[linux.git] / net / ipv6 / addrlabel.c

/*
 * IPv6 Address Label subsystem
 * for the IPv6 "Default" Source Address Selection
 *
 * Copyright (C)2007 USAGI/WIDE Project
 */
/*
 * Author:
 * 	YOSHIFUJI Hideaki @ USAGI/WIDE Project <[email protected]>
 */

#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/rcupdate.h>
#include <linux/in6.h>
#include <linux/slab.h>
#include <net/addrconf.h>
#include <linux/if_addrlabel.h>
#include <linux/netlink.h>
#include <linux/rtnetlink.h>

#if 0
#define ADDRLABEL(x...) printk(x)
#else
#define ADDRLABEL(x...) do { ; } while(0)
#endif

/*
 * Policy Table
 */
struct ip6addrlbl_entry
{
#ifdef CONFIG_NET_NS
	struct net *lbl_net;
#endif
	struct in6_addr prefix;
	int prefixlen;
	int ifindex;
	int addrtype;
	u32 label;
	struct hlist_node list;
	atomic_t refcnt;
	struct rcu_head rcu;
};

static struct ip6addrlbl_table
{
	struct hlist_head head;
	spinlock_t lock;
	u32 seq;
} ip6addrlbl_table;

static inline
struct net *ip6addrlbl_net(const struct ip6addrlbl_entry *lbl)
{
	return read_pnet(&lbl->lbl_net);
}

/*
 * Default policy table (RFC3484 + extensions)
 *
 * prefix		addr_type	label
 * -------------------------------------------------------------------------
 * ::1/128		LOOPBACK	0
 * ::/0			N/A		1
 * 2002::/16		N/A		2
 * ::/96		COMPATv4	3
 * ::ffff:0:0/96	V4MAPPED	4
 * fc00::/7		N/A		5		ULA (RFC 4193)
 * 2001::/32		N/A		6		Teredo (RFC 4380)
 * 2001:10::/28		N/A		7		ORCHID (RFC 4843)
 *
 * Note: 0xffffffff is used if we do not have any policies.
 */

#define IPV6_ADDR_LABEL_DEFAULT	0xffffffffUL

static const __net_initdata struct ip6addrlbl_init_table
{
	const struct in6_addr *prefix;
	int prefixlen;
	u32 label;
} ip6addrlbl_init_table[] = {
	{	/* ::/0 */
		.prefix = &in6addr_any,
		.label = 1,
	},{	/* fc00::/7 */
		.prefix = &(struct in6_addr){{{ 0xfc }}},
		.prefixlen = 7,
		.label = 5,
	},{	/* 2002::/16 */
		.prefix = &(struct in6_addr){{{ 0x20, 0x02 }}},
		.prefixlen = 16,
		.label = 2,
	},{	/* 2001::/32 */
		.prefix = &(struct in6_addr){{{ 0x20, 0x01 }}},
		.prefixlen = 32,
		.label = 6,
	},{	/* 2001:10::/28 */
		.prefix = &(struct in6_addr){{{ 0x20, 0x01, 0x00, 0x10 }}},
		.prefixlen = 28,
		.label = 7,
	},{	/* ::ffff:0:0 */
		.prefix = &(struct in6_addr){{{ [10] = 0xff, [11] = 0xff }}},
		.prefixlen = 96,
		.label = 4,
	},{	/* ::/96 */
		.prefix = &in6addr_any,
		.prefixlen = 96,
		.label = 3,
	},{	/* ::1/128 */
		.prefix = &in6addr_loopback,
		.prefixlen = 128,
		.label = 0,
	}
};

/* Object management */
static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p)
{
#ifdef CONFIG_NET_NS
	release_net(p->lbl_net);
#endif
	kfree(p);
}

static void ip6addrlbl_free_rcu(struct rcu_head *h)
{
	ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu));
}

static inline int ip6addrlbl_hold(struct ip6addrlbl_entry *p)
{
	return atomic_inc_not_zero(&p->refcnt);
}

static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p)
{
	if (atomic_dec_and_test(&p->refcnt))
		call_rcu(&p->rcu, ip6addrlbl_free_rcu);
}

/* Find label */
static int __ip6addrlbl_match(struct net *net,
			      struct ip6addrlbl_entry *p,
			      const struct in6_addr *addr,
			      int addrtype, int ifindex)
{
	if (!net_eq(ip6addrlbl_net(p), net))
		return 0;
	if (p->ifindex && p->ifindex != ifindex)
		return 0;
	if (p->addrtype && p->addrtype != addrtype)
		return 0;
	if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
		return 0;
	return 1;
}

static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
						  const struct in6_addr *addr,
						  int type, int ifindex)
{
	struct hlist_node *pos;
	struct ip6addrlbl_entry *p;
	hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) {
		if (__ip6addrlbl_match(net, p, addr, type, ifindex))
			return p;
	}
	return NULL;
}

u32 ipv6_addr_label(struct net *net,
		    const struct in6_addr *addr, int type, int ifindex)
{
	u32 label;
	struct ip6addrlbl_entry *p;

	type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;

	rcu_read_lock();
	p = __ipv6_addr_label(net, addr, type, ifindex);
	label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
	rcu_read_unlock();

	ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
		  __func__, addr, type, ifindex, label);

	return label;
}

/* allocate one entry */
static struct ip6addrlbl_entry *ip6addrlbl_alloc(struct net *net,
						 const struct in6_addr *prefix,
						 int prefixlen, int ifindex,
						 u32 label)
{
	struct ip6addrlbl_entry *newp;
	int addrtype;

	ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
		  __func__, prefix, prefixlen, ifindex, (unsigned int)label);

	addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);

	switch (addrtype) {
	case IPV6_ADDR_MAPPED:
		if (prefixlen > 96)
			return ERR_PTR(-EINVAL);
		if (prefixlen < 96)
			addrtype = 0;
		break;
	case IPV6_ADDR_COMPATv4:
		if (prefixlen != 96)
			addrtype = 0;
		break;
	case IPV6_ADDR_LOOPBACK:
		if (prefixlen != 128)
			addrtype = 0;
		break;
	}

	newp = kmalloc(sizeof(*newp), GFP_KERNEL);
	if (!newp)
		return ERR_PTR(-ENOMEM);

	ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
	newp->prefixlen = prefixlen;
	newp->ifindex = ifindex;
	newp->addrtype = addrtype;
	newp->label = label;
	INIT_HLIST_NODE(&newp->list);
#ifdef CONFIG_NET_NS
	newp->lbl_net = hold_net(net);
#endif
	atomic_set(&newp->refcnt, 1);
	return newp;
}

/* add a label */
static int __ip6addrlbl_add(struct ip6addrlbl_entry *newp, int replace)
{
	int ret = 0;

	ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n",
			__func__,
			newp, replace);

	if (hlist_empty(&ip6addrlbl_table.head)) {
		hlist_add_head_rcu(&newp->list, &ip6addrlbl_table.head);
	} else {
		struct hlist_node *pos, *n;
		struct ip6addrlbl_entry *p = NULL;
		hlist_for_each_entry_safe(p, pos, n,
					  &ip6addrlbl_table.head, list) {
			if (p->prefixlen == newp->prefixlen &&
			    net_eq(ip6addrlbl_net(p), ip6addrlbl_net(newp)) &&
			    p->ifindex == newp->ifindex &&
			    ipv6_addr_equal(&p->prefix, &newp->prefix)) {
				if (!replace) {
					ret = -EEXIST;
					goto out;
				}
				hlist_replace_rcu(&p->list, &newp->list);
				ip6addrlbl_put(p);
				goto out;
			} else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
				   (p->prefixlen < newp->prefixlen)) {
				hlist_add_before_rcu(&newp->list, &p->list);
				goto out;
			}
		}
		hlist_add_after_rcu(&p->list, &newp->list);
	}
out:
	if (!ret)
		ip6addrlbl_table.seq++;
	return ret;
}

/* add a label */
static int ip6addrlbl_add(struct net *net,
			  const struct in6_addr *prefix, int prefixlen,
			  int ifindex, u32 label, int replace)
{
	struct ip6addrlbl_entry *newp;
	int ret = 0;

	ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
		  __func__, prefix, prefixlen, ifindex, (unsigned int)label,
		  replace);

	newp = ip6addrlbl_alloc(net, prefix, prefixlen, ifindex, label);
	if (IS_ERR(newp))
		return PTR_ERR(newp);
	spin_lock(&ip6addrlbl_table.lock);
	ret = __ip6addrlbl_add(newp, replace);
	spin_unlock(&ip6addrlbl_table.lock);
	if (ret)
		ip6addrlbl_free(newp);
	return ret;
}

/* remove a label */
static int __ip6addrlbl_del(struct net *net,
			    const struct in6_addr *prefix, int prefixlen,
			    int ifindex)
{
	struct ip6addrlbl_entry *p = NULL;
	struct hlist_node *pos, *n;
	int ret = -ESRCH;

	ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
		  __func__, prefix, prefixlen, ifindex);

	hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) {
		if (p->prefixlen == prefixlen &&
		    net_eq(ip6addrlbl_net(p), net) &&
		    p->ifindex == ifindex &&
		    ipv6_addr_equal(&p->prefix, prefix)) {
			hlist_del_rcu(&p->list);
			ip6addrlbl_put(p);
			ret = 0;
			break;
		}
	}
	return ret;
}

static int ip6addrlbl_del(struct net *net,
			  const struct in6_addr *prefix, int prefixlen,
			  int ifindex)
{
	struct in6_addr prefix_buf;
	int ret;

	ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
		  __func__, prefix, prefixlen, ifindex);

	ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
	spin_lock(&ip6addrlbl_table.lock);
	ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
	spin_unlock(&ip6addrlbl_table.lock);
	return ret;
}

/* add default label */
static int __net_init ip6addrlbl_net_init(struct net *net)
{
	int err = 0;
	int i;

