1// SPDX-License-Identifier: GPL-2.0
2/*
3 * IPv6 Address Label subsystem
4 * for the IPv6 "Default" Source Address Selection
5 *
6 * Copyright (C)2007 USAGI/WIDE Project
7 */
8/*
9 * Author:
10 * YOSHIFUJI Hideaki @ USAGI/WIDE Project <yoshfuji@linux-ipv6.org>
11 */
12
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
23/*
24 * Policy Table
25 */
26struct ip6addrlbl_entry {
27 struct in6_addr prefix;
28 int prefixlen;
29 int ifindex;
30 int addrtype;
31 u32 label;
32 struct hlist_node list;
33 struct rcu_head rcu;
34};
35
36/*
37 * Default policy table (RFC6724 + extensions)
38 *
39 * prefix addr_type label
40 * -------------------------------------------------------------------------
41 * ::1/128 LOOPBACK 0
42 * ::/0 N/A 1
43 * 2002::/16 N/A 2
44 * ::/96 COMPATv4 3
45 * ::ffff:0:0/96 V4MAPPED 4
46 * fc00::/7 N/A 5 ULA (RFC 4193)
47 * 2001::/32 N/A 6 Teredo (RFC 4380)
48 * 2001:10::/28 N/A 7 ORCHID (RFC 4843)
49 * fec0::/10 N/A 11 Site-local
50 * (deprecated by RFC3879)
51 * 3ffe::/16 N/A 12 6bone
52 *
53 * Note: 0xffffffff is used if we do not have any policies.
54 * Note: Labels for ULA and 6to4 are different from labels listed in RFC6724.
55 */
56
57#define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
58
59static const __net_initconst struct ip6addrlbl_init_table
60{
61 const struct in6_addr *prefix;
62 int prefixlen;
63 u32 label;
64} ip6addrlbl_init_table[] = {
65 { /* ::/0 */
66 .prefix = &in6addr_any,
67 .label = 1,
68 }, { /* fc00::/7 */
69 .prefix = &(struct in6_addr){ { { 0xfc } } } ,
70 .prefixlen = 7,
71 .label = 5,
72 }, { /* fec0::/10 */
73 .prefix = &(struct in6_addr){ { { 0xfe, 0xc0 } } },
74 .prefixlen = 10,
75 .label = 11,
76 }, { /* 2002::/16 */
77 .prefix = &(struct in6_addr){ { { 0x20, 0x02 } } },
78 .prefixlen = 16,
79 .label = 2,
80 }, { /* 3ffe::/16 */
81 .prefix = &(struct in6_addr){ { { 0x3f, 0xfe } } },
82 .prefixlen = 16,
83 .label = 12,
84 }, { /* 2001::/32 */
85 .prefix = &(struct in6_addr){ { { 0x20, 0x01 } } },
86 .prefixlen = 32,
87 .label = 6,
88 }, { /* 2001:10::/28 */
89 .prefix = &(struct in6_addr){ { { 0x20, 0x01, 0x00, 0x10 } } },
90 .prefixlen = 28,
91 .label = 7,
92 }, { /* ::ffff:0:0 */
93 .prefix = &(struct in6_addr){ { { [10] = 0xff, [11] = 0xff } } },
94 .prefixlen = 96,
95 .label = 4,
96 }, { /* ::/96 */
97 .prefix = &in6addr_any,
98 .prefixlen = 96,
99 .label = 3,
100 }, { /* ::1/128 */
101 .prefix = &in6addr_loopback,
102 .prefixlen = 128,
103 .label = 0,
104 }
105};
106
107/* Find label */
108static bool __ip6addrlbl_match(const struct ip6addrlbl_entry *p,
109 const struct in6_addr *addr,
110 int addrtype, int ifindex)
111{
112 if (p->ifindex && p->ifindex != ifindex)
113 return false;
114 if (p->addrtype && p->addrtype != addrtype)
115 return false;
116 if (!ipv6_prefix_equal(addr1: addr, addr2: &p->prefix, prefixlen: p->prefixlen))
117 return false;
118 return true;
119}
120
121static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
122 const struct in6_addr *addr,
123 int type, int ifindex)
124{
125 struct ip6addrlbl_entry *p;
126
127 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
128 if (__ip6addrlbl_match(p, addr, addrtype: type, ifindex))
129 return p;
130 }
131 return NULL;
132}
133
134u32 ipv6_addr_label(struct net *net,
135 const struct in6_addr *addr, int type, int ifindex)
136{
137 u32 label;
138 struct ip6addrlbl_entry *p;
139
140 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
141
142 rcu_read_lock();
143 p = __ipv6_addr_label(net, addr, type, ifindex);
