Re: [PATCH v2 02/15] kvm: x86: Introduce KVM APICv state

[Alexander Graf <graf@xxxxxxxxxx>]

On 15.08.19 18:25, Suthikulpanit, Suravee wrote:
> Currently, after a VM boots with APICv enabled, it could go into
> the following states:
>    * activated   = VM is running w/ APICv
>    * deactivated = VM deactivate APICv (temporary)
>    * disabled    = VM deactivate APICv (permanent)
>
> Introduce KVM APICv state enum to help keep track of the APICv states
> along with a new variable struct kvm_arch.apicv_state to store
> the current state.
>
> Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@xxxxxxx>
> ---
>   arch/x86/include/asm/kvm_host.h | 11 +++++++++++
>   arch/x86/kvm/x86.c              | 14 +++++++++++++-
>   2 files changed, 24 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 56bc702..04d7066 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -845,6 +845,15 @@ enum kvm_irqchip_mode {
>   	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
>   };
>   
> +/*
> + * KVM assumes all vcpus in a VM operate in the same mode.
> + */
> +enum kvm_apicv_state {
> +	APICV_DISABLED,		/* Disabled (such as for Hyper-V case) */
> +	APICV_DEACTIVATED,	/* Deactivated tempoerary */

typo

I'm also not sure the name is 100% obvious. How about something like 
"suspended" or "paused"?

> +	APICV_ACTIVATED,	/* Default status when APICV is enabled */
> +};
> +
>   struct kvm_arch {
>   	unsigned long n_used_mmu_pages;
>   	unsigned long n_requested_mmu_pages;
> @@ -873,6 +882,8 @@ struct kvm_arch {
>   	struct kvm_apic_map *apic_map;
>   
>   	bool apic_access_page_done;
> +	struct mutex apicv_lock;
> +	enum kvm_apicv_state apicv_state;
>   
>   	gpa_t wall_clock;
>   
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 7daf0dd..f9c3f63 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -4584,6 +4584,8 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
>   		kvm->arch.irqchip_mode = KVM_IRQCHIP_SPLIT;
>   		kvm->arch.nr_reserved_ioapic_pins = cap->args[0];
>   		r = 0;
> +		if (kvm_x86_ops->get_enable_apicv(kvm))
> +			kvm->arch.apicv_state = APICV_ACTIVATED;
>   split_irqchip_unlock:
>   		mutex_unlock(&kvm->lock);
>   		break;
> @@ -4701,6 +4703,8 @@ long kvm_arch_vm_ioctl(struct file *filp,
>   		/* Write kvm->irq_routing before enabling irqchip_in_kernel. */
>   		smp_wmb();
>   		kvm->arch.irqchip_mode = KVM_IRQCHIP_KERNEL;
> +		if (kvm_x86_ops->get_enable_apicv(kvm))
> +			kvm->arch.apicv_state = APICV_ACTIVATED;
>   	create_irqchip_unlock:
>   		mutex_unlock(&kvm->lock);
>   		break;
> @@ -9150,13 +9154,18 @@ int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
>   		goto fail_free_pio_data;
>   
>   	if (irqchip_in_kernel(vcpu->kvm)) {
> -		vcpu->arch.apicv_active = kvm_x86_ops->get_enable_apicv(vcpu->kvm);

Why are you moving this into a locked section?

>   		r = kvm_create_lapic(vcpu, lapic_timer_advance_ns);
>   		if (r < 0)
>   			goto fail_mmu_destroy;
>   	} else
>   		static_key_slow_inc(&kvm_no_apic_vcpu);
>   
> +	mutex_lock(&vcpu->kvm->arch.apicv_lock);
> +	if (irqchip_in_kernel(vcpu->kvm) &&
> +	    vcpu->kvm->arch.apicv_state == APICV_ACTIVATED)
> +		vcpu->arch.apicv_active = kvm_x86_ops->get_enable_apicv(vcpu->kvm);
> +	mutex_unlock(&vcpu->kvm->arch.apicv_lock);
> +
>   	vcpu->arch.mce_banks = kzalloc(KVM_MAX_MCE_BANKS * sizeof(u64) * 4,
>   				       GFP_KERNEL_ACCOUNT);
>   	if (!vcpu->arch.mce_banks) {
> @@ -9255,6 +9264,9 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
>   	kvm_page_track_init(kvm);
>   	kvm_mmu_init_vm(kvm);
>   
> +	/* APICV initialization */
> +	mutex_init(&kvm->arch.apicv_lock);

In fact, the whole lock story is not part of the patch description :).


Alex

> +
>   	if (kvm_x86_ops->vm_init)
>   		return kvm_x86_ops->vm_init(kvm);
>