	ADDRLABEL(KERN_DEBUG "%s()\n", __func__);

	for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
		int ret = ip6addrlbl_add(net,
					 ip6addrlbl_init_table[i].prefix,
					 ip6addrlbl_init_table[i].prefixlen,
					 0,
					 ip6addrlbl_init_table[i].label, 0);
		/* XXX: should we free all rules when we catch an error? */
		if (ret && (!err || err != -ENOMEM))
			err = ret;
	}
	return err;
}

static void __net_exit ip6addrlbl_net_exit(struct net *net)
{
	struct ip6addrlbl_entry *p = NULL;
	struct hlist_node *pos, *n;

	/* Remove all labels belonging to the exiting net */
	spin_lock(&ip6addrlbl_table.lock);
	hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) {
		if (net_eq(ip6addrlbl_net(p), net)) {
			hlist_del_rcu(&p->list);
			ip6addrlbl_put(p);
		}
	}
	spin_unlock(&ip6addrlbl_table.lock);
}

static struct pernet_operations ipv6_addr_label_ops = {
	.init = ip6addrlbl_net_init,
	.exit = ip6addrlbl_net_exit,
};

int __init ipv6_addr_label_init(void)
{
	spin_lock_init(&ip6addrlbl_table.lock);

	return register_pernet_subsys(&ipv6_addr_label_ops);
}

void ipv6_addr_label_cleanup(void)
{
	unregister_pernet_subsys(&ipv6_addr_label_ops);
}

static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
	[IFAL_ADDRESS]		= { .len = sizeof(struct in6_addr), },
	[IFAL_LABEL]		= { .len = sizeof(u32), },
};

static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
			     void *arg)
{
	struct net *net = sock_net(skb->sk);
	struct ifaddrlblmsg *ifal;
	struct nlattr *tb[IFAL_MAX+1];
	struct in6_addr *pfx;
	u32 label;
	int err = 0;

	err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy);
	if (err < 0)
		return err;

	ifal = nlmsg_data(nlh);

	if (ifal->ifal_family != AF_INET6 ||
	    ifal->ifal_prefixlen > 128)
		return -EINVAL;

	if (!tb[IFAL_ADDRESS])
		return -EINVAL;

	pfx = nla_data(tb[IFAL_ADDRESS]);
	if (!pfx)
		return -EINVAL;

	if (!tb[IFAL_LABEL])
		return -EINVAL;
	label = nla_get_u32(tb[IFAL_LABEL]);
	if (label == IPV6_ADDR_LABEL_DEFAULT)
		return -EINVAL;

	switch(nlh->nlmsg_type) {
	case RTM_NEWADDRLABEL:
		if (ifal->ifal_index &&
		    !__dev_get_by_index(net, ifal->ifal_index))
			return -EINVAL;

		err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
				     ifal->ifal_index, label,
				     nlh->nlmsg_flags & NLM_F_REPLACE);
		break;
	case RTM_DELADDRLABEL:
		err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
				     ifal->ifal_index);
		break;
	default:
		err = -EOPNOTSUPP;
	}
	return err;
}

static inline void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
				     int prefixlen, int ifindex, u32 lseq)
{
	struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
	ifal->ifal_family = AF_INET6;
	ifal->ifal_prefixlen = prefixlen;
	ifal->ifal_flags = 0;
	ifal->ifal_index = ifindex;
	ifal->ifal_seq = lseq;
};

static int ip6addrlbl_fill(struct sk_buff *skb,
			   struct ip6addrlbl_entry *p,
			   u32 lseq,
			   u32 pid, u32 seq, int event,
			   unsigned int flags)
{
	struct nlmsghdr *nlh = nlmsg_put(skb, pid, seq, event,
					 sizeof(struct ifaddrlblmsg), flags);
	if (!nlh)
		return -EMSGSIZE;

	ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);

	if (nla_put(skb, IFAL_ADDRESS, 16, &p->prefix) < 0 ||
	    nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
		nlmsg_cancel(skb, nlh);
		return -EMSGSIZE;
	}

	return nlmsg_end(skb, nlh);
}

static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
	struct net *net = sock_net(skb->sk);
	struct ip6addrlbl_entry *p;
	struct hlist_node *pos;
	int idx = 0, s_idx = cb->args[0];
	int err;

	rcu_read_lock();
	hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) {
		if (idx >= s_idx &&
		    net_eq(ip6addrlbl_net(p), net)) {
			if ((err = ip6addrlbl_fill(skb, p,
						   ip6addrlbl_table.seq,
						   NETLINK_CB(cb->skb).pid,
						   cb->nlh->nlmsg_seq,
						   RTM_NEWADDRLABEL,
						   NLM_F_MULTI)) <= 0)
				break;
		}
		idx++;
	}
	rcu_read_unlock();
	cb->args[0] = idx;
	return skb->len;
}

static inline int ip6addrlbl_msgsize(void)
{
	return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
		+ nla_total_size(16)	/* IFAL_ADDRESS */
		+ nla_total_size(4);	/* IFAL_LABEL */
}

