Re: [PATCH 2/2] tracepoint: introduce *_noupdate APIs.

[Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>]

* Lai Jiangshan (laijs@cn.fujitsu.com) wrote:
> 
> new APIs separate tracepoint_probe_register(),
> tracepoint_probe_unregister() into 2 steps. The first step of them
> is just update tracepoint_entry, not connect or disconnect.
> 
> this patch introduce tracepoint_probe_update_all() for update all.
> 
> these APIs are very useful for registering a lots of probes
> but just update once only. and a very important thing is that
> *_noupdate APIs do not require module_mutex.
> 
> Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>

[...]

> +/**
> + * tracepoint_probe_update_all -  update tracepoints
> + */
> +void tracepoint_probe_update_all(void)
> +{
> +	LIST_HEAD(release_probes);
> +	struct tp_probes *pos, *next;
> +
> +	mutex_lock(&tracepoints_mutex);
> +	if (!need_update) {
> +		mutex_unlock(&tracepoints_mutex);
> +		return;
> +	}
> +	if (!list_empty(&old_probes))
> +		list_replace_init(&old_probes, &release_probes);
> +	need_update = 0;
> +	mutex_unlock(&tracepoints_mutex);
> +
> +	tracepoint_update_probes();

I think the read-side of this release_probes list should be protected by
the tracepoints_mutex too. Two concurrent tracepoint_probe_update_all()
could cause havoc here :


mutex_lock(&tracepoints_mutex);

> +	list_for_each_entry_safe(pos, next, &release_probes, u.list) {
> +		list_del(&pos->u.list);
> +		call_rcu_sched(&pos->u.rcu, rcu_free_old_probes);
> +	}

mutex_unlock(&tracepoints_mutex);

?

The rest looks good.

Mathieu

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68