Re: 2.6.8.1-mm2

[Ingo Molnar <mingo@elte.hu>]

* Rusty Russell <rusty@rustcorp.com.au> wrote:

> Nathan, can you revert that, and apply this?  This actually fixes the
> might_sleep problem, and should fix at least the problem Vatsa saw. 
> If it doesn't solve your problem, we need to look again.

i've attached a much simpler replacement: dont allow CPU hotplug during
self-reap.

	Ingo

DESC

disable preemption in the self-reap codepath, as such tasks may not be
on the tasklist anymore and CPU-hotplug relies on the tasklist to
migrate tasks.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

--- linux/kernel/exit.c.orig	
+++ linux/kernel/exit.c	
@@ -25,6 +25,7 @@
 #include <linux/proc_fs.h>
 #include <linux/mempolicy.h>
 #include <linux/perfctr.h>
+#include <linux/cpu.h>
 
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
@@ -780,8 +781,14 @@ static void exit_notify(struct task_stru
 
 	/* If the process is dead, release it - nobody will wait for it */
 	if (state == TASK_DEAD) {
+		lock_cpu_hotplug();
 		release_task(tsk);
 		write_lock_irq(&tasklist_lock);
+		/*
+		 * No preemption may happen from this point on,
+		 * or CPU hotplug (and task exit) breaks:
+		 */
+		unlock_cpu_hotplug();
 		tsk->state = state;
 		_raw_write_unlock(&tasklist_lock);
 		local_irq_enable();