From: Steven Rostedt <rostedt@goodmis.org>
To: Abhishek Sagar <sagar.abhishek@gmail.com>
Cc: linux-kernel@vger.kernel.org, Ingo Molnar <mingo@elte.hu>,
Thomas Gleixner <tglx@linutronix.de>,
Peter Zijlstra <peterz@infradead.org>,
Andrew Morton <akpm@linux-foundation.org>,
Linus Torvalds <torvalds@linux-foundation.org>,
David Miller <davem@davemloft.net>,
Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>,
Roland McGrath <roland@redhat.com>,
Ulrich Drepper <drepper@redhat.com>,
Rusty Russell <rusty@rustcorp.com.au>,
Jeremy Fitzhardinge <jeremy@goop.org>,
Gregory Haskins <ghaskins@novell.com>,
Arnaldo Carvalho de Melo <acme@redhat.com>,
"Luis Claudio R. Goncalves" <lclaudio@uudg.org>,
Clark Williams <williams@redhat.com>
Subject: Re: [PATCH 0/5] ftrace: to kill a daemon
Date: Sat, 9 Aug 2008 09:01:36 -0400 (EDT) [thread overview]
Message-ID: <alpine.DEB.1.10.0808090854170.24140@gandalf.stny.rr.com> (raw)
In-Reply-To: <863e9df20808090248h3adc2ac3mdb3217fe2876ab3b@mail.gmail.com>
On Sat, 9 Aug 2008, Abhishek Sagar wrote:
> On Thu, Aug 7, 2008 at 11:50 PM, Steven Rostedt <rostedt@goodmis.org> wrote:
> > You see, the reason for this is that for ftrace to maintain performance
> > when configured in but disabled, it would need to change all the
> > locations that called "mcount" (enabled with the gcc -pg option) into
> > nops. The "-pg" option in gcc sets up a function profiler to call this
> > function called "mcount". If you simply have "mcount" return, it will
> > still add 15 to 18% overhead in performance. Changing all the calls to
> > nops moved the overhead into noise.
> >
> > To get rid of this, I had the mcount code record the location that called
> > it. Later, the "ftraced" daemon would wake up and look to see if
> > any new functions were recorded. If so, it would call kstop_machine
> > and convert the calls to "nops". We needed kstop_machine because bad
> > things happen on SMP if you modify code that happens to be in the
> > instruction cache of another CPU.
>
> Is this new framework needed for x86 specific reasons only? From what
> I gathered here, ftraced defers mcount patching simply because there's
> no way to update a 5-byte nop atomically. If so, why can't mcount site
> patching be left to arch specific ftrace code? For !SMP or archs which
> generate word sized mcount branch calls (e.g, ARM) is there really no
> way to patch mcount sites synchronously from inside ftrace_record_ip
> by disabling interrupts?
There's two topics in this thread.
1) the x86 issue of the 5 byte instruction. The problem with x86 is that on
some CPUs the nop used consists of two nops to fill the 5 bytes. There is
no way to change that atomically. The workarounds for this is the arch
specific ftrace_pre_enable() that will make sure no process is about to
execute the second part of that nop.
2) Getting rid of the daemon. The daemon is used to patch the code
dynamically later on bootup. Now an arch may or may not be able to modify
code in SMP, but I've been told that this is dangerous to do even on PPC.
Dynamically modifying text that might be in the pipeline on another CPU
may or may not be dangerous on all archs.
The fix here is to convert the mcount calls to nops at boot up. This is
really ideal on all archs. This means we know ever mcount call, and we get
rid of the requirement that we need to run the code once before we can
trace it.
The kstop_machine is now only left at the start and stop of tracing.
