[PATCH] kthread: fix kthread_bind() race

[KOSAKI Motohiro]

Peter Zijlstra reported kthread_bind() has a race. It doesn't hold
any locks and set PF_THREAD_BOUND _after_ setting cpus_allowed. So,
following race can be happen.

  CPU0                           CPU1
  ----------------------------------------------------
  do_set_cpus_allowed()
				sys_sched_setaffinity()
  p->flags |= PF_THREAD_BOUND;

The solution is to take either rq lock or pi_lock. They prevent
a race because set_cpus_allowed_ptr() take both locks. This patch
choose to use latter way.

Reported-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
---
 kernel/kthread.c |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/kernel/kthread.c b/kernel/kthread.c
index 4ba7ccc..92e3083 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -196,15 +196,20 @@ EXPORT_SYMBOL(kthread_create_on_node);
  */
 void kthread_bind(struct task_struct *p, unsigned int cpu)
 {
+	unsigned long flags;
+
 	/* Must have done schedule() in kthread() before we set_task_cpu */
 	if (!wait_task_inactive(p, TASK_UNINTERRUPTIBLE)) {
 		WARN_ON(1);
 		return;
 	}

+	/* protect from a race against set_cpus_allowed_ptr() */
+	raw_spin_lock_irqsave(&p->pi_lock, flags);
 	/* It's safe because the task is inactive. */
 	do_set_cpus_allowed(p, cpumask_of(cpu));
 	p->flags |= PF_THREAD_BOUND;
+	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
 }
 EXPORT_SYMBOL(kthread_bind);

-- 
1.7.3.1




--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/