[PATCH 0/3] tracing: introduce TRACE_EVENT_NONE and use it

[PATCH 0/3] tracing: introduce TRACE_EVENT_NONE and use it

[Yafang Shao]

In this patchset, I introduce a new macro TRACE_EVENT_NONE(), which will
define a tracepoint as a do-nothing inline function.
	#define TRACE_EVENT_NONE(name, proto)			\
		static inline void trace_##name(proto)		\
		{ }						\
		static inline bool trace_##name##_enabled(void)	\
		{						\
			return false;				\
		}

Let's take some example for why this macro is needed.

- sched
The tracepoints trace_sched_stat_{iowait, blocked, wait, sleep} should
be not exposed to user if CONFIG_SCHEDSTATS is not set.
We can place #ifdef in the kernel/sched/fair.c to fix it, but we don't want
to sprinkle #ifdef.
So with TRACE_EVENT_NONE(), we can fix it in include/trace/events/sched.h.

- rcu
When CONFIG_RCU_TRACE is not set, some rcu tracepoints are define as
do-nothing macro without validating arguments, that is not proper.
We should validate the arguments.

Yafang Shao (3):
  tracing: introduce TRACE_EVENT_NONE()
  sched/fair: do not expose some tracepoints to user if
    CONFIG_SCHEDSTATS is not set
  rcu: validate arguments for rcu tracepoints

 include/linux/tracepoint.h   |  8 +++++
 include/trace/define_trace.h |  4 +++
 include/trace/events/rcu.h   | 84 ++++++++++++++++++++++++++++----------------
 include/trace/events/sched.h | 14 ++++++++
 kernel/rcu/rcu.h             |  9 ++---
 kernel/rcu/tree.c            |  8 ++---
 6 files changed, 86 insertions(+), 41 deletions(-)

-- 
1.8.3.1

[PATCH 1/3] tracing: introduce TRACE_EVENT_NONE()

[Yafang Shao]

Sometimes we want define a tracepoint as a do-nothing function.
So I introduce this TRACE_EVENT_NONE() for this kind of usage.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
---
 include/linux/tracepoint.h   | 8 ++++++++
 include/trace/define_trace.h | 4 ++++
 2 files changed, 12 insertions(+)

diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
index 9c31865..d00c5e6 100644
--- a/include/linux/tracepoint.h
+++ b/include/linux/tracepoint.h
@@ -548,4 +548,12 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p)
 
 #define TRACE_EVENT_PERF_PERM(event, expr...)
 
+#define TRACE_EVENT_NONE(name, proto)				\
+	static inline void trace_##name(proto)			\
+	{ }							\
+	static inline bool trace_##name##_enabled(void)		\
+	{							\
+		return false;					\
+	}
+
 #endif /* ifdef TRACE_EVENT (see note above) */
diff --git a/include/trace/define_trace.h b/include/trace/define_trace.h
index cb30c55..a833b6f 100644
--- a/include/trace/define_trace.h
+++ b/include/trace/define_trace.h
@@ -46,6 +46,9 @@
 		assign, print, reg, unreg)			\
 	DEFINE_TRACE_FN(name, reg, unreg)
 
+#undef TRACE_EVENT_NONE
+#define TRACE_EVENT_NONE(name, proto)
+
 #undef DEFINE_EVENT
 #define DEFINE_EVENT(template, name, proto, args) \
 	DEFINE_TRACE(name)
@@ -102,6 +105,7 @@
 #undef TRACE_EVENT_FN
 #undef TRACE_EVENT_FN_COND
 #undef TRACE_EVENT_CONDITION
+#undef TRACE_EVENT_NONE
 #undef DECLARE_EVENT_CLASS
 #undef DEFINE_EVENT
 #undef DEFINE_EVENT_FN
-- 
1.8.3.1

[PATCH 2/3] sched/fair: do not expose some tracepoints to user if CONFIG_SCHEDSTATS is not set

[Yafang Shao]

The tracepoints trace_sched_stat_{iowait, blocked, wait, sleep} should
be not exposed to user if CONFIG_SCHEDSTATS is not set.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
---
 include/trace/events/sched.h | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
index 9a4bdfa..4686c7f 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -336,6 +336,7 @@ static inline long __trace_sched_switch_state(bool preempt, struct task_struct *
 		  __entry->pid, __entry->old_pid)
 );
 
