Re: [PATCH] KVM: X86: Avoid to handle first-time write when updating the pv stuffs each time

[Paolo Bonzini <pbonzini@redhat.com>]

On 13/11/2017 07:47, Wanpeng Li wrote:
> From: Wanpeng Li <wanpeng.li@hotmail.com>
> 
> There is a logic to handle first-time write when updating the 
> pvclock/wall clock/steal time shared memory pages each time, 
> actually we should do this logic during pv stuffs setup if we 
> suspect the version-field can't be guranteed to be initialized 
> to an even number by the guest. This patch fixes it by handling 
> the first-time write of pvclock/steal time during setup since 
> the update is frequent, and keeping the wall clock since it is 
> rare updating.

The cost is almost zero, and the code is smaller without this patch, so
I think it's a little better to leave things as is.

Paolo

> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Radim Krčmář <rkrcmar@redhat.com>
> Cc: Liran Alon <liran.alon@oracle.com>
> Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
> ---
>  arch/x86/kvm/x86.c | 29 ++++++++++++++++++++++-------
>  1 file changed, 22 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 4552427..19311e0 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -1833,9 +1833,6 @@ static void kvm_setup_pvclock_page(struct kvm_vcpu *v)
>  	 */
>  	BUILD_BUG_ON(offsetof(struct pvclock_vcpu_time_info, version) != 0);
>  
> -	if (guest_hv_clock.version & 1)
> -		++guest_hv_clock.version;  /* first time write, random junk */
> -
>  	vcpu->hv_clock.version = guest_hv_clock.version + 1;
>  	kvm_write_guest_cached(v->kvm, &vcpu->pv_time,
>  				&vcpu->hv_clock,
> @@ -2126,9 +2123,6 @@ static void record_steal_time(struct kvm_vcpu *vcpu)
>  
>  	vcpu->arch.st.steal.preempted = 0;
>  
> -	if (vcpu->arch.st.steal.version & 1)
> -		vcpu->arch.st.steal.version += 1;  /* first time write, random junk */
> -
>  	vcpu->arch.st.steal.version += 1;
>  
>  	kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.st.stime,
> @@ -2256,8 +2250,19 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
>  		     &vcpu->arch.pv_time, data & ~1ULL,
>  		     sizeof(struct pvclock_vcpu_time_info)))
>  			vcpu->arch.pv_time_enabled = false;
> -		else
> +		else {
> +			struct pvclock_vcpu_time_info guest_hv_clock;
> +
>  			vcpu->arch.pv_time_enabled = true;
> +			if (unlikely(kvm_read_guest_cached(vcpu->kvm, &vcpu->arch.pv_time,
> +				&guest_hv_clock, sizeof(guest_hv_clock))))
> +				break;
> +			if (guest_hv_clock.version & 1)
> +				++guest_hv_clock.version;  /* first time write, random junk */
> +			kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.pv_time,
> +				&vcpu->arch.hv_clock,
> +				sizeof(vcpu->arch.hv_clock.version));
> +		}
>  
>  		break;
>  	}
> @@ -2283,6 +2288,16 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
>  		if (!(data & KVM_MSR_ENABLED))
>  			break;
>  
> +		if (unlikely(kvm_read_guest_cached(vcpu->kvm, &vcpu->arch.st.stime,
> +			&vcpu->arch.st.steal, sizeof(struct kvm_steal_time))))
> +			break;
> +
> +		if (vcpu->arch.st.steal.version & 1)
> +			vcpu->arch.st.steal.version += 1;  /* first time write, random junk */
> +
> +		kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.st.stime,
> +			&vcpu->arch.st.steal, sizeof(struct kvm_steal_time));
> +
>  		kvm_make_request(KVM_REQ_STEAL_UPDATE, vcpu);
>  
>  		break;
>