Re: [PATCH 10/13] tracing/uprobes: Fetch args before reserving a ring buffer

[Namhyung Kim <namhyung@kernel.org>]

Hi Steven,

On Thu, 22 Aug 2013 21:08:30 -0400, Steven Rostedt wrote:
> On Fri, 23 Aug 2013 07:57:15 +0800
> "zhangwei(Jovi)" <jovi.zhangwei@huawei.com> wrote:
>
>
>> > 
>> > What about creating a per cpu buffer when uprobes are registered, and
>> > delete them when they are finished? Basically what trace_printk() does
>> > if it detects that there are users of trace_printk() in the kernel.
>> > Note, it does not deallocate them when finished, as it is never
>> > finished until reboot ;-)
>> > 
>> > -- Steve
>> >
>> I also thought out this approach, but the issue is we cannot fetch user
>> memory into per-cpu buffer, because use per-cpu buffer should under
>> preempt disabled, and fetching user memory could sleep.
>
> Actually, we could create a per_cpu mutex to match the per_cpu buffers.
> This is not unlike what we do in -rt.
>
> 	int cpu;
> 	struct mutex *mutex;
> 	void *buf;
>
>
> 	/*
> 	 * Use per cpu buffers for fastest access, but we might migrate
> 	 * So the mutex makes sure we have sole access to it.
> 	 */
>
> 	cpu = raw_smp_processor_id();
> 	mutex = per_cpu(uprobe_cpu_mutex, cpu);
> 	buf = per_cpu(uprobe_cpu_buffer, cpu);
>
> 	mutex_lock(mutex);
> 	store_trace_args(..., buf,...);
> 	mutex_unlock(mutex);
>

Great!  I'll go with this approach.  Is it OK to you, Masami?

Thanks,
Namhyung