public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
To: Steven Rostedt <rostedt@goodmis.org>
Cc: David Howells <dhowells@redhat.com>,
	"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	akpm@linux-foundation.org, Ingo Molnar <mingo@elte.hu>,
	Peter Zijlstra <a.p.zijlstra@chello.nl>,
	linux-kernel@vger.kernel.org, Nicolas Pitre <nico@cam.org>,
	Ralf Baechle <ralf@linux-mips.org>,
	benh@kernel.crashing.org, paulus@samba.org,
	David Miller <davem@davemloft.net>,
	Ingo Molnar <mingo@redhat.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	linux-arch@vger.kernel.org
Subject: Re: [RFC patch 08/18] cnt32_to_63 should use smp_rmb()
Date: Fri, 7 Nov 2008 14:18:34 -0500	[thread overview]
Message-ID: <20081107191833.GA31809@Krystal> (raw)
In-Reply-To: <alpine.DEB.1.10.0811071229570.28548@gandalf.stny.rr.com>

* Steven Rostedt (rostedt@goodmis.org) wrote:
> 
> On Fri, 7 Nov 2008, Mathieu Desnoyers wrote:
> > 
> >   __m_cnt_hi
> >  is read before
> >   mmio cnt_lo read
> > 
> > for the detailed reasons explained in my previous discussion with
> > Nicolas here :
> > http://lkml.org/lkml/2008/10/21/1
> > 
> > I use smp_rmb() to do this on SMP systems (hrm, actually, a rmb() could
> > be required so it works also on UP systems safely wrt interrupts).
> 
> smp_rmb turns into a compiler barrier on UP and should prevent the below
> description.
> 

Ah, right, preserving program order on UP should be enough. smp_rmb()
then.

Thanks,

Mathieu

> -- Steve
> 
> > 
> > The write side is between the hardware counter, which is assumed to
> > increment monotonically between each read, and the value __m_cnt_hi
> > updated by the CPU. I don't see where we could put a wmb() there.
> > 
> > Without barrier, the smp race looks as follow :
> > 
> > 
> > CPU    A                                    B
> >                                             read hw cnt low (0xFFFFFFFA)
> >        read __m_cnt_hi (0x80000000)
> >        read hw cnt low (0x00000001)
> >        (wrap detected :
> >         (s32)(0x80000000 ^ 0x1) < 0)
> >        write __m_cnt_hi = 0x00000001
> >                                             read __m_cnt_hi (0x00000001)
> >        return 0x0000000100000001
> >                                             (wrap detected :
> >                                              (s32)(0x00000001 ^ 0xFFFFFFFA) < 0)
> >                                             write __m_cnt_hi = 0x80000001
> >                                             return 0x80000001FFFFFFFA
> >                                                                   (time jumps)
> > 
> > And UP interrupt race :
> > 
> >        Thread context                       Interrupts
> >        read hw cnt low (0xFFFFFFFA)
> >                                             read __m_cnt_hi (0x80000000)
> >                                             read hw cnt low (0x00000001)
> >                                             (wrap detected :
> >                                              (s32)(0x80000000 ^ 0x1) < 0)
> >                                             write __m_cnt_hi = 0x00000001
> >        read __m_cnt_hi (0x00000001)
> >                                             return 0x0000000100000001
> >        (wrap detected :
> >         (s32)(0x00000001 ^ 0xFFFFFFFA) < 0)
> >        write __m_cnt_hi = 0x80000001
> >        return 0x80000001FFFFFFFA
> >                 (time jumps)
> > 

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68

  reply	other threads:[~2008-11-07 19:18 UTC|newest]

