netfilter: disable defrag once its no longer needed

When I changed defrag hooks to no longer get registered by default I
intentionally made it so that registration can only be un-done by unloading
the nf_defrag_ipv4/6 module.

In hindsight this was too conservative; there is no reason to keep defrag
on while there is no feature dependency anymore.

Moreover, this won't work if user isn't allowed to remove nf_defrag module.

This adds the disable() functions for both ipv4 and ipv6 and calls them
from conntrack, TPROXY and the xtables socket module.

ipvs isn't converted here, it will behave as before this patch and
will need module removal.

Signed-off-by: Florian Westphal <[e-mail]>
Signed-off-by: Pablo Neira Ayuso <[e-mail]>

include/net/netfilter/ipv4/nf_defrag_ipv4.h

···
 #define _NF_DEFRAG_IPV4_H
 
 struct net;
-int nf_defrag_ipv4_enable(struct net *);
+int nf_defrag_ipv4_enable(struct net *net);
+void nf_defrag_ipv4_disable(struct net *net);
 
 #endif /* _NF_DEFRAG_IPV4_H */

include/net/netfilter/ipv6/nf_defrag_ipv6.h

···
 #include <linux/skbuff.h>
 #include <linux/types.h>
 
-int nf_defrag_ipv6_enable(struct net *);
+int nf_defrag_ipv6_enable(struct net *net);
+void nf_defrag_ipv6_disable(struct net *net);
 
 int nf_ct_frag6_init(void);
 void nf_ct_frag6_cleanup(void);

net/ipv4/netfilter/nf_defrag_ipv4.c

···
 	struct defrag4_pernet *nf_defrag = net_generic(net, defrag4_pernet_id);
 	int err = 0;
 
-	might_sleep();
-
-	if (nf_defrag->users)
-		return 0;
-
 	mutex_lock(&defrag4_mutex);
-	if (nf_defrag->users)
+	if (nf_defrag->users == UINT_MAX) {
+		err = -EOVERFLOW;
 		goto out_unlock;
+	}
+
+	if (nf_defrag->users) {
+		nf_defrag->users++;
+		goto out_unlock;
+	}
 
 	err = nf_register_net_hooks(net, ipv4_defrag_ops,
 				    ARRAY_SIZE(ipv4_defrag_ops));
···
 	return err;
 }
 EXPORT_SYMBOL_GPL(nf_defrag_ipv4_enable);
+
+void nf_defrag_ipv4_disable(struct net *net)
+{
+	struct defrag4_pernet *nf_defrag = net_generic(net, defrag4_pernet_id);
+
+	mutex_lock(&defrag4_mutex);
+	if (nf_defrag->users) {
+		nf_defrag->users--;
+		if (nf_defrag->users == 0)
+			nf_unregister_net_hooks(net, ipv4_defrag_ops,
+						ARRAY_SIZE(ipv4_defrag_ops));
+	}
+
+	mutex_unlock(&defrag4_mutex);
+}
+EXPORT_SYMBOL_GPL(nf_defrag_ipv4_disable);
 
 module_init(nf_defrag_init);
 module_exit(nf_defrag_fini);

net/ipv6/netfilter/nf_defrag_ipv6_hooks.c

···
 	struct nft_ct_frag6_pernet *nf_frag = net_generic(net, nf_frag_pernet_id);
 	int err = 0;
 
-	might_sleep();
-
-	if (nf_frag->users)
-		return 0;
-
 	mutex_lock(&defrag6_mutex);
-	if (nf_frag->users)
+	if (nf_frag->users == UINT_MAX) {
+		err = -EOVERFLOW;
 		goto out_unlock;
+	}
+
+	if (nf_frag->users) {
+		nf_frag->users++;
+		goto out_unlock;
+	}
 
