Re: [PATCH 1/2] x86/microcode/intel: Refresh the revisions that determine old_microcode

[Andrew Cooper]

On 19/08/2025 8:07 pm, Dave Hansen wrote:
> On 8/19/25 11:28, Pawan Gupta wrote:
>>> Alas, these deletions are documented as well as everything else is in
>>> the changelog...
>> Should this file reflect those deletions as well? As an example, if an
>> ancient part gets removed from the microcode repo, it may still be worth
>> for Linux to keep the record of its last microcode version.
> <sigh>
>
> There's even a "Removed Platforms" section in the microcode repo
> changelogs that gets copied and pasted for each release. But it's not
> consistently updated as platforms are removed.
>
> But, thanks to the magic of git, we can just look for the removed files:
>
> 06-55-06

CLX B0 stepping, notably gaining MSR_TSX_FORCE_ABORT rather than
MSR_TSX_CTRL.  It was a giant muddle for mitigations.

> 06-cf-01

EMR.  Stepping 2 still available.

> 06-8f-04
> 06-8f-05
> 06-8f-06

SPR.  Steppings 7 and 8 still available.

These are weird.  06-8f-08 has two blobs in it with disjoint platform
flags (this is fine).

However, all 3 blobs between -07 and -08 all have sigtables covering
everything back to stepping 4.

Arguably that's a bug for the intel form of the microcode (which is
indexed on stepping), but it's just some extraneous metadata.  The rest
of it is suggesting that SPR steppings 4-8 had compatible-enough
microcode for them all to be patched by the same blob.

> 06-ba-08

RPL. The repo has steppings 2 and 3, so I wonder what RPL stepping 8 is.

> 06-86-04
> 06-86-05

SNR.  All steppings dropped, although it's apparently a mobile
basestation only, so I guess it's not running Linux.

~Andrew