-- Steve
next prev parent reply other threads:[~2008-08-09 13:01 UTC|newest]
Thread overview: 107+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-08-07 18:20 Steven Rostedt
2008-08-07 18:20 ` [PATCH 1/5] ftrace: create __mcount_loc section Steven Rostedt
2008-08-07 18:20 ` [PATCH 2/5] ftrace: mcount call site on boot nops core Steven Rostedt
2008-08-07 18:20 ` [PATCH 3/5] ftrace: enable mcount recording for modules Steven Rostedt
2008-08-08 6:43 ` Rusty Russell
2008-08-08 12:51 ` Steven Rostedt
2008-08-07 18:20 ` [PATCH 4/5] ftrace: rebuild everything on change to FTRACE_MCOUNT_RECORD Steven Rostedt
2008-08-07 18:20 ` [PATCH 5/5] ftrace: enable using mcount recording on x86 Steven Rostedt
2008-08-07 18:47 ` [PATCH 0/5] ftrace: to kill a daemon Mathieu Desnoyers
2008-08-07 20:42 ` Steven Rostedt
2008-08-08 17:22 ` Mathieu Desnoyers
2008-08-08 17:36 ` Steven Rostedt
2008-08-08 17:46 ` Mathieu Desnoyers
2008-08-08 18:13 ` Steven Rostedt
2008-08-08 18:15 ` Peter Zijlstra
2008-08-08 18:21 ` Mathieu Desnoyers
2008-08-08 18:41 ` Steven Rostedt
2008-08-08 19:04 ` Linus Torvalds
2008-08-08 19:05 ` Mathieu Desnoyers
2008-08-08 23:38 ` Steven Rostedt
2008-08-09 0:23 ` Andi Kleen
2008-08-09 0:36 ` Steven Rostedt
2008-08-09 0:47 ` Jeremy Fitzhardinge
2008-08-09 0:51 ` Linus Torvalds
2008-08-09 1:25 ` Steven Rostedt
2008-08-13 6:31 ` Mathieu Desnoyers
2008-08-13 15:38 ` Mathieu Desnoyers
2008-08-13 17:52 ` Efficient x86 and x86_64 NOP microbenchmarks Mathieu Desnoyers
2008-08-13 18:27 ` Linus Torvalds
2008-08-13 18:41 ` Andi Kleen
2008-08-13 18:45 ` Avi Kivity
2008-08-13 18:51 ` Andi Kleen
2008-08-13 18:56 ` Avi Kivity
2008-08-13 19:30 ` Mathieu Desnoyers
2008-08-13 19:37 ` Andi Kleen
2008-08-13 20:01 ` Mathieu Desnoyers
2008-08-13 23:41 ` [RFC PATCH] x86 alternatives : fix LOCK_PREFIX race with preemptible kernel and CPU hotplug Mathieu Desnoyers
2008-08-14 0:01 ` H. Peter Anvin
2008-08-14 1:13 ` Mathieu Desnoyers
2008-08-14 1:22 ` Jeremy Fitzhardinge
2008-08-14 1:26 ` Roland McGrath
2008-08-14 1:49 ` Mathieu Desnoyers
2008-08-14 3:35 ` Jeremy Fitzhardinge
2008-08-14 15:18 ` Mathieu Desnoyers
2008-08-14 16:10 ` Linus Torvalds
2008-08-14 16:13 ` H. Peter Anvin
2008-08-14 16:58 ` Mathieu Desnoyers
2008-08-14 17:05 ` Jeremy Fitzhardinge
2008-08-14 17:30 ` Mathieu Desnoyers
2008-08-14 17:43 ` Jeremy Fitzhardinge
2008-08-14 18:37 ` H. Peter Anvin
2008-08-14 18:53 ` Mathieu Desnoyers
2008-08-14 19:29 ` Jeremy Fitzhardinge
2008-08-14 20:31 ` Mathieu Desnoyers
2008-08-14 20:39 ` H. Peter Anvin
2008-08-14 21:46 ` Jeremy Fitzhardinge
2008-08-14 22:26 ` H. Peter Anvin
2008-08-14 17:17 ` H. Peter Anvin
2008-08-14 18:09 ` Mathieu Desnoyers
2008-08-14 19:49 ` Mathieu Desnoyers
2008-08-14 17:04 ` Jeremy Fitzhardinge
2008-08-14 17:18 ` H. Peter Anvin
2008-08-14 17:28 ` Jeremy Fitzhardinge
