From: Peter Zijlstra <peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
To: Andy Lutomirski <luto-kltTT9wpgjJwATOyAt5JVQ@public.gmane.org>
Cc: Mathieu Desnoyers
<mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>,
"Paul E. McKenney"
<paulmck-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>,
Ingo Molnar <mingo-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>,
Paul Turner <commonly-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>,
Andi Kleen <andi-Vw/NltI1exuRpAAqCnN02g@public.gmane.org>,
Chris Lameter <cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org>,
Dave Watson <davejwatson-b10kYP2dOMg@public.gmane.org>,
Josh Triplett <josh-iaAMLnmF4UmaiuxdJuQwMA@public.gmane.org>,
Linux API <linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
"linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org"
<linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
Andrew Hunter <ahh-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>,
Linus Torvalds
<torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
Subject: Re: [RFC PATCH 0/3] restartable sequences v2: fast user-space percpu critical sections
Date: Fri, 8 Apr 2016 08:41:36 +0200 [thread overview]
Message-ID: <20160408064136.GJ3448@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <CALCETrVGo1Di3qamxx1NAFUSN_o=-HnYRDpeVp7zrQEBwe5u-g-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
On Thu, Apr 07, 2016 at 09:43:33AM -0700, Andy Lutomirski wrote:
> enter the critical section:
> 1:
> movq %[cpu], %%r12
> movq {address of counter for our cpu}, %%r13
> movq {some fresh value}, (%%r13)
> cmpq %[cpu], %%r12
> jne 1b
This is inherently racy; your forgot the detail of 'some fresh value',
but since you want to avoid collisions you really want an increment.
But load-store archs cannot do that. Or rather, they need to do:
load Rn, $event
add Rn, Rn, 1
store $event, Rn
But if they're preempted in the middle, two threads will collide and
generate the _same_ increment. Comparing CPU numbers will not fix that.
WARNING: multiple messages have this Message-ID (diff)
From: Peter Zijlstra <peterz@infradead.org>
To: Andy Lutomirski <luto@amacapital.net>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
Ingo Molnar <mingo@redhat.com>, Paul Turner <commonly@gmail.com>,
Andi Kleen <andi@firstfloor.org>, Chris Lameter <cl@linux.com>,
Dave Watson <davejwatson@fb.com>,
Josh Triplett <josh@joshtriplett.org>,
Linux API <linux-api@vger.kernel.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
Andrew Hunter <ahh@google.com>,
Linus Torvalds <torvalds@linux-foundation.org>
Subject: Re: [RFC PATCH 0/3] restartable sequences v2: fast user-space percpu critical sections
Date: Fri, 8 Apr 2016 08:41:36 +0200 [thread overview]
Message-ID: <20160408064136.GJ3448@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <CALCETrVGo1Di3qamxx1NAFUSN_o=-HnYRDpeVp7zrQEBwe5u-g@mail.gmail.com>
On Thu, Apr 07, 2016 at 09:43:33AM -0700, Andy Lutomirski wrote:
> enter the critical section:
> 1:
> movq %[cpu], %%r12
> movq {address of counter for our cpu}, %%r13
> movq {some fresh value}, (%%r13)
> cmpq %[cpu], %%r12
> jne 1b
This is inherently racy; your forgot the detail of 'some fresh value',
but since you want to avoid collisions you really want an increment.
But load-store archs cannot do that. Or rather, they need to do:
load Rn, $event
add Rn, Rn, 1
store $event, Rn
But if they're preempted in the middle, two threads will collide and
generate the _same_ increment. Comparing CPU numbers will not fix that.
