On Tue, 2011-02-08 at 15:37 +0100, Thomas Gleixner wrote:
> On Tue, 8 Feb 2011, Thomas Gleixner wrote:
>
> > On Mon, 7 Feb 2011, James Bottomley wrote:
> >
> > > On Mon, 2011-02-07 at 15:26 -0600, James Bottomley wrote:
> > > > Unfortunately, it's panicing on boot in txn_alloc (that's a CPU
> > > > interrupt allocation in the iosapic), so I'm going to debug further.
> > >
> > > Following up with Thomas over IRC, the attached is the finally final
> > > patch he came up with to fix the panic and sort out the compile issues.
> > > I've verified this on an iosapic system (pa8800) and on dino/gsc (C360)
> > > could someone try it on an astro system?
> > >
> > > Thanks,
> > >
> > > James
> > >
> > > diff --git a/arch/parisc/kernel/irq.c b/arch/parisc/kernel/irq.c
> > > index 9eea164..cb450e1 100644
> > > --- a/arch/parisc/kernel/irq.c
> > > +++ b/arch/parisc/kernel/irq.c
> > > @@ -113,7 +113,7 @@ int cpu_check_affinity(struct irq_data *d, const struct cpumask *dest)
> > > int cpu_dest;
> > >
> > > /* timer and ipi have to always be received on all CPUs */
> > > - if (CHECK_IRQ_PER_CPU(irq)) {
> > > + if (CHECK_IRQ_PER_CPU(irq_to_desc(d->irq)->status)) {
> >
> > For extended fun I just noticed, that nothing ever sets IRQ_PER_CPU in
> > desc->status for all of parisc.
> >
> > This check (using irq) never evaluated to true due to:
> >
> > #define IRQ_PER_CPU 0x00010000 /* IRQ is per CPU */
> >
> > I doubt that irq can be that large on parisc :)
> >
> > The other place evaluates to false as well, because nothing ever sets
> > the damn flag.
> >
> > If you want to set it please use: set_irq_status_flags(irq, IRQ_PER_CPU);
> > otherwise just kill that PER_CPU thingy :)
>
> Gah. You set it indirect via IRQF_PERCPU with setup/request_irq.
>
> Staring too long into that code makes my brain fuzzy.
Yes ... know the feeling over the years trying to keep up with the
various irq schemes. Got to say, it's a lot easier now: The way parisc
chips accept interrupts is essentially MSI. Trying to live in the old
world modelled on the x86 PIC wasn't fun.
So did you want me to roll all the fixes together and apply them, or is
there something else that needs to happen?
James
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
