Re: [Xen-devel] [PATCH RFC 16/35] ARM64 / ACPI: Parse GTDT to initialize timer

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Thu, 5 Feb 2015, Ian Campbell wrote:
> On Wed, 2015-02-04 at 21:51 +0000, Julien Grall wrote:
> > > +    res = platform_init_time();
> > 
> > The platform code is DT-centrict.
> 
> This is an interesting point. Given the stated goals and reasons for
> having ACPI on ARM it seems to me that in general nothing under
> xen/arch/arm/platforms/* should ever be needed, and certainly not things
> like platform time init which we should insist have been setup correctly
> by the platform.

At the moment I don't think that ACPI on ARM is in the state where we
can get away without anything under platforms.


> It's possible that we might need some concept of ACPI quirks at some
> point, but lets have a simple rule for now:
> 
>       * The initial set of core ACPI patches for Xen should not ever
>         call into any platform specific code.

I think that is a good rule for the initial implementation.
If/When we need platform files on acpi systems, we might want to
introduce a new separate struct platform_desc for them anyway.


> If there turns out to be a need to call platform code in this initial
> series then it must be in a separate patch towards the end of the
> series, after all the core code has been introduced. The patch
> description should contain a detailed justification as to why this
> platform specific code is required.
> 
> I think it is probable that any ACPI quirks should be separate from the
> (DT-based) xen/arch/arm/platforms/* stuff, but lets see what, if
> anything, turns out to be needed.

I agree


> Perhaps platform_* should all grow an "ASSERT(!in acpi mode)" (or
> "ASSERT(in dt mode)") to help enforce this.
> 
> Ian.
> 

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel