Skip to contentSkip to conversation

CFS vs cpu hotplug

BUG Applied Dmitry Adamushko 3 participants 3 messages Jul 9, 2008 – Jul 10, 2008

lore
Share & tools
lore
#0 lore raw

hm, while looking at this code again...

Ingo,

I think we may have a race between try_to_wake_up() and migrate_live_tasks() -> move_task_off_dead_cpu() when the later one may end up looping endlessly.

Subject: sched: prevent a potentially endless loop in move_task_off_dead_cpu()

Interrupts are enabled on other CPUs when migration_call(CPU_DEAD, ...) is called so we may get a race between try_to_wake_up() and migrate_live_tasks() -> move_task_off_dead_cpu(). The former one may push a task out of a dead CPU causing the later one to loop endlessly.

---

kernel/sched.c +3 -1
index 94ead43..9397b87 100644
@@ -5621,8 +5621,10 @@ static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
 
 	double_rq_lock(rq_src, rq_dest);
 	/* Already moved. */
-	if (task_cpu(p) != src_cpu)
+	if (task_cpu(p) != src_cpu) {
+		ret = 1;
 		goto out;
+	}
 	/* Affinity changed (again). */
 	if (!cpu_isset(dest_cpu, p->cpus_allowed))
 		goto out;

---

#1 lore raw
Quoted 10 lines · Dmitry Adamushko

hm, while looking at this code again...

Ingo,

I think we may have a race between try_to_wake_up() and migrate_live_tasks() -> move_task_off_dead_cpu() when the later one may end up looping endlessly.

Quoted 5 lines

Subject: sched: prevent a potentially endless loop in move_task_off_dead_cpu()

Interrupts are enabled on other CPUs when migration_call(CPU_DEAD, ...) is called so we may get a race between try_to_wake_up() and migrate_live_tasks() -> move_task_off_dead_cpu(). The former one may push a task out of a dead CPU causing the later one to loop endlessly.

That's exactly what explains a dump I got yesterday. Thanks for fixing! :)

Will apply your patch and let you know if it fixes the problem. (may take until next week unfortunately).

Quoted 22 lines

---

kernel/sched.c +3 -2
index 94ead43..9397b87 100644
@@ -5621,8 +5621,10 @@ static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
 
 	double_rq_lock(rq_src, rq_dest);
 	/* Already moved. */
-	if (task_cpu(p) != src_cpu)
+	if (task_cpu(p) != src_cpu) {
+		ret = 1;
 		goto out;
+	}
 	/* Affinity changed (again). */
 	if (!cpu_isset(dest_cpu, p->cpus_allowed))
 		goto out;
 
---
Applied
#2 lore raw
Quoted 11 lines · Heiko Carstens
Quoted 8 lines

Subject: sched: prevent a potentially endless loop in move_task_off_dead_cpu()

Interrupts are enabled on other CPUs when migration_call(CPU_DEAD, ...) is called so we may get a race between try_to_wake_up() and migrate_live_tasks() -> move_task_off_dead_cpu(). The former one may push a task out of a dead CPU causing the later one to loop endlessly.

That's exactly what explains a dump I got yesterday. Thanks for fixing! :)

applied to tip/sched/urgent via the commit below - lets see whether we can still get it into v2.6.26.

Ingo

---------------->
commit dc7fab8b3bb388c57c6c4a43ba68c8a32ca25204

Author: Dmitry Adamushko <[email protected]>
Date:   Thu Jul 10 00:32:40 2008 +0200

sched: fix cpu hotplug

I think we may have a race between try_to_wake_up() and migrate_live_tasks() -> move_task_off_dead_cpu() when the later one may end up looping endlessly.

Interrupts are enabled on other CPUs when migration_call(CPU_DEAD, ...) is called so we may get a race between try_to_wake_up() and migrate_live_tasks() -> move_task_off_dead_cpu(). The former one may push a task out of a dead CPU causing the later one to loop endlessly.

Heiko Carstens observed:

| That's exactly what explains a dump I got yesterday. Thanks for fixing! :)

Cc Lai Jiangshan <[email protected]>
Cc Heiko Carstens <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Avi Kivity <[email protected]>
Cc Andrew Morton <[email protected]>
kernel/sched.c +3 -1
index 94ead43..9397b87 100644
@@ -5621,8 +5621,10 @@ static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
 
 	double_rq_lock(rq_src, rq_dest);
 	/* Already moved. */
-	if (task_cpu(p) != src_cpu)
+	if (task_cpu(p) != src_cpu) {
+		ret = 1;
 		goto out;
+	}
 	/* Affinity changed (again). */
 	if (!cpu_isset(dest_cpu, p->cpus_allowed))
 		goto out;