Re: [PATCH] KVM: VMX: Cache IA32_DEBUGCTL in memory

[Wanpeng Li]

2017-11-29 16:48 GMT+08:00 Paolo Bonzini <pbonzini@xxxxxxxxxx>:
> On 29/11/2017 07:07, Wanpeng Li wrote:
>> From: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
>>
>> MSR_IA32_DEBUGCTLMSR is zeroed on VMEXIT, so it is saved/restored
>> each time during world switch. Jim from Google pointed out that
>> when running schbench in L2, vmx_vcpu_run will occupy 4% cpu time,
>> and the 25% of vmx_vcpu_run cpu time is occupied by get_debugctlmsr().
>> This patch caches the host IA32_DEBUGCTL MSR and saves/restores
>> the host IA32_DEBUGCTL msr when guest/host switches to avoid to
>> save/restore each time during world switch.
>>
>> Suggested-by: Jim Mattson <jmattson@xxxxxxxxxx>
>> Cc: Jim Mattson <jmattson@xxxxxxxxxx>
>> Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx>
>> Cc: Radim KrÄmÃÅ <rkrcmar@xxxxxxxxxx>
>> Signed-off-by: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
>
> The update_debugctlmsr should stay in vmx_vcpu_run so that tracing
> features work correctly.  However, the get_debugctlmsr indeed can be

The tracing can't run except vCPU is schedule out, so why
update_debugctlmsr should stay in vmx_vcpu_run?

Regards,
Wanpeng Li

> moved to vmx_vcpu_load.
>
>> ---
>>  arch/x86/include/asm/kvm_host.h |  1 +
>>  arch/x86/kvm/vmx.c              | 11 +++++------
>>  2 files changed, 6 insertions(+), 6 deletions(-)
>>
>> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
>> index 63d34bc..c904250 100644
>> --- a/arch/x86/include/asm/kvm_host.h
>> +++ b/arch/x86/include/asm/kvm_host.h
>> @@ -619,6 +619,7 @@ struct kvm_vcpu_arch {
>>       unsigned long dr7;
>>       unsigned long eff_db[KVM_NR_DB_REGS];
>>       unsigned long guest_debug_dr7;
>> +     unsigned long debugctlmsr;
>
> Please rename to host_debugctlmsr and place it in struct vcpu_vmx.
>
> Thanks,
>
> Paolo
>
>>       u64 msr_platform_info;
>>       u64 msr_misc_features_enables;
>>
>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>> index 8c7e816..b167bba 100644
>> --- a/arch/x86/kvm/vmx.c
>> +++ b/arch/x86/kvm/vmx.c
>> @@ -2326,6 +2326,7 @@ static void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
>>
>>       vmx_vcpu_pi_load(vcpu, cpu);
>>       vmx->host_pkru = read_pkru();
>> +     vcpu->arch.debugctlmsr = get_debugctlmsr();
>>  }
>>
>>  static void vmx_vcpu_pi_put(struct kvm_vcpu *vcpu)
>> @@ -2347,6 +2348,9 @@ static void vmx_vcpu_put(struct kvm_vcpu *vcpu)
>>       vmx_vcpu_pi_put(vcpu);
>>
>>       __vmx_load_host_state(to_vmx(vcpu));
>> +     /* MSR_IA32_DEBUGCTLMSR is zeroed on vmexit. Restore it if needed */
>> +     if (vcpu->arch.debugctlmsr)
>> +             update_debugctlmsr(vcpu->arch.debugctlmsr);
>>  }
>>
>>  static bool emulation_required(struct kvm_vcpu *vcpu)
>> @@ -9346,7 +9350,7 @@ static void vmx_arm_hv_timer(struct kvm_vcpu *vcpu)
>>  static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
>>  {
>>       struct vcpu_vmx *vmx = to_vmx(vcpu);
>> -     unsigned long debugctlmsr, cr3, cr4;
>> +     unsigned long cr3, cr4;
>>
>>       /* Record the guest's net vcpu time for enforced NMI injections. */
>>       if (unlikely(!enable_vnmi &&
>> @@ -9399,7 +9403,6 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
>>               __write_pkru(vcpu->arch.pkru);
>>
>>       atomic_switch_perf_msrs(vmx);
>> -     debugctlmsr = get_debugctlmsr();
>>
>>       vmx_arm_hv_timer(vcpu);
>>
>> @@ -9509,10 +9512,6 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
>>  #endif
>>             );
>>
>> -     /* MSR_IA32_DEBUGCTLMSR is zeroed on vmexit. Restore it if needed */
>> -     if (debugctlmsr)
>> -             update_debugctlmsr(debugctlmsr);
>> -
>>  #ifndef CONFIG_X86_64
>>       /*
>>        * The sysexit path does not restore ds/es, so we must set them to
>>
>