On 25/05/2017 08:56, Zhuangyanying wrote: > From: ZhuangYanying <ann.zhuangyanying@xxxxxxxxxx> > > When spin_lock_irqsave() deadlock occurs inside the guest, vcpu threads, > other than the lock-holding one, would enter into S state because of > pvspinlock. Then inject NMI via libvirt API "inject-nmi", the NMI could > not be injected into vm. > > The reason is: > 1 It sets nmi_queued to 1 when calling ioctl KVM_NMI in qemu, and sets > cpu->kvm_vcpu_dirty to true in do_inject_external_nmi() meanwhile. > 2 It sets nmi_queued to 0 in process_nmi(), before entering guest, because > cpu->kvm_vcpu_dirty is true. > > It's not enough just to check nmi_queued to decide whether to stay in > vcpu_block() or not. NMI should be injected immediately at any situation. > Add checking KVM_REQ_NMI request plus with nmi_queued in > vm_vcpu_has_events(). > > Signed-off-by: Zhuang Yanying <ann.zhuangyanying@xxxxxxxxxx> > --- > arch/x86/kvm/x86.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > index 02363e3..2d15708 100644 > --- a/arch/x86/kvm/x86.c > +++ b/arch/x86/kvm/x86.c > @@ -8394,7 +8394,9 @@ static inline bool kvm_vcpu_has_events(struct kvm_vcpu *vcpu) > if (vcpu->arch.pv.pv_unhalted) > return true; > > - if (atomic_read(&vcpu->arch.nmi_queued)) > + if ((kvm_test_request(KVM_REQ_NMI, vcpu) || > + atomic_read(&vcpu->arch.nmi_queued)) && > + kvm_x86_ops->nmi_allowed(vcpu)) It's the other way round; nmi_pending was correct in your v1. Testing KVM_REQ_NMI replaces nmi_queued, while the bugfix tests "vcpu->arch.nmi_pending && kvm_x86_ops->nmi_allowed(vcpu)". The new est has to match the existing test in inject_pending_event. In fact, we also need to do the same change for SMIs, i.e. kvm_test_request(KVM_REQ_SMI, vcpu) || (vcpu->arch.smi_pending && !is_smm(vcpu)) Thanks, Paolo > return true; > > if (kvm_test_request(KVM_REQ_SMI, vcpu)) >
