2017-11-02 10:38:00

by Miroslav Benes

[permalink] [raw]
Subject: Re: [PATCH v3 1/2] livepatch: send a fake signal to all blocking tasks

On Wed, 1 Nov 2017, Oleg Nesterov wrote:

> On 11/01, Petr Mladek wrote:
> >
> > On Tue 2017-10-31 12:48:52, Miroslav Benes wrote:
> > > + if (task->flags & PF_KTHREAD) {
> > > + /*
> > > + * Wake up a kthread which still has not been migrated.
> > > + */
> > > + wake_up_process(task);
> >
> > I have just noticed that freezer used wake_up_state(p, TASK_INTERRUPTIBLE);
> > IMHO, we should do so as well.
>
> I won't argue, but...
>
> > wake_up_process() wakes also tasks in TASK_UNINTERRUPTIBLE state.
> > These might not be ready for an unexpected wakeup. For example,
> > see concat_dev_erase() in drivers/mtd/mtdcontact.c.
>
> I'd say that concat_dev_erase() should be fixed, any code should be ready
> for spurious wakeup.

I agree.

> Note also that wake_up_state(TASK_INTERRUPTIBLE) won't wakeup the TASK_IDLE
> kthreads, and most of the kthreads which use TASK_INTERRUPTIBLE should use
> TASK_IDLE today, because in most cases TASK_INTERRUPTIBLE was used to not
> contribute to loadavg.

Yes. Unfortunately, we have TASK_IDLE for more than two years now and
nothing much has happened yet. TASK_IDLE is still used sporadically. I'd
like to be on the safe side with livepatch and given that
TASK_INTERRUPTIBLE loops should be prepared for spurious wakeups by
definition, I think it is better to have wake_up_state(TASK_INTERRUPTIBLE)
there. At least till all "concat_dev_erase" beauties are fixed (sigh).

Miroslav

From 1582882798193826211@xxx Wed Nov 01 16:44:30 +0000 2017
X-GM-THRID: 1582773844473281428
X-Gmail-Labels: Inbox,Category Forums