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>
23 #define ADDRLABEL(x...) printk(x)
25 #define ADDRLABEL(x...) do { ; } while (0)
31 struct ip6addrlbl_entry {
32 struct in6_addr prefix;
37 struct hlist_node list;
42 * Default policy table (RFC6724 + extensions)
44 * prefix addr_type label
45 * -------------------------------------------------------------------------
50 * ::ffff:0:0/96 V4MAPPED 4
51 * fc00::/7 N/A 5 ULA (RFC 4193)
52 * 2001::/32 N/A 6 Teredo (RFC 4380)
53 * 2001:10::/28 N/A 7 ORCHID (RFC 4843)
54 * fec0::/10 N/A 11 Site-local
55 * (deprecated by RFC3879)
56 * 3ffe::/16 N/A 12 6bone
58 * Note: 0xffffffff is used if we do not have any policies.
59 * Note: Labels for ULA and 6to4 are different from labels listed in RFC6724.
62 #define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
64 static const __net_initconst struct ip6addrlbl_init_table
66 const struct in6_addr *prefix;
69 } ip6addrlbl_init_table[] = {
71 .prefix = &in6addr_any,
74 .prefix = &(struct in6_addr){ { { 0xfc } } } ,
78 .prefix = &(struct in6_addr){ { { 0xfe, 0xc0 } } },
82 .prefix = &(struct in6_addr){ { { 0x20, 0x02 } } },
86 .prefix = &(struct in6_addr){ { { 0x3f, 0xfe } } },
90 .prefix = &(struct in6_addr){ { { 0x20, 0x01 } } },
93 }, { /* 2001:10::/28 */
94 .prefix = &(struct in6_addr){ { { 0x20, 0x01, 0x00, 0x10 } } },
98 .prefix = &(struct in6_addr){ { { [10] = 0xff, [11] = 0xff } } },
102 .prefix = &in6addr_any,
106 .prefix = &in6addr_loopback,
113 static bool __ip6addrlbl_match(const struct ip6addrlbl_entry *p,
114 const struct in6_addr *addr,
115 int addrtype, int ifindex)
117 if (p->ifindex && p->ifindex != ifindex)
119 if (p->addrtype && p->addrtype != addrtype)
121 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
126 static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
127 const struct in6_addr *addr,
128 int type, int ifindex)
130 struct ip6addrlbl_entry *p;
132 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
133 if (__ip6addrlbl_match(p, addr, type, ifindex))
139 u32 ipv6_addr_label(struct net *net,
140 const struct in6_addr *addr, int type, int ifindex)
143 struct ip6addrlbl_entry *p;
145 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
148 p = __ipv6_addr_label(net, addr, type, ifindex);
149 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
152 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
153 __func__, addr, type, ifindex, label);
158 /* allocate one entry */
159 static struct ip6addrlbl_entry *ip6addrlbl_alloc(const struct in6_addr *prefix,
160 int prefixlen, int ifindex,
163 struct ip6addrlbl_entry *newp;
166 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
167 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
169 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
172 case IPV6_ADDR_MAPPED:
174 return ERR_PTR(-EINVAL);
178 case IPV6_ADDR_COMPATv4:
182 case IPV6_ADDR_LOOPBACK:
183 if (prefixlen != 128)
188 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
190 return ERR_PTR(-ENOMEM);
192 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
193 newp->prefixlen = prefixlen;
194 newp->ifindex = ifindex;
195 newp->addrtype = addrtype;
197 INIT_HLIST_NODE(&newp->list);
202 static int __ip6addrlbl_add(struct net *net, struct ip6addrlbl_entry *newp,
205 struct ip6addrlbl_entry *last = NULL, *p = NULL;
206 struct hlist_node *n;
209 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", __func__, newp,
212 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
213 if (p->prefixlen == newp->prefixlen &&
214 p->ifindex == newp->ifindex &&
215 ipv6_addr_equal(&p->prefix, &newp->prefix)) {
220 hlist_replace_rcu(&p->list, &newp->list);
223 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
224 (p->prefixlen < newp->prefixlen)) {
225 hlist_add_before_rcu(&newp->list, &p->list);
231 hlist_add_behind_rcu(&newp->list, &last->list);
233 hlist_add_head_rcu(&newp->list, &net->ipv6.ip6addrlbl_table.head);
236 net->ipv6.ip6addrlbl_table.seq++;
241 static int ip6addrlbl_add(struct net *net,
242 const struct in6_addr *prefix, int prefixlen,
243 int ifindex, u32 label, int replace)
245 struct ip6addrlbl_entry *newp;
248 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
249 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
252 newp = ip6addrlbl_alloc(prefix, prefixlen, ifindex, label);
254 return PTR_ERR(newp);
255 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
256 ret = __ip6addrlbl_add(net, newp, replace);
257 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
264 static int __ip6addrlbl_del(struct net *net,
265 const struct in6_addr *prefix, int prefixlen,
268 struct ip6addrlbl_entry *p = NULL;
269 struct hlist_node *n;
272 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
273 __func__, prefix, prefixlen, ifindex);
275 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
276 if (p->prefixlen == prefixlen &&
277 p->ifindex == ifindex &&
278 ipv6_addr_equal(&p->prefix, prefix)) {
279 hlist_del_rcu(&p->list);
288 static int ip6addrlbl_del(struct net *net,
289 const struct in6_addr *prefix, int prefixlen,
292 struct in6_addr prefix_buf;
295 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
296 __func__, prefix, prefixlen, ifindex);
298 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
