On Fri, 2018-01-05 at 13:56 +0000, Woodhouse, David wrote:
> 
> At some point during this whole painful mess, I had come to the
> conclusion that having relocations in altinstr didn't work, and that's
> why I had X86_xx_NO_RETPOLINE instead of X86_xx_RETPOLINE. I now think
> that something else was wrong when I was testing that, and relocs in
> altinstr do work. So sure, X86_FEATURE_RETPOLINE ought to work. I can
> change that round, and it's simpler for the IBRS patch set to take it
> into account and set it when appropriate.

+bpetkov

Nope, alternatives are broken. Only a jmp as the *first* opcode of
altinstr gets handled by recompute_jump(), while any subsequent insn is
just copied untouched.

To fix that and handle every instruction, the alternative code would
need to know about instruction lengths. I think we need to stick with
the inverted X86_FEATURE_NO_RETPOLINE flag for the moment, and not tie
it to a complex bugfix there.