+#ifdef CONFIG_SCHEDSTATS
 /*
  * XXX the below sched_stat tracepoints only apply to SCHED_OTHER/BATCH/IDLE
  *     adding sched_stat support to SCHED_FIFO/RR would be welcome.
@@ -395,6 +396,19 @@ static inline long __trace_sched_switch_state(bool preempt, struct task_struct *
 	     TP_PROTO(struct task_struct *tsk, u64 delay),
 	     TP_ARGS(tsk, delay));
 
+#else /* CONFIG_SCHEDSTATS */
+
+TRACE_EVENT_NONE(sched_stat_wait,
+		 TP_PROTO(struct task_struct *tsk, u64 delay));
+TRACE_EVENT_NONE(sched_stat_sleep,
+		 TP_PROTO(struct task_struct *tsk, u64 delay));
+TRACE_EVENT_NONE(sched_stat_iowait,
+		 TP_PROTO(struct task_struct *tsk, u64 delay));
+TRACE_EVENT_NONE(sched_stat_blocked,
+		 TP_PROTO(struct task_struct *tsk, u64 delay));
+
+#endif /* CONFIG_SCHEDSTATS */
+
 /*
  * Tracepoint for accounting runtime (time the task is executing
  * on a CPU).
-- 
1.8.3.1

Re: [PATCH 2/3] sched/fair: do not expose some tracepoints to user if CONFIG_SCHEDSTATS is not set

[Steven Rostedt]

On Tue, 26 Mar 2019 09:40:06 +0800
Yafang Shao <laoar.shao@gmail.com> wrote:

> The tracepoints trace_sched_stat_{iowait, blocked, wait, sleep} should
> be not exposed to user if CONFIG_SCHEDSTATS is not set.
> 
> Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
> ---
>  include/trace/events/sched.h | 14 ++++++++++++++
>  1 file changed, 14 insertions(+)
> 
> diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
> index 9a4bdfa..4686c7f 100644
> --- a/include/trace/events/sched.h
> +++ b/include/trace/events/sched.h
> @@ -336,6 +336,7 @@ static inline long __trace_sched_switch_state(bool preempt, struct task_struct *
>  		  __entry->pid, __entry->old_pid)
>  );
>  
> +#ifdef CONFIG_SCHEDSTATS

Instead of doing this #ifdef here, perhaps we should have at the top of the file:

#ifdef CONFIG_SCHEDSTATS
# define TRACE_EVENT_SCHEDSTAT TRACE_EVENT
#else
# define TRACE_EVENT_SCHEDSTAT TRACE_EVENT_NOP
#endif

And then switch the sched stat tracepoints from TRACE_EVENT to
use TRACE_EVENT_SCHEDSTAT.

And if my magic macro hat is working while sitting on the shelf (don't
have time to put it on), then the TRACE_EVENT_SCHEDSTAT will work like
a TRACE_EVENT when enabled, and a TRACE_EVENT_NOP when not.

Can you give that a try?

If this works, you don't need to duplicate all the parameters again.

-- Steve

>  /*
>   * XXX the below sched_stat tracepoints only apply to
> SCHED_OTHER/BATCH/IDLE
>   *     adding sched_stat support to SCHED_FIFO/RR would be welcome.
> @@ -395,6 +396,19 @@ static inline long
> __trace_sched_switch_state(bool preempt, struct task_struct *
> TP_PROTO(struct task_struct *tsk, u64 delay), TP_ARGS(tsk, delay));
>  
> +#else /* CONFIG_SCHEDSTATS */
> +
> +TRACE_EVENT_NONE(sched_stat_wait,
> +		 TP_PROTO(struct task_struct *tsk, u64 delay));
> +TRACE_EVENT_NONE(sched_stat_sleep,
> +		 TP_PROTO(struct task_struct *tsk, u64 delay));
> +TRACE_EVENT_NONE(sched_stat_iowait,
> +		 TP_PROTO(struct task_struct *tsk, u64 delay));
> +TRACE_EVENT_NONE(sched_stat_blocked,
> +		 TP_PROTO(struct task_struct *tsk, u64 delay));
> +
> +#endif /* CONFIG_SCHEDSTATS */
> +
>  /*
>   * Tracepoint for accounting runtime (time the task is executing
>   * on a CPU).

