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>
22 #include <linux/refcount.h>
25 #define ADDRLABEL(x...) printk(x)
27 #define ADDRLABEL(x...) do { ; } while (0)
33 struct ip6addrlbl_entry {
34 possible_net_t lbl_net;
35 struct in6_addr prefix;
40 struct hlist_node list;
45 static struct ip6addrlbl_table
47 struct hlist_head head;
53 struct net *ip6addrlbl_net(const struct ip6addrlbl_entry *lbl)
55 return read_pnet(&lbl->lbl_net);
59 * Default policy table (RFC6724 + extensions)
61 * prefix addr_type label
62 * -------------------------------------------------------------------------
67 * ::ffff:0:0/96 V4MAPPED 4
68 * fc00::/7 N/A 5 ULA (RFC 4193)
69 * 2001::/32 N/A 6 Teredo (RFC 4380)
70 * 2001:10::/28 N/A 7 ORCHID (RFC 4843)
71 * fec0::/10 N/A 11 Site-local
72 * (deprecated by RFC3879)
73 * 3ffe::/16 N/A 12 6bone
75 * Note: 0xffffffff is used if we do not have any policies.
76 * Note: Labels for ULA and 6to4 are different from labels listed in RFC6724.
79 #define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
81 static const __net_initconst struct ip6addrlbl_init_table
83 const struct in6_addr *prefix;
86 } ip6addrlbl_init_table[] = {
88 .prefix = &in6addr_any,
91 .prefix = &(struct in6_addr){ { { 0xfc } } } ,
95 .prefix = &(struct in6_addr){ { { 0xfe, 0xc0 } } },
99 .prefix = &(struct in6_addr){ { { 0x20, 0x02 } } },
103 .prefix = &(struct in6_addr){ { { 0x3f, 0xfe } } },
107 .prefix = &(struct in6_addr){ { { 0x20, 0x01 } } },
110 }, { /* 2001:10::/28 */
111 .prefix = &(struct in6_addr){ { { 0x20, 0x01, 0x00, 0x10 } } },
114 }, { /* ::ffff:0:0 */
115 .prefix = &(struct in6_addr){ { { [10] = 0xff, [11] = 0xff } } },
119 .prefix = &in6addr_any,
123 .prefix = &in6addr_loopback,
129 /* Object management */
130 static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p)
135 static void ip6addrlbl_free_rcu(struct rcu_head *h)
137 ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu));
140 static bool ip6addrlbl_hold(struct ip6addrlbl_entry *p)
142 return refcount_inc_not_zero(&p->refcnt);
145 static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p)
147 if (refcount_dec_and_test(&p->refcnt))
148 call_rcu(&p->rcu, ip6addrlbl_free_rcu);
152 static bool __ip6addrlbl_match(struct net *net,
153 const struct ip6addrlbl_entry *p,
154 const struct in6_addr *addr,
155 int addrtype, int ifindex)
157 if (!net_eq(ip6addrlbl_net(p), net))
159 if (p->ifindex && p->ifindex != ifindex)
161 if (p->addrtype && p->addrtype != addrtype)
163 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
168 static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
169 const struct in6_addr *addr,
170 int type, int ifindex)
172 struct ip6addrlbl_entry *p;
173 hlist_for_each_entry_rcu(p, &ip6addrlbl_table.head, list) {
174 if (__ip6addrlbl_match(net, p, addr, type, ifindex))
180 u32 ipv6_addr_label(struct net *net,
181 const struct in6_addr *addr, int type, int ifindex)
184 struct ip6addrlbl_entry *p;
186 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
189 p = __ipv6_addr_label(net, addr, type, ifindex);
190 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
193 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
194 __func__, addr, type, ifindex, label);
199 /* allocate one entry */
200 static struct ip6addrlbl_entry *ip6addrlbl_alloc(struct net *net,
201 const struct in6_addr *prefix,
202 int prefixlen, int ifindex,
205 struct ip6addrlbl_entry *newp;
208 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
209 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
211 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
214 case IPV6_ADDR_MAPPED:
216 return ERR_PTR(-EINVAL);
220 case IPV6_ADDR_COMPATv4:
224 case IPV6_ADDR_LOOPBACK:
225 if (prefixlen != 128)
230 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
232 return ERR_PTR(-ENOMEM);
