Re: [PATCH] ftrace: fix current_tracer error return

["Frédéric Weisbecker" <fweisbec@gmail.com>]

2008/10/28 Steven Rostedt <rostedt@goodmis.org>:
>
> On Tue, 28 Oct 2008, Fr?d?ric Weisbecker wrote:
>
>> 2008/10/28 Steven Rostedt <rostedt@goodmis.org>:
>> >
>> > On Tue, 28 Oct 2008, Fr?d?ric Weisbecker wrote:
>> >>
>> >> An that corrects the bug when I failed to assign the sched_switch tracer.
>> >> Thanks Steven!
>> >>
>> >> Tested-by: Frederic Weisbecker <fweisbec@gmail.com>
>> >
>> > Do you still have issues with the function tracer in the latest linux-tip?
>> > If so, can you rebuild it from scratch, send me the config again, and
>> > explain in detail what the issue is.
>> >
>> > Thanks,
>>
>> I think I've caught it. See the patch I just Cc'ed you:
>> ftrace: perform an initialization for ftrace to enable it
>
> Ah! I see the issue.
>
> My test scripts automatically set or unset /proc/sys/kernel/ftrace_enabled
>
> I believe I wanted that to be default off when dynamic ftrace was not
> active. Currently, it really shouldn't matter, because there's a check to
> make sure that ftrace has a registered function before calling it.
>
> That was just my preference. But if you find it less confusing, then
> having ftrace_enabled as default on even when dynamic ftrace is not
> enabled, is fine with me.
>
> -- Steve
>
>

Thanks. I guess this bug was there for several days. But since I
always used dynamic ftrace, I've not
seen it until yesterday. (BTW I didn't tested with dynamic since this
patch). I had one other bug that made the set of
function tracer blocking during the write on current_trace with
dynamic ftrace. I will have a look on it.

Actually one thing that makes me a bit confused is the use of these
two vars: ftrace_enabled and ftrace_disabled.
They are not used for the same purpose but their name suggest the same
role.... :-)