1 // SPDX-License-Identifier: GPL-2.0
3 * IPv6 Address Label subsystem
4 * for the IPv6 "Default" Source Address Selection
6 * Copyright (C)2007 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki @ USAGI/WIDE Project <yoshfuji@linux-ipv6.org>
13 #include <linux/kernel.h>
14 #include <linux/list.h>
15 #include <linux/rcupdate.h>
16 #include <linux/in6.h>
17 #include <linux/slab.h>
18 #include <net/addrconf.h>
19 #include <linux/if_addrlabel.h>
20 #include <linux/netlink.h>
21 #include <linux/rtnetlink.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;
43 * Default policy table (RFC6724 + extensions)
45 * prefix addr_type label
46 * -------------------------------------------------------------------------
51 * ::ffff:0:0/96 V4MAPPED 4
52 * fc00::/7 N/A 5 ULA (RFC 4193)
53 * 2001::/32 N/A 6 Teredo (RFC 4380)
54 * 2001:10::/28 N/A 7 ORCHID (RFC 4843)
55 * fec0::/10 N/A 11 Site-local
56 * (deprecated by RFC3879)
57 * 3ffe::/16 N/A 12 6bone
59 * Note: 0xffffffff is used if we do not have any policies.
60 * Note: Labels for ULA and 6to4 are different from labels listed in RFC6724.
63 #define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
65 static const __net_initconst struct ip6addrlbl_init_table
67 const struct in6_addr *prefix;
70 } ip6addrlbl_init_table[] = {
72 .prefix = &in6addr_any,
75 .prefix = &(struct in6_addr){ { { 0xfc } } } ,
79 .prefix = &(struct in6_addr){ { { 0xfe, 0xc0 } } },
83 .prefix = &(struct in6_addr){ { { 0x20, 0x02 } } },
87 .prefix = &(struct in6_addr){ { { 0x3f, 0xfe } } },
91 .prefix = &(struct in6_addr){ { { 0x20, 0x01 } } },
94 }, { /* 2001:10::/28 */
95 .prefix = &(struct in6_addr){ { { 0x20, 0x01, 0x00, 0x10 } } },
99 .prefix = &(struct in6_addr){ { { [10] = 0xff, [11] = 0xff } } },
103 .prefix = &in6addr_any,
107 .prefix = &in6addr_loopback,
114 static bool __ip6addrlbl_match(const struct ip6addrlbl_entry *p,
115 const struct in6_addr *addr,
116 int addrtype, int ifindex)
118 if (p->ifindex && p->ifindex != ifindex)
120 if (p->addrtype && p->addrtype != addrtype)
122 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
127 static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
128 const struct in6_addr *addr,
129 int type, int ifindex)
131 struct ip6addrlbl_entry *p;
133 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
134 if (__ip6addrlbl_match(p, addr, type, ifindex))
140 u32 ipv6_addr_label(struct net *net,
141 const struct in6_addr *addr, int type, int ifindex)
144 struct ip6addrlbl_entry *p;
146 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
149 p = __ipv6_addr_label(net, addr, type, ifindex);
150 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
153 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
154 __func__, addr, type, ifindex, label);
159 /* allocate one entry */
160 static struct ip6addrlbl_entry *ip6addrlbl_alloc(const struct in6_addr *prefix,
161 int prefixlen, int ifindex,
164 struct ip6addrlbl_entry *newp;
167 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
168 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
170 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
173 case IPV6_ADDR_MAPPED:
175 return ERR_PTR(-EINVAL);
179 case IPV6_ADDR_COMPATv4:
183 case IPV6_ADDR_LOOPBACK:
184 if (prefixlen != 128)
189 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
191 return ERR_PTR(-ENOMEM);
193 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
194 newp->prefixlen = prefixlen;
195 newp->ifindex = ifindex;
196 newp->addrtype = addrtype;
198 INIT_HLIST_NODE(&newp->list);
203 static int __ip6addrlbl_add(struct net *net, struct ip6addrlbl_entry *newp,
206 struct ip6addrlbl_entry *last = NULL, *p = NULL;
207 struct hlist_node *n;
210 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", __func__, newp,
213 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
214 if (p->prefixlen == newp->prefixlen &&
215 p->ifindex == newp->ifindex &&
216 ipv6_addr_equal(&p->prefix, &newp->prefix)) {
221 hlist_replace_rcu(&p->list, &newp->list);
224 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
225 (p->prefixlen < newp->prefixlen)) {
226 hlist_add_before_rcu(&newp->list, &p->list);
232 hlist_add_behind_rcu(&newp->list, &last->list);
234 hlist_add_head_rcu(&newp->list, &net->ipv6.ip6addrlbl_table.head);
