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, 30 Mar 2022 00:04:41 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Wanpeng Li <kernellwp@gmail.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>
Subject: Re: [PATCH RESEND 2/5] KVM: X86: Add guest interrupt disable state
 support
Message-ID: <YkOembt1lvTEJrx0@google.com>
References: <1648216709-44755-1-git-send-email-wanpengli@tencent.com>
 <1648216709-44755-3-git-send-email-wanpengli@tencent.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1648216709-44755-3-git-send-email-wanpengli@tencent.com>
Precedence: bulk

On Fri, Mar 25, 2022, Wanpeng Li wrote:
> From: Wanpeng Li <wanpengli@tencent.com>
> 
> Let's get the information whether or not guests disable interruptions.

This is missing critical information for _why_.  It took me some staring to
understand that this allows querying IRQs from a _different_ vCPU, which needs
caching on VMX due to the need to do a VMREAD.

> Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
> ---
>  arch/x86/include/asm/kvm_host.h | 1 +
>  arch/x86/kvm/x86.c              | 3 +++
>  2 files changed, 4 insertions(+)
> 
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 50f011a7445a..8e05cbfa9827 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -861,6 +861,7 @@ struct kvm_vcpu_arch {
>  		bool preempt_count_enabled;
>  		struct gfn_to_hva_cache preempt_count_cache;
>  	} pv_pc;
> +	bool irq_disabled;

This is going to at best be confusing, and at worst lead to bugs  The flag is
valid if and only if the vCPU is not loaded.  I don't have a clever answer, but
this needs to have some form of guard to (a) clarify when it's valid and (b) actively
prevent misuse.