Hi Drew,
On Sat, Jul 01, 2017 at 06:26:54PM +0200, Andrew Jones wrote:
> kvm_pmu_overflow_set() is called from perf's interrupt handler,
> making the call of kvm_vgic_inject_irq() from it introduced with
> "KVM: arm/arm64: PMU: remove request-less vcpu kick" a really bad
> idea, as it's quite easy to try and retake a lock that the
> interrupted context is already holding. The fix is to use a vcpu
> kick, leaving the interrupt injection to kvm_pmu_sync_hwstate(),
> like it was doing before the refactoring. We don't just revert,
> though, because before the kick was request-less, leaving the vcpu
> exposed to the request-less vcpu kick race, and also because the
> kick was used unnecessarily from register access handlers.
I have some patches that makes it possible to inject virtual IRQs to the
vgic from interrupt context (my timer improvement series), but it's
probably better to merge this as a fix even though.
>
> Signed-off-by: Andrew Jones <drjones@xxxxxxxxxx>
Reviewed-by: Christoffer Dall <cdall@xxxxxxxxxx>
> ---
> arch/arm64/kvm/sys_regs.c | 2 +-
> include/kvm/arm_pmu.h | 2 --
> virt/kvm/arm/pmu.c | 43 +++++++++++++++----------------------------
> 3 files changed, 16 insertions(+), 31 deletions(-)
>
> diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
> index 77862881ae86..2e070d3baf9f 100644
> --- a/arch/arm64/kvm/sys_regs.c
> +++ b/arch/arm64/kvm/sys_regs.c
> @@ -764,7 +764,7 @@ static bool access_pmovs(struct kvm_vcpu *vcpu, struct sys_reg_params *p,
> if (p->is_write) {
> if (r->CRm & 0x2)
> /* accessing PMOVSSET_EL0 */
> - kvm_pmu_overflow_set(vcpu, p->regval & mask);
> + vcpu_sys_reg(vcpu, PMOVSSET_EL0) |= (p->regval & mask);
> else
> /* accessing PMOVSCLR_EL0 */
> vcpu_sys_reg(vcpu, PMOVSSET_EL0) &= ~(p->regval & mask);
> diff --git a/include/kvm/arm_pmu.h b/include/kvm/arm_pmu.h
> index f6e030617467..f87fe20fcb05 100644
> --- a/include/kvm/arm_pmu.h
> +++ b/include/kvm/arm_pmu.h
> @@ -48,7 +48,6 @@ void kvm_pmu_vcpu_reset(struct kvm_vcpu *vcpu);
> void kvm_pmu_vcpu_destroy(struct kvm_vcpu *vcpu);
> void kvm_pmu_disable_counter(struct kvm_vcpu *vcpu, u64 val);
> void kvm_pmu_enable_counter(struct kvm_vcpu *vcpu, u64 val);
> -void kvm_pmu_overflow_set(struct kvm_vcpu *vcpu, u64 val);
> void kvm_pmu_flush_hwstate(struct kvm_vcpu *vcpu);
> void kvm_pmu_sync_hwstate(struct kvm_vcpu *vcpu);
> bool kvm_pmu_should_notify_user(struct kvm_vcpu *vcpu);
> @@ -86,7 +85,6 @@ static inline void kvm_pmu_vcpu_reset(struct kvm_vcpu *vcpu) {}
> static inline void kvm_pmu_vcpu_destroy(struct kvm_vcpu *vcpu) {}
> static inline void kvm_pmu_disable_counter(struct kvm_vcpu *vcpu, u64 val) {}
> static inline void kvm_pmu_enable_counter(struct kvm_vcpu *vcpu, u64 val) {}
> -static inline void kvm_pmu_overflow_set(struct kvm_vcpu *vcpu, u64 val) {}
