Re: [Xenomai-help] ARM gpio pins interrupts

[Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>]

Henri Roosen wrote:
> On Thu, Jan 20, 2011 at 4:12 PM, Gilles Chanteperdrix
> <gilles.chanteperdrix@xenomai.org> wrote:
>> Henri Roosen wrote:
>>> Hi,
>>>
>>> I have an ARM based platform which is similar to the at91sam9263ek
>>> eval board. One of the differences is that touchscreen irq is on pin
>>> PB31 instead of PA15. The 'arch/arm/mach-at91/board-sam9263ek.c' has
>>> been changed for this.
>>>
>>> Touch works when using the kernel without Xenomai.
>>> When using the same kernel with Xenomai patches, touch doesn't work. I
>>> get only 2 interrupts on the GPIO pin (cat /proc/interrupts).
>>>
>>> Anyone an idea why the Xenomai enabled kernel stops generating
>>> interrupts on the GPIO pins while the non-Xenomai enabled kernel
>>> doesn't?
>> Xenomai does not change anything. However, the I-pipe patch changes the
>> way interrupts are handled. Now, we have to figure out exactly how this
>> is an issue.
>>
>> Is this GPIO a muxed GPIO?
> 
> Pin PB31 is a muxed GPIO IRQ.
> 
>> As a first test, could you try commenting out the
>> #define __IPIPE_FEATURE_PIC_MUTE
>> in arch/arm/mach-at91/include/mach/irqs.h ?
>>
> 
> Commenting out the define has no difference; still get 2 irq's.
> 
>> Have you tried commenting out the line you show us which disables
>> interrupts? Can you explain why this interrupt line is using
>> handle_simple_IRQ if it is not a muxed GPIO?
> 
> I tried not disabling the IRQ and not setting the irq_disabled flag in
> the isr, but that unfortunately always resulted in unbalanced enables
> for the IRQ and finally a lock up of the system. Have to dig into the
> driver code a little deeper if I want to balance it.
> Interesting is that I get more IRQ's than just 2. Even many more than
> I expect for one touch (>100) which tells me the comment in the
> driver's isr makes sense. And which also indicates this problem is
> likely related to disabling the irq here.
> 
> (BTW: tried ipipe 1.18-01 which shows the same problem)

If you comment out the disable_irq, you should comment out the
"enable_irq" counterpart. You can also try and replace handle_simple_irq
with handle_level_irq, to see if it changes anything. Though,
handle_simple_irq seems to be the rule for muxed gpios, but other
drivers probably do not try and disable them in their handlers.

-- 
                                                                Gilles.