next prev parent reply other threads:[~2016-04-08 6:41 UTC|newest]
Thread overview: 69+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-10-27 23:56 [RFC PATCH 0/3] restartable sequences v2: fast user-space percpu critical sections Paul Turner
2015-10-27 23:56 ` Paul Turner
2015-10-27 23:56 ` [RFC PATCH v2 1/3] restartable sequences: user-space per-cpu " Paul Turner
[not found] ` <20151027235653.16059.8933.stgit-G8L5E6GV2z5XSTzz+wBt03oUN1GumTyQ7j82oEJ37pA@public.gmane.org>
2015-11-19 16:38 ` Johannes Berg
2015-11-19 16:38 ` Johannes Berg
2015-12-11 12:56 ` Mathieu Desnoyers
2015-12-11 12:56 ` Mathieu Desnoyers
2015-10-27 23:57 ` [RFC PATCH v2 2/3] restartable sequences: x86 ABI Paul Turner
[not found] ` <20151027235705.16059.63268.stgit-G8L5E6GV2z5XSTzz+wBt03oUN1GumTyQ7j82oEJ37pA@public.gmane.org>
2015-10-28 5:03 ` Peter Zijlstra
2015-10-28 5:03 ` Peter Zijlstra
[not found] ` <20151028050314.GC11242-IIpfhp3q70xmmu7s1q4rt2t3HXsI98Cx0E9HWUfgJXw@public.gmane.org>
2015-10-28 5:19 ` Paul Turner
2015-10-28 5:19 ` Paul Turner
2015-12-11 13:30 ` Mathieu Desnoyers
2015-12-11 13:30 ` Mathieu Desnoyers
2015-10-27 23:57 ` [RFC PATCH v2 3/3] restartable sequences: basic self-tests Paul Turner
[not found] ` <20151027235716.16059.47610.stgit-G8L5E6GV2z5XSTzz+wBt03oUN1GumTyQ7j82oEJ37pA@public.gmane.org>
2016-04-05 20:33 ` Mathieu Desnoyers
2016-04-05 20:33 ` Mathieu Desnoyers
[not found] ` <1276514010.46061.1459888406999.JavaMail.zimbra-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
2016-04-06 7:43 ` Peter Zijlstra
2016-04-06 7:43 ` Peter Zijlstra
[not found] ` <20160406074309.GE3430-ndre7Fmf5hadTX5a5knrm8zTDFooKrT+cvkQGrU6aU0@public.gmane.org>
2016-04-06 13:39 ` Mathieu Desnoyers
2016-04-06 13:39 ` Mathieu Desnoyers
[not found] ` <528054829.46502.1459949962537.JavaMail.zimbra-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
2016-04-06 19:25 ` Peter Zijlstra
2016-04-06 19:25 ` Peter Zijlstra
[not found] ` <20151027235635.16059.11630.stgit-G8L5E6GV2z5XSTzz+wBt03oUN1GumTyQ7j82oEJ37pA@public.gmane.org>
2015-10-28 14:44 ` [RFC PATCH 0/3] restartable sequences v2: fast user-space percpu critical sections Dave Watson
2015-10-28 14:44 ` Dave Watson
2015-12-11 12:05 ` Mathieu Desnoyers
2015-12-11 12:05 ` Mathieu Desnoyers
[not found] ` <1070636085.232143.1449835536723.JavaMail.zimbra-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
2015-12-11 13:39 ` Mathieu Desnoyers
2015-12-11 13:39 ` Mathieu Desnoyers
2016-04-06 15:56 ` Andy Lutomirski
2016-04-07 12:02 ` Peter Zijlstra
[not found] ` <20160407120254.GY3448-ndre7Fmf5hadTX5a5knrm8zTDFooKrT+cvkQGrU6aU0@public.gmane.org>
2016-04-07 14:35 ` Andy Lutomirski
2016-04-07 14:35 ` Andy Lutomirski
[not found] ` <CALCETrV0vcYcnBrs0axykJD=_BM28wKWVMG6bMzK8zh8R3m5fg-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2016-04-07 15:24 ` Peter Zijlstra
2016-04-07 15:24 ` Peter Zijlstra
[not found] ` <20160407152432.GZ3448-ndre7Fmf5hadTX5a5knrm8zTDFooKrT+cvkQGrU6aU0@public.gmane.org>
2016-04-07 15:39 ` Peter Zijlstra
2016-04-07 15:39 ` Peter Zijlstra
