[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <86f5e43d-e2fd-6cce-70b8-cdf01b7ebed4@amd.com>
Date: Fri, 11 Sep 2020 16:44:09 -0500
From: Tom Lendacky <thomas.lendacky@....com>
To: Borislav Petkov <bp@...en8.de>,
Krish Sadhukhan <krish.sadhukhan@...cle.com>
Cc: kvm@...r.kernel.org, pbonzini@...hat.com, jmattson@...gle.com,
tglx@...utronix.de, mingo@...hat.com, x86@...nel.org,
sean.j.christopherson@...el.com, vkuznets@...hat.com,
wanpengli@...cent.com, joro@...tes.org,
dave.hansen@...ux.intel.com, luto@...nel.org, peterz@...radead.org,
linux-kernel@...r.kernel.org, hpa@...or.com
Subject: Re: [PATCH 2/4 v3] x86: AMD: Add hardware-enforced cache coherency as
a CPUID feature
On 9/11/20 4:33 PM, Borislav Petkov wrote:
> + Tom.
>
> On Fri, Sep 11, 2020 at 07:25:59PM +0000, Krish Sadhukhan wrote:
>> +#define X86_FEATURE_HW_CACHE_COHERENCY (11*32+ 7) /* AMD hardware-enforced cache coherency */
>
> so before you guys paint the bikeshed all kinds of colors :), Tom (CCed)
> is digging out the official name. (If it is even uglier, we might keep
> on bikeshedding...).
I believe the official name is something like CoherencyEnforced. Since
it's under the 0x8000001f leaf (AMD Secure Encryption), it means that
coherency is enforced between the same physical address when referenced
with or without the encryption bit (bare metal or in a guest).
So I kind of like the X86_FEATURE_SME_COHERENT suggestion by Dave, even if
it also applies to SEV.
Thanks,
Tom
>
> Once you have that, add the "" after the comment - like
> X86_FEATURE_FENCE_SWAPGS_USER, for example, so that it doesn't show in
> /proc/cpuinfo as luserspace doesn't care about hw coherency between enc
> memory.
>
> Thx.
>
Powered by blists - more mailing lists