linux-kernel - Re: [RFC 01/14] x86/apic: Add new driver for Secure AVIC

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: <d35bd29e-d00a-4da9-9ab3-1273ed1bf6c2@amd.com>
Date: Wed, 9 Oct 2024 19:20:54 +0530
From: Neeraj Upadhyay <Neeraj.Upadhyay@....com>
To: Tom Lendacky <thomas.lendacky@....com>, Borislav Petkov <bp@...en8.de>
Cc: linux-kernel@...r.kernel.org, tglx@...utronix.de, mingo@...hat.com,
 dave.hansen@...ux.intel.com, nikunj@....com, Santosh.Shukla@....com,
 Vasant.Hegde@....com, Suravee.Suthikulpanit@....com, David.Kaplan@....com,
 x86@...nel.org, hpa@...or.com, peterz@...radead.org, seanjc@...gle.com,
 pbonzini@...hat.com, kvm@...r.kernel.org
Subject: Re: [RFC 01/14] x86/apic: Add new driver for Secure AVIC



On 10/9/2024 6:45 PM, Tom Lendacky wrote:
> On 10/9/24 01:01, Neeraj Upadhyay wrote:
>>
>>
>> On 10/9/2024 10:53 AM, Borislav Petkov wrote:
>>> On Wed, Oct 09, 2024 at 07:26:55AM +0530, Neeraj Upadhyay wrote:
>>>> As SECURE_AVIC feature is not supported (as reported by snp_get_unsupported_features())
>>>> by guest at this patch in the series, it is added to SNP_FEATURES_IMPL_REQ here. The bit
>>>> value within SNP_FEATURES_IMPL_REQ hasn't changed with this change as the same bit pos
>>>> was part of MSR_AMD64_SNP_RESERVED_MASK before this patch. In patch 14 SECURE_AVIC guest
>>>> support is indicated by guest.
>>>
>>> So what's the point of adding it to SNP_FEATURES_IMPL_REQ here? What does that
>>> do at all in this patch alone? Why is this change needed in here?
>>>
>>
>> Before this patch, if hypervisor enables Secure AVIC  (reported in sev_status), guest would
>> terminate in snp_check_features(). The reason for this is, SNP_FEATURES_IMPL_REQ had the Secure
>> AVIC bit set before this patch, as that bit was part of MSR_AMD64_SNP_RESERVED_MASK 
>> GENMASK_ULL(63, 18).
>>
>> #define SNP_FEATURES_IMPL_REQ	(MSR_AMD64_SNP_VTOM |			\
>> 				 ...
>> 				 MSR_AMD64_SNP_RESERVED_MASK)
>>
>>
>>
>> Adding MSR_AMD64_SNP_SECURE_AVIC_BIT (bit 18) to SNP_FEATURES_IMPL_REQ in this patch
>> keeps that behavior intact as now with this change MSR_AMD64_SNP_RESERVED_MASK becomes
>> GENMASK_ULL(63, 19).
>>
>>
>>> IOW, why don't you do all the feature bit handling in the last patch, where it
>>> all belongs logically?
>>>
>>
>> If we do that, then hypervisor could have enabled Secure AVIC support and the guest
>> code at this patch won't catch the missing guest-support early and it can result in some
>> unknown failures at later point during guest boot.
> 
> Won't the SNP_RESERVED_MASK catch it? You are just renaming the bit
> position value, right? It was a 1 before and is still a 1. So the guest
> will terminate if the hypervisor sets the Secure AVIC bit both before
> and after this patch, right?
> 

Yes that is right. SNP_RESERVED_MASK catches it before this patch. My reply to Boris
above was for the case if we move setting of MSR_AMD64_SNP_SECURE_AVIC_ENABLED in
SNP_FEATURES_IMPL_REQ  from this patch to patch 14.


- Neeraj

> Thanks,
> Tom
> 
>>
>>
>> - Neeraj
>>
>>> In the last patch you can start *testing* for
>>> MSR_AMD64_SNP_SECURE_AVIC_ENABLED *and* enforce it with SNP_FEATURES_PRESENT.
>>>