static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr* nlh,
			  void *arg)
{
	struct net *net = sock_net(in_skb->sk);
	struct ifaddrlblmsg *ifal;
	struct nlattr *tb[IFAL_MAX+1];
	struct in6_addr *addr;
	u32 lseq;
	int err = 0;
	struct ip6addrlbl_entry *p;
	struct sk_buff *skb;

	err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy);
	if (err < 0)
		return err;

	ifal = nlmsg_data(nlh);

	if (ifal->ifal_family != AF_INET6 ||
	    ifal->ifal_prefixlen != 128)
		return -EINVAL;

	if (ifal->ifal_index &&
	    !__dev_get_by_index(net, ifal->ifal_index))
		return -EINVAL;

	if (!tb[IFAL_ADDRESS])
		return -EINVAL;

	addr = nla_data(tb[IFAL_ADDRESS]);
	if (!addr)
		return -EINVAL;

	rcu_read_lock();
	p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
	if (p && ip6addrlbl_hold(p))
		p = NULL;
	lseq = ip6addrlbl_table.seq;
	rcu_read_unlock();

	if (!p) {
		err = -ESRCH;
		goto out;
	}

	if (!(skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL))) {
		ip6addrlbl_put(p);
		return -ENOBUFS;
	}

	err = ip6addrlbl_fill(skb, p, lseq,
			      NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
			      RTM_NEWADDRLABEL, 0);

	ip6addrlbl_put(p);

