locking/qspinlock/x86: Fix performance regression under unaccelerated VMs
authorPeter Zijlstra <peterz@infradead.org>
Fri, 4 Sep 2015 15:25:23 +0000 (17:25 +0200)
committerIngo Molnar <mingo@kernel.org>
Fri, 11 Sep 2015 05:49:42 +0000 (07:49 +0200)
commit43b3f02899f74ae9914a39547cc5492156f0027a
tree0db5018c95f8d243b974f42abda1c0c618abeee5
parentedcd591c77a48da753456f92daf8bb50fe9bac93
locking/qspinlock/x86: Fix performance regression under unaccelerated VMs

Dave ran into horrible performance on a VM without PARAVIRT_SPINLOCKS
set and Linus noted that the test-and-set implementation was retarded.

One should spin on the variable with a load, not a RMW.

While there, remove 'queued' from the name, as the lock isn't queued
at all, but a simple test-and-set.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Dave Chinner <david@fromorbit.com>
Tested-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Waiman Long <Waiman.Long@hp.com>
Cc: stable@vger.kernel.org # v4.2+
Link: http://lkml.kernel.org/r/20150904152523.GR18673@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/include/asm/qspinlock.h
include/asm-generic/qspinlock.h
kernel/locking/qspinlock.c