From: Wanpeng Li <wanpengli@tencent.com> SDM mentioned that, RDPMC: IF (((CR4.PCE = 1) or (CPL = 0) or (CR0.PE = 0)) and (ECX indicates a supported counter)) THEN EAX := counter[31:0]; EDX := ZeroExtend(counter[MSCB:32]); ELSE (* ECX is not valid or CR4.PCE is 0 and CPL is 1, 2, or 3 and CR0.PE is 1 *) #GP(0); FI; Let's add the CR0.PE is 1 checking to rdpmc emulate. Signed-off-by: Wanpeng Li <wanpengli@tencent.com> --- arch/x86/kvm/emulate.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c index 9a144ca8e146..ab7ec569e8c9 100644 --- a/arch/x86/kvm/emulate.c +++ b/arch/x86/kvm/emulate.c @@ -4213,6 +4213,7 @@ static int check_rdtsc(struct x86_emulate_ctxt *ctxt) static int check_rdpmc(struct x86_emulate_ctxt *ctxt) { u64 cr4 = ctxt->ops->get_cr(ctxt, 4); + u64 cr0 = ctxt->ops->get_cr(ctxt, 0); u64 rcx = reg_read(ctxt, VCPU_REGS_RCX); /* @@ -4222,7 +4223,7 @@ static int check_rdpmc(struct x86_emulate_ctxt *ctxt) if (enable_vmware_backdoor && is_vmware_backdoor_pmc(rcx)) return X86EMUL_CONTINUE; - if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) || + if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt) && (cr0 & X86_CR0_PE)) || ctxt->ops->check_pmc(ctxt, rcx)) return emulate_gp(ctxt, 0); -- 2.25.1
From: Wanpeng Li <wanpengli@tencent.com> SDM section 18.2.3 mentioned that: "IA32_PERF_GLOBAL_OVF_CTL MSR allows software to clear overflow indicator(s) of any general-purpose or fixed-function counters via a single WRMSR." It is R/W mentioned by SDM, we read this msr on bare-metal during perf testing, the value is always 0 for CLX/SKX boxes on hands. Let's fill get_msr MSR_CORE_PERF_GLOBAL_OVF_CTRL w/ 0 as hardware behavior. Signed-off-by: Wanpeng Li <wanpengli@tencent.com> --- Btw, xen also fills get_msr MSR_CORE_PERF_GLOBAL_OVF_CTRL 0. arch/x86/kvm/vmx/pmu_intel.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c index 10cc4f65c4ef..47260a8563f9 100644 --- a/arch/x86/kvm/vmx/pmu_intel.c +++ b/arch/x86/kvm/vmx/pmu_intel.c @@ -365,7 +365,7 @@ static int intel_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info) msr_info->data = pmu->global_ctrl; return 0; case MSR_CORE_PERF_GLOBAL_OVF_CTRL: - msr_info->data = pmu->global_ovf_ctrl; + msr_info->data = 0; return 0; default: if ((pmc = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0)) || -- 2.25.1
From: Wanpeng Li <wanpengli@tencent.com> The overhead of kvm_vcpu_kick() is huge since expensive rcu/memory barrier etc operations in rcuwait_wake_up(). It is worse when local delivery since the vCPU is scheduled and we still suffer from this. We can observe 12us+ for kvm_vcpu_kick() in kvm_pmu_deliver_pmi() path by ftrace before the patch and 6us+ after the optimization. Signed-off-by: Wanpeng Li <wanpengli@tencent.com> --- arch/x86/kvm/lapic.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c index 76fb00921203..ec6997187c6d 100644 --- a/arch/x86/kvm/lapic.c +++ b/arch/x86/kvm/lapic.c @@ -1120,7 +1120,8 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode, case APIC_DM_NMI: result = 1; kvm_inject_nmi(vcpu); - kvm_vcpu_kick(vcpu); + if (vcpu != kvm_get_running_vcpu()) + kvm_vcpu_kick(vcpu); break; case APIC_DM_INIT: -- 2.25.1
Wanpeng Li <kernellwp@gmail.com> writes: > From: Wanpeng Li <wanpengli@tencent.com> > > The overhead of kvm_vcpu_kick() is huge since expensive rcu/memory > barrier etc operations in rcuwait_wake_up(). It is worse when local > delivery since the vCPU is scheduled and we still suffer from this. > We can observe 12us+ for kvm_vcpu_kick() in kvm_pmu_deliver_pmi() > path by ftrace before the patch and 6us+ after the optimization. > > Signed-off-by: Wanpeng Li <wanpengli@tencent.com> > --- > arch/x86/kvm/lapic.