299 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
300 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
301 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
305 /* add default label */
306 static int __net_init ip6addrlbl_net_init(struct net *net)
311 ADDRLABEL(KERN_DEBUG "%s\n", __func__);
313 spin_lock_init(&net->ipv6.ip6addrlbl_table.lock);
314 INIT_HLIST_HEAD(&net->ipv6.ip6addrlbl_table.head);
316 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
317 int ret = ip6addrlbl_add(net,
318 ip6addrlbl_init_table[i].prefix,
319 ip6addrlbl_init_table[i].prefixlen,
321 ip6addrlbl_init_table[i].label, 0);
322 /* XXX: should we free all rules when we catch an error? */
323 if (ret && (!err || err != -ENOMEM))
329 static void __net_exit ip6addrlbl_net_exit(struct net *net)
331 struct ip6addrlbl_entry *p = NULL;
332 struct hlist_node *n;
334 /* Remove all labels belonging to the exiting net */
335 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
336 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
337 hlist_del_rcu(&p->list);
340 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
343 static struct pernet_operations ipv6_addr_label_ops = {
344 .init = ip6addrlbl_net_init,
345 .exit = ip6addrlbl_net_exit,
348 int __init ipv6_addr_label_init(void)
350 return register_pernet_subsys(&ipv6_addr_label_ops);
353 void ipv6_addr_label_cleanup(void)
355 unregister_pernet_subsys(&ipv6_addr_label_ops);
358 static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
359 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
360 [IFAL_LABEL] = { .len = sizeof(u32), },
363 static bool addrlbl_ifindex_exists(struct net *net, int ifindex)
366 struct net_device *dev;
369 dev = dev_get_by_index_rcu(net, ifindex);
375 static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
376 struct netlink_ext_ack *extack)
378 struct net *net = sock_net(skb->sk);
379 struct ifaddrlblmsg *ifal;
380 struct nlattr *tb[IFAL_MAX+1];
381 struct in6_addr *pfx;
385 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
390 ifal = nlmsg_data(nlh);
392 if (ifal->ifal_family != AF_INET6 ||
393 ifal->ifal_prefixlen > 128)
396 if (!tb[IFAL_ADDRESS])
398 pfx = nla_data(tb[IFAL_ADDRESS]);
402 label = nla_get_u32(tb[IFAL_LABEL]);
403 if (label == IPV6_ADDR_LABEL_DEFAULT)
406 switch (nlh->nlmsg_type) {
407 case RTM_NEWADDRLABEL:
408 if (ifal->ifal_index &&
409 !addrlbl_ifindex_exists(net, ifal->ifal_index))
412 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
413 ifal->ifal_index, label,
414 nlh->nlmsg_flags & NLM_F_REPLACE);
416 case RTM_DELADDRLABEL:
417 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
426 static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
427 int prefixlen, int ifindex, u32 lseq)
429 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
430 ifal->ifal_family = AF_INET6;
431 ifal->ifal_prefixlen = prefixlen;
432 ifal->ifal_flags = 0;
433 ifal->ifal_index = ifindex;
434 ifal->ifal_seq = lseq;
437 static int ip6addrlbl_fill(struct sk_buff *skb,
438 struct ip6addrlbl_entry *p,
440 u32 portid, u32 seq, int event,
443 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, event,
444 sizeof(struct ifaddrlblmsg), flags);
448 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
450 if (nla_put_in6_addr(skb, IFAL_ADDRESS, &p->prefix) < 0 ||
451 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
452 nlmsg_cancel(skb, nlh);
460 static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
462 struct net *net = sock_net(skb->sk);
463 struct ip6addrlbl_entry *p;
464 int idx = 0, s_idx = cb->args[0];
468 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
470 err = ip6addrlbl_fill(skb, p,
471 net->ipv6.ip6addrlbl_table.seq,
472 NETLINK_CB(cb->skb).portid,
486 static inline int ip6addrlbl_msgsize(void)
488 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
489 + nla_total_size(16) /* IFAL_ADDRESS */
490 + nla_total_size(4); /* IFAL_LABEL */
493 static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
494 struct netlink_ext_ack *extack)
496 struct net *net = sock_net(in_skb->sk);
497 struct ifaddrlblmsg *ifal;
498 struct nlattr *tb[IFAL_MAX+1];
499 struct in6_addr *addr;
502 struct ip6addrlbl_entry *p;
505 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
510 ifal = nlmsg_data(nlh);
512 if (ifal->ifal_family != AF_INET6 ||
513 ifal->ifal_prefixlen != 128)
516 if (ifal->ifal_index &&
517 !addrlbl_ifindex_exists(net, ifal->ifal_index))
520 if (!tb[IFAL_ADDRESS])
522 addr = nla_data(tb[IFAL_ADDRESS]);
524 skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL);
531 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
532 lseq = net->ipv6.ip6addrlbl_table.seq;
534 err = ip6addrlbl_fill(skb, p, lseq,
535 NETLINK_CB(in_skb).portid,
537 RTM_NEWADDRLABEL, 0);
541 WARN_ON(err == -EMSGSIZE);
544 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
549 void __init ipv6_addr_label_rtnl_register(void)
551 __rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel,
552 NULL, RTNL_FLAG_DOIT_UNLOCKED);
553 __rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel,
554 NULL, RTNL_FLAG_DOIT_UNLOCKED);
555 __rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get,
556 ip6addrlbl_dump, RTNL_FLAG_DOIT_UNLOCKED);