The gp_seq value can come from either of rdp, rsp or rnp. This makes the
rcu_grace_period tracepoint extremely confusing since once if left
constantly applying effort to reason about what a gp_seq means.
Only the rsp is the global source of truth (most accurate GP info). The
rnp can be off by ~1 and the rdp can be off by way more. Add some more
context to traces to clarify where it comes from.
While this can be resolved using some trace post-processing, such post-
processing does need extra maintenance of scripts. Just make it easy by
adding more context in the trace itself.
Signed-off-by: Joel Fernandes (Google) <[email protected]>
---
include/trace/events/rcu.h | 12 ++++++++----
kernel/rcu/tree.c | 32 ++++++++++++++++----------------
kernel/rcu/tree_plugin.h | 4 ++--
3 files changed, 26 insertions(+), 22 deletions(-)
diff --git a/include/trace/events/rcu.h b/include/trace/events/rcu.h
index 155b5cb43cfd3..bf2d9f2e25f85 100644
--- a/include/trace/events/rcu.h
+++ b/include/trace/events/rcu.h
@@ -68,24 +68,28 @@ TRACE_EVENT(rcu_utilization,
*/
TRACE_EVENT_RCU(rcu_grace_period,
- TP_PROTO(const char *rcuname, unsigned long gp_seq, const char *gpevent),
+ TP_PROTO(const char *rcuname, const char *gp_seq_src,
+ unsigned long gp_seq, const char *gpevent),
- TP_ARGS(rcuname, gp_seq, gpevent),
+ TP_ARGS(rcuname, gp_seq_src, gp_seq, gpevent),
TP_STRUCT__entry(
__field(const char *, rcuname)
+ __field(const char *, gp_seq_src)
__field(long, gp_seq)
__field(const char *, gpevent)
),
TP_fast_assign(
__entry->rcuname = rcuname;
+ __entry->gp_seq_src = gp_seq_src;
__entry->gp_seq = (long)gp_seq;
__entry->gpevent = gpevent;
),
- TP_printk("%s %ld %s",
- __entry->rcuname, __entry->gp_seq, __entry->gpevent)
+ TP_printk("%s %s_gp_seq=%ld %s",
+ __entry->rcuname, __entry->gp_seq_src,
+ __entry->gp_seq, __entry->gpevent)
);
/*
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index d0988a1c1079d..b625811f529fa 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -1367,7 +1367,7 @@ static bool rcu_start_this_gp(struct rcu_node *rnp_start, struct rcu_data *rdp,
trace_rcu_this_gp(rnp, rdp, gp_seq_req, TPS("NoGPkthread"));
goto unlock_out;
}
- trace_rcu_grace_period(rcu_state.name, data_race(rcu_state.gp_seq), TPS("newreq"));
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), data_race(rcu_state.gp_seq), TPS("newreq"));
ret = true; /* Caller must wake GP kthread. */
unlock_out:
/* Push furthest requested GP to leaf node and rcu_data structure. */
@@ -1464,9 +1464,9 @@ static bool rcu_accelerate_cbs(struct rcu_node *rnp, struct rcu_data *rdp)
/* Trace depending on how much we were able to accelerate. */
if (rcu_segcblist_restempty(&rdp->cblist, RCU_WAIT_TAIL))
- trace_rcu_grace_period(rcu_state.name, gp_seq_req, TPS("AccWaitCB"));
+ trace_rcu_grace_period(rcu_state.name, TPS("req"), gp_seq_req, TPS("AccWaitCB"));
else
- trace_rcu_grace_period(rcu_state.name, gp_seq_req, TPS("AccReadyCB"));
+ trace_rcu_grace_period(rcu_state.name, TPS("req"), gp_seq_req, TPS("AccReadyCB"));
return ret;
}
@@ -1566,7 +1566,7 @@ static bool __note_gp_changes(struct rcu_node *rnp, struct rcu_data *rdp)
if (!offloaded)
ret = rcu_advance_cbs(rnp, rdp); /* Advance CBs. */
rdp->core_needs_qs = false;
- trace_rcu_grace_period(rcu_state.name, rdp->gp_seq, TPS("cpuend"));
+ trace_rcu_grace_period(rcu_state.name, TPS("rdp"), rdp->gp_seq, TPS("cpuend"));
} else {
if (!offloaded)
ret = rcu_accelerate_cbs(rnp, rdp); /* Recent CBs. */
@@ -1582,7 +1582,7 @@ static bool __note_gp_changes(struct rcu_node *rnp, struct rcu_data *rdp)
* set up to detect a quiescent state, otherwise don't
* go looking for one.
