>From e96262150a513ce3d54ff221d4ace8aeec96e0bf Mon Sep 17 00:00:00 2001
From: Tejun Heo <[email protected]>
Date: Wed, 12 Jun 2013 13:37:42 -0700
percpu_ref_get/put() are using preempt_disable/enable() while
percpu_ref_kill() is using plain call_rcu() instead of
call_rcu_sched(). This is buggy as grace periods of the two may not
match. Fix it by using plain RCU in percpu_ref_get/put().
(I suggested using sched RCU in the first place but there's no actual
benefit in doing so unless we're gonna introduce different variants
of get/put to be called while preemption is alredy disabled, which we
definitely shouldn't.)
Signed-off-by: Tejun Heo <[email protected]>
Reported-by: Rusty Russell <[email protected]>
---
include/linux/percpu-refcount.h | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/include/linux/percpu-refcount.h b/include/linux/percpu-refcount.h
index 24b31ef..abe1411 100644
--- a/include/linux/percpu-refcount.h
+++ b/include/linux/percpu-refcount.h
@@ -85,7 +85,7 @@ static inline void percpu_ref_get(struct percpu_ref *ref)
{
unsigned __percpu *pcpu_count;
- preempt_disable();
+ rcu_read_lock();
pcpu_count = ACCESS_ONCE(ref->pcpu_count);
@@ -94,7 +94,7 @@ static inline void percpu_ref_get(struct percpu_ref *ref)
else
atomic_inc(&ref->count);
- preempt_enable();
+ rcu_read_unlock();
}
/**
@@ -107,7 +107,7 @@ static inline void percpu_ref_put(struct percpu_ref *ref)
{
unsigned __percpu *pcpu_count;
- preempt_disable();
+ rcu_read_lock();
pcpu_count = ACCESS_ONCE(ref->pcpu_count);
@@ -116,7 +116,7 @@ static inline void percpu_ref_put(struct percpu_ref *ref)
else if (unlikely(atomic_dec_and_test(&ref->count)))
ref->release(ref);
- preempt_enable();
+ rcu_read_unlock();
}
#endif
--
1.8.2.1
On Wed, Jun 12, 2013 at 01:40:32PM -0700, Tejun Heo wrote:
> From e96262150a513ce3d54ff221d4ace8aeec96e0bf Mon Sep 17 00:00:00 2001
> From: Tejun Heo <[email protected]>
> Date: Wed, 12 Jun 2013 13:37:42 -0700
>
> percpu_ref_get/put() are using preempt_disable/enable() while
> percpu_ref_kill() is using plain call_rcu() instead of
> call_rcu_sched(). This is buggy as grace periods of the two may not
> match. Fix it by using plain RCU in percpu_ref_get/put().
>
> (I suggested using sched RCU in the first place but there's no actual
> benefit in doing so unless we're gonna introduce different variants
> of get/put to be called while preemption is alredy disabled, which we
> definitely shouldn't.)
>
> Signed-off-by: Tejun Heo <[email protected]>
> Reported-by: Rusty Russell <[email protected]>
Acked-by: Kent Overstreet <[email protected]>
On Wed, Jun 12, 2013 at 01:40:32PM -0700, Tejun Heo wrote:
> From e96262150a513ce3d54ff221d4ace8aeec96e0bf Mon Sep 17 00:00:00 2001
> From: Tejun Heo <[email protected]>
> Date: Wed, 12 Jun 2013 13:37:42 -0700
>
> percpu_ref_get/put() are using preempt_disable/enable() while
> percpu_ref_kill() is using plain call_rcu() instead of
> call_rcu_sched(). This is buggy as grace periods of the two may not
> match. Fix it by using plain RCU in percpu_ref_get/put().
>
> (I suggested using sched RCU in the first place but there's no actual
> benefit in doing so unless we're gonna introduce different variants
> of get/put to be called while preemption is alredy disabled, which we
> definitely shouldn't.)
>
> Signed-off-by: Tejun Heo <[email protected]>
> Reported-by: Rusty Russell <[email protected]>
Applied to percpu/for-3.11.
Thanks.
--
tejun