Re: [PATCH] xfs: fix error handling in xfs_bmap_extents_to_btree

[Dave Chinner <david@fromorbit.com>]

On Sun, Sep 30, 2018 at 08:34:44AM -0700, Darrick J. Wong wrote:
> On Sun, Sep 30, 2018 at 04:18:07PM +1000, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@redhat.com>
> > 
> > Commit 01239d77b9dd ("xfs: fix a null pointer dereference in
> > xfs_bmap_extents_to_btree") attempted to fix a null pointer
> > dreference when a fuzzing corruption of some kind was found.
> > This fix was flawed, resulting in assert failures like:
.....
> > @@ -732,9 +732,10 @@ xfs_bmap_extents_to_btree(
> >  	xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, 1L);
> >  	abp = xfs_btree_get_bufl(mp, tp, args.fsbno, 0);
> >  	if (!abp) {
> 
> This reminds me to ask, has anyone made progress converting the
> {get,read}_buf functions (and associated callers) to return error codes?

Not me. I had a brief thought about this when looking at this,
but I think we really need to start from the bottom up (i.e. from
the xfs_buf_* functions first) before moving to the higher level
code.

> (I wasn't expecting that as a part of this fix; I'm simply recollecting
> my marbles.)
> 
> Anyway, this looks ok to me.
> 
> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>

Thanks Darrick, I'll add it to the fixes branch (it's been on top of
the merged tree I've been testing over the weekend) and get it out.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com