[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <aLtNJZo2o/i9AyqX@intel.com>
Date: Sat, 6 Sep 2025 04:50:45 +0800
From: Chao Gao <chao.gao@...el.com>
To: John Allen <john.allen@....com>
CC: <kvm@...r.kernel.org>, <linux-kernel@...r.kernel.org>, <x86@...nel.org>,
<seanjc@...gle.com>, <pbonzini@...hat.com>, <dave.hansen@...el.com>,
<rick.p.edgecombe@...el.com>, <mlevitsk@...hat.com>,
<weijiang.yang@...el.com>, <bp@...en8.de>, <dave.hansen@...ux.intel.com>,
<hpa@...or.com>, <mingo@...hat.com>, <tglx@...utronix.de>,
<thomas.lendacky@....com>
Subject: Re: [PATCH v3 5/5] KVM: SVM: Enable shadow stack virtualization for
SVM
>> >- /* KVM doesn't yet support CET virtualization for SVM. */
>> >- kvm_cpu_cap_clear(X86_FEATURE_SHSTK);
>> >- kvm_cpu_cap_clear(X86_FEATURE_IBT);
>>
>> IIUC, IBT should be cleared because KVM doesn't support IBT for SVM.
>
>Yeah, I wondered about this. The reason I chose to not clear this is
>because we don't explicitly clear other features that are not supported
>on AMD hardware AFAICT.
Your series doesn't enable IBT for SVM. If future AMD CPUs add IBT support,
this KVM running on those CPUs will inadvertently advertise IBT support.
>Is there a reason we should clear this and not
>other unsupported features?
I think they should be cleared if they require any KVM enabling beyond just
adding the CPUID bits. At the very least, we can handle IBT correctly.
Powered by blists - more mailing lists