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, 04 Jan 2023 13:20:13 -0800
In-Reply-To: <1cacbda18e3c7dcccd92a7390b0ca7f4ba073f85.1667110240.git.isaku.yamahata@intel.com>
Mime-Version: 1.0
Message-ID: <diqz5ydmov3m.fsf@google.com>
Subject: Re: [PATCH v10 098/108] KVM: TDX: Implement callbacks for MSR
 operations for TDX
From:   Ackerley Tng <ackerleytng@google.com>
To:     isaku.yamahata@intel.com
Cc:     kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
        isaku.yamahata@intel.com, isaku.yamahata@gmail.com,
        pbonzini@redhat.com, erdemaktas@google.com, seanjc@google.com,
        sagis@google.com, dmatlack@google.com
Content-Type: text/plain; charset="UTF-8"; format=flowed; delsp=yes
Precedence: bulk


I believe we should also have a handler for .msr_filter_changed.

Without an .msr_filter_changed handler, a host crash can occur if we
first set up a vcpu for the TD, and then set an MSR filter.

If we first set up a vcpu for the TD, and then set an MSR filter, upon
vcpu_enter_guest, the .msr_filter_changed handler (currently
vmx_msr_filter_changed()) will be invoked. to_vmx(vcpu) interprets the
containing struct of struct kvm_vcpu to be a struct vcpu_vmx instead of
a struct vcpu_tdx.

In my case, I was working on a selftest and the missing handler caused a
NULL dereference in vmx_disable_intercept_for_msr() because
vmx->vmcs01.msr_bitmap is NULL.