2 * IPv6 Address Label subsystem
3 * for the IPv6 "Default" Source Address Selection
5 * Copyright (C)2007 USAGI/WIDE Project
9 * YOSHIFUJI Hideaki @ USAGI/WIDE Project <yoshfuji@linux-ipv6.org>
12 #include <linux/kernel.h>
13 #include <linux/list.h>
14 #include <linux/rcupdate.h>
15 #include <linux/in6.h>
16 #include <linux/slab.h>
17 #include <net/addrconf.h>
18 #include <linux/if_addrlabel.h>
19 #include <linux/netlink.h>
20 #include <linux/rtnetlink.h>
21 #include <linux/refcount.h>
24 #define ADDRLABEL(x...) printk(x)
26 #define ADDRLABEL(x...) do { ; } while (0)
32 struct ip6addrlbl_entry {
33 struct in6_addr prefix;
38 struct hlist_node list;
44 * Default policy table (RFC6724 + extensions)
46 * prefix addr_type label
47 * -------------------------------------------------------------------------
52 * ::ffff:0:0/96 V4MAPPED 4
53 * fc00::/7 N/A 5 ULA (RFC 4193)
54 * 2001::/32 N/A 6 Teredo (RFC 4380)
55 * 2001:10::/28 N/A 7 ORCHID (RFC 4843)
56 * fec0::/10 N/A 11 Site-local
57 * (deprecated by RFC3879)
58 * 3ffe::/16 N/A 12 6bone
60 * Note: 0xffffffff is used if we do not have any policies.
61 * Note: Labels for ULA and 6to4 are different from labels listed in RFC6724.
64 #define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
66 static const __net_initconst struct ip6addrlbl_init_table
68 const struct in6_addr *prefix;
71 } ip6addrlbl_init_table[] = {
73 .prefix = &in6addr_any,
76 .prefix = &(struct in6_addr){ { { 0xfc } } } ,
80 .prefix = &(struct in6_addr){ { { 0xfe, 0xc0 } } },
84 .prefix = &(struct in6_addr){ { { 0x20, 0x02 } } },
88 .prefix = &(struct in6_addr){ { { 0x3f, 0xfe } } },
92 .prefix = &(struct in6_addr){ { { 0x20, 0x01 } } },
95 }, { /* 2001:10::/28 */
96 .prefix = &(struct in6_addr){ { { 0x20, 0x01, 0x00, 0x10 } } },
100 .prefix = &(struct in6_addr){ { { [10] = 0xff, [11] = 0xff } } },
104 .prefix = &in6addr_any,
108 .prefix = &in6addr_loopback,
114 /* Object management */
115 static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p)
120 static void ip6addrlbl_free_rcu(struct rcu_head *h)
122 ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu));
125 static bool ip6addrlbl_hold(struct ip6addrlbl_entry *p)
127 return refcount_inc_not_zero(&p->refcnt);
130 static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p)
132 if (refcount_dec_and_test(&p->refcnt))
133 call_rcu(&p->rcu, ip6addrlbl_free_rcu);
137 static bool __ip6addrlbl_match(const struct ip6addrlbl_entry *p,
138 const struct in6_addr *addr,
139 int addrtype, int ifindex)
141 if (p->ifindex && p->ifindex != ifindex)
143 if (p->addrtype && p->addrtype != addrtype)
145 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
150 static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
151 const struct in6_addr *addr,
152 int type, int ifindex)
154 struct ip6addrlbl_entry *p;
156 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
157 if (__ip6addrlbl_match(p, addr, type, ifindex))
163 u32 ipv6_addr_label(struct net *net,
164 const struct in6_addr *addr, int type, int ifindex)
167 struct ip6addrlbl_entry *p;
169 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
172 p = __ipv6_addr_label(net, addr, type, ifindex);
173 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
176 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
177 __func__, addr, type, ifindex, label);
182 /* allocate one entry */
183 static struct ip6addrlbl_entry *ip6addrlbl_alloc(const struct in6_addr *prefix,
184 int prefixlen, int ifindex,
187 struct ip6addrlbl_entry *newp;
190 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
191 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
193 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
196 case IPV6_ADDR_MAPPED:
198 return ERR_PTR(-EINVAL);
202 case IPV6_ADDR_COMPATv4:
206 case IPV6_ADDR_LOOPBACK:
207 if (prefixlen != 128)
