[deliverable/linux.git] / net / netfilter / nf_conntrack_proto.c

/* L3/L4 protocol support for nf_conntrack. */

/* (C) 1999-2001 Paul `Rusty' Russell
 * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org>
 * (C) 2003,2004 USAGI/WIDE Project <http://www.linux-ipv6.org>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */

#include <linux/types.h>
#include <linux/netfilter.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/mutex.h>
#include <linux/vmalloc.h>
#include <linux/stddef.h>
#include <linux/err.h>
#include <linux/percpu.h>
#include <linux/notifier.h>
#include <linux/kernel.h>
#include <linux/netdevice.h>
#include <linux/rtnetlink.h>

#include <net/netfilter/nf_conntrack.h>
#include <net/netfilter/nf_conntrack_l3proto.h>
#include <net/netfilter/nf_conntrack_l4proto.h>
#include <net/netfilter/nf_conntrack_core.h>

static struct nf_conntrack_l4proto __rcu **nf_ct_protos[PF_MAX] __read_mostly;
struct nf_conntrack_l3proto __rcu *nf_ct_l3protos[AF_MAX] __read_mostly;
EXPORT_SYMBOL_GPL(nf_ct_l3protos);

static DEFINE_MUTEX(nf_ct_proto_mutex);

#ifdef CONFIG_SYSCTL
static int
nf_ct_register_sysctl(struct net *net,
		      struct ctl_table_header **header,
		      const char *path,
		      struct ctl_table *table)
{
	if (*header == NULL) {
		*header = register_net_sysctl(net, path, table);
		if (*header == NULL)
			return -ENOMEM;
	}

	return 0;
}

static void
nf_ct_unregister_sysctl(struct ctl_table_header **header,
			struct ctl_table **table,
			unsigned int users)
{
	if (users > 0)
		return;

	unregister_net_sysctl_table(*header);
	kfree(*table);
	*header = NULL;
	*table = NULL;
}
#endif

struct nf_conntrack_l4proto *
__nf_ct_l4proto_find(u_int16_t l3proto, u_int8_t l4proto)
{
	if (unlikely(l3proto >= AF_MAX || nf_ct_protos[l3proto] == NULL))
		return &nf_conntrack_l4proto_generic;

	return rcu_dereference(nf_ct_protos[l3proto][l4proto]);
}
EXPORT_SYMBOL_GPL(__nf_ct_l4proto_find);

/* this is guaranteed to always return a valid protocol helper, since
 * it falls back to generic_protocol */
struct nf_conntrack_l3proto *
nf_ct_l3proto_find_get(u_int16_t l3proto)
{
	struct nf_conntrack_l3proto *p;

	rcu_read_lock();
	p = __nf_ct_l3proto_find(l3proto);
	if (!try_module_get(p->me))
		p = &nf_conntrack_l3proto_generic;
	rcu_read_unlock();

	return p;
}
EXPORT_SYMBOL_GPL(nf_ct_l3proto_find_get);

void nf_ct_l3proto_put(struct nf_conntrack_l3proto *p)
{
	module_put(p->me);
}
EXPORT_SYMBOL_GPL(nf_ct_l3proto_put);

int
nf_ct_l3proto_try_module_get(unsigned short l3proto)
{
	int ret;
	struct nf_conntrack_l3proto *p;

retry:	p = nf_ct_l3proto_find_get(l3proto);
	if (p == &nf_conntrack_l3proto_generic) {
		ret = request_module("nf_conntrack-%d", l3proto);
		if (!ret)
			goto retry;

		return -EPROTOTYPE;
	}

	return 0;
}
EXPORT_SYMBOL_GPL(nf_ct_l3proto_try_module_get);

void nf_ct_l3proto_module_put(unsigned short l3proto)
{
	struct nf_conntrack_l3proto *p;

	/* rcu_read_lock not necessary since the caller holds a reference, but
	 * taken anyways to avoid lockdep warnings in __nf_ct_l3proto_find()
	 */
	rcu_read_lock();
	p = __nf_ct_l3proto_find(l3proto);
	module_put(p->me);
	rcu_read_unlock();
}
EXPORT_SYMBOL_GPL(nf_ct_l3proto_module_put);

struct nf_conntrack_l4proto *
nf_ct_l4proto_find_get(u_int16_t l3num, u_int8_t l4num)
{
	struct nf_conntrack_l4proto *p;

	rcu_read_lock();
	p = __nf_ct_l4proto_find(l3num, l4num);
	if (!try_module_get(p->me))
		p = &nf_conntrack_l4proto_generic;
	rcu_read_unlock();

	return p;
}
EXPORT_SYMBOL_GPL(nf_ct_l4proto_find_get);

void nf_ct_l4proto_put(struct nf_conntrack_l4proto *p)
{
	module_put(p->me);
}
EXPORT_SYMBOL_GPL(nf_ct_l4proto_put);

static int kill_l3proto(struct nf_conn *i, void *data)
{
	return nf_ct_l3num(i) == ((struct nf_conntrack_l3proto *)data)->l3proto;
}

static int kill_l4proto(struct nf_conn *i, void *data)
{
	struct nf_conntrack_l4proto *l4proto;
	l4proto = (struct nf_conntrack_l4proto *)data;
	return nf_ct_protonum(i) == l4proto->l4proto &&
	       nf_ct_l3num(i) == l4proto->l3proto;
}

static struct nf_ip_net *nf_ct_l3proto_net(struct net *net,
					   struct nf_conntrack_l3proto *l3proto)
{
	if (l3proto->l3proto == PF_INET)
		return &net->ct.nf_ct_proto;
	else
		return NULL;
}

static int nf_ct_l3proto_register_sysctl(struct net *net,
					 struct nf_conntrack_l3proto *l3proto)
{
	int err = 0;
	struct nf_ip_net *in = nf_ct_l3proto_net(net, l3proto);
	/* nf_conntrack_l3proto_ipv6 doesn't support sysctl */
	if (in == NULL)
		return 0;