2008-08-14 17:31 ` H. Peter Anvin
2008-08-14 17:46 ` Mathieu Desnoyers
2008-08-14 17:49 ` Jeremy Fitzhardinge
2008-08-14 17:55 ` Mathieu Desnoyers
2008-08-14 18:59 ` Gregory Haskins
2008-08-15 21:34 ` Efficient x86 and x86_64 NOP microbenchmarks Steven Rostedt
2008-08-15 21:51 ` Andi Kleen
2008-08-13 19:16 ` Mathieu Desnoyers
2008-08-09 0:51 ` [PATCH 0/5] ftrace: to kill a daemon Steven Rostedt
2008-08-09 0:53 ` Roland McGrath
2008-08-09 1:13 ` Andi Kleen
2008-08-09 1:19 ` Andi Kleen
2008-08-09 1:30 ` Steven Rostedt
2008-08-09 1:55 ` Andi Kleen
2008-08-09 2:03 ` Steven Rostedt
2008-08-09 2:23 ` Andi Kleen
2008-08-09 4:12 ` Steven Rostedt
2008-08-09 0:30 ` Steven Rostedt
2008-08-11 18:21 ` Mathieu Desnoyers
2008-08-11 19:28 ` Steven Rostedt
2008-08-08 19:08 ` Jeremy Fitzhardinge
2008-08-11 2:41 ` Rusty Russell
2008-08-11 12:33 ` Steven Rostedt
2008-08-07 21:11 ` Jeremy Fitzhardinge
2008-08-07 21:29 ` Steven Rostedt
2008-08-07 22:26 ` Roland McGrath
2008-08-08 1:21 ` Steven Rostedt
2008-08-08 1:24 ` Steven Rostedt
2008-08-08 1:56 ` Steven Rostedt
2008-08-08 7:22 ` Peter Zijlstra
2008-08-08 11:31 ` Steven Rostedt
2008-08-08 4:54 ` Sam Ravnborg
2008-08-09 9:48 ` Abhishek Sagar
2008-08-09 13:01 ` Steven Rostedt [this message]
2008-08-09 15:01 ` Abhishek Sagar
2008-08-09 15:37 ` Steven Rostedt
2008-08-09 17:14 ` Abhishek Sagar
[not found] <aYipy-5FM-9@gated-at.bofh.it>
2008-08-07 21:28 ` Bodo Eggert
2008-08-07 21:24 ` Jeremy Fitzhardinge
2008-08-07 21:35 ` Steven Rostedt
[not found] ` <aYiIP-6bN-11@gated-at.bofh.it>
[not found] ` <aYkAT-15Q-1@gated-at.bofh.it>
[not found] ` <aYDX5-fa-31@gated-at.bofh.it>
[not found] ` <aYE6H-tg-21@gated-at.bofh.it>
[not found] ` <aYEgj-Fc-17@gated-at.bofh.it>
[not found] ` <aYEJq-1xB-23@gated-at.bofh.it>
[not found] ` <aYET8-1L7-17@gated-at.bofh.it>
[not found] ` <aYFcq-2f2-11@gated-at.bofh.it>
[not found] ` <aYFvP-2Vc-25@gated-at.bofh.it>
[not found] ` <aYJIZ-15G-7@gated-at.bofh.it>
[not found] ` <aYKvs-2gQ-7@gated-at.bofh.it>
[not found] ` <aYKF1-2qs-7@gated-at.bofh.it>
2008-08-09 11:50 ` Bodo Eggert
2008-08-09 13:02 ` Steven Rostedt
2008-08-09 14:25 ` Steven Rostedt
2008-08-09 14:42 ` Bodo Eggert
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=alpine.DEB.1.10.0808090854170.24140@gandalf.stny.rr.com \
--to=rostedt@goodmis.org \
--cc=acme@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=davem@davemloft.net \
--cc=drepper@redhat.com \
--cc=ghaskins@novell.com \
--cc=jeremy@goop.org \
--cc=lclaudio@uudg.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mathieu.desnoyers@polymtl.ca \
--cc=mingo@elte.hu \
--cc=peterz@infradead.org \
--cc=roland@redhat.com \
--cc=rusty@rustcorp.com.au \
--cc=sagar.abhishek@gmail.com \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
--cc=williams@redhat.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®