[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 4 Apr 2007 23:19:48 +0530
From: Srivatsa Vaddagiri <vatsa@...ibm.com>
To: Oleg Nesterov <oleg@...sign.ru>
Cc: Gautham R Shenoy <ego@...ibm.com>, akpm@...ux-foundation.org,
paulmck@...ibm.com, torvalds@...ux-foundation.org,
linux-kernel@...r.kernel.org, "Rafael J. Wysocki" <rjw@...k.pl>,
mingo@...e.hu, dipankar@...ibm.com, dino@...ibm.com,
masami.hiramatsu.pt@...achi.com
Subject: Re: [PATCH 7/8] Clean up workqueue.c with respect to the freezer based cpu-hotplug
On Wed, Apr 04, 2007 at 07:28:28PM +0400, Oleg Nesterov wrote:
> I changed my mind :) The problem is general, I am starting to believe
> it is better to change kthread_stop().
yes i agree. Although is some cases like destroy_workqueue, we need to
mark the target thread non-freezable way before we call kthread_stop (as
you pointed out).
> > I suspected that we cannot modify p->flags just like that. How abt
> > moving freezer exemption bits to a separate field, which is protected by
> > task_lock?
>
> Probably yes... In that case it makes sense to move PF_FREEZER_SKIP/PF_FROZEN
> to the new field as well.
I wonder if there are some reserved fields in task_struct which we can
reuse here ..
> Perhaps we can ignore this problem for now. Freezer is not 100% reliable
> anyway. For example,
>
> worker_thread:
>
> for (;;) {
> try_to_freeze();
>
> prepare_to_wait();
> if (...)
> schedule();
> finish_wait();
> }
>
> This is racy, we can miss freeze_process()->signal_wake_up() if it happens
> between try_to_freeze() and prepare_to_wait(). We have to check TIF_FREEZE
> before entering schedule() if we want to fix this race.
Yes that needs a fix as well. Oh dear, freezer is so fragile to break!
> Should we? I don't know. This will uglify the code, and the probability
> of this race is very low.
Would be nice to fix IMO. Atleast serves to show "how to make your code
freezer friendly".
--
Regards,
vatsa
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists