Re: [PATCH 2/3] xfs: complain if anyone tries to create a too-large buffer log item

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Wed, Jan 08, 2020 at 12:51:31AM -0800, Christoph Hellwig wrote:
> On Tue, Jan 07, 2020 at 08:18:19PM -0800, Darrick J. Wong wrote:
> > diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c
> > index 3984779e5911..bfbe8a5b8959 100644
> > --- a/fs/xfs/xfs_buf_item.c
> > +++ b/fs/xfs/xfs_buf_item.c
> > @@ -761,18 +761,25 @@ xfs_buf_item_init(
> >  	 * buffer. This makes the implementation as simple as possible.
> >  	 */
> >  	error = xfs_buf_item_get_format(bip, bp->b_map_count);
> > -	ASSERT(error == 0);
> > -	if (error) {	/* to stop gcc throwing set-but-unused warnings */
> > -		kmem_cache_free(xfs_buf_item_zone, bip);
> > -		return error;
> > +	if (error) {
> > +		xfs_err(mp, "could not allocate buffer item, err=%d", error);
> > +		goto err;
> >  	}
> 
> The error handling here is weird, as xfs_buf_item_get_format can't fail
> to start with.  I'd rather see a prep patch removing the bogus check
> for the kmem_zalloc and change the return value from
> xfs_buf_item_get_format to void.
> 
> Otherwise the patch looks good.

Ok.

--D