I'm looking at reducing the interrupt overhead for virtualized guests: some workloads spend a large part of their time processing interrupts. On kvm, an EOI write from the guest causes an expensive exit to host; we avoid this using shared memory. The patches work fine on my boxes. See individual patches for perf tests. You need to patch qemu to whitelist the kvm feature. qemu patch was sent separately. The patches are against Linus's master and apply to kvm.git cleanly. The last patch in the series, supplying the host part, also depends on the ISR optimization patch that I have for convenience included in the series (patch 2), I also included a documentation patch (patch 1) - it is here since it clarifies patch 2. This revision does not yet address Thomas's idea of reworking the APIC page handling. Changes to this optimization would require reworking this last patch in the series. The rest of the patchset has not changed significantly since v2. Thanks, MST changes from v8: Document the use of reboot notifier: it is used to make sure kexec works correctly changes from v7: Address Avi's comments: make bit 1 in the new MSR reserved, require an aligned address remove unused macros from header replace __align by BUILD_BUG_ON and a comment Whitespace fix changes from v6: Address Marcelo's comments: marcelo's comments for isr optimization: isr_cache -> highest_* marcelo's comments for host side eoi: rename isr_cache->highest_* kvm eoi msr doc: fix typo pointed out by marcelo isr optimization: add comment to address marcelo's request kvm: don't make lapic attention check unlikely. at marcelo's request eoi host side: remove unlikely annotations Changes from v5: Clear PV EOI when we cancel interrupts. Pointed out by Gleb. Always set ISR cache when we inject an interrupt. Suggested by Ronen Hod. Changes from v4: Turn off PV EOI on each exit. Turn it back on when safe. Suggested by Avi. Address bug with nested interrupts pointed out by Marcelo. Changes from v3: Address review comments by Marcelo: Multiple cosmetic changes eoi -> pv_eoi Added multiple comments Changes from v2: Kill guest with GP on an illegal MSR value Add documentation Changes from v1: Add host side patch to series Remove kvm-specific __test_and_clear_bit, document that x86 one does what we want already Clear msr on cpu unplug Michael S. Tsirkin (8): kvm: document lapic regs field kvm: optimize ISR lookups kvm_para: guest side for eoi avoidance x86/bitops: note on __test_and_clear_bit atomicity kvm: eoi msi documentation kvm: only sync when attention bits set kvm: rearrange injection cancelling code kvm: host side for eoi optimization Documentation/virtual/kvm/msr.txt | 34 +++++++ arch/x86/include/asm/bitops.h | 7 ++ arch/x86/include/asm/kvm_host.h | 12 +++ arch/x86/include/asm/kvm_para.h | 7 ++ arch/x86/kernel/kvm.c | 57 ++++++++++- arch/x86/kvm/cpuid.c | 1 + arch/x86/kvm/lapic.c | 194 +++++++++++++++++++++++++++++++++++- arch/x86/kvm/lapic.h | 11 ++ arch/x86/kvm/trace.h | 34 +++++++ arch/x86/kvm/x86.c | 20 +++- 10 files changed, 363 insertions(+), 14 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