144 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
145 rcu_read_unlock();
146
147 net_dbg_ratelimited("%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n", __func__, addr, type,
148 ifindex, label);
149
150 return label;
151}
152
153/* allocate one entry */
154static struct ip6addrlbl_entry *ip6addrlbl_alloc(const struct in6_addr *prefix,
155 int prefixlen, int ifindex,
156 u32 label)
157{
158 struct ip6addrlbl_entry *newp;
159 int addrtype;
160
161 net_dbg_ratelimited("%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n", __func__,
162 prefix, prefixlen, ifindex, (unsigned int)label);
163
164 addrtype = ipv6_addr_type(addr: prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
165
166 switch (addrtype) {
167 case IPV6_ADDR_MAPPED:
168 if (prefixlen > 96)
169 return ERR_PTR(error: -EINVAL);
170 if (prefixlen < 96)
171 addrtype = 0;
172 break;
173 case IPV6_ADDR_COMPATv4:
174 if (prefixlen != 96)
175 addrtype = 0;
176 break;
177 case IPV6_ADDR_LOOPBACK:
178 if (prefixlen != 128)
179 addrtype = 0;
180 break;
181 }
182
183 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
184 if (!newp)
185 return ERR_PTR(error: -ENOMEM);
186
187 ipv6_addr_prefix(pfx: &newp->prefix, addr: prefix, plen: prefixlen);
188 newp->prefixlen = prefixlen;
189 newp->ifindex = ifindex;
190 newp->addrtype = addrtype;
191 newp->label = label;
192 INIT_HLIST_NODE(h: &newp->list);
193 return newp;
194}
195
196/* add a label */
197static int __ip6addrlbl_add(struct net *net, struct ip6addrlbl_entry *newp,
198 int replace)
199{
200 struct ip6addrlbl_entry *last = NULL, *p = NULL;
201 struct hlist_node *n;
202 int ret = 0;
203
204 net_dbg_ratelimited("%s(newp=%p, replace=%d)\n", __func__, newp, replace);
205
206 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
207 if (p->prefixlen == newp->prefixlen &&
208 p->ifindex == newp->ifindex &&
209 ipv6_addr_equal(a1: &p->prefix, a2: &newp->prefix)) {
210 if (!replace) {
211 ret = -EEXIST;
212 goto out;
213 }
214 hlist_replace_rcu(old: &p->list, new: &newp->list);
215 kfree_rcu(p, rcu);
216 goto out;
217 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
218 (p->prefixlen < newp->prefixlen)) {
219 hlist_add_before_rcu(n: &newp->list, next: &p->list);
220 goto out;
221 }
222 last = p;
223 }
224 if (last)
225 hlist_add_behind_rcu(n: &newp->list, prev: &last->list);
226 else
227 hlist_add_head_rcu(n: &newp->list, h: &net->ipv6.ip6addrlbl_table.head);
228out:
229 if (!ret)
230 WRITE_ONCE(net->ipv6.ip6addrlbl_table.seq,
231 net->ipv6.ip6addrlbl_table.seq + 1);
232 return ret;
233}
234
235/* add a label */
236static int ip6addrlbl_add(struct net *net,
237 const struct in6_addr *prefix, int prefixlen,
238 int ifindex, u32 label, int replace)
239{
240 struct ip6addrlbl_entry *newp;
241 int ret = 0;
242
243 net_dbg_ratelimited("%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
244 __func__, prefix, prefixlen, ifindex, (unsigned int)label, replace);
245
246 newp = ip6addrlbl_alloc(prefix, prefixlen, ifindex, label);
247 if (IS_ERR(ptr: newp))
248 return PTR_ERR(ptr: newp);
249 spin_lock(lock: &net->ipv6.ip6addrlbl_table.lock);
250 ret = __ip6addrlbl_add(net, newp, replace);
251 spin_unlock(lock: &net->ipv6.ip6addrlbl_table.lock);
252 if (ret)
253 kfree(objp: newp);
254 return ret;
255}
256
257/* remove a label */
258static int __ip6addrlbl_del(struct net *net,
259 const struct in6_addr *prefix, int prefixlen,
260 int ifindex)
261{
262 struct ip6addrlbl_entry *p = NULL;
