On Wednesday, 11 April 2007 16:36, Oleg Nesterov wrote:
> On 04/11, Gautham R Shenoy wrote:
> > On Wed, Apr 11, 2007 at 03:48:05PM +0400, Oleg Nesterov wrote:
> > > On 04/11, Gautham R Shenoy wrote:
> > > >
> > > > On Wed, Apr 11, 2007 at 12:13:34PM +0200, Rafael J. Wysocki wrote:
> > > > >
> > > > > It should be calling try_to_freeze() somewhere anyway. We may need
> > > > > to freeze
> > > > > all tasks in some cases.
> > > >
> > > > How about
> > > > for (;;) {
> > > > try_to_freeze();
> > > >
> > > > ?
> > >
> > > Why?
> >
> > If some event (defintely NOT cpu hotplug) needs this thread frozen.
> >
> > >
> > > > This change allows us to make all the worker threads freezeable by
> > > > default.
> > > > >From cpu-hotplug perspective, helper_wq was the only singlethreaded
> > > > non-freezeable workqueue.
> > >
> > > I think Eric's patch is what you need. We should _not_ freeze kthreadd(),
> > > we
> > > need kthread_create() after freezing. Now it doesn't depend on
> > > workqueues, we
> > > can freeze them all, single-thread or not.
> > >
> >
> > These were my exact thoughts.
>
> Sorry, I misunderstood your message.
>
> Yes, we can freeze it with FE_HOTPLUG_CPU. In that case wait_event()
> should also check !freezing(), and try_to_freeze() should be called
> after case wait_event().
>
> On the other hand, if "kthreadd" does not sleep on kthread_create_work,
> we have another unfrozen process waiting for kthread_create_info.done.
> So, is there any practical reason why kthreadd() should explicitely go
> to refrigerator?
Good question. Right now, there probably is not any.
Greetings,
Rafael
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
