[openib-general] [patch] mad.c, agent.c spinlocking on UP

[Roland Dreier]

    Sean> I'd vote to remove the BUG_ONs, versus adding #ifdef.

That seems fine to me.  Maybe adding a comment in agent.c similar to
what mad.c says ("Assumes ib_mad_port_list_lock is being held") is all
we really need, something like this:

Index: agent.c
===================================================================
--- agent.c	(revision 1249)
+++ agent.c	(working copy)
@@ -36,6 +36,9 @@
 extern kmem_cache_t *ib_mad_cache;
 
 
+/*
+ * Caller must hold ib_agent_port_list_lock.
+ */
 static inline struct ib_agent_port_private *
 __ib_get_agent_port(struct ib_device *device, int port_num,
 		    struct ib_mad_agent *mad_agent)