Re: [PATCH v3 edk2-platforms 03/23] Platform/Raspberry/Pi3: Add ACPI tables

[Ard Biesheuvel <ard.biesheuvel@linaro.org>]

On Tue, 29 Jan 2019 at 13:54, Pete Batard <pete@akeo.ie> wrote:
>
> Hi Ard, thanks for the reviews,
>
> On 2019.01.28 13:24, Ard Biesheuvel wrote:
> > A couple of general remarks on these tables.
> >
> > - AFAICT (but is is difficult to tell from the binary dumps), these
> > are based on ACPI 5.0, which predates the introduction of ARM support
> > into the spec. That essentially makes them Windows-only, and not spec
> > based, which is unfortunate. However, since this SoC does not have an
> > interrupt controller permitted by ACPI, that actually does not matter
> > a lot in practice.
>
> Publicly, these tables are only meant to boot Windows and we got them
> from Microsoft. So they are not exactly meant to be ACPI compliant (for
> instance, AFAIK, there's no definition of ACPI that includes the RPI
> interrupt controller).
>
> As such, we really don't want to touch these because we have run into
> nightmarish situation where Windows wouldn't boot at all after we tried
> to alter them.
>
> In other words, if we are to try to change much in here, we will
> probably need to have Microsoft's involvement...
>
> > - _CID fields should be properly formatted ACPI or PNP handles, you
> > can't put arbitrary strings in there (like BCMAUXSPI or VC4)
>
> The Microsoft Windows drivers expect those exact IDs, so that's not
> something we can alter either. Once again, these ACPI tables were not
> designed to be ACPI compliant, and making them so that late in the game
> is going to break a lot of things downstream, so we don't believe that
> we can do that at this stage.
>

No, I kind of figured that. It is just very unfortunate that we have
two completely separate descriptions for Windows vs Linux, so that is
why I brought it up.

>
> I will apply your other request for ACPI (drop the "-mcmodel=tiny"
> workaround in [BuildOptions]), as well as the changes you requested for
> 1/2/19 and, if you are okay with having our hands tied with regards to
> ACPI compliance, submit a v4 later on today.
>

Sure, although I was expecting Leif to chime in as well. Leif?