2014-06-15 13:12:12

by Peter Zijlstra

Subject: [PATCH 06/11] qspinlock: Optimize pending bit

XXX: merge into the pending bit patch..

It is possible so observe the pending bit without the locked bit when
the last owner has just released but the pending owner has not yet
taken ownership.

In this case we would normally queue -- because the pending bit is
already taken. However, in this case the pending bit is guaranteed to
be released 'soon', therefore wait for it and avoid queueing.

Signed-off-by: Peter Zijlstra <[email protected]>
---
kernel/locking/qspinlock.c | 10 ++++++++++
1 file changed, 10 insertions(+)

Index: linux-2.6/kernel/locking/qspinlock.c
===================================================================
--- linux-2.6.orig/kernel/locking/qspinlock.c
+++ linux-2.6/kernel/locking/qspinlock.c
@@ -226,6 +226,16 @@ void queue_spin_lock_slowpath(struct qsp
BUILD_BUG_ON(CONFIG_NR_CPUS >= (1U << _Q_TAIL_CPU_BITS));

/*
+ * wait for in-progress pending->locked hand-overs
+ *
+ * 0,1,0 -> 0,0,1
+ */
+ if (val == _Q_PENDING_VAL) {
+ while ((val = atomic_read(&lock->val)) == _Q_PENDING_VAL)
+ cpu_relax();
+ }
+
+ /*
* trylock || pending
*
* 0,0,0 -> 0,0,1 ; trylock

2014-06-18 11:43:16

by Paolo Bonzini

Subject: Re: [PATCH 06/11] qspinlock: Optimize pending bit

Il 15/06/2014 14:47, Peter Zijlstra ha scritto:
> XXX: merge into the pending bit patch..

Agree, or if not move it right after the pending bit patch, before the
NR_CPUS optimization.

Paolo

> It is possible so observe the pending bit without the locked bit when
> the last owner has just released but the pending owner has not yet
> taken ownership.
>
> In this case we would normally queue -- because the pending bit is
> already taken. However, in this case the pending bit is guaranteed to
> be released 'soon', therefore wait for it and avoid queueing.
>
> Signed-off-by: Peter Zijlstra <[email protected]>