Browse Source

hung_task debugging: Add tracepoint to report the hang

Currently check_hung_task() prints a warning if it detects the
problem, but it is not convenient to watch the system logs if
user-space wants to be notified about the hang.

Add the new trace_sched_process_hang() into check_hung_task(),
this way a user-space monitor can easily wait for the hang and
potentially resolve a problem.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Dave Sullivan <dsulliva@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/20131019161828.GA7439@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Oleg Nesterov 11 years ago
parent
commit
6a716c90a5
2 changed files with 23 additions and 0 deletions
  1. 19 0
      include/trace/events/sched.h
  2. 4 0
      kernel/hung_task.c

+ 19 - 0
include/trace/events/sched.h

@@ -424,6 +424,25 @@ TRACE_EVENT(sched_pi_setprio,
 			__entry->oldprio, __entry->newprio)
 );
 
+#ifdef CONFIG_DETECT_HUNG_TASK
+TRACE_EVENT(sched_process_hang,
+	TP_PROTO(struct task_struct *tsk),
+	TP_ARGS(tsk),
+
+	TP_STRUCT__entry(
+		__array( char,	comm,	TASK_COMM_LEN	)
+		__field( pid_t,	pid			)
+	),
+
+	TP_fast_assign(
+		memcpy(__entry->comm, tsk->comm, TASK_COMM_LEN);
+		__entry->pid = tsk->pid;
+	),
+
+	TP_printk("comm=%s pid=%d", __entry->comm, __entry->pid)
+);
+#endif /* CONFIG_DETECT_HUNG_TASK */
+
 #endif /* _TRACE_SCHED_H */
 
 /* This part must be outside protection */

+ 4 - 0
kernel/hung_task.c

@@ -16,6 +16,7 @@
 #include <linux/export.h>
 #include <linux/sysctl.h>
 #include <linux/utsname.h>
+#include <trace/events/sched.h>
 
 /*
  * The number of tasks checked:
@@ -92,6 +93,9 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
 		t->last_switch_count = switch_count;
 		return;
 	}
+
+	trace_sched_process_hang(t);
+
 	if (!sysctl_hung_task_warnings)
 		return;
 	sysctl_hung_task_warnings--;