Re: [PATCH] tracing: add trace_event_read_lock()

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Mon, May 18, 2009 at 03:59:31PM +0200, Frederic Weisbecker wrote:
> On Mon, May 18, 2009 at 07:35:34PM +0800, Lai Jiangshan wrote:
> > 
> > I found that there is nothing to protect event_hash in
> > ftrace_find_event().
> 
> Actually, rcu protects it, but not enough. We have neither
> synchronize_rcu() nor rcu_read_lock.
> 
> So we protect against concurrent hlist accesses.
> But the event can be removed when a module is unloaded,
> and that can happen between the time we get the event output
> callback and the time we actually use it.

I will ask the stupid question...  Would invoking rcu_barrier() in the
module-exit function take care of this?  The rcu_barrier() primitive
waits for all in-flight RCU callbacks to complete execution.

							Thanx, Paul

> > RCU can also protect ftrace_find_event(), but it will add
> > latency to kernel. So this fix converts 'trace_event_mutex'
> > to a read/write semaphore, and adds trace_event_read_lock()
> > to protect ftrace_find_event().
> > 
> > Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
> > ---
> > diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> > index 665a915..4ef8052 100644
> > --- a/kernel/trace/trace.c
> > +++ b/kernel/trace/trace.c
> > @@ -1569,12 +1569,14 @@ static void *s_start(struct seq_file *m, loff_t *pos)
> >  		p = s_next(m, p, &l);
> >  	}
> >  
> > +	trace_event_read_lock();
> >  	return p;
> >  }
> >  
> >  static void s_stop(struct seq_file *m, void *p)
> >  {
> >  	atomic_dec(&trace_record_cmdline_disabled);
> > +	trace_event_read_unlock();
> >  }
> >  
> >  static void print_lat_help_header(struct seq_file *m)
> > @@ -1817,6 +1819,7 @@ static int trace_empty(struct trace_iterator *iter)
> >  	return 1;
> >  }
> >  
> > +/*  Called with trace_event_read_lock() held. */
> >  static enum print_line_t print_trace_line(struct trace_iterator *iter)
> >  {
> >  	enum print_line_t ret;
> > @@ -3008,6 +3011,7 @@ waitagain:
> >  	       offsetof(struct trace_iterator, seq));
> >  	iter->pos = -1;
> >  
> > +	trace_event_read_lock();
> >  	while (find_next_entry_inc(iter) != NULL) {
> >  		enum print_line_t ret;
> >  		int len = iter->seq.len;
> > @@ -3024,6 +3028,7 @@ waitagain:
> >  		if (iter->seq.len >= cnt)
> >  			break;
> >  	}
> > +	trace_event_read_unlock();
> >  	/* Now copy what we have to the user */
> >  	sret = trace_seq_to_user(&iter->seq, ubuf, cnt);
> > @@ -3146,6 +3151,8 @@ static ssize_t tracing_splice_read_pipe(struct file *filp,
> >  		goto out_err;
> >  	}
> >  
> > +	trace_event_read_lock();
> > +
> >  	/* Fill as many pages as possible. */
> >  	for (i = 0, rem = len; i < PIPE_BUFFERS && rem; i++) {
> >  		pages[i] = alloc_page(GFP_KERNEL);
> > @@ -3168,6 +3175,7 @@ static ssize_t tracing_splice_read_pipe(struct file *filp,
> >  		trace_seq_init(&iter->seq);
> >  	}
> >  
> > +	trace_event_read_unlock();
> >  	mutex_unlock(&iter->mutex);
> >  
> >  	spd.nr_pages = i;
> > diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
> > index 489c0e8..7136420 100644
> > --- a/kernel/trace/trace_output.c
> > +++ b/kernel/trace/trace_output.c
> > @@ -14,7 +14,7 @@
> >  /* must be a power of 2 */
> >  #define EVENT_HASHSIZE	128
> >  
