linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Ingo Molnar <mingo@elte.hu>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: "Zhang, Yanmin" <yanmin_zhang@linux.intel.com>,
	Andi Kleen <andi@firstfloor.org>, Matthew Wilcox <matthew@wil.cx>,
	LKML <linux-kernel@vger.kernel.org>,
	Alexander Viro <viro@ftp.linux.org.uk>,
	Andrew Morton <akpm@linux-foundation.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	"H. Peter Anvin" <hpa@zytor.com>
Subject: [patch] speed up / fix the new generic semaphore code (fix AIM7 40% regression with 2.6.26-rc1)
Date: Thu, 8 May 2008 14:01:30 +0200	[thread overview]
Message-ID: <20080508120130.GA2860@elte.hu> (raw)
In-Reply-To: <alpine.LFD.1.10.0805072115190.3024@woody.linux-foundation.org>


* Linus Torvalds <torvalds@linux-foundation.org> wrote:

> > > That said, "idle 0%" is easy when you spin. Do you also have 
> > > actual performance numbers?
> >
> > Yes. My conclusion is based on the actual number. cpu idle 0% is 
> > just a behavior it should be.
> 
> Thanks, that's all I wanted to verify.
> 
> I'll leave this overnight, and see if somebody has come up with some 
> smart and wonderful patch. And if not, I think I'll apply mine as 
> "known to fix a regression", and we can perhaps then improve on things 
> further from there.

hey, i happen to have such a smart and wonderful patch =B-)

i reproduced the AIM7 workload and can confirm Yanmin's findings that 
-.26-rc1 regresses over .25 - by over 67% here.

Looking at the workload i found and fixed what i believe to be the real 
bug causing the AIM7 regression: it was inefficient wakeup / scheduling 
/ locking behavior of the new generic semaphore code, causing suboptimal 
performance.

The problem comes from the following code. The new semaphore code does 
this on down():

        spin_lock_irqsave(&sem->lock, flags);
        if (likely(sem->count > 0))
                sem->count--;
        else
                __down(sem);
        spin_unlock_irqrestore(&sem->lock, flags);

and this on up():

        spin_lock_irqsave(&sem->lock, flags);
        if (likely(list_empty(&sem->wait_list)))
                sem->count++;
        else
                __up(sem);
        spin_unlock_irqrestore(&sem->lock, flags);

where __up() does:

        list_del(&waiter->list);
        waiter->up = 1;
        wake_up_process(waiter->task);

and where __down() does this in essence:

        list_add_tail(&waiter.list, &sem->wait_list);
        waiter.task = task;
        waiter.up = 0;
        for (;;) {
                [...]
                spin_unlock_irq(&sem->lock);
                timeout = schedule_timeout(timeout);
                spin_lock_irq(&sem->lock);
                if (waiter.up)
                        return 0;
        }

the fastpath looks good and obvious, but note the following property of 
the contended path: if there's a task on the ->wait_list, the up() of 
the current owner will "pass over" ownership to that waiting task, in a 
wake-one manner, via the waiter->up flag and by removing the waiter from 
the wait list.

That is all and fine in principle, but as implemented in 
kernel/semaphore.c it also creates a nasty, hidden source of contention!

The contention comes from the following property of the new semaphore 
code: the new owner owns the semaphore exclusively, even if it is not 
running yet.

So if the old owner, even if just a few instructions later, does a 
down() [lock_kernel()] again, it will be blocked and will have to wait 
on the new owner to eventually be scheduled (possibly on another CPU)! 
Or if another other task gets to lock_kernel() sooner than the "new 
owner" scheduled, it will be blocked unnecessarily and for a very long 
time when there are 2000 tasks running.

I.e. the implementation of the new semaphores code does wake-one and 
lock ownership in a very restrictive way - it does not allow 
opportunistic re-locking of the lock at all and keeps the scheduler from 
picking task order intelligently.

This kind of scheduling, with 2000 AIM7 processes running, creates awful 
cross-scheduling between those 2000 tasks, causes reduced parallelism, a 
throttled runqueue length and a lot of idle time. With increasing number 
of CPUs it causes an exponentially worse behavior in AIM7, as the chance 
for a newly woken new-owner task to actually run anytime soon is less 
and less likely.

