Commit-ID: 9261dd0da6c6432f08670719069449c6efe4f7a9
Gitweb: http://git.kernel.org/tip/9261dd0da6c6432f08670719069449c6efe4f7a9
Author: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
AuthorDate: Wed, 14 Aug 2013 16:24:26 -0700
Committer: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
CommitDate: Mon, 23 Sep 2013 09:18:13 -0700
rcu: Add tracing for rcuo no-CBs CPU wakeup handshake
Lost wakeups from call_rcu() to the rcuo kthreads can result in hangs
that are difficult to diagnose. This commit therefore adds tracing to
help pin down the cause of these hangs.
Reported-by: Clark Williams <williams@xxxxxxxxxx>
Reported-by: Carsten Emde <C.Emde@xxxxxxxxx>
Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
[ paulmck: Add const per kbuild test robot's advice. ]
---
include/trace/events/rcu.h | 37 +++++++++++++++++++++++++++++++++++++
kernel/rcutree_plugin.h | 14 +++++++++++++-
2 files changed, 50 insertions(+), 1 deletion(-)
diff --git a/include/trace/events/rcu.h b/include/trace/events/rcu.h
index 98466c6..4301cd9 100644
--- a/include/trace/events/rcu.h
+++ b/include/trace/events/rcu.h
@@ -172,6 +172,42 @@ TRACE_EVENT(rcu_grace_period_init,
);
/*
+ * Tracepoint for RCU no-CBs CPU callback handoffs. This event is intended
+ * to assist debugging of these handoffs.
+ *
+ * The first argument is the name of the RCU flavor, and the second is
+ * the number of the offloaded CPU are extracted. The third and final
+ * argument is a string as follows:
+ *
+ * "WakeEmpty": Wake rcuo kthread, first CB to empty list.
+ * "WakeOvf": Wake rcuo kthread, CB list is huge.
+ * "WakeNot": Don't wake rcuo kthread.
+ * "WakeNotPoll": Don't wake rcuo kthread because it is polling.
+ * "WokeEmpty": rcuo kthread woke to find empty list.
+ * "WokeNonEmpty": rcuo kthread woke to find non-empty list.
+ */
+TRACE_EVENT(rcu_nocb_wake,
+
+ TP_PROTO(const char *rcuname, int cpu, const char *reason),
+
+ TP_ARGS(rcuname, cpu, reason),
+
+ TP_STRUCT__entry(
+ __field(const char *, rcuname)
+ __field(int, cpu)
+ __field(const char *, reason)
+ ),
+
+ TP_fast_assign(
+ __entry->rcuname = rcuname;
+ __entry->cpu = cpu;
+ __entry->reason = reason;
+ ),
+
+ TP_printk("%s %d %s", __entry->rcuname, __entry->cpu, __entry->reason)
+);
+
+/*
* Tracepoint for tasks blocking within preemptible-RCU read-side
* critical sections. Track the type of RCU (which one day might
* include SRCU), the grace-period number that the task is blocking
@@ -667,6 +703,7 @@ TRACE_EVENT(rcu_barrier,
#define trace_rcu_future_grace_period(rcuname, gpnum, completed, c, \
level, grplo, grphi, event) \
do { } while (0)
+#define trace_rcu_nocb_wake(rcuname, cpu, reason) do { } while (0)
#define trace_rcu_preempt_task(rcuname, pid, gpnum) do { } while (0)
#define trace_rcu_unlock_preempted_task(rcuname, gpnum, pid) do { } while (0)
#define trace_rcu_quiescent_state_report(rcuname, gpnum, mask, qsmask, level, \
diff --git a/kernel/rcutree_plugin.h b/kernel/rcutree_plugin.h
index 130c97b..f4ed24b 100644
--- a/kernel/rcutree_plugin.h
+++ b/kernel/rcutree_plugin.h
@@ -2108,15 +2108,22 @@ static void __call_rcu_nocb_enqueue(struct rcu_data *rdp,
/* If we are not being polled and there is a kthread, awaken it ... */
t = ACCESS_ONCE(rdp->nocb_kthread);
- if (rcu_nocb_poll | !t)
+ if (rcu_nocb_poll | !t) {
+ trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
+ TPS("WakeNotPoll"));
return;
+ }
len = atomic_long_read(&rdp->nocb_q_count);
if (old_rhpp == &rdp->nocb_head) {
wake_up(&rdp->nocb_wq); /* ... only if queue was empty ... */
rdp->qlen_last_fqs_check = 0;
+ trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeEmpty"));
} else if (len > rdp->qlen_last_fqs_check + qhimark) {
wake_up_process(t); /* ... or if many callbacks queued. */
rdp->qlen_last_fqs_check = LONG_MAX / 2;
+ trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeOvf"));
+ } else {
+ trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeNot"));
}
return;
}
@@ -2233,10 +2240,15 @@ static int rcu_nocb_kthread(void *arg)
wait_event_interruptible(rdp->nocb_wq, rdp->nocb_head);
list = ACCESS_ONCE(rdp->nocb_head);
if (!list) {
+ if (!rcu_nocb_poll)
+ trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
+ TPS("WokeEmpty"));
schedule_timeout_interruptible(1);
flush_signals(current);
continue;
}
+ trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
+ TPS("WokeNonEmpty"));
/*
* Extract queued callbacks, update counts, and wait
--
To unsubscribe from this list: send the line "unsubscribe linux-tip-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
|