korg/linux
0
0
Fork 0
mirror of https://mirrors.bfsu.edu.cn/git/linux.git synced 2025-01-02 11:54:36 +08:00
Code Issues Packages Projects Releases Wiki Activity
e79b91bb3c
linux/arch/x86/kvm
History
Michael Roth e79b91bb3c KVM: SVM: use vmsave/vmload for saving/restoring additional host state 
Using a guest workload which simply issues 'hlt' in a tight loop to
generate VMEXITs, it was observed (on a recent EPYC processor) that a
significant amount of the VMEXIT overhead measured on the host was the
result of MSR reads/writes in svm_vcpu_load/svm_vcpu_put according to
perf:

  67.49%--kvm_arch_vcpu_ioctl_run
          |
          |--23.13%--vcpu_put
          |          kvm_arch_vcpu_put
          |          |
          |          |--21.31%--native_write_msr
          |          |
          |           --1.27%--svm_set_cr4
          |
          |--16.11%--vcpu_load
          |          |
          |           --15.58%--kvm_arch_vcpu_load
          |                     |
          |                     |--13.97%--svm_set_cr4
          |                     |          |
          |                     |          |--12.64%--native_read_msr

Most of these MSRs relate to 'syscall'/'sysenter' and segment bases, and
can be saved/restored using 'vmsave'/'vmload' instructions rather than
explicit MSR reads/writes. In doing so there is a significant reduction
in the svm_vcpu_load/svm_vcpu_put overhead measured for the above
workload:

  50.92%--kvm_arch_vcpu_ioctl_run
          |
          |--19.28%--disable_nmi_singlestep
          |
          |--13.68%--vcpu_load
          |          kvm_arch_vcpu_load
          |          |
          |          |--9.19%--svm_set_cr4
          |          |          |
          |          |           --6.44%--native_read_msr
          |          |
          |           --3.55%--native_write_msr
          |
          |--6.05%--kvm_inject_nmi
          |--2.80%--kvm_sev_es_mmio_read
          |--2.19%--vcpu_put
          |          |
          |           --1.25%--kvm_arch_vcpu_put
          |                     native_write_msr

Quantifying this further, if we look at the raw cycle counts for a
normal iteration of the above workload (according to 'rdtscp'),
kvm_arch_vcpu_ioctl_run() takes ~4600 cycles from start to finish with
the current behavior. Using 'vmsave'/'vmload', this is reduced to
~2800 cycles, a savings of 39%.

While this approach doesn't seem to manifest in any noticeable
improvement for more realistic workloads like UnixBench, netperf, and
kernel builds, likely due to their exit paths generally involving IO
with comparatively high latencies, it does improve overall overhead
of KVM_RUN significantly, which may still be noticeable for certain
situations. It also simplifies some aspects of the code.

With this change, explicit save/restore is no longer needed for the
following host MSRs, since they are documented[1] as being part of the
VMCB State Save Area:

  MSR_STAR, MSR_LSTAR, MSR_CSTAR,
  MSR_SYSCALL_MASK, MSR_KERNEL_GS_BASE,
  MSR_IA32_SYSENTER_CS,
  MSR_IA32_SYSENTER_ESP,
  MSR_IA32_SYSENTER_EIP,
  MSR_FS_BASE, MSR_GS_BASE

and only the following MSR needs individual handling in
svm_vcpu_put/svm_vcpu_load:

  MSR_TSC_AUX

We could drop the host_save_user_msrs array/loop and instead handle
MSR read/write of MSR_TSC_AUX directly, but we leave that for now as
a potential follow-up.

Since 'vmsave'/'vmload' also handles the LDTR and FS/GS segment
registers (and associated hidden state)[2], some of the code
previously used to handle this is no longer needed, so we drop it
as well.

The first public release of the SVM spec[3] also documents the same
handling for the host state in question, so we make these changes
unconditionally.

Also worth noting is that we 'vmsave' to the same page that is
subsequently used by 'vmrun' to record some host additional state. This
is okay, since, in accordance with the spec[2], the additional state
written to the page by 'vmrun' does not overwrite any fields written by
'vmsave'. This has also been confirmed through testing (for the above
CPU, at least).

[1] AMD64 Architecture Programmer's Manual, Rev 3.33, Volume 2, Appendix B, Table B-2
[2] AMD64 Architecture Programmer's Manual, Rev 3.31, Volume 3, Chapter 4, VMSAVE/VMLOAD
[3] Secure Virtual Machine Architecture Reference Manual, Rev 3.01

Suggested-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Message-Id: <20210202190126.2185715-2-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2021-02-04 05:27:34 -05:00
..
mmu KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
svm KVM: SVM: use vmsave/vmload for saving/restoring additional host state 2021-02-04 05:27:34 -05:00
vmx KVM: VMX: Use the kernel's version of VMXOFF 2021-02-04 05:27:33 -05:00
cpuid.c KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
cpuid.h KVM: x86: reinstate vendor-agnostic check on SPEC_CTRL cpuid bits 2020-12-11 19:05:13 -05:00
debugfs.c KVM: let kvm_destroy_vm_debugfs clean up vCPU debugfs directories 2020-06-04 11:00:54 -04:00
emulate.c KVM: X86: Rename DR6_INIT to DR6_ACTIVE_LOW 2021-02-04 05:27:27 -05:00
hyperv.c KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
hyperv.h KVM: x86: hyper-v: allow KVM_GET_SUPPORTED_HV_CPUID as a system ioctl 2020-11-15 09:49:11 -05:00
i8254.c kvm: i8254: remove redundant assignment to pointer s 2020-06-11 12:35:18 -04:00
i8254.h License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
i8259.c KVM: x86: Refactor picdev_write() to prevent Spectre-v1/L1TF attacks 2020-01-27 19:59:37 +01:00
ioapic.c KVM: ioapic: break infinite recursion on lazy EOI 2020-10-24 04:42:06 -04:00
ioapic.h kvm/x86: Remove redundant function implementations 2020-05-27 13:11:10 -04:00
irq_comm.c x86/kvm: Use msi_msg shadow structs 2020-10-28 20:26:26 +01:00
irq.c KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
irq.h kvm/x86: Remove redundant function implementations 2020-05-27 13:11:10 -04:00
Kconfig KVM: SVM: Add support for SEV-ES capability in KVM 2020-12-14 11:09:31 -05:00
kvm_cache_regs.h KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
kvm_emulate.h ARM: 2020-04-02 15:13:15 -07:00
lapic.c KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
lapic.h KVM: Stop using deprecated jump label APIs 2021-02-04 05:27:29 -05:00
Makefile KVM: X86: Implement ring-based dirty memory tracking 2020-11-15 09:49:15 -05:00
mmu.h KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
mtrr.c KVM: x86: remove bogus #GP injection 2020-12-15 05:18:30 -05:00
pmu.c KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
pmu.h KVM: vmx/pmu: Release guest LBR event via lazy release mechanism 2021-02-04 05:27:26 -05:00
trace.h KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
tss.h License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
x86.c KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00
x86.h KVM: x86: use static calls to reduce kvm_x86_ops overhead 2021-02-04 05:27:30 -05:00