On Tue, Sep 27, 2016 at 11:20:12PM +0200, Paolo Bonzini wrote:
> Since bf9f6ac8d749 ("KVM: Update Posted-Interrupts Descriptor when vCPU
> is blocked", 2015-09-18) the posted interrupt descriptor is checked
> unconditionally for PIR.ON. Therefore we don't need KVM_REQ_EVENT to
> trigger the scan and, if NMIs or SMIs are not involved, we can avoid
> the complicated event injection path.
>
> However, there is a race between vmx_deliver_posted_interrupt and
> vcpu_enter_guest. Fix it by disabling interrupts before vcpu->mode is
> set to IN_GUEST_MODE.
Could you describe the race a bit more please?
I'm surprised that locally disabling irqs
fixes a race with a different CPUs.
>
> Calling kvm_vcpu_kick if PIR.ON=1 is also useless, though it has been
> there since APICv was introduced.
>
> Signed-off-by: Paolo Bonzini <pbonz...@redhat.com>
> ---
> arch/x86/kvm/lapic.c | 2 --
> arch/x86/kvm/vmx.c | 8 +++++---
> arch/x86/kvm/x86.c | 9 +++++++--
> 3 files changed, 12 insertions(+), 7 deletions(-)
>
> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> index 63a442aefc12..be8b7ad56dd1 100644
> --- a/arch/x86/kvm/lapic.c
> +++ b/arch/x86/kvm/lapic.c
> @@ -356,8 +356,6 @@ void kvm_apic_update_irr(struct kvm_vcpu *vcpu, u32 *pir)
> struct kvm_lapic *apic = vcpu->arch.apic;
>
> __kvm_apic_update_irr(pir, apic->regs);
> -
> - kvm_make_request(KVM_REQ_EVENT, vcpu);
> }
> EXPORT_SYMBOL_GPL(kvm_apic_update_irr);
>
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index b33eee395b00..207b9aa32915 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -4844,9 +4844,11 @@ static void vmx_deliver_posted_interrupt(struct
> kvm_vcpu *vcpu, int vector)
> if (pi_test_and_set_pir(vector, &vmx->pi_desc))
> return;
>
> - r = pi_test_and_set_on(&vmx->pi_desc);
> - kvm_make_request(KVM_REQ_EVENT, vcpu);
> - if (r || !kvm_vcpu_trigger_posted_interrupt(vcpu))
so the logic here was reversed - kick was not called
when PIR=OFF previously. It's a bug but what was it's effect?
I'm guessing - extra latency where VCPU won't get
woken up correctly?
> + /* If a previous notification has sent the IPI, nothing to do. */
> + if (pi_test_and_set_on(&vmx->pi_desc))
> + return;
> +
> + if (!kvm_vcpu_trigger_posted_interrupt(vcpu))
> kvm_vcpu_kick(vcpu);
> }
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 3ee8a91a78c3..604cfbfc5bee 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -6658,6 +6658,13 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
> kvm_x86_ops->prepare_guest_switch(vcpu);
> if (vcpu->fpu_active)
> kvm_load_guest_fpu(vcpu);
> +
> + /*
> + * Disable IRQs before setting IN_GUEST_MODE, so that
> + * posted interrupts with vcpu->mode == IN_GUEST_MODE
> + * always result in virtual interrupt delivery.
> + */
> + local_irq_disable();
> vcpu->mode = IN_GUEST_MODE;
>
> srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
> @@ -6671,8 +6678,6 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
> */
> smp_mb__after_srcu_read_unlock();
>
> - local_irq_disable();
> -
> if (vcpu->mode == EXITING_GUEST_MODE || vcpu->requests
> || need_resched() || signal_pending(current)) {
> vcpu->mode = OUTSIDE_GUEST_MODE;
> --
> 1.8.3.1
>