Note that it takes just a tiny bit of contention for the 'new-semaphore 
catastrophy' to happen: the wakeup latencies get added to whatever small 
contention there is, and quickly snowball out of control!

I believe Yanmin's findings and numbers support this analysis too.

The best fix for this problem is to use the same scheduling logic that 
the kernel/mutex.c code uses: keep the wake-one behavior (that is OK and 
wanted because we do not want to over-schedule), but also allow 
opportunistic locking of the lock even if a wakee is already "in 
flight".

The patch below implements this new logic. With this patch applied the 
AIM7 regression is largely fixed on my quad testbox:

  # v2.6.25 vanilla:
  ..................
  Tasks   Jobs/Min        JTI     Real    CPU     Jobs/sec/task
  2000    56096.4         91      207.5   789.7   0.4675
  2000    55894.4         94      208.2   792.7   0.4658

  # v2.6.26-rc1-166-gc0a1811 vanilla:
  ...................................
  Tasks   Jobs/Min        JTI     Real    CPU     Jobs/sec/task
  2000    33230.6         83      350.3   784.5   0.2769
  2000    31778.1         86      366.3   783.6   0.2648

  # v2.6.26-rc1-166-gc0a1811 + semaphore-speedup:
  ...............................................
  Tasks   Jobs/Min        JTI     Real    CPU     Jobs/sec/task
  2000    55707.1         92      209.0   795.6   0.4642
  2000    55704.4         96      209.0   796.0   0.4642

i.e. a 67% speedup. We are now back to within 1% of the v2.6.25 
performance levels and have zero idle time during the test, as expected.

Btw., interactivity also improved dramatically with the fix - for 
example console-switching became almost instantaneous during this 
workload (which after all is running 2000 tasks at once!), without the 
patch it was stuck for a minute at times.

I also ran Linus's spinlock-BKL patch as well:

  # v2.6.26-rc1-166-gc0a1811 + Linus-BKL-spinlock-patch:
  ......................................................
  Tasks   Jobs/Min        JTI     Real    CPU     Jobs/sec/task
  2000    55889.0         92      208.3   793.3   0.4657
  2000    55891.7         96      208.3   793.3   0.4658

it is about 0.3% faster - but note that that is within the general noise 
levels of this test. I'd expect Linus's spinlock-BKL patch to give some 
small speedup because the BKL acquire times are short and 2000 tasks 
running all at once really increases the context-switching cost and most 
BKL contentions are within the cost of context-switching.

But i believe the better solution for that is to remove BKL use from all 
hotpaths, not to hide some of its costs by reintroducing it as a 
spinlock. Reintroducing the spinlock based BKL would have other 
disadvantages as well: it could reintroduce per-CPU-ness assumptions in 
BKL-using code and other complications as well. It's also not a very 
realistic workload - with 2000 tasks running the system was barely 
serviceable.

I'd much rather make BKL costs more apparent and more visible - but 50% 
regression was of course too much. But 0.3% for a 2000-tasks workload, 
which is near the noise level ... is acceptable i think - especially as 
this discussion has now reinvigorated the remove-the-BKL discussions and 
patches.

Linus, we can do your spinlock-BKL patch too if you feel strongly about 
it, but i'd rather not - we fought so hard for the preemptible BKL :-)

The spinlock-based-BKL patch only worked around the real problem i 
believe, because it eliminated the use of the suboptimal new semaphore 
code: with spinlocks there's no scheduling at all, so the wakeup/locking 
bug of the new semaphore code did not apply. It was not about any 
fastpath overhead AFAICS. [we'd have seen that with the 
CONFIG_PREEMPT_BKL=y code as well, which has been the default setting 
since v2.6.8.]

There's another nice side-effect of this speedup patch, the new generic 
semaphore code got even smaller:

   text    data     bss     dec     hex filename
   1241       0       0    1241     4d9 semaphore.o.before
   1207       0       0    1207     4b7 semaphore.o.after

(because the waiter.up complication got removed.)

Longer-term we should look into using the mutex code for the generic 
semaphore code as well - but i's not easy due to legacies and it's 
outside of the scope of v2.6.26 and outside the scope of this patch as 
well.

Hm?