#if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
	if (in->ctl_table != NULL) {
		err = nf_ct_register_sysctl(net,
					    &in->ctl_table_header,
					    l3proto->ctl_table_path,
					    in->ctl_table);
		if (err < 0) {
			kfree(in->ctl_table);
			in->ctl_table = NULL;
		}
	}
#endif
	return err;
}

static void nf_ct_l3proto_unregister_sysctl(struct net *net,
					    struct nf_conntrack_l3proto *l3proto)
{
	struct nf_ip_net *in = nf_ct_l3proto_net(net, l3proto);

	if (in == NULL)
		return;
#if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
	if (in->ctl_table_header != NULL)
		nf_ct_unregister_sysctl(&in->ctl_table_header,
					&in->ctl_table,
					0);
#endif
}

static int
nf_conntrack_l3proto_register_net(struct nf_conntrack_l3proto *proto)
{
	int ret = 0;
	struct nf_conntrack_l3proto *old;

	if (proto->l3proto >= AF_MAX)
		return -EBUSY;

	if (proto->tuple_to_nlattr && !proto->nlattr_tuple_size)
		return -EINVAL;

	mutex_lock(&nf_ct_proto_mutex);
	old = rcu_dereference_protected(nf_ct_l3protos[proto->l3proto],
					lockdep_is_held(&nf_ct_proto_mutex));
	if (old != &nf_conntrack_l3proto_generic) {
		ret = -EBUSY;
		goto out_unlock;
	}

	if (proto->nlattr_tuple_size)
		proto->nla_size = 3 * proto->nlattr_tuple_size();

	rcu_assign_pointer(nf_ct_l3protos[proto->l3proto], proto);

out_unlock:
	mutex_unlock(&nf_ct_proto_mutex);
	return ret;

}

int nf_conntrack_l3proto_register(struct net *net,
				  struct nf_conntrack_l3proto *proto)
{
	int ret = 0;

	if (proto->init_net) {
		ret = proto->init_net(net);
		if (ret < 0)
			return ret;
	}

	ret = nf_ct_l3proto_register_sysctl(net, proto);
	if (ret < 0)
		return ret;

	if (net == &init_net) {
		ret = nf_conntrack_l3proto_register_net(proto);
		if (ret < 0)
			nf_ct_l3proto_unregister_sysctl(net, proto);
	}

	return ret;
}
EXPORT_SYMBOL_GPL(nf_conntrack_l3proto_register);

static void
nf_conntrack_l3proto_unregister_net(struct nf_conntrack_l3proto *proto)
{
	BUG_ON(proto->l3proto >= AF_MAX);

	mutex_lock(&nf_ct_proto_mutex);
	BUG_ON(rcu_dereference_protected(nf_ct_l3protos[proto->l3proto],
					 lockdep_is_held(&nf_ct_proto_mutex)
					 ) != proto);
	rcu_assign_pointer(nf_ct_l3protos[proto->l3proto],
			   &nf_conntrack_l3proto_generic);
	mutex_unlock(&nf_ct_proto_mutex);

	synchronize_rcu();
}

void nf_conntrack_l3proto_unregister(struct net *net,
				     struct nf_conntrack_l3proto *proto)
{
	if (net == &init_net)
		nf_conntrack_l3proto_unregister_net(proto);

	nf_ct_l3proto_unregister_sysctl(net, proto);

	/* Remove all contrack entries for this protocol */
	rtnl_lock();
	nf_ct_iterate_cleanup(net, kill_l3proto, proto);
	rtnl_unlock();
}
EXPORT_SYMBOL_GPL(nf_conntrack_l3proto_unregister);

static struct nf_proto_net *nf_ct_l4proto_net(struct net *net,
					      struct nf_conntrack_l4proto *l4proto)
{
	switch (l4proto->l4proto) {
	case IPPROTO_TCP:
		return (struct nf_proto_net *)&net->ct.nf_ct_proto.tcp;
	case IPPROTO_UDP:
		return (struct nf_proto_net *)&net->ct.nf_ct_proto.udp;
	case IPPROTO_ICMP:
		return (struct nf_proto_net *)&net->ct.nf_ct_proto.icmp;
	case IPPROTO_ICMPV6:
		return (struct nf_proto_net *)&net->ct.nf_ct_proto.icmpv6;
	case 255: /* l4proto_generic */
		return (struct nf_proto_net *)&net->ct.nf_ct_proto.generic;
	default:
		if (l4proto->net_id)
			return net_generic(net, *l4proto->net_id);
		else
			return NULL;
	}
	return NULL;
}

static
int nf_ct_l4proto_register_sysctl(struct net *net,
				  struct nf_proto_net *pn,
				  struct nf_conntrack_l4proto *l4proto)
{
	int err = 0;

