Re: [PATCH 5/7] KVM: x86: Move kvm_set_apic_base() implementation to lapic.c (from x86.c)

[Sean Christopherson]

On Mon, Oct 14, 2024, Sean Christopherson wrote:
> On Mon, Oct 14, 2024, Kai Huang wrote:
> > On Wed, 2024-10-09 at 11:17 -0700, Sean Christopherson wrote:
> > > Move kvm_set_apic_base() to lapic.c so that the bulk of KVM's local APIC
> > > code resides in lapic.c, regardless of whether or not KVM is emulating the
> > > local APIC in-kernel.  This will also allow making various helpers visible
> > > only to lapic.c.
> > > 
> > > No functional change intended.
> > > 
> > > Signed-off-by: Sean Christopherson <seanjc@xxxxxxxxxx>
> > > ---
> > >  arch/x86/kvm/lapic.c | 21 +++++++++++++++++++++
> > >  arch/x86/kvm/x86.c   | 21 ---------------------
> > >  2 files changed, 21 insertions(+), 21 deletions(-)
> > > 
> > > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> > > index fe30f465611f..6239cfd89aad 100644
> > > --- a/arch/x86/kvm/lapic.c
> > > +++ b/arch/x86/kvm/lapic.c
> > > @@ -2628,6 +2628,27 @@ void kvm_lapic_set_base(struct kvm_vcpu *vcpu, u64 value)
> > >  	}
> > >  }
> > >  
> > > +int kvm_set_apic_base(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> > > +{
> > > +	enum lapic_mode old_mode = kvm_get_apic_mode(vcpu);
> > > +	enum lapic_mode new_mode = kvm_apic_mode(msr_info->data);
> > > +	u64 reserved_bits = kvm_vcpu_reserved_gpa_bits_raw(vcpu) | 0x2ff |
> > > +		(guest_cpuid_has(vcpu, X86_FEATURE_X2APIC) ? 0 : X2APIC_ENABLE);
> > > +
> > > +	if ((msr_info->data & reserved_bits) != 0 || new_mode == LAPIC_MODE_INVALID)
> > > +		return 1;
> > > +	if (!msr_info->host_initiated) {
> > > +		if (old_mode == LAPIC_MODE_X2APIC && new_mode == LAPIC_MODE_XAPIC)
> > > +			return 1;
> > > +		if (old_mode == LAPIC_MODE_DISABLED && new_mode == LAPIC_MODE_X2APIC)
> > > +			return 1;
> > > +	}
> > > +
> > > +	kvm_lapic_set_base(vcpu, msr_info->data);
> > > +	kvm_recalculate_apic_map(vcpu->kvm);
> > > +	return 0;
> > > +}
> > 
> > Nit:
> > 
> > It is a little bit weird to use 'struct msr_data *msr_info' as function
> > parameter if kvm_set_apic_base() is in lapic.c.  Maybe we can change to take
> > apic_base and host_initialized directly.
> > 
> > A side gain is we can get rid of using the 'struct msr_data apic_base_msr' local
> > variable in __set_sregs_common() when calling kvm_apic_set_base():
> 
> Ooh, nice.  I agree, it'd be better to pass in separate parameters.
> 
> Gah, and looking at this with fresh eyes reminded me why I even started poking at
> this code in the first place.  Patch 1's changelog does a poor job of calling it
> out,

Duh, because patch 1 doesn't change any of that.  KVM already skips setting the
map DIRTY if neither MSR_IA32_APICBASE_ENABLE nor X2APIC_ENABLE is toggled.  So
it's really just the (IIRC, rare) collision with an already-dirty map that's nice
to avoid.

> but the main impetus for this series was to avoid kvm_recalculate_apic_map()
> when doing KVM_SET_SREGS without modifying APIC_BASE.  That's _mostly_ handled by
> patch 1, but it doesn't completely fix things because if the map is already DIRTY,
> then KVM will unnecessarily fall into kvm_recalculate_apic_map()'s slow path, even
> though some other vCPU/task is responsible for refreshing the calculation.
> 
> I'll send a v2 with your suggested change, a better changelog for patch 1, and
> another patch at the end to short-circuit kvm_apic_set_base() (not just the inner
> helper) if the new value is the same as the old value.
> 
> Thanks Kai!
> 
> > static int __set_sregs_common(...)
> > {
> >         struct msr_data apic_base_msr;
> > 	...
> > 
> >         apic_base_msr.data = sregs->apic_base;
> >         apic_base_msr.host_initiated = true;
> >         if (kvm_set_apic_base(vcpu, &apic_base_msr))
> >                 return -EINVAL;
> > 	...
> > }
> >