Hi,
<...>
>> I'm curious why do you think that it is broken now.
>> You are getting a data bus error by any chance?
>>
>
> after a pci write, the pci read return bogus values.
Can you be more specific please? The new code reads bogus values from every
configuration registers or only from some of them?
> However, I go through this tomorrow.
Ok.
<...>
>>> -I never hit the pci controller bug, any steps to replicate?
>>
>> Hm, weird. Your devices are based on AR7240 or or AR7241?
>>
>> Regards,
>> Gabor
>>
>
> AR7241 AH-4A
Yeah, that is different, my Bullet 5M uses an AR7240. I have retested the code
with and without the workaround on an AR7241 based board and it is working in
both cases. So it seems that the AR7241 is not affected. I will change the patch
to use the workaround only if the kernel is running on AR7240.
> Maybe it is better to split ar71xx and ar72xx pci support completly?
Sorry, what do you mean? It is separated already. The common functions are
'pcibios_map_irq', 'pcibios_plat_dev_init' and 'ath79_register_pci'. None of
these functions are related to accessing of the PCI devices. Additionally, the
'ath79_register_pci' function calls the appropriate 'ar7{1x,24}x_pcibios_init'
depending on the actual SoC, so the AR71XX specific PCI controller code never
runs on AR724X.
-Gabor
