+)
> >
> > diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
> > index 64fd847..4bffea4 100644
> > --- a/fs/xfs/libxfs/xfs_btree.c
> > +++ b/fs/xfs/libxfs/xfs_btree.c
> > @@ -1814,6 +1814,10 @@ xfs_btree_lookup(
> >
> > XFS_BTREE_STATS_INC(cur, lookup);
> >
> > + /* No such thing as a zero-level tree. */
> > + if (cur->bc_nlevels == 0)
> > + return -EFSCORRUPTED;
> > +
> > block = NULL;
> > keyno = 0;
> >
> >
> > _______________________________________________
> > xfs mailing list
> > xfs@xxxxxxxxxxx
> > http://oss.sgi.com/mailman/listinfo/xfs
|