*/
- trace_rcu_grace_period(rcu_state.name, rnp->gp_seq, TPS("cpustart"));
+ trace_rcu_grace_period(rcu_state.name, TPS("rnp"), rnp->gp_seq, TPS("cpustart"));
need_qs = !!(rnp->qsmask & rdp->grpmask);
rdp->cpu_no_qs.b.norm = need_qs;
rdp->core_needs_qs = need_qs;
@@ -1683,7 +1683,7 @@ static bool rcu_gp_init(void)
/* Record GP times before starting GP, hence rcu_seq_start(). */
rcu_seq_start(&rcu_state.gp_seq);
ASSERT_EXCLUSIVE_WRITER(rcu_state.gp_seq);
- trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq, TPS("start"));
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq, TPS("start"));
raw_spin_unlock_irq_rcu_node(rnp);
/*
@@ -1851,7 +1851,7 @@ static void rcu_gp_fqs_loop(void)
WRITE_ONCE(rcu_state.jiffies_kick_kthreads,
jiffies + (j ? 3 * j : 2));
}
- trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
TPS("fqswait"));
rcu_state.gp_state = RCU_GP_WAIT_FQS;
ret = swait_event_idle_timeout_exclusive(
@@ -1866,7 +1866,7 @@ static void rcu_gp_fqs_loop(void)
/* If time for quiescent-state forcing, do it. */
if (!time_after(rcu_state.jiffies_force_qs, jiffies) ||
(gf & RCU_GP_FLAG_FQS)) {
- trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
TPS("fqsstart"));
rcu_gp_fqs(first_gp_fqs);
gf = 0;
@@ -1874,7 +1874,7 @@ static void rcu_gp_fqs_loop(void)
first_gp_fqs = false;
gf = rcu_state.cbovld ? RCU_GP_FLAG_OVLD : 0;
}
- trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
TPS("fqsend"));
cond_resched_tasks_rcu_qs();
WRITE_ONCE(rcu_state.gp_activity, jiffies);
@@ -1885,7 +1885,7 @@ static void rcu_gp_fqs_loop(void)
cond_resched_tasks_rcu_qs();
WRITE_ONCE(rcu_state.gp_activity, jiffies);
WARN_ON(signal_pending(current));
- trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
TPS("fqswaitsig"));
ret = 1; /* Keep old FQS timing. */
j = jiffies;
@@ -1968,7 +1968,7 @@ static void rcu_gp_cleanup(void)
raw_spin_lock_irq_rcu_node(rnp); /* GP before ->gp_seq update. */
/* Declare grace period done, trace first to use old GP number. */
- trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq, TPS("end"));
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq, TPS("end"));
rcu_seq_end(&rcu_state.gp_seq);
ASSERT_EXCLUSIVE_WRITER(rcu_state.gp_seq);
rcu_state.gp_state = RCU_GP_IDLE;
@@ -1985,7 +1985,7 @@ static void rcu_gp_cleanup(void)
if ((offloaded || !rcu_accelerate_cbs(rnp, rdp)) && needgp) {
WRITE_ONCE(rcu_state.gp_flags, RCU_GP_FLAG_INIT);
WRITE_ONCE(rcu_state.gp_req_activity, jiffies);
- trace_rcu_grace_period(rcu_state.name,
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"),
rcu_state.gp_seq,
TPS("newreq"));
} else {
@@ -2005,7 +2005,7 @@ static int __noreturn rcu_gp_kthread(void *unused)
/* Handle grace-period start. */
for (;;) {
- trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
TPS("reqwait"));
rcu_state.gp_state = RCU_GP_WAIT_GPS;
swait_event_idle_exclusive(rcu_state.gp_wq,
@@ -2019,7 +2019,7 @@ static int __noreturn rcu_gp_kthread(void *unused)
cond_resched_tasks_rcu_qs();
WRITE_ONCE(rcu_state.gp_activity, jiffies);
WARN_ON(signal_pending(current));
- trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
TPS("reqwaitsig"));
}
@@ -2263,7 +2263,7 @@ int rcutree_dying_cpu(unsigned int cpu)
return 0;
blkd = !!(rnp->qsmask & rdp->grpmask);
