Re: [RFC 05/10] x86/speculation: Add basic IBRS support infrastructure

[David Woodhouse]

On Wed, 2018-01-24 at 13:29 +0100, Peter Zijlstra wrote:
> 
> Yes :/
> 
> We could look at extending x86_cpu_id and x86_match_cpu with a stepping
> option I suppose, but that might be lots of churn.

That goes all the way to mod_deviceinfo, and would be horrid.

We could add an x86_match_cpu_stepping() function, I suppose? But I'm
mostly trying to avoid depending on other stuff like that, for patches
which are going to need to be backported to all the stable kernels.

I'd much rather do it this way and then if we see another use case for
it (that commit you mentioned could be nicer, I suppose), consolidate
into a single stepping-capable lookup function in a later "cleanup".

Attachment: smime.p7s
Description: S/MIME cryptographic signature