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 possible_net_t lbl_net;
34 struct in6_addr prefix;
39 struct hlist_node list;
44 static struct ip6addrlbl_table
46 struct hlist_head head;
52 struct net *ip6addrlbl_net(const struct ip6addrlbl_entry *lbl)
54 return read_pnet(&lbl->lbl_net);
58 * Default policy table (RFC6724 + extensions)
60 * prefix addr_type label
61 * -------------------------------------------------------------------------
66 * ::ffff:0:0/96 V4MAPPED 4
67 * fc00::/7 N/A 5 ULA (RFC 4193)
68 * 2001::/32 N/A 6 Teredo (RFC 4380)
69 * 2001:10::/28 N/A 7 ORCHID (RFC 4843)
70 * fec0::/10 N/A 11 Site-local
71 * (deprecated by RFC3879)
72 * 3ffe::/16 N/A 12 6bone
74 * Note: 0xffffffff is used if we do not have any policies.
75 * Note: Labels for ULA and 6to4 are different from labels listed in RFC6724.
78 #define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
80 static const __net_initconst struct ip6addrlbl_init_table
82 const struct in6_addr *prefix;
85 } ip6addrlbl_init_table[] = {
87 .prefix = &in6addr_any,
90 .prefix = &(struct in6_addr){ { { 0xfc } } } ,
94 .prefix = &(struct in6_addr){ { { 0xfe, 0xc0 } } },
98 .prefix = &(struct in6_addr){ { { 0x20, 0x02 } } },
102 .prefix = &(struct in6_addr){ { { 0x3f, 0xfe } } },
106 .prefix = &(struct in6_addr){ { { 0x20, 0x01 } } },
109 }, { /* 2001:10::/28 */
110 .prefix = &(struct in6_addr){ { { 0x20, 0x01, 0x00, 0x10 } } },
113 }, { /* ::ffff:0:0 */
114 .prefix = &(struct in6_addr){ { { [10] = 0xff, [11] = 0xff } } },
118 .prefix = &in6addr_any,
122 .prefix = &in6addr_loopback,
128 /* Object management */
129 static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p)
134 static void ip6addrlbl_free_rcu(struct rcu_head *h)
136 ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu));
139 static bool ip6addrlbl_hold(struct ip6addrlbl_entry *p)
141 return refcount_inc_not_zero(&p->refcnt);
144 static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p)
146 if (refcount_dec_and_test(&p->refcnt))
147 call_rcu(&p->rcu, ip6addrlbl_free_rcu);
151 static bool __ip6addrlbl_match(struct net *net,
152 const struct ip6addrlbl_entry *p,
153 const struct in6_addr *addr,
154 int addrtype, int ifindex)
156 if (!net_eq(ip6addrlbl_net(p), net))
158 if (p->ifindex && p->ifindex != ifindex)
160 if (p->addrtype && p->addrtype != addrtype)
162 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
167 static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
168 const struct in6_addr *addr,
169 int type, int ifindex)
171 struct ip6addrlbl_entry *p;
172 hlist_for_each_entry_rcu(p, &ip6addrlbl_table.head, list) {
173 if (__ip6addrlbl_match(net, p, addr, type, ifindex))
179 u32 ipv6_addr_label(struct net *net,
180 const struct in6_addr *addr, int type, int ifindex)
183 struct ip6addrlbl_entry *p;
185 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
188 p = __ipv6_addr_label(net, addr, type, ifindex);
189 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
192 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
193 __func__, addr, type, ifindex, label);
198 /* allocate one entry */
199 static struct ip6addrlbl_entry *ip6addrlbl_alloc(struct net *net,
200 const struct in6_addr *prefix,
201 int prefixlen, int ifindex,
204 struct ip6addrlbl_entry *newp;
207 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
208 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
210 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
213 case IPV6_ADDR_MAPPED:
215 return ERR_PTR(-EINVAL);
219 case IPV6_ADDR_COMPATv4:
223 case IPV6_ADDR_LOOPBACK:
224 if (prefixlen != 128)
229 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
231 return ERR_PTR(-ENOMEM);
233 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
234 newp->prefixlen = prefixlen;
