Re: [PATCH 2/3] KVM/Hyper-V: Add new KVM cap KVM_CAP_HYPERV_DIRECT_TLBFLUSH

[Vitaly Kuznetsov <vkuznets@redhat.com>]

lantianyu1986@gmail.com writes:

> From: Tianyu Lan <Tianyu.Lan@microsoft.com>
>
> This patch adds new KVM cap KVM_CAP_HYPERV_DIRECT_TLBFLUSH and let
> user space to enable direct tlb flush function when only Hyper-V
> hypervsior capability is exposed to VM. This patch also adds
> enable_direct_tlbflush callback in the struct kvm_x86_ops and
> platforms may use it to implement direct tlb flush support.
>
> Signed-off-by: Tianyu Lan <Tianyu.Lan@microsoft.com>
> ---
>  Documentation/virtual/kvm/api.txt | 10 ++++++++++
>  arch/x86/include/asm/kvm_host.h   |  2 ++
>  arch/x86/kvm/x86.c                |  8 ++++++++
>  include/uapi/linux/kvm.h          |  1 +
>  4 files changed, 21 insertions(+)
>
> diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
> index 2cd6250b2896..45308ed6dd75 100644
> --- a/Documentation/virtual/kvm/api.txt
> +++ b/Documentation/virtual/kvm/api.txt
> @@ -5289,3 +5289,13 @@ Architectures: x86
>  This capability indicates that KVM supports paravirtualized Hyper-V IPI send
>  hypercalls:
>  HvCallSendSyntheticClusterIpi, HvCallSendSyntheticClusterIpiEx.
> +8.21 KVM_CAP_HYPERV_DIRECT_TLBFLUSH
> +
> +Architecture: x86
> +
> +This capability indicates that KVM supports Hyper-V direct tlb flush function.
> +User space should enable this feature only when Hyper-V hypervisor capability
> +is exposed to guest and KVM profile is hided. Both Hyper-V and KVM hypercalls
> +use RAX and RCX registers to pass parameters. If KVM hypercall is exposed
> +to L2 guest with direct tlbflush enabled, Hyper-V may mistake KVM hypercall
> +for Hyper-V tlb flush Hypercall due to paremeter register overlap.

First, we need to explicitly state that this is for KVM on Hyper-V and
second, that this disables normal hypercall handling by KVM.

My take:

This capability indicates that KVM running on top of Hyper-V hypervisor
enables Direct TLB flush for its guests meaning that TLB flush
hypercalls are handled by Level 1 hypervisor (Hyper-V) bypassing KVM. 
Due to the different ABI for hypercall parameters between Hyper-V and
KVM, enabling this capability effectively disables all hypercall
handling by KVM (as some KVM hypercall may be mistakenly treated as TLB
flush hypercalls by Hyper-C) so userspace should disable KVM
identification in CPUID.

I think we should also enforce this somehow leaving only Hyper-V style
hypercalls handling (for Windows guests) in place.

> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 0cc5b611a113..667d154e89d4 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -1205,6 +1205,8 @@ struct kvm_x86_ops {
>  	uint16_t (*nested_get_evmcs_version)(struct kvm_vcpu *vcpu);
>  
>  	bool (*need_emulation_on_page_fault)(struct kvm_vcpu *vcpu);
> +
> +	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
>  };
>  
>  struct kvm_arch_async_pf {
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 9d7b9e6a0939..a9d8ee7f7bf0 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -3183,6 +3183,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
>  		r = kvm_x86_ops->get_nested_state ?
>  			kvm_x86_ops->get_nested_state(NULL, NULL, 0) : 0;
>  		break;
> +	case KVM_CAP_HYPERV_DIRECT_TLBFLUSH:
> +		r = kvm_x86_ops->enable_direct_tlbflush ? 1 : 0;
> +		break;
>  	default:
>  		break;
>  	}
> @@ -3953,6 +3956,11 @@ static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
>  				r = -EFAULT;
>  		}
>  		return r;
> +	case KVM_CAP_HYPERV_DIRECT_TLBFLUSH:
> +		if (!kvm_x86_ops->enable_direct_tlbflush)
> +			return -ENOTTY;
> +
> +		return kvm_x86_ops->enable_direct_tlbflush(vcpu);
>  
>  	default:
>  		return -EINVAL;
> diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
> index a7c19540ce21..cb959bc925b1 100644
> --- a/include/uapi/linux/kvm.h
> +++ b/include/uapi/linux/kvm.h
> @@ -996,6 +996,7 @@ struct kvm_ppc_resize_hpt {
>  #define KVM_CAP_ARM_PTRAUTH_ADDRESS 171
>  #define KVM_CAP_ARM_PTRAUTH_GENERIC 172
>  #define KVM_CAP_PMU_EVENT_FILTER 173
> +#define KVM_CAP_HYPERV_DIRECT_TLBFLUSH 174
>  
>  #ifdef KVM_CAP_IRQ_ROUTING

-- 
Vitaly