	if (err < 0) {
		WARN_ON(err == -EMSGSIZE);
		kfree_skb(skb);
		goto out;
	}

	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
out:
	return err;
}

void __init ipv6_addr_label_rtnl_register(void)
{
	__rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel, NULL);
	__rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel, NULL);
	__rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get, ip6addrlbl_dump);
}
Commit	Line	Data
2a8cc6c8 YH	1	/*
	2	* IPv6 Address Label subsystem
	3	* for the IPv6 "Default" Source Address Selection
	4	*
	5	* Copyright (C)2007 USAGI/WIDE Project
	6	*/
	7	/*
	8	* Author:
	9	* YOSHIFUJI Hideaki @ USAGI/WIDE Project <[email protected]>
	10	*/
	11
	12	#include <linux/kernel.h>
	13	#include <linux/list.h>
	14	#include <linux/rcupdate.h>
	15	#include <linux/in6.h>
5a0e3ad6	16	#include <linux/slab.h>
2a8cc6c8 YH	17	#include <net/addrconf.h>
	18	#include <linux/if_addrlabel.h>
	19	#include <linux/netlink.h>
	20	#include <linux/rtnetlink.h>
	21
	22	#if 0
	23	#define ADDRLABEL(x...) printk(x)
	24	#else
	25	#define ADDRLABEL(x...) do { ; } while(0)
	26	#endif
	27
	28	/*
	29	* Policy Table
	30	*/
	31	struct ip6addrlbl_entry
	32	{
3de23255 BT	33	#ifdef CONFIG_NET_NS
	34	struct net *lbl_net;
	35	#endif
2a8cc6c8 YH	36	struct in6_addr prefix;
	37	int prefixlen;
	38	int ifindex;
	39	int addrtype;
	40	u32 label;
	41	struct hlist_node list;
	42	atomic_t refcnt;
	43	struct rcu_head rcu;
	44	};
	45
	46	static struct ip6addrlbl_table
	47	{
	48	struct hlist_head head;
	49	spinlock_t lock;
	50	u32 seq;
	51	} ip6addrlbl_table;
	52
3de23255 BT	53	static inline
	54	struct net ip6addrlbl_net(const struct ip6addrlbl_entry lbl)
	55	{
c2d9ba9b	56	return read_pnet(&lbl->lbl_net);
3de23255 BT	57	}
3de23255 BT	58
2a8cc6c8 YH	59	/*
	60	* Default policy table (RFC3484 + extensions)
	61	*
	62	* prefix addr_type label
	63	* -------------------------------------------------------------------------
	64	* ::1/128 LOOPBACK 0
	65	* ::/0 N/A 1
	66	* 2002::/16 N/A 2
	67	* ::/96 COMPATv4 3
	68	* ::ffff:0:0/96 V4MAPPED 4
	69	* fc00::/7 N/A 5 ULA (RFC 4193)
	70	* 2001::/32 N/A 6 Teredo (RFC 4380)
5fe47b8a	71	* 2001:10::/28 N/A 7 ORCHID (RFC 4843)
2a8cc6c8 YH	72	*
	73	* Note: 0xffffffff is used if we do not have any policies.
	74	*/
	75
	76	#define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
	77
3de23255	78	static const __net_initdata struct ip6addrlbl_init_table
2a8cc6c8 YH	79	{
	80	const struct in6_addr *prefix;
	81	int prefixlen;
	82	u32 label;
	83	} ip6addrlbl_init_table[] = {
	84	{ /* ::/0 */
	85	.prefix = &in6addr_any,
	86	.label = 1,
	87	},{ /* fc00::/7 */
	88	.prefix = &(struct in6_addr){{{ 0xfc }}},
	89	.prefixlen = 7,
	90	.label = 5,
	91	},{ /* 2002::/16 */
	92	.prefix = &(struct in6_addr){{{ 0x20, 0x02 }}},
	93	.prefixlen = 16,
	94	.label = 2,
	95	},{ /* 2001::/32 */
	96	.prefix = &(struct in6_addr){{{ 0x20, 0x01 }}},
	97	.prefixlen = 32,
	98	.label = 6,
5fe47b8a JMT	99	},{ /* 2001:10::/28 */
	100	.prefix = &(struct in6_addr){{{ 0x20, 0x01, 0x00, 0x10 }}},
	101	.prefixlen = 28,
	102	.label = 7,
2a8cc6c8 YH	103	},{ /* ::ffff:0:0 */
	104	.prefix = &(struct in6_addr){{{ [10] = 0xff, [11] = 0xff }}},
	105	.prefixlen = 96,
	106	.label = 4,
	107	},{ /* ::/96 */
	108	.prefix = &in6addr_any,
	109	.prefixlen = 96,
	110	.label = 3,
	111	},{ /* ::1/128 */
	112	.prefix = &in6addr_loopback,
	113	.prefixlen = 128,
	114	.label = 0,
	115	}
	116	};
	117
	118	/* Object management */
	119	static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p)
	120	{
3de23255 BT	121	#ifdef CONFIG_NET_NS
	122	release_net(p->lbl_net);
	123	#endif
2a8cc6c8 YH	124	kfree(p);
	125	}
	126
85040bcb YH	127	static void ip6addrlbl_free_rcu(struct rcu_head *h)
	128	{
	129	ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu));
	130	}
	131
2a8cc6c8 YH	132	static inline int ip6addrlbl_hold(struct ip6addrlbl_entry *p)
	133	{
	134	return atomic_inc_not_zero(&p->refcnt);
	135	}
	136
	137	static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p)
	138	{
	139	if (atomic_dec_and_test(&p->refcnt))
85040bcb	140	call_rcu(&p->rcu, ip6addrlbl_free_rcu);
2a8cc6c8 YH	141	}
	142
	143	/* Find label */
3de23255 BT	144	static int __ip6addrlbl_match(struct net *net,
3de23255 BT	145	struct ip6addrlbl_entry *p,
2a8cc6c8 YH	146	const struct in6_addr *addr,
	147	int addrtype, int ifindex)
	148	{
3de23255 BT	149	if (!net_eq(ip6addrlbl_net(p), net))
3de23255 BT	150	return 0;