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c > index 76fb00921203..ec6997187c6d 100644 > --- a/arch/x86/kvm/lapic.c > +++ b/arch/x86/kvm/lapic.c > @@ -1120,7 +1120,8 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode, > case APIC_DM_NMI: > result = 1; > kvm_inject_nmi(vcpu); > - kvm_vcpu_kick(vcpu); > + if (vcpu != kvm_get_running_vcpu()) > + kvm_vcpu_kick(vcpu); Out of curiosity, can this be converted into a generic optimization for kvm_vcpu_kick() instead? I.e. if kvm_vcpu_kick() is called for the currently running vCPU, there's almost nothing to do, especially when we already have a request pending, right? (I didn't put too much though to it) > break; > > case APIC_DM_INIT: -- Vitaly
cc Andi, On 8/10/2021 5:57 pm, Wanpeng Li wrote: > From: Wanpeng Li <wanpengli@tencent.com> > > SDM section 18.2.3 mentioned that: > > "IA32_PERF_GLOBAL_OVF_CTL MSR allows software to clear overflow indicator(s) of > any general-purpose or fixed-function counters via a single WRMSR." > > It is R/W mentioned by SDM, we read this msr on bare-metal during perf testing, > the value is always 0 for CLX/SKX boxes on hands. Let's fill get_msr > MSR_CORE_PERF_GLOBAL_OVF_CTRL w/ 0 as hardware behavior. > > Signed-off-by: Wanpeng Li <wanpengli@tencent.com> > --- > Btw, xen also fills get_msr MSR_CORE_PERF_GLOBAL_OVF_CTRL 0. > > arch/x86/kvm/vmx/pmu_intel.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c > index 10cc4f65c4ef..47260a8563f9 100644 > --- a/arch/x86/kvm/vmx/pmu_intel.c > +++ b/arch/x86/kvm/vmx/pmu_intel.c > @@ -365,7 +365,7 @@ static int intel_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info) > msr_info->data = pmu->global_ctrl; > return 0; > case MSR_CORE_PERF_GLOBAL_OVF_CTRL: > - msr_info->data = pmu->global_ovf_ctrl; > + msr_info->data = 0; Tested-by: Like Xu <likexu@tencent.com> Further, better to drop 'u64 global_ovf_ctrl' directly. > return 0; > default: > if ((pmc = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0)) || >
On Fri, 8 Oct 2021 at 18:52, Vitaly Kuznetsov <vkuznets@redhat.com> wrote:
>
> Wanpeng Li <kernellwp@gmail.com> writes:
>
> > From: Wanpeng Li <wanpengli@tencent.com>
> >
> > The overhead of kvm_vcpu_kick() is huge since expensive rcu/memory
> > barrier etc operations in rcuwait_wake_up(). It is worse when local
> > delivery since the vCPU is scheduled and we still suffer from this.
> > We can observe 12us+ for kvm_vcpu_kick() in kvm_pmu_deliver_pmi()
> > path by ftrace before the patch and 6us+ after the optimization.
> >
> > Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
> > ---
> > arch/x86/kvm/lapic.c | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> > index 76fb00921203..ec6997187c6d 100644
> > --- a/arch/x86/kvm/lapic.c
> > +++ b/arch/x86/kvm/lapic.c
> > @@ -1120,7 +1120,8 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
> > case APIC_DM_NMI:
> > result = 1;
> > kvm_inject_nmi(vcpu);
> > - kvm_vcpu_kick(vcpu);
> > + if (vcpu != kvm_get_running_vcpu())
> > + kvm_vcpu_kick(vcpu);
>
> Out of curiosity,
>
> can this be converted into a generic optimization for kvm_vcpu_kick()
> instead? I.e. if kvm_vcpu_kick() is called for the currently running
> vCPU, there's almost nothing to do, especially when we already have a
> request pending, right? (I didn't put too much though to it)
I thought about it before, I will do it in the next version since you
also vote for it. :)
Wanpeng
