From: Prarit Bhargava <prarit@redhat.com>
Date: Fri, 5 Nov 2010 18:08:36 -0400
Subject: [kernel] add stop_machine barrier to fix lock contention
Message-id: <4CD44824.9040307@redhat.com>
Patchwork-id: 29054
O-Subject: Re: [RHEL5 BZ 634454 PATCH]: add barrier in stop_machine() to fix
	lock	contention
Bugzilla: 634454
RH-Acked-by: Vivek Goyal <vgoyal@redhat.com>
RH-Acked-by: Don Dutile <ddutile@redhat.com>

Add smp_mb() to stop_machine() after yield() is called.  This avoids a
potential deadlocks over locks when stop_machine() is called.

There is a detailed explanation of how this arose in the code in the BZ.

Successfully tested by me on an AMD Dinar system and an Intel Nehalem-EP
(see comments 97 and 99 in the BZ).

Resolves BZ 634454.

Signed-off-by: Jarod Wilson <jarod@redhat.com>

diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c
index 618363a..ff6dd9f 100644
--- a/kernel/stop_machine.c
+++ b/kernel/stop_machine.c
@@ -102,8 +102,12 @@ static int stop_machine(void)
 	}
 
 	/* Wait for them all to come to life. */
-	while (atomic_read(&stopmachine_thread_ack) != stopmachine_num_threads)
+	while (atomic_read(&stopmachine_thread_ack) != stopmachine_num_threads){
 		yield();
+		/* add in a barrier to avoid kstopmachine
+		   causing lock starvation issues for other threads */
+		smp_mb();
+	}
 
 	/* If some failed, kill them all. */
 	if (ret < 0) {