235 newp->ifindex = ifindex;
236 newp->addrtype = addrtype;
238 INIT_HLIST_NODE(&newp->list);
239 write_pnet(&newp->lbl_net, net);
240 refcount_set(&newp->refcnt, 1);
245 static int __ip6addrlbl_add(struct ip6addrlbl_entry *newp, int replace)
247 struct hlist_node *n;
248 struct ip6addrlbl_entry *last = NULL, *p = NULL;
251 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", __func__, newp,
254 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) {
255 if (p->prefixlen == newp->prefixlen &&
256 net_eq(ip6addrlbl_net(p), ip6addrlbl_net(newp)) &&
257 p->ifindex == newp->ifindex &&
258 ipv6_addr_equal(&p->prefix, &newp->prefix)) {
263 hlist_replace_rcu(&p->list, &newp->list);
266 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
267 (p->prefixlen < newp->prefixlen)) {
268 hlist_add_before_rcu(&newp->list, &p->list);
274 hlist_add_behind_rcu(&newp->list, &last->list);
276 hlist_add_head_rcu(&newp->list, &ip6addrlbl_table.head);
279 ip6addrlbl_table.seq++;
284 static int ip6addrlbl_add(struct net *net,
285 const struct in6_addr *prefix, int prefixlen,
286 int ifindex, u32 label, int replace)
288 struct ip6addrlbl_entry *newp;
291 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
292 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
295 newp = ip6addrlbl_alloc(net, prefix, prefixlen, ifindex, label);
297 return PTR_ERR(newp);
298 spin_lock(&ip6addrlbl_table.lock);
299 ret = __ip6addrlbl_add(newp, replace);
300 spin_unlock(&ip6addrlbl_table.lock);
302 ip6addrlbl_free(newp);
307 static int __ip6addrlbl_del(struct net *net,
308 const struct in6_addr *prefix, int prefixlen,
311 struct ip6addrlbl_entry *p = NULL;
312 struct hlist_node *n;
315 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
316 __func__, prefix, prefixlen, ifindex);
318 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) {
319 if (p->prefixlen == prefixlen &&
320 net_eq(ip6addrlbl_net(p), net) &&
321 p->ifindex == ifindex &&
322 ipv6_addr_equal(&p->prefix, prefix)) {
323 hlist_del_rcu(&p->list);
332 static int ip6addrlbl_del(struct net *net,
333 const struct in6_addr *prefix, int prefixlen,
336 struct in6_addr prefix_buf;
339 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
340 __func__, prefix, prefixlen, ifindex);
342 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
343 spin_lock(&ip6addrlbl_table.lock);
344 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
345 spin_unlock(&ip6addrlbl_table.lock);
349 /* add default label */
350 static int __net_init ip6addrlbl_net_init(struct net *net)
355 ADDRLABEL(KERN_DEBUG "%s\n", __func__);
357 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
358 int ret = ip6addrlbl_add(net,
359 ip6addrlbl_init_table[i].prefix,
360 ip6addrlbl_init_table[i].prefixlen,
362 ip6addrlbl_init_table[i].label, 0);
363 /* XXX: should we free all rules when we catch an error? */
364 if (ret && (!err || err != -ENOMEM))
370 static void __net_exit ip6addrlbl_net_exit(struct net *net)
372 struct ip6addrlbl_entry *p = NULL;
373 struct hlist_node *n;
375 /* Remove all labels belonging to the exiting net */
376 spin_lock(&ip6addrlbl_table.lock);
377 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) {
378 if (net_eq(ip6addrlbl_net(p), net)) {
379 hlist_del_rcu(&p->list);
383 spin_unlock(&ip6addrlbl_table.lock);
386 static struct pernet_operations ipv6_addr_label_ops = {
387 .init = ip6addrlbl_net_init,
388 .exit = ip6addrlbl_net_exit,
391 int __init ipv6_addr_label_init(void)
393 spin_lock_init(&ip6addrlbl_table.lock);
395 return register_pernet_subsys(&ipv6_addr_label_ops);
398 void ipv6_addr_label_cleanup(void)
400 unregister_pernet_subsys(&ipv6_addr_label_ops);
403 static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
404 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
