[RFC PATCH 0/4] Add ACPI support for HiSilicon PCIe Host Controllers

[Gabriele Paoloni]

Hi Arnd, Sinan

> -----Original Message-----
> From: Sinan Kaya [mailto:okaya at codeaurora.org]
> Sent: 08 February 2016 14:12
> To: Arnd Bergmann; linux-arm-kernel at lists.infradead.org
> Cc: Gabriele Paoloni; Lorenzo.Pieralisi at arm.com; jcm at redhat.com;
> tn at semihalf.com; linux-pci at vger.kernel.org; Linuxarm; xuwei (O); linux-
> kernel at vger.kernel.org; linux-acpi at vger.kernel.org; Wangzhou (B);
> liudongdong (C); Guohanjun (Hanjun Guo); bhelgaas at google.com;
> zhangjukuo; Liguozhu (Kenneth); qiujiang
> Subject: Re: [RFC PATCH 0/4] Add ACPI support for HiSilicon PCIe Host
> Controllers
> 
> On 2/8/2016 8:55 AM, Arnd Bergmann wrote:
> > I haven't really followed what is going on with ACPI. Do you expect
> > to see future machines come out that are not just implementing SBSA
> > but that still need to run ACPI? I thought this was just a hack
> > for some early machines that only run with ACPI but are not actually
> > compliant.

Well from our side (HiSilicon) we're trying to move away from non fully 
ECAM platforms, so from us in the long term I don't expect too many quirks,
but I don't know about the other vendors.

Obviously the reason why Tomasz implemented the quirks is to fit non 
fully ECAM HW and to allow custom HW init; this is why I thought better
to have the ACPI version in the same dir as the DT (maybe we can create 
an ACPI sub-dir in drivers/pci/host ?)


> >
> > 	Arnd
> 
> I agree. We shouldn't be playing with half-baked ACPI solutions. We
> have seen
> two variants already that claim to be ACPI compliant yet they do not
> tie into
> anything inside ACPICA.
> 
> The correct route is to use Tomasz's ACPI PCI root bridge driver and
> use the ACPI
> framework.
> 
> If a platform has quirks, Tomasz's patches allow vendors add quirks
> too.
> 
> The combination of PCI host bridge driver + ACPI hack is not right.

If you look at my patchset  you can see that I didn't do any hack,

I just used the framework provided by Tomasz patchset.

The discussion here is more about the code location for the quirks.
Since the configuration read/write and the HW init sequences can be
similar between the ACPI variant and DT variant I thought it make 
sense to have them in "drivers/pci/host"

Thanks

Gab

> 
> --
> Sinan Kaya
> Qualcomm Technologies, Inc. on behalf of Qualcomm Innovation Center,
> Inc.
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a
> Linux Foundation Collaborative Project