237 WRITE_ONCE(net->ipv6.ip6addrlbl_table.seq,
238 net->ipv6.ip6addrlbl_table.seq + 1);
243 static int ip6addrlbl_add(struct net *net,
244 const struct in6_addr *prefix, int prefixlen,
245 int ifindex, u32 label, int replace)
247 struct ip6addrlbl_entry *newp;
250 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
251 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
254 newp = ip6addrlbl_alloc(prefix, prefixlen, ifindex, label);
256 return PTR_ERR(newp);
257 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
258 ret = __ip6addrlbl_add(net, newp, replace);
259 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
266 static int __ip6addrlbl_del(struct net *net,
267 const struct in6_addr *prefix, int prefixlen,
270 struct ip6addrlbl_entry *p = NULL;
271 struct hlist_node *n;
274 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
275 __func__, prefix, prefixlen, ifindex);
277 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
278 if (p->prefixlen == prefixlen &&
279 p->ifindex == ifindex &&
280 ipv6_addr_equal(&p->prefix, prefix)) {
281 hlist_del_rcu(&p->list);
290 static int ip6addrlbl_del(struct net *net,
291 const struct in6_addr *prefix, int prefixlen,
294 struct in6_addr prefix_buf;
297 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
298 __func__, prefix, prefixlen, ifindex);
300 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
301 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
302 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
303 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
307 /* add default label */
308 static int __net_init ip6addrlbl_net_init(struct net *net)
310 struct ip6addrlbl_entry *p = NULL;
311 struct hlist_node *n;
315 ADDRLABEL(KERN_DEBUG "%s\n", __func__);
317 spin_lock_init(&net->ipv6.ip6addrlbl_table.lock);
318 INIT_HLIST_HEAD(&net->ipv6.ip6addrlbl_table.head);
320 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
321 err = ip6addrlbl_add(net,
322 ip6addrlbl_init_table[i].prefix,
323 ip6addrlbl_init_table[i].prefixlen,
325 ip6addrlbl_init_table[i].label, 0);
327 goto err_ip6addrlbl_add;
332 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
333 hlist_del_rcu(&p->list);
339 static void __net_exit ip6addrlbl_net_exit(struct net *net)
341 struct ip6addrlbl_entry *p = NULL;
342 struct hlist_node *n;
344 /* Remove all labels belonging to the exiting net */
345 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
346 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
347 hlist_del_rcu(&p->list);
350 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
353 static struct pernet_operations ipv6_addr_label_ops = {
354 .init = ip6addrlbl_net_init,
355 .exit = ip6addrlbl_net_exit,
358 int __init ipv6_addr_label_init(void)
360 return register_pernet_subsys(&ipv6_addr_label_ops);
363 void ipv6_addr_label_cleanup(void)
365 unregister_pernet_subsys(&ipv6_addr_label_ops);
368 static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
369 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
370 [IFAL_LABEL] = { .len = sizeof(u32), },
373 static bool addrlbl_ifindex_exists(struct net *net, int ifindex)
376 struct net_device *dev;
379 dev = dev_get_by_index_rcu(net, ifindex);
385 static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
386 struct netlink_ext_ack *extack)
388 struct net *net = sock_net(skb->sk);
389 struct ifaddrlblmsg *ifal;
390 struct nlattr *tb[IFAL_MAX+1];
391 struct in6_addr *pfx;
395 err = nlmsg_parse_deprecated(nlh, sizeof(*ifal), tb, IFAL_MAX,
396 ifal_policy, extack);
400 ifal = nlmsg_data(nlh);
402 if (ifal->ifal_family != AF_INET6 ||
403 ifal->ifal_prefixlen > 128)
406 if (!tb[IFAL_ADDRESS])
408 pfx = nla_data(tb[IFAL_ADDRESS]);
412 label = nla_get_u32(tb[IFAL_LABEL]);
413 if (label == IPV6_ADDR_LABEL_DEFAULT)
416 switch (nlh->nlmsg_type) {
417 case RTM_NEWADDRLABEL:
418 if (ifal->ifal_index &&
419 !addrlbl_ifindex_exists(net, ifal->ifal_index))
422 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
423 ifal->ifal_index, label,
424 nlh->nlmsg_flags & NLM_F_REPLACE);
426 case RTM_DELADDRLABEL:
427 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
436 static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
