Re: [PATCH 2/2] xfs: fix an integer overflow error in xfs_growfs_rt

["Darrick J. Wong" <djwong@kernel.org>]

On Wed, Jul 14, 2021 at 11:12:15AM +1000, Dave Chinner wrote:
> On Mon, Jul 12, 2021 at 03:07:31PM -0700, Darrick J. Wong wrote:
> > From: Darrick J. Wong <djwong@kernel.org>
> > 
> > During a realtime grow operation, we run a single transaction for each
> > rt bitmap block added to the filesystem.  This means that each step has
> > to be careful to increase sb_rblocks appropriately.
> > 
> > Fix the integer overflow error in this calculation that can happen when
> > the extent size is very large.  Found by running growfs to add a rt
> > volume to a filesystem formatted with a 1g rt extent size.
> > 
> > Signed-off-by: Darrick J. Wong <djwong@kernel.org>
> > ---
> >  fs/xfs/xfs_rtalloc.c |    2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > 
> > diff --git a/fs/xfs/xfs_rtalloc.c b/fs/xfs/xfs_rtalloc.c
> > index 8920bce4fb0a..a47d43c30283 100644
> > --- a/fs/xfs/xfs_rtalloc.c
> > +++ b/fs/xfs/xfs_rtalloc.c
> > @@ -1019,7 +1019,7 @@ xfs_growfs_rt(
> >  		nsbp->sb_rbmblocks = bmbno + 1;
> >  		nsbp->sb_rblocks =
> >  			XFS_RTMIN(nrblocks,
> > -				  nsbp->sb_rbmblocks * NBBY *
> > +				  (xfs_rfsblock_t)nsbp->sb_rbmblocks * NBBY *
> >  				  nsbp->sb_blocksize * nsbp->sb_rextsize);
> >  		nsbp->sb_rextents = nsbp->sb_rblocks;
> >  		do_div(nsbp->sb_rextents, nsbp->sb_rextsize);
> 
> Oh, that's just nasty code.  This needs a comment explaining that the
> cast is to avoid an overflow, otherwise someone will come along
> later and remove the "unnecessary" cast.
> 
> Alternatively, because we do "nsbp->sb_rbmblocks = bmbno + 1;" a
> couple of lines above, this could be done differently without the
> need for a cast. Make bmbno a xfs_rfsblock_t, and simply write the
> code as:
> 
> 		nsbp->sb_rblocks = min(nrblocks,
> 					(bmbno + 1) * NBBY *
> 					nsbp->sb_blocksize * nsbp->sb_rextsize);
> 		nsbp->sb_rbmblocks = bmbno + 1;

I like that, it'll get changed in the next revision.

> Notes for future cleanup:
> 
> #define XFS_RTMIN(a,b) ((a) < (b) ? (a) : (b))
> 
> Needs to die.

Heh, yes.

--D

> 
> Cheers,
> 
> Dave.
> -- 
> Dave Chinner
> david@fromorbit.com