#ifdef CONFIG_SYSCTL
	if (pn->ctl_table != NULL) {
		err = nf_ct_register_sysctl(net,
					    &pn->ctl_table_header,
					    "net/netfilter",
					    pn->ctl_table);
		if (err < 0) {
			if (!pn->users) {
				kfree(pn->ctl_table);
				pn->ctl_table = NULL;
			}
			goto out;
		}
	}
#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
	if (l4proto->l3proto != AF_INET6 && pn->ctl_compat_table != NULL) {
		err = nf_ct_register_sysctl(net,
					    &pn->ctl_compat_header,
					    "net/ipv4/netfilter",
					    pn->ctl_compat_table);
		if (err == 0)
			goto out;

		nf_ct_kfree_compat_sysctl_table(pn);
		nf_ct_unregister_sysctl(&pn->ctl_table_header,
					&pn->ctl_table,
					pn->users);
	}
#endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
out:
#endif /* CONFIG_SYSCTL */
	return err;
}

static
void nf_ct_l4proto_unregister_sysctl(struct net *net,
				     struct nf_proto_net *pn,
				     struct nf_conntrack_l4proto *l4proto)
{
#ifdef CONFIG_SYSCTL
	if (pn->ctl_table_header != NULL)
		nf_ct_unregister_sysctl(&pn->ctl_table_header,
					&pn->ctl_table,
					pn->users);

#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
	if (l4proto->l3proto != AF_INET6 && pn->ctl_compat_header != NULL)
		nf_ct_unregister_sysctl(&pn->ctl_compat_header,
					&pn->ctl_compat_table,
					0);
#endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
#endif /* CONFIG_SYSCTL */
}

/* FIXME: Allow NULL functions and sub in pointers to generic for
   them. --RR */
static int
nf_conntrack_l4proto_register_net(struct nf_conntrack_l4proto *l4proto)
{
	int ret = 0;

	if (l4proto->l3proto >= PF_MAX)
		return -EBUSY;

	if ((l4proto->to_nlattr && !l4proto->nlattr_size)
		|| (l4proto->tuple_to_nlattr && !l4proto->nlattr_tuple_size))
		return -EINVAL;

	mutex_lock(&nf_ct_proto_mutex);
	if (!nf_ct_protos[l4proto->l3proto]) {
		/* l3proto may be loaded latter. */
		struct nf_conntrack_l4proto __rcu **proto_array;
		int i;

		proto_array = kmalloc(MAX_NF_CT_PROTO *
				      sizeof(struct nf_conntrack_l4proto *),
				      GFP_KERNEL);
		if (proto_array == NULL) {
			ret = -ENOMEM;
			goto out_unlock;
		}

		for (i = 0; i < MAX_NF_CT_PROTO; i++)
			RCU_INIT_POINTER(proto_array[i], &nf_conntrack_l4proto_generic);

		/* Before making proto_array visible to lockless readers,
		 * we must make sure its content is committed to memory.
		 */
		smp_wmb();

		nf_ct_protos[l4proto->l3proto] = proto_array;
	} else if (rcu_dereference_protected(
			nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
			lockdep_is_held(&nf_ct_proto_mutex)
			) != &nf_conntrack_l4proto_generic) {
		ret = -EBUSY;
		goto out_unlock;
	}

	l4proto->nla_size = 0;
	if (l4proto->nlattr_size)
		l4proto->nla_size += l4proto->nlattr_size();
	if (l4proto->nlattr_tuple_size)
		l4proto->nla_size += 3 * l4proto->nlattr_tuple_size();

	rcu_assign_pointer(nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
			   l4proto);
out_unlock:
	mutex_unlock(&nf_ct_proto_mutex);
	return ret;
}

int nf_conntrack_l4proto_register(struct net *net,
				  struct nf_conntrack_l4proto *l4proto)
{
	int ret = 0;
	struct nf_proto_net *pn = NULL;

	if (l4proto->init_net) {
		ret = l4proto->init_net(net, l4proto->l3proto);
		if (ret < 0)
			goto out;
	}

	pn = nf_ct_l4proto_net(net, l4proto);
	if (pn == NULL)
		goto out;

	ret = nf_ct_l4proto_register_sysctl(net, pn, l4proto);
	if (ret < 0)
		goto out;

	if (net == &init_net) {
		ret = nf_conntrack_l4proto_register_net(l4proto);
		if (ret < 0) {
			nf_ct_l4proto_unregister_sysctl(net, pn, l4proto);
			goto out;
		}
	}

	pn->users++;
out:
	return ret;
}
EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_register);

static void
nf_conntrack_l4proto_unregister_net(struct nf_conntrack_l4proto *l4proto)
{
	BUG_ON(l4proto->l3proto >= PF_MAX);

	mutex_lock(&nf_ct_proto_mutex);
	BUG_ON(rcu_dereference_protected(
			nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
			lockdep_is_held(&nf_ct_proto_mutex)
			) != l4proto);
	rcu_assign_pointer(nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
			   &nf_conntrack_l4proto_generic);
	mutex_unlock(&nf_ct_proto_mutex);

	synchronize_rcu();
}

void nf_conntrack_l4proto_unregister(struct net *net,
				     struct nf_conntrack_l4proto *l4proto)
{
	struct nf_proto_net *pn = NULL;

	if (net == &init_net)
		nf_conntrack_l4proto_unregister_net(l4proto);

	pn = nf_ct_l4proto_net(net, l4proto);
	if (pn == NULL)
		return;

	pn->users--;
	nf_ct_l4proto_unregister_sysctl(net, pn, l4proto);