234 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
235 newp->prefixlen = prefixlen;
236 newp->ifindex = ifindex;
237 newp->addrtype = addrtype;
239 INIT_HLIST_NODE(&newp->list);
240 write_pnet(&newp->lbl_net, net);
241 refcount_set(&newp->refcnt, 1);
246 static int __ip6addrlbl_add(struct ip6addrlbl_entry *newp, int replace)
248 struct hlist_node *n;
249 struct ip6addrlbl_entry *last = NULL, *p = NULL;
252 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", __func__, newp,
255 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) {
256 if (p->prefixlen == newp->prefixlen &&
257 net_eq(ip6addrlbl_net(p), ip6addrlbl_net(newp)) &&
258 p->ifindex == newp->ifindex &&
259 ipv6_addr_equal(&p->prefix, &newp->prefix)) {
264 hlist_replace_rcu(&p->list, &newp->list);
267 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
268 (p->prefixlen < newp->prefixlen)) {
269 hlist_add_before_rcu(&newp->list, &p->list);
275 hlist_add_behind_rcu(&newp->list, &last->list);
277 hlist_add_head_rcu(&newp->list, &ip6addrlbl_table.head);
280 ip6addrlbl_table.seq++;
285 static int ip6addrlbl_add(struct net *net,
286 const struct in6_addr *prefix, int prefixlen,
287 int ifindex, u32 label, int replace)
289 struct ip6addrlbl_entry *newp;
292 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
293 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
296 newp = ip6addrlbl_alloc(net, prefix, prefixlen, ifindex, label);
298 return PTR_ERR(newp);
299 spin_lock(&ip6addrlbl_table.lock);
300 ret = __ip6addrlbl_add(newp, replace);
301 spin_unlock(&ip6addrlbl_table.lock);
303 ip6addrlbl_free(newp);
308 static int __ip6addrlbl_del(struct net *net,
309 const struct in6_addr *prefix, int prefixlen,
312 struct ip6addrlbl_entry *p = NULL;
313 struct hlist_node *n;
316 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
317 __func__, prefix, prefixlen, ifindex);
319 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) {
320 if (p->prefixlen == prefixlen &&
321 net_eq(ip6addrlbl_net(p), net) &&
322 p->ifindex == ifindex &&
323 ipv6_addr_equal(&p->prefix, prefix)) {
324 hlist_del_rcu(&p->list);
333 static int ip6addrlbl_del(struct net *net,
334 const struct in6_addr *prefix, int prefixlen,
337 struct in6_addr prefix_buf;
340 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
341 __func__, prefix, prefixlen, ifindex);
343 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
344 spin_lock(&ip6addrlbl_table.lock);
345 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
346 spin_unlock(&ip6addrlbl_table.lock);
350 /* add default label */
351 static int __net_init ip6addrlbl_net_init(struct net *net)
356 ADDRLABEL(KERN_DEBUG "%s\n", __func__);
358 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
359 int ret = ip6addrlbl_add(net,
360 ip6addrlbl_init_table[i].prefix,
361 ip6addrlbl_init_table[i].prefixlen,
363 ip6addrlbl_init_table[i].label, 0);
364 /* XXX: should we free all rules when we catch an error? */
365 if (ret && (!err || err != -ENOMEM))
371 static void __net_exit ip6addrlbl_net_exit(struct net *net)
373 struct ip6addrlbl_entry *p = NULL;
374 struct hlist_node *n;
376 /* Remove all labels belonging to the exiting net */
377 spin_lock(&ip6addrlbl_table.lock);
378 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) {
379 if (net_eq(ip6addrlbl_net(p), net)) {
380 hlist_del_rcu(&p->list);
384 spin_unlock(&ip6addrlbl_table.lock);
387 static struct pernet_operations ipv6_addr_label_ops = {
388 .init = ip6addrlbl_net_init,
389 .exit = ip6addrlbl_net_exit,
392 int __init ipv6_addr_label_init(void)
394 spin_lock_init(&ip6addrlbl_table.lock);
396 return register_pernet_subsys(&ipv6_addr_label_ops);
399 void ipv6_addr_label_cleanup(void)
401 unregister_pernet_subsys(&ipv6_addr_label_ops);
404 static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
