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
36 struct in6_addr prefix;
41 struct hlist_node list;
46 static struct ip6addrlbl_table
48 struct hlist_head head;
54 struct net *ip6addrlbl_net(const struct ip6addrlbl_entry *lbl)
64 * Default policy table (RFC3484 + extensions)
66 * prefix addr_type label
67 * -------------------------------------------------------------------------
72 * ::ffff:0:0/96 V4MAPPED 4
73 * fc00::/7 N/A 5 ULA (RFC 4193)
74 * 2001::/32 N/A 6 Teredo (RFC 4380)
75 * 2001:10::/28 N/A 7 ORCHID (RFC 4843)
77 * Note: 0xffffffff is used if we do not have any policies.
80 #define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
82 static const __net_initdata struct ip6addrlbl_init_table
84 const struct in6_addr *prefix;
87 } ip6addrlbl_init_table[] = {
89 .prefix = &in6addr_any,
92 .prefix = &(struct in6_addr){{{ 0xfc }}},
96 .prefix = &(struct in6_addr){{{ 0x20, 0x02 }}},
100 .prefix = &(struct in6_addr){{{ 0x20, 0x01 }}},
103 },{ /* 2001:10::/28 */
104 .prefix = &(struct in6_addr){{{ 0x20, 0x01, 0x00, 0x10 }}},
108 .prefix = &(struct in6_addr){{{ [10] = 0xff, [11] = 0xff }}},
112 .prefix = &in6addr_any,
116 .prefix = &in6addr_loopback,
122 /* Object management */
123 static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p)
126 release_net(p->lbl_net);
131 static void ip6addrlbl_free_rcu(struct rcu_head *h)
133 ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu));
136 static inline int ip6addrlbl_hold(struct ip6addrlbl_entry *p)
138 return atomic_inc_not_zero(&p->refcnt);
141 static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p)
143 if (atomic_dec_and_test(&p->refcnt))
144 call_rcu(&p->rcu, ip6addrlbl_free_rcu);
148 static int __ip6addrlbl_match(struct net *net,
149 struct ip6addrlbl_entry *p,
150 const struct in6_addr *addr,
151 int addrtype, int ifindex)
153 if (!net_eq(ip6addrlbl_net(p), net))
155 if (p->ifindex && p->ifindex != ifindex)
157 if (p->addrtype && p->addrtype != addrtype)
159 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
164 static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
165 const struct in6_addr *addr,
166 int type, int ifindex)
168 struct hlist_node *pos;
169 struct ip6addrlbl_entry *p;
170 hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) {
171 if (__ip6addrlbl_match(net, p, addr, type, ifindex))
177 u32 ipv6_addr_label(struct net *net,
178 const struct in6_addr *addr, int type, int ifindex)
181 struct ip6addrlbl_entry *p;
183 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
186 p = __ipv6_addr_label(net, addr, type, ifindex);
187 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
190 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
191 __func__, addr, type, ifindex, label);
196 /* allocate one entry */
197 static struct ip6addrlbl_entry *ip6addrlbl_alloc(struct net *net,
198 const struct in6_addr *prefix,
199 int prefixlen, int ifindex,
202 struct ip6addrlbl_entry *newp;
205 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
206 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
208 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
211 case IPV6_ADDR_MAPPED:
213 return ERR_PTR(-EINVAL);
217 case IPV6_ADDR_COMPATv4:
221 case IPV6_ADDR_LOOPBACK:
222 if (prefixlen != 128)
227 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
229 return ERR_PTR(-ENOMEM);
231 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
232 newp->prefixlen = prefixlen;
233 newp->ifindex = ifindex;
234 newp->addrtype = addrtype;
236 INIT_HLIST_NODE(&newp->list);
238 newp->lbl_net = hold_net(net);
240 atomic_set(&newp->refcnt, 1);
245 static int __ip6addrlbl_add(struct ip6addrlbl_entry *newp, int replace)
249 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n",
253 if (hlist_empty(&ip6addrlbl_table.head)) {
254 hlist_add_head_rcu(&newp->list, &ip6addrlbl_table.head);
256 struct hlist_node *pos, *n;
257 struct ip6addrlbl_entry *p = NULL;
258 hlist_for_each_entry_safe(p, pos, n,
259 &ip6addrlbl_table.head, list) {
260 if (p->prefixlen == newp->prefixlen &&
261 net_eq(ip6addrlbl_net(p), ip6addrlbl_net(newp)) &&
262 p->ifindex == newp->ifindex &&
263 ipv6_addr_equal(&p->prefix, &newp->prefix)) {
268 hlist_replace_rcu(&p->list, &newp->list);
271 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
272 (p->prefixlen < newp->prefixlen)) {
273 hlist_add_before_rcu(&newp->list, &p->list);
277 hlist_add_after_rcu(&p->list, &newp->list);
281 ip6addrlbl_table.seq++;
286 static int ip6addrlbl_add(struct net *net,
287 const struct in6_addr *prefix, int prefixlen,
288 int ifindex, u32 label, int replace)
290 struct ip6addrlbl_entry *newp;
293 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
294 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
297 newp = ip6addrlbl_alloc(net, prefix, prefixlen, ifindex, label);
299 return PTR_ERR(newp);
300 spin_lock(&ip6addrlbl_table.lock);
301 ret = __ip6addrlbl_add(newp, replace);
302 spin_unlock(&ip6addrlbl_table.lock);
304 ip6addrlbl_free(newp);
309 static int __ip6addrlbl_del(struct net *net,
310 const struct in6_addr *prefix, int prefixlen,
313 struct ip6addrlbl_entry *p = NULL;
314 struct hlist_node *pos, *n;
317 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
318 __func__, prefix, prefixlen, ifindex);
320 hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) {
321 if (p->prefixlen == prefixlen &&
