Re: [RFC][PATCH 2/2] Freezer: Try to handle killable tasks

[Pavel Machek <pavel@suse.cz>]

On Wed 2008-05-07 20:41:19, Rafael J. Wysocki wrote:
> On Wednesday, 7 of May 2008, Matthew Wilcox wrote:
> > On Wed, May 07, 2008 at 12:07:55AM +0200, Rafael J. Wysocki wrote:
> > > The introduction of TASK_KILLABLE allows the freezer to work in some situation
> > > that it could not handle before.
> > > 
> > > Make the freezer handle killable tasks and add try_to_freeze() in some places
> > > where it is safe to freeze a (killable) task.  Introduce the
> > > wait_event_killable_freezable() macro to be used wherever the freezing of
> > > a waiting killable task is desirable.
> > 
> > Why do you say that TASK_KILLABLE allows the freezer to work in some
> > situations where it couldn't before?  If something's using one of the
> > killable functions, it means that it knows how to clean up and unwind
> > gracefully if the task receives a fatal signal.  I don't understand what
> > connection there is to the freezer.
> 
> The reason why we don't freeze uninterruptible tasks is that we don't know
> why they are in that state.  If one of tasks is uninterruptible for a
> relatively long time, that may indicate a non-recoverable error making it
> dangerous to put the system into a sleep state.  If the task is killable,
> though, the situation is recoverable.

....but the task may still hold some locks, so we can't "just freeze
it".
								Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html