Follow x86's primary MMU, which hasn't flushed TLBs when clearing Accessed bits for 10+ years, and skip all TLB flushes when aging SPTEs in response to a clear_flush_young() mmu_notifier event. As documented in x86's ptep_clear_flush_young(), the probability and impact of "bad" reclaim due to stale A-bit information is relatively low, whereas the performance cost of TLB flushes is relatively high. I.e. the cost of flushing TLBs outweighs the benefits. On KVM x86, the cost of TLB flushes is even higher, as KVM doesn't batch TLB flushes for mmu_notifier events (KVM's mmu_notifier contract with MM makes it all but impossible), and sending IPIs forces all running vCPUs to go through a VM-Exit => VM-Enter roundtrip. Furthermore, MGLRU aging of secondary MMUs is expected to use flush-less mmu_notifiers, i.e. flushing for the !MGLRU will make even less sense, and will be actively confusing as it wouldn't be clear why KVM "needs" to flush TLBs for legacy LRU aging, but not for MGLRU aging. Cc: James Houghton <jthoughton@xxxxxxxxxx> Cc: Yan Zhao <yan.y.zhao@xxxxxxxxx> Link: https://lore.kernel.org/all/20240926013506.860253-18-jthoughton@xxxxxxxxxx Signed-off-by: Sean Christopherson <seanjc@xxxxxxxxxx> --- arch/x86/kvm/Kconfig | 1 + arch/x86/kvm/mmu/spte.h | 5 ++--- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/arch/x86/kvm/Kconfig b/arch/x86/kvm/Kconfig index f09f13c01c6b..1ed1e4f5d51c 100644 --- a/arch/x86/kvm/Kconfig +++ b/arch/x86/kvm/Kconfig @@ -22,6 +22,7 @@ config KVM_X86 depends on X86_LOCAL_APIC select KVM_COMMON select KVM_GENERIC_MMU_NOTIFIER + select KVM_ELIDE_TLB_FLUSH_IF_YOUNG select HAVE_KVM_IRQCHIP select HAVE_KVM_PFNCACHE select HAVE_KVM_DIRTY_RING_TSO diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h index e90cc401c168..8b09a0d60ea6 100644 --- a/arch/x86/kvm/mmu/spte.h +++ b/arch/x86/kvm/mmu/spte.h @@ -470,9 +470,8 @@ static inline bool is_mmu_writable_spte(u64 spte) * needs to flush at the time the SPTEs is modified, before dropping mmu_lock. * * Don't flush if the Accessed bit is cleared, as access tracking tolerates - * false negatives, and the one path that does care about TLB flushes, - * kvm_mmu_notifier_clear_flush_young(), flushes if a young SPTE is found, i.e. - * doesn't rely on lower helpers to detect the need to flush. + * false negatives, e.g. KVM x86 omits TLB flushes even when aging SPTEs for a + * mmu_notifier.clear_flush_young() event. * * Lastly, don't flush if the Dirty bit is cleared, as KVM unconditionally * flushes when enabling dirty logging (see kvm_mmu_slot_apply_flags()), and -- 2.47.0.rc1.288.g06298d1525-goog
