Re: [PATCH -v3 5/8] fsnotify: unified filesystem notification backend

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Tue, 2008-11-25 at 12:21 -0500, Eric Paris wrote:
> +struct fsnotify_group *fsnotify_find_group(unsigned int group_num, unsigned long mask, struct fsnotify_ops *ops)
> +{
> +       struct fsnotify_group *group_iter;
> +       struct fsnotify_group *group = NULL;
> +
> +       mutex_lock(&fsnotify_grp_mutex);
> +       list_for_each_entry_rcu(group_iter, &fsnotify_groups, group_list) {
> +               if (group_iter->group_num == group_num) {
> +                       if ((group_iter->mask == mask) &&
> +                           (group_iter->ops == ops)) {
> +                               fsnotify_get_group(group_iter);
> +                               group = group_iter;
> +                       } else
> +                               group = ERR_PTR(-EEXIST);
> +                       goto out;
> +               }
> +       }
> +
> +       group = kmalloc(sizeof(struct fsnotify_group), GFP_KERNEL);
> +       if (!group) {
> +               group = ERR_PTR(-ENOMEM);
> +               goto out;
> +       }
> +
> +       atomic_set(&group->refcnt, 1);
> +
> +       group->group_num = group_num;
> +       group->mask = mask;
> +
> +       mutex_init(&group->notification_mutex);
> +       INIT_LIST_HEAD(&group->notification_list);
> +       init_waitqueue_head(&group->notification_waitq);
> +
> +       group->ops = ops;
> +
> +       /* add it */
> +       list_add_rcu(&group->group_list, &fsnotify_groups);
> +
> +out:
> +       mutex_unlock(&fsnotify_grp_mutex);
> +       fsnotify_recalc_global_mask();
> +       return group;
> +}

Can't you do a lockless lookup and handle the insertion race?

Also, since it creates the object if its not found, _find_ might not be
the best name, how about obtain?