Re: [Qemu-devel] [PATCH] hw/arm/virt-acpi-build: drop _ADR entry from SPCR

[Shannon Zhao]

On 2015/8/6 20:28, Andrew Jones wrote:
> On Thu, Aug 06, 2015 at 12:24:27PM +0100, Leif Lindholm wrote:
> > >The _ADR entry in SPCR is optional and redundant. The same information
> > >is already provided in _CRS (which is mandatory).
> > >
> > >Signed-off-by: Leif Lindholm<address@hidden>
> > >---
> > >
> > >So, this _ADR entry is only consumed by a set of not-widely-circulated
> > >patches for the Linux kernel. And while the ARM Server Base Boot
> > >Requirements specification mandates SPCR, it does not mandate this _ADR
> > >entry.
> > >
> > >In the interest of not propagating non-standard extensions, I would be
> > >really happy if we could consider dropping this from 2.4.
> > >I do realize that this is a completely unreasonable request this late
> > >in the release process, but I only spotted this yesterday, and it is a
> > >very isolated change with very quantifiable effects.
> > >
> > >The patch 
> > athttps://git.linaro.org/leg/acpi/leg-kernel.git/commitdiff/46eeec7b7332bdd104941703696d3812afd934c8
> > >converts the non-upstream kernel SPCR handling code to use the _CRS
> > >information instead.
> Grr... So when I saw how the kernel (the original non-upstream patch)
> was using ADR, I presumed that that was a documented behavior. I guess
> I should have confirmed that...
>
> While the kernel change makes sense, I'm not sure we want this QEMU
> patch, as there*are*  kernels already using ADR. In the least I wouldn't
> want to get burned twice, so I'd prefer to see the SPCR code actually
> get into Linux first this time. That would also allow us to point at
> something when we start breaking guests.

I agree. It would be better after the kernel patch get into upstream kernel.

> Actually, why was ADR used
> the first time? It would be good to know the story there in case there
> as a valid reason.

--
Shannon