Sasha, I actually hit these when I was testing with a low OSM_UMAD_MAX_PENDING value. I was a bit unsure of what to do in these cases if p_physp was NULL. What I do in this patch seems reasonable as the fabric does route but perhaps you could double check me?
Thanks,
Ira
>From 310a6cceca946fcc86f7bec28dfdeba77d011fc5 Mon Sep 17 00:00:00 2001
From: Ira K. Weiny <[EMAIL PROTECTED]>
Date: Wed, 20 Feb 2008 13:53:51 -0800
Subject: [PATCH] Fix 2 potential core dumps now that osm_node_get_physp_ptr can
return NULL
Signed-off-by: Ira K. Weiny <[EMAIL PROTECTED]>
---
opensm/opensm/osm_node.c | 2 ++
opensm/opensm/osm_ucast_updn.c | 4 ++++
2 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/opensm/opensm/osm_node.c b/opensm/opensm/osm_node.c
index 85ea3c9..843f7c8 100644
--- a/opensm/opensm/osm_node.c
+++ b/opensm/opensm/osm_node.c
@@ -285,6 +285,8 @@ osm_node_t *osm_node_get_remote_node(IN osm_node_t * const
p_node,
osm_physp_t *p_remote_physp;
p_physp = osm_node_get_physp_ptr(p_node, port_num);
+ if (!p_physp)
+ return (NULL);
if (!osm_physp_has_any_link(p_physp))
return (NULL);
diff --git a/opensm/opensm/osm_ucast_updn.c b/opensm/opensm/osm_ucast_updn.c
index 76b94cb..a7d61aa 100644
--- a/opensm/opensm/osm_ucast_updn.c
+++ b/opensm/opensm/osm_ucast_updn.c
@@ -377,6 +377,10 @@ updn_subn_rank(IN unsigned num_guids,
/* Current port fetched in order to get remote side */
p_physp =
osm_node_get_physp_ptr(p_sw->p_node, port_num);
+
+ if (!p_physp)
+ continue;
+
p_remote_physp = p_physp->p_remote_physp;
/*
--
1.5.1
0002-Fix-2-potential-core-dumps-now-that-osm_node_get_phy.patch
Description: Binary data
_______________________________________________ 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