322 net_eq(ip6addrlbl_net(p), net) &&
323 p->ifindex == ifindex &&
324 ipv6_addr_equal(&p->prefix, prefix)) {
325 hlist_del_rcu(&p->list);
334 static int ip6addrlbl_del(struct net *net,
335 const struct in6_addr *prefix, int prefixlen,
338 struct in6_addr prefix_buf;
341 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
342 __func__, prefix, prefixlen, ifindex);
344 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
345 spin_lock(&ip6addrlbl_table.lock);
346 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
347 spin_unlock(&ip6addrlbl_table.lock);
351 /* add default label */
352 static int __net_init ip6addrlbl_net_init(struct net *net)
357 ADDRLABEL(KERN_DEBUG "%s()\n", __func__);
359 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
360 int ret = ip6addrlbl_add(net,
361 ip6addrlbl_init_table[i].prefix,
362 ip6addrlbl_init_table[i].prefixlen,
364 ip6addrlbl_init_table[i].label, 0);
365 /* XXX: should we free all rules when we catch an error? */
366 if (ret && (!err || err != -ENOMEM))
372 static void __net_exit ip6addrlbl_net_exit(struct net *net)
374 struct ip6addrlbl_entry *p = NULL;
375 struct hlist_node *pos, *n;
377 /* Remove all labels belonging to the exiting net */
378 spin_lock(&ip6addrlbl_table.lock);
379 hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) {
380 if (net_eq(ip6addrlbl_net(p), net)) {
381 hlist_del_rcu(&p->list);
385 spin_unlock(&ip6addrlbl_table.lock);
388 static struct pernet_operations ipv6_addr_label_ops = {
389 .init = ip6addrlbl_net_init,
390 .exit = ip6addrlbl_net_exit,
393 int __init ipv6_addr_label_init(void)
395 spin_lock_init(&ip6addrlbl_table.lock);
397 return register_pernet_subsys(&ipv6_addr_label_ops);
400 static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
401 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
402 [IFAL_LABEL] = { .len = sizeof(u32), },
405 static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
408 struct net *net = sock_net(skb->sk);
409 struct ifaddrlblmsg *ifal;
410 struct nlattr *tb[IFAL_MAX+1];
411 struct in6_addr *pfx;
415 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy);
419 ifal = nlmsg_data(nlh);
421 if (ifal->ifal_family != AF_INET6 ||
422 ifal->ifal_prefixlen > 128)
425 if (!tb[IFAL_ADDRESS])
428 pfx = nla_data(tb[IFAL_ADDRESS]);
434 label = nla_get_u32(tb[IFAL_LABEL]);
435 if (label == IPV6_ADDR_LABEL_DEFAULT)
438 switch(nlh->nlmsg_type) {
439 case RTM_NEWADDRLABEL:
440 if (ifal->ifal_index &&
441 !__dev_get_by_index(net, ifal->ifal_index))
444 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
445 ifal->ifal_index, label,
446 nlh->nlmsg_flags & NLM_F_REPLACE);
448 case RTM_DELADDRLABEL:
449 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
458 static inline void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
459 int prefixlen, int ifindex, u32 lseq)
461 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
462 ifal->ifal_family = AF_INET6;
463 ifal->ifal_prefixlen = prefixlen;
464 ifal->ifal_flags = 0;
465 ifal->ifal_index = ifindex;
466 ifal->ifal_seq = lseq;
469 static int ip6addrlbl_fill(struct sk_buff *skb,
470 struct ip6addrlbl_entry *p,
472 u32 pid, u32 seq, int event,
475 struct nlmsghdr *nlh = nlmsg_put(skb, pid, seq, event,
476 sizeof(struct ifaddrlblmsg), flags);
480 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
482 if (nla_put(skb, IFAL_ADDRESS, 16, &p->prefix) < 0 ||
483 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
484 nlmsg_cancel(skb, nlh);
488 return nlmsg_end(skb, nlh);
491 static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
493 struct net *net = sock_net(skb->sk);
494 struct ip6addrlbl_entry *p;
495 struct hlist_node *pos;
496 int idx = 0, s_idx = cb->args[0];
500 hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) {
502 net_eq(ip6addrlbl_net(p), net)) {
503 if ((err = ip6addrlbl_fill(skb, p,
504 ip6addrlbl_table.seq,
505 NETLINK_CB(cb->skb).pid,
518 static inline int ip6addrlbl_msgsize(void)
520 return (NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
521 + nla_total_size(16) /* IFAL_ADDRESS */
522 + nla_total_size(4) /* IFAL_LABEL */
526 static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr* nlh,
529 struct net *net = sock_net(in_skb->sk);
530 struct ifaddrlblmsg *ifal;
531 struct nlattr *tb[IFAL_MAX+1];
532 struct in6_addr *addr;
535 struct ip6addrlbl_entry *p;
538 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy);
542 ifal = nlmsg_data(nlh);
544 if (ifal->ifal_family != AF_INET6 ||
545 ifal->ifal_prefixlen != 128)
548 if (ifal->ifal_index &&
549 !__dev_get_by_index(net, ifal->ifal_index))
552 if (!tb[IFAL_ADDRESS])
555 addr = nla_data(tb[IFAL_ADDRESS]);
560 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
561 if (p && ip6addrlbl_hold(p))
563 lseq = ip6addrlbl_table.seq;
571 if (!(skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL))) {
576 err = ip6addrlbl_fill(skb, p, lseq,
577 NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
578 RTM_NEWADDRLABEL, 0);
583 WARN_ON(err == -EMSGSIZE);
588 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
593 void __init ipv6_addr_label_rtnl_register(void)
595 __rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel, NULL);
596 __rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel, NULL);
597 __rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get, ip6addrlbl_dump);