- trace_rcu_grace_period(rcu_state.name, READ_ONCE(rnp->gp_seq),
+ trace_rcu_grace_period(rcu_state.name, TPS("rsp"), READ_ONCE(rnp->gp_seq),
blkd ? TPS("cpuofl") : TPS("cpuofl-bgp"));
return 0;
}
@@ -3875,7 +3875,7 @@ int rcutree_prepare_cpu(unsigned int cpu)
rdp->core_needs_qs = false;
rdp->rcu_iw_pending = false;
rdp->rcu_iw_gp_seq = rdp->gp_seq - 1;
- trace_rcu_grace_period(rcu_state.name, rdp->gp_seq, TPS("cpuonl"));
+ trace_rcu_grace_period(rcu_state.name, TPS("rdp"), rdp->gp_seq, TPS("cpuonl"));
raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
rcu_prepare_kthreads(cpu);
rcu_spawn_cpu_nocb_kthread(cpu);
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index 982fc5be52698..7df8cd968f7fd 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -262,7 +262,7 @@ static void rcu_qs(void)
{
RCU_LOCKDEP_WARN(preemptible(), "rcu_qs() invoked with preemption enabled!!!\n");
if (__this_cpu_read(rcu_data.cpu_no_qs.s)) {
- trace_rcu_grace_period(TPS("rcu_preempt"),
+ trace_rcu_grace_period(TPS("rcu_preempt"), TPS("rdp"),
__this_cpu_read(rcu_data.gp_seq),
TPS("cpuqs"));
__this_cpu_write(rcu_data.cpu_no_qs.b.norm, false);
@@ -788,7 +788,7 @@ static void rcu_qs(void)
RCU_LOCKDEP_WARN(preemptible(), "rcu_qs() invoked with preemption enabled!!!");
if (!__this_cpu_read(rcu_data.cpu_no_qs.s))
return;
- trace_rcu_grace_period(TPS("rcu_sched"),
+ trace_rcu_grace_period(TPS("rcu_sched"), TPS("rdp"),
__this_cpu_read(rcu_data.gp_seq), TPS("cpuqs"));
__this_cpu_write(rcu_data.cpu_no_qs.b.norm, false);
if (!__this_cpu_read(rcu_data.cpu_no_qs.b.exp))
--
2.27.0.111.gc72c7da667-goog
On Thu, Jun 18, 2020 at 09:36:41PM -0400, Joel Fernandes (Google) wrote:
[...]
> @@ -2019,7 +2019,7 @@ static int __noreturn rcu_gp_kthread(void *unused)
> cond_resched_tasks_rcu_qs();
> WRITE_ONCE(rcu_state.gp_activity, jiffies);
> WARN_ON(signal_pending(current));
> - trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
> + trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
> TPS("reqwaitsig"));
> }
>
> @@ -2263,7 +2263,7 @@ int rcutree_dying_cpu(unsigned int cpu)
> return 0;
>
> blkd = !!(rnp->qsmask & rdp->grpmask);
> - trace_rcu_grace_period(rcu_state.name, READ_ONCE(rnp->gp_seq),
> + trace_rcu_grace_period(rcu_state.name, TPS("rsp"), READ_ONCE(rnp->gp_seq),
This should be: TPS("rnp") :-(
Happy to fix it up and resend if you'd like. Thanks!
On Thu, Jun 18, 2020 at 10:07:18PM -0400, Joel Fernandes wrote:
> On Thu, Jun 18, 2020 at 09:36:41PM -0400, Joel Fernandes (Google) wrote:
> [...]
> > @@ -2019,7 +2019,7 @@ static int __noreturn rcu_gp_kthread(void *unused)
> > cond_resched_tasks_rcu_qs();
> > WRITE_ONCE(rcu_state.gp_activity, jiffies);
> > WARN_ON(signal_pending(current));
> > - trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
> > + trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
> > TPS("reqwaitsig"));
> > }
> >
> > @@ -2263,7 +2263,7 @@ int rcutree_dying_cpu(unsigned int cpu)
> > return 0;
> >
> > blkd = !!(rnp->qsmask & rdp->grpmask);
> > - trace_rcu_grace_period(rcu_state.name, READ_ONCE(rnp->gp_seq),
> > + trace_rcu_grace_period(rcu_state.name, TPS("rsp"), READ_ONCE(rnp->gp_seq),
>
> This should be: TPS("rnp") :-(
>
> Happy to fix it up and resend if you'd like. Thanks!