Re: [PATCH 2/3] sched/fair: do not expose some tracepoints to user if CONFIG_SCHEDSTATS is not set

[Yafang Shao]

On Tue, Mar 26, 2019 at 10:21 AM Steven Rostedt <rostedt@goodmis.org> wrote:
>
> On Tue, 26 Mar 2019 09:40:06 +0800
> Yafang Shao <laoar.shao@gmail.com> wrote:
>
> > The tracepoints trace_sched_stat_{iowait, blocked, wait, sleep} should
> > be not exposed to user if CONFIG_SCHEDSTATS is not set.
> >
> > Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
> > ---
> >  include/trace/events/sched.h | 14 ++++++++++++++
> >  1 file changed, 14 insertions(+)
> >
> > diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
> > index 9a4bdfa..4686c7f 100644
> > --- a/include/trace/events/sched.h
> > +++ b/include/trace/events/sched.h
> > @@ -336,6 +336,7 @@ static inline long __trace_sched_switch_state(bool preempt, struct task_struct *
> >                 __entry->pid, __entry->old_pid)
> >  );
> >
> > +#ifdef CONFIG_SCHEDSTATS
>
> Instead of doing this #ifdef here, perhaps we should have at the top of the file:
>
> #ifdef CONFIG_SCHEDSTATS
> # define TRACE_EVENT_SCHEDSTAT TRACE_EVENT
> #else
> # define TRACE_EVENT_SCHEDSTAT TRACE_EVENT_NOP
> #endif
>
> And then switch the sched stat tracepoints from TRACE_EVENT to
> use TRACE_EVENT_SCHEDSTAT.
>
> And if my magic macro hat is working while sitting on the shelf (don't
> have time to put it on), then the TRACE_EVENT_SCHEDSTAT will work like
> a TRACE_EVENT when enabled, and a TRACE_EVENT_NOP when not.
>
> Can you give that a try?
>
> If this works, you don't need to duplicate all the parameters again.
>

You proposal seems like a better solution.
I will have a try.
Thanks for your suggestion.

-Yafang

>
> >  /*
> >   * XXX the below sched_stat tracepoints only apply to
> > SCHED_OTHER/BATCH/IDLE
> >   *     adding sched_stat support to SCHED_FIFO/RR would be welcome.
> > @@ -395,6 +396,19 @@ static inline long
> > __trace_sched_switch_state(bool preempt, struct task_struct *
> > TP_PROTO(struct task_struct *tsk, u64 delay), TP_ARGS(tsk, delay));
> >
> > +#else /* CONFIG_SCHEDSTATS */
> > +
> > +TRACE_EVENT_NONE(sched_stat_wait,
> > +              TP_PROTO(struct task_struct *tsk, u64 delay));
> > +TRACE_EVENT_NONE(sched_stat_sleep,
> > +              TP_PROTO(struct task_struct *tsk, u64 delay));
> > +TRACE_EVENT_NONE(sched_stat_iowait,
> > +              TP_PROTO(struct task_struct *tsk, u64 delay));
> > +TRACE_EVENT_NONE(sched_stat_blocked,
> > +              TP_PROTO(struct task_struct *tsk, u64 delay));
> > +
> > +#endif /* CONFIG_SCHEDSTATS */
> > +
> >  /*
> >   * Tracepoint for accounting runtime (time the task is executing
> >   * on a CPU).
>

[PATCH 3/3] rcu: validate arguments for rcu tracepoints

[Yafang Shao]

When CONFIG_RCU_TRACE is not set, all these tracepoints are define as
do-nothing macro.
We'd better make those inline functions that take proper arguments.

As RCU_TRACE() is defined as do-nothing marco as well when
CONFIG_RCU_TRACE is not set, so we can clean it up.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
---
 include/trace/events/rcu.h | 84 +++++++++++++++++++++++++++++-----------------
 kernel/rcu/rcu.h           |  9 ++---
 kernel/rcu/tree.c          |  8 ++---
 3 files changed, 60 insertions(+), 41 deletions(-)

diff --git a/include/trace/events/rcu.h b/include/trace/events/rcu.h
index f0c4d10..357e1b6 100644
--- a/include/trace/events/rcu.h
+++ b/include/trace/events/rcu.h
@@ -750,36 +750,60 @@
 
