linux-kernel - Re: [PATCH v2 3/5] KVM: x86: Use common 'enable

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <aa5ef24e-6235-ad25-2f01-580efd2f1bbb@redhat.com>
Date:   Mon, 24 May 2021 19:02:18 +0200
From:   Paolo Bonzini <pbonzini@...hat.com>
To:     Sean Christopherson <seanjc@...gle.com>
Cc:     Vitaly Kuznetsov <vkuznets@...hat.com>, kvm@...r.kernel.org,
        Wanpeng Li <wanpengli@...cent.com>,
        Jim Mattson <jmattson@...gle.com>,
        Maxim Levitsky <mlevitsk@...hat.com>,
        Kechen Lu <kechenl@...dia.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 3/5] KVM: x86: Use common 'enable_apicv' variable for
 both APICv and AVIC

On 24/05/21 18:52, Sean Christopherson wrote:
> On Mon, May 24, 2021, Paolo Bonzini wrote:
>> On 18/05/21 22:39, Sean Christopherson wrote:
>>>> +/* enable / disable AVIC */
>>>> +static int avic;
>>>> +module_param(avic, int, 0444);
>>> We should opportunistically make avic a "bool".
>>>
>>
>> And also:
>>
>> diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
>> index 11714c22c9f1..48cb498ff070 100644
>> --- a/arch/x86/kvm/svm/svm.c
>> +++ b/arch/x86/kvm/svm/svm.c
>> @@ -185,9 +185,12 @@ module_param(vls, int, 0444);
>>   static int vgif = true;
>>   module_param(vgif, int, 0444);
>> -/* enable / disable AVIC */
>> -static int avic;
>> -module_param(avic, int, 0444);
>> +/*
>> + * enable / disable AVIC.  Because the defaults differ for APICv
>> + * support between VMX and SVM we cannot use module_param_named.
>> + */
>> +static bool avic;
>> +module_param(avic, bool, 0444);
>>   bool __read_mostly dump_invalid_vmcb;
>>   module_param(dump_invalid_vmcb, bool, 0644);
>> @@ -1013,11 +1016,7 @@ static __init int svm_hardware_setup(void)
>>   			nrips = false;
>>   	}
>> -	if (!npt_enabled || !boot_cpu_has(X86_FEATURE_AVIC))
>> -		avic = false;
>> -
>> -	/* 'enable_apicv' is common between VMX/SVM but the defaults differ */
>> -	enable_apicv = avic;
>> +	enable_apicv = avic && npt_enabled && boot_cpu_has(X86_FEATURE_AVIC);
>>   	if (enable_apicv) {
>>   		pr_info("AVIC enabled\n");
>>
>> The "if" can come back when AVIC is enabled by default.
> 
> But "avic" is connected to the module param, even if it's off by default its
> effective value should be reflected in sysfs.  E.g. the user may incorrectly
> think AVIC is in use if they set avic=1 but the CPU doesn't support AVIC.
> Forcing the user to check /proc/cpuinfo or look for "AVIC enabled" in dmesg is
> kludgy at best.

Indeed -- I even tested the above, but only before realizing that 
module_param_named would change the default.  So for now this needs to
be "enable_apicv = avic = ...", and later it can become just

	enable_apicv &= npt_enabled && boot_cpu_has(X86_FEATURE_AVIC);

Paolo