Re: [RFC] [PATCH 1/3] OMAP4: Keyboard Controller Support

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Tue, May 11, 2010 at 07:53:23AM -0700, Kevin Hilman wrote:
> Dmitry Torokhov <dmitry.torokhov@gmail.com> writes:
> 
> > On Tue, May 11, 2010 at 12:03:44AM -0500, Arce, Abraham wrote:
> >> Hi again Dmitry,
> >> 
> >> > No worries, although at first I was surprised that Trilok spoke exactly
> >> > the same words I did ;)
> >> > 
> >> 
> >> :)
> >> 
> >> > > > > > +
> >> > > > > > +/* Interrupt thread handler thread */
> >> > > > > > +
> >> > > > > > +static irqreturn_t omap_keypad_threaded(int irq, void *dev_id)
> >> > > > > > +{
> >> > > > >
> >> > > > > Why is iti threaded? I fo not see anything that will sleep.
> >> > >
> >> > >
> >> > > It was implemented based on previous comments...
> >> > >
> >> > 
> >> > Would you point me to that comment? Like I said, I do not see anything
> >> > that would possibly sleep in this routine so you don't need to use
> >> > threaded interrupt.
> >> 
> >> http://www.mail-archive.com/linux-omap@vger.kernel.org/msg26352.html
> >> 
> >
> > Thanks.
> >
> > I think Kevin meant "use theaded IRQ, wherever possible" [if we need
> > to sleep in interrupt handler].
> 
> Actually, even interrupts that don't sleep can use threaded IRQs.  I
> prefer to see threaded IRQs wherever possible.  Especially since we're
> moving towards a world where all interrupts are run with interrupts
> disabled, using threaded IRQs minimizes interrupts-off critical
> sections.
> 

I think in this case threaded IRQ would just add unnecessary overhead.
There are no scanning delays, just a few register reads and writes.
Input core will take some cycles propagating the events but it disables
interrupts anyway. Setting up a separate thread and scheduling does not
make much sense here.

Also I am not sure if arches with large number of interrupts would want
to move to all threaded interrupts model.

-- 
Dmitry