From: Andrea Parri <andrea.parri@amarulasolutions.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>,
Yongji Xie <elohimes@gmail.com>,
mingo@redhat.com, will.deacon@arm.com,
linux-kernel@vger.kernel.org, xieyongji@baidu.com,
zhangyu31@baidu.com, liuqi16@baidu.com, yuanlinsi01@baidu.com,
nixun@baidu.com, lilin24@baidu.com, longman@redhat.com
Subject: Re: [RFC] locking/rwsem: Avoid issuing wakeup before setting the reader waiter to nil
Date: Fri, 30 Nov 2018 10:30:29 +0100 [thread overview]
Message-ID: <20181130093029.GA6299@andrea> (raw)
In-Reply-To: <20181129221714.GF11632@hirez.programming.kicks-ass.net>
On Thu, Nov 29, 2018 at 11:17:14PM +0100, Peter Zijlstra wrote:
> On Thu, Nov 29, 2018 at 01:34:21PM -0800, Davidlohr Bueso wrote:
> > I messed up something such that waiman was not in the thread. Ccing.
> >
> > > On Thu, 29 Nov 2018, Waiman Long wrote:
> > >
> > > > That can be costly for x86 which will now have 2 locked instructions.
> > >
> > > Yeah, and when used as an actual queue we should really start to notice.
> > > Some users just have a single task in the wake_q because avoiding the cost
> > > of wake_up_process() with locks held is significant.
> > >
> > > How about instead of adding the barrier before the cmpxchg, we do it
> > > in the failed branch, right before we return. This is the uncommon
> > > path.
> > >
> > > Thanks,
> > > Davidlohr
> > >
> > > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > > index 091e089063be..0d844a18a9dc 100644
> > > --- a/kernel/sched/core.c
> > > +++ b/kernel/sched/core.c
> > > @@ -408,8 +408,14 @@ void wake_q_add(struct wake_q_head *head, struct task_struct *task)
> > > * This cmpxchg() executes a full barrier, which pairs with the full
> > > * barrier executed by the wakeup in wake_up_q().
> > > */
> > > - if (cmpxchg(&node->next, NULL, WAKE_Q_TAIL))
> > > + if (cmpxchg(&node->next, NULL, WAKE_Q_TAIL)) {
> > > + /*
> > > + * Ensure, that when the cmpxchg() fails, the corresponding
> > > + * wake_up_q() will observe our prior state.
> > > + */
> > > + smp_mb__after_atomic();
> > > return;
> > > + }
>
> So wake_up_q() does:
>
> wake_up_q():
> node->next = NULL;
> /* implied smp_mb */
> wake_up_process();
>
> So per the cross your variables 'rule', this side then should do:
>
> wake_q_add():
> /* wake_cond = true */
> smp_mb()
> cmpxchg_relaxed(&node->next, ...);
>
> So that the ordering pivots around node->next.
>
> Either we see NULL and win the cmpxchg (in which case we'll do the
> wakeup later) or, when we fail the cmpxchg, we must observe what came
> before the failure.
>
> If it wasn't so damn late, I'd try and write a litmus test for this,
> because now I'm starting to get confused -- also probably because it's
> late.
The above description suggests:
C wake_up_q-wake_q_add
{
int next = 0;
int y = 0;
}
P0(int *next, int *y)
{
int r0;
/* in wake_up_q() */
WRITE_ONCE(*next, 1); /* node->next = NULL */
smp_mb(); /* implied by wake_up_process() */
r0 = READ_ONCE(*y);
}
P1(int *next, int *y)
{
int r1;
/* in wake_q_add() */
WRITE_ONCE(*y, 1); /* wake_cond = true */
smp_mb__before_atomic();
r1 = cmpxchg_relaxed(next, 1, 2);
}
exists (0:r0=0 /\ 1:r1=0)
This "exists" clause cannot be satisfied according to the LKMM:
Test wake_up_q-wake_q_add Allowed
States 3
0:r0=0; 1:r1=1;
0:r0=1; 1:r1=0;
0:r0=1; 1:r1=1;
No
Witnesses
Positive: 0 Negative: 3
Condition exists (0:r0=0 /\ 1:r1=0)
Observation wake_up_q-wake_q_add Never 0 3
Time wake_up_q-wake_q_add 0.00
Hash=72d85545f97ef7fd35c8928259225ee0
(TBH, I'm not sure what "y" (you denoted it "wake_cond") is pointing to
here/is modeling, but I might have missed some previous remarks...)