	Ingo

Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/semaphore.c |   36 ++++++++++++++++--------------------
 1 file changed, 16 insertions(+), 20 deletions(-)

Index: linux/kernel/semaphore.c
===================================================================
--- linux.orig/kernel/semaphore.c
+++ linux/kernel/semaphore.c
@@ -54,10 +54,9 @@ void down(struct semaphore *sem)
 	unsigned long flags;
 
 	spin_lock_irqsave(&sem->lock, flags);
-	if (likely(sem->count > 0))
-		sem->count--;
-	else
+	if (unlikely(sem->count <= 0))
 		__down(sem);
+	sem->count--;
 	spin_unlock_irqrestore(&sem->lock, flags);
 }
 EXPORT_SYMBOL(down);
@@ -77,10 +76,10 @@ int down_interruptible(struct semaphore 
 	int result = 0;
 
 	spin_lock_irqsave(&sem->lock, flags);
-	if (likely(sem->count > 0))
-		sem->count--;
-	else
+	if (unlikely(sem->count <= 0))
 		result = __down_interruptible(sem);
+	if (!result)
+		sem->count--;
 	spin_unlock_irqrestore(&sem->lock, flags);
 
 	return result;
@@ -103,10 +102,10 @@ int down_killable(struct semaphore *sem)
 	int result = 0;
 
 	spin_lock_irqsave(&sem->lock, flags);
-	if (likely(sem->count > 0))
-		sem->count--;
-	else
+	if (unlikely(sem->count <= 0))
 		result = __down_killable(sem);
+	if (!result)
+		sem->count--;
 	spin_unlock_irqrestore(&sem->lock, flags);
 
 	return result;
@@ -157,10 +156,10 @@ int down_timeout(struct semaphore *sem, 
 	int result = 0;
 
 	spin_lock_irqsave(&sem->lock, flags);
-	if (likely(sem->count > 0))
-		sem->count--;
-	else
+	if (unlikely(sem->count <= 0))
 		result = __down_timeout(sem, jiffies);
+	if (!result)
+		sem->count--;
 	spin_unlock_irqrestore(&sem->lock, flags);
 
 	return result;
@@ -179,9 +178,8 @@ void up(struct semaphore *sem)
 	unsigned long flags;
 
 	spin_lock_irqsave(&sem->lock, flags);
-	if (likely(list_empty(&sem->wait_list)))
-		sem->count++;
-	else
+	sem->count++;
+	if (unlikely(!list_empty(&sem->wait_list)))
 		__up(sem);
 	spin_unlock_irqrestore(&sem->lock, flags);
 }
@@ -192,7 +190,6 @@ EXPORT_SYMBOL(up);
 struct semaphore_waiter {
 	struct list_head list;
 	struct task_struct *task;
-	int up;
 };
 
 /*
@@ -206,11 +203,11 @@ static inline int __sched __down_common(
 	struct task_struct *task = current;
 	struct semaphore_waiter waiter;
 
-	list_add_tail(&waiter.list, &sem->wait_list);
 	waiter.task = task;
-	waiter.up = 0;
 
 	for (;;) {
+		list_add_tail(&waiter.list, &sem->wait_list);
+
 		if (state == TASK_INTERRUPTIBLE && signal_pending(task))
 			goto interrupted;
 		if (state == TASK_KILLABLE && fatal_signal_pending(task))
@@ -221,7 +218,7 @@ static inline int __sched __down_common(
 		spin_unlock_irq(&sem->lock);
 		timeout = schedule_timeout(timeout);
 		spin_lock_irq(&sem->lock);
-		if (waiter.up)
+		if (sem->count > 0)
 			return 0;
 	}
 
@@ -259,6 +256,5 @@ static noinline void __sched __up(struct
 	struct semaphore_waiter *waiter = list_first_entry(&sem->wait_list,
 						struct semaphore_waiter, list);
 	list_del(&waiter->list);
-	waiter->up = 1;
 	wake_up_process(waiter->task);
 }

  reply	other threads:[~2008-05-08 12:02 UTC|newest]

