netfilter: x_tables: never register tables by default

For historical reasons x_tables still register tables by default in the
initial namespace.
Only newly created net namespaces add the hook on demand.

This means that the init_net always pays hook cost, even if no filtering
rules are added (e.g. only used inside a single netns).

Note that the hooks are added even when 'iptables -L' is called.
This is because there is no way to tell 'iptables -A' and 'iptables -L'
apart at kernel level.

The only solution would be to register the table, but delay hook
registration until the first rule gets added (or policy gets changed).

That however means that counters are not hooked either, so 'iptables -L'
would always show 0-counters even when traffic is flowing which might be
unexpected.

This keeps table and hook registration consistent with what is already done
in non-init netns: first iptables(-save) invocation registers both table
and hooks.

This applies the same solution adopted for ebtables.
All tables register a template that contains the l3 family, the name
and a constructor function that is called when the initial table has to
be added.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

1 files changed, 80 insertions, 18 deletions

diff --git a/net/netfilter/x_tables.c b/net/netfilter/x_tables.c
index 84e58ee501a4..25524e393349 100644
--- a/net/netfilter/x_tables.c
+++ b/net/netfilter/x_tables.c
@@ -39,6 +39,20 @@ MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module");
 #define XT_PCPU_BLOCK_SIZE 4096
 #define XT_MAX_TABLE_SIZE	(512 * 1024 * 1024)
 
+struct xt_template {
+	struct list_head list;
+
+	/* called when table is needed in the given netns */
+	int (*table_init)(struct net *net);
+
+	struct module *me;
+
+	/* A unique name... */
+	char name[XT_TABLE_MAXNAMELEN];
+};
+
+static struct list_head xt_templates[NFPROTO_NUMPROTO];
+
 struct xt_pernet {
 	struct list_head tables[NFPROTO_NUMPROTO];
 };
@@ -1221,48 +1235,43 @@ struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af,
 				    const char *name)
 {
 	struct xt_pernet *xt_net = net_generic(net, xt_pernet_id);
-	struct xt_table *t, *found = NULL;
+	struct module *owner = NULL;
+	struct xt_template *tmpl;
+	struct xt_table *t;
 
 	mutex_lock(&xt[af].mutex);
 	list_for_each_entry(t, &xt_net->tables[af], list)
 		if (strcmp(t->name, name) == 0 && try_module_get(t->me))
 			return t;
 
-	if (net == &init_net)
-		goto out;
-
-	/* Table doesn't exist in this netns, re-try init */
-	xt_net = net_generic(&init_net, xt_pernet_id);
-	list_for_each_entry(t, &xt_net->tables[af], list) {
+	/* Table doesn't exist in this netns, check larval list */
+	list_for_each_entry(tmpl, &xt_templates[af], list) {
 		int err;
 
-		if (strcmp(t->name, name))
+		if (strcmp(tmpl->name, name))
 			continue;
-		if (!try_module_get(t->me))
+		if (!try_module_get(tmpl->me))
 			goto out;
+
+		owner = tmpl->me;
+
 		mutex_unlock(&xt[af].mutex);
-		err = t->table_init(net);
+		err = tmpl->table_init(net);
 		if (err < 0) {
-			module_put(t->me);
+			module_put(owner);
 			return ERR_PTR(err);
 		}
 
-		found = t;
-
 		mutex_lock(&xt[af].mutex);
 		break;
 	}
 
-	if (!found)
-		goto out;
-
-	xt_net = net_generic(net, xt_pernet_id);
 	/* and once again: */
 	list_for_each_entry(t, &xt_net->tables[af], list)
 		if (strcmp(t->name, name) == 0)
 			return t;
 
-	module_put(found->me);
+	module_put(owner);
  out:
 	mutex_unlock(&xt[af].mutex);
 	return ERR_PTR(-ENOENT);
@@ -1749,6 +1758,58 @@ xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn)
 }
 EXPORT_SYMBOL_GPL(xt_hook_ops_alloc);
 
+int xt_register_template(const struct xt_table *table,
+			 int (*table_init)(struct net *net))
+{
+	int ret = -EEXIST, af = table->af;
+	struct xt_template *t;
+
+	mutex_lock(&xt[af].mutex);
+
+	list_for_each_entry(t, &xt_templates[af], list) {
+		if (WARN_ON_ONCE(strcmp(table->name, t->name) == 0))
+			goto out_unlock;
+	}
+
+	ret = -ENOMEM;
+	t = kzalloc(sizeof(*t), GFP_KERNEL);
+	if (!t)
+		goto out_unlock;
+
+	BUILD_BUG_ON(sizeof(t->name) != sizeof(table->name));
+
+	strscpy(t->name, table->name, sizeof(t->name));
+	t->table_init = table_init;
+	t->me = table->me;
+	list_add(&t->list, &xt_templates[af]);
+	ret = 0;
+out_unlock:
+	mutex_unlock(&xt[af].mutex);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(xt_register_template);
+
+void xt_unregister_template(const struct xt_table *table)
+{
+	struct xt_template *t;
+	int af = table->af;
+
+	mutex_lock(&xt[af].mutex);
+	list_for_each_entry(t, &xt_templates[af], list) {
+		if (strcmp(table->name, t->name))
+			continue;
+
+		list_del(&t->list);
+		mutex_unlock(&xt[af].mutex);
+		kfree(t);
+		return;
+	}
+
+	mutex_unlock(&xt[af].mutex);
+	WARN_ON_ONCE(1);
+}
+EXPORT_SYMBOL_GPL(xt_unregister_template);
+
 int xt_proto_init(struct net *net, u_int8_t af)
 {
 #ifdef CONFIG_PROC_FS
@@ -1937,6 +1998,7 @@ static int __init xt_init(void)
 #endif
 		INIT_LIST_HEAD(&xt[i].target);
 		INIT_LIST_HEAD(&xt[i].match);
+		INIT_LIST_HEAD(&xt_templates[i]);
 	}
 	rv = register_pernet_subsys(&xt_net_ops);
 	if (rv < 0)