Re: [edk2-devel] [PATCH 0/2] UefiCpuPkg/Library: Fix bug in MpInitLib

["Duran, Leo" <leo.duran@amd.com>]

> -----Original Message-----
> From: Ni, Ray [mailto:ray.ni@intel.com]
> Sent: Wednesday, February 26, 2020 2:57 AM
> To: Laszlo Ersek <lersek@redhat.com>; devel@edk2.groups.io; Duran, Leo
> <leo.duran@amd.com>; Wu, Hao A <hao.a.wu@intel.com>; Fu, Siyuan
> <siyuan.fu@intel.com>
> Cc: Dong, Eric <eric.dong@intel.com>
> Subject: RE: [edk2-devel] [PATCH 0/2] UefiCpuPkg/Library: Fix bug in
> MpInitLib
> 
> Leo,
> 
> > > BTW, reading the PlatformId MSR was already being done by
> > > MicrocodeDetect(), but it never affected AMD-based platforms as the
> > > flow never gets that far, since the Detect routine bails out early when it
> finds the size of the patch is zero.
> 
> You are saying that PlatformId MSR access is not performed by CPU in old
> code because of the zero size uCode.
> But now with Hao or Siyuan's change, the PlatformId MSR access is always
> performed even when there is no uCode. It sounds like a regression to
> optimization to me.
> Did you evaluate the path to avoid accessing PlatformID MSR when uCode
> doesn't exist? So that the API to detect AMD processor is not needed at all.
[Duran, Leo] 
Hi Ray,
I think your summary is pretty accurate, except that I'd say that avoiding a READ from the PlatformId MSR
should happen solely based on the fact that the MSR simply does not exist on AMD processors.
Then as a result of that,  the usage of the PlatformId (as it relates to microcode or anything else) must then be dealt with separately.

To that end, I think I covered all cases where the MSR is being read, and also where PlatformId is being used.
(I also added comments for each case)

Thanks,
Leo.

> 
> Thanks,
> Ray