From: Dipankar Sarma <dipankar@in.ibm.com>
To: Jesse Barnes <jbarnes@engr.sgi.com>
Cc: William Lee Irwin III <wli@holomorphy.com>,
Jack Steiner <steiner@sgi.com>,
linux-kernel@vger.kernel.org, edwardsg@sgi.com,
levon@movementarian.org
Subject: Re: contention on profile_lock
Date: Fri, 5 Nov 2004 03:46:55 +0530 [thread overview]
Message-ID: <20041104221655.GA4099@in.ibm.com> (raw)
In-Reply-To: <200411041355.27228.jbarnes@engr.sgi.com>
On Thu, Nov 04, 2004 at 01:55:27PM -0800, Jesse Barnes wrote:
> On Thursday, November 4, 2004 12:49 pm, Jesse Barnes wrote:
> > John pointed out that this breaks modules. Would registering and
>
> static void timer_stop(void)
> {
> - unregister_profile_notifier(&timer_notifier);
> + /* Tear down the callback pointer after sync_kernel */
> + synchronize_kernel();
> + oprofile_timer_notify = NULL;
> }
>
>
> +/* Oprofile timer tick hook */
> +int (*oprofile_timer_notify)(struct pt_regs *);
> +
> static atomic_t *prof_buffer;
> static unsigned long prof_len, prof_shift;
> static int prof_on;
> @@ -168,38 +172,6 @@
> return err;
> }
>
> -void profile_hook(struct pt_regs * regs)
> -{
> - read_lock(&profile_lock);
> - notifier_call_chain(&profile_listeners, 0, regs);
> - read_unlock(&profile_lock);
> -}
> -
> -EXPORT_SYMBOL_GPL(register_profile_notifier);
> -EXPORT_SYMBOL_GPL(unregister_profile_notifier);
> EXPORT_SYMBOL_GPL(task_handoff_register);
> EXPORT_SYMBOL_GPL(task_handoff_unregister);
>
> @@ -394,8 +366,8 @@
>
> void profile_tick(int type, struct pt_regs *regs)
> {
> - if (type == CPU_PROFILING)
> - profile_hook(regs);
> + if (type == CPU_PROFILING && oprofile_timer_notify)
> + oprofile_timer_notify(regs);
> if (!user_mode(regs) && cpu_isset(smp_processor_id(), prof_cpu_mask))
> profile_hit(type, (void *)profile_pc(regs));
> }
Or you could just do -
void profile_tick(int type, struct pt_regs *regs)
{
int (*timer_notify)(struct pt_regs *);
timer_notify = oprofile_timer_notify;
smp_read_barrier_depends();
if (type == CPU_PROFILING && timer_notify) {
timer_notify(regs);
}
if (!user_mode(regs) && cpu_isset(smp_processor_id(), prof_cpu_mask))
profile_hit(type, (void *)profile_pc(regs));
}
And avoid the synchronize_kernel(). But I think the synchronize_kernel()
thing is cleaner.
I am looking at the notifier chain mechanism itself and its
lack of proper locking. That is a different story.
Thanks
Dipankar
next prev parent reply other threads:[~2004-11-04 22:35 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2004-11-02 19:52 contention on profile_lock Jesse Barnes
2004-11-02 20:02 ` Jack Steiner
2004-11-02 21:42 ` Jesse Barnes
2004-11-04 19:56 ` Jesse Barnes
2004-11-04 20:12 ` William Lee Irwin III
2004-11-04 20:49 ` Jesse Barnes
2004-11-04 21:51 ` John Levon
2004-11-04 22:08 ` Jesse Barnes
2004-11-04 21:55 ` Jesse Barnes
2004-11-04 22:16 ` Dipankar Sarma [this message]
2004-11-04 22:21 ` John Levon
2004-11-04 22:27 ` Jesse Barnes
2004-11-04 22:52 ` John Levon
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=20041104221655.GA4099@in.ibm.com \
--to=dipankar@in.ibm.com \
--cc=edwardsg@sgi.com \
--cc=jbarnes@engr.sgi.com \
--cc=levon@movementarian.org \
--cc=linux-kernel@vger.kernel.org \
--cc=steiner@sgi.com \
--cc=wli@holomorphy.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