 	err = nf_register_net_hooks(net, ipv6_defrag_ops,
 				    ARRAY_SIZE(ipv6_defrag_ops));
···
 	return err;
 }
 EXPORT_SYMBOL_GPL(nf_defrag_ipv6_enable);
+
+void nf_defrag_ipv6_disable(struct net *net)
+{
+	struct nft_ct_frag6_pernet *nf_frag = net_generic(net, nf_frag_pernet_id);
+
+	mutex_lock(&defrag6_mutex);
+	if (nf_frag->users) {
+		nf_frag->users--;
+		if (nf_frag->users == 0)
+			nf_unregister_net_hooks(net, ipv6_defrag_ops,
+						ARRAY_SIZE(ipv6_defrag_ops));
+	}
+	mutex_unlock(&defrag6_mutex);
+}
+EXPORT_SYMBOL_GPL(nf_defrag_ipv6_disable);
 
 module_init(nf_defrag_init);
 module_exit(nf_defrag_fini);

net/netfilter/nf_conntrack_proto.c

···
 	mutex_lock(&nf_ct_proto_mutex);
 	switch (nfproto) {
 	case NFPROTO_IPV4:
-		if (cnet->users4 && (--cnet->users4 == 0))
+		if (cnet->users4 && (--cnet->users4 == 0)) {
 			nf_unregister_net_hooks(net, ipv4_conntrack_ops,
 						ARRAY_SIZE(ipv4_conntrack_ops));
+			nf_defrag_ipv4_disable(net);
+		}
 		break;
 #if IS_ENABLED(CONFIG_IPV6)
 	case NFPROTO_IPV6:
-		if (cnet->users6 && (--cnet->users6 == 0))
+		if (cnet->users6 && (--cnet->users6 == 0)) {
 			nf_unregister_net_hooks(net, ipv6_conntrack_ops,
 						ARRAY_SIZE(ipv6_conntrack_ops));
+			nf_defrag_ipv6_disable(net);
+		}
 		break;
 #endif
 	case NFPROTO_BRIDGE:

net/netfilter/nft_tproxy.c

···
 	return 0;
 }
 
+static void nft_tproxy_destroy(const struct nft_ctx *ctx,
+			       const struct nft_expr *expr)
+{
+	const struct nft_tproxy *priv = nft_expr_priv(expr);
+
+	switch (priv->family) {
+	case NFPROTO_IPV4:
+		nf_defrag_ipv4_disable(ctx->net);
+		break;
+#if IS_ENABLED(CONFIG_NF_TABLES_IPV6)
+	case NFPROTO_IPV6:
+		nf_defrag_ipv6_disable(ctx->net);
+		break;
+#endif
+	case NFPROTO_UNSPEC:
+		nf_defrag_ipv4_disable(ctx->net);
+#if IS_ENABLED(CONFIG_NF_TABLES_IPV6)
+		nf_defrag_ipv6_disable(ctx->net);
+#endif
+		break;
+	}
+}
+
 static int nft_tproxy_dump(struct sk_buff *skb,
 			   const struct nft_expr *expr)
 {
···
 	.size		= NFT_EXPR_SIZE(sizeof(struct nft_tproxy)),
 	.eval		= nft_tproxy_eval,
 	.init		= nft_tproxy_init,
+	.destroy	= nft_tproxy_destroy,
 	.dump		= nft_tproxy_dump,
 };
 

net/netfilter/xt_TPROXY.c

···
 	pr_info_ratelimited("Can be used only with -p tcp or -p udp\n");
 	return -EINVAL;
 }
+
+static void tproxy_tg6_destroy(const struct xt_tgdtor_param *par)
+{
+	nf_defrag_ipv6_disable(par->net);
+}
 #endif
 
 static int tproxy_tg4_check(const struct xt_tgchk_param *par)
···
 	return -EINVAL;
 }
 