Thread overview: 140+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-05-06  5:48 AIM7 40% regression with 2.6.26-rc1 Zhang, Yanmin
2008-05-06 11:18 ` Matthew Wilcox
2008-05-06 11:44 ` Ingo Molnar
2008-05-06 12:09   ` Matthew Wilcox
2008-05-06 16:23     ` Matthew Wilcox
2008-05-06 16:36       ` Linus Torvalds
2008-05-06 16:42         ` Matthew Wilcox
2008-05-06 16:39           ` Alan Cox
2008-05-06 16:51             ` Matthew Wilcox
2008-05-06 16:45               ` Alan Cox
2008-05-06 17:42               ` Linus Torvalds
2008-05-06 20:28           ` Linus Torvalds
2008-05-06 16:44         ` J. Bruce Fields
2008-05-06 17:21       ` Andrew Morton
2008-05-06 17:31         ` Matthew Wilcox
2008-05-06 17:49           ` Ingo Molnar
2008-05-06 18:07             ` Andrew Morton
2008-05-11 11:11               ` Matthew Wilcox
2008-05-06 17:39         ` Ingo Molnar
2008-05-07  6:49           ` Zhang, Yanmin
2008-05-06 17:45         ` Linus Torvalds
2008-05-07 16:38         ` Matthew Wilcox
2008-05-07 16:55           ` Linus Torvalds
2008-05-07 17:08             ` Linus Torvalds
2008-05-07 17:16               ` Andrew Morton
2008-05-07 17:27                 ` Linus Torvalds
2008-05-07 17:22               ` Ingo Molnar
2008-05-07 17:25                 ` Ingo Molnar
2008-05-07 17:31                 ` Linus Torvalds
2008-05-07 17:47                   ` Linus Torvalds
2008-05-07 17:49                   ` Ingo Molnar
2008-05-07 18:02                     ` Linus Torvalds
2008-05-07 18:17                       ` Ingo Molnar
2008-05-07 18:27                         ` Linus Torvalds
2008-05-07 18:43                           ` Ingo Molnar
2008-05-07 19:01                             ` Linus Torvalds
2008-05-07 19:09                               ` Ingo Molnar
2008-05-07 19:24                               ` Matthew Wilcox
2008-05-07 19:44                                 ` Linus Torvalds
2008-05-07 20:00                                   ` Oi. NFS people. Read this Matthew Wilcox
2008-05-07 22:10                                     ` Trond Myklebust
2008-05-09  1:43                                       ` J. Bruce Fields
2008-05-08  3:24       ` AIM7 40% regression with 2.6.26-rc1 Zhang, Yanmin
2008-05-08  3:34         ` Linus Torvalds
2008-05-08  4:37           ` Zhang, Yanmin
2008-05-08 14:58             ` Linus Torvalds
2008-05-07  2:11   ` Zhang, Yanmin
2008-05-07  3:41     ` Zhang, Yanmin
2008-05-07  3:59       ` Andrew Morton
2008-05-07  4:46         ` Zhang, Yanmin
2008-05-07  6:26       ` Ingo Molnar
2008-05-07  6:28         ` Ingo Molnar
2008-05-07  7:05           ` Zhang, Yanmin
2008-05-07 11:00       ` Andi Kleen
2008-05-07 11:46         ` Matthew Wilcox
2008-05-07 12:21           ` Andi Kleen
2008-05-07 14:36             ` Linus Torvalds
2008-05-07 14:35               ` Alan Cox
2008-05-07 15:00                 ` Linus Torvalds
2008-05-07 15:02                   ` Linus Torvalds
2008-05-07 14:57               ` Andi Kleen
2008-05-07 15:31                 ` Andrew Morton
2008-05-07 16:22                   ` Matthew Wilcox
2008-05-07 15:19               ` Linus Torvalds
2008-05-07 17:14                 ` Ingo Molnar
2008-05-08  2:44                 ` Zhang, Yanmin
2008-05-08  3:29                   ` Linus Torvalds
2008-05-08  4:08                     ` Zhang, Yanmin
2008-05-08  4:17                       ` Linus Torvalds
2008-05-08 12:01                         ` Ingo Molnar [this message]
2008-05-08 12:28                           ` [patch] speed up / fix the new generic semaphore code (fix AIM7 40% regression with 2.6.26-rc1) Ingo Molnar
2008-05-08 14:43                             ` Ingo Molnar
2008-05-08 15:10                               ` [git pull] scheduler fixes Ingo Molnar
2008-05-08 15:33                                 ` Adrian Bunk
2008-05-08 15:41                                   ` Ingo Molnar
2008-05-08 19:42                                     ` Adrian Bunk
2008-05-11 11:03                                 ` Matthew Wilcox
2008-05-11 11:14                                   ` Matthew Wilcox
2008-05-11 11:48                                   ` Matthew Wilcox
2008-05-11 12:50                                     ` Ingo Molnar
2008-05-11 12:52                                       ` Ingo Molnar
2008-05-11 13:02                                         ` Matthew Wilcox
2008-05-11 13:26                                           ` Matthew Wilcox
2008-05-11 14:00                                             ` Ingo Molnar
