Hello Dmitry,
> --- a/net/mac802154/iface.c
> +++ b/net/mac802154/iface.c
> @@ -669,7 +669,7 @@ ieee802154_if_add(struct ieee802154_local *local, const char *name,
> goto err;
>
> mutex_lock(&local->iflist_mtx);
> - list_add_tail_rcu(&sdata->list, &local->interfaces);
> + list_add_tail(&sdata->list, &local->interfaces);
> mutex_unlock(&local->iflist_mtx);
>
> return ndev;
> @@ -683,11 +683,13 @@ void ieee802154_if_remove(struct ieee802154_sub_if_data *sdata)
> {
> ASSERT_RTNL();
>
> + if (test_and_set_bit(SDATA_STATE_REMOVED, &sdata->state))
> + return;
> +
> mutex_lock(&sdata->local->iflist_mtx);
> - list_del_rcu(&sdata->list);
> + list_del(&sdata->list);
> mutex_unlock(&sdata->local->iflist_mtx);
>
> - synchronize_rcu();
> unregister_netdevice(sdata->dev);
> }
>
> @@ -697,6 +699,8 @@ void ieee802154_remove_interfaces(struct ieee802154_local *local)
>
> mutex_lock(&local->iflist_mtx);
> list_for_each_entry_safe(sdata, tmp, &local->interfaces, list) {
> + if (test_and_set_bit(SDATA_STATE_REMOVED, &sdata->state))
> + continue;
> list_del(&sdata->list);
Why not just enclose this list_del() within a mutex_lock(iflist_mtx)
like the others? Would probably make more sense and prevent the use of
yet another protection mechanism? Is there anything preventing the use
of this mutex here?
Thanks,
Miquèl
