Re: [PATCH] xen/arm: Handle translated addresses for hardware domains in GICv2

[Ian Campbell <ian.campbell@citrix.com>]

On Tue, 2015-02-24 at 16:21 +0000, Julien Grall wrote:
> On 24/02/15 16:01, Ian Campbell wrote:
> > On Mon, 2015-02-23 at 16:30 +0000, Julien Grall wrote:
> >> I'm wondering if it's worth to backport this patch to Xen 4.5?
> > 
> > Maybe, I doubt 4.5 is being used on platforms which suffer from this
> > issue, but it's pretty un-intrusive I suppose.
> 
> There is no reason to consider only the set of platforms we support
> upstream.

No, but it is a factor: we shouldn't be risking destabilising things for
those platforms which were supported out of the box on 4.5 for the
benefit of platforms which were and are not.

Of course risk is a judgement call, and as I said this change is pretty
unobtrusive (read: probably not that risky), hence "maybe" rather than
"no".

>  We should take into account people who want to ship a product
> based on Xen 4.5.

Again, this is a factor but not the most important one.

>  This release aims to work anywhere by only adding few
> lines of code in the platform code.

And so is this.

> Furthermore, the current code rely on how the device has been written.
> For instance the calxeda device tree is using ranges and, therefore, the
> regs contains offset.
> Fortunately, the offset and the translated address is the same.
> 
> > I certainly wouldn't want to backport a fix to gic v2 only.
> 
> As we didn't backport other patches related to GICv3, it would not make
> much sense to backport a similar fix for GICv3..

We should consider each patch/series on its own merits. The other GICv3
fixes were far more invasive and complicated than what is being done
here.

Assuming the v3 fix is similar to the v2 one here either both are
justified or neither.

Ian.