Re: [PATCH] mac802154: fix interface deletion

[Miquel Raynal <miquel.raynal@bootlin.com>]

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