Re: [PATCH v3 01/13] x86/retpoline: Add initial retpoline support

[Paul Turner]

On Thu, Jan 04, 2018 at 10:40:23AM -0800, Andi Kleen wrote:
> > Clearly Paul's approach to retpoline without lfence is faster.
> > I'm guessing it wasn't shared with amazon/intel until now and
> > this set of patches going to adopt it, right?
> > 
> > Paul, could you share a link to a set of alternative gcc patches
> > that do retpoline similar to llvm diff ?
> 
> I don't think it's a good idea to use any sequence not signed
> off by CPU designers and extensively tested. 

I can confirm that "pause; jmp" has been previously reviewed by your side.

That said, again, per the other email, once we have guaranteed that speculative
execution will reach this point, beyond the guarantee that it does something
"safe" the choice of sequence here is a performance detail rather than
correctness.

> 
> While another one may work for most tests, it could always fail in
> some corner case.
> 
> Right now we have the more heavy weight one and I would
> suggest to stay with that one for now. Then worry
> about more optimizations later.
> 
> Correctness first.
> 
> -Andi