2a8cc6c8 YH	151	if (p->ifindex && p->ifindex != ifindex)
	152	return 0;
	153	if (p->addrtype && p->addrtype != addrtype)
	154	return 0;
	155	if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
	156	return 0;
	157	return 1;
	158	}
	159
3de23255 BT	160	static struct ip6addrlbl_entry __ipv6_addr_label(struct net net,
3de23255 BT	161	const struct in6_addr *addr,
2a8cc6c8 YH	162	int type, int ifindex)
	163	{
	164	struct hlist_node *pos;
	165	struct ip6addrlbl_entry *p;
	166	hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) {
3de23255	167	if (__ip6addrlbl_match(net, p, addr, type, ifindex))
2a8cc6c8 YH	168	return p;
	169	}
	170	return NULL;
	171	}
	172
3de23255 BT	173	u32 ipv6_addr_label(struct net *net,
3de23255 BT	174	const struct in6_addr *addr, int type, int ifindex)
2a8cc6c8 YH	175	{
	176	u32 label;
	177	struct ip6addrlbl_entry *p;
	178
	179	type &= IPV6_ADDR_MAPPED \| IPV6_ADDR_COMPATv4 \| IPV6_ADDR_LOOPBACK;
	180
	181	rcu_read_lock();
3de23255	182	p = __ipv6_addr_label(net, addr, type, ifindex);
2a8cc6c8 YH	183	label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
	184	rcu_read_unlock();
	185
5b095d98	186	ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
0c6ce78a	187	__func__, addr, type, ifindex, label);
2a8cc6c8 YH	188
	189	return label;
	190	}
	191
	192	/* allocate one entry */
3de23255 BT	193	static struct ip6addrlbl_entry ip6addrlbl_alloc(struct net net,
3de23255 BT	194	const struct in6_addr *prefix,
40fee36e YH	195	int prefixlen, int ifindex,
40fee36e YH	196	u32 label)
2a8cc6c8 YH	197	{
	198	struct ip6addrlbl_entry *newp;
	199	int addrtype;
	200
5b095d98	201	ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
0c6ce78a	202	__func__, prefix, prefixlen, ifindex, (unsigned int)label);
2a8cc6c8 YH	203
	204	addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED \| IPV6_ADDR_COMPATv4 \| IPV6_ADDR_LOOPBACK);
	205
	206	switch (addrtype) {
	207	case IPV6_ADDR_MAPPED:
	208	if (prefixlen > 96)
	209	return ERR_PTR(-EINVAL);
	210	if (prefixlen < 96)
	211	addrtype = 0;
	212	break;
	213	case IPV6_ADDR_COMPATv4:
	214	if (prefixlen != 96)
	215	addrtype = 0;
	216	break;
	217	case IPV6_ADDR_LOOPBACK:
	218	if (prefixlen != 128)
	219	addrtype = 0;
	220	break;
	221	}
	222
	223	newp = kmalloc(sizeof(*newp), GFP_KERNEL);
	224	if (!newp)
	225	return ERR_PTR(-ENOMEM);
	226
	227	ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
	228	newp->prefixlen = prefixlen;
	229	newp->ifindex = ifindex;
	230	newp->addrtype = addrtype;
	231	newp->label = label;
	232	INIT_HLIST_NODE(&newp->list);
3de23255 BT	233	#ifdef CONFIG_NET_NS
	234	newp->lbl_net = hold_net(net);
	235	#endif
2a8cc6c8 YH	236	atomic_set(&newp->refcnt, 1);
	237	return newp;
	238	}
	239
	240	/* add a label */
40fee36e	241	static int __ip6addrlbl_add(struct ip6addrlbl_entry *newp, int replace)
2a8cc6c8 YH	242	{
	243	int ret = 0;
	244
	245	ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n",
0dc47877	246	__func__,
2a8cc6c8 YH	247	newp, replace);
	248
	249	if (hlist_empty(&ip6addrlbl_table.head)) {
	250	hlist_add_head_rcu(&newp->list, &ip6addrlbl_table.head);
	251	} else {
	252	struct hlist_node pos, n;
	253	struct ip6addrlbl_entry *p = NULL;
	254	hlist_for_each_entry_safe(p, pos, n,
	255	&ip6addrlbl_table.head, list) {
	256	if (p->prefixlen == newp->prefixlen &&
3de23255	257	net_eq(ip6addrlbl_net(p), ip6addrlbl_net(newp)) &&
2a8cc6c8 YH	258	p->ifindex == newp->ifindex &&
	259	ipv6_addr_equal(&p->prefix, &newp->prefix)) {
	260	if (!replace) {
	261	ret = -EEXIST;
	262	goto out;
	263	}
	264	hlist_replace_rcu(&p->list, &newp->list);
	265	ip6addrlbl_put(p);
2a8cc6c8 YH	266	goto out;
	267	} else if ((p->prefixlen == newp->prefixlen && !p->ifindex) \|\|
	268	(p->prefixlen < newp->prefixlen)) {
	269	hlist_add_before_rcu(&newp->list, &p->list);
	270	goto out;
	271	}
	272	}
	273	hlist_add_after_rcu(&p->list, &newp->list);
	274	}
	275	out:
	276	if (!ret)
	277	ip6addrlbl_table.seq++;
	278	return ret;
	279	}
	280
	281	/* add a label */
3de23255 BT	282	static int ip6addrlbl_add(struct net *net,
3de23255 BT	283	const struct in6_addr *prefix, int prefixlen,
40fee36e	284	int ifindex, u32 label, int replace)
2a8cc6c8 YH	285	{
	286	struct ip6addrlbl_entry *newp;
	287	int ret = 0;
	288
5b095d98	289	ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
0c6ce78a HH	290	__func__, prefix, prefixlen, ifindex, (unsigned int)label,
0c6ce78a HH	291	replace);
2a8cc6c8	292