	/* Remove all contrack entries for this protocol */
	rtnl_lock();
	nf_ct_iterate_cleanup(net, kill_l4proto, l4proto);
	rtnl_unlock();
}
EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_unregister);

int nf_conntrack_proto_init(struct net *net)
{
	unsigned int i;
	int err;
	struct nf_proto_net *pn = nf_ct_l4proto_net(net,
					&nf_conntrack_l4proto_generic);

	err = nf_conntrack_l4proto_generic.init_net(net,
					nf_conntrack_l4proto_generic.l3proto);
	if (err < 0)
		return err;
	err = nf_ct_l4proto_register_sysctl(net,
					    pn,
					    &nf_conntrack_l4proto_generic);
	if (err < 0)
		return err;

	if (net == &init_net) {
		for (i = 0; i < AF_MAX; i++)
			rcu_assign_pointer(nf_ct_l3protos[i],
					   &nf_conntrack_l3proto_generic);
	}

	pn->users++;
	return 0;
}

void nf_conntrack_proto_fini(struct net *net)
{
	unsigned int i;
	struct nf_proto_net *pn = nf_ct_l4proto_net(net,
					&nf_conntrack_l4proto_generic);

	pn->users--;
	nf_ct_l4proto_unregister_sysctl(net,
					pn,
					&nf_conntrack_l4proto_generic);
	if (net == &init_net) {
		/* free l3proto protocol tables */
		for (i = 0; i < PF_MAX; i++)
			kfree(nf_ct_protos[i]);
	}
}
Commit	Line	Data
8f03dea5 MJ	1	/* L3/L4 protocol support for nf_conntrack. */
	2
	3	/* (C) 1999-2001 Paul `Rusty' Russell
	4	* (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org>
	5	* (C) 2003,2004 USAGI/WIDE Project <http://www.linux-ipv6.org>
	6	*
	7	* This program is free software; you can redistribute it and/or modify
	8	* it under the terms of the GNU General Public License version 2 as
	9	* published by the Free Software Foundation.
	10	*/
	11
	12	#include <linux/types.h>
	13	#include <linux/netfilter.h>
	14	#include <linux/module.h>
5a0e3ad6	15	#include <linux/slab.h>
d62f9ed4	16	#include <linux/mutex.h>
8f03dea5 MJ	17	#include <linux/vmalloc.h>
	18	#include <linux/stddef.h>
	19	#include <linux/err.h>
	20	#include <linux/percpu.h>
8f03dea5 MJ	21	#include <linux/notifier.h>
	22	#include <linux/kernel.h>
	23	#include <linux/netdevice.h>
efb9a8c2	24	#include <linux/rtnetlink.h>
8f03dea5 MJ	25
	26	#include <net/netfilter/nf_conntrack.h>
	27	#include <net/netfilter/nf_conntrack_l3proto.h>
605dcad6	28	#include <net/netfilter/nf_conntrack_l4proto.h>
8f03dea5 MJ	29	#include <net/netfilter/nf_conntrack_core.h>
8f03dea5 MJ	30
0906a372 AB	31	static struct nf_conntrack_l4proto __rcu **nf_ct_protos[PF_MAX] __read_mostly;
0906a372 AB	32	struct nf_conntrack_l3proto __rcu *nf_ct_l3protos[AF_MAX] __read_mostly;
13b18339	33	EXPORT_SYMBOL_GPL(nf_ct_l3protos);
8f03dea5	34
b19caa0c	35	static DEFINE_MUTEX(nf_ct_proto_mutex);
d62f9ed4	36
b19caa0c	37	#ifdef CONFIG_SYSCTL
d62f9ed4	38	static int
2c352f44 G	39	nf_ct_register_sysctl(struct net *net,
	40	struct ctl_table_header **header,
	41	const char *path,
fa34fff5	42	struct ctl_table *table)
d62f9ed4 PM	43	{
d62f9ed4 PM	44	if (*header == NULL) {
2c352f44	45	*header = register_net_sysctl(net, path, table);
d62f9ed4 PM	46	if (*header == NULL)
	47	return -ENOMEM;
	48	}
2c352f44	49
d62f9ed4 PM	50	return 0;
	51	}
	52
	53	static void
	54	nf_ct_unregister_sysctl(struct ctl_table_header **header,
2c352f44	55	struct ctl_table **table,
fa34fff5	56	unsigned int users)
d62f9ed4	57	{
fa34fff5	58	if (users > 0)
d62f9ed4	59	return;
b3fd3ffe	60
5dd3df10	61	unregister_net_sysctl_table(*header);
2c352f44	62	kfree(*table);
d62f9ed4	63	*header = NULL;
2c352f44	64	*table = NULL;
d62f9ed4 PM	65	}
	66	#endif
	67
605dcad6 MJ	68	struct nf_conntrack_l4proto *
605dcad6 MJ	69	__nf_ct_l4proto_find(u_int16_t l3proto, u_int8_t l4proto)
8f03dea5 MJ	70	{
8f03dea5 MJ	71	if (unlikely(l3proto >= AF_MAX \|\| nf_ct_protos[l3proto] == NULL))
605dcad6	72	return &nf_conntrack_l4proto_generic;
8f03dea5	73
923f4902	74	return rcu_dereference(nf_ct_protos[l3proto][l4proto]);
8f03dea5	75	}
13b18339	76	EXPORT_SYMBOL_GPL(__nf_ct_l4proto_find);
8f03dea5 MJ	77
	78	/* this is guaranteed to always return a valid protocol helper, since
