[Cluster-devel] [bug report] GFS2: Use resizable hash table for glocks

[Dan Carpenter <dan.carpenter@oracle.com>]

On Wed, Dec 14, 2016 at 08:41:57AM -0500, Bob Peterson wrote:
> Hi Dan,
> 
> ----- Original Message -----
> | Hello Bob Peterson,
> | 
> | The patch 88ffbf3e037e: "GFS2: Use resizable hash table for glocks"
> | from Mar 16, 2015, leads to the following static checker warning:
> | 
> | 	fs/gfs2/glock.c:1813 gfs2_glock_iter_next()
> | 	error: 'gi->gl' dereferencing possible ERR_PTR()
> | 
> | fs/gfs2/glock.c
> |   1803  static void gfs2_glock_iter_next(struct gfs2_glock_iter *gi)
> |   1804  {
> |   1805          do {
> |   1806                  gi->gl = rhashtable_walk_next(&gi->hti);
> |   1807                  if (IS_ERR(gi->gl)) {
> |   1808                          if (PTR_ERR(gi->gl) == -EAGAIN)
> |   1809                                  continue;
> | 
> | This continue was probably intended to go to the top of the loop, but
> | it's a do while loop so it actually drops down
> | 
> |   1810                          gi->gl = NULL;
> |   1811                  }
> |   1812          /* Skip entries for other sb and dead entries */
> |   1813          } while ((gi->gl) && ((gi->sdp != gi->gl->gl_name.ln_sbd) ||
> |                                                   ^^^^^^^^
> | to here where we dereference gi->gl.  It's weird that Smatch is only
> | complaining about this now though...
> | 
> |   1814
> |   __lockref_is_dead(&gi->gl->gl_lockref)));
> |   1815  }
> | 
> | regards,
> | dan carpenter
>  
> Yes, that looks like a bug. Do you have a patch or should I patch it?
> 

Could you patch it?

regards,
dan carpenter