3de23255	293	newp = ip6addrlbl_alloc(net, prefix, prefixlen, ifindex, label);
2a8cc6c8 YH	294	if (IS_ERR(newp))
	295	return PTR_ERR(newp);
	296	spin_lock(&ip6addrlbl_table.lock);
	297	ret = __ip6addrlbl_add(newp, replace);
	298	spin_unlock(&ip6addrlbl_table.lock);
	299	if (ret)
	300	ip6addrlbl_free(newp);
	301	return ret;
	302	}
	303
	304	/* remove a label */
3de23255 BT	305	static int __ip6addrlbl_del(struct net *net,
3de23255 BT	306	const struct in6_addr *prefix, int prefixlen,
40fee36e	307	int ifindex)
2a8cc6c8 YH	308	{
	309	struct ip6addrlbl_entry *p = NULL;
	310	struct hlist_node pos, n;
	311	int ret = -ESRCH;
	312
5b095d98	313	ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
0c6ce78a	314	__func__, prefix, prefixlen, ifindex);
2a8cc6c8 YH	315
	316	hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) {
	317	if (p->prefixlen == prefixlen &&
3de23255	318	net_eq(ip6addrlbl_net(p), net) &&
2a8cc6c8 YH	319	p->ifindex == ifindex &&
	320	ipv6_addr_equal(&p->prefix, prefix)) {
	321	hlist_del_rcu(&p->list);
	322	ip6addrlbl_put(p);
2a8cc6c8 YH	323	ret = 0;
	324	break;
	325	}
	326	}
	327	return ret;
	328	}
	329
3de23255 BT	330	static int ip6addrlbl_del(struct net *net,
3de23255 BT	331	const struct in6_addr *prefix, int prefixlen,
40fee36e	332	int ifindex)
2a8cc6c8 YH	333	{
	334	struct in6_addr prefix_buf;
	335	int ret;
	336
5b095d98	337	ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
0c6ce78a	338	__func__, prefix, prefixlen, ifindex);
2a8cc6c8 YH	339
	340	ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
	341	spin_lock(&ip6addrlbl_table.lock);
3de23255	342	ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
2a8cc6c8 YH	343	spin_unlock(&ip6addrlbl_table.lock);
	344	return ret;
	345	}
	346
	347	/* add default label */
3de23255	348	static int __net_init ip6addrlbl_net_init(struct net *net)
2a8cc6c8 YH	349	{
	350	int err = 0;
	351	int i;
	352
0dc47877	353	ADDRLABEL(KERN_DEBUG "%s()\n", __func__);
2a8cc6c8 YH	354
2a8cc6c8 YH	355	for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
3de23255 BT	356	int ret = ip6addrlbl_add(net,
3de23255 BT	357	ip6addrlbl_init_table[i].prefix,
2a8cc6c8 YH	358	ip6addrlbl_init_table[i].prefixlen,
	359	0,
	360	ip6addrlbl_init_table[i].label, 0);
	361	/* XXX: should we free all rules when we catch an error? */
	362	if (ret && (!err \|\| err != -ENOMEM))
	363	err = ret;
	364	}
	365	return err;
	366	}
	367
3de23255 BT	368	static void __net_exit ip6addrlbl_net_exit(struct net *net)
	369	{
	370	struct ip6addrlbl_entry *p = NULL;
	371	struct hlist_node pos, n;
	372
	373	/* Remove all labels belonging to the exiting net */
	374	spin_lock(&ip6addrlbl_table.lock);
	375	hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) {
	376	if (net_eq(ip6addrlbl_net(p), net)) {
	377	hlist_del_rcu(&p->list);
	378	ip6addrlbl_put(p);
	379	}
	380	}
	381	spin_unlock(&ip6addrlbl_table.lock);
	382	}
	383
	384	static struct pernet_operations ipv6_addr_label_ops = {
	385	.init = ip6addrlbl_net_init,
	386	.exit = ip6addrlbl_net_exit,
	387	};
	388
2a8cc6c8 YH	389	int __init ipv6_addr_label_init(void)
	390	{
	391	spin_lock_init(&ip6addrlbl_table.lock);
	392
3de23255	393	return register_pernet_subsys(&ipv6_addr_label_ops);
2a8cc6c8 YH	394	}
2a8cc6c8 YH	395
2cc6d2bf NH	396	void ipv6_addr_label_cleanup(void)
	397	{
	398	unregister_pernet_subsys(&ipv6_addr_label_ops);
	399	}
	400
2a8cc6c8 YH	401	static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
	402	[IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
	403	[IFAL_LABEL] = { .len = sizeof(u32), },
	404	};
	405
	406	static int ip6addrlbl_newdel(struct sk_buff skb, struct nlmsghdr nlh,
	407	void *arg)
	408	{
3b1e0a65	409	struct net *net = sock_net(skb->sk);
2a8cc6c8 YH	410	struct ifaddrlblmsg *ifal;
	411	struct nlattr *tb[IFAL_MAX+1];
	412	struct in6_addr *pfx;
	413	u32 label;
	414	int err = 0;
	415
	416	err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy);
	417	if (err < 0)
	418	return err;
	419
	420	ifal = nlmsg_data(nlh);
	421
	422	if (ifal->ifal_family != AF_INET6 \|\|
	423	ifal->ifal_prefixlen > 128)
	424	return -EINVAL;
	425
2a8cc6c8 YH	426	if (!tb[IFAL_ADDRESS])
	427	return -EINVAL;
	428
	429	pfx = nla_data(tb[IFAL_ADDRESS]);
	430	if (!pfx)
	431	return -EINVAL;
	432
	433	if (!tb[IFAL_LABEL])
	434	return -EINVAL;
	435	label = nla_get_u32(tb[IFAL_LABEL]);
	436	if (label == IPV6_ADDR_LABEL_DEFAULT)
	437	return -EINVAL;
	438
	439	switch(nlh->nlmsg_type) {
	440	case RTM_NEWADDRLABEL:
