ipmi: handle run_to_completion properly in deliver_recv_msg()
authorJiri Kosina <jkosina@suse.cz>
Wed, 26 May 2010 21:43:53 +0000 (14:43 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 27 May 2010 16:12:50 +0000 (09:12 -0700)
If run_to_completion flag is set, it means that we are running in a
single-threaded mode, and thus no locks are held.

This fixes a deadlock when IPMI notifier is being called during panic.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Acked-by: Corey Minyard <minyard@acm.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
drivers/char/ipmi/ipmi_si_intf.c

index f052c481327a55062a29aedb353fcbf87a117842..35603dd4e6c5276911dec0b3a11464a152beb8a5 100644 (file)
@@ -323,9 +323,14 @@ static void deliver_recv_msg(struct smi_info *smi_info,
 {
        /* Deliver the message to the upper layer with the lock
           released. */
-       spin_unlock(&(smi_info->si_lock));
-       ipmi_smi_msg_received(smi_info->intf, msg);
-       spin_lock(&(smi_info->si_lock));
+
+       if (smi_info->run_to_completion) {
+               ipmi_smi_msg_received(smi_info->intf, msg);
+       } else {
+               spin_unlock(&(smi_info->si_lock));
+               ipmi_smi_msg_received(smi_info->intf, msg);
+               spin_lock(&(smi_info->si_lock));
+       }
 }
 
 static void return_hosed_msg(struct smi_info *smi_info, int cCode)