Re: [RFC][PATCH 6/6] objtool,x86: Rewrite retpoline thunk calls

[Peter Zijlstra]

On Sat, Feb 20, 2021 at 01:39:20AM +0100, Borislav Petkov wrote:
> On Fri, Feb 19, 2021 at 11:01:58PM +0100, Peter Zijlstra wrote:
> > We could, but it so happens Joerg is also wanting negative features. 
> 
> Juergen.

Argh! I should stick to jgross. Sorry.

> > So I was thikning that perhaps we can convince Boris they're not
> > really all that aweful after all :-)
> 
> Well, I'm not crazy about this, TBH - I totally agree with Josh:
> 
> "with objtool generating code, it's going to be a maze to figure out
> where the generated code is coming from"
> 
> and without a real good reason to do this, what's the point? I know,
> because we can. :-)

Well:

 - straight line execution is always better than a round-trip to
   somewhere else, no matter how trivial.
 - supposely EIBRS (yeah, I know, there's a paper out there) should
   result in no longer using retpolines.
 - I really, as in _REALLY_ don't want to do a CET enabled retpoline
 - IOW, retpolines should be on their way out (knock on wood)
 - doing this was fun :-)
 - this stuff was mostly trivial make work stuff I could do with a head
   full of snot and a headache.
 - if we had negative alternatives objtool doesn't need to actually
   rewrite code in this case. It could simply emit alternative entries
   and call it a day.
 - objtool already rewrites code
 - I have more cases for objtool to rewrite code (I'll see if I can
   rebase and post that this weekend -- no promises).
 - also https://lkml.kernel.org/r/20200625200235.GQ4781@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx