32-rc1 aka 32-rc2: warning at manage.c:361 (set_irq_wake), matrix-keypad related?

[Dmitry Torokhov]

On Sat, Jan 02, 2010 at 02:56:09PM +0100, Pavel Machek wrote:
> Hi!
> 
> > > > We ignore the failure right now in the mainline but that causes stack
> > > > traces on resume as we trying to disable not enabled wakeup GPIOs. That
> > > > was original Pavel's complaint.
> > > 
> > > Yep...
> > > 
> > > I'd say that BUG() simply should not trigger if wakeup can not be
> > > enabled/disabled for particular source...?
> > 
> > Pavel,
> > 
> > Could you please try the patch below and let me know if it fixes the
> > problem for you?
> 
> (sorry for long delay).
> 
> With 33-rc1+your patch, warnings are no longer there. 
> 

Thank you for testing.

> I still have one on irq191, but that's gpio button driver.
> 
> But please, hold the patch for few more days. I may have cleaner/nicer
> solution. It seems that zaurus is doing something very stupid in low layers.

Sorry, it is already upstream. However I do think that the change is
needed. As long as set_irq_wake() may return an error the driver should
handle such error. Especially matrix_keypad that operates with large
number of GPIOs but has only one knob for enabling wakeup.

-- 
Dmitry