Re: [PATCH RFC] tracepoint: Introduce tracepoint callbacks executing with preempt on
From: Paul E. McKenney
Date: Fri Apr 27 2018 - 13:04:14 EST
On Fri, Apr 27, 2018 at 10:00:48AM -0700, Paul E. McKenney wrote:
> On Fri, Apr 27, 2018 at 12:46:41PM -0400, Steven Rostedt wrote:
> > On Fri, 27 Apr 2018 09:45:54 -0700
> > "Paul E. McKenney" <paulmck@xxxxxxxxxxxxxxxxxx> wrote:
> > > > > That shouldn't be needed. For the rcu_read_lock_sched case, there is a
> > > > > preempt_disable which needs to be a notrace, but for the srcu one,
> > > > > since we don't do that, I think it should be fine.
> > > >
> > > > Actually, I think I may agree here too. Because the _notrace is for
> > > > function tracing, and it shouldn't affect it. If people don't want it
> > > > traced, they could add those functions to the list in the notrace file.
> > >
> > > OK, feel free to ignore my notrace srcu_read_lock() patch, then. ;-)
> > Of course I wasn't thinking about the lockdep tracepoints that Joel
> > mentioned, which happens to be the reason for all this discussion in
> > the first place :-) Now I think we do need it. (OK, I can keep
> > changing my mind, can't I?).
> You can, but at some point I start applying heavy-duty hysteresis. ;-)
> So the current thought (as of your having sent the above email) is that
> we need notrace versions of srcu_read_lock() and srcu_read_unlock(),
> but not for __srcu_read_lock() and __srcu_read_unlock(), correct?
And Joel noted offline that I messed up srcu_read_unlock_notrace(),
so here is an updated patch with that fixed.
diff --git a/include/linux/srcu.h b/include/linux/srcu.h
index 91494d7e8e41..3e72a291c401 100644
@@ -195,6 +195,16 @@ static inline int srcu_read_lock(struct srcu_struct *sp) __acquires(sp)
+/* Used by tracing, cannot be traced and cannot invoke lockdep. */
+static inline notrace int
+srcu_read_lock_notrace(struct srcu_struct *sp) __acquires(sp)
+ int retval;
+ retval = __srcu_read_lock(sp);
+ return retval;
* srcu_read_unlock - unregister a old reader from an SRCU-protected structure.
* @sp: srcu_struct in which to unregister the old reader.
@@ -209,6 +219,13 @@ static inline void srcu_read_unlock(struct srcu_struct *sp, int idx)
+/* Used by tracing, cannot be traced and cannot call lockdep. */
+static inline notrace void
+srcu_read_unlock_notrace(struct srcu_struct *sp, int idx) __releases(sp)
+ __srcu_read_unlock(sp, idx);
* smp_mb__after_srcu_read_unlock - ensure full ordering after srcu_read_unlock