[ofa-general] ofed1.2.5rc2 and intel mpi error

[Jack Morgenstein]

On Tuesday 05 February 2008 19:17, Sean Hefty wrote:
> >There was a bug in the local sa implementation (the informinfo query did
> >not incorporate a required change).
> >
> >This bug has been fixed, and the fix will be in the next OFED daily build.
> >
> >The fix will also be in RC4.
> 
> Thanks, Jack
> 
Thanks for not getting annoyed.  The fix was a simple one, and I wanted the fix
to make it into OFED 1.3 RC4.

- Jack

P.S., this change is also needed for anyone using the driver built in to kernels 2.6.23 and above.

P.P.S., I'll post the changed patch now.  I used quilt, though, to generate the patch (rather
than git) -- it was MUCH easier for me, but the side effect is many cosmetic changes.

The only relevant change though is (in patch file 
kernel_patches/fixes/sean_local_sa_1_notifications.patch):


@@ -1025,13 +1030,10 @@ index 6469406..369fe60 100644
 +      if (!query)
 +              return -ENOMEM;
 +
-+      query->sa_query.mad_buf = ib_create_send_mad(agent, 1, 0,
-+                                                   0, IB_MGMT_SA_HDR,
-+                                                   IB_MGMT_SA_DATA, gfp_mask);
-+      if (!query->sa_query.mad_buf) {
-+              ret = -ENOMEM;
++      query->sa_query.port     = port;
++      ret = alloc_mad(&query->sa_query, gfp_mask);
++      if (ret)
 +              goto err1;
-+      }