On Sun, Sep 04, 2016 at 09:04:59PM +0200, Robert Jarzmik wrote:

And retested with my 3 patches on top of it. Everything works fine :
 - CF insertion is correctly detected now !
 - AT/2 keyboard interrupts fire, keys are there, etc ...
 - the AT/2 warning is now gone

The pxa_cplds_irqs patch will make it work as before, better than today for
sure. I think edge type interrupts (such as sa1111), if they come close enough,
might be lost with the current implementation, but anyway, that's for another
time.

Good news, but I suspect CF probably isn't working as it should do (see
my reply to the clock alias patch.)

There is one patch in my serie that I left up to you, as I wasn't very sure
about the accuracy of my commit message, nor been able to write down all the
requirements implied on sa1111 with this additional test.

That's basically what I have here at the moment, but since I made the
suggestion, I've been thinking about moving it into the IRQ
initialisation instead.

Also, I think it's a genirq problem that you can install a chained
handler on an interrupt without a registered chip - and then when an
irq chip comes along later, the IRQ isn't unmasked.  We should either
not allow a chained handler to be installed in that situation, /or/
we should unmask chained IRQs when the chip comes along later.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.