Thread overview: 118+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-11-07  5:23 [RFC patch 00/18] Trace Clock v2 Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 01/18] get_cycles() : kconfig HAVE_GET_CYCLES Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 02/18] get_cycles() : x86 HAVE_GET_CYCLES Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 03/18] get_cycles() : sparc64 HAVE_GET_CYCLES Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 04/18] get_cycles() : powerpc64 HAVE_GET_CYCLES Mathieu Desnoyers
2008-11-07 14:56   ` Josh Boyer
2008-11-07 18:14     ` Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 05/18] get_cycles() : MIPS HAVE_GET_CYCLES_32 Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 06/18] Trace clock generic Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 07/18] Trace clock core Mathieu Desnoyers
2008-11-07  5:52   ` Andrew Morton
2008-11-07  6:16     ` Mathieu Desnoyers
2008-11-07  6:26       ` Andrew Morton
2008-11-07 16:12         ` Mathieu Desnoyers
2008-11-07 16:19           ` Andrew Morton
2008-11-07 18:16             ` Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 08/18] cnt32_to_63 should use smp_rmb() Mathieu Desnoyers
2008-11-07  6:05   ` Andrew Morton
2008-11-07  8:12     ` Nicolas Pitre
2008-11-07  8:38       ` Andrew Morton
2008-11-07 11:20         ` David Howells
2008-11-07 15:01           ` Nicolas Pitre
2008-11-07 15:50             ` Andrew Morton
2008-11-07 16:21               ` David Howells
2008-11-07 16:29                 ` Andrew Morton
2008-11-07 17:10                   ` David Howells
2008-11-07 17:26                     ` Andrew Morton
2008-11-07 18:00                       ` Mathieu Desnoyers
2008-11-07 18:21                         ` Andrew Morton
2008-11-07 18:30                           ` Harvey Harrison
2008-11-07 18:42                             ` Mathieu Desnoyers
2008-11-07 18:33                           ` Mathieu Desnoyers
2008-11-07 18:36                           ` Linus Torvalds
2008-11-07 18:45                             ` Andrew Morton
2008-11-07 16:47               ` Nicolas Pitre
2008-11-07 16:55                 ` David Howells
2008-11-07 17:21                 ` Andrew Morton
2008-11-07 20:03                   ` Nicolas Pitre
2008-11-07 16:07             ` David Howells
2008-11-07 16:47           ` Mathieu Desnoyers
2008-11-07 17:04             ` David Howells
2008-11-07 17:17               ` Mathieu Desnoyers
2008-11-07 23:27                 ` David Howells
2008-11-07 20:11             ` Russell King
2008-11-07 21:36               ` Mathieu Desnoyers
2008-11-07 22:18                 ` Russell King
2008-11-07 22:36                   ` Mathieu Desnoyers
2008-11-07 23:41                   ` David Howells
2008-11-08  0:15                     ` Russell King
2008-11-08  0:45                       ` David Howells
2008-11-08 15:24                       ` Nicolas Pitre
2008-11-08 23:20                         ` [PATCH] clarify usage expectations for cnt32_to_63() Nicolas Pitre
2008-11-09  2:25                           ` Mathieu Desnoyers
2008-11-09  2:54                             ` Nicolas Pitre
2008-11-09  5:06                               ` Nicolas Pitre
2008-11-09  5:27                                 ` [PATCH v2] " Nicolas Pitre
2008-11-09  6:48                                   ` Mathieu Desnoyers
2008-11-09 13:34                                     ` Nicolas Pitre
2008-11-09 13:43                                       ` Russell King
2008-11-09 16:22                                       ` Mathieu Desnoyers
2008-11-10  4:20                                         ` Nicolas Pitre
2008-11-10  4:42                                           ` Andrew Morton
2008-11-10 21:34                                             ` Nicolas Pitre
2008-11-10 21:58                                               ` Andrew Morton
2008-11-10 23:15                                                 ` Nicolas Pitre
2008-11-10 23:22                                                   ` Andrew Morton
2008-11-10 23:38                                                     ` Steven Rostedt
2008-11-11  0:26                                                     ` Nicolas Pitre
2008-11-11 18:28                                                       ` [PATCH] convert cnt32_to_63 to inline Mathieu Desnoyers
2008-11-11 19:13                                                         ` Russell King
2008-11-11 20:11                                                           ` Mathieu Desnoyers
2008-11-11 21:51                                                             ` Russell King
2008-11-12  3:48                                                               ` Mathieu Desnoyers
2008-11-11 21:00                                                         ` Nicolas Pitre
2008-11-11 21:13                                                           ` Russell King
2008-11-11 22:31                                                         ` David Howells
2008-11-11 22:37                                                           ` Peter Zijlstra
2008-11-12  1:13                                                             ` Steven Rostedt
2008-11-07 11:03       ` [RFC patch 08/18] cnt32_to_63 should use smp_rmb() David Howells
2008-11-07 16:51         ` Mathieu Desnoyers
2008-11-07 20:18           ` Nicolas Pitre
2008-11-07 23:55             ` David Howells
2008-11-07 10:59     ` David Howells
2008-11-07 10:55   ` David Howells
2008-11-07 17:09     ` Mathieu Desnoyers
2008-11-07 17:33       ` Steven Rostedt
2008-11-07 19:18         ` Mathieu Desnoyers [this message]
2008-11-07 19:32           ` Peter Zijlstra
2008-11-07 20:02             ` Mathieu Desnoyers
2008-11-07 20:45               ` Mathieu Desnoyers
2008-11-07 20:54                 ` Paul E. McKenney
2008-11-07 21:04                   ` Steven Rostedt
2008-11-08  0:34                     ` Paul E. McKenney
2008-11-07 21:16                   ` Mathieu Desnoyers
2008-11-07 23:50         ` David Howells
2008-11-08  0:55           ` Steven Rostedt
2008-11-09 11:51             ` David Howells
2008-11-09 14:31               ` Steven Rostedt
2008-11-09 16:18               ` Mathieu Desnoyers
2008-11-07 20:08       ` Steven Rostedt
2008-11-07 20:55         ` Paul E. McKenney
2008-11-07 21:27         ` Mathieu Desnoyers
2008-11-07 20:36       ` Nicolas Pitre
2008-11-07 20:55         ` Mathieu Desnoyers
2008-11-07 21:22           ` Nicolas Pitre
2008-11-07  5:23 ` [RFC patch 09/18] Powerpc : Trace clock Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 10/18] Sparc64 " Mathieu Desnoyers
2008-11-07  5:45   ` David Miller
2008-11-07  5:23 ` [RFC patch 11/18] LTTng timestamp sh Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 12/18] LTTng - TSC synchronicity test Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 13/18] x86 : remove arch-specific tsc_sync.c Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 14/18] MIPS use tsc_sync.c Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 15/18] MIPS : export hpt frequency for trace_clock Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 16/18] MIPS create empty sync_core() Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 17/18] MIPS : Trace clock Mathieu Desnoyers
2008-11-07 11:53   ` Peter Zijlstra
2008-11-07 17:44     ` Mathieu Desnoyers
2008-11-07  5:23 ` [RFC patch 18/18] x86 trace clock 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=20081107191833.GA31809@Krystal \
    --to=mathieu.desnoyers@polymtl.ca \
    --cc=a.p.zijlstra@chello.nl \
    --cc=akpm@linux-foundation.org \
    --cc=benh@kernel.crashing.org \
    --cc=davem@davemloft.net \
    --cc=dhowells@redhat.com \
    --cc=linux-arch@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@elte.hu \
    --cc=mingo@redhat.com \
    --cc=nico@cam.org \
    --cc=paulmck@linux.vnet.ibm.com \
    --cc=paulus@samba.org \
    --cc=ralf@linux-mips.org \
    --cc=rostedt@goodmis.org \
    --cc=tglx@linutronix.de \
    --cc=torvalds@linux-foundation.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox