Re: [PATCH v10 07/20] KVM: arm64: PMU: Recreate events after reset

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Deferred PMU event recreation corrupts PMU counter reset values by writing back stale perf_event counts into freshly reset system registers.
--

commit a8710b2a9555c20c3e81b98a0509dda9d3d93f2e
Author: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>

KVM: arm64: PMU: Recreate events after reset

This commit defers the destruction and recreation of PMU events until the
vCPU thread runs, setting a flag during system register reset to stop and
reprogram the backing perf events.

> diff --git a/arch/arm64/kvm/pmu-emul.c b/arch/arm64/kvm/pmu-emul.c
> index a5ccb6ae44d5f..8530cc907b56c 100644
> --- a/arch/arm64/kvm/pmu-emul.c
> +++ b/arch/arm64/kvm/pmu-emul.c
[ ... ]
> @@ -900,7 +906,16 @@ u64 kvm_pmu_get_pmceid(struct kvm_vcpu *vcpu, bool pmceid1)
>  
>  void kvm_vcpu_reload_pmu(struct kvm_vcpu *vcpu)
>  {
> +	struct kvm_pmu *pmu = &vcpu->arch.pmu;
>  	u64 mask = kvm_pmu_implemented_counter_mask(vcpu);
> +	int i;
> +
> +	if (pmu->events_need_recreate) {
> +		for (i = 0; i < KVM_ARMV8_PMU_MAX_COUNTERS; i++)
> +			kvm_pmu_stop_counter(kvm_vcpu_idx_to_pmc(vcpu, i));

[Severity: High]
Does this corrupt the architectural reset state of the PMU counters?

If a guest VM triggers a warm reset, kvm_reset_sys_regs() resets the PMU
counters to their architectural reset state and then calls
kvm_pmu_request_recreate().

On the next vCPU entry, kvm_vcpu_reload_pmu() iterates over all PMCs and calls
kvm_pmu_stop_counter(). However, kvm_pmu_stop_counter() is designed for dynamic
disablement. It preserves the counter by reading the current sysreg value,
adding the accumulated perf_event count, and writing the sum back to the
sysreg:

arch/arm64/kvm/pmu-emul.c:kvm_pmu_stop_counter() {
    ...
    val = kvm_pmu_get_pmc_value(pmc);
    reg = counter_index_to_reg(pmc->idx);
    __vcpu_assign_sys_reg(vcpu, reg, val);
    ...
}

Will this sequence overwrite the freshly reset counter value with the sum of
the reset value and the stale perf_event count from the previous run?

> +
> +		pmu->events_need_recreate = false;
> +	}
>  
>  	__vcpu_rmw_sys_reg(vcpu, PMOVSSET_EL0, &=, mask);
>  	__vcpu_rmw_sys_reg(vcpu, PMINTENSET_EL1, &=, mask);
> diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
> index 42b6705bc21d1..1a57e07cec9a4 100644
> --- a/arch/arm64/kvm/sys_regs.c
> +++ b/arch/arm64/kvm/sys_regs.c
> @@ -5390,7 +5390,7 @@ void kvm_reset_sys_regs(struct kvm_vcpu *vcpu)
>  	set_bit(KVM_ARCH_FLAG_ID_REGS_INITIALIZED, &kvm->arch.flags);
>  
>  	if (kvm_vcpu_has_pmu(vcpu))
> -		kvm_make_request(KVM_REQ_RELOAD_PMU, vcpu);
> +		kvm_pmu_request_recreate(vcpu);
>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260914-hybrid-v10-0-92788053a691@rsg.ci.i.u-tokyo.ac.jp?part=7