I queued and pushed 1/2 and 2/2, but again, I am still not at all
convinced by 3/3. If you want to make RCU trace output human
readable, post-processing will be needed.
Thanx, Paul
On Fri, Jun 19, 2020 at 10:40:01AM -0700, Paul E. McKenney wrote:
> On Thu, Jun 18, 2020 at 10:07:18PM -0400, Joel Fernandes wrote:
> > On Thu, Jun 18, 2020 at 09:36:41PM -0400, Joel Fernandes (Google) wrote:
> > [...]
> > > @@ -2019,7 +2019,7 @@ static int __noreturn rcu_gp_kthread(void *unused)
> > > cond_resched_tasks_rcu_qs();
> > > WRITE_ONCE(rcu_state.gp_activity, jiffies);
> > > WARN_ON(signal_pending(current));
> > > - trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
> > > + trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
> > > TPS("reqwaitsig"));
> > > }
> > >
> > > @@ -2263,7 +2263,7 @@ int rcutree_dying_cpu(unsigned int cpu)
> > > return 0;
> > >
> > > blkd = !!(rnp->qsmask & rdp->grpmask);
> > > - trace_rcu_grace_period(rcu_state.name, READ_ONCE(rnp->gp_seq),
> > > + trace_rcu_grace_period(rcu_state.name, TPS("rsp"), READ_ONCE(rnp->gp_seq),
> >
> > This should be: TPS("rnp") :-(
> >
> > Happy to fix it up and resend if you'd like. Thanks!
>
> I queued and pushed 1/2 and 2/2.
Thanks!
> but again, I am still not at all
> convinced by 3/3. If you want to make RCU trace output human
> readable, post-processing will be needed.
Or I could post-process the code before building it since the pattern seems
easy to parse ;-)
- Joel
On Sat, Jun 20, 2020 at 01:36:08AM -0400, Joel Fernandes wrote:
> On Fri, Jun 19, 2020 at 10:40:01AM -0700, Paul E. McKenney wrote:
> > On Thu, Jun 18, 2020 at 10:07:18PM -0400, Joel Fernandes wrote:
> > > On Thu, Jun 18, 2020 at 09:36:41PM -0400, Joel Fernandes (Google) wrote:
> > > [...]
> > > > @@ -2019,7 +2019,7 @@ static int __noreturn rcu_gp_kthread(void *unused)
> > > > cond_resched_tasks_rcu_qs();
> > > > WRITE_ONCE(rcu_state.gp_activity, jiffies);
> > > > WARN_ON(signal_pending(current));
> > > > - trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq,
> > > > + trace_rcu_grace_period(rcu_state.name, TPS("rsp"), rcu_state.gp_seq,
> > > > TPS("reqwaitsig"));
> > > > }
> > > >
> > > > @@ -2263,7 +2263,7 @@ int rcutree_dying_cpu(unsigned int cpu)
> > > > return 0;
> > > >
> > > > blkd = !!(rnp->qsmask & rdp->grpmask);
> > > > - trace_rcu_grace_period(rcu_state.name, READ_ONCE(rnp->gp_seq),
> > > > + trace_rcu_grace_period(rcu_state.name, TPS("rsp"), READ_ONCE(rnp->gp_seq),
> > >
> > > This should be: TPS("rnp") :-(
> > >
> > > Happy to fix it up and resend if you'd like. Thanks!
> >
> > I queued and pushed 1/2 and 2/2.
>
> Thanks!
>
> > but again, I am still not at all
> > convinced by 3/3. If you want to make RCU trace output human
> > readable, post-processing will be needed.
>
> Or I could post-process the code before building it since the pattern seems
> easy to parse ;-)
For this one thing, perhaps. For most other information of interest,
doing so in-kernel would not be so good, for example, from a
lock-contention viewpoint.
Thanx, Paul