Re: [PATCH 6/6] KVM: VMX: enable IPI virtualization

From: Paolo Bonzini <pbonzini@redhat.com>
To: Zeng Guang <guang.zeng@intel.com>,
	Sean Christopherson <seanjc@google.com>,
	Vitaly Kuznetsov <vkuznets@redhat.com>,
	Wanpeng Li <wanpengli@tencent.com>,
	Jim Mattson <jmattson@google.com>, Joerg Roedel <joro@8bytes.org>,
	"kvm@vger.kernel.org" <kvm@vger.kernel.org>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	"Luck, Tony" <tony.luck@intel.com>,
	Kan Liang <kan.liang@linux.intel.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	"H. Peter Anvin" <hpa@zytor.com>,
	Kim Phillips <kim.phillips@amd.com>,
	Jarkko Sakkinen <jarkko@kernel.org>,
	Jethro Beekman <jethro@fortanix.com>,
	"Huang, Kai" <kai.huang@intel.com>
Cc: "x86@kernel.org" <x86@kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"Hu, Robert" <robert.hu@intel.com>,
	"Gao, Chao" <chao.gao@intel.com>,
	Emanuele Giuseppe Esposito <eesposit@redhat.com>
Subject: Re: [PATCH 6/6] KVM: VMX: enable IPI virtualization
Date: Mon, 19 Jul 2021 15:58:17 +0200	[thread overview]
Message-ID: <223af27a-2412-40f6-f4a6-e0a662041855@redhat.com> (raw)
In-Reply-To: <949abcb7-5f24-2107-a089-5e6c1bee8cf2@intel.com>

On 19/07/21 14:38, Zeng Guang wrote:
>> Understood, but in practice all uses of vmx->ipiv_active are
>> guarded by kvm_vcpu_apicv_active so they are always reached with
>> vmx->ipiv_active == enable_ipiv.
>> 
>> The one above instead seems wrong and should just use enable_ipiv.
> 
> enable_ipiv associate with "IPI virtualization" setting in tertiary
> exec controls and enable_apicv which depends on cpu_has_vmx_apicv().
> kvm_vcpu_apicv_active still can be false even if enable_ipiv is true,
> e.g. in case irqchip not emulated in kernel.

Right, kvm_vcpu_apicv_active *is* set in init_vmcs.  But there's an 
     "if (kvm_vcpu_apicv_active(&vmx->vcpu))" above.  You can just stick

	if (enable_ipicv)
		install_pid(vmx);

inside there.  As to the other occurrences of vmx->ipiv_active, look here:

> +	if (!kvm_vcpu_apicv_active(vcpu))
> +		return;
> +
> +	if ((!kvm_arch_has_assigned_device(vcpu->kvm) ||
> +		!irq_remapping_cap(IRQ_POSTING_CAP)) &&
> +		!to_vmx(vcpu)->ipiv_active)
>  		return;
>  

This one can be enable_ipiv because APICv must be active.

> +	if (!kvm_vcpu_apicv_active(vcpu))
> +		return 0;
> +
> +	/* Put vCPU into a list and set NV to wakeup vector if it is
> +	 * one of the following cases:
> +	 * 1. any assigned device is in use.
> +	 * 2. IPI virtualization is enabled.
> +	 */
> +	if ((!kvm_arch_has_assigned_device(vcpu->kvm) ||
> +		!irq_remapping_cap(IRQ_POSTING_CAP)) && !to_vmx(vcpu)->ipiv_active)
>  		return 0;

This one can be !enable_ipiv because APICv must be active.

> 
> @@ -3870,6 +3877,8 @@ static void vmx_update_msr_bitmap_x2apic(struct kvm_vcpu *vcpu, u8 mode)
>  		vmx_enable_intercept_for_msr(vcpu, X2APIC_MSR(APIC_TMCCT), MSR_TYPE_RW);
>  		vmx_disable_intercept_for_msr(vcpu, X2APIC_MSR(APIC_EOI), MSR_TYPE_W);
>  		vmx_disable_intercept_for_msr(vcpu, X2APIC_MSR(APIC_SELF_IPI), MSR_TYPE_W);
> +		vmx_set_intercept_for_msr(vcpu, X2APIC_MSR(APIC_ICR),
> +				MSR_TYPE_RW, !to_vmx(vcpu)->ipiv_active);
>  	}
>  }

Is inside "if (mode & MSR_BITMAP_MODE_X2APIC_APICV)" so APICv must be 
activ; so it can be enable_ipiv as well.

In conclusion, you do not need vmx->ipiv_active.

Paolo