RE: [PATCH] x86: Prefer MWAIT over HALT on AMD processors

From: Thomas Gleixner
Date: Wed Apr 06 2022 - 12:29:35 EST


On Tue, Apr 05 2022 at 20:40, Mario Limonciello wrote:

> [Public]

Really useful information that your post is public while some of the
earlier posts in this _public_ thread were marked '[AMD confidential]'.

>> >> This seem _bit_ at odds with the commit message (and the AMD SSBD
>> >> whitepaper):
>> >>
>> >>> Add the necessary synchronization logic for AMD family 17H.
>> >> So, is X86_FEATURE_ZEN for family==0x17, or family>=0x17?
>> > There are Zen family CPUs and APUs from family 0x19. Perhaps at the
>> > time of the whitepaper there weren't yet though.
>> Is this a gap in the documentation, then? Is there some documentation
>> of the availability of SSBD mitigations on family 0x19?
> It looks like a misinterpretation of the document.

Not at all. The document does not mention family 19h at all. So where is
the misinterpretation?

Dave was asking for documentation for family 0x19, right?

> Notice it mentions in Non-architectural MSRs explicitly:
> "some models of family 17h have logic that allow loads to bypass older stores
> but lack the architectural SPEC_CTRL or VIRT_SPEC_CTR"

That's relevant to Dave's question in which way?

> But that is not for all family 17h nor for family 19h. You can see earlier in
> the document the method to detect presence for SSBD which applies to the
> rest of 17h and 19h.

We know how to read this document. But this document does not mention
anything else than family 17h. So what are you talking about?

> That code in amd_set_core_ssb_state is only used for one of the
> mitigation codepaths without MSR support, not for all Zen CPUs.

Again, how is that relevant to the legitimate question whether that
document applies to family 17h only or to 17h+ which includes 19h?

We need to make a decison about what X86_FEATURE_ZEN means. Is it that
hard to give an authoritive answer?