| Message ID | 20190920212509.2578-1-aarcange@redhat.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | KVM monolithic v1 | expand |
On Fri, Sep 20, 2019 at 05:24:52PM -0400, Andrea Arcangeli wrote: > Andrea Arcangeli (17): > x86: spec_ctrl: fix SPEC_CTRL initialization after kexec > KVM: monolithic: x86: convert the kvm_x86_ops methods to external > functions > KVM: monolithic: x86: handle the request_immediate_exit variation > KVM: monolithic: x86: convert the kvm_pmu_ops methods to external > functions > KVM: monolithic: x86: enable the kvm_x86_ops external functions > KVM: monolithic: x86: enable the kvm_pmu_ops external functions > KVM: monolithic: x86: adjust the section prefixes > KVM: monolithic: adjust the section prefixes in the KVM common code > KVM: monolithic: x86: remove kvm.ko IMO, the conversion to a monolithic module should occur immediately, i.e. "KVM: monolithic: x86: remove kvm.ko" should be patch 01/nn. Removing kvm_x86_ops and kvm_pmu_ops isn't a preqrequisite to making KVM a monolothic module, rather they're enhancements that are made possible *because* KVM is a monolithic module. With that ordering, I suspect the convert->enable->use of kvm_x86_ops can be collapsed into a single patch. > KVM: monolithic: x86: use the external functions instead of kvm_x86_ops > KVM: monolithic: x86: remove exports > KVM: monolithic: remove exports from KVM common code > KVM: monolithic: x86: drop the kvm_pmu_ops structure > KVM: monolithic: x86: inline more exit handlers in vmx.c > KVM: retpolines: x86: eliminate retpoline from vmx.c exit handlers > KVM: retpolines: x86: eliminate retpoline from svm.c exit handlers > x86: retpolines: eliminate retpoline from msr event handlers
Hello, This patchset micro optimizes the vmexit to increase performance by dropping the kvm.ko kernel module. All common KVM code gets linked twice into kvm-intel and kvm-amd with the only cons of using more disk space, but the pros of CPU (and RAM) runtime benefits. This improves the vmexit performance by two digits percent on microbenchmarks with the spectre_v2 default mitigation on both VMX and SVM. With spectre_v2=off or with CPUs with IBRS_ALL in ARCH_CAPABILITIES this still improve performance but it's more of the order of 1%. We'll still have to deal with CPUs without IBRS_ALL for a decade and reducing the vmexit latency is important to pass certain benchmarks with workloads that happen to trigger frequent vmexits without having to set spectre_v2=off in the host (which at least in theory would make the host kernel vulnerable from a spectre v2 attack from the guest, even through hyperthreading). The first patch 1/17 should be splitted off from this series and it's intended to be merged separately, it's included here only to avoid any possible erroneous measurement if using kexec for testing, in turn if using kexec it's recommended to include it in the baseline measurements too. A git clonable branch for quick testing is available here: https://git.kernel.org/pub/scm/linux/kernel/git/andrea/aa.git/log/?h=kvm-mono1 Thanks, Andrea Andrea Arcangeli (17): x86: spec_ctrl: fix SPEC_CTRL initialization after kexec KVM: monolithic: x86: convert the kvm_x86_ops methods to external functions KVM: monolithic: x86: handle the request_immediate_exit variation KVM: monolithic: x86: convert the kvm_pmu_ops methods to external functions KVM: monolithic: x86: enable the kvm_x86_ops external functions KVM: monolithic: x86: enable the kvm_pmu_ops external functions KVM: monolithic: x86: adjust the section prefixes KVM: monolithic: adjust the section prefixes in the KVM common code KVM: monolithic: x86: remove kvm.ko KVM: monolithic: x86: use the external functions instead of kvm_x86_ops KVM: monolithic: x86: remove exports KVM: monolithic: remove exports from KVM common code KVM: monolithic: x86: drop the kvm_pmu_ops structure KVM: monolithic: x86: inline more exit handlers in vmx.c KVM: retpolines: x86: eliminate retpoline from vmx.c exit handlers KVM: retpolines: x86: eliminate retpoline from svm.c exit handlers x86: retpolines: eliminate retpoline from msr event handlers arch/x86/events/intel/core.c | 11 + arch/x86/include/asm/kvm_host.h | 15 +- arch/x86/include/asm/kvm_ops.h | 166 ++++++++ arch/x86/include/asm/msr-index.h | 2 + arch/x86/kernel/cpu/bugs.c | 20 +- arch/x86/kvm/Makefile | 5 +- arch/x86/kvm/cpuid.c | 27 +- arch/x86/kvm/hyperv.c | 8 +- arch/x86/kvm/irq.c | 4 - arch/x86/kvm/irq_comm.c | 2 - arch/x86/kvm/kvm_cache_regs.h | 10 +- arch/x86/kvm/lapic.c | 44 +- arch/x86/kvm/mmu.c | 50 +-- arch/x86/kvm/mmu.h | 4 +- arch/x86/kvm/mtrr.c | 2 - arch/x86/kvm/pmu.c | 27 +- arch/x86/kvm/pmu.h | 21 +- arch/x86/kvm/pmu_amd.c | 15 +- arch/x86/kvm/pmu_amd_ops.c | 68 ++++ arch/x86/kvm/pmu_ops.h | 22 + arch/x86/kvm/svm.c | 19 +- arch/x86/kvm/svm_ops.c | 672 ++++++++++++++++++++++++++++++ arch/x86/kvm/trace.h | 4 +- arch/x86/kvm/vmx/pmu_intel.c | 17 +- arch/x86/kvm/vmx/pmu_intel_ops.c | 68 ++++ arch/x86/kvm/vmx/vmx.c | 36 +- arch/x86/kvm/vmx/vmx_ops.c | 675 +++++++++++++++++++++++++++++++ arch/x86/kvm/x86.c | 409 +++++++------------ arch/x86/kvm/x86.h | 2 +- virt/kvm/eventfd.c | 1 - virt/kvm/kvm_main.c | 71 +--- 31 files changed, 1982 insertions(+), 515 deletions(-) create mode 100644 arch/x86/include/asm/kvm_ops.h create mode 100644 arch/x86/kvm/pmu_amd_ops.c create mode 100644 arch/x86/kvm/pmu_ops.h create mode 100644 arch/x86/kvm/svm_ops.c create mode 100644 arch/x86/kvm/vmx/pmu_intel_ops.c create mode 100644 arch/x86/kvm/vmx/vmx_ops.c