(From the original mail)

Steven Rostedt wrote:
quoted text
>
> Along came Gregory Haskins, who was bickering about having ftrace enabl=
ed
quoted text
> on a production -rt kernel. I told him the reasons that this would be b=
ad
quoted text
> and then he started thinking out loud, and suggesting wild ideas, like
> patching gcc!
>  =20
^^^^^^^^^^

Ingo Molnar wrote:
quoted text
> Eventually gcc should be extended to provide a separate section for=20
> instrumentation patch sites, instead of us having to disassemble the=20
> object code.=20

:)

I obviously agree with this, so +1

Though, tbh, at the time I suggested it I didn't think of Steve's idea=20
to post-process which was quite clever. But I do agree that having gcc=20
do it will probably save some build time since it will probably be=20
trivial for it to do this when already processing -pg. It would have the =

added benefit of letting the arch specific toolchain do the arch=20
specific work (though I think Steve's solution capitalizes on the=20
toolchain extensively as it is).

The biggest downside is that we would have an external dependency on gcc =

for the feature, but I guess the kernel already has some of those anyway =

(e.g. the stack overflow guard feature, etc). We could always fall back=20
on Steve's post-processing if the toolchain lacks the feature.

-Greg