[ARM] head.S change broke platform device registration?

On Tue, Dec 04, 2012 at 10:48:56PM +0100, Marko Kati? wrote:

I have included the complete dmesg log of vanilla 3.7.0-rc7 in my
previous mail.
Here's a relevant snippet of it:

Linux version 3.7.0-rc7+ (dromede at dromedary) (gcc version 4.7.2 (GCC)
) #63 PREEMPT Fri Nov 30 13:49:35 CET 2012
CPU: XScale-PXA270 [69054117] revision 7 (ARMv5TE), cr=0000397f
CPU: VIVT data cache, VIVT instruction cache
Machine: SHARP Akita
Memory policy: ECC disabled, Data cache writeback
BUG: mapping for 0x00000000 at 0xff000000 out of vmalloc space

....

Sharp Scoop Device found at 0x10800000 -> 0xc4846000


It does seem that the kernel boots with the correct platform id.
I doubt that the second scoop device somehow got registered
and blocked those gpio numbers. It would fail to register, this would
be visible in dmesg output. Also, the second scoop device starts at
0x08800040. So the above registered scoop device is scoop device 1.

Ok, so the correct boot ID throws my theory out, but I don't see
anything else that would claim those GPIOs and prevent the MAX device
registering its GPIOs.

You're going to have to boot -rc7, mount debugfs and read
/sys/kernel/debug/gpio to find out what is claiming those GPIOs that
the MAX device wants to use.  There is only _one_ other device for PXA
for Sharp devices which is hard-coded to the same GPIO numbers and that's
the Scoop 2 device - but that's not registered for the !machine_is_akita()
case.

Everything you've reported points to machine_is_akita() being false, which
it won't be given your Machine: line above.

So.. I don't know, and no one can explain the behaviour you're seeing.