When in master state and performing light sweeps, openSM ignores
other subnet managers in master state and fails to perform
handovers or relinquish control. Handovers and relinquishing
master is only performed after a heavy sweep. This can result
in two subnet managers being in master state, seeing each other
SM in master state, but both choose to do nothing about it.
This patch initiates a heavy sweep when another master subnet
manager is found during a light sweep. This is sufficient to
start a handover or relinquish scenario.
Signed-off-by: Albert Chu <ch...@llnl.gov>
---
opensm/osm_sminfo_rcv.c | 18 ++++++++++++++++--
1 files changed, 16 insertions(+), 2 deletions(-)
diff --git a/opensm/osm_sminfo_rcv.c b/opensm/osm_sminfo_rcv.c
index 66eb886..d4ae7c9 100644
--- a/opensm/osm_sminfo_rcv.c
+++ b/opensm/osm_sminfo_rcv.c
@@ -385,8 +385,22 @@ static void smi_rcv_process_get_sm(IN osm_sm_t * sm,
osm_sm_state_mgr_signal_master_is_alive(sm);
else {
/* This is a response we got while sweeping the
subnet.
- We will handle a case of handover needed
later on, when the sweep
- is done and all SMs are recongnized. */
+ *
+ * If this is during a heavy sweep, we will
handle a case of
+ * handover needed later on, when the sweep is
done and all
+ * SMs are recognized.
+ *
+ * If this is during a light sweep, initiate a
heavy sweep
+ * to initiate handover scenarios.
+ *
+ * Note that it does not matter if the remote
SM is lower
+ * or higher priority. If it is lower
priority, we must
+ * wait for it HANDOVER. If it is higher
priority, we need
+ * to HANDOVER to it. Both cases are handled
after doing
+ * a heavy sweep.
+ */
+ if (light_sweep)
+ sm->p_subn->force_heavy_sweep = TRUE;
}
break;
case IB_SMINFO_STATE_STANDBY:
--
1.7.1
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
