Re: [for-next][PATCH 4/4] ftrace: Add comment to why rcu_dereference_sched() is open coded

[Joel Fernandes <joel@joelfernandes.org>]

On Wed, Feb 05, 2020 at 06:33:49AM -0500, Steven Rostedt wrote:
> 
> Paul and Joel,
> 
> Care to ack this patch (or complain about it ;-) ?
> 
> -- Steve
> 
> 
> On Wed, 05 Feb 2020 05:49:33 -0500
> Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> > From: "Steven Rostedt (VMware)" <rostedt@goodmis.org>
> > 
> > Because the function graph tracer can execute in sections where RCU is not
> > "watching", the rcu_dereference_sched() for the has needs to be open coded.
> > This is fine because the RCU "flavor" of the ftrace hash is protected by
> > its own RCU handling (it does its own little synchronization on every CPU
> > and does not rely on RCU sched).
> > 
> > Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
> > ---
> >  kernel/trace/trace.h | 10 ++++++++++
> >  1 file changed, 10 insertions(+)
> > 
> > diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h
> > index 022def96d307..8c52f5de9384 100644
> > --- a/kernel/trace/trace.h
> > +++ b/kernel/trace/trace.h
> > @@ -975,6 +975,11 @@ static inline int ftrace_graph_addr(struct ftrace_graph_ent *trace)
> >  
> >  	preempt_disable_notrace();
> >  
> > +	/*
> > +	 * Have to open code "rcu_dereference_sched()" because the
> > +	 * function graph tracer can be called when RCU is not
> > +	 * "watching".
> > +	 */
> >  	hash = rcu_dereference_protected(ftrace_graph_hash, !preemptible());
> >  
> >  	if (ftrace_hash_empty(hash)) {
> > @@ -1022,6 +1027,11 @@ static inline int ftrace_graph_notrace_addr(unsigned long addr)
> >  
> >  	preempt_disable_notrace();
> >  
> > +	/*
> > +	 * Have to open code "rcu_dereference_sched()" because the
> > +	 * function graph tracer can be called when RCU is not
> > +	 * "watching".
> > +	 */
> >  	notrace_hash = rcu_dereference_protected(ftrace_graph_notrace_hash,
> >  						 !preemptible());
> >  

Could you paste the stack here when RCU is not watching? In trace event code
IIRC we call rcu_enter_irqs_on() to have RCU temporarily watch, since that
code can be called from idle loop. Should we doing the same here as well?

thanks,

 - Joel