263 struct hlist_node *n;
264 int ret = -ESRCH;
265
266 net_dbg_ratelimited("%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n", __func__, prefix,
267 prefixlen, ifindex);
268
269 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
270 if (p->prefixlen == prefixlen &&
271 p->ifindex == ifindex &&
272 ipv6_addr_equal(a1: &p->prefix, a2: prefix)) {
273 hlist_del_rcu(n: &p->list);
274 kfree_rcu(p, rcu);
275 ret = 0;
276 break;
277 }
278 }
279 return ret;
280}
281
282static int ip6addrlbl_del(struct net *net,
283 const struct in6_addr *prefix, int prefixlen,
284 int ifindex)
285{
286 struct in6_addr prefix_buf;
287 int ret;
288
289 net_dbg_ratelimited("%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n", __func__, prefix,
290 prefixlen, ifindex);
291
292 ipv6_addr_prefix(pfx: &prefix_buf, addr: prefix, plen: prefixlen);
293 spin_lock(lock: &net->ipv6.ip6addrlbl_table.lock);
294 ret = __ip6addrlbl_del(net, prefix: &prefix_buf, prefixlen, ifindex);
295 spin_unlock(lock: &net->ipv6.ip6addrlbl_table.lock);
296 return ret;
297}
298
299/* add default label */
300static int __net_init ip6addrlbl_net_init(struct net *net)
301{
302 struct ip6addrlbl_entry *p = NULL;
303 struct hlist_node *n;
304 int err;
305 int i;
306
307 spin_lock_init(&net->ipv6.ip6addrlbl_table.lock);
308 INIT_HLIST_HEAD(&net->ipv6.ip6addrlbl_table.head);
309
310 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
311 err = ip6addrlbl_add(net,
312 prefix: ip6addrlbl_init_table[i].prefix,
313 prefixlen: ip6addrlbl_init_table[i].prefixlen,
314 ifindex: 0,
315 label: ip6addrlbl_init_table[i].label, replace: 0);
316 if (err)
317 goto err_ip6addrlbl_add;
318 }
319 return 0;
320
321err_ip6addrlbl_add:
322 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
323 hlist_del_rcu(n: &p->list);
324 kfree_rcu(p, rcu);
325 }
326 return err;
327}
328
329static void __net_exit ip6addrlbl_net_exit(struct net *net)
330{
331 struct ip6addrlbl_entry *p = NULL;
332 struct hlist_node *n;
333
334 /* Remove all labels belonging to the exiting net */
335 spin_lock(lock: &net->ipv6.ip6addrlbl_table.lock);
336 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
337 hlist_del_rcu(n: &p->list);
338 kfree_rcu(p, rcu);
339 }
340 spin_unlock(lock: &net->ipv6.ip6addrlbl_table.lock);
341}
342
343static struct pernet_operations ipv6_addr_label_ops = {
344 .init = ip6addrlbl_net_init,
345 .exit = ip6addrlbl_net_exit,
346};
347
348int __init ipv6_addr_label_init(void)
349{
350 return register_pernet_subsys(&ipv6_addr_label_ops);
351}
352
353void ipv6_addr_label_cleanup(void)
354{
355 unregister_pernet_subsys(&ipv6_addr_label_ops);
356}
357
358static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
359 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
360 [IFAL_LABEL] = { .len = sizeof(u32), },
361};
362
363static bool addrlbl_ifindex_exists(struct net *net, int ifindex)
364{
365
366 struct net_device *dev;
367
368 rcu_read_lock();
369 dev = dev_get_by_index_rcu(net, ifindex);
370 rcu_read_unlock();
371
372 return dev != NULL;
373}
374
375static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
376 struct netlink_ext_ack *extack)
377{
378 struct net *net = sock_net(sk: skb->sk);
379 struct ifaddrlblmsg *ifal;
380 struct nlattr *tb[IFAL_MAX+1];
381 struct in6_addr *pfx;
382 u32 label;
383 int err = 0;
384
385 err = nlmsg_parse_deprecated(nlh, hdrlen: sizeof(*ifal), tb, IFAL_MAX,
386 policy: ifal_policy, extack);
387 if (err < 0)
388 return err;
389
390 ifal = nlmsg_data(nlh);
391
