mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Waiman Long <longman@redhat.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: will.deacon@arm.com, mingo@kernel.org,
	linux-kernel@vger.kernel.org, andrea.parri@amarulasolutions.com,
	tglx@linutronix.de
Subject: Re: [RFC][PATCH 0/3] locking/qspinlock: Improve determinism for x86
Date: Wed, 26 Sep 2018 19:21:49 -0400	[thread overview]
Message-ID: <5c492b85-365f-e64d-2230-f479903591d8@redhat.com> (raw)
In-Reply-To: <20180926175112.GA5254@hirez.programming.kicks-ass.net>

On 09/26/2018 01:51 PM, Peter Zijlstra wrote:
> On Wed, Sep 26, 2018 at 12:20:09PM -0400, Waiman Long wrote:
>> On 09/26/2018 07:01 AM, Peter Zijlstra wrote:
>>> Back when Will did his qspinlock determinism patches, we were left with one
>>> cmpxchg loop on x86 due to the use of atomic_fetch_or(). Will proposed a nifty
>>> trick:
>>>
>>>   http://lkml.kernel.org/r/20180409145409.GA9661@arm.com
>>>
>>> But at the time we didn't pursue it. This series implements that and argues for
>>> its correctness. In particular it places an smp_mb__after_atomic() in
>>> between the two operations, which forces the load to come after the
>>> store (which is free on x86 anyway).
>>>
>>> In particular this ordering ensures a concurrent unlock cannot trigger
>>> the uncontended handoff. Also it ensures that if the xchg() happens
>>> after a (successful) trylock, we must observe that LOCKED bit.
>> When you said "concurrent unlock cannot trigger the uncontended
>> handoff", are you saying the current code has this uncontended handoff
>> problem or just when comparing to doing a load first followed by xchg()?
> In particular I'm talking about this:
>
> locked:
> 	/*
> 	 * claim the lock:
> 	 *
> 	 * n,0,0 -> 0,0,1 : lock, uncontended
> 	 * *,0,0 -> *,0,1 : lock, contended
> 	 *
> 	 * If the queue head is the only one in the queue (lock value == tail)
> 	 * and nobody is pending, clear the tail code and grab the lock.
> 	 * Otherwise, we only need to grab the lock.
> 	 */
> 	/*
> 	 * In the PV case we might already have _Q_LOCKED_VAL set, because
> 	 * of lock stealing; therefore we must also allow:
> 	 *
> 	 * n,0,1 -> 0,0,1
> 	 *
> 	 * Note: at this point: (val & _Q_PENDING_MASK) == 0, because of the
> 	 *       above wait condition, therefore any concurrent setting of
> 	 *       PENDING will make the uncontended transition fail.
> 	 */
> 	if ((val & _Q_TAIL_MASK) == tail) {
> 		if (atomic_try_cmpxchg_relaxed(&lock->val, &val, _Q_LOCKED_VAL))
> 			goto release; /* No contention */
> 	}
>
>
> Here queue head acquires the lock and we take the "uncontended" path. If
> we set PENDING prior to this, the cmpxchg above will fail, and the later
> load will observe the tail.

Thanks for the clarification. I was wondering if you are talking about a
latent bug in the code. So this is not the case here.

Cheers,
Longman

      reply	other threads:[~2018-09-26 23:21 UTC|newest]

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-09-26 11:01 Peter Zijlstra
2018-09-26 11:01 ` [RFC][PATCH 1/3] locking/qspinlock: Re-order code Peter Zijlstra
2018-10-01 17:17   ` Will Deacon
2018-09-26 11:01 ` [RFC][PATCH 2/3] locking/qspinlock: Rework some comments Peter Zijlstra
2018-10-01 17:17   ` Will Deacon
2018-10-01 19:10     ` Peter Zijlstra
2018-10-02 13:20       ` Will Deacon
2018-10-02 13:43         ` Peter Zijlstra
2018-09-26 11:01 ` [RFC][PATCH 3/3] locking/qspinlock: Optimize for x86 Peter Zijlstra
2018-09-26 16:30   ` Waiman Long
2018-09-26 17:54     ` Peter Zijlstra
2018-09-27  7:29       ` Peter Zijlstra
2018-09-26 20:52   ` Andrea Parri
2018-09-27  7:17     ` Peter Zijlstra
2018-09-27  7:47       ` Andrea Parri
2018-09-27  7:59         ` Peter Zijlstra
2018-09-27  8:13           ` Andrea Parri
2018-09-27  8:57             ` Peter Zijlstra
2018-09-27 12:16   ` David Laight
2018-10-01 17:17   ` Will Deacon
2018-10-01 20:00     ` Peter Zijlstra
2018-10-02 13:19       ` Will Deacon
2018-10-02 14:14         ` Peter Zijlstra
2018-10-02 12:31     ` Andrea Parri
2018-10-02 13:22       ` Will Deacon
2018-10-02 13:44         ` Andrea Parri
2018-09-26 15:01 ` [RFC][PATCH 0/3] locking/qspinlock: Improve determinism " Sebastian Andrzej Siewior
2018-09-26 15:08   ` Thomas Gleixner
2018-09-26 15:38     ` Sebastian Andrzej Siewior
2018-09-26 16:20 ` Waiman Long
2018-09-26 17:51   ` Peter Zijlstra
2018-09-26 23:21     ` Waiman Long [this message]

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=5c492b85-365f-e64d-2230-f479903591d8@redhat.com \
    --to=longman@redhat.com \
    --cc=andrea.parri@amarulasolutions.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@kernel.org \
    --cc=peterz@infradead.org \
    --cc=tglx@linutronix.de \
    --cc=will.deacon@arm.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

all inboxes | Powered by JetHome®