Re: [PATCH 2/4] x86/amd_nb: add support for newer PCI topologies

[Bjorn Helgaas <helgaas@kernel.org>]

On Mon, Nov 05, 2018 at 10:56:50PM +0100, Borislav Petkov wrote:
> On Mon, Nov 05, 2018 at 03:45:37PM -0600, Bjorn Helgaas wrote:
> > amd_nb.c prevents us from achieving that goal.  These patches don't
> > add new functionality; they merely describe minor topographical
> > differences in new hardware.  We usually try to do that in a more
> > generic way, e.g., via an ACPI method, so the new platform can update
> > the ACPI method and use an old, already-qualified, already-shipped
> > kernel.
> > 
> > I'm not strenuously objecting to these because this isn't a *huge*
> > deal, but I suspect it is a source of friction for distros that don't
> > want to update and requalify their software for every new platform.
> 
> Err, how is this any different from adding distro support for a new CPU
> family?
> 
> This is basically the same thing. When distros add support for new
> hardware, they have to backport patches for upstream. These PCI devices
> which are part of the CPU are part of that hardware enablement.
> 
> So there's no way around doing that enablement. I don't think you can do
> "old distro, new hardware" stuff without *some* hw enablement.

Just depends on how hard you want to work on defining abstract
interfaces.

This isn't some complicated new device where the programming model
changed on the new CPU.  This is a thermometer that was already
supported.  ACPI provides plenty of functionality that could be used
to support this generically, e.g., see drivers/acpi/thermal.c,
drivers/thermal/int340x_thermal/processor_thermal_device.c, etc.

But maybe there's some real value in the nitty-gritty device-specific
code in amd_nb.c.  If so, I guess you're stuck with updates like this
and negotiating with the distros to do backports and new releases.

Bjorn