	79	* it falls back to generic_protocol */
8f03dea5 MJ	80	struct nf_conntrack_l3proto *
	81	nf_ct_l3proto_find_get(u_int16_t l3proto)
	82	{
	83	struct nf_conntrack_l3proto *p;
	84
923f4902	85	rcu_read_lock();
8f03dea5 MJ	86	p = __nf_ct_l3proto_find(l3proto);
8f03dea5 MJ	87	if (!try_module_get(p->me))
605dcad6	88	p = &nf_conntrack_l3proto_generic;
923f4902	89	rcu_read_unlock();
8f03dea5 MJ	90
	91	return p;
	92	}
13b18339	93	EXPORT_SYMBOL_GPL(nf_ct_l3proto_find_get);
8f03dea5 MJ	94
	95	void nf_ct_l3proto_put(struct nf_conntrack_l3proto *p)
	96	{
	97	module_put(p->me);
	98	}
13b18339	99	EXPORT_SYMBOL_GPL(nf_ct_l3proto_put);
8f03dea5 MJ	100
	101	int
	102	nf_ct_l3proto_try_module_get(unsigned short l3proto)
	103	{
	104	int ret;
	105	struct nf_conntrack_l3proto *p;
	106
	107	retry: p = nf_ct_l3proto_find_get(l3proto);
605dcad6	108	if (p == &nf_conntrack_l3proto_generic) {
8f03dea5 MJ	109	ret = request_module("nf_conntrack-%d", l3proto);
	110	if (!ret)
	111	goto retry;
	112
	113	return -EPROTOTYPE;
	114	}
	115
	116	return 0;
	117	}
13b18339	118	EXPORT_SYMBOL_GPL(nf_ct_l3proto_try_module_get);
8f03dea5 MJ	119
	120	void nf_ct_l3proto_module_put(unsigned short l3proto)
	121	{
	122	struct nf_conntrack_l3proto *p;
	123
3b254c54 PM	124	/* rcu_read_lock not necessary since the caller holds a reference, but
	125	* taken anyways to avoid lockdep warnings in __nf_ct_l3proto_find()
	126	*/
	127	rcu_read_lock();
8f03dea5	128	p = __nf_ct_l3proto_find(l3proto);
8f03dea5	129	module_put(p->me);
3b254c54	130	rcu_read_unlock();
8f03dea5	131	}
13b18339	132	EXPORT_SYMBOL_GPL(nf_ct_l3proto_module_put);
8f03dea5	133
c1ebd7df PNA	134	struct nf_conntrack_l4proto *
	135	nf_ct_l4proto_find_get(u_int16_t l3num, u_int8_t l4num)
	136	{
	137	struct nf_conntrack_l4proto *p;
	138
	139	rcu_read_lock();
	140	p = __nf_ct_l4proto_find(l3num, l4num);
	141	if (!try_module_get(p->me))
	142	p = &nf_conntrack_l4proto_generic;
	143	rcu_read_unlock();
	144
	145	return p;
	146	}
	147	EXPORT_SYMBOL_GPL(nf_ct_l4proto_find_get);
	148
	149	void nf_ct_l4proto_put(struct nf_conntrack_l4proto *p)
	150	{
	151	module_put(p->me);
	152	}
	153	EXPORT_SYMBOL_GPL(nf_ct_l4proto_put);
	154
8f03dea5 MJ	155	static int kill_l3proto(struct nf_conn i, void data)
8f03dea5 MJ	156	{
5e8fbe2a	157	return nf_ct_l3num(i) == ((struct nf_conntrack_l3proto *)data)->l3proto;
8f03dea5 MJ	158	}
8f03dea5 MJ	159
605dcad6	160	static int kill_l4proto(struct nf_conn i, void data)
8f03dea5	161	{
605dcad6 MJ	162	struct nf_conntrack_l4proto *l4proto;
605dcad6 MJ	163	l4proto = (struct nf_conntrack_l4proto *)data;
5e8fbe2a PM	164	return nf_ct_protonum(i) == l4proto->l4proto &&
5e8fbe2a PM	165	nf_ct_l3num(i) == l4proto->l3proto;
8f03dea5 MJ	166	}
8f03dea5 MJ	167
524a53e5 G	168	static struct nf_ip_net nf_ct_l3proto_net(struct net net,
524a53e5 G	169	struct nf_conntrack_l3proto *l3proto)
d62f9ed4	170	{
524a53e5 G	171	if (l3proto->l3proto == PF_INET)
	172	return &net->ct.nf_ct_proto;
	173	else
	174	return NULL;
	175	}
d62f9ed4	176
524a53e5 G	177	static int nf_ct_l3proto_register_sysctl(struct net *net,
	178	struct nf_conntrack_l3proto *l3proto)
	179	{
	180	int err = 0;
	181	struct nf_ip_net *in = nf_ct_l3proto_net(net, l3proto);
	182	/* nf_conntrack_l3proto_ipv6 doesn't support sysctl */
	183	if (in == NULL)
	184	return 0;
	185
	186	#if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
	187	if (in->ctl_table != NULL) {
	188	err = nf_ct_register_sysctl(net,
	189	&in->ctl_table_header,
d62f9ed4	190	l3proto->ctl_table_path,
fa34fff5	191	in->ctl_table);
524a53e5 G	192	if (err < 0) {
	193	kfree(in->ctl_table);
	194	in->ctl_table = NULL;
	195	}
d62f9ed4	196	}
d62f9ed4 PM	197	#endif
	198	return err;
	199	}
	200
524a53e5 G	201	static void nf_ct_l3proto_unregister_sysctl(struct net *net,
524a53e5 G	202	struct nf_conntrack_l3proto *l3proto)
d62f9ed4	203	{
524a53e5 G	204	struct nf_ip_net *in = nf_ct_l3proto_net(net, l3proto);
	205
	206	if (in == NULL)
	207	return;
