Skip to contentSkip to conversation

sched: teach migrate_disable about atomic contexts

PATCH In review Peter Zijlstra 2 participants 2 messages Sep 2, 2011 – Sep 6, 2011 Scheduler

lore
Share & tools
lore

Peter Zijlstra

author maintainer
#0 lore raw

PATCH 1/1 sched: teach migrate_disable about atomic contexts

Subject: sched: teach migrate_disable about atomic contexts
From: Peter Zijlstra <[email protected]>
Date: Fri Sep 02 14:29:27 CEST 2011

 <NMI>  [<ffffffff812dafd8>] spin_bug+0x94/0xa8
 [<ffffffff812db07f>] do_raw_spin_lock+0x43/0xea
 [<ffffffff814fa9be>] _raw_spin_lock_irqsave+0x6b/0x85
 [<ffffffff8106ff9e>] ? migrate_disable+0x75/0x12d
 [<ffffffff81078aaf>] ? pin_current_cpu+0x36/0xb0
 [<ffffffff8106ff9e>] migrate_disable+0x75/0x12d
 [<ffffffff81115b9d>] pagefault_disable+0xe/0x1f
 [<ffffffff81047027>] copy_from_user_nmi+0x74/0xe6
 [<ffffffff810489d7>] perf_callchain_user+0xf3/0x135

Now clearly we can't go around taking locks from NMI context, cure this by short-circuiting migrate_disable() when we're in an atomic context already.

Add some extra debugging to avoid things like:

  preempt_disable()
  migrate_disable();

  preempt_enable();
  migrate_enable();
Signed-off-by Peter Zijlstra <[email protected]>
Link http://lkml.kernel.org/n/[email protected] open here
---
 include/linux/sched.h |    3 +++
 kernel/sched.c        |   21 +++++++++++++++++++++
 2 files changed, 24 insertions(+)
kernel/sched.c +21 -0
Index: linux-2.6/kernel/sched.c
===================================================================
@@ -6135,6 +6135,17 @@ void migrate_disable(void)
 	unsigned long flags;
 	struct rq *rq;
 
+	if (in_atomic()) {
+#ifdef CONFIG_SCHED_DEBUG
+		p->migrate_disable_atomic++;
+#endif
+		return;
+	}
+
+#ifdef CONFIG_SCHED_DEBUG
+	WARN_ON_ONCE(p->migrate_disable_atomic);
+#endif
+
 	preempt_disable();
 	if (p->migrate_disable) {
 		p->migrate_disable++;

@@ -6183,6 +6194,16 @@ void migrate_enable(void)
 	unsigned long flags;
 	struct rq *rq;
 
+	if (in_atomic()) {
+#ifdef CONFIG_SCHED_DEBUG
+		p->migrate_disable_atomic--;
+#endif
+		return;
+	}
+
+#ifdef CONFIG_SCHED_DEBUG
+	WARN_ON_ONCE(p->migrate_disable_atomic);
+#endif
 	WARN_ON_ONCE(p->migrate_disable <= 0);
 
 	preempt_disable();
include/linux/sched.h +3 -0
Index: linux-2.6/include/linux/sched.h
===================================================================
@@ -1264,6 +1264,9 @@ struct task_struct {
 	unsigned int policy;
 #ifdef CONFIG_PREEMPT_RT_FULL
 	int migrate_disable;
+#ifdef CONFIG_SCHED_DEBUG
+	int migrate_disable_atomic;
+#endif
 #endif
 	cpumask_t cpus_allowed;
Tested-by on v1 1/1
#1 lore raw

Em Fri, Sep 02, 2011 at 02:41:37PM +0200, Peter Zijlstra escreveu:

Quoted 20 lines

Subject: sched: teach migrate_disable about atomic contexts
From: Peter Zijlstra <[email protected]>
Date: Fri Sep 02 14:29:27 CEST 2011

 <NMI>  [<ffffffff812dafd8>] spin_bug+0x94/0xa8
 [<ffffffff812db07f>] do_raw_spin_lock+0x43/0xea
 [<ffffffff814fa9be>] _raw_spin_lock_irqsave+0x6b/0x85
 [<ffffffff8106ff9e>] ? migrate_disable+0x75/0x12d
 [<ffffffff81078aaf>] ? pin_current_cpu+0x36/0xb0
 [<ffffffff8106ff9e>] migrate_disable+0x75/0x12d
 [<ffffffff81115b9d>] pagefault_disable+0xe/0x1f
 [<ffffffff81047027>] copy_from_user_nmi+0x74/0xe6
 [<ffffffff810489d7>] perf_callchain_user+0xf3/0x135

Now clearly we can't go around taking locks from NMI context, cure this by short-circuiting migrate_disable() when we're in an atomic context already.

Add some extra debugging to avoid things like:

Quoted 8 lines
  preempt_disable()
  migrate_disable();

  preempt_enable();
  migrate_enable();
Signed-off-by Peter Zijlstra <[email protected]>
Link http://lkml.kernel.org/n/[email protected] open here

Now I can collect callchains with perf on linux RT, thanks.

Tested-by Arnaldo Carvalho de Melo <[email protected]>

- Arnaldo