> static inline void kvm_pmu_flush_hwstate(struct kvm_vcpu *vcpu) {}
> static inline void kvm_pmu_sync_hwstate(struct kvm_vcpu *vcpu) {}
> static inline bool kvm_pmu_should_notify_user(struct kvm_vcpu *vcpu)
> diff --git a/virt/kvm/arm/pmu.c b/virt/kvm/arm/pmu.c
> index fc8a723ff387..8a9c42366db7 100644
> --- a/virt/kvm/arm/pmu.c
> +++ b/virt/kvm/arm/pmu.c
> @@ -203,11 +203,15 @@ static u64 kvm_pmu_overflow_status(struct kvm_vcpu *vcpu)
> return reg;
> }
>
> -static void kvm_pmu_check_overflow(struct kvm_vcpu *vcpu)
> +static void kvm_pmu_update_state(struct kvm_vcpu *vcpu)
> {
> struct kvm_pmu *pmu = &vcpu->arch.pmu;
> - bool overflow = !!kvm_pmu_overflow_status(vcpu);
> + bool overflow;
> +
> + if (!kvm_arm_pmu_v3_ready(vcpu))
> + return;
>
> + overflow = !!kvm_pmu_overflow_status(vcpu);
> if (pmu->irq_level == overflow)
> return;
>
> @@ -215,33 +219,11 @@ static void kvm_pmu_check_overflow(struct kvm_vcpu *vcpu)
>
> if (likely(irqchip_in_kernel(vcpu->kvm))) {
> int ret = kvm_vgic_inject_irq(vcpu->kvm, vcpu->vcpu_id,
> - pmu->irq_num, overflow,
> - &vcpu->arch.pmu);
> + pmu->irq_num, overflow, pmu);
> WARN_ON(ret);
> }
> }
>
> -/**
> - * kvm_pmu_overflow_set - set PMU overflow interrupt
> - * @vcpu: The vcpu pointer
> - * @val: the value guest writes to PMOVSSET register
> - */
> -void kvm_pmu_overflow_set(struct kvm_vcpu *vcpu, u64 val)
> -{
> - if (val == 0)
> - return;
> -
> - vcpu_sys_reg(vcpu, PMOVSSET_EL0) |= val;
> - kvm_pmu_check_overflow(vcpu);
> -}
> -
> -static void kvm_pmu_update_state(struct kvm_vcpu *vcpu)
> -{
> - if (!kvm_arm_pmu_v3_ready(vcpu))
> - return;
> - kvm_pmu_check_overflow(vcpu);
> -}
> -
> bool kvm_pmu_should_notify_user(struct kvm_vcpu *vcpu)
> {
> struct kvm_pmu *pmu = &vcpu->arch.pmu;
> @@ -303,7 +285,7 @@ static inline struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc)
> }
>
> /**
> - * When perf event overflows, call kvm_pmu_overflow_set to set overflow status.
> + * When the perf event overflows, set the overflow status and inform the vcpu.
> */
> static void kvm_pmu_perf_overflow(struct perf_event *perf_event,
> struct perf_sample_data *data,
> @@ -313,7 +295,12 @@ static void kvm_pmu_perf_overflow(struct perf_event *perf_event,
> struct kvm_vcpu *vcpu = kvm_pmc_to_vcpu(pmc);
> int idx = pmc->idx;
>
> - kvm_pmu_overflow_set(vcpu, BIT(idx));
> + vcpu_sys_reg(vcpu, PMOVSSET_EL0) |= BIT(idx);
> +
> + if (kvm_pmu_overflow_status(vcpu)) {
> + kvm_make_request(KVM_REQ_IRQ_PENDING, vcpu);
> + kvm_vcpu_kick(vcpu);
> + }
> }
>
> /**
> @@ -341,7 +328,7 @@ void kvm_pmu_software_increment(struct kvm_vcpu *vcpu, u64 val)
> reg = lower_32_bits(reg);
> vcpu_sys_reg(vcpu, PMEVCNTR0_EL0 + i) = reg;
> if (!reg)
> - kvm_pmu_overflow_set(vcpu, BIT(i));
> + vcpu_sys_reg(vcpu, PMOVSSET_EL0) |= BIT(i);
> }
> }
> }
> --
> 2.9.4
>
