[PATCH v3 4/4] ftrace: add tracepoint for itimer

[Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>]

Those tracepoints are wanted and useful:
1: We can detect a itimer's delay
2: We can monitor the lifecycle and behaviors of a itimer

Thus they help in analysing and debuging.

Great thanks to Thomas for giving me so many valuable advices.

There have two case: 
case 1, if we use ITIMER_REAL itimer:
Example ftrace output:
  main-1517  [000]   288.826521: hrtimer_start: timer=ce807544 func=it_real_fn expires=289327664443 ns softexpires=289327664443 ns
  main-1517  [000]   288.826524: itimer_state: which=0 expires=0 it_value=0.500 it_interval=0.0
  main-1517  [000]   288.827313: hrtimer_expire: timer=ce807544 now=289327958161 ns
  main-1517  [000]   288.827314: hrtimer_cancel: timer=ce807544
  main-1517  [000]   288.827315: itimer_expire: pid=1517 which=0 now=0
  main-1517  [000]   288.827319: hrtimer_callback_done: timer=ce807544
  main-1517  [000]   288.827463: itimer_state: which=0 expires=0 it_value=0.0 it_interval=0.0

ITIMER_REAL itime use hrtimer to calculate time, so we can cooperate with
hrtimer's tracepoint to get it's delay.

from this output information, the process set up a ITIMER_REAL
iteimr(which=0), use a hrtime which address is ce807544, we can get the
delay is 289327958161-289327664443=293718 ns

case 2, if we use ITIMER_VIRTUAL/ITIMER_PROF itimer:
Example ftrace output:
  main-5409  [001] 12763.874835: itimer_state: which=1 expires=2 it_value=0.500 it_interval=0.500
  main-5409  [001] 12763.883495: itimer_expire: pid=5409 which=1 now=2
  main-5409  [001] 12763.883712: itimer_state: which=1 expires=0 it_value=0.0 it_interval=0.0
  
from this, we can know the itimer delay is 2-2=0 jiffies.

Changelog:
v1->v2:
Remove ktime_to_ns() in TP_fast_assign()
v2->v3:
Below changes are all base on Thomas's suggestion:
1: Remove ITIMER_REAL's timer address and 'state' from trace_itimer_state()
2: Don't save task's name in trace_itimer_state()
3: Save task's pid instead of task's name in trace_itimer_expire()

Signed-off-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
---
 include/trace/events/timer.h |   52 ++++++++++++++++++++++++++++++++++++++++++
 kernel/itimer.c              |    5 +++-
 kernel/posix-cpu-timers.c    |    3 ++
 3 files changed, 59 insertions(+), 1 deletions(-)

diff --git a/include/trace/events/timer.h b/include/trace/events/timer.h
index 8b22b12..0a77f34 100644
--- a/include/trace/events/timer.h
+++ b/include/trace/events/timer.h
@@ -237,6 +237,58 @@ TRACE_EVENT(hrtimer_cancel,
 	TP_printk("timer=%p", __entry->timer)
 );
 
+TRACE_EVENT(itimer_state,
+
+	TP_PROTO(int which, struct itimerval *value, cputime_t expires),
+
+	TP_ARGS(which, value, expires),
+
+	TP_STRUCT__entry(
+		__field(	int,		which		)
+		__field(	cputime_t,	expires		)
+		__field(	long,		value_sec	)
+		__field(	long,		value_usec	)
+		__field(	long,		interval_sec	)
+		__field(	long,		interval_usec	)
+	),
+
+	TP_fast_assign(
+		__entry->which		= which;
+		__entry->expires	= expires;
+		__entry->value_sec	= value->it_value.tv_sec;
+		__entry->value_usec	= value->it_value.tv_usec;
+		__entry->interval_sec	= value->it_interval.tv_sec;
+		__entry->interval_usec	= value->it_interval.tv_usec;
+	),
+
+	TP_printk("which=%d expires=%lu it_value=%lu.%lu it_interval=%lu.%lu",
+		  __entry->which, __entry->expires,
+		  __entry->value_sec, __entry->value_usec,
+		  __entry->interval_sec, __entry->interval_usec)
+);
+
+TRACE_EVENT(itimer_expire,
+
+	TP_PROTO(int which, struct pid *pid, cputime_t now),
+
+	TP_ARGS(which, pid, now),
+
+	TP_STRUCT__entry(
+		__field( int ,		which	)
+		__field( int,		pid	)
+		__field( cputime_t,	now	)
+	),
+
+	TP_fast_assign(
+		__entry->which	= which;
+		__entry->now	= now;
+		__entry->pid	= pid_vnr(pid);
+	),
+
+	TP_printk("pid=%d which=%d now=%lu", __entry->pid, __entry->which,
+		  __entry->now)
+);
+
 #endif /*  _TRACE_TIMER_H */
 
