[Android-virt] [PATCH] ARM: test for PMU feature on v7 (v2 with typo fix)

[Will Deacon]

On Mon, Mar 26, 2012 at 10:50:35AM +0100, Peter Maydell wrote:
> On 26 March 2012 10:40, Will Deacon <will.deacon at arm.com> wrote:
> > On Fri, Mar 23, 2012 at 11:17:07PM +0000, Rusty Russell wrote:
> >>
> >> OK, I see that argument, but I don't like it.  eg. the Cortex A-15 is
> >> defined to have a PMUv2 with 6 counters, and below you're arguing we
> >> should implement zero of them.
> >
> > Well, I think you should actually implement all 6 of them if that's the
> > target which you're emulating.
> 
> The trouble with this line of thought is that it leads to the requirement
> to support nested virtualization and virtualization of TrustZone, neither
> of which are actually particularly useful or simple to do. At some point
> we're going to end up saying that the "A15" we expose to the guest is
> not an exact match to the real hardware, that it's missing certain
> features (and hopefully that the lack of those features is advertised
> via the feature registers). So the question is where you draw the line...

I'd be inclined to say that, at least for kvm, you should draw the line at
the point where the guest doesn't need modifications to run in the virtual
environment. That is, everything used by Linux is emulated enough for the
kernel to run happily. Clearly you can improve on that in the areas which
are considered to be most critical.

Aside: given that guests have to run non-secure, how would you emulate a
secure platform without emulating the corresponding TrustZone API? Or is
that one of the reasons why you're currently targetting Versatile Express?

Will