Re: [PATCH v5] KVM x86/xen: add an override for PVCLOCK_TSC_STABLE_BIT

[Sean Christopherson <seanjc@google.com>]

On Thu, Nov 30, 2023, David Woodhouse wrote:
> On Thu, 2023-11-30 at 16:41 +0000, Paul Durrant wrote:
> > On 30/11/2023 16:36, Sean Christopherson wrote:
> > > +Andrew
> > > 
> > > On Thu, Nov 02, 2023, Paul Durrant wrote:
> > > > From: Paul Durrant <pdurrant@amazon.com>
> > > > 
> > > > Unless explicitly told to do so (by passing 'clocksource=tsc' and
> > > > 'tsc=stable:socket', and then jumping through some hoops concerning
> > > > potential CPU hotplug) Xen will never use TSC as its clocksource.
> > > > Hence, by default, a Xen guest will not see PVCLOCK_TSC_STABLE_BIT set
> > > > in either the primary or secondary pvclock memory areas. This has
> > > > led to bugs in some guest kernels which only become evident if
> > > > PVCLOCK_TSC_STABLE_BIT *is* set in the pvclocks. Hence, to support
> > > > such guests, give the VMM a new Xen HVM config flag to tell KVM to
> > > > forcibly clear the bit in the Xen pvclocks.
> > > 
> > > ...
> > > 
> > > > diff --git a/Documentation/virt/kvm/api.rst b/Documentation/virt/kvm/api.rst
> > > > index 7025b3751027..a9bdd25826d1 100644
> > > > --- a/Documentation/virt/kvm/api.rst
> > > > +++ b/Documentation/virt/kvm/api.rst
> > > > @@ -8374,6 +8374,7 @@ PVHVM guests. Valid flags are::
> > > >     #define KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL            (1 << 4)
> > > >     #define KVM_XEN_HVM_CONFIG_EVTCHN_SEND              (1 << 5)
> > > >     #define KVM_XEN_HVM_CONFIG_RUNSTATE_UPDATE_FLAG     (1 << 6)
> > > > +  #define KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE      (1 << 7)
> > > 
> > > Does Xen actually support PVCLOCK_TSC_STABLE_BIT?  I.e. do we need new uAPI to
> > > fix this, or can/should KVM simply _never_ set PVCLOCK_TSC_STABLE_BIT for Xen
> > > clocks?  At a glance, PVCLOCK_TSC_STABLE_BIT looks like it was added as a purely
> > > Linux/KVM-only thing.
> > 
> > It's certainly tested in arch/x86/xen/time.c, in 
> > xen_setup_vsyscall_time_info() and xen_time_init(), so I'd guess it is 
> > considered to be supported.
> 
> And yes, Xen does set it, if you jump through the right hoops to make
> Xen actually use the TSC as its clocksource.
> 
> The new uAPI is just a single bit in the KVM_XEN_HVM_CONFIG
> capabilities; I think it's reasonable enough.

Yeah, I was just hoping that maybe we could squeak by without it.  I'll get this
queued up next week, purely because I try to avoid (but often fail) pushing to
-next on Fridays.