On 2017/03/10 11:54AM, Steven Rostedt wrote:

On Fri, 10 Mar 2017 21:38:53 +0530
"Naveen N. Rao" [off-list ref] wrote:

quoted

On 2017/03/10 10:45AM, Steven Rostedt wrote:

quoted

On Thu, 02 Mar 2017 20:38:53 +1100
Michael Ellerman [off-list ref] wrote:

quoted

quoted

quoted

So if we drop that we're left with ftrace.S - which seems perfect to me.  

Yeah, I agree. But then there's the problem that ftrace.c and ftrace.S
will get the same ftrace.o. Maybe make it ftrace-hook.S ?  

I've avoided that issue by naming the files ftrace_32.S and ftrace_64.S 
(which gets further split up).

That's what I looked at doing for x86 as well. But not all archs have
32 / 64 splits. Should we look to have something that all archs can be
consistent with?

I don't have a strong opinion about this, but I feel that x86 can simply 
use ftrace_64.S, seeing as the current name is mcount_64.S.

Other architectures could do something similar too, or fall back to 
ftrace_hook.S. That way, all ftrace low-level code can simply be 
referred to as arch/*/ftrace_*.S

- Naveen