Re: [PATCH 02/11] [BUGFIX] ftrace, kprobes: Fix a deadlock on ftrace_regex_lock

[Masami Hiramatsu]

(2013/05/10 2:08), Steven Rostedt wrote:
> On Thu, 2013-05-09 at 12:34 -0400, Steven Rostedt wrote:
>> On Thu, 2013-05-09 at 12:27 -0400, Steven Rostedt wrote:
>>
>>> We probably should have a better way to initialize this. As there are 26
>>> ftrace_ops currently in the kernel (and this patch doesn't cover all of
>>> them). Maybe have the first time its registered to initialize it.

Oh, I missed many of them :(

>> Crap, but it can be used before that. Hmm, I guess all ftrace functions
>> will need to check that flag first. We do something similar for rt_mutex
>> in -rt.
> 
> I added this on top of your patch. I kept the INIT_REGEX_LOCK as it's
> only local to ftrace.c and wont spread further. Also, the
> ftrace_list_end ftrace_ops is just a place holder (needed for race
> conditions that can have function tracers call its stub), so it does not
> need to be initialized. If anything tries to grab its mutex, that's a
> bug anyway.
> 
> What do you think?

Hmm, would we really need to have the additional flag?
I mean, do we better force ftrace user to use ftrace_ops_init before
calling such functions as mutex itself does?

Thank you,

-- 
Masami HIRAMATSU
IT Management Research Dept. Linux Technology Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@xxxxxxxxxxx


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/