	208	#if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
	209	if (in->ctl_table_header != NULL)
	210	nf_ct_unregister_sysctl(&in->ctl_table_header,
	211	&in->ctl_table,
fa34fff5	212	0);
d62f9ed4 PM	213	#endif
	214	}
	215
524a53e5 G	216	static int
524a53e5 G	217	nf_conntrack_l3proto_register_net(struct nf_conntrack_l3proto *proto)
8f03dea5 MJ	218	{
8f03dea5 MJ	219	int ret = 0;
0e60ebe0	220	struct nf_conntrack_l3proto *old;
8f03dea5	221
0661cca9 PM	222	if (proto->l3proto >= AF_MAX)
0661cca9 PM	223	return -EBUSY;
ae5718fb	224
d0dba725 HE	225	if (proto->tuple_to_nlattr && !proto->nlattr_tuple_size)
	226	return -EINVAL;
	227
b19caa0c	228	mutex_lock(&nf_ct_proto_mutex);
0e60ebe0 ED	229	old = rcu_dereference_protected(nf_ct_l3protos[proto->l3proto],
	230	lockdep_is_held(&nf_ct_proto_mutex));
	231	if (old != &nf_conntrack_l3proto_generic) {
8f03dea5	232	ret = -EBUSY;
ae5718fb	233	goto out_unlock;
8f03dea5	234	}
d62f9ed4	235
d0dba725 HE	236	if (proto->nlattr_tuple_size)
	237	proto->nla_size = 3 * proto->nlattr_tuple_size();
	238
0661cca9	239	rcu_assign_pointer(nf_ct_l3protos[proto->l3proto], proto);
8f03dea5	240
ae5718fb	241	out_unlock:
b19caa0c	242	mutex_unlock(&nf_ct_proto_mutex);
8f03dea5	243	return ret;
524a53e5	244
8f03dea5 MJ	245	}
8f03dea5 MJ	246
524a53e5 G	247	int nf_conntrack_l3proto_register(struct net *net,
524a53e5 G	248	struct nf_conntrack_l3proto *proto)
8f03dea5	249	{
524a53e5 G	250	int ret = 0;
524a53e5 G	251
fa0f61f0 G	252	if (proto->init_net) {
	253	ret = proto->init_net(net);
	254	if (ret < 0)
	255	return ret;
	256	}
524a53e5	257
fa0f61f0	258	ret = nf_ct_l3proto_register_sysctl(net, proto);
524a53e5 G	259	if (ret < 0)
	260	return ret;
	261
fa0f61f0 G	262	if (net == &init_net) {
fa0f61f0 G	263	ret = nf_conntrack_l3proto_register_net(proto);
524a53e5	264	if (ret < 0)
fa0f61f0	265	nf_ct_l3proto_unregister_sysctl(net, proto);
524a53e5	266	}
fa0f61f0 G	267
fa0f61f0 G	268	return ret;
524a53e5 G	269	}
524a53e5 G	270	EXPORT_SYMBOL_GPL(nf_conntrack_l3proto_register);
678d6675	271
524a53e5 G	272	static void
	273	nf_conntrack_l3proto_unregister_net(struct nf_conntrack_l3proto *proto)
	274	{
fe3eb20c	275	BUG_ON(proto->l3proto >= AF_MAX);
ae5718fb	276
b19caa0c	277	mutex_lock(&nf_ct_proto_mutex);
0e60ebe0 ED	278	BUG_ON(rcu_dereference_protected(nf_ct_l3protos[proto->l3proto],
	279	lockdep_is_held(&nf_ct_proto_mutex)
	280	) != proto);
923f4902 PM	281	rcu_assign_pointer(nf_ct_l3protos[proto->l3proto],
923f4902 PM	282	&nf_conntrack_l3proto_generic);
b19caa0c	283	mutex_unlock(&nf_ct_proto_mutex);
8f03dea5	284
0661cca9	285	synchronize_rcu();
524a53e5 G	286	}
	287
	288	void nf_conntrack_l3proto_unregister(struct net *net,
	289	struct nf_conntrack_l3proto *proto)
	290	{
	291	if (net == &init_net)
	292	nf_conntrack_l3proto_unregister_net(proto);
	293
	294	nf_ct_l3proto_unregister_sysctl(net, proto);
d62f9ed4	295
8f03dea5	296	/* Remove all contrack entries for this protocol */
efb9a8c2	297	rtnl_lock();
524a53e5	298	nf_ct_iterate_cleanup(net, kill_l3proto, proto);
efb9a8c2	299	rtnl_unlock();
8f03dea5	300	}
13b18339	301	EXPORT_SYMBOL_GPL(nf_conntrack_l3proto_unregister);
8f03dea5	302
2c352f44 G	303	static struct nf_proto_net nf_ct_l4proto_net(struct net net,
	304	struct nf_conntrack_l4proto *l4proto)
	305	{
15f585bd	306	switch (l4proto->l4proto) {
d2ba1fde G	307	case IPPROTO_TCP:
d2ba1fde G	308	return (struct nf_proto_net *)&net->ct.nf_ct_proto.tcp;
0ce490ad G	309	case IPPROTO_UDP:
0ce490ad G	310	return (struct nf_proto_net *)&net->ct.nf_ct_proto.udp;
4b626b9c G	311	case IPPROTO_ICMP:
4b626b9c G	312	return (struct nf_proto_net *)&net->ct.nf_ct_proto.icmp;
7080ba09 G	313	case IPPROTO_ICMPV6:
7080ba09 G	314	return (struct nf_proto_net *)&net->ct.nf_ct_proto.icmpv6;
15f585bd G	315	case 255: /* l4proto_generic */
	316	return (struct nf_proto_net *)&net->ct.nf_ct_proto.generic;
	317	default:
	318	if (l4proto->net_id)
	319	return net_generic(net, *l4proto->net_id);
	320	else
	321	return NULL;
	322	}
