Re: [RFC Patch 3/3] KVM/x86: Add tlb_remote_flush callback support for vmcs

From: Vitaly Kuznetsov
Date: Tue Jun 12 2018 - 11:12:14 EST


Tianyu Lan <Tianyu.Lan@xxxxxxxxxxxxx> writes:

> Register tlb_remote_flush callback for vmcs when hyperv capability of
> nested guest mapping flush is detected. The interface can help to reduce
> overhead when flush ept table among vcpus for nested VM. The tradition way
> is to send IPIs to all affected vcpus and executes INVEPT on each vcpus.
> It will trigger several vmexits for IPI and INVEPT emulation. Hyperv provides
> such hypercall to do flush for all vcpus.
>
> Signed-off-by: Lan Tianyu <Tianyu.Lan@xxxxxxxxxxxxx>
> ---
> arch/x86/kvm/vmx.c | 15 +++++++++++++++
> 1 file changed, 15 insertions(+)
>
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index e50beb76d846..6cb241c05690 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -4737,6 +4737,17 @@ static inline void __vmx_flush_tlb(struct kvm_vcpu *vcpu, int vpid,
> }
> }
>
> +static int vmx_remote_flush_tlb(struct kvm *kvm)
> +{
> + struct kvm_vcpu *vcpu = kvm_get_vcpu(kvm, 0);
> +
> + if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
> + return -1;

Why vcpu0? Can arch.mmu.root_hpa-s differ across vCPUs? What happens if
they do?

> +
> + return hyperv_flush_guest_mapping(construct_eptp(vcpu,
> + vcpu->arch.mmu.root_hpa));
> +}

The 'vmx_remote_flush_tlb' name looks generic enough but it is actually
Hyper-V-specific. I'd suggest renaming to something like
hv_remote_flush_tlb().

> +
> static void vmx_flush_tlb(struct kvm_vcpu *vcpu, bool invalidate_gpa)
> {
> __vmx_flush_tlb(vcpu, to_vmx(vcpu)->vpid, invalidate_gpa);
> @@ -7495,6 +7506,10 @@ static __init int hardware_setup(void)
> if (enable_ept && !cpu_has_vmx_ept_2m_page())
> kvm_disable_largepages();
>
> + if (ms_hyperv.nested_features & HV_X64_NESTED_GUSET_MAPPING_FLUSH
> + && enable_ept)
> + kvm_x86_ops->tlb_remote_flush = vmx_remote_flush_tlb;
> +
> if (!cpu_has_vmx_ple()) {
> ple_gap = 0;
> ple_window = 0;

--
Vitaly