]> bbs.cooldavid.org Git - net-next-2.6.git/commitdiff
workqueue: Add basic tracepoints to track workqueue execution
authorArjan van de Ven <arjan@linux.intel.com>
Sat, 21 Aug 2010 20:07:26 +0000 (13:07 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 21 Aug 2010 20:19:37 +0000 (13:19 -0700)
With the introduction of the new unified work queue thread pools,
we lost one feature: It's no longer possible to know which worker
is causing the CPU to wake out of idle. The result is that PowerTOP
now reports a lot of "kworker/a:b" instead of more readable results.

This patch adds a pair of tracepoints to the new workqueue code,
similar in style to the timer/hrtimer tracepoints.

With this pair of tracepoints, the next PowerTOP can correctly
report which work item caused the wakeup (and how long it took):

Interrupt (43)            i915      time   3.51ms    wakeups 141
Work      ieee80211_iface_work      time   0.81ms    wakeups  29
Work              do_dbs_timer      time   0.55ms    wakeups  24
Process                   Xorg      time  21.36ms    wakeups   4
Timer    sched_rt_period_timer      time   0.01ms    wakeups   1

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/trace/events/workqueue.h [new file with mode: 0644]
kernel/workqueue.c

diff --git a/include/trace/events/workqueue.h b/include/trace/events/workqueue.h
new file mode 100644 (file)
index 0000000..49682d7
--- /dev/null
@@ -0,0 +1,62 @@
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM workqueue
+
+#if !defined(_TRACE_WORKQUEUE_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_WORKQUEUE_H
+
+#include <linux/tracepoint.h>
+#include <linux/workqueue.h>
+
+/**
+ * workqueue_execute_start - called immediately before the workqueue callback
+ * @work:      pointer to struct work_struct
+ *
+ * Allows to track workqueue execution.
+ */
+TRACE_EVENT(workqueue_execute_start,
+
+       TP_PROTO(struct work_struct *work),
+
+       TP_ARGS(work),
+
+       TP_STRUCT__entry(
+               __field( void *,        work    )
+               __field( void *,        function)
+       ),
+
+       TP_fast_assign(
+               __entry->work           = work;
+               __entry->function       = work->func;
+       ),
+
+       TP_printk("work struct %p: function %pf", __entry->work, __entry->function)
+);
+
+/**
+ * workqueue_execute_end - called immediately before the workqueue callback
+ * @work:      pointer to struct work_struct
+ *
+ * Allows to track workqueue execution.
+ */
+TRACE_EVENT(workqueue_execute_end,
+
+       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)
+);
+
+
+#endif /*  _TRACE_WORKQUEUE_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>
index 2994a0e3a61cc2cfdeb382936394456875a4c01c..8bd600c020e5cdf5f2454681bc2e89fcc99c47d0 100644 (file)
@@ -35,6 +35,9 @@
 #include <linux/lockdep.h>
 #include <linux/idr.h>
 
+#define CREATE_TRACE_POINTS
+#include <trace/events/workqueue.h>
+
 #include "workqueue_sched.h"
 
 enum {
@@ -1790,7 +1793,13 @@ static void process_one_work(struct worker *worker, struct work_struct *work)
        work_clear_pending(work);
        lock_map_acquire(&cwq->wq->lockdep_map);
        lock_map_acquire(&lockdep_map);
+       trace_workqueue_execute_start(work);
        f(work);
+       /*
+        * While we must be careful to not use "work" after this, the trace
+        * point will only record its address.
+        */
+       trace_workqueue_execute_end(work);
        lock_map_release(&lockdep_map);
        lock_map_release(&cwq->wq->lockdep_map);