Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 6 Jul 2022 18:44:47 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Vitaly Kuznetsov <vkuznets@redhat.com>
Cc:     kvm@vger.kernel.org, Paolo Bonzini <pbonzini@redhat.com>,
        Anirudh Rayabharam <anrayabh@linux.microsoft.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Maxim Levitsky <mlevitsk@redhat.com>,
        linux-hyperv@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 15/28] KVM: VMX: Check
 CPU_BASED_{INTR,NMI}_WINDOW_EXITING in setup_vmcs_config()
Message-ID: <YsXYH43clik02kZ9@google.com>
References: <20220629150625.238286-1-vkuznets@redhat.com>
 <20220629150625.238286-16-vkuznets@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220629150625.238286-16-vkuznets@redhat.com>
Precedence: bulk

On Wed, Jun 29, 2022, Vitaly Kuznetsov wrote:
> CPU_BASED_{INTR,NMI}_WINDOW_EXITING controls are toggled dynamically by
> vmx_enable_{irq,nmi}_window, handle_interrupt_window(), handle_nmi_window()
> but setup_vmcs_config() doesn't check their existence. Add the check and
> filter the controls out in vmx_exec_control().
> 
> No (real) functional change intended as all existing CPUs supporting
> VMX are supposed to have these controls.
> 
> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
> ---
>  arch/x86/kvm/vmx/vmx.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> index da8bbba38d0e..89a3bbafa5af 100644
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -2487,7 +2487,9 @@ static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf,
>  	      CPU_BASED_MWAIT_EXITING |
>  	      CPU_BASED_MONITOR_EXITING |
>  	      CPU_BASED_INVLPG_EXITING |
> -	      CPU_BASED_RDPMC_EXITING;
> +	      CPU_BASED_RDPMC_EXITING |
> +	      CPU_BASED_INTR_WINDOW_EXITING |
> +	      CPU_BASED_NMI_WINDOW_EXITING;

Requiring NMI_WINDOW_EXITING is wrong, KVM doesn't use NMI_WINDOW_EXITING if
enable_vnmi == false:

	if (!enable_vnmi ||
	    vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_STI) {
		vmx_enable_irq_window(vcpu);
		return;
	}

	exec_controls_setbit(to_vmx(vcpu), CPU_BASED_NMI_WINDOW_EXITING);

And enable_vnmi is cleared if PIN_BASED_VIRTUAL_NMIS are unsupported:

	if (!cpu_has_virtual_nmis())
		enable_vnmi = 0;

I suspect CPUs that don't support VIRTUAL_NMIS also don't support NMI_WINDOW_EXITING,
and KVM explicitly supports such CPUs.  See commit 8a1b43922d0d ("kvm: vmx: Reinstate
support for CPUs without virtual NMI").

Making NMI_WINDOW_EXITING optional and then adding it to cpu_has_virtual_nmis()
seems like the correct approach.

diff --git a/arch/x86/kvm/vmx/capabilities.h b/arch/x86/kvm/vmx/capabilities.h
index 069d8d298e1d..c34890e9fc2a 100644
--- a/arch/x86/kvm/vmx/capabilities.h
+++ b/arch/x86/kvm/vmx/capabilities.h
@@ -82,7 +82,8 @@ static inline bool cpu_has_vmx_basic_inout(void)

 static inline bool cpu_has_virtual_nmis(void)
 {
-       return vmcs_config.pin_based_exec_ctrl & PIN_BASED_VIRTUAL_NMIS;
+       return vmcs_config.pin_based_exec_ctrl & PIN_BASED_VIRTUAL_NMIS &&
+              vmcs_config.cpu_based_exec_ctrl & CPU_BASED_NMI_WINDOW_EXITING;
 }

 static inline bool cpu_has_vmx_preemption_timer(void)


>  	opt = CPU_BASED_TPR_SHADOW |
>  	      CPU_BASED_USE_MSR_BITMAPS |
> @@ -4300,6 +4302,10 @@ static u32 vmx_exec_control(struct vcpu_vmx *vmx)
>  {
>  	u32 exec_control = vmcs_config.cpu_based_exec_ctrl;
>  
> +	/* INTR_WINDOW_EXITING and NMI_WINDOW_EXITING are toggled dynamically */
> +	exec_control &= ~(CPU_BASED_INTR_WINDOW_EXITING |
> +			  CPU_BASED_NMI_WINDOW_EXITING);
> +
>  	if (vmx->vcpu.arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)
>  		exec_control &= ~CPU_BASED_MOV_DR_EXITING;
>  
> -- 
> 2.35.3
>