Loading...
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#if 0
24#define ADDRLABEL(x...) printk(x)
25#else
26#define ADDRLABEL(x...) do { ; } while (0)
27#endif
28
29/*
30 * Policy Table
31 */
32struct ip6addrlbl_entry {
33 struct in6_addr prefix;
34 int prefixlen;
35 int ifindex;
36 int addrtype;
37 u32 label;
38 struct hlist_node list;
39 struct rcu_head rcu;
40};
41
42/*
43 * Default policy table (RFC6724 + extensions)
44 *
45 * prefix addr_type label
46 * -------------------------------------------------------------------------
47 * ::1/128 LOOPBACK 0
48 * ::/0 N/A 1
49 * 2002::/16 N/A 2
50 * ::/96 COMPATv4 3
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
58 *
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.
61 */
62
63#define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
64
65static const __net_initconst struct ip6addrlbl_init_table
66{
67 const struct in6_addr *prefix;
68 int prefixlen;
69 u32 label;
70} ip6addrlbl_init_table[] = {
71 { /* ::/0 */
72 .prefix = &in6addr_any,
73 .label = 1,
74 }, { /* fc00::/7 */
75 .prefix = &(struct in6_addr){ { { 0xfc } } } ,
76 .prefixlen = 7,
77 .label = 5,
78 }, { /* fec0::/10 */
79 .prefix = &(struct in6_addr){ { { 0xfe, 0xc0 } } },
80 .prefixlen = 10,
81 .label = 11,
82 }, { /* 2002::/16 */
83 .prefix = &(struct in6_addr){ { { 0x20, 0x02 } } },
84 .prefixlen = 16,
85 .label = 2,
86 }, { /* 3ffe::/16 */
87 .prefix = &(struct in6_addr){ { { 0x3f, 0xfe } } },
88 .prefixlen = 16,
89 .label = 12,
90 }, { /* 2001::/32 */
91 .prefix = &(struct in6_addr){ { { 0x20, 0x01 } } },
92 .prefixlen = 32,
93 .label = 6,
94 }, { /* 2001:10::/28 */
95 .prefix = &(struct in6_addr){ { { 0x20, 0x01, 0x00, 0x10 } } },
96 .prefixlen = 28,
97 .label = 7,
98 }, { /* ::ffff:0:0 */
99 .prefix = &(struct in6_addr){ { { [10] = 0xff, [11] = 0xff } } },
100 .prefixlen = 96,
101 .label = 4,
102 }, { /* ::/96 */
103 .prefix = &in6addr_any,
104 .prefixlen = 96,
105 .label = 3,
106 }, { /* ::1/128 */
107 .prefix = &in6addr_loopback,
108 .prefixlen = 128,
109 .label = 0,
110 }
111};
112
113/* Find label */
114static bool __ip6addrlbl_match(const struct ip6addrlbl_entry *p,
115 const struct in6_addr *addr,
116 int addrtype, int ifindex)
117{
118 if (p->ifindex && p->ifindex != ifindex)
119 return false;
120 if (p->addrtype && p->addrtype != addrtype)
121 return false;
122 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
123 return false;
124 return true;
125}
126
127static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
128 const struct in6_addr *addr,
129 int type, int ifindex)
130{
131 struct ip6addrlbl_entry *p;
132
133 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
134 if (__ip6addrlbl_match(p, addr, type, ifindex))
135 return p;
136 }
137 return NULL;
138}
139
140u32 ipv6_addr_label(struct net *net,
141 const struct in6_addr *addr, int type, int ifindex)
142{
143 u32 label;
144 struct ip6addrlbl_entry *p;
145
146 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
147
148 rcu_read_lock();
149 p = __ipv6_addr_label(net, addr, type, ifindex);
150 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
151 rcu_read_unlock();
152
153 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
154 __func__, addr, type, ifindex, label);
155
156 return label;
157}
158
159/* allocate one entry */
160static struct ip6addrlbl_entry *ip6addrlbl_alloc(const struct in6_addr *prefix,
161 int prefixlen, int ifindex,
162 u32 label)
163{
164 struct ip6addrlbl_entry *newp;
165 int addrtype;
166
167 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
168 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
169
170 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
171
172 switch (addrtype) {
173 case IPV6_ADDR_MAPPED:
174 if (prefixlen > 96)
175 return ERR_PTR(-EINVAL);
176 if (prefixlen < 96)
177 addrtype = 0;
178 break;
179 case IPV6_ADDR_COMPATv4:
180 if (prefixlen != 96)
181 addrtype = 0;
182 break;
183 case IPV6_ADDR_LOOPBACK:
184 if (prefixlen != 128)
185 addrtype = 0;
186 break;
187 }
188
189 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
190 if (!newp)
191 return ERR_PTR(-ENOMEM);
192
193 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
194 newp->prefixlen = prefixlen;
195 newp->ifindex = ifindex;
196 newp->addrtype = addrtype;
197 newp->label = label;
198 INIT_HLIST_NODE(&newp->list);
199 return newp;
200}
201
202/* add a label */
