Re: Scheduler Bug (set_cpus_allowed)

[Andrea Arcangeli]

Hello,

On Sat, Jun 08, 2002 at 01:20:04AM +0200, J.A. Magallon wrote:
> 
> On 2002.06.07 Robert Love wrote:
> >
> >Anyhow, with this issue, I guess we need to fix it... I'll send a patch
> >to Linus.
> >
> 
> Plz, could you also post it in the list ? -aa will need also...

So far I merged this Ingo's fix into my tree, I will overview the other
parts soon (as worse next week).

diff -urNp frozen-ref/include/linux/sched_runqueue.h frozen/include/linux/sched_runqueue.h
--- frozen-ref/include/linux/sched_runqueue.h	Tue Jun 11 03:03:35 2002
+++ frozen/include/linux/sched_runqueue.h	Tue Jun 11 03:00:46 2002
@@ -57,7 +57,6 @@ struct prio_array {
  */
 struct runqueue {
 	spinlock_t lock;
-	spinlock_t frozen;
 	unsigned long nr_running, nr_switches, expired_timestamp;
 	long quiescent; /* RCU */
 	task_t *curr, *idle;
diff -urNp frozen-ref/kernel/sched.c frozen/kernel/sched.c
--- frozen-ref/kernel/sched.c	Tue Jun 11 03:03:35 2002
+++ frozen/kernel/sched.c	Tue Jun 11 03:01:29 2002
@@ -352,7 +352,7 @@ void sched_exit(task_t * p)
 #if CONFIG_SMP
 asmlinkage void schedule_tail(task_t *prev)
 {
-	spin_unlock_irq(&this_rq()->frozen);
+	spin_unlock_irq(&this_rq()->lock);
 }
 #endif
 
@@ -762,9 +762,6 @@ switch_tasks:
 	if (likely(prev != next)) {
 		rq->nr_switches++;
 		rq->curr = next;
-		spin_lock(&rq->frozen);
-		spin_unlock(&rq->lock);
-
 		context_switch(prev, next);
 
 		/*
@@ -774,10 +771,8 @@ switch_tasks:
 		 */
 		smp_mb();
 		rq = this_rq();
-		spin_unlock_irq(&rq->frozen);
-	} else {
-		spin_unlock_irq(&rq->lock);
 	}
+	spin_unlock_irq(&rq->lock);
 
 	reacquire_kernel_lock(current);
 	if (need_resched())
@@ -1539,7 +1534,6 @@ void __init sched_init(void)
 		rq->active = rq->arrays;
 		rq->expired = rq->arrays + 1;
 		spin_lock_init(&rq->lock);
-		spin_lock_init(&rq->frozen);
 		INIT_LIST_HEAD(&rq->migration_queue);
 
 		for (j = 0; j < 2; j++) {

Andrea