 #else /* #ifdef CONFIG_RCU_TRACE */
 
-#define trace_rcu_grace_period(rcuname, gp_seq, gpevent) do { } while (0)
-#define trace_rcu_future_grace_period(rcuname, gp_seq, gp_seq_req, \
-				      level, grplo, grphi, event) \
-				      do { } while (0)
-#define trace_rcu_grace_period_init(rcuname, gp_seq, level, grplo, grphi, \
-				    qsmask) do { } while (0)
-#define trace_rcu_exp_grace_period(rcuname, gqseq, gpevent) \
-	do { } while (0)
-#define trace_rcu_exp_funnel_lock(rcuname, level, grplo, grphi, gpevent) \
-	do { } while (0)
-#define trace_rcu_nocb_wake(rcuname, cpu, reason) do { } while (0)
-#define trace_rcu_preempt_task(rcuname, pid, gp_seq) do { } while (0)
-#define trace_rcu_unlock_preempted_task(rcuname, gp_seq, pid) do { } while (0)
-#define trace_rcu_quiescent_state_report(rcuname, gp_seq, mask, qsmask, level, \
-					 grplo, grphi, gp_tasks) do { } \
-	while (0)
-#define trace_rcu_fqs(rcuname, gp_seq, cpu, qsevent) do { } while (0)
-#define trace_rcu_dyntick(polarity, oldnesting, newnesting, dyntick) do { } while (0)
-#define trace_rcu_callback(rcuname, rhp, qlen_lazy, qlen) do { } while (0)
-#define trace_rcu_kfree_callback(rcuname, rhp, offset, qlen_lazy, qlen) \
-	do { } while (0)
-#define trace_rcu_batch_start(rcuname, qlen_lazy, qlen, blimit) \
-	do { } while (0)
-#define trace_rcu_invoke_callback(rcuname, rhp) do { } while (0)
-#define trace_rcu_invoke_kfree_callback(rcuname, rhp, offset) do { } while (0)
-#define trace_rcu_batch_end(rcuname, callbacks_invoked, cb, nr, iit, risk) \
-	do { } while (0)
-#define trace_rcu_torture_read(rcutorturename, rhp, secs, c_old, c) \
-	do { } while (0)
-#define trace_rcu_barrier(name, s, cpu, cnt, done) do { } while (0)
+TRACE_EVENT_NONE(rcu_grace_period,
+	TP_PROTO(const char *rcuname, unsigned long gp_seq,
+		 const char *gpevent));
+TRACE_EVENT_NONE(rcu_future_grace_period,
+	TP_PROTO(const char *rcuname, unsigned long gp_seq,
+		 unsigned long gp_seq_req, u8 level, int grplo,
+		 int grphi, const char *gpevent));
+TRACE_EVENT_NONE(rcu_grace_period_init,
+	TP_PROTO(const char *rcuname, unsigned long gp_seq, u8 level,
+		 int grplo, int grphi, unsigned long qsmask));
+TRACE_EVENT_NONE(rcu_exp_grace_period,
+	TP_PROTO(const char *rcuname, unsigned long gpseq,
+		 const char *gpevent));
+TRACE_EVENT_NONE(rcu_exp_funnel_lock,
+	TP_PROTO(const char *rcuname, u8 level, int grplo, int grphi,
+		 const char *gpevent));
+TRACE_EVENT_NONE(rcu_nocb_wake,
+	TP_PROTO(const char *rcuname, int cpu, const char *reason));
+TRACE_EVENT_NONE(rcu_preempt_task,
+	TP_PROTO(const char *rcuname, int pid, unsigned long gp_seq));
+TRACE_EVENT_NONE(rcu_unlock_preempted_task,
+	TP_PROTO(const char *rcuname, unsigned long gp_seq, int pid));
+TRACE_EVENT_NONE(rcu_quiescent_state_report,
+	TP_PROTO(const char *rcuname, unsigned long gp_seq,
+		 unsigned long mask, unsigned long qsmask,
+		 u8 level, int grplo, int grphi, int gp_tasks));
+TRACE_EVENT_NONE(rcu_fqs,
+	TP_PROTO(const char *rcuname, unsigned long gp_seq, int cpu,
+		 const char *qsevent));
+TRACE_EVENT_NONE(rcu_dyntick,
+	TP_PROTO(const char *polarity, long oldnesting, long newnesting,
+		 atomic_t dynticks));
+TRACE_EVENT_NONE(rcu_callback,
+	TP_PROTO(const char *rcuname, struct rcu_head *rhp, long qlen_lazy,
+		 long qlen));
+TRACE_EVENT_NONE(rcu_kfree_callback,
+	TP_PROTO(const char *rcuname, struct rcu_head *rhp,
+		 unsigned long offset, long qlen_lazy, long qlen));
+TRACE_EVENT_NONE(rcu_batch_start,
+	TP_PROTO(const char *rcuname, long qlen_lazy, long qlen, long blimit));
+TRACE_EVENT_NONE(rcu_invoke_callback,
+	TP_PROTO(const char *rcuname, struct rcu_head *rhp));
+TRACE_EVENT_NONE(rcu_invoke_kfree_callback,
+	TP_PROTO(const char *rcuname, struct rcu_head *rhp,
+		 unsigned long offset));
+TRACE_EVENT_NONE(rcu_batch_end,
+	TP_PROTO(const char *rcuname, int callbacks_invoked,
+		 char cb, char nr, char iit, char risk));
+TRACE_EVENT_NONE(rcu_torture_read,
+	TP_PROTO(const char *rcutorturename, struct rcu_head *rhp,
+		 unsigned long secs, unsigned long c_old, unsigned long c));
+TRACE_EVENT_NONE(rcu_barrier,
+	TP_PROTO(const char *rcuname, const char *s, int cpu, int cnt,
+		 unsigned long done));
 
