Re: [PATCH v2] KVM: x86: inhibit APICv when KVM_GUESTDBG_BLOCKIRQ active

[Vitaly Kuznetsov <vkuznets@redhat.com>]

Maxim Levitsky <mlevitsk@redhat.com> writes:

> On Wed, 2021-11-03 at 14:28 +0100, Vitaly Kuznetsov wrote:
>> Maxim Levitsky <mlevitsk@redhat.com> writes:
>> 
>> > KVM_GUESTDBG_BLOCKIRQ relies on interrupts being injected using
>> > standard kvm's inject_pending_event, and not via APICv/AVIC.
>> > 
>> > Since this is a debug feature, just inhibit it while it
>> > is in use.
>> > 
>> > Fixes: 61e5f69ef0837 ("KVM: x86: implement KVM_GUESTDBG_BLOCKIRQ")
>> > 
>> > Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
>> > ---
>> >  arch/x86/include/asm/kvm_host.h | 1 +
>> >  arch/x86/kvm/svm/avic.c         | 3 ++-
>> >  arch/x86/kvm/vmx/vmx.c          | 3 ++-
>> >  arch/x86/kvm/x86.c              | 3 +++
>> >  4 files changed, 8 insertions(+), 2 deletions(-)
>> > 
>> > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
>> > index 88fce6ab4bbd7..8f6e15b95a4d8 100644
>> > --- a/arch/x86/include/asm/kvm_host.h
>> > +++ b/arch/x86/include/asm/kvm_host.h
>> > @@ -1034,6 +1034,7 @@ struct kvm_x86_msr_filter {
>> >  #define APICV_INHIBIT_REASON_IRQWIN     3
>> >  #define APICV_INHIBIT_REASON_PIT_REINJ  4
>> >  #define APICV_INHIBIT_REASON_X2APIC	5
>> > +#define APICV_INHIBIT_REASON_BLOCKIRQ	6
>> >  
>> >  struct kvm_arch {
>> >  	unsigned long n_used_mmu_pages;
>> > diff --git a/arch/x86/kvm/svm/avic.c b/arch/x86/kvm/svm/avic.c
>> > index 8052d92069e01..affc0ea98d302 100644
>> > --- a/arch/x86/kvm/svm/avic.c
>> > +++ b/arch/x86/kvm/svm/avic.c
>> > @@ -904,7 +904,8 @@ bool svm_check_apicv_inhibit_reasons(ulong bit)
>> >  			  BIT(APICV_INHIBIT_REASON_NESTED) |
>> >  			  BIT(APICV_INHIBIT_REASON_IRQWIN) |
>> >  			  BIT(APICV_INHIBIT_REASON_PIT_REINJ) |
>> > -			  BIT(APICV_INHIBIT_REASON_X2APIC);
>> > +			  BIT(APICV_INHIBIT_REASON_X2APIC) |
>> > +			  BIT(APICV_INHIBIT_REASON_BLOCKIRQ);
>> >  
>> >  	return supported & BIT(bit);
>> >  }
>> > diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
>> > index 71f54d85f104c..e4fc9ff7cd944 100644
>> > --- a/arch/x86/kvm/vmx/vmx.c
>> > +++ b/arch/x86/kvm/vmx/vmx.c
>> > @@ -7565,7 +7565,8 @@ static void hardware_unsetup(void)
>> >  static bool vmx_check_apicv_inhibit_reasons(ulong bit)
>> >  {
>> >  	ulong supported = BIT(APICV_INHIBIT_REASON_DISABLE) |
>> > -			  BIT(APICV_INHIBIT_REASON_HYPERV);
>> > +			  BIT(APICV_INHIBIT_REASON_HYPERV) |
>> > +			  BIT(APICV_INHIBIT_REASON_BLOCKIRQ);
>> >  
>> >  	return supported & BIT(bit);
>> >  }
>> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> > index ac83d873d65b0..dccf927baa4dd 100644
>> > --- a/arch/x86/kvm/x86.c
>> > +++ b/arch/x86/kvm/x86.c
>> > @@ -10747,6 +10747,9 @@ int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
>> >  	if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
>> >  		vcpu->arch.singlestep_rip = kvm_get_linear_rip(vcpu);
>> >  
>> > +	kvm_request_apicv_update(vcpu->kvm,
>> > +				 !(vcpu->guest_debug & KVM_GUESTDBG_BLOCKIRQ),
>> > +				 APICV_INHIBIT_REASON_BLOCKIRQ);
>> >  	/*
>> >  	 * Trigger an rflags update that will inject or remove the trace
>> >  	 * flags.
>> 
>> This fixes the problem for me!
>> 
>> Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
>
> Cool!
>> 
>
> Now that I think about it, since guest debug flags are per-vcpu, this code won't
> work if there are multiple vCPUs and you enable KVM_GUESTDBG_BLOCKIRQ on all of them
> and then disable on this flag on just one of vCPUs, because this code will re-enable APICv/AVIC in this case.
> A counter is needed, like you did in synic/autoeoi case.
>

Right, I completely forgot about this peculiarity!

-- 
Vitaly