Re: [RFC PATCH v5 0/6] KVM: x86: add per-vCPU exits disable capability

[Zhi Wang <zhi.wang.linux@gmail.com>]

On Wed, 18 Jan 2023 10:30:03 +0200
Zhi Wang <zhi.wang.linux@gmail.com> wrote:

Hi:

No sure why the test never finishes on my testing machine. Will take a look
later today.

My CPU: 
model name      : Intel(R) Core(TM) i7-8700 CPU @ 3.20GHz

branch kvm.git/master top commit:
310bc39546a435c83cc27a0eba878afac0d74714

-----

[inno@inno-lk-server x86_64]$ time ./disable_exits_test
VM-scoped tests start
Halter vCPU thread started
vCPU thread running vCPU 0
Halter vCPU thread reported its first HLT executed after 1 seconds.
vCPU thread running vCPU 1
Halter vCPU had 10 halt exits
Guest records 10 HLTs executed, waked 9 times
Halter vCPU thread started
vCPU thread running vCPU 0



^C

real    19m0.923s
user    37m56.512s
sys     0m1.086s

-----

> On Fri, 13 Jan 2023 22:01:08 +0000
> Kechen Lu <kechenl@nvidia.com> wrote:
> 
> Hi:
> 
> checkpatch.pl throws a lot of warning and errors when I was trying
> this series. Can you fix them?
> 
> total: 470 errors, 22 warnings, 464 lines checked
> 
> > Summary
> > ===========
> > Introduce support of vCPU-scoped ioctl with KVM_CAP_X86_DISABLE_EXITS
> > cap for disabling exits to enable finer-grained VM exits disabling
> > on per vCPU scales instead of whole guest. This patch series enabled
> > the vCPU-scoped exits control and toggling.
> > 
> > Motivation
> > ============
> > In use cases like Windows guest running heavy CPU-bound
> > workloads, disabling HLT VM-exits could mitigate host sched ctx switch
> > overhead. Simply HLT disabling on all vCPUs could bring
> > performance benefits, but if no pCPUs reserved for host threads, could
> > happened to the forced preemption as host does not know the time to do
> > the schedule for other host threads want to run. With this patch, we
> > could only disable part of vCPUs HLT exits for one guest, this still
> > keeps performance benefits, and also shows resiliency to host stressing
> > workload running at the same time.
> > 
> > Performance and Testing
> > =========================
> > In the host stressing workload experiment with Windows guest heavy
> > CPU-bound workloads, it shows good resiliency and having the ~3%
> > performance improvement. E.g. Passmark running in a Windows guest
> > with this patch disabling HLT exits on only half of vCPUs still
> > showing 2.4% higher main score v/s baseline.
> > 
> > Tested everything on AMD machines.
> > 
> > v4->v5 :
> > - Drop the usage of KVM request, keep the VM-scoped exits disable
> >   as the existing design, and only allow per-vCPU settings to
> >   override the per-VM settings (Sean Christopherson)
> > - Refactor the disable exits selftest without introducing any
> >   new prerequisite patch, tests per-vCPU exits disable and overrides,
> >   and per-VM exits disable
> > 
> > v3->v4 (Chao Gao) :
> > - Use kvm vCPU request KVM_REQ_DISABLE_EXIT to perform the arch
> >   VMCS updating (patch 5)
> > - Fix selftests redundant arguments (patch 7)
> > - Merge overlapped fix bits from patch 4 to patch 3
> > 
> > v2->v3 (Sean Christopherson) :
> > - Reject KVM_CAP_X86_DISABLE_EXITS if userspace disable MWAIT exits
> >   when MWAIT is not allowed in guest (patch 3)
> > - Make userspace able to re-enable previously disabled exits (patch 4)
> > - Add mwait/pause/cstate exits flag toggling instead of only hlt
> >   exits (patch 5)
> > - Add selftests for KVM_CAP_X86_DISABLE_EXITS (patch 7)
> > 
> > v1->v2 (Sean Christopherson) :
> > - Add explicit restriction for VM-scoped exits disabling to be called
> >   before vCPUs creation (patch 1)
> > - Use vCPU ioctl instead of 64bit vCPU bitmask (patch 5), and make exits
> >   disable flags check purely for vCPU instead of VM (patch 2)
> > 
> > Best Regards,
> > Kechen
> > 
> > Kechen Lu (3):
> >   KVM: x86: Move *_in_guest power management flags to vCPU scope
> >   KVM: x86: add vCPU scoped toggling for disabled exits
> >   KVM: selftests: Add tests for VM and vCPU cap
> >     KVM_CAP_X86_DISABLE_EXITS
> > 
> > Sean Christopherson (3):
> >   KVM: x86: only allow exits disable before vCPUs created
> >   KVM: x86: Reject disabling of MWAIT interception when not allowed
> >   KVM: x86: Let userspace re-enable previously disabled exits
> > 
> >  Documentation/virt/kvm/api.rst                |   8 +-
> >  arch/x86/include/asm/kvm-x86-ops.h            |   1 +
> >  arch/x86/include/asm/kvm_host.h               |   7 +
> >  arch/x86/kvm/cpuid.c                          |   4 +-
> >  arch/x86/kvm/lapic.c                          |   7 +-
> >  arch/x86/kvm/svm/nested.c                     |   4 +-
> >  arch/x86/kvm/svm/svm.c                        |  42 +-
> >  arch/x86/kvm/vmx/vmx.c                        |  53 +-
> >  arch/x86/kvm/x86.c                            |  69 ++-
> >  arch/x86/kvm/x86.h                            |  16 +-
> >  include/uapi/linux/kvm.h                      |   4 +-
> >  tools/testing/selftests/kvm/Makefile          |   1 +
> >  .../selftests/kvm/x86_64/disable_exits_test.c | 457 ++++++++++++++++++
> >  13 files changed, 626 insertions(+), 47 deletions(-)
> >  create mode 100644 tools/testing/selftests/kvm/x86_64/disable_exits_test.c
> > 
>