Currently, with VHE, KVM enables the EL0 event counting for the guest on vcpu_load() or KVM enables it as a part of the PMU register emulation process, when needed. However, in the migration case (with VHE), the same handling is lacking. So, enable it on the first KVM_RUN with VHE (after the migration) when needed.
Fixes: d0c94c49792c ("KVM: arm64: Restore PMU configuration on first run") Cc: stable@vger.kernel.org Signed-off-by: Reiji Watanabe reijiw@google.com --- arch/arm64/kvm/pmu-emul.c | 1 + arch/arm64/kvm/sys_regs.c | 1 - 2 files changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/pmu-emul.c b/arch/arm64/kvm/pmu-emul.c index c243b10f3e15..5eca0cdd961d 100644 --- a/arch/arm64/kvm/pmu-emul.c +++ b/arch/arm64/kvm/pmu-emul.c @@ -558,6 +558,7 @@ void kvm_pmu_handle_pmcr(struct kvm_vcpu *vcpu, u64 val) for_each_set_bit(i, &mask, 32) kvm_pmu_set_pmc_value(kvm_vcpu_idx_to_pmc(vcpu, i), 0, true); } + kvm_vcpu_pmu_restore_guest(vcpu); }
static bool kvm_pmu_counter_is_enabled(struct kvm_pmc *pmc) diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index 1b2c161120be..34688918c811 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -794,7 +794,6 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p, if (!kvm_supports_32bit_el0()) val |= ARMV8_PMU_PMCR_LC; kvm_pmu_handle_pmcr(vcpu, val); - kvm_vcpu_pmu_restore_guest(vcpu); } else { /* PMCR.P & PMCR.C are RAZ */ val = __vcpu_sys_reg(vcpu, PMCR_EL0)
On Tue, 28 Mar 2023 04:47:25 +0100, Reiji Watanabe reijiw@google.com wrote:
Currently, with VHE, KVM enables the EL0 event counting for the guest on vcpu_load() or KVM enables it as a part of the PMU register emulation process, when needed. However, in the migration case (with VHE), the same handling is lacking. So, enable it on the first KVM_RUN with VHE (after the migration) when needed.
It wasn't completely clear to me how the migration case was affected by this until I started digging into the call stack:
At load-time, the PMCR_EL0 effects haven't been propagated yet (the events haven't been created, as this is what kvm_pmu_handle_pmcr() does on first run). So there is an ordering inversion between kvm_pmu_handle_pmcr() and kvm_vcpu_pmu_restore_guest().
Moving the latter call into the former fixes the issue, completely emulating an extra write to PMCR_EL0.
I think it would be worth capturing some of the above in the commit message so that it doesn't get lost...
Fixes: d0c94c49792c ("KVM: arm64: Restore PMU configuration on first run") Cc: stable@vger.kernel.org Signed-off-by: Reiji Watanabe reijiw@google.com
arch/arm64/kvm/pmu-emul.c | 1 + arch/arm64/kvm/sys_regs.c | 1 - 2 files changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/pmu-emul.c b/arch/arm64/kvm/pmu-emul.c index c243b10f3e15..5eca0cdd961d 100644 --- a/arch/arm64/kvm/pmu-emul.c +++ b/arch/arm64/kvm/pmu-emul.c @@ -558,6 +558,7 @@ void kvm_pmu_handle_pmcr(struct kvm_vcpu *vcpu, u64 val) for_each_set_bit(i, &mask, 32) kvm_pmu_set_pmc_value(kvm_vcpu_idx_to_pmc(vcpu, i), 0, true); }
- kvm_vcpu_pmu_restore_guest(vcpu);
} static bool kvm_pmu_counter_is_enabled(struct kvm_pmc *pmc) diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index 1b2c161120be..34688918c811 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -794,7 +794,6 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p, if (!kvm_supports_32bit_el0()) val |= ARMV8_PMU_PMCR_LC; kvm_pmu_handle_pmcr(vcpu, val);
kvm_vcpu_pmu_restore_guest(vcpu);
With the nitpicking above addressed, and should this go into 6.3 as a fix:
Reviewed-by: Marc Zyngier maz@kernel.org
I can otherwise take it into 6.4, depending on what Oliver decides to do.
Thanks,
M.
Hi Marc,
On Tue, Mar 28, 2023 at 12:08:31PM +0100, Marc Zyngier wrote:
On Tue, 28 Mar 2023 04:47:25 +0100, Reiji Watanabe reijiw@google.com wrote:
Currently, with VHE, KVM enables the EL0 event counting for the guest on vcpu_load() or KVM enables it as a part of the PMU register emulation process, when needed. However, in the migration case (with VHE), the same handling is lacking. So, enable it on the first KVM_RUN with VHE (after the migration) when needed.
It wasn't completely clear to me how the migration case was affected by this until I started digging into the call stack:
At load-time, the PMCR_EL0 effects haven't been propagated yet (the events haven't been created, as this is what kvm_pmu_handle_pmcr() does on first run). So there is an ordering inversion between kvm_pmu_handle_pmcr() and kvm_vcpu_pmu_restore_guest().
Moving the latter call into the former fixes the issue, completely emulating an extra write to PMCR_EL0.
I think it would be worth capturing some of the above in the commit message so that it doesn't get lost...
I agree with that. I will add the explanation in the commit message, and will post v2.
Fixes: d0c94c49792c ("KVM: arm64: Restore PMU configuration on first run") Cc: stable@vger.kernel.org Signed-off-by: Reiji Watanabe reijiw@google.com
arch/arm64/kvm/pmu-emul.c | 1 + arch/arm64/kvm/sys_regs.c | 1 - 2 files changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/pmu-emul.c b/arch/arm64/kvm/pmu-emul.c index c243b10f3e15..5eca0cdd961d 100644 --- a/arch/arm64/kvm/pmu-emul.c +++ b/arch/arm64/kvm/pmu-emul.c @@ -558,6 +558,7 @@ void kvm_pmu_handle_pmcr(struct kvm_vcpu *vcpu, u64 val) for_each_set_bit(i, &mask, 32) kvm_pmu_set_pmc_value(kvm_vcpu_idx_to_pmc(vcpu, i), 0, true); }
- kvm_vcpu_pmu_restore_guest(vcpu);
} static bool kvm_pmu_counter_is_enabled(struct kvm_pmc *pmc) diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index 1b2c161120be..34688918c811 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -794,7 +794,6 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p, if (!kvm_supports_32bit_el0()) val |= ARMV8_PMU_PMCR_LC; kvm_pmu_handle_pmcr(vcpu, val);
kvm_vcpu_pmu_restore_guest(vcpu);With the nitpicking above addressed, and should this go into 6.3 as a fix:
Reviewed-by: Marc Zyngier maz@kernel.org
Thank you! Reiji
I can otherwise take it into 6.4, depending on what Oliver decides to do.
Thanks,
M.
-- Without deviation from the norm, progress is not possible.
linux-stable-mirror@lists.linaro.org
-
Marc Zyngier -
Reiji Watanabe