On Fri, 8 Oct 2021 at 19:02, Like Xu <like.xu.linux@gmail.com> wrote: > > cc Andi, > > On 8/10/2021 5:57 pm, Wanpeng Li wrote: > > From: Wanpeng Li <wanpengli@tencent.com> > > > > SDM section 18.2.3 mentioned that: > > > > "IA32_PERF_GLOBAL_OVF_CTL MSR allows software to clear overflow indicator(s) of > > any general-purpose or fixed-function counters via a single WRMSR." > > > > It is R/W mentioned by SDM, we read this msr on bare-metal during perf testing, > > the value is always 0 for CLX/SKX boxes on hands. Let's fill get_msr > > MSR_CORE_PERF_GLOBAL_OVF_CTRL w/ 0 as hardware behavior. > > > > Signed-off-by: Wanpeng Li <wanpengli@tencent.com> > > --- > > Btw, xen also fills get_msr MSR_CORE_PERF_GLOBAL_OVF_CTRL 0. > > > > arch/x86/kvm/vmx/pmu_intel.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c > > index 10cc4f65c4ef..47260a8563f9 100644 > > --- a/arch/x86/kvm/vmx/pmu_intel.c > > +++ b/arch/x86/kvm/vmx/pmu_intel.c > > @@ -365,7 +365,7 @@ static int intel_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info) > > msr_info->data = pmu->global_ctrl; > > return 0; > > case MSR_CORE_PERF_GLOBAL_OVF_CTRL: > > - msr_info->data = pmu->global_ovf_ctrl; > > + msr_info->data = 0; > > Tested-by: Like Xu <likexu@tencent.com> Thanks. > Further, better to drop 'u64 global_ovf_ctrl' directly. Good suggestion. :) Wanpeng
The shortlog makes it sound like "inject a #GP if CR0.PE=1", i.e. unconditionally inject #GP for RDMPC in protected mode. Maybe "Don't inject #GP when emulating RDMPC if CR0.PE=0"? On Fri, Oct 08, 2021, Wanpeng Li wrote: > From: Wanpeng Li <wanpengli@tencent.com> > > SDM mentioned that, RDPMC: > > IF (((CR4.PCE = 1) or (CPL = 0) or (CR0.PE = 0)) and (ECX indicates a supported counter)) > THEN > EAX := counter[31:0]; > EDX := ZeroExtend(counter[MSCB:32]); > ELSE (* ECX is not valid or CR4.PCE is 0 and CPL is 1, 2, or 3 and CR0.PE is 1 *) > #GP(0); > FI; > > Let's add the CR0.PE is 1 checking to rdpmc emulate. > > Signed-off-by: Wanpeng Li <wanpengli@tencent.com> > --- > arch/x86/kvm/emulate.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c > index 9a144ca8e146..ab7ec569e8c9 100644 > --- a/arch/x86/kvm/emulate.c > +++ b/arch/x86/kvm/emulate.c > @@ -4213,6 +4213,7 @@ static int check_rdtsc(struct x86_emulate_ctxt *ctxt) > static int check_rdpmc(struct x86_emulate_ctxt *ctxt) > { > u64 cr4 = ctxt->ops->get_cr(ctxt, 4); > + u64 cr0 = ctxt->ops->get_cr(ctxt, 0); > u64 rcx = reg_read(ctxt, VCPU_REGS_RCX); > > /* > @@ -4222,7 +4223,7 @@ static int check_rdpmc(struct x86_emulate_ctxt *ctxt) > if (enable_vmware_backdoor && is_vmware_backdoor_pmc(rcx)) > return X86EMUL_CONTINUE; > > - if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) || > + if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt) && (cr0 & X86_CR0_PE)) || I don't think it's possible for CPL to be >0 if CR0.PE=0, e.g. we could probably WARN in the #GP path. Realistically it doesn't add value though, so maybe just add a blurb in the changelog saying this isn't strictly necessary? > ctxt->ops->check_pmc(ctxt, rcx)) > return emulate_gp(ctxt, 0); > > -- > 2.25.1 >
On Fri, Oct 08, 2021, Wanpeng Li wrote: > On Fri, 8 Oct 2021 at 18:52, Vitaly Kuznetsov <vkuznets@redhat.com> wrote: > > > > Wanpeng Li <kernellwp@gmail.com> writes: > > > > > From: Wanpeng Li <wanpengli@tencent.com> > > > > > > The overhead of kvm_vcpu_kick() is huge since expensive rcu/memory > > > barrier etc operations in rcuwait_wake_up(). It is worse when local Memory barriers on x86 are just compiler barriers. The only meaningful overhead is the locked transaction in rcu_read_lock() => preempt_disable(). I suspect the performance benefit from this patch comes either comes from avoiding a second lock when disabling preemption again for get_cpu(), or by avoiding the cmpxchg() in kvm_vcpu_exiting_guest_mode(). > > > delivery since the vCPU is scheduled and we still suffer from this. > > > We can observe 12us+ for kvm_vcpu_kick() in kvm_pmu_deliver_pmi() > > > path by ftrace before the patch and 6us+ after the optimization. Those numbers seem off, I wouldn't expect a few locks to take 6us. > > > Signed-off-by: Wanpeng Li <wanpengli@tencent.com> > > > --- > > > arch/x86/kvm/lapic.c | 3 ++- > > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > > > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c > > > index 76fb00921203..ec6997187c6d 100644 > > > --- a/arch/x86/kvm/lapic.c > > > +++ b/arch/x86/kvm/lapic.c > > > @@ -1120,7 +1120,8 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode, > > > case APIC_DM_NMI: > > > result = 1; > > > kvm_inject_nmi(vcpu); > > > - kvm_vcpu_kick(vcpu); > > > + if (vcpu != kvm_get_running_vcpu()) > > > + kvm_vcpu_kick(vcpu); > > > > Out of curiosity, > > > > can this be converted into a generic optimization for kvm_vcpu_kick() > > instead? I.e. if kvm_vcpu_kick() is called for the currently running > > vCPU, there's almost nothing to do, especially when we already have a > > request pending, right? (I didn't put too much though to it) > > I thought about it before, I will do it in the next version since you > also vote for it. :) Adding a kvm_get_running_vcpu() check before kvm_vcpu_wake_up() in kvm_vcpu_kick() is not functionally correct as it's possible to reach kvm_cpu_kick() from (soft) IRQ context, e.g. hrtimer => apic_timer_expired() and pi_wakeup_handler(). If the kick occurs after prepare_to_rcuwait() and the final kvm_vcpu_check_block(), but before the vCPU is scheduled out, then the kvm_vcpu_wake_up() is required to wake the vCPU, even if it is the current running vCPU. The extra check might also degrade performance for many cases since the full kick path would need to disable preemption three times, though if the overhead is from x86's cmpxchg() then it's a moot point. I think we'd want something like this to avoid extra preempt_disable() as well as the cmpxchg() when @vcpu is the running vCPU. diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c index 8b7dc6e89fd7..f148a7d2a8b9 100644 --- a/virt/kvm/kvm_main.c +++ b/virt/kvm/kvm_main.c @@ -3349,8 +3349,15 @@ void kvm_vcpu_kick(struct kvm_vcpu *vcpu) { int me, cpu; - if (kvm_vcpu_wake_up(vcpu)) - return; + me = get_cpu(); + + if (rcuwait_active(&vcpu->wait) && kvm_vcpu_wake_up(vcpu)) + goto out; + + if (vcpu == __this_cpu_read(kvm_running_vcpu)) { + WARN_ON_ONCE(vcpu->mode == IN_GUEST_MODE); + goto out; + } /* * Note, the vCPU could get migrated to a different pCPU at any point @@ -3359,12 +3366,12 @@ void kvm_vcpu_kick(struct kvm_vcpu *vcpu) * IPI is to force the vCPU to leave IN_GUEST_MODE, and migrating the * vCPU also requires it to leave IN_GUEST_MODE. */ - me = get_cpu(); if (kvm_arch_vcpu_should_kick(vcpu)) { cpu = READ_ONCE(vcpu->cpu); if (cpu != me && (unsigned)cpu < nr_cpu_ids && cpu_online(cpu)) smp_send_reschedule(cpu); } +out: put_cpu(); } EXPORT_SYMBOL_GPL(kvm_vcpu_kick);
On Fri, 8 Oct 2021 at 23:20, Sean Christopherson <seanjc@google.com> wrote: > > The shortlog makes it sound like "inject a #GP if CR0.PE=1", i.e. unconditionally > inject #GP for RDMPC in protected mode. Maybe "Don't inject #GP when emulating > RDMPC if CR0.PE=0"? > Agreed. > On Fri, Oct 08, 2021, Wanpeng Li wrote: > > From: Wanpeng Li <wanpengli@tencent.com> > > > > SDM mentioned that, RDPMC: > > > > IF (((CR4.PCE = 1) or (CPL = 0) or (CR0.PE = 0)) and (ECX indicates a supported counter)) > > THEN > > EAX := counter[31:0]; > > EDX := ZeroExtend(counter[MSCB:32]); > > ELSE (* ECX is not valid or CR4.PCE is 0 and CPL is 1, 2, or 3 and CR0.PE is 1 *) > > #GP(0); > > FI; > > > > Let's add the CR0.PE is 1 checking to rdpmc emulate. > > > > Signed-off-by: Wanpeng Li <wanpengli@tencent.com> > > --- > > arch/x86/kvm/emulate.c | 3 ++- > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c > > index 9a144ca8e146..ab7ec569e8c9 100644 > > --- a/arch/x86/kvm/emulate.c > > +++ b/arch/x86/kvm/emulate.c > > @@ -4213,6 +4213,7 @@ static int check_rdtsc(struct x86_emulate_ctxt *ctxt) > > static int check_rdpmc(struct x86_emulate_ctxt *ctxt) > > { > > u64 cr4 = ctxt->ops->get_cr(ctxt, 4); > > + u64 