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@redhat.com>
Reported-by: Carsten Emde <C.Emde@osadl.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
[ paulmck: Add const per kbuild test robot's advice. ]
This commit is contained in:
Paul E. McKenney 2013-08-14 16:24:26 -07:00
parent bb311eccbd
commit 9261dd0da6
2 changed files with 50 additions and 1 deletions

View File

@ -171,6 +171,42 @@ TRACE_EVENT(rcu_grace_period_init,
__entry->grplo, __entry->grphi, __entry->qsmask)
);
/*
* 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
@ -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, \

View File

@ -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