 #endif /* #else #ifdef CONFIG_RCU_TRACE */
 
diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h
index a393e24..2778e44 100644
--- a/kernel/rcu/rcu.h
+++ b/kernel/rcu/rcu.h
@@ -24,11 +24,6 @@
 #define __LINUX_RCU_H
 
 #include <trace/events/rcu.h>
-#ifdef CONFIG_RCU_TRACE
-#define RCU_TRACE(stmt) stmt
-#else /* #ifdef CONFIG_RCU_TRACE */
-#define RCU_TRACE(stmt)
-#endif /* #else #ifdef CONFIG_RCU_TRACE */
 
 /* Offset to allow for unmatched rcu_irq_{enter,exit}(). */
 #define DYNTICK_IRQ_NONIDLE	((LONG_MAX / 2) + 1)
@@ -229,12 +224,12 @@ static inline bool __rcu_reclaim(const char *rn, struct rcu_head *head)
 
 	rcu_lock_acquire(&rcu_callback_map);
 	if (__is_kfree_rcu_offset(offset)) {
-		RCU_TRACE(trace_rcu_invoke_kfree_callback(rn, head, offset);)
+		trace_rcu_invoke_kfree_callback(rn, head, offset);
 		kfree((void *)head - offset);
 		rcu_lock_release(&rcu_callback_map);
 		return true;
 	} else {
-		RCU_TRACE(trace_rcu_invoke_callback(rn, head);)
+		trace_rcu_invoke_callback(rn, head);
 		f = head->func;
 		WRITE_ONCE(head->func, (rcu_callback_t)0L);
 		f(head);
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 9180158..d2ad39f 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -2329,14 +2329,14 @@ static void rcu_report_qs_rnp(unsigned long mask, struct rcu_node *rnp,
  */
 int rcutree_dying_cpu(unsigned int cpu)
 {
-	RCU_TRACE(bool blkd;)
-	RCU_TRACE(struct rcu_data *rdp = this_cpu_ptr(&rcu_data);)
-	RCU_TRACE(struct rcu_node *rnp = rdp->mynode;)
+	bool blkd;
+	struct rcu_data *rdp = this_cpu_ptr(&rcu_data);
+	struct rcu_node *rnp = rdp->mynode;
 
 	if (!IS_ENABLED(CONFIG_HOTPLUG_CPU))
 		return 0;
 
-	RCU_TRACE(blkd = !!(rnp->qsmask & rdp->grpmask);)
+	blkd = !!(rnp->qsmask & rdp->grpmask);
 	trace_rcu_grace_period(rcu_state.name, rnp->gp_seq,
 			       blkd ? TPS("cpuofl") : TPS("cpuofl-bgp"));
 	return 0;
-- 
1.8.3.1

Re: [PATCH 3/3] rcu: validate arguments for rcu tracepoints

[Paul E. McKenney]

On Tue, Mar 26, 2019 at 09:40:07AM +0800, Yafang Shao wrote:
> When CONFIG_RCU_TRACE is not set, all these tracepoints are define as
> do-nothing macro.
> We'd better make those inline functions that take proper arguments.
> 
> As RCU_TRACE() is defined as do-nothing marco as well when
> CONFIG_RCU_TRACE is not set, so we can clean it up.
> 
> Signed-off-by: Yafang Shao <laoar.shao@gmail.com>

With the naming changes suggested by Steven:

Acked-by: Paul E. McKenney <paulmck@linux.ibm.com>

This area of the code should be relatively free from conflicts, so please
feel free to take it with the tracing updates.