	323	return NULL;
2c352f44 G	324	}
	325
	326	static
	327	int nf_ct_l4proto_register_sysctl(struct net *net,
fa34fff5	328	struct nf_proto_net *pn,
2c352f44	329	struct nf_conntrack_l4proto *l4proto)
d62f9ed4 PM	330	{
	331	int err = 0;
	332
	333	#ifdef CONFIG_SYSCTL
2c352f44 G	334	if (pn->ctl_table != NULL) {
	335	err = nf_ct_register_sysctl(net,
	336	&pn->ctl_table_header,
f99e8f71	337	"net/netfilter",
fa34fff5	338	pn->ctl_table);
2c352f44 G	339	if (err < 0) {
	340	if (!pn->users) {
	341	kfree(pn->ctl_table);
	342	pn->ctl_table = NULL;
	343	}
a999e683	344	goto out;
2c352f44	345	}
d62f9ed4	346	}
a999e683	347	#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
2c352f44 G	348	if (l4proto->l3proto != AF_INET6 && pn->ctl_compat_table != NULL) {
	349	err = nf_ct_register_sysctl(net,
	350	&pn->ctl_compat_header,
f99e8f71	351	"net/ipv4/netfilter",
fa34fff5	352	pn->ctl_compat_table);
a999e683 PM	353	if (err == 0)
a999e683 PM	354	goto out;
2c352f44	355
f28997e2	356	nf_ct_kfree_compat_sysctl_table(pn);
2c352f44 G	357	nf_ct_unregister_sysctl(&pn->ctl_table_header,
2c352f44 G	358	&pn->ctl_table,
fa34fff5	359	pn->users);
a999e683 PM	360	}
	361	#endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
	362	out:
933a41e7	363	#endif /* CONFIG_SYSCTL */
d62f9ed4 PM	364	return err;
	365	}
	366
2c352f44 G	367	static
2c352f44 G	368	void nf_ct_l4proto_unregister_sysctl(struct net *net,
fa34fff5	369	struct nf_proto_net *pn,
2c352f44	370	struct nf_conntrack_l4proto *l4proto)
d62f9ed4 PM	371	{
d62f9ed4 PM	372	#ifdef CONFIG_SYSCTL
2c352f44 G	373	if (pn->ctl_table_header != NULL)
	374	nf_ct_unregister_sysctl(&pn->ctl_table_header,
	375	&pn->ctl_table,
fa34fff5	376	pn->users);
2c352f44	377
a999e683	378	#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
2c352f44 G	379	if (l4proto->l3proto != AF_INET6 && pn->ctl_compat_header != NULL)
	380	nf_ct_unregister_sysctl(&pn->ctl_compat_header,
	381	&pn->ctl_compat_table,
fa34fff5	382	0);
a999e683	383	#endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
933a41e7	384	#endif /* CONFIG_SYSCTL */
d62f9ed4 PM	385	}
d62f9ed4 PM	386
8f03dea5 MJ	387	/* FIXME: Allow NULL functions and sub in pointers to generic for
8f03dea5 MJ	388	them. --RR */
2c352f44 G	389	static int
2c352f44 G	390	nf_conntrack_l4proto_register_net(struct nf_conntrack_l4proto *l4proto)
8f03dea5 MJ	391	{
	392	int ret = 0;
	393
0661cca9 PM	394	if (l4proto->l3proto >= PF_MAX)
0661cca9 PM	395	return -EBUSY;
ae5718fb	396
d0dba725 HE	397	if ((l4proto->to_nlattr && !l4proto->nlattr_size)
	398	\|\| (l4proto->tuple_to_nlattr && !l4proto->nlattr_tuple_size))
	399	return -EINVAL;
	400
b19caa0c	401	mutex_lock(&nf_ct_proto_mutex);
c6a1e615	402	if (!nf_ct_protos[l4proto->l3proto]) {
8f03dea5	403	/* l3proto may be loaded latter. */
c5d277d2	404	struct nf_conntrack_l4proto __rcu **proto_array;
8f03dea5 MJ	405	int i;
8f03dea5 MJ	406
c6a1e615 PM	407	proto_array = kmalloc(MAX_NF_CT_PROTO *
	408	sizeof(struct nf_conntrack_l4proto *),
	409	GFP_KERNEL);
8f03dea5 MJ	410	if (proto_array == NULL) {
8f03dea5 MJ	411	ret = -ENOMEM;
b19caa0c	412	goto out_unlock;
8f03dea5	413	}
c6a1e615	414
8f03dea5	415	for (i = 0; i < MAX_NF_CT_PROTO; i++)
c5d277d2	416	RCU_INIT_POINTER(proto_array[i], &nf_conntrack_l4proto_generic);
d817d29d ED	417
	418	/* Before making proto_array visible to lockless readers,
	419	* we must make sure its content is committed to memory.
	420	*/
	421	smp_wmb();
	422
c6a1e615	423	nf_ct_protos[l4proto->l3proto] = proto_array;
0e60ebe0 ED	424	} else if (rcu_dereference_protected(
	425	nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
	426	lockdep_is_held(&nf_ct_proto_mutex)
	427	) != &nf_conntrack_l4proto_generic) {
c6a1e615 PM	428	ret = -EBUSY;
c6a1e615 PM	429	goto out_unlock;
8f03dea5 MJ	430	}
8f03dea5 MJ	431
d0dba725 HE	432	l4proto->nla_size = 0;
	433	if (l4proto->nlattr_size)
	434	l4proto->nla_size += l4proto->nlattr_size();
	435	if (l4proto->nlattr_tuple_size)