392 if (ifal->ifal_family != AF_INET6 ||
393 ifal->ifal_prefixlen > 128)
394 return -EINVAL;
395
396 if (!tb[IFAL_ADDRESS])
397 return -EINVAL;
398 pfx = nla_data(nla: tb[IFAL_ADDRESS]);
399
400 if (!tb[IFAL_LABEL])
401 return -EINVAL;
402 label = nla_get_u32(nla: tb[IFAL_LABEL]);
403 if (label == IPV6_ADDR_LABEL_DEFAULT)
404 return -EINVAL;
405
406 switch (nlh->nlmsg_type) {
407 case RTM_NEWADDRLABEL:
408 if (ifal->ifal_index &&
409 !addrlbl_ifindex_exists(net, ifindex: ifal->ifal_index))
410 return -EINVAL;
411
412 err = ip6addrlbl_add(net, prefix: pfx, prefixlen: ifal->ifal_prefixlen,
413 ifindex: ifal->ifal_index, label,
414 replace: nlh->nlmsg_flags & NLM_F_REPLACE);
415 break;
416 case RTM_DELADDRLABEL:
417 err = ip6addrlbl_del(net, prefix: pfx, prefixlen: ifal->ifal_prefixlen,
418 ifindex: ifal->ifal_index);
419 break;
420 default:
421 err = -EOPNOTSUPP;
422 }
423 return err;
424}
425
426static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
427 int prefixlen, int ifindex, u32 lseq)
428{
429 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
430 ifal->ifal_family = AF_INET6;
431 ifal->__ifal_reserved = 0;
432 ifal->ifal_prefixlen = prefixlen;
433 ifal->ifal_flags = 0;
434 ifal->ifal_index = ifindex;
435 ifal->ifal_seq = lseq;
436};
437
438static int ip6addrlbl_fill(struct sk_buff *skb,
439 const struct ip6addrlbl_entry *p,
440 u32 lseq,
441 u32 portid, u32 seq, int event,
442 unsigned int flags)
443{
444 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, type: event,
445 payload: sizeof(struct ifaddrlblmsg), flags);
446 if (!nlh)
447 return -EMSGSIZE;
448
449 ip6addrlbl_putmsg(nlh, prefixlen: p->prefixlen, ifindex: p->ifindex, lseq);
450
451 if (nla_put_in6_addr(skb, attrtype: IFAL_ADDRESS, addr: &p->prefix) < 0 ||
452 nla_put_u32(skb, attrtype: IFAL_LABEL, value: p->label) < 0) {
453 nlmsg_cancel(skb, nlh);
454 return -EMSGSIZE;
455 }
456
457 nlmsg_end(skb, nlh);
458 return 0;
459}
460
461static int ip6addrlbl_valid_dump_req(const struct nlmsghdr *nlh,
462 struct netlink_ext_ack *extack)
463{
464 struct ifaddrlblmsg *ifal;
465
466 ifal = nlmsg_payload(nlh, len: sizeof(*ifal));
467 if (!ifal) {
468 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address label dump request");
469 return -EINVAL;
470 }
471
472 if (ifal->__ifal_reserved || ifal->ifal_prefixlen ||
473 ifal->ifal_flags || ifal->ifal_index || ifal->ifal_seq) {
474 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address label dump request");
475 return -EINVAL;
476 }
477
478 if (nlmsg_attrlen(nlh, hdrlen: sizeof(*ifal))) {
479 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header for address label dump request");
480 return -EINVAL;
481 }
482
483 return 0;
484}
485
486static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
487{
488 const struct nlmsghdr *nlh = cb->nlh;
489 struct net *net = sock_net(sk: skb->sk);
490 struct ip6addrlbl_entry *p;
491 int idx = 0, s_idx = cb->args[0];
492 int err = 0;
493 u32 lseq;
494
495 if (cb->strict_check) {
496 err = ip6addrlbl_valid_dump_req(nlh, extack: cb->extack);
497 if (err < 0)
498 return err;
499 }
500
501 rcu_read_lock();
502 lseq = READ_ONCE(net->ipv6.ip6addrlbl_table.seq);
503 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
504 if (idx >= s_idx) {
505 err = ip6addrlbl_fill(skb, p,
506 lseq,
507 NETLINK_CB(cb->skb).portid,
508 seq: nlh->nlmsg_seq,
509 RTM_NEWADDRLABEL,
510 NLM_F_MULTI);
511 if (err < 0)
512 break;
513 }
514 idx++;
515 }
516 rcu_read_unlock();
517 cb->args[0] = idx;