2016-04-07 15:44 ` Andy Lutomirski
2016-04-07 15:44 ` Andy Lutomirski
[not found] ` <CALCETrU5ZL6Jajc=9up-j86vY_Xtt-gTFjdQE0sB0d=d-CJZ6A-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2016-04-07 15:53 ` Peter Zijlstra
2016-04-07 15:53 ` Peter Zijlstra
[not found] ` <20160407155312.GA3448-ndre7Fmf5hadTX5a5knrm8zTDFooKrT+cvkQGrU6aU0@public.gmane.org>
2016-04-07 16:43 ` Andy Lutomirski
2016-04-07 16:43 ` Andy Lutomirski
[not found] ` <CALCETrVGo1Di3qamxx1NAFUSN_o=-HnYRDpeVp7zrQEBwe5u-g-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2016-04-07 20:11 ` Peter Zijlstra
2016-04-07 20:11 ` Peter Zijlstra
[not found] ` <20160407201156.GC3448-ndre7Fmf5hadTX5a5knrm8zTDFooKrT+cvkQGrU6aU0@public.gmane.org>
2016-04-07 22:05 ` Andy Lutomirski
2016-04-07 22:05 ` Andy Lutomirski
[not found] ` <CALCETrXVReuuGGKW6EOV7tFFaK9RbwWxYvKdpUdvU=MpDaOtsQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2016-04-08 1:11 ` Mathieu Desnoyers
2016-04-08 1:11 ` Mathieu Desnoyers
2016-04-08 1:21 ` Andy Lutomirski
2016-04-08 2:05 ` Mathieu Desnoyers
2016-04-08 2:05 ` Mathieu Desnoyers
2016-04-08 17:46 ` Mathieu Desnoyers
[not found] ` <65466698.51122.1460137589499.JavaMail.zimbra-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
2016-04-08 21:16 ` Andy Lutomirski
2016-04-08 21:16 ` Andy Lutomirski
2016-04-08 21:25 ` Linus Torvalds
2016-04-08 21:25 ` Linus Torvalds
[not found] ` <CA+55aFwqJmTy+Nz0k9N_2zsms51meTFMdvYYW5VHdiOq8Jjr7Q-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2016-04-10 14:07 ` Mathieu Desnoyers
2016-04-10 14:07 ` Mathieu Desnoyers
2016-04-08 11:02 ` Peter Zijlstra
2016-04-08 11:02 ` Peter Zijlstra
[not found] ` <20160408110232.GP3448-ndre7Fmf5hadTX5a5knrm8zTDFooKrT+cvkQGrU6aU0@public.gmane.org>
2016-04-08 15:57 ` Andy Lutomirski
2016-04-08 15:57 ` Andy Lutomirski
2016-04-08 6:41 ` Peter Zijlstra [this message]
2016-04-08 6:41 ` Peter Zijlstra
[not found] ` <20160408064136.GJ3448-ndre7Fmf5hadTX5a5knrm8zTDFooKrT+cvkQGrU6aU0@public.gmane.org>
2016-04-08 15:58 ` Andy Lutomirski
2016-04-08 15:58 ` Andy Lutomirski
2016-04-11 21:55 ` Mathieu Desnoyers
2016-04-11 21:55 ` Mathieu Desnoyers
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=20160408064136.GJ3448@twins.programming.kicks-ass.net \
--to=peterz-wegcikhe2lqwvfeawa7xhq@public.gmane.org \
--cc=ahh-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org \
--cc=andi-Vw/NltI1exuRpAAqCnN02g@public.gmane.org \
--cc=cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org \
--cc=commonly-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org \
--cc=davejwatson-b10kYP2dOMg@public.gmane.org \
--cc=josh-iaAMLnmF4UmaiuxdJuQwMA@public.gmane.org \
--cc=linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=luto-kltTT9wpgjJwATOyAt5JVQ@public.gmane.org \
--cc=mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org \
--cc=mingo-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \
--cc=paulmck-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org \
--cc=torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.