0771275b FW	441	if (ifal->ifal_index &&
	442	!__dev_get_by_index(net, ifal->ifal_index))
	443	return -EINVAL;
	444
3de23255	445	err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
2a8cc6c8 YH	446	ifal->ifal_index, label,
	447	nlh->nlmsg_flags & NLM_F_REPLACE);
	448	break;
	449	case RTM_DELADDRLABEL:
3de23255	450	err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
2a8cc6c8 YH	451	ifal->ifal_index);
	452	break;
	453	default:
	454	err = -EOPNOTSUPP;
	455	}
	456	return err;
	457	}
	458
	459	static inline void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
	460	int prefixlen, int ifindex, u32 lseq)
	461	{
	462	struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
	463	ifal->ifal_family = AF_INET6;
	464	ifal->ifal_prefixlen = prefixlen;
	465	ifal->ifal_flags = 0;
	466	ifal->ifal_index = ifindex;
	467	ifal->ifal_seq = lseq;
	468	};
	469
	470	static int ip6addrlbl_fill(struct sk_buff *skb,
	471	struct ip6addrlbl_entry *p,
	472	u32 lseq,
	473	u32 pid, u32 seq, int event,
	474	unsigned int flags)
	475	{
	476	struct nlmsghdr *nlh = nlmsg_put(skb, pid, seq, event,
	477	sizeof(struct ifaddrlblmsg), flags);
	478	if (!nlh)
	479	return -EMSGSIZE;
	480
	481	ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
	482
	483	if (nla_put(skb, IFAL_ADDRESS, 16, &p->prefix) < 0 \|\|
	484	nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
	485	nlmsg_cancel(skb, nlh);
	486	return -EMSGSIZE;
	487	}
	488
	489	return nlmsg_end(skb, nlh);
	490	}
	491
	492	static int ip6addrlbl_dump(struct sk_buff skb, struct netlink_callback cb)
	493	{
3b1e0a65	494	struct net *net = sock_net(skb->sk);
2a8cc6c8 YH	495	struct ip6addrlbl_entry *p;
	496	struct hlist_node *pos;
	497	int idx = 0, s_idx = cb->args[0];
	498	int err;
	499
	500	rcu_read_lock();
	501	hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) {
3de23255 BT	502	if (idx >= s_idx &&
3de23255 BT	503	net_eq(ip6addrlbl_net(p), net)) {
2a8cc6c8 YH	504	if ((err = ip6addrlbl_fill(skb, p,
	505	ip6addrlbl_table.seq,
	506	NETLINK_CB(cb->skb).pid,
	507	cb->nlh->nlmsg_seq,
	508	RTM_NEWADDRLABEL,
	509	NLM_F_MULTI)) <= 0)
	510	break;
	511	}
	512	idx++;
	513	}
	514	rcu_read_unlock();
	515	cb->args[0] = idx;
	516	return skb->len;
	517	}
	518
	519	static inline int ip6addrlbl_msgsize(void)
	520	{
a02cec21	521	return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
2a8cc6c8	522	+ nla_total_size(16) /* IFAL_ADDRESS */
a02cec21	523	+ nla_total_size(4); /* IFAL_LABEL */
2a8cc6c8 YH	524	}
	525
	526	static int ip6addrlbl_get(struct sk_buff in_skb, struct nlmsghdr nlh,
	527	void *arg)
	528	{
3b1e0a65	529	struct net *net = sock_net(in_skb->sk);
2a8cc6c8 YH	530	struct ifaddrlblmsg *ifal;
	531	struct nlattr *tb[IFAL_MAX+1];
	532	struct in6_addr *addr;
	533	u32 lseq;
	534	int err = 0;
	535	struct ip6addrlbl_entry *p;
	536	struct sk_buff *skb;
	537
	538	err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy);
	539	if (err < 0)
	540	return err;
	541
	542	ifal = nlmsg_data(nlh);
	543
	544	if (ifal->ifal_family != AF_INET6 \|\|
	545	ifal->ifal_prefixlen != 128)
	546	return -EINVAL;
	547
	548	if (ifal->ifal_index &&
3de23255	549	!__dev_get_by_index(net, ifal->ifal_index))
2a8cc6c8 YH	550	return -EINVAL;
	551
	552	if (!tb[IFAL_ADDRESS])
	553	return -EINVAL;
	554
	555	addr = nla_data(tb[IFAL_ADDRESS]);
	556	if (!addr)
	557	return -EINVAL;
	558
	559	rcu_read_lock();
3de23255	560	p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
2a8cc6c8 YH	561	if (p && ip6addrlbl_hold(p))
	562	p = NULL;
	563	lseq = ip6addrlbl_table.seq;
	564	rcu_read_unlock();
	565
	566	if (!p) {
	567	err = -ESRCH;
	568	goto out;
	569	}
	570
	571	if (!(skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL))) {
	572	ip6addrlbl_put(p);
	573	return -ENOBUFS;
	574	}
	575
	576	err = ip6addrlbl_fill(skb, p, lseq,
	577	NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
	578	RTM_NEWADDRLABEL, 0);
	579
	580	ip6addrlbl_put(p);
	581
	582	if (err < 0) {
	583	WARN_ON(err == -EMSGSIZE);
	584	kfree_skb(skb);
	585	goto out;
	586	}
	587
3de23255	588	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2a8cc6c8 YH	589	out:
	590	return err;
	591	}
	592
	593	void __init ipv6_addr_label_rtnl_register(void)
	594	{
	595	__rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel, NULL);
	596	__rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel, NULL);
	597	__rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get, ip6addrlbl_dump);
	598	}
	599