On Tue, Sep 13, 2022, Suthikulpanit, Suravee wrote:
> Hi Sean
>
> On 9/2/2022 7:22 PM, Sean Christopherson wrote:
> > Disable the optimized APIC logical map if multiple vCPUs are aliased to
> > the same logical ID. Architecturally, all CPUs whose logical ID matches
> > the MDA are supposed to receive the interrupt; overwriting existing map
> > entries can result in missed IPIs.
> >
> > Fixes: 1e08ec4a130e ("KVM: optimize apic interrupt delivery")
> > Signed-off-by: Sean Christopherson <seanjc@xxxxxxxxxx>
> > Reviewed-by: Maxim Levitsky <mlevitsk@xxxxxxxxxx>
> > ---
> > arch/x86/kvm/lapic.c | 5 +++--
> > 1 file changed, 3 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> > index 6b2f538b8fd0..75748c380ceb 100644
> > --- a/arch/x86/kvm/lapic.c
> > +++ b/arch/x86/kvm/lapic.c
> > @@ -303,12 +303,13 @@ void kvm_recalculate_apic_map(struct kvm *kvm)
> > if (!mask)
> > continue;
> > - if (!is_power_of_2(mask)) {
> > + ldr = ffs(mask) - 1;
> > + if (!is_power_of_2(mask) || cluster[ldr]) {
>
> Should this be checking if the cluster[ldr] is pointing to the same struct
> apic instead? For example:
>
> if (!is_power_of_2(mask) || cluster[ldr] != apic)
>
> From my observation, the kvm_recalculate_apic_map() can be called many
> times, and the cluster[ldr] could have already been assigned from the
> previous invocation. So, as long as it is the same, it should be okay.
No, because cluster[ldr] can never match "apic". kvm_recalculate_apic_map()
creates and populates a _new_ kvm_apic_map every time, it doesn't do an in-place
update of the current map.
The loop containing this code is:
kvm_for_each_vcpu(i, vcpu, kvm) {
struct kvm_lapic *apic = vcpu->arch.apic;
...
}
so it's impossible for cluster[ldr] to hold the current "apic", because this is
the first and only iteration that processes the current "apic".
