Re: [PATCH 1/1] tracing, workqueuetrace: Make workqueue tracepoints use TRACE_EVENT macro

[Frederic Weisbecker <fweisbec@gmail.com>]

On Mon, Apr 13, 2009 at 10:44:27AM +0900, KOSAKI Motohiro wrote:
> Hi,
> 
> hehe, I also have similar patch in my local patch queue ;)
> 
> > +TRACE_EVENT(workqueue_insertion,
> > +
> > +	TP_PROTO(struct task_struct *wq_thread, struct work_struct *work),
> > +
> > +	TP_ARGS(wq_thread, work),
> > +
> > +	TP_STRUCT__entry(
> > +		__array(	char,			thread_comm,	TASK_COMM_LEN	)
> > +		__field(	pid_t,			thread_pid			)
> > +		__field(	struct work_struct *,	work				)
> > +		__field(	work_func_t,		func				)
> > +	),
> > +
> > +	TP_fast_assign(
> > +		memcpy(__entry->thread_comm, wq_thread->comm, TASK_COMM_LEN);
> > +		__entry->thread_pid	= wq_thread->pid;
> > +		__entry->work		= work;
> > +		__entry->func		= work->func;
> > +	),
> > +
> > +	TP_printk("thread=%s:%d work=%p func=%p", __entry->thread_comm,
> > +		__entry->thread_pid, __entry->work, __entry->func)
> > +);
> 
> please don't display raw kernel pointer.
> 
> work: unnecessary. it's internal information and func name provide enough
>       information.
> func: %pF can print pritty format



Ah, you already commented out it, sorry, I haven't seen you answer :)
(Reminder for me: always check the patches with threaded view).



> 
> 
> > +
> > +TRACE_EVENT(workqueue_execution,
> > +
> > +	TP_PROTO(struct task_struct *wq_thread, struct work_struct *work),
> > +
> > +	TP_ARGS(wq_thread, work),
> > +
> > +	TP_STRUCT__entry(
> > +		__array(	char,			thread_comm,	TASK_COMM_LEN	)
> > +		__field(	pid_t,			thread_pid			)
> > +		__field(	struct work_struct *,	work				)
> > +		__field(	work_func_t,		func				)
> > +	),
> > +
> > +	TP_fast_assign(
> > +		memcpy(__entry->thread_comm, wq_thread->comm, TASK_COMM_LEN);
> > +		__entry->thread_pid	= wq_thread->pid;
> > +		__entry->work		= work;
> > +		__entry->func		= work->func;
> > +	),
> > +
> > +	TP_printk("thread=%s:%d work=%p func=%p", __entry->thread_comm,
> > +		__entry->thread_pid, __entry->work, __entry->func)
> > +);
> 
> ditto.
> 
> and, I plan to rename this tracepoint to workqueue_handler_entry()
> and create workqueue_handler_exit().
>
> It's because stupid driver comsume many cputime in workqueue thread,
> it make delay to after work. We need mesure it.



Nice idea!

Frederic.

 
> 
> > +
> > +/* Trace the creation of one workqueue thread on a cpu */
> > +TRACE_EVENT(workqueue_creation,
> > +
> > +	TP_PROTO(struct task_struct *wq_thread, int cpu),
> > +
> > +	TP_ARGS(wq_thread, cpu),
> > +
> > +	TP_STRUCT__entry(
> > +		__array(	char,	thread_comm,	TASK_COMM_LEN	)
> > +		__field(	pid_t,	thread_pid			)
> > +		__field(	int,	cpu				)
> > +	),
> > +
> > +	TP_fast_assign(
> > +		memcpy(__entry->thread_comm, wq_thread->comm, TASK_COMM_LEN);
> > +		__entry->thread_pid	= wq_thread->pid;
> > +		__entry->cpu		= cpu;
> > +	),
> > +
> > +	TP_printk("thread=%s:%d cpu=%d", __entry->thread_comm,
> > +		__entry->thread_pid, __entry->cpu)
> > +);
> > +
> > +TRACE_EVENT(workqueue_destruction,
> > +
> > +	TP_PROTO(struct task_struct *wq_thread),
> > +
> > +	TP_ARGS(wq_thread),
> > +
> > +	TP_STRUCT__entry(
> > +		__array(	char,	thread_comm,	TASK_COMM_LEN	)
> > +		__field(	pid_t,	thread_pid			)
> > +	),
> > +
> > +	TP_fast_assign(
> > +		memcpy(__entry->thread_comm, wq_thread->comm, TASK_COMM_LEN);
> > +		__entry->thread_pid	= wq_thread->pid;
> > +	),
> > +
> > +	TP_printk("thread=%s:%d", __entry->thread_comm, __entry->thread_pid)
> > +);
> > +
> > +#undef TRACE_SYSTEM
> > -- 
> > 1.5.5.3
> > 
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > Please read the FAQ at  http://www.tux.org/lkml/
> 
> 
>