518 return err;
519}
520
521static inline int ip6addrlbl_msgsize(void)
522{
523 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
524 + nla_total_size(payload: 16) /* IFAL_ADDRESS */
525 + nla_total_size(payload: 4); /* IFAL_LABEL */
526}
527
528static int ip6addrlbl_valid_get_req(struct sk_buff *skb,
529 const struct nlmsghdr *nlh,
530 struct nlattr **tb,
531 struct netlink_ext_ack *extack)
532{
533 struct ifaddrlblmsg *ifal;
534 int i, err;
535
536 ifal = nlmsg_payload(nlh, len: sizeof(*ifal));
537 if (!ifal) {
538 NL_SET_ERR_MSG_MOD(extack, "Invalid header for addrlabel get request");
539 return -EINVAL;
540 }
541
542 if (!netlink_strict_get_check(skb))
543 return nlmsg_parse_deprecated(nlh, hdrlen: sizeof(*ifal), tb,
544 IFAL_MAX, policy: ifal_policy, extack);
545
546 if (ifal->__ifal_reserved || ifal->ifal_flags || ifal->ifal_seq) {
547 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for addrlabel get request");
548 return -EINVAL;
549 }
550
551 err = nlmsg_parse_deprecated_strict(nlh, hdrlen: sizeof(*ifal), tb, IFAL_MAX,
552 policy: ifal_policy, extack);
553 if (err)
554 return err;
555
556 for (i = 0; i <= IFAL_MAX; i++) {
557 if (!tb[i])
558 continue;
559
560 switch (i) {
561 case IFAL_ADDRESS:
562 break;
563 default:
564 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in addrlabel get request");
565 return -EINVAL;
566 }
567 }
568
569 return 0;
570}
571
572static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
573 struct netlink_ext_ack *extack)
574{
575 struct net *net = sock_net(sk: in_skb->sk);
576 struct ifaddrlblmsg *ifal;
577 struct nlattr *tb[IFAL_MAX+1];
578 struct in6_addr *addr;
579 u32 lseq;
580 int err = 0;
581 struct ip6addrlbl_entry *p;
582 struct sk_buff *skb;
583
584 err = ip6addrlbl_valid_get_req(skb: in_skb, nlh, tb, extack);
585 if (err < 0)
586 return err;
587
588 ifal = nlmsg_data(nlh);
589
590 if (ifal->ifal_family != AF_INET6 ||
591 ifal->ifal_prefixlen != 128)
592 return -EINVAL;
593
594 if (ifal->ifal_index &&
595 !addrlbl_ifindex_exists(net, ifindex: ifal->ifal_index))
596 return -EINVAL;
597
598 if (!tb[IFAL_ADDRESS])
599 return -EINVAL;
600 addr = nla_data(nla: tb[IFAL_ADDRESS]);
601
602 skb = nlmsg_new(payload: ip6addrlbl_msgsize(), GFP_KERNEL);
603 if (!skb)
604 return -ENOBUFS;
605
606 err = -ESRCH;
607
608 rcu_read_lock();
609 p = __ipv6_addr_label(net, addr, type: ipv6_addr_type(addr), ifindex: ifal->ifal_index);
610 lseq = READ_ONCE(net->ipv6.ip6addrlbl_table.seq);
611 if (p)
612 err = ip6addrlbl_fill(skb, p, lseq,
613 NETLINK_CB(in_skb).portid,
614 seq: nlh->nlmsg_seq,
615 RTM_NEWADDRLABEL, flags: 0);
616 rcu_read_unlock();
617
618 if (err < 0) {
619 WARN_ON(err == -EMSGSIZE);
620 kfree_skb(skb);
621 } else {
622 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
623 }
624 return err;
625}
626
627static const struct rtnl_msg_handler ipv6_adddr_label_rtnl_msg_handlers[] __initconst_or_module = {
628 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDRLABEL,
629 .doit = ip6addrlbl_newdel, .flags = RTNL_FLAG_DOIT_UNLOCKED},
630 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDRLABEL,
631 .doit = ip6addrlbl_newdel, .flags = RTNL_FLAG_DOIT_UNLOCKED},
632 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDRLABEL,
633 .doit = ip6addrlbl_get, .dumpit = ip6addrlbl_dump,
634 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
635};
636
637int __init ipv6_addr_label_rtnl_register(void)
638{
639 return rtnl_register_many(ipv6_adddr_label_rtnl_msg_handlers);
640}
641