On Tue, 2022-03-01 at 17:47 +0700, Suravee Suthikulpanit wrote:
> Hi Maxim,
>
> On 2/25/22 12:18 AM, Maxim Levitsky wrote:
> > On Sun, 2022-02-20 at 20:19 -0600, Suravee Suthikulpanit wrote:
> > > xAVIC and x2AVIC modes can support diffferent number of vcpus.
> > > Update existing logics to support each mode accordingly.
> > >
> > > Also, modify the maximum physical APIC ID for AVIC to 255 to reflect
> > > the actual value supported by the architecture.
> > >
> > > Signed-off-by: Suravee Suthikulpanit<suravee.suthikulpanit@xxxxxxx>
> > > ---
> > > arch/x86/include/asm/svm.h | 12 +++++++++---
> > > arch/x86/kvm/svm/avic.c | 8 +++++---
> > > 2 files changed, 14 insertions(+), 6 deletions(-)
> > >
> > > diff --git a/arch/x86/include/asm/svm.h b/arch/x86/include/asm/svm.h
> > > index 7a7a2297165b..681a348a9365 100644
> > > --- a/arch/x86/include/asm/svm.h
> > > +++ b/arch/x86/include/asm/svm.h
> > > @@ -250,10 +250,16 @@ enum avic_ipi_failure_cause {
> > >
> > >
> > > /*
> > > - * 0xff is broadcast, so the max index allowed for physical APIC ID
> > > - * table is 0xfe. APIC IDs above 0xff are reserved.
> > > + * For AVIC, the max index allowed for physical APIC ID
> > > + * table is 0xff (255).
> > > */
> > > -#define AVIC_MAX_PHYSICAL_ID_COUNT 0xff
> > > +#define AVIC_MAX_PHYSICAL_ID 0XFFULL
> > > +
> > > +/*
> > > + * For x2AVIC, the max index allowed for physical APIC ID
> > > + * table is 0x1ff (511).
> > > + */
> > > +#define X2AVIC_MAX_PHYSICAL_ID 0x1FFUL
> > Yep, physid page can't hold more entries...
> >
> > This brings the inventible question of what to do when a VM has more
> > that 512 vCPUs...
> >
> > With AVIC, since it is xapic, it would be easy - xapic supports up to
> > 254 CPUs.
>
> Actually, 255 vCPUs.
Sorry for off-by-one mistake - just remembered that 0xFF is reserved,
but then 255 is already 1 less that 256.
>
> > But with x2apic, there is no such restriction on max 512 CPUs,
> > thus it is legal to create a VM with x2apic and more that 512 CPUs,
> > and x2AVIC won't work well in this case.
> >
> > I guess AVIC_IPI_FAILURE_INVALID_TARGET, has to be extened to support those
> > cases, even with loss of performance, or we need to inhibit x2AVIC.
>
> In case of x2APIC-enabled guest w/ vCPU exceeding the max APIC ID (512) limit,
> the ioctl operation for KVM_CREATE_VCPU will fail. For QEMU, this would
> exit with error code. Would this be sufficient?
Yes, this is the best.
Best regards,
Maxim Levitsky
>
> Regards,
> Suravee
>
>
>
