On Wed, Jul 8, 2020 at 1:02 AM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> On Tue, 7 Jul 2020 18:21:17 +0900
> Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:
>
> > ccflags-remove-$(CONFIG_FUNCTION_TRACER) += $(CC_FLAGS_FTRACE)
> >
> > exists here in sub-directories of lib/ to keep the behavior of
> > commit 2464a609ded0 ("ftrace: do not trace library functions").
> >
> > Since that commit, not only the objects in lib/ but also the ones in
> > the sub-directories are excluded from ftrace (although the commit
> > description did not explicitly mention this).
> >
> > However, most of library functions in sub-directories are not so hot.
> > Re-add them to ftrace.
>
> I'm OK with this change, but note, it wasn't just the hot path that I
> disabled ftrace on lib for, but some of these calls are done very early
> at boot up. It may have been PowerPC that I was stumbling over. The
> issue is that they would call mcount before the kernel was mapped
> properly, and the system would crash.
>
> My PowerPC box no longer boots so I can't test this anymore. But a lot
> has changed since 2008, and all this may very well be OK.
>
> -- Steve
That's why I split this into two commits
so that we can do git-bisect and
revert the second one in case of a regression.
Anyway, we have some more time to test this in linux-next
(and somebody reports an issue, if any).
--
Best Regards
Masahiro Yamada