 /* This part must be outside protection */
diff --git a/kernel/itimer.c b/kernel/itimer.c
index 58762f7..7e9ee89 100644
--- a/kernel/itimer.c
+++ b/kernel/itimer.c
@@ -12,6 +12,7 @@
 #include <linux/time.h>
 #include <linux/posix-timers.h>
 #include <linux/hrtimer.h>
+#include <trace/events/timer.h>
 
 #include <asm/uaccess.h>
 
@@ -123,6 +124,7 @@ enum hrtimer_restart it_real_fn(struct hrtimer *timer)
 	struct signal_struct *sig =
 		container_of(timer, struct signal_struct, real_timer);
 
+	trace_itimer_expire(ITIMER_REAL, sig->leader_pid, 0);
 	kill_pid_info(SIGALRM, SEND_SIG_PRIV, sig->leader_pid);
 
 	return HRTIMER_NORESTART;
@@ -139,7 +141,7 @@ int do_setitimer(int which, struct itimerval *value, struct itimerval *ovalue)
 	struct task_struct *tsk = current;
 	struct hrtimer *timer;
 	ktime_t expires;
-	cputime_t cval, cinterval, nval, ninterval;
+	cputime_t cval, cinterval, nval = 0, ninterval;
 
 	/*
 	 * Validate the timevals in value.
@@ -220,6 +222,7 @@ again:
 	default:
 		return -EINVAL;
 	}
+	trace_itimer_state(which, value, nval);
 	return 0;
 }
 
diff --git a/kernel/posix-cpu-timers.c b/kernel/posix-cpu-timers.c
index bece7c0..63e93bf 100644
--- a/kernel/posix-cpu-timers.c
+++ b/kernel/posix-cpu-timers.c
@@ -8,6 +8,7 @@
 #include <linux/math64.h>
 #include <asm/uaccess.h>
 #include <linux/kernel_stat.h>
+#include <trace/events/timer.h>
 
 /*
  * Called after updating RLIMIT_CPU to set timer expiration if necessary.
@@ -1160,6 +1161,7 @@ static void check_process_timers(struct task_struct *tsk,
 				sig->it_prof_expires = cputime_add(
 					sig->it_prof_expires, ptime);
 			}
+			trace_itimer_expire(ITIMER_PROF, sig->leader_pid, ptime);
 			__group_send_sig_info(SIGPROF, SEND_SIG_PRIV, tsk);
 		}
 		if (!cputime_eq(sig->it_prof_expires, cputime_zero) &&
@@ -1176,6 +1178,7 @@ static void check_process_timers(struct task_struct *tsk,
 				sig->it_virt_expires = cputime_add(
 					sig->it_virt_expires, utime);
 			}
+			trace_itimer_expire(ITIMER_VIRTUAL, sig->leader_pid, utime);
 			__group_send_sig_info(SIGVTALRM, SEND_SIG_PRIV, tsk);
 		}
 		if (!cputime_eq(sig->it_virt_expires, cputime_zero) &&
-- 
1.6.1.2