> > -static DEFINE_MUTEX(trace_event_mutex);
> > +static DECLARE_RWSEM(trace_event_mutex);
> >  static struct hlist_head event_hash[EVENT_HASHSIZE] __read_mostly;
> >  
> >  static int next_event_type = __TRACE_LAST_TYPE + 1;
> > @@ -466,6 +466,7 @@ static int task_state_char(unsigned long state)
> >   * @type: the type of event to look for
> >   *
> >   * Returns an event of type @type otherwise NULL
> > + * Called with trace_event_read_lock() held.
> >   */
> >  struct trace_event *ftrace_find_event(int type)
> >  {
> > @@ -475,7 +476,7 @@ struct trace_event *ftrace_find_event(int type)
> >  
> >  	key = type & (EVENT_HASHSIZE - 1);
> >  
> > -	hlist_for_each_entry_rcu(event, n, &event_hash[key], node) {
> > +	hlist_for_each_entry(event, n, &event_hash[key], node) {
> >  		if (event->type == type)
> >  			return event;
> >  	}
> > @@ -513,6 +514,16 @@ static int trace_search_list(struct list_head **list)
> >  	return last + 1;
> >  }
> >  
> > +void trace_event_read_lock(void)
> > +{
> > +	down_read(&trace_event_mutex);
> > +}
> > +
> > +void trace_event_read_unlock(void)
> > +{
> > +	up_read(&trace_event_mutex);
> > +}
> > +
> >  /**
> >   * register_ftrace_event - register output for an event type
> >   * @event: the event type to register
> > @@ -533,7 +544,7 @@ int register_ftrace_event(struct trace_event *event)
> >  	unsigned key;
> >  	int ret = 0;
> >  
> > -	mutex_lock(&trace_event_mutex);
> > +	down_write(&trace_event_mutex);
> >  
> >  	if (WARN_ON(!event))
> >  		goto out;
> > @@ -581,11 +592,11 @@ int register_ftrace_event(struct trace_event *event)
> >  
> >  	key = event->type & (EVENT_HASHSIZE - 1);
> >  
> > -	hlist_add_head_rcu(&event->node, &event_hash[key]);
> > +	hlist_add_head(&event->node, &event_hash[key]);
> >  
> >  	ret = event->type;
> >   out:
> > -	mutex_unlock(&trace_event_mutex);
> > +	up_write(&trace_event_mutex);
> >  
> >  	return ret;
> >  }
> > @@ -597,10 +608,10 @@ EXPORT_SYMBOL_GPL(register_ftrace_event);
> >   */
> >  int unregister_ftrace_event(struct trace_event *event)
> >  {
> > -	mutex_lock(&trace_event_mutex);
> > +	down_write(&trace_event_mutex);
> >  	hlist_del(&event->node);
> >  	list_del(&event->list);
> > -	mutex_unlock(&trace_event_mutex);
> > +	up_write(&trace_event_mutex);
> >  
> >  	return 0;
> >  }
> > diff --git a/kernel/trace/trace_output.h b/kernel/trace/trace_output.h
> > index 6e220a8..ac240e7 100644
> > --- a/kernel/trace/trace_output.h
> > +++ b/kernel/trace/trace_output.h
> > @@ -20,6 +20,8 @@ extern int seq_print_user_ip(struct trace_seq *s, struct mm_struct *mm,
> >  extern int trace_print_context(struct trace_iterator *iter);
> >  extern int trace_print_lat_context(struct trace_iterator *iter);
> >  
> > +extern void trace_event_read_lock(void);
> > +extern void trace_event_read_unlock(void);
> >  extern struct trace_event *ftrace_find_event(int type);
> >  
> >  extern enum print_line_t trace_nop_print(struct trace_iterator *iter,
> 
> 
> 
> It could be more fine grained. We could have a per event rwsem, and also place the
> protected read section only in trace_print_entry() which is the only racy window.
> 
> But I'm not sure it's that worthy since event removal is a rare thing.
> 
> So I guess this patch is fine.
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/