On Thu, Jun 20, 2013 at 10:34:39AM +0200, Paolo Bonzini wrote:
> Il 20/06/2013 09:31, Gleb Natapov ha scritto:
> >> I agree with you on the potential problems but I think we are completely
> >> ignoring the "non-migration" use case. These users will probably benefit
> >> from a correct value of (virtual) msr_platform_info. And it appears, the
> >> easiest way to give both options to the user is using a new module_param,
> >> say ignore_platform_info.
> >>
> > Migration is important part of virtualization. PMU emulation currently
> > is in demo status because migration is not implemented yet, but at least
> > it is possible to implement it.
>
> But there's a lot more to do for migration than just moving the MSR
> contents from one machine to another. We need to support customizing
> the values of CPUID leaf 0ah, so that PMU migration can work if you
> specify "lowest common denominator" CPU models.
>
> I haven't looked at it yet because QEMU support for this is not in sight.
>
Eduardo working on it.
> >> Scenarios :
> >> 1. ignore_platform_info = 0 (Default). Inject #GP if application tries to
> >> read this MSR.
> >> 2. ignore_platform_info = 1. User wants to read the calculated value, her
> >> environment doesn't require migration.
> >> 3. ignore_msrs = 1. If this is set, we always return 0 and application will
> >> hopefully resort to a workaround.
> >>
> > Module flag is global for all VMs on a host. Implementing it like this
> > will guaranty that the feature will not be used in production ever.
> > ignore_msrs exists only for developers to quickly check if a problem goes
> > away if some MSR does not #GP, never as a real solution.
> >
> > To make it somewhat useful the flag should be per-vm and exposed to all
> > layers up to a management. Management is the one who enables it per VM
> > basis and guaranties that VM with the feature enabled is never live
> > migrated. Frankly IMO it will be another management knob that will never
> > be set.
>
> I agree.
>
> I think it's fine to apply Bandan's patches. It's just one more thing
> to care about when migrating machines that use the PMU. And I hope that
> Intel will add TSC scaling sooner or later, which will fix the issue
> automatically. Hear, Jun! :)
>
I am find with adding things without migration support if we know how
migration can be fixed, in this case we are adding something that cannot
be fixed without HW support and, as such, should be enabled only on
hypothetical future HW that will support required feature.
So enabling it by default is wrong, and I listed pros and cons of global
ignore_platform_info option and per vm enable option above.
--
Gleb.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