203static int __ip6addrlbl_add(struct net *net, struct ip6addrlbl_entry *newp,
204 int replace)
205{
206 struct ip6addrlbl_entry *last = NULL, *p = NULL;
207 struct hlist_node *n;
208 int ret = 0;
209
210 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", __func__, newp,
211 replace);
212
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)) {
217 if (!replace) {
218 ret = -EEXIST;
219 goto out;
220 }
221 hlist_replace_rcu(&p->list, &newp->list);
222 kfree_rcu(p, rcu);
223 goto out;
224 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
225 (p->prefixlen < newp->prefixlen)) {
226 hlist_add_before_rcu(&newp->list, &p->list);
227 goto out;
228 }
229 last = p;
230 }
231 if (last)
232 hlist_add_behind_rcu(&newp->list, &last->list);
233 else
234 hlist_add_head_rcu(&newp->list, &net->ipv6.ip6addrlbl_table.head);
235out:
236 if (!ret)
237 WRITE_ONCE(net->ipv6.ip6addrlbl_table.seq,
238 net->ipv6.ip6addrlbl_table.seq + 1);
239 return ret;
240}
241
242/* add a label */
243static int ip6addrlbl_add(struct net *net,
244 const struct in6_addr *prefix, int prefixlen,
245 int ifindex, u32 label, int replace)
246{
247 struct ip6addrlbl_entry *newp;
248 int ret = 0;
249
250 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
251 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
252 replace);
253
254 newp = ip6addrlbl_alloc(prefix, prefixlen, ifindex, label);
255 if (IS_ERR(newp))
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);
260 if (ret)
261 kfree(newp);
262 return ret;
263}
264
265/* remove a label */
266static int __ip6addrlbl_del(struct net *net,
267 const struct in6_addr *prefix, int prefixlen,
268 int ifindex)
269{
270 struct ip6addrlbl_entry *p = NULL;
271 struct hlist_node *n;
272 int ret = -ESRCH;
273
274 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
275 __func__, prefix, prefixlen, ifindex);
276
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);
282 kfree_rcu(p, rcu);
283 ret = 0;
284 break;
285 }
286 }
287 return ret;
288}
289
290static int ip6addrlbl_del(struct net *net,
291 const struct in6_addr *prefix, int prefixlen,
292 int ifindex)
293{
294 struct in6_addr prefix_buf;
295 int ret;
296
297 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
298 __func__, prefix, prefixlen, ifindex);
299
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);
304 return ret;
305}
306
307/* add default label */
308static int __net_init ip6addrlbl_net_init(struct net *net)
309{
310 struct ip6addrlbl_entry *p = NULL;
311 struct hlist_node *n;
312 int err;
313 int i;
314
315 ADDRLABEL(KERN_DEBUG "%s\n", __func__);
316
317 spin_lock_init(&net->ipv6.ip6addrlbl_table.lock);
318 INIT_HLIST_HEAD(&net->ipv6.ip6addrlbl_table.head);
319
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,
324 0,
325 ip6addrlbl_init_table[i].label, 0);
326 if (err)
327 goto err_ip6addrlbl_add;
328 }
329 return 0;
330
331err_ip6addrlbl_add:
332 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
333 hlist_del_rcu(&p->list);
334 kfree_rcu(p, rcu);
335 }
336 return err;
337}
338
339static void __net_exit ip6addrlbl_net_exit(struct net *net)
340{
341 struct ip6addrlbl_entry *p = NULL;
342 struct hlist_node *n;
343
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);
348 kfree_rcu(p, rcu);
349 }
350 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
351}
352
353static struct pernet_operations ipv6_addr_label_ops = {
354 .init = ip6addrlbl_net_init,
355 .exit = ip6addrlbl_net_exit,
356};
357
358int __init ipv6_addr_label_init(void)
359{
360 return register_pernet_subsys(&ipv6_addr_label_ops);
361}
362
363void ipv6_addr_label_cleanup(void)
364{
365 unregister_pernet_subsys(&ipv6_addr_label_ops);
366}
367
368static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
369 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
370 [IFAL_LABEL] = { .len = sizeof(u32), },
371};
372
373static bool addrlbl_ifindex_exists(struct net *net, int ifindex)
374{
375
376 struct net_device *dev;
377
378 rcu_read_lock();
379 dev = dev_get_by_index_rcu(net, ifindex);
380 rcu_read_unlock();
381
382 return dev != NULL;
383}
384
385static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
386 struct netlink_ext_ack *extack)
387{
388 struct net *net = sock_net(skb->sk);
389 struct ifaddrlblmsg *ifal;
390 struct nlattr *tb[IFAL_MAX+1];
391 struct in6_addr *pfx;
392 u32 label;
393 int err = 0;
394
395 err = nlmsg_parse_deprecated(nlh, sizeof(*ifal), tb, IFAL_MAX,
396 ifal_policy, extack);
