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;
Message-ID: <cd7fcdec-bf2c-ca27-4355-ce56ab9538d9@redhat.com>
Date:   Wed, 2 Nov 2022 17:02:36 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.4.0
Subject: Re: [PATCH 6/7] KVM: SVM: move MSR_IA32_SPEC_CTRL save/restore to
 assembly
Content-Language: en-US
To:     Josh Poimboeuf <jpoimboe@kernel.org>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        jmattson@google.com, seanjc@google.com
References: <20221028230723.3254250-1-pbonzini@redhat.com>
 <20221028230723.3254250-7-pbonzini@redhat.com>
 <20221102152814.lmuzib5472zsaroy@treble>
From:   Paolo Bonzini <pbonzini@redhat.com>
In-Reply-To: <20221102152814.lmuzib5472zsaroy@treble>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 11/2/22 16:28, Josh Poimboeuf wrote:
> On Fri, Oct 28, 2022 at 07:07:22PM -0400, Paolo Bonzini wrote:
>> +++ b/arch/x86/kvm/svm/svm.c
>> @@ -3918,10 +3918,21 @@ static noinstr void svm_vcpu_enter_exit(struct kvm_vcpu *vcpu)
>>   	struct vcpu_svm *svm = to_svm(vcpu);
>>   	unsigned long vmcb_pa = svm->current_vmcb->pa;
>>   
>> +	/*
>> +	 * For non-nested case:
>> +	 * If the L01 MSR bitmap does not intercept the MSR, then we need to
>> +	 * save it.
>> +	 *
>> +	 * For nested case:
>> +	 * If the L02 MSR bitmap does not intercept the MSR, then we need to
>> +	 * save it.
>> +	 */
>> +	bool spec_ctrl_intercepted = msr_write_intercepted(vcpu, MSR_IA32_SPEC_CTRL);
> 
> This triggers a warning:
> 
>    vmlinux.o: warning: objtool: svm_vcpu_enter_exit+0x3d: call to svm_msrpm_offset() leaves .noinstr.text section
> 
> svm_vcpu_enter_exit() is noinstr, but it's calling
> msr_write_intercepted() which is not.

I suspect I didn't see it because it's inlined here, but it has to be 
fixed indeed.

> That's why in the VMX code I did the call to msr_write_intercepted() (in
> __vmx_vcpu_run_flags) before calling vmx_vcpu_enter_exit().

Yes, it's easy to do the same and do it in svm_vcpu_run().

Paolo