Re: [PATCH 1/2] trace_events_filter: use rcu_assign_pointer() when setting ftrace_event_call->filter

[Steven Rostedt <rostedt@goodmis.org>]

[ Added Paul to Cc ]

On Tue, 2011-11-22 at 17:46 -0800, Tejun Heo wrote:
> ftrace_event_call->filter is sched RCU protected but didn't use
> rcu_assign_pointer().  Fix it.

Is it really needed? Maybe just for documentation but I'm not sure this
use is required because all use cases have synchronize_sched() used,
which is a big hammer compared to the rcu_assign_pointer().

> 
> TODO: Add proper __rcu annotation to call->filter and all its users.
> 
> Signed-off-by: Tejun Heo <tj@kernel.org>
> ---
>  kernel/trace/trace_events_filter.c |    6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> Index: work/kernel/trace/trace_events_filter.c
> ===================================================================
> --- work.orig/kernel/trace/trace_events_filter.c
> +++ work/kernel/trace/trace_events_filter.c
> @@ -1686,7 +1686,7 @@ static int replace_system_preds(struct e
>  		 * replace the filter for the call.
>  		 */
>  		filter = call->filter;
> -		call->filter = filter_item->filter;
> +		rcu_assign_pointer(call->filter, filter_item->filter);

We update filter here, and then call synchronize_sched() before we free
the filter_item->filter.

>  		filter_item->filter = filter;
>  
>  		fail = false;
> @@ -1741,7 +1741,7 @@ int apply_event_filter(struct ftrace_eve
>  		filter = call->filter;
>  		if (!filter)
>  			goto out_unlock;
> -		call->filter = NULL;
> +		rcu_assign_pointer(call->filter, NULL);
>  		/* Make sure the filter is not being used */

Again you can see that synchronize_sched() is called here.

>  		synchronize_sched();
>  		__free_filter(filter);
> @@ -1782,7 +1782,7 @@ out:
>  	 * string
>  	 */
>  	tmp = call->filter;
> -	call->filter = filter;
> +	rcu_assign_pointer(call->filter, filter);

We only call synchronize_sched if call->filter wasn't NULL, because we
are going to free tmp. We need to make sure all users are done with tmp
before we free it.

>  	if (tmp) {
>  		/* Make sure the call is done with the filter */
>  		synchronize_sched();

Thus my question is, do we really need to add the rcu_assign_pointer().
I have no problem if we only do so to document that this is an rcu sched
protected variable. But it should be commented as such.

-- Steve