linux-kernel - Re: [PATCH v9 2/6] KVM: x86: Virtualize CR4.LAM

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:   Wed, 7 Jun 2023 12:55:27 +0800
From:   Binbin Wu <binbin.wu@...ux.intel.com>
To:     "Huang, Kai" <kai.huang@...el.com>
Cc:     "kvm@...r.kernel.org" <kvm@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "robert.hu@...ux.intel.com" <robert.hu@...ux.intel.com>,
        "pbonzini@...hat.com" <pbonzini@...hat.com>,
        "Christopherson,, Sean" <seanjc@...gle.com>,
        "Gao, Chao" <chao.gao@...el.com>,
        "David.Laight@...LAB.COM" <David.Laight@...LAB.COM>
Subject: Re: [PATCH v9 2/6] KVM: x86: Virtualize CR4.LAM_SUP



On 6/7/2023 11:40 AM, Huang, Kai wrote:
> On Tue, 2023-06-06 at 17:18 +0800, Binbin Wu wrote:
>> Move CR4.LAM_SUP out of CR4_RESERVED_BITS and its reservation depends on vcpu
>> supporting LAM feature or not. Leave the bit intercepted to avoid vmread every
>> time when KVM fetches its value, with the expectation that guest won't toggle
>> the bit frequently.
> KVM only needs to do vmread once to cache guest's CR4, and presumable vmread is
> a lot cheaper than a VMEXIT.  So I don't see the value of intercepting it if
> there's no need to do.
Here is the discussion about the general rule of interception of CR4 bit.
Sean mentioned:  "As a base
rule, KVM intercepts CR4 bits unless there's a reason not to, e.g. if 
the CR4 bit
in question is written frequently by real guests and/or never consumed 
by KVM."
https://lore.kernel.org/all/Y7xA53sLxCwzfvgD@google.com/

And CR4.LAM_SUP value will be used to determin the LAM mode when apply 
LAM masking in instruction emulations / VMExit handlers,
and if the bit is passed-through, it will be a vmread in these pathes.

>
> But presumably I think we cannot allow guest to own this bit because KVM wants
> to return a valid CR4 if LAM isn't exposed to guest?  Otherwise guest can still
> set this bit even LAM isn't exposed to guest.
>
> Am I missing something?
Right, this is also a reason why the CR4.LAM_SUP bit should be intercepted.
Will update the justification.
I suppose this reason is enough for justification, will remove the 
performance part in changelog.

Thanks.
>
> If not, your justification of intercepting this bit isn't correct and needs
> update.
>