2020-01-13 22:53:59

by Daniel Jordan

Subject: [PATCH 1/2] workqueue: add worker function to workqueue_execute_end tracepoint

It's surprising that workqueue_execute_end includes only the work when
its counterpart workqueue_execute_start has both the work and the worker
function.

You can't set a tracing filter or trigger based on the function, and
postprocessing scripts interested in specific functions are harder to
write since they have to remember the work from _start and match it up
with the same field in _end.

Add the function name, taking care to use the copy stashed in the
worker since the work is no longer safe to touch.

Signed-off-by: Daniel Jordan <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: Lai Jiangshan <[email protected]>
Cc: [email protected]
---
include/trace/events/workqueue.h | 19 ++++++++++++++++---
kernel/workqueue.c | 2 +-
2 files changed, 17 insertions(+), 4 deletions(-)

diff --git a/include/trace/events/workqueue.h b/include/trace/events/workqueue.h
index e172549283be..bdfc53e7f82f 100644
--- a/include/trace/events/workqueue.h
+++ b/include/trace/events/workqueue.h
@@ -108,14 +108,27 @@ TRACE_EVENT(workqueue_execute_start,
/**
* workqueue_execute_end - called immediately after the workqueue callback
* @work: pointer to struct work_struct
+ * @function: pointer to worker function
*
* Allows to track workqueue execution.
*/
-DEFINE_EVENT(workqueue_work, workqueue_execute_end,
+TRACE_EVENT(workqueue_execute_end,

- TP_PROTO(struct work_struct *work),
+ TP_PROTO(struct work_struct *work, work_func_t function),

- TP_ARGS(work)
+ TP_ARGS(work, function),
+
+ TP_STRUCT__entry(
+ __field( void *, work )
+ __field( void *, function)
+ ),
+
+ TP_fast_assign(
+ __entry->work = work;
+ __entry->function = function;
+ ),
+
+ TP_printk("work struct %p: function %ps", __entry->work, __entry->function)
);

#endif /* _TRACE_WORKQUEUE_H */
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index cfc923558e04..4bdfa270a37b 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -2266,7 +2266,7 @@ __acquires(&pool->lock)
* While we must be careful to not use "work" after this, the trace
* point will only record its address.
*/
- trace_workqueue_execute_end(work);
+ trace_workqueue_execute_end(work, worker->current_func);
lock_map_release(&lockdep_map);
lock_map_release(&pwq->wq->lockdep_map);

--
2.24.1

2020-01-13 22:54:15

by Daniel Jordan

Subject: [PATCH 2/2] workqueue: remove workqueue_work event class

The trace event class workqueue_work now has only one consumer, so get
rid of it. No functional change.

Signed-off-by: Daniel Jordan <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: Lai Jiangshan <[email protected]>
Cc: [email protected]
---
include/trace/events/workqueue.h | 31 ++++++++++++-------------------
1 file changed, 12 insertions(+), 19 deletions(-)

diff --git a/include/trace/events/workqueue.h b/include/trace/events/workqueue.h
index bdfc53e7f82f..9b8ae961acc5 100644
--- a/include/trace/events/workqueue.h
+++ b/include/trace/events/workqueue.h
@@ -8,23 +8,6 @@
#include <linux/tracepoint.h>
#include <linux/workqueue.h>

-DECLARE_EVENT_CLASS(workqueue_work,
-
- TP_PROTO(struct work_struct *work),
-
- TP_ARGS(work),
-
- TP_STRUCT__entry(
- __field( void *, work )
- ),
-
- TP_fast_assign(
- __entry->work = work;
- ),
-
- TP_printk("work struct %p", __entry->work)
-);
-
struct pool_workqueue;

/**
@@ -73,11 +56,21 @@ TRACE_EVENT(workqueue_queue_work,
* which happens immediately after queueing unless @max_active limit
* is reached.
*/
-DEFINE_EVENT(workqueue_work, workqueue_activate_work,
+TRACE_EVENT(workqueue_activate_work,

TP_PROTO(struct work_struct *work),

- TP_ARGS(work)
+ TP_ARGS(work),
+
+ TP_STRUCT__entry(
+ __field( void *, work )
+ ),
+
+ TP_fast_assign(
+ __entry->work = work;
+ ),
+
+ TP_printk("work struct %p", __entry->work)
);

/**
--
2.24.1

2020-01-15 16:05:02

by Tejun Heo

Subject: Re: [PATCH 2/2] workqueue: remove workqueue_work event class

On Mon, Jan 13, 2020 at 05:52:40PM -0500, Daniel Jordan wrote:
> The trace event class workqueue_work now has only one consumer, so get
> rid of it. No functional change.
>
> Signed-off-by: Daniel Jordan <[email protected]>
> Cc: Tejun Heo <[email protected]>
> Cc: Lai Jiangshan <[email protected]>
> Cc: [email protected]

Applied 1-2 to wq/for-5.6.

Thanks.

--
tejun