	436	l4proto->nla_size += 3 * l4proto->nlattr_tuple_size();
	437
c6a1e615 PM	438	rcu_assign_pointer(nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
c6a1e615 PM	439	l4proto);
8f03dea5	440	out_unlock:
b19caa0c	441	mutex_unlock(&nf_ct_proto_mutex);
8f03dea5 MJ	442	return ret;
	443	}
	444
2c352f44 G	445	int nf_conntrack_l4proto_register(struct net *net,
2c352f44 G	446	struct nf_conntrack_l4proto *l4proto)
8f03dea5	447	{
2c352f44	448	int ret = 0;
fa34fff5	449	struct nf_proto_net *pn = NULL;
2c352f44	450
fa0f61f0	451	if (l4proto->init_net) {
f1caad27	452	ret = l4proto->init_net(net, l4proto->l3proto);
fa0f61f0	453	if (ret < 0)
fa34fff5	454	goto out;
fa0f61f0	455	}
678d6675	456
fa34fff5 G	457	pn = nf_ct_l4proto_net(net, l4proto);
	458	if (pn == NULL)
	459	goto out;
	460
	461	ret = nf_ct_l4proto_register_sysctl(net, pn, l4proto);
2c352f44	462	if (ret < 0)
fa34fff5	463	goto out;
2c352f44	464
fa0f61f0 G	465	if (net == &init_net) {
fa0f61f0 G	466	ret = nf_conntrack_l4proto_register_net(l4proto);
fa34fff5 G	467	if (ret < 0) {
	468	nf_ct_l4proto_unregister_sysctl(net, pn, l4proto);
	469	goto out;
	470	}
fa0f61f0 G	471	}
fa0f61f0 G	472
fa34fff5 G	473	pn->users++;
fa34fff5 G	474	out:
fa0f61f0	475	return ret;
2c352f44 G	476	}
	477	EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_register);
	478
	479	static void
	480	nf_conntrack_l4proto_unregister_net(struct nf_conntrack_l4proto *l4proto)
	481	{
fe3eb20c	482	BUG_ON(l4proto->l3proto >= PF_MAX);
ae5718fb	483
b19caa0c	484	mutex_lock(&nf_ct_proto_mutex);
0e60ebe0 ED	485	BUG_ON(rcu_dereference_protected(
	486	nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
	487	lockdep_is_held(&nf_ct_proto_mutex)
	488	) != l4proto);
923f4902 PM	489	rcu_assign_pointer(nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
923f4902 PM	490	&nf_conntrack_l4proto_generic);
b19caa0c	491	mutex_unlock(&nf_ct_proto_mutex);
8f03dea5	492
0661cca9	493	synchronize_rcu();
2c352f44	494	}
d62f9ed4	495
2c352f44 G	496	void nf_conntrack_l4proto_unregister(struct net *net,
	497	struct nf_conntrack_l4proto *l4proto)
	498	{
fa34fff5 G	499	struct nf_proto_net *pn = NULL;
fa34fff5 G	500
2c352f44 G	501	if (net == &init_net)
	502	nf_conntrack_l4proto_unregister_net(l4proto);
	503
fa34fff5 G	504	pn = nf_ct_l4proto_net(net, l4proto);
	505	if (pn == NULL)
	506	return;
	507
	508	pn->users--;
	509	nf_ct_l4proto_unregister_sysctl(net, pn, l4proto);
	510
8f03dea5	511	/* Remove all contrack entries for this protocol */
efb9a8c2	512	rtnl_lock();
2c352f44	513	nf_ct_iterate_cleanup(net, kill_l4proto, l4proto);
efb9a8c2	514	rtnl_unlock();
8f03dea5	515	}
13b18339	516	EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_unregister);
ac5357eb	517
15f585bd	518	int nf_conntrack_proto_init(struct net *net)
ac5357eb PM	519	{
	520	unsigned int i;
	521	int err;
fa34fff5 G	522	struct nf_proto_net *pn = nf_ct_l4proto_net(net,
	523	&nf_conntrack_l4proto_generic);
	524
f1caad27 G	525	err = nf_conntrack_l4proto_generic.init_net(net,
f1caad27 G	526	nf_conntrack_l4proto_generic.l3proto);
15f585bd G	527	if (err < 0)
	528	return err;
	529	err = nf_ct_l4proto_register_sysctl(net,
fa34fff5	530	pn,
15f585bd	531	&nf_conntrack_l4proto_generic);
ac5357eb PM	532	if (err < 0)
	533	return err;
	534
15f585bd G	535	if (net == &init_net) {
	536	for (i = 0; i < AF_MAX; i++)
	537	rcu_assign_pointer(nf_ct_l3protos[i],
	538	&nf_conntrack_l3proto_generic);
	539	}
fa34fff5 G	540
fa34fff5 G	541	pn->users++;
ac5357eb PM	542	return 0;
	543	}
	544
15f585bd	545	void nf_conntrack_proto_fini(struct net *net)
ac5357eb PM	546	{
ac5357eb PM	547	unsigned int i;
fa34fff5 G	548	struct nf_proto_net *pn = nf_ct_l4proto_net(net,
	549	&nf_conntrack_l4proto_generic);
	550
	551	pn->users--;
15f585bd	552	nf_ct_l4proto_unregister_sysctl(net,
fa34fff5	553	pn,
15f585bd G	554	&nf_conntrack_l4proto_generic);
	555	if (net == &init_net) {
	556	/* free l3proto protocol tables */
	557	for (i = 0; i < PF_MAX; i++)
	558	kfree(nf_ct_protos[i]);
	559	}
ac5357eb	560	}