> ---
>  include/trace/events/rcu.h | 84 +++++++++++++++++++++++++++++-----------------
>  kernel/rcu/rcu.h           |  9 ++---
>  kernel/rcu/tree.c          |  8 ++---
>  3 files changed, 60 insertions(+), 41 deletions(-)
> 
> diff --git a/include/trace/events/rcu.h b/include/trace/events/rcu.h
> index f0c4d10..357e1b6 100644
> --- a/include/trace/events/rcu.h
> +++ b/include/trace/events/rcu.h
> @@ -750,36 +750,60 @@
> 
>  #else /* #ifdef CONFIG_RCU_TRACE */
> 
> -#define trace_rcu_grace_period(rcuname, gp_seq, gpevent) do { } while (0)
> -#define trace_rcu_future_grace_period(rcuname, gp_seq, gp_seq_req, \
> -				      level, grplo, grphi, event) \
> -				      do { } while (0)
> -#define trace_rcu_grace_period_init(rcuname, gp_seq, level, grplo, grphi, \
> -				    qsmask) do { } while (0)
> -#define trace_rcu_exp_grace_period(rcuname, gqseq, gpevent) \
> -	do { } while (0)
> -#define trace_rcu_exp_funnel_lock(rcuname, level, grplo, grphi, gpevent) \
> -	do { } while (0)
> -#define trace_rcu_nocb_wake(rcuname, cpu, reason) do { } while (0)
> -#define trace_rcu_preempt_task(rcuname, pid, gp_seq) do { } while (0)
> -#define trace_rcu_unlock_preempted_task(rcuname, gp_seq, pid) do { } while (0)
> -#define trace_rcu_quiescent_state_report(rcuname, gp_seq, mask, qsmask, level, \
> -					 grplo, grphi, gp_tasks) do { } \
> -	while (0)
> -#define trace_rcu_fqs(rcuname, gp_seq, cpu, qsevent) do { } while (0)
> -#define trace_rcu_dyntick(polarity, oldnesting, newnesting, dyntick) do { } while (0)
> -#define trace_rcu_callback(rcuname, rhp, qlen_lazy, qlen) do { } while (0)
> -#define trace_rcu_kfree_callback(rcuname, rhp, offset, qlen_lazy, qlen) \
> -	do { } while (0)
> -#define trace_rcu_batch_start(rcuname, qlen_lazy, qlen, blimit) \
> -	do { } while (0)
> -#define trace_rcu_invoke_callback(rcuname, rhp) do { } while (0)
> -#define trace_rcu_invoke_kfree_callback(rcuname, rhp, offset) do { } while (0)
> -#define trace_rcu_batch_end(rcuname, callbacks_invoked, cb, nr, iit, risk) \
> -	do { } while (0)
> -#define trace_rcu_torture_read(rcutorturename, rhp, secs, c_old, c) \
> -	do { } while (0)
> -#define trace_rcu_barrier(name, s, cpu, cnt, done) do { } while (0)
> +TRACE_EVENT_NONE(rcu_grace_period,
> +	TP_PROTO(const char *rcuname, unsigned long gp_seq,
> +		 const char *gpevent));
> +TRACE_EVENT_NONE(rcu_future_grace_period,
> +	TP_PROTO(const char *rcuname, unsigned long gp_seq,
> +		 unsigned long gp_seq_req, u8 level, int grplo,
> +		 int grphi, const char *gpevent));
> +TRACE_EVENT_NONE(rcu_grace_period_init,
> +	TP_PROTO(const char *rcuname, unsigned long gp_seq, u8 level,
> +		 int grplo, int grphi, unsigned long qsmask));
> +TRACE_EVENT_NONE(rcu_exp_grace_period,
> +	TP_PROTO(const char *rcuname, unsigned long gpseq,
> +		 const char *gpevent));
> +TRACE_EVENT_NONE(rcu_exp_funnel_lock,
> +	TP_PROTO(const char *rcuname, u8 level, int grplo, int grphi,
> +		 const char *gpevent));
> +TRACE_EVENT_NONE(rcu_nocb_wake,
> +	TP_PROTO(const char *rcuname, int cpu, const char *reason));
> +TRACE_EVENT_NONE(rcu_preempt_task,
> +	TP_PROTO(const char *rcuname, int pid, unsigned long gp_seq));
> +TRACE_EVENT_NONE(rcu_unlock_preempted_task,
> +	TP_PROTO(const char *rcuname, unsigned long gp_seq, int pid));
> +TRACE_EVENT_NONE(rcu_quiescent_state_report,
> +	TP_PROTO(const char *rcuname, unsigned long gp_seq,
