linux-kernel - Re: [PATCH] selftests: KVM: AMD Nested SVM test infrastructure

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]

Date:   Tue, 21 Jan 2020 13:50:07 +0100
From:   Auger Eric <eric.auger@...hat.com>
To:     Paolo Bonzini <pbonzini@...hat.com>,
        Vitaly Kuznetsov <vkuznets@...hat.com>
Cc:     thuth@...hat.com, drjones@...hat.com, eric.auger.pro@...il.com,
        linux-kernel@...r.kernel.org, kvm@...r.kernel.org
Subject: Re: [PATCH] selftests: KVM: AMD Nested SVM test infrastructure

Hi Vitaly, Paolo,

On 1/21/20 1:17 PM, Paolo Bonzini wrote:
> On 21/01/20 12:12, Auger Eric wrote:
>>>> +
>>>> +static struct test tests[] = {
>>>> +	/* name, supported, custom setup, l2 code, exit code, custom check, finished */
>>>> +	{"vmmcall", NULL, NULL, l2_vmcall, SVM_EXIT_VMMCALL},
>>>> +	{"vmrun", NULL, NULL, l2_vmrun, SVM_EXIT_VMRUN},
>>>> +	{"CR3 read intercept", NULL, prepare_cr3_intercept, l2_cr3_read, SVM_EXIT_READ_CR3},
>>>> +};
>>> selftests are usualy not that well structured :-) E.g. we don't have
>>> sub-tests and a way to specify which one to run so there is a single
>>> flow when everything is being executed. I'd suggest to keep things as
>>> simple as possibe (especially in the basic 'svm' test).
>> In this case the differences between the tests is very tiny. One line on
>> L2 and one line on L1 to check the exit status. I wondered whether it
>> deserves to have separate test files for that. I did not intend to run
>> the subtests separately nor to add many more subtests but rather saw all
>> of them as a single basic test. More complex tests would be definitively
>> separate.
> 
> I would just leave this deeper kind of test to kvm-unit-tests and keep
> selftests for API tests.  So this would mean basically only keep (and
> inline) the vmmcall test.

OK this makes sense. I implemented those 3 basic tests as a proof of
concept but this definitively overlaps with kvm-unit-tests coverage. I
will focus on new tests and leverage the kselftest framework instead.

Thanks

Eric
> 
> Paolo
>