Andrea
>
> In any case, I think you patch is 'wrong' because it puts the barrier on
> the wrong side of the cmpxchg() (after, as opposed to before).
next prev parent reply other threads:[~2018-11-30 9:30 UTC|newest]
Thread overview: 50+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-11-29 12:50 [RFC] locking/rwsem: Avoid issuing wakeup before setting the reader waiter to nil Yongji Xie
2018-11-29 13:12 ` Peter Zijlstra
2018-11-29 13:44 ` Peter Zijlstra
2018-11-29 14:02 ` Yongji Xie
2018-11-29 18:43 ` Davidlohr Bueso
2018-11-29 18:49 ` Waiman Long
2018-11-29 15:21 ` Waiman Long
2018-11-29 15:29 ` Waiman Long
2018-11-29 16:06 ` Peter Zijlstra
2018-11-29 17:02 ` Waiman Long
2018-11-29 17:27 ` Peter Zijlstra
2018-11-29 17:58 ` Waiman Long
2018-11-29 18:13 ` Peter Zijlstra
2018-11-29 18:17 ` Davidlohr Bueso
2018-11-29 18:08 ` Peter Zijlstra
2018-11-29 18:26 ` Waiman Long
2018-11-29 18:31 ` Will Deacon
2018-11-29 18:34 ` Waiman Long
2018-11-29 22:05 ` Peter Zijlstra
2018-11-30 9:34 ` 答复: " Liu,Qi(ACU-T1)
2018-11-30 14:15 ` Peter Zijlstra
2018-11-29 21:30 ` Davidlohr Bueso
2018-11-29 21:34 ` Davidlohr Bueso
2018-11-29 22:17 ` Peter Zijlstra
2018-11-30 9:30 ` Andrea Parri [this message]
2018-12-03 5:31 ` [PATCH -tip] kernel/sched,wake_q: Branch predict wake_q_add() cmpxchg Davidlohr Bueso
2018-12-03 16:10 ` Waiman Long
2019-01-21 11:28 ` [tip:locking/core] sched/wake_q: Add branch prediction hint to " tip-bot for Davidlohr Bueso
2018-12-10 15:12 ` [RFC] locking/rwsem: Avoid issuing wakeup before setting the reader waiter to nil Yongji Xie
2018-12-17 11:37 ` Peter Zijlstra
2018-12-17 13:12 ` Yongji Xie
2019-01-07 14:35 ` Waiman Long
2019-01-07 15:31 ` Peter Zijlstra
2019-01-07 15:35 ` Waiman Long
2018-12-17 20:53 ` Davidlohr Bueso
2018-12-18 13:10 ` Peter Zijlstra
2018-12-18 13:14 ` Peter Zijlstra
2018-12-18 17:27 ` Davidlohr Bueso
2018-12-18 18:54 ` [PATCH v2] sched/wake_q: Reduce reference counting for special users Davidlohr Bueso
2018-12-18 19:17 ` Waiman Long
2018-12-18 19:30 ` Davidlohr Bueso
2018-12-18 19:39 ` Davidlohr Bueso
2018-12-18 19:53 ` [PATCH v4] " Davidlohr Bueso
2018-12-18 20:35 ` Waiman Long
2019-01-21 16:02 ` Davidlohr Bueso
2019-01-22 8:55 ` Peter Zijlstra
2019-02-04 8:57 ` [tip:locking/core] " tip-bot for Davidlohr Bueso
2019-02-07 19:30 ` Davidlohr Bueso
2019-02-12 14:14 ` Daniel Vacek
2019-01-21 11:28 ` [tip:locking/core] locking/rwsem: Fix (possible) missed wakeup tip-bot for Xie Yongji
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=20181130093029.GA6299@andrea \
--to=andrea.parri@amarulasolutions.com \
--cc=dave@stgolabs.net \
--cc=elohimes@gmail.com \
--cc=lilin24@baidu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=liuqi16@baidu.com \
--cc=longman@redhat.com \
--cc=mingo@redhat.com \
--cc=nixun@baidu.com \
--cc=peterz@infradead.org \
--cc=will.deacon@arm.com \
--cc=xieyongji@baidu.com \
--cc=yuanlinsi01@baidu.com \
--cc=zhangyu31@baidu.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).