212 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
214 return ERR_PTR(-ENOMEM);
216 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
217 newp->prefixlen = prefixlen;
218 newp->ifindex = ifindex;
219 newp->addrtype = addrtype;
221 INIT_HLIST_NODE(&newp->list);
222 refcount_set(&newp->refcnt, 1);
227 static int __ip6addrlbl_add(struct net *net, struct ip6addrlbl_entry *newp,
230 struct ip6addrlbl_entry *last = NULL, *p = NULL;
231 struct hlist_node *n;
234 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", __func__, newp,
237 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
238 if (p->prefixlen == newp->prefixlen &&
239 p->ifindex == newp->ifindex &&
240 ipv6_addr_equal(&p->prefix, &newp->prefix)) {
245 hlist_replace_rcu(&p->list, &newp->list);
248 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
249 (p->prefixlen < newp->prefixlen)) {
250 hlist_add_before_rcu(&newp->list, &p->list);
256 hlist_add_behind_rcu(&newp->list, &last->list);
258 hlist_add_head_rcu(&newp->list, &net->ipv6.ip6addrlbl_table.head);
261 net->ipv6.ip6addrlbl_table.seq++;
266 static int ip6addrlbl_add(struct net *net,
267 const struct in6_addr *prefix, int prefixlen,
268 int ifindex, u32 label, int replace)
270 struct ip6addrlbl_entry *newp;
273 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
274 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
277 newp = ip6addrlbl_alloc(prefix, prefixlen, ifindex, label);
279 return PTR_ERR(newp);
280 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
281 ret = __ip6addrlbl_add(net, newp, replace);
282 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
284 ip6addrlbl_free(newp);
289 static int __ip6addrlbl_del(struct net *net,
290 const struct in6_addr *prefix, int prefixlen,
293 struct ip6addrlbl_entry *p = NULL;
294 struct hlist_node *n;
297 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
298 __func__, prefix, prefixlen, ifindex);
300 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
301 if (p->prefixlen == prefixlen &&
302 p->ifindex == ifindex &&
303 ipv6_addr_equal(&p->prefix, prefix)) {
304 hlist_del_rcu(&p->list);
313 static int ip6addrlbl_del(struct net *net,
314 const struct in6_addr *prefix, int prefixlen,
317 struct in6_addr prefix_buf;
320 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
321 __func__, prefix, prefixlen, ifindex);
323 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
324 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
325 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
326 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
330 /* add default label */
331 static int __net_init ip6addrlbl_net_init(struct net *net)
336 ADDRLABEL(KERN_DEBUG "%s\n", __func__);
338 spin_lock_init(&net->ipv6.ip6addrlbl_table.lock);
339 INIT_HLIST_HEAD(&net->ipv6.ip6addrlbl_table.head);
341 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
342 int ret = ip6addrlbl_add(net,
343 ip6addrlbl_init_table[i].prefix,
344 ip6addrlbl_init_table[i].prefixlen,
346 ip6addrlbl_init_table[i].label, 0);
347 /* XXX: should we free all rules when we catch an error? */
348 if (ret && (!err || err != -ENOMEM))
354 static void __net_exit ip6addrlbl_net_exit(struct net *net)
356 struct ip6addrlbl_entry *p = NULL;
357 struct hlist_node *n;
359 /* Remove all labels belonging to the exiting net */
360 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
361 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
362 hlist_del_rcu(&p->list);
365 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
368 static struct pernet_operations ipv6_addr_label_ops = {
369 .init = ip6addrlbl_net_init,
370 .exit = ip6addrlbl_net_exit,
373 int __init ipv6_addr_label_init(void)
375 return register_pernet_subsys(&ipv6_addr_label_ops);
378 void ipv6_addr_label_cleanup(void)
380 unregister_pernet_subsys(&ipv6_addr_label_ops);