2008-05-11 14:18                                               ` Matthew Wilcox
2008-05-11 14:42                                                 ` Ingo Molnar
2008-05-11 14:48                                                   ` Matthew Wilcox
2008-05-11 15:19                                                     ` Ingo Molnar
2008-05-11 15:29                                                       ` Matthew Wilcox
2008-05-13 14:11                                                         ` Ingo Molnar
2008-05-13 14:21                                                           ` Matthew Wilcox
2008-05-13 14:42                                                             ` Ingo Molnar
2008-05-13 15:28                                                               ` Matthew Wilcox
2008-05-13 17:13                                                                 ` Ingo Molnar
2008-05-13 17:22                                                                   ` Linus Torvalds
2008-05-13 21:05                                                                     ` Ingo Molnar
2008-05-11 13:54                                           ` Ingo Molnar
2008-05-11 14:22                                             ` Matthew Wilcox
2008-05-11 14:32                                               ` Ingo Molnar
2008-05-11 14:46                                                 ` Matthew Wilcox
2008-05-11 16:47                                                 ` Linus Torvalds
2008-05-11 13:01                                   ` Ingo Molnar
2008-05-11 13:06                                     ` Matthew Wilcox
2008-05-11 13:45                                       ` Ingo Molnar
2008-05-11 14:10                                   ` Sven Wegener
2008-05-08 16:02                             ` [patch] speed up / fix the new generic semaphore code (fix AIM7 40% regression with 2.6.26-rc1) Linus Torvalds
2008-05-08 18:30                               ` Linus Torvalds
2008-05-08 20:19                                 ` Ingo Molnar
2008-05-08 20:27                                   ` Linus Torvalds
2008-05-08 21:45                                     ` Ingo Molnar
2008-05-08 22:02                                       ` Ingo Molnar
2008-05-08 22:55                                       ` Linus Torvalds
2008-05-08 23:07                                         ` Linus Torvalds
2008-05-08 23:14                                           ` Linus Torvalds
2008-05-08 23:16                                         ` Alan Cox
2008-05-08 23:33                                           ` Linus Torvalds
2008-05-08 23:27                                             ` Alan Cox
2008-05-09  6:50                                             ` Ingo Molnar
2008-05-09  8:29                                             ` Andi Kleen
2008-05-08 13:20                           ` Matthew Wilcox
2008-05-08 15:01                             ` Ingo Molnar
2008-05-08 13:56                           ` Arjan van de Ven
2008-05-08  6:43                   ` AIM7 40% regression with 2.6.26-rc1 Ingo Molnar
2008-05-08  6:48                     ` Andrew Morton
2008-05-08  7:14                     ` Zhang, Yanmin
2008-05-08  7:39                       ` Ingo Molnar
2008-05-08  8:44                         ` Zhang, Yanmin
2008-05-08  9:21                           ` Ingo Molnar
2008-05-08  9:29                             ` Ingo Molnar
2008-05-08  9:30                             ` Zhang, Yanmin
2008-05-07 16:20               ` Ingo Molnar
2008-05-07 16:35                 ` Linus Torvalds
2008-05-07 17:05                   ` Ingo Molnar
2008-05-07 17:24                     ` Linus Torvalds
2008-05-07 17:36                       ` Ingo Molnar
2008-05-07 17:55                         ` Linus Torvalds
2008-05-07 17:59                           ` Matthew Wilcox
2008-05-07 18:17                             ` Linus Torvalds
2008-05-07 18:49                               ` Ingo Molnar
2008-05-07 13:59         ` Alan Cox

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20080508120130.GA2860@elte.hu \
    --to=mingo@elte.hu \
    --cc=akpm@linux-foundation.org \
    --cc=andi@firstfloor.org \
    --cc=hpa@zytor.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=matthew@wil.cx \
    --cc=tglx@linutronix.de \
    --cc=torvalds@linux-foundation.org \
    --cc=viro@ftp.linux.org.uk \
    --cc=yanmin_zhang@linux.intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).