From: Nicholas Piggin Date: Tue, 23 Nov 2021 09:51:48 +0000 (+1000) Subject: KVM: PPC: Book3S HV: Don't always save PMU for guest capable of nesting X-Git-Url: http://git.osdn.net/view?a=commitdiff_plain;h=d3c8a2d3740d93778ea102d4c781746d284177bf;p=uclinux-h8%2Flinux.git KVM: PPC: Book3S HV: Don't always save PMU for guest capable of nesting Provide a config option that controls the workaround added by commit 63279eeb7f93 ("KVM: PPC: Book3S HV: Always save guest pmu for guest capable of nesting"). The option defaults to y for now, but is expected to go away within a few releases. Nested capable guests running with the earlier commit 178266389794 ("KVM: PPC: Book3S HV Nested: Reflect guest PMU in-use to L0 when guest SPRs are live") will now indicate the PMU in-use status of their guests, which means the parent does not need to unconditionally save the PMU for nested capable guests. After this latest round of performance optimisations, this option costs about 540 cycles or 10% entry/exit performance on a POWER9 nested-capable guest. Signed-off-by: Nicholas Piggin Reviewed-by: Fabiano Rosas References: 178266389794 ("KVM: PPC: Book3S HV Nested: Reflect guest PMU in-use to L0 when guest SPRs are live") Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20211123095231.1036501-11-npiggin@gmail.com --- diff --git a/arch/powerpc/kvm/Kconfig b/arch/powerpc/kvm/Kconfig index ff581d70f20c..6a58532300c5 100644 --- a/arch/powerpc/kvm/Kconfig +++ b/arch/powerpc/kvm/Kconfig @@ -130,6 +130,21 @@ config KVM_BOOK3S_HV_EXIT_TIMING If unsure, say N. +config KVM_BOOK3S_HV_NESTED_PMU_WORKAROUND + bool "Nested L0 host workaround for L1 KVM host PMU handling bug" if EXPERT + depends on KVM_BOOK3S_HV_POSSIBLE + default !EXPERT + help + Old nested HV capable Linux guests have a bug where they don't + reflect the PMU in-use status of their L2 guest to the L0 host + while the L2 PMU registers are live. This can result in loss + of L2 PMU register state, causing perf to not work correctly in + L2 guests. + + Selecting this option for the L0 host implements a workaround for + those buggy L1s which saves the L2 state, at the cost of performance + in all nested-capable guest entry/exit. + config KVM_BOOKE_HV bool diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c index a683ee5f420a..6e760f48bbaf 100644 --- a/arch/powerpc/kvm/book3s_hv.c +++ b/arch/powerpc/kvm/book3s_hv.c @@ -4047,8 +4047,14 @@ static int kvmhv_p9_guest_entry(struct kvm_vcpu *vcpu, u64 time_limit, vcpu->arch.vpa.dirty = 1; save_pmu = lp->pmcregs_in_use; } - /* Must save pmu if this guest is capable of running nested guests */ - save_pmu |= nesting_enabled(vcpu->kvm); + if (IS_ENABLED(CONFIG_KVM_BOOK3S_HV_NESTED_PMU_WORKAROUND)) { + /* + * Save pmu if this guest is capable of running nested guests. + * This is option is for old L1s that do not set their + * lppaca->pmcregs_in_use properly when entering their L2. + */ + save_pmu |= nesting_enabled(vcpu->kvm); + } kvmhv_save_guest_pmu(vcpu, save_pmu); #ifdef CONFIG_PPC_PSERIES