On Monday 18 February 2008 20:28, Michael Brooks wrote: > Commit 2aec5c602c6a44e2a3a173339a9ab94549658e4b on 2008-02-05 modified > the SA notification patch to use alloc_mad() from within > sa_query.c:ib_sa_informinfo_query(), but didn't also modify the function > to use the associated free_mad() call on exit, resulting in a lost > kref_put() on the AH structure. > > Patch follows: > > diff --git a/kernel_patches/fixes/sean_local_sa_1_notifications.patch > b/kernel_patches/fixes/sean_local_sa_1_notifications.patch > index 9d272ff..e710897 100644 > --- a/kernel_patches/fixes/sean_local_sa_1_notifications.patch > +++ b/kernel_patches/fixes/sean_local_sa_1_notifications.patch > @@ -1061,7 +1061,7 @@ Index: > ofa_1_3_dev_kernel/drivers/infiniband/core/sa_query.c > +err2: > + *sa_query = NULL; > + ib_sa_client_put(query->sa_query.client); > -+ ib_free_send_mad(query->sa_query.mad_buf); > ++ free_mad(&query->sa_query); > +err1: > + kfree(query); > + return ret;
Thanks for catching this one before RC5!! - Jack _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
