[PATCH 2/3] workqueue: Add workqueue_mayday and workqueue_rescued tracepoints

[Aaron Tomlin]

When a worker pool fails to create a new worker thread within
MAYDAY_INTERVAL (e.g., typically under severe memory pressure where
memory reclaim paths depend on pending work items), send_mayday()
signals distress to the workqueue's rescuer thread. The rescuer then
takes over processing the pending work items via assign_rescuer_work().

While pwq->stats[PWQ_STAT_MAYDAY] and pwq->stats[PWQ_STAT_RESCUED] track
these occurrences cumulatively, there is currently no event-driven
mechanism to observe exactly when mayday distress occurs or which work
items require rescue.

Add two new tracepoints namely workqueue_mayday and workqueue_rescued to
make distress and rescuer execution easily observable (e.g., via Ftrace
or eBPF).

Signed-off-by: Aaron Tomlin <atomlin@xxxxxxxxxxx>
---
 include/trace/events/workqueue.h | 68 ++++++++++++++++++++++++++++++++
 kernel/workqueue.c               |  2 +
 2 files changed, 70 insertions(+)

diff --git a/include/trace/events/workqueue.h b/include/trace/events/workqueue.h
index ef0f3a4b73a8..013cfa472f6d 100644
--- a/include/trace/events/workqueue.h
+++ b/include/trace/events/workqueue.h
@@ -165,6 +165,74 @@ TRACE_EVENT(workqueue_cpu_intensive,
 		  __entry->cpu, __entry->duration_us)
 );
 
+/**
+ * workqueue_mayday - called when a pool_workqueue sends mayday to rescuer
+ * @pwq:	pointer to struct pool_workqueue
+ *
+ * This event occurs when a worker pool fails to create a new worker
+ * within MAYDAY_INTERVAL and requests the workqueue's rescuer thread to
+ * process pending works.
+ */
+TRACE_EVENT(workqueue_mayday,
+
+	TP_PROTO(struct pool_workqueue *pwq),
+
+	TP_ARGS(pwq),
+
+	TP_STRUCT__entry(
+		__string( workqueue,	pwq->wq->name	)
+		__field( int,		pool_id		)
+		__field( int,		cpu		)
+		__field( int,		nr_active	)
+	),
+
+	TP_fast_assign(
+		__assign_str(workqueue);
+		__entry->pool_id	= pwq->pool->id;
+		__entry->cpu		= pwq->pool->cpu;
+		__entry->nr_active	= pwq->nr_active;
+	),
+
+	TP_printk("workqueue=%s pool_id=%d cpu=%d nr_active=%d",
+		  __get_str(workqueue), __entry->pool_id, __entry->cpu,
+		  __entry->nr_active)
+);
+
+/**
+ * workqueue_rescued - called when a work item is assigned to a rescuer
+ * @pwq:	pointer to struct pool_workqueue
+ * @work:	pointer to struct work_struct
+ * @function:	pointer to worker function
+ *
+ * This event occurs when a work item is claimed by a rescuer thread
+ * to guarantee forward progress.
+ */
+TRACE_EVENT(workqueue_rescued,
+
+	TP_PROTO(struct pool_workqueue *pwq, struct work_struct *work,
+		 work_func_t function),
+
+	TP_ARGS(pwq, work, function),
+
+	TP_STRUCT__entry(
+		__field( void *,	work		)
+		__field( void *,	function	)
+		__string( workqueue,	pwq->wq->name	)
+		__field( int,		cpu		)
+	),
+
+	TP_fast_assign(
+		__entry->work		= work;
+		__entry->function	= function;
+		__assign_str(workqueue);
+		__entry->cpu		= pwq->pool->cpu;
+	),
+
+	TP_printk("work struct=%p function=%ps workqueue=%s cpu=%d",
+		  __entry->work, __entry->function, __get_str(workqueue),
+		  __entry->cpu)
+);
+
 #endif /*  _TRACE_WORKQUEUE_H */
 
 /* This part must be outside protection */
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 6ee644155b82..6f6fe2068389 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -3122,6 +3122,7 @@ static void send_mayday(struct pool_workqueue *pwq)
 		list_add_tail(&pwq->mayday_node, &wq->maydays);
 		wake_up_process(wq->rescuer->task);
 		pwq->stats[PWQ_STAT_MAYDAY]++;
+		trace_workqueue_mayday(pwq);
 	}
 }
 
@@ -3619,6 +3620,7 @@ static bool assign_rescuer_work(struct pool_workqueue *pwq, struct worker *rescu
 	list_for_each_entry_safe_from(work, n, &pool->worklist, entry) {
 		if (get_work_pwq(work) == pwq && assign_work(work, rescuer, &n)) {
 			pwq->stats[PWQ_STAT_RESCUED]++;
+			trace_workqueue_rescued(pwq, work, work->func);
 			/* put the cursor for next search */
 			list_move_tail(&cursor->entry, &n->entry);
 			return true;
-- 
2.55.0