cr0 = ctxt->ops->get_cr(ctxt, 0); > > u64 rcx = reg_read(ctxt, VCPU_REGS_RCX); > > > > /* > > @@ -4222,7 +4223,7 @@ static int check_rdpmc(struct x86_emulate_ctxt *ctxt) > > if (enable_vmware_backdoor && is_vmware_backdoor_pmc(rcx)) > > return X86EMUL_CONTINUE; > > > > - if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) || > > + if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt) && (cr0 & X86_CR0_PE)) || > > I don't think it's possible for CPL to be >0 if CR0.PE=0, e.g. we could probably > WARN in the #GP path. Realistically it doesn't add value though, so maybe just > add a blurb in the changelog saying this isn't strictly necessary? Do it in v2. Wanpeng
On Fri, 8 Oct 2021 at 23:59, Sean Christopherson <seanjc@google.com> wrote: > > On Fri, Oct 08, 2021, Wanpeng Li wrote: > > On Fri, 8 Oct 2021 at 18:52, Vitaly Kuznetsov <vkuznets@redhat.com> wrote: > > > > > > Wanpeng Li <kernellwp@gmail.com> writes: > > > > > > > From: Wanpeng Li <wanpengli@tencent.com> > > > > > > > > The overhead of kvm_vcpu_kick() is huge since expensive rcu/memory > > > > barrier etc operations in rcuwait_wake_up(). It is worse when local > > Memory barriers on x86 are just compiler barriers. The only meaningful overhead > is the locked transaction in rcu_read_lock() => preempt_disable(). I suspect the > performance benefit from this patch comes either comes from avoiding a second > lock when disabling preemption again for get_cpu(), or by avoiding the cmpxchg() > in kvm_vcpu_exiting_guest_mode(). > > > > > delivery since the vCPU is scheduled and we still suffer from this. > > > > We can observe 12us+ for kvm_vcpu_kick() in kvm_pmu_deliver_pmi() > > > > path by ftrace before the patch and 6us+ after the optimization. > > Those numbers seem off, I wouldn't expect a few locks to take 6us. Maybe the ftrace introduces more overhead. > > > > > Signed-off-by: Wanpeng Li <wanpengli@tencent.com> > > > > --- > > > > arch/x86/kvm/lapic.c | 3 ++- > > > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > > > > > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c > > > > index 76fb00921203..ec6997187c6d 100644 > > > > --- a/arch/x86/kvm/lapic.c > > > > +++ b/arch/x86/kvm/lapic.c > > > > @@ -1120,7 +1120,8 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode, > > > > case APIC_DM_NMI: > > > > result = 1; > > > > kvm_inject_nmi(vcpu); > > > > - kvm_vcpu_kick(vcpu); > > > > + if (vcpu != kvm_get_running_vcpu()) > > > > + kvm_vcpu_kick(vcpu); > > > > > > Out of curiosity, > > > > > > can this be converted into a generic optimization for kvm_vcpu_kick() > > > instead? I.e. if kvm_vcpu_kick() is called for the currently running > > > vCPU, there's almost nothing to do, especially when we already have a > > > request pending, right? (I didn't put too much though to it) > > > > I thought about it before, I will do it in the next version since you > > also vote for it. :) > > Adding a kvm_get_running_vcpu() check before kvm_vcpu_wake_up() in kvm_vcpu_kick() > is not functionally correct as it's possible to reach kvm_cpu_kick() from (soft) > IRQ context, e.g. hrtimer => apic_timer_expired() and pi_wakeup_handler(). If > the kick occurs after prepare_to_rcuwait() and the final kvm_vcpu_check_block(), > but before the vCPU is scheduled out, then the kvm_vcpu_wake_up() is required to > wake the vCPU, even if it is the current running vCPU. Good point. > > The extra check might also degrade performance for many cases since the full kick > path would need to disable preemption three times, though if the overhead is from > x86's cmpxchg() then it's a moot point. > > I think we'd want something like this to avoid extra preempt_disable() as well > as the cmpxchg() when @vcpu is the running vCPU. Do it in v2, thanks for the suggestion. Wanpeng