Messages in this thread | | | Date | Mon, 18 May 2009 17:35:05 -0700 | From | "Paul E. McKenney" <> | Subject | Re: [PATCH] tracing: add trace_event_read_lock() |
| |
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/
| |