> +		 unsigned long mask, unsigned long qsmask,
> +		 u8 level, int grplo, int grphi, int gp_tasks));
> +TRACE_EVENT_NONE(rcu_fqs,
> +	TP_PROTO(const char *rcuname, unsigned long gp_seq, int cpu,
> +		 const char *qsevent));
> +TRACE_EVENT_NONE(rcu_dyntick,
> +	TP_PROTO(const char *polarity, long oldnesting, long newnesting,
> +		 atomic_t dynticks));
> +TRACE_EVENT_NONE(rcu_callback,
> +	TP_PROTO(const char *rcuname, struct rcu_head *rhp, long qlen_lazy,
> +		 long qlen));
> +TRACE_EVENT_NONE(rcu_kfree_callback,
> +	TP_PROTO(const char *rcuname, struct rcu_head *rhp,
> +		 unsigned long offset, long qlen_lazy, long qlen));
> +TRACE_EVENT_NONE(rcu_batch_start,
> +	TP_PROTO(const char *rcuname, long qlen_lazy, long qlen, long blimit));
> +TRACE_EVENT_NONE(rcu_invoke_callback,
> +	TP_PROTO(const char *rcuname, struct rcu_head *rhp));
> +TRACE_EVENT_NONE(rcu_invoke_kfree_callback,
> +	TP_PROTO(const char *rcuname, struct rcu_head *rhp,
> +		 unsigned long offset));
> +TRACE_EVENT_NONE(rcu_batch_end,
> +	TP_PROTO(const char *rcuname, int callbacks_invoked,
> +		 char cb, char nr, char iit, char risk));
> +TRACE_EVENT_NONE(rcu_torture_read,
> +	TP_PROTO(const char *rcutorturename, struct rcu_head *rhp,
> +		 unsigned long secs, unsigned long c_old, unsigned long c));
> +TRACE_EVENT_NONE(rcu_barrier,
> +	TP_PROTO(const char *rcuname, const char *s, int cpu, int cnt,
> +		 unsigned long done));
> 
>  #endif /* #else #ifdef CONFIG_RCU_TRACE */
> 
> diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h
> index a393e24..2778e44 100644
> --- a/kernel/rcu/rcu.h
> +++ b/kernel/rcu/rcu.h
> @@ -24,11 +24,6 @@
>  #define __LINUX_RCU_H
> 
>  #include <trace/events/rcu.h>
> -#ifdef CONFIG_RCU_TRACE
> -#define RCU_TRACE(stmt) stmt
> -#else /* #ifdef CONFIG_RCU_TRACE */
> -#define RCU_TRACE(stmt)
> -#endif /* #else #ifdef CONFIG_RCU_TRACE */
> 
>  /* Offset to allow for unmatched rcu_irq_{enter,exit}(). */
>  #define DYNTICK_IRQ_NONIDLE	((LONG_MAX / 2) + 1)
> @@ -229,12 +224,12 @@ static inline bool __rcu_reclaim(const char *rn, struct rcu_head *head)
> 
>  	rcu_lock_acquire(&rcu_callback_map);
>  	if (__is_kfree_rcu_offset(offset)) {
> -		RCU_TRACE(trace_rcu_invoke_kfree_callback(rn, head, offset);)
> +		trace_rcu_invoke_kfree_callback(rn, head, offset);
>  		kfree((void *)head - offset);
>  		rcu_lock_release(&rcu_callback_map);
>  		return true;
>  	} else {
> -		RCU_TRACE(trace_rcu_invoke_callback(rn, head);)
> +		trace_rcu_invoke_callback(rn, head);
>  		f = head->func;
>  		WRITE_ONCE(head->func, (rcu_callback_t)0L);
>  		f(head);
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index 9180158..d2ad39f 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -2329,14 +2329,14 @@ static void rcu_report_qs_rnp(unsigned long mask, struct rcu_node *rnp,
>   */
>  int rcutree_dying_cpu(unsigned int cpu)
>  {
> -	RCU_TRACE(bool blkd;)
> -	RCU_TRACE(struct rcu_data *rdp = this_cpu_ptr(&rcu_data);)
> -	RCU_TRACE(struct rcu_node *rnp = rdp->mynode;)
> +	bool blkd;
> +	struct rcu_data *rdp = this_cpu_ptr(&rcu_data);
> +	struct rcu_node *rnp = rdp->mynode;
> 
>  	if (!IS_ENABLED(CONFIG_HOTPLUG_CPU))
>  		return 0;
> 
> -	RCU_TRACE(blkd = !!(rnp->qsmask & rdp->grpmask);)
> +	blkd = !!(rnp->qsmask & rdp->grpmask);
>  	trace_rcu_grace_period(rcu_state.name, rnp->gp_seq,
>  			       blkd ? TPS("cpuofl") : TPS("cpuofl-bgp"));
>  	return 0;
> -- 
> 1.8.3.1
> 

