Re: [PATCH V1 11/11] arm64, pci, acpi: Support for ACPI based PCI hostbridge init

[David Daney <ddaney@caviumnetworks.com>]

On 11/03/2015 07:19 AM, Hanjun Guo wrote:
> On 11/03/2015 10:15 PM, Lorenzo Pieralisi wrote:
>> On Wed, Oct 28, 2015 at 02:46:37PM -0400, Sinan Kaya wrote:
>>
>> [...]
>>
>>>> -int raw_pci_write(unsigned int domain, unsigned int bus,
>>>> -        unsigned int devfn, int reg, int len, u32 val)
>>>> +struct pci_ops pci_root_ops = {
>>>> +    .map_bus = pci_mcfg_dev_base,
>>>> +    .read = pci_generic_config_read,
>>>> +    .write = pci_generic_config_write,
>>>
>>>
>>> Can you change these with pci_generic_config_read32 and
>>> pci_generic_config_write32? We have some targets that can only do 32
>>> bits PCI config space access.
>>
>> No.
>>
>> http://www.spinics.net/lists/linux-pci/msg44869.html
>>
>> Can you be a bit more specific please ?
>>
>> Sigh. Looks like we have to start adding platform specific quirks even
>> before we merged the generic ACPI PCIe host controller implementation.
>
> Cc Gab, Zhou, and Dondong who upstream the hip05 (designware) PCIe host
> support.
>
> I think so, some platform may not support ECAM for root complex,
> which needs special handling of access config space, we may need
> to consider those cases.
>

Yes, it is indeed true.  For example, some Cavium ThunderX processors 
fall into this category.

Some options I thought of are:

  o Use DECLARE_ACPI_MCFG_FIXUP() in the kernel to supply the needed 
config space accessors.

  o Define additional root_device_ids that imply the needed config space 
accessors.


> Thanks
> Hanjun