405 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
406 [IFAL_LABEL] = { .len = sizeof(u32), },
409 static bool addrlbl_ifindex_exists(struct net *net, int ifindex)
412 struct net_device *dev;
415 dev = dev_get_by_index_rcu(net, ifindex);
421 static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
422 struct netlink_ext_ack *extack)
424 struct net *net = sock_net(skb->sk);
425 struct ifaddrlblmsg *ifal;
426 struct nlattr *tb[IFAL_MAX+1];
427 struct in6_addr *pfx;
431 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
436 ifal = nlmsg_data(nlh);
438 if (ifal->ifal_family != AF_INET6 ||
439 ifal->ifal_prefixlen > 128)
442 if (!tb[IFAL_ADDRESS])
444 pfx = nla_data(tb[IFAL_ADDRESS]);
448 label = nla_get_u32(tb[IFAL_LABEL]);
449 if (label == IPV6_ADDR_LABEL_DEFAULT)
452 switch (nlh->nlmsg_type) {
453 case RTM_NEWADDRLABEL:
454 if (ifal->ifal_index &&
455 !addrlbl_ifindex_exists(net, ifal->ifal_index))
458 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
459 ifal->ifal_index, label,
460 nlh->nlmsg_flags & NLM_F_REPLACE);
462 case RTM_DELADDRLABEL:
463 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
472 static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
473 int prefixlen, int ifindex, u32 lseq)
475 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
476 ifal->ifal_family = AF_INET6;
477 ifal->ifal_prefixlen = prefixlen;
478 ifal->ifal_flags = 0;
479 ifal->ifal_index = ifindex;
480 ifal->ifal_seq = lseq;
483 static int ip6addrlbl_fill(struct sk_buff *skb,
484 struct ip6addrlbl_entry *p,
486 u32 portid, u32 seq, int event,
489 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, event,
490 sizeof(struct ifaddrlblmsg), flags);
494 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
496 if (nla_put_in6_addr(skb, IFAL_ADDRESS, &p->prefix) < 0 ||
497 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
498 nlmsg_cancel(skb, nlh);
506 static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
508 struct net *net = sock_net(skb->sk);
509 struct ip6addrlbl_entry *p;
510 int idx = 0, s_idx = cb->args[0];
514 hlist_for_each_entry_rcu(p, &ip6addrlbl_table.head, list) {
516 net_eq(ip6addrlbl_net(p), net)) {
517 err = ip6addrlbl_fill(skb, p,
518 ip6addrlbl_table.seq,
519 NETLINK_CB(cb->skb).portid,
533 static inline int ip6addrlbl_msgsize(void)
535 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
536 + nla_total_size(16) /* IFAL_ADDRESS */
537 + nla_total_size(4); /* IFAL_LABEL */
540 static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
541 struct netlink_ext_ack *extack)
543 struct net *net = sock_net(in_skb->sk);
544 struct ifaddrlblmsg *ifal;
545 struct nlattr *tb[IFAL_MAX+1];
546 struct in6_addr *addr;
549 struct ip6addrlbl_entry *p;
552 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
557 ifal = nlmsg_data(nlh);
559 if (ifal->ifal_family != AF_INET6 ||
560 ifal->ifal_prefixlen != 128)
563 if (ifal->ifal_index &&
564 !addrlbl_ifindex_exists(net, ifal->ifal_index))
567 if (!tb[IFAL_ADDRESS])
569 addr = nla_data(tb[IFAL_ADDRESS]);
572 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
573 if (p && !ip6addrlbl_hold(p))
575 lseq = ip6addrlbl_table.seq;
583 skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL);
589 err = ip6addrlbl_fill(skb, p, lseq,
590 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
591 RTM_NEWADDRLABEL, 0);
596 WARN_ON(err == -EMSGSIZE);
601 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
606 void __init ipv6_addr_label_rtnl_register(void)
608 __rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel,
609 NULL, RTNL_FLAG_DOIT_UNLOCKED);
610 __rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel,
611 NULL, RTNL_FLAG_DOIT_UNLOCKED);
612 __rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get,
613 ip6addrlbl_dump, RTNL_FLAG_DOIT_UNLOCKED);