Re: [PATCH] stop_machine: Mark functions as notrace

From: Paul E. McKenney
Date: Wed Oct 21 2020 - 11:45:04 EST


On Wed, Oct 21, 2020 at 10:15:22AM -0400, Steven Rostedt wrote:
> On Wed, 21 Oct 2020 10:12:16 -0400
> Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> > > Fixes: 4ecf0a43e729 ("processor: get rid of cpu_relax_yield")
> > > Fixes: 366237e7b083 ("stop_machine: Provide RCU quiescent state in
> > > multi_cpu_stop()")
> >
> > I really do not like to add "notrace" to core functions because a single
> > architecture has issues with it. Why does RISCV have problems with these
> > functions but no other architecture does?
> >
> > NACK from me until it is shown that these are issues for a broader set of
> > architectures.
>
> After looking at the two above fixes, I take back my NACK ;-)
>
> One of them duplicates an already notraced function, so that looks fine.
> The other makes a static function global, which could cause issues as well.
>
> After further review:
>
> Acked-by: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>

If someone else would like to take this:

Acked-by: Paul E. McKenney <paulmck@xxxxxxxxxx>

Or let me know if you would like me to take it, target v5.11.

Thanx, Paul