437 int prefixlen, int ifindex, u32 lseq)
439 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
440 ifal->ifal_family = AF_INET6;
441 ifal->__ifal_reserved = 0;
442 ifal->ifal_prefixlen = prefixlen;
443 ifal->ifal_flags = 0;
444 ifal->ifal_index = ifindex;
445 ifal->ifal_seq = lseq;
448 static int ip6addrlbl_fill(struct sk_buff *skb,
449 const struct ip6addrlbl_entry *p,
451 u32 portid, u32 seq, int event,
454 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, event,
455 sizeof(struct ifaddrlblmsg), flags);
459 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
461 if (nla_put_in6_addr(skb, IFAL_ADDRESS, &p->prefix) < 0 ||
462 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
463 nlmsg_cancel(skb, nlh);
471 static int ip6addrlbl_valid_dump_req(const struct nlmsghdr *nlh,
472 struct netlink_ext_ack *extack)
474 struct ifaddrlblmsg *ifal;
476 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifal))) {
477 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address label dump request");
481 ifal = nlmsg_data(nlh);
482 if (ifal->__ifal_reserved || ifal->ifal_prefixlen ||
483 ifal->ifal_flags || ifal->ifal_index || ifal->ifal_seq) {
484 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address label dump request");
488 if (nlmsg_attrlen(nlh, sizeof(*ifal))) {
489 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header for address label dump request");
496 static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
498 const struct nlmsghdr *nlh = cb->nlh;
499 struct net *net = sock_net(skb->sk);
500 struct ip6addrlbl_entry *p;
501 int idx = 0, s_idx = cb->args[0];
505 if (cb->strict_check) {
506 err = ip6addrlbl_valid_dump_req(nlh, cb->extack);
512 lseq = READ_ONCE(net->ipv6.ip6addrlbl_table.seq);
513 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
515 err = ip6addrlbl_fill(skb, p,
517 NETLINK_CB(cb->skb).portid,
531 static inline int ip6addrlbl_msgsize(void)
533 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
534 + nla_total_size(16) /* IFAL_ADDRESS */
535 + nla_total_size(4); /* IFAL_LABEL */
538 static int ip6addrlbl_valid_get_req(struct sk_buff *skb,
539 const struct nlmsghdr *nlh,
541 struct netlink_ext_ack *extack)
543 struct ifaddrlblmsg *ifal;
546 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifal))) {
547 NL_SET_ERR_MSG_MOD(extack, "Invalid header for addrlabel get request");
551 if (!netlink_strict_get_check(skb))
552 return nlmsg_parse_deprecated(nlh, sizeof(*ifal), tb,
553 IFAL_MAX, ifal_policy, extack);
555 ifal = nlmsg_data(nlh);
556 if (ifal->__ifal_reserved || ifal->ifal_flags || ifal->ifal_seq) {
557 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for addrlabel get request");
561 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifal), tb, IFAL_MAX,
562 ifal_policy, extack);
566 for (i = 0; i <= IFAL_MAX; i++) {
574 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in addrlabel get request");
582 static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
583 struct netlink_ext_ack *extack)
585 struct net *net = sock_net(in_skb->sk);
586 struct ifaddrlblmsg *ifal;
587 struct nlattr *tb[IFAL_MAX+1];
588 struct in6_addr *addr;
591 struct ip6addrlbl_entry *p;
594 err = ip6addrlbl_valid_get_req(in_skb, nlh, tb, extack);
598 ifal = nlmsg_data(nlh);
600 if (ifal->ifal_family != AF_INET6 ||
601 ifal->ifal_prefixlen != 128)
604 if (ifal->ifal_index &&
605 !addrlbl_ifindex_exists(net, ifal->ifal_index))
608 if (!tb[IFAL_ADDRESS])
610 addr = nla_data(tb[IFAL_ADDRESS]);
612 skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL);
619 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
620 lseq = READ_ONCE(net->ipv6.ip6addrlbl_table.seq);
622 err = ip6addrlbl_fill(skb, p, lseq,
623 NETLINK_CB(in_skb).portid,
625 RTM_NEWADDRLABEL, 0);
629 WARN_ON(err == -EMSGSIZE);
632 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
637 int __init ipv6_addr_label_rtnl_register(void)
641 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDRLABEL,
643 NULL, RTNL_FLAG_DOIT_UNLOCKED);
646 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDRLABEL,
648 NULL, RTNL_FLAG_DOIT_UNLOCKED);
651 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDRLABEL,
653 ip6addrlbl_dump, RTNL_FLAG_DOIT_UNLOCKED |
654 RTNL_FLAG_DUMP_UNLOCKED);