Re: sh: intc - add support for SH7760

[Magnus D. <mag...@gm...>]

Hi again Manuel,

On 7/25/07, Manuel Lauss <ma...@ro...> wrote:
> On Wed, Jul 25, 2007 at 11:16:39PM +0900, Magnus Damm wrote:
> > On 7/25/07, Manuel Lauss <ma...@ro...> wrote:
> > > On Wed, Jul 25, 2007 at 10:17:39PM +0900, Magnus Damm wrote:
> > > > sh: intc - add support for SH7760
> > >
> > > > +     INTC_GROUP(MMCIF, MMCIF0, MMCIF1, MMCIF2, MMCIF3),
> > >
> > > With this patch my mmcif driver not longer works, since no
> > > mmcif irq is generated.  In the driver I only request
> > > MMCIF0-2; MMCIF3 is unused. Does the grouping above mask all
> > > grouped IRQs unless all have been enabled/requested?
> >
> > It should be possible to request a single interrupt only.
> >
> > The grouping is there because the priority value is shared between all
> > MMCIF interrupts. We need some way to tell the intc code which
> > interrupt sources that are bundled together, and that's what the
> > groups are used for.
> >
> > The sh7780 device supports both masking with both priorities and
> > bitmaps. The common intc code should prefer to use the per-interrupt
> > source bitmap for MMCIF0-3 over the priority group to mask and unmask.
> > So interrupts should be masked per-interrupt source for MMCIF.
> >
> > So.. are things working as before, or have i mistakenly introduced
> > some new behavior? =)
>
> New behavior -- unless all 4 of the grouped MMCIF irqs are requested,
> none trigger. They show up in /proc/interrupts but count does not change.

I have two ideas:

1. Please try to add MMCIF3 to mask_registers[]. I must have missed
that one. The "0" after MMCIF2 should be replaced with MMCIF3. This is
a pretty obvious miss that I made when I created the tables from the
data sheet. I hope that this fix helps.

2. If number 1 above doesn't help, please try replacing all MMCIF0-3
with 0 in mask_registers[]. That should force masking via the priority
register.

Could you please check if the ideas above help? Number 1 is preferred over 2.

Thanks!

/ magnus