[PATCH 1/2] PF_DEAD: cleanup usage

[Oleg Nesterov <oleg@tv-sign.ru>]

schedule() checks PF_DEAD on every context switch, and sets ->state = EXIT_DEAD
to ensure that exited task will be deactivated.

I think it is better to set EXIT_DEAD in do_exit(), along with PF_DEAD flag.

It is safe to do without task_rq() locking, because concurrent try_to_wake_up()
can't change task's ->state: the 'state' argument of try_to_wake_up() can't have
EXIT_DEAD bit. And in case when try_to_wake_up() sees stale value of ->state ==
TASK_RUNNING it will do nothing.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>

--- 2.6.15-rc2/kernel/exit.c~4_DEAD	2005-11-23 19:33:27.000000000 +0300
+++ 2.6.15-rc2/kernel/exit.c	2005-11-24 19:14:29.000000000 +0300
@@ -875,6 +875,7 @@ fastcall NORET_TYPE void do_exit(long co
 	preempt_disable();
 	BUG_ON(tsk->flags & PF_DEAD);
 	tsk->flags |= PF_DEAD;
+	tsk->state = EXIT_DEAD;
 
 	schedule();
 	BUG();
--- 2.6.15-rc2/kernel/sched.c~4_DEAD	2005-11-22 19:35:52.000000000 +0300
+++ 2.6.15-rc2/kernel/sched.c	2005-11-24 19:13:34.000000000 +0300
@@ -3000,9 +3000,6 @@ need_resched_nonpreemptible:
 
 	spin_lock_irq(&rq->lock);
 
-	if (unlikely(prev->flags & PF_DEAD))
-		prev->state = EXIT_DEAD;
-
 	switch_count = &prev->nivcsw;
 	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
 		switch_count = &prev->nvcsw;