Message ID | 20180725163057.13948-1-vkuznets@redhat.com (mailing list archive) |
---|---|
Headers | show |
Series | KVM: nVMX: Enlightened VMCS for Hyper-V on KVM | expand |
On 25/07/2018 18:30, Vitaly Kuznetsov wrote: > Changes since v2: > - Rebase to the current kvm/queue. > - Simplify clean fields check in prepare_vmcs02{,_full} and > copy_enlightened_to_vmcs12() by resetting the clean fields mask in > nested_vmx_handle_enlightened_vmptrld() when we switch from one L2 guest > to another [Paolo Bonzini] > - Cope with the newly introduced vmx_get_nested_state() by returning > -ENOTSUPP when eVMCS is in use. We'll do proper nested state save/restore > later. Looks good, but we have to do something about live migration. Perhaps you can write a testcase analogous to the new state_test, that tests eVMCS in general before we have support in QEMU. Thanks, Paolo > Description: > > This is an initial implementation of Enlightened VMCS for nested Hyper-V on > KVM. Using it helps to spare 1500 cpu cycles for nested vmexit (tight cpuid > loop in WS2016 with Hyper-V role on KVM: 15200 cycles -> 13700 cycles). > > Ladi Prosek (1): > KVM: hyperv: define VP assist page helpers > > Vitaly Kuznetsov (5): > KVM: VMX: refactor evmcs_sanitize_exec_ctrls() > KVM: nVMX: add KVM_CAP_HYPERV_ENLIGHTENED_VMCS capability > KVM: nVMX: add enlightened VMCS state > KVM: nVMX: implement enlightened VMPTRLD and VMCLEAR > KVM: nVMX: optimize prepare_vmcs02{,_full} for Enlightened VMCS case > > arch/x86/include/asm/kvm_host.h | 3 + > arch/x86/kvm/hyperv.c | 23 +- > arch/x86/kvm/hyperv.h | 4 + > arch/x86/kvm/lapic.c | 4 +- > arch/x86/kvm/lapic.h | 2 +- > arch/x86/kvm/svm.c | 9 + > arch/x86/kvm/vmx.c | 824 +++++++++++++++++++++++++++++++++------- > arch/x86/kvm/x86.c | 17 +- > include/uapi/linux/kvm.h | 1 + > 9 files changed, 737 insertions(+), 150 deletions(-) >
Paolo Bonzini <pbonzini@redhat.com> writes: > On 25/07/2018 18:30, Vitaly Kuznetsov wrote: >> Changes since v2: >> - Rebase to the current kvm/queue. >> - Simplify clean fields check in prepare_vmcs02{,_full} and >> copy_enlightened_to_vmcs12() by resetting the clean fields mask in >> nested_vmx_handle_enlightened_vmptrld() when we switch from one L2 guest >> to another [Paolo Bonzini] >> - Cope with the newly introduced vmx_get_nested_state() by returning >> -ENOTSUPP when eVMCS is in use. We'll do proper nested state save/restore >> later. > > Looks good, but we have to do something about live migration. Sure, I don't actually see any fundumental problems when eVMCS is in use, however, I'd like to enable migration when we know that it works so I was going to wait for your nested virtualization state migration series to land and gain Qemu support before starting to experiment. "One bite at a time". > Perhaps you can write a testcase analogous to the new state_test, that > tests eVMCS in general before we have support in QEMU. I'll take a look, thanks for the pointer!
On 02/08/2018 15:22, Vitaly Kuznetsov wrote: >> Looks good, but we have to do something about live migration. > Sure, > > I don't actually see any fundumental problems when eVMCS is in use, > however, I'd like to enable migration when we know that it works so I > was going to wait for your nested virtualization state migration series > to land and gain Qemu support before starting to experiment. "One bite > at a time". > Don't worry about QEMU - working with tools/testing/selftests/kvm/ is much simpler (as long as a similar test exists to what you're doing). Paolo
Just make clean and it will work. Will debug it. Paolo ----- Paolo Bonzini <pbonzini@redhat.com> ha scritto: > On 02/08/2018 15:22, Vitaly Kuznetsov wrote: > >> Looks good, but we have to do something about live migration. > > Sure, > > > > I don't actually see any fundumental problems when eVMCS is in use, > > however, I'd like to enable migration when we know that it works so I > > was going to wait for your nested virtualization state migration series > > to land and gain Qemu support before starting to experiment. "One bite > > at a time". > > > > Don't worry about QEMU - working with tools/testing/selftests/kvm/ is > much simpler (as long as a similar test exists to what you're doing). > > Paolo