On 03/06/16 09:55, Lorenzo Pieralisi wrote:
> On Fri, May 27, 2016 at 01:25:14PM +0100, Marc Zyngier wrote:
>
> [...]
>
>>> + } else {
>>> + bus = pci_scan_root_bus(&pdev->dev, 0,
>>> + &rockchip_pcie_ops, port, &res);
>>> + }
>>> + if (!bus)
>>> + return -ENOMEM;
>>> +
>>> + if (!pci_has_flag(PCI_PROBE_ONLY)) {
>>
>> Why do you have catter for the PCI_PROBE_ONLY case? Nobody should ever
>> use that for properly implemented HW.
>
> I think that's just copy and paste and it is a useless check given
> that the only way we can set that flag on ARM/ARM64 is through DT
> (of_pci_check_probe_only()) and I doubt that systems probing this
> driver really require a PCI_PROBE_ONLY set-up.
>
> So, unless you can explain to us why it is really needed, please
> remove the:
>
> if (!pci_has_flag(PCI_PROBE_ONLY))
>
> check.
Agreed. Maybe we should add a comment somewhere saying that this is
deprecated on arm/arm64, and only useful to slightly dumb virtualization
environments (kvmtool being the prime example).
Thanks,
M.
--
Jazz is not dead. It just smells funny...
