On Mon, 2016-08-01 at 15:08 +0200, Andrew Jones wrote: > > I'm not sure a warning is enough: if I start a guest and > > explicitly ask for a PMU, I expect it to be there, or for > > the guest not to start at all. How does x86 behave in this > > regard? > > Peter had a good suggestion for this. We need to wrap the property > addition in an arm_feature check like the has_el3 property. That will > remove it from all cpu types that don't support it.
Wouldn't that mean that you'd be unable to use -cpu foo,pmu=off if CPU model 'foo' doesn't support a PMU? I'd expect that to work. I've played around with this a bit on x86 and it doesn't look like it necessarily behaves the way I'd expect it to, either, so maybe this is just a case of my expectations being unreasonable? :) -- Andrea Bolognani / Red Hat / Virtualization
