Re: [RFC PATCH] tracing/kprobe: Wait for disabling all running kprobe handlers

From: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>,
	Frederic Weisbecker <fweisbec@gmail.com>,
	Oleg Nesterov <oleg@redhat.com>,
	linux-kernel@vger.kernel.org, Ingo Molnar <mingo@redhat.com>,
	Arnaldo Carvalho de Melo <acme@ghostprotocols.net>,
	Andrew Morton <akpm@linux-foundation.org>,
	jovi.zhangwei@huawei.com, Jiri Olsa <jolsa@redhat.com>,
	Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Subject: Re: [RFC PATCH] tracing/kprobe: Wait for disabling all running kprobe handlers
Date: Tue, 09 Jul 2013 17:50:07 +0900	[thread overview]
Message-ID: <51DBCEBF.7010608@hitachi.com> (raw)
In-Reply-To: <20130709082149.GK25631@dyad.programming.kicks-ass.net>

(2013/07/09 17:21), Peter Zijlstra wrote:
> On Tue, Jul 09, 2013 at 05:20:17PM +0900, Masami Hiramatsu wrote:
>> (2013/07/09 17:07), Peter Zijlstra wrote:
>>> On Tue, Jul 09, 2013 at 05:01:45PM +0900, Masami Hiramatsu wrote:
>>>> +	if (wait) {
>>>> +		/*
>>>> +		 * synchronize with kprobe_trace_func/kretprobe_trace_func
>>>> +		 * to ensure disabled (all running handlers are finished)
>>>> +		 */
>>>> +		synchronize_sched();
>>>> +		kfree(link);	/* Ignored if link == NULL */
>>>> +	}
>>>
>>> What's not clear to me from this comment is if we're only waiting for kfree()?
>>> In that case shouldn't we use call_rcu() to free the thing? Or do we need the
>>> sync_sched() for other purposes as well?
>>
>> No, this is not only for kfree, but also to ensure completing
>> disabling process, because trace_remove_event_call() supposes
>> that for releasing event_call related objects (Those objects
>> will be accessed in the handlers).
> 
> Then may I kindly suggest you clarify this in the comment? :-)
> 

Ah, right! I'll update it :)

Thank you!

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