Re: [PATCH 0/3] tracing: introduce TRACE_EVENT_NONE and use it

[Steven Rostedt]

On Tue, 26 Mar 2019 09:40:04 +0800
Yafang Shao <laoar.shao@gmail.com> wrote:

> In this patchset, I introduce a new macro TRACE_EVENT_NONE(), which will
> define a tracepoint as a do-nothing inline function.
> 	#define TRACE_EVENT_NONE(name, proto)			\
> 		static inline void trace_##name(proto)		\
> 		{ }						\
> 		static inline bool trace_##name##_enabled(void)	\
> 		{						\
> 			return false;				\
> 		}
> 

BTW, I prefer a name of TRACE_EVENT_DISABLED()

Which shows that it is disabled, or TRACE_EVENT_UNDEF(), or
TRACE_EVENT_NOP(). Something that states it is turned off. When I first
saw NONE() I thought it was another attempt to introduce zero parameter
trace events.


> Let's take some example for why this macro is needed.
> 
> - sched
> The tracepoints trace_sched_stat_{iowait, blocked, wait, sleep} should
> be not exposed to user if CONFIG_SCHEDSTATS is not set.
> We can place #ifdef in the kernel/sched/fair.c to fix it, but we
> don't want to sprinkle #ifdef.

But you still sprinkle #ifdef in the header, but the part I don't like
is the need to duplicate the prototypes and the like.


> So with TRACE_EVENT_NONE(), we can fix it in
> include/trace/events/sched.h.
> 
> - rcu
> When CONFIG_RCU_TRACE is not set, some rcu tracepoints are define as
> do-nothing macro without validating arguments, that is not proper.
> We should validate the arguments.
> 
> Yafang Shao (3):
>   tracing: introduce TRACE_EVENT_NONE()
>   sched/fair: do not expose some tracepoints to user if
>     CONFIG_SCHEDSTATS is not set
>   rcu: validate arguments for rcu tracepoints
> 

See my reply in the next patch.

-- Steve

Re: [PATCH 0/3] tracing: introduce TRACE_EVENT_NONE and use it

[Yafang Shao]

On Tue, Mar 26, 2019 at 10:17 AM Steven Rostedt <rostedt@goodmis.org> wrote:
>
> On Tue, 26 Mar 2019 09:40:04 +0800
> Yafang Shao <laoar.shao@gmail.com> wrote:
>
> > In this patchset, I introduce a new macro TRACE_EVENT_NONE(), which will
> > define a tracepoint as a do-nothing inline function.
> >       #define TRACE_EVENT_NONE(name, proto)                   \
> >               static inline void trace_##name(proto)          \
> >               { }                                             \
> >               static inline bool trace_##name##_enabled(void) \
> >               {                                               \
> >                       return false;                           \
> >               }
> >
>
> BTW, I prefer a name of TRACE_EVENT_DISABLED()
>
> Which shows that it is disabled, or TRACE_EVENT_UNDEF(), or
> TRACE_EVENT_NOP(). Something that states it is turned off. When I first
> saw NONE() I thought it was another attempt to introduce zero parameter
> trace events.
>

I prefer TRACE_EVENT_NOP.
I will change it.

Thanks
Yafang

>
> > Let's take some example for why this macro is needed.
> >
> > - sched
> > The tracepoints trace_sched_stat_{iowait, blocked, wait, sleep} should
> > be not exposed to user if CONFIG_SCHEDSTATS is not set.
> > We can place #ifdef in the kernel/sched/fair.c to fix it, but we
> > don't want to sprinkle #ifdef.
>
> But you still sprinkle #ifdef in the header, but the part I don't like
> is the need to duplicate the prototypes and the like.
>
>
> > So with TRACE_EVENT_NONE(), we can fix it in
> > include/trace/events/sched.h.
> >
> > - rcu
> > When CONFIG_RCU_TRACE is not set, some rcu tracepoints are define as
> > do-nothing macro without validating arguments, that is not proper.
> > We should validate the arguments.
> >
> > Yafang Shao (3):
> >   tracing: introduce TRACE_EVENT_NONE()
> >   sched/fair: do not expose some tracepoints to user if
> >     CONFIG_SCHEDSTATS is not set
> >   rcu: validate arguments for rcu tracepoints
> >
>
> See my reply in the next patch.
>
> -- Steve