405 [IFAL_LABEL] = { .len = sizeof(u32), },
408 static bool addrlbl_ifindex_exists(struct net *net, int ifindex)
411 struct net_device *dev;
414 dev = dev_get_by_index_rcu(net, ifindex);
420 static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
421 struct netlink_ext_ack *extack)
423 struct net *net = sock_net(skb->sk);
424 struct ifaddrlblmsg *ifal;
425 struct nlattr *tb[IFAL_MAX+1];
426 struct in6_addr *pfx;
430 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
435 ifal = nlmsg_data(nlh);
437 if (ifal->ifal_family != AF_INET6 ||
438 ifal->ifal_prefixlen > 128)
441 if (!tb[IFAL_ADDRESS])
443 pfx = nla_data(tb[IFAL_ADDRESS]);
447 label = nla_get_u32(tb[IFAL_LABEL]);
448 if (label == IPV6_ADDR_LABEL_DEFAULT)
451 switch (nlh->nlmsg_type) {
452 case RTM_NEWADDRLABEL:
453 if (ifal->ifal_index &&
454 !addrlbl_ifindex_exists(net, ifal->ifal_index))
457 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
458 ifal->ifal_index, label,
459 nlh->nlmsg_flags & NLM_F_REPLACE);
461 case RTM_DELADDRLABEL:
462 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
471 static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
472 int prefixlen, int ifindex, u32 lseq)
474 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
475 ifal->ifal_family = AF_INET6;
476 ifal->ifal_prefixlen = prefixlen;
477 ifal->ifal_flags = 0;
478 ifal->ifal_index = ifindex;
479 ifal->ifal_seq = lseq;
482 static int ip6addrlbl_fill(struct sk_buff *skb,
483 struct ip6addrlbl_entry *p,
485 u32 portid, u32 seq, int event,
488 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, event,
489 sizeof(struct ifaddrlblmsg), flags);
493 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
495 if (nla_put_in6_addr(skb, IFAL_ADDRESS, &p->prefix) < 0 ||
496 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
497 nlmsg_cancel(skb, nlh);
505 static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
507 struct net *net = sock_net(skb->sk);
508 struct ip6addrlbl_entry *p;
509 int idx = 0, s_idx = cb->args[0];
513 hlist_for_each_entry_rcu(p, &ip6addrlbl_table.head, list) {
515 net_eq(ip6addrlbl_net(p), net)) {
516 err = ip6addrlbl_fill(skb, p,
517 ip6addrlbl_table.seq,
518 NETLINK_CB(cb->skb).portid,
532 static inline int ip6addrlbl_msgsize(void)
534 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
535 + nla_total_size(16) /* IFAL_ADDRESS */
536 + nla_total_size(4); /* IFAL_LABEL */
539 static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
540 struct netlink_ext_ack *extack)
542 struct net *net = sock_net(in_skb->sk);
543 struct ifaddrlblmsg *ifal;
544 struct nlattr *tb[IFAL_MAX+1];
545 struct in6_addr *addr;
548 struct ip6addrlbl_entry *p;
551 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
556 ifal = nlmsg_data(nlh);
558 if (ifal->ifal_family != AF_INET6 ||
559 ifal->ifal_prefixlen != 128)
562 if (ifal->ifal_index &&
563 !addrlbl_ifindex_exists(net, ifal->ifal_index))
566 if (!tb[IFAL_ADDRESS])
568 addr = nla_data(tb[IFAL_ADDRESS]);
571 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
572 if (p && !ip6addrlbl_hold(p))
574 lseq = ip6addrlbl_table.seq;
582 skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL);
588 err = ip6addrlbl_fill(skb, p, lseq,
589 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
590 RTM_NEWADDRLABEL, 0);
595 WARN_ON(err == -EMSGSIZE);
600 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
605 void __init ipv6_addr_label_rtnl_register(void)
607 __rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel,
608 NULL, RTNL_FLAG_DOIT_UNLOCKED);
609 __rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel,
610 NULL, RTNL_FLAG_DOIT_UNLOCKED);
611 __rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get,
612 ip6addrlbl_dump, RTNL_FLAG_DOIT_UNLOCKED);