+static void tproxy_tg4_destroy(const struct xt_tgdtor_param *par)
+{
+	nf_defrag_ipv4_disable(par->net);
+}
+
 static struct xt_target tproxy_tg_reg[] __read_mostly = {
 	{
 		.name		= "TPROXY",
···
 		.revision	= 0,
 		.targetsize	= sizeof(struct xt_tproxy_target_info),
 		.checkentry	= tproxy_tg4_check,
+		.destroy	= tproxy_tg4_destroy,
 		.hooks		= 1 << NF_INET_PRE_ROUTING,
 		.me		= THIS_MODULE,
 	},
···
 		.revision	= 1,
 		.targetsize	= sizeof(struct xt_tproxy_target_info_v1),
 		.checkentry	= tproxy_tg4_check,
+		.destroy	= tproxy_tg4_destroy,
 		.hooks		= 1 << NF_INET_PRE_ROUTING,
 		.me		= THIS_MODULE,
 	},
···
 		.revision	= 1,
 		.targetsize	= sizeof(struct xt_tproxy_target_info_v1),
 		.checkentry	= tproxy_tg6_check,
+		.destroy	= tproxy_tg6_destroy,
 		.hooks		= 1 << NF_INET_PRE_ROUTING,
 		.me		= THIS_MODULE,
 	},

net/netfilter/xt_socket.c

···
 	return 0;
 }
 
+static void socket_mt_destroy(const struct xt_mtdtor_param *par)
+{
+	if (par->family == NFPROTO_IPV4)
+		nf_defrag_ipv4_disable(par->net);
+	else if (par->family == NFPROTO_IPV6)
+		nf_defrag_ipv4_disable(par->net);
+}
+
 static struct xt_match socket_mt_reg[] __read_mostly = {
 	{
 		.name		= "socket",
···
 		.revision	= 1,
 		.family		= NFPROTO_IPV4,
 		.match		= socket_mt4_v1_v2_v3,
+		.destroy	= socket_mt_destroy,
 		.checkentry	= socket_mt_v1_check,
 		.matchsize	= sizeof(struct xt_socket_mtinfo1),
 		.hooks		= (1 << NF_INET_PRE_ROUTING) |
···
 		.match		= socket_mt6_v1_v2_v3,
 		.checkentry	= socket_mt_v1_check,
 		.matchsize	= sizeof(struct xt_socket_mtinfo1),
+		.destroy	= socket_mt_destroy,
 		.hooks		= (1 << NF_INET_PRE_ROUTING) |
 				  (1 << NF_INET_LOCAL_IN),
 		.me		= THIS_MODULE,
···
 		.family		= NFPROTO_IPV4,
 		.match		= socket_mt4_v1_v2_v3,
 		.checkentry	= socket_mt_v2_check,
+		.destroy	= socket_mt_destroy,
 		.matchsize	= sizeof(struct xt_socket_mtinfo1),
 		.hooks		= (1 << NF_INET_PRE_ROUTING) |
 				  (1 << NF_INET_LOCAL_IN),
···
 		.family		= NFPROTO_IPV6,
 		.match		= socket_mt6_v1_v2_v3,
 		.checkentry	= socket_mt_v2_check,
+		.destroy	= socket_mt_destroy,
 		.matchsize	= sizeof(struct xt_socket_mtinfo1),
 		.hooks		= (1 << NF_INET_PRE_ROUTING) |
 				  (1 << NF_INET_LOCAL_IN),
···
 		.family		= NFPROTO_IPV4,
 		.match		= socket_mt4_v1_v2_v3,
 		.checkentry	= socket_mt_v3_check,
+		.destroy	= socket_mt_destroy,
 		.matchsize	= sizeof(struct xt_socket_mtinfo1),
 		.hooks		= (1 << NF_INET_PRE_ROUTING) |
 				  (1 << NF_INET_LOCAL_IN),
···
 		.family		= NFPROTO_IPV6,
 		.match		= socket_mt6_v1_v2_v3,
 		.checkentry	= socket_mt_v3_check,
+		.destroy	= socket_mt_destroy,
 		.matchsize	= sizeof(struct xt_socket_mtinfo1),
 		.hooks		= (1 << NF_INET_PRE_ROUTING) |
 				  (1 << NF_INET_LOCAL_IN),