383 static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
384 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
385 [IFAL_LABEL] = { .len = sizeof(u32), },
388 static bool addrlbl_ifindex_exists(struct net *net, int ifindex)
391 struct net_device *dev;
394 dev = dev_get_by_index_rcu(net, ifindex);
400 static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
401 struct netlink_ext_ack *extack)
403 struct net *net = sock_net(skb->sk);
404 struct ifaddrlblmsg *ifal;
405 struct nlattr *tb[IFAL_MAX+1];
406 struct in6_addr *pfx;
410 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
415 ifal = nlmsg_data(nlh);
417 if (ifal->ifal_family != AF_INET6 ||
418 ifal->ifal_prefixlen > 128)
421 if (!tb[IFAL_ADDRESS])
423 pfx = nla_data(tb[IFAL_ADDRESS]);
427 label = nla_get_u32(tb[IFAL_LABEL]);
428 if (label == IPV6_ADDR_LABEL_DEFAULT)
431 switch (nlh->nlmsg_type) {
432 case RTM_NEWADDRLABEL:
433 if (ifal->ifal_index &&
434 !addrlbl_ifindex_exists(net, ifal->ifal_index))
437 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
438 ifal->ifal_index, label,
439 nlh->nlmsg_flags & NLM_F_REPLACE);
441 case RTM_DELADDRLABEL:
442 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
451 static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
452 int prefixlen, int ifindex, u32 lseq)
454 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
455 ifal->ifal_family = AF_INET6;
456 ifal->ifal_prefixlen = prefixlen;
457 ifal->ifal_flags = 0;
458 ifal->ifal_index = ifindex;
459 ifal->ifal_seq = lseq;
462 static int ip6addrlbl_fill(struct sk_buff *skb,
463 struct ip6addrlbl_entry *p,
465 u32 portid, u32 seq, int event,
468 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, event,
469 sizeof(struct ifaddrlblmsg), flags);
473 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
475 if (nla_put_in6_addr(skb, IFAL_ADDRESS, &p->prefix) < 0 ||
476 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
477 nlmsg_cancel(skb, nlh);
485 static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
487 struct net *net = sock_net(skb->sk);
488 struct ip6addrlbl_entry *p;
489 int idx = 0, s_idx = cb->args[0];
493 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
495 err = ip6addrlbl_fill(skb, p,
496 net->ipv6.ip6addrlbl_table.seq,
497 NETLINK_CB(cb->skb).portid,
511 static inline int ip6addrlbl_msgsize(void)
513 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
514 + nla_total_size(16) /* IFAL_ADDRESS */
515 + nla_total_size(4); /* IFAL_LABEL */
518 static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
519 struct netlink_ext_ack *extack)
521 struct net *net = sock_net(in_skb->sk);
522 struct ifaddrlblmsg *ifal;
523 struct nlattr *tb[IFAL_MAX+1];
524 struct in6_addr *addr;
527 struct ip6addrlbl_entry *p;
530 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
535 ifal = nlmsg_data(nlh);
537 if (ifal->ifal_family != AF_INET6 ||
538 ifal->ifal_prefixlen != 128)
541 if (ifal->ifal_index &&
542 !addrlbl_ifindex_exists(net, ifal->ifal_index))
545 if (!tb[IFAL_ADDRESS])
547 addr = nla_data(tb[IFAL_ADDRESS]);
550 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
551 if (p && !ip6addrlbl_hold(p))
553 lseq = net->ipv6.ip6addrlbl_table.seq;
561 skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL);
567 err = ip6addrlbl_fill(skb, p, lseq,
568 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
569 RTM_NEWADDRLABEL, 0);
574 WARN_ON(err == -EMSGSIZE);
579 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
584 void __init ipv6_addr_label_rtnl_register(void)
586 __rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel,
587 NULL, RTNL_FLAG_DOIT_UNLOCKED);
588 __rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel,
589 NULL, RTNL_FLAG_DOIT_UNLOCKED);
590 __rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get,
591 ip6addrlbl_dump, RTNL_FLAG_DOIT_UNLOCKED);