397 if (err < 0)
398 return err;
399
400 ifal = nlmsg_data(nlh);
401
402 if (ifal->ifal_family != AF_INET6 ||
403 ifal->ifal_prefixlen > 128)
404 return -EINVAL;
405
406 if (!tb[IFAL_ADDRESS])
407 return -EINVAL;
408 pfx = nla_data(tb[IFAL_ADDRESS]);
409
410 if (!tb[IFAL_LABEL])
411 return -EINVAL;
412 label = nla_get_u32(tb[IFAL_LABEL]);
413 if (label == IPV6_ADDR_LABEL_DEFAULT)
414 return -EINVAL;
415
416 switch (nlh->nlmsg_type) {
417 case RTM_NEWADDRLABEL:
418 if (ifal->ifal_index &&
419 !addrlbl_ifindex_exists(net, ifal->ifal_index))
420 return -EINVAL;
421
422 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
423 ifal->ifal_index, label,
424 nlh->nlmsg_flags & NLM_F_REPLACE);
425 break;
426 case RTM_DELADDRLABEL:
427 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
428 ifal->ifal_index);
429 break;
430 default:
431 err = -EOPNOTSUPP;
432 }
433 return err;
434}
435
436static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
437 int prefixlen, int ifindex, u32 lseq)
438{
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;
446};
447
448static int ip6addrlbl_fill(struct sk_buff *skb,
449 const struct ip6addrlbl_entry *p,
450 u32 lseq,
451 u32 portid, u32 seq, int event,
452 unsigned int flags)
453{
454 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, event,
455 sizeof(struct ifaddrlblmsg), flags);
456 if (!nlh)
457 return -EMSGSIZE;
458
459 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
460
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);
464 return -EMSGSIZE;
465 }
466
467 nlmsg_end(skb, nlh);
468 return 0;
469}
470
471static int ip6addrlbl_valid_dump_req(const struct nlmsghdr *nlh,
472 struct netlink_ext_ack *extack)
473{
474 struct ifaddrlblmsg *ifal;
475
476 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifal))) {
477 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address label dump request");
478 return -EINVAL;
479 }
480
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");
485 return -EINVAL;
486 }
487
488 if (nlmsg_attrlen(nlh, sizeof(*ifal))) {
489 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header for address label dump request");
490 return -EINVAL;
491 }
492
493 return 0;
494}
495
496static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
497{
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];
502 int err = 0;
503 u32 lseq;
504
505 if (cb->strict_check) {
506 err = ip6addrlbl_valid_dump_req(nlh, cb->extack);
507 if (err < 0)
508 return err;
509 }
510
511 rcu_read_lock();
512 lseq = READ_ONCE(net->ipv6.ip6addrlbl_table.seq);
513 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
514 if (idx >= s_idx) {
515 err = ip6addrlbl_fill(skb, p,
516 lseq,
517 NETLINK_CB(cb->skb).portid,
518 nlh->nlmsg_seq,
519 RTM_NEWADDRLABEL,
520 NLM_F_MULTI);
521 if (err < 0)
522 break;
523 }
524 idx++;
525 }
526 rcu_read_unlock();
527 cb->args[0] = idx;
528 return err;
529}
530
531static inline int ip6addrlbl_msgsize(void)
532{
533 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
534 + nla_total_size(16) /* IFAL_ADDRESS */
535 + nla_total_size(4); /* IFAL_LABEL */
536}
537
538static int ip6addrlbl_valid_get_req(struct sk_buff *skb,
539 const struct nlmsghdr *nlh,
540 struct nlattr **tb,
541 struct netlink_ext_ack *extack)
542{
543 struct ifaddrlblmsg *ifal;
544 int i, err;
545
546 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifal))) {
547 NL_SET_ERR_MSG_MOD(extack, "Invalid header for addrlabel get request");
548 return -EINVAL;
549 }
550
551 if (!netlink_strict_get_check(skb))
552 return nlmsg_parse_deprecated(nlh, sizeof(*ifal), tb,
553 IFAL_MAX, ifal_policy, extack);
554
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");
558 return -EINVAL;
559 }
560
561 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifal), tb, IFAL_MAX,
562 ifal_policy, extack);
563 if (err)
564 return err;
565
566 for (i = 0; i <= IFAL_MAX; i++) {
567 if (!tb[i])
568 continue;
569
570 switch (i) {
571 case IFAL_ADDRESS:
572 break;
573 default:
574 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in addrlabel get request");
575 return -EINVAL;
576 }
577 }
578
579 return 0;
580}
581
582static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
583 struct netlink_ext_ack *extack)
584{
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;
589 u32 lseq;
590 int err = 0;
591 struct ip6addrlbl_entry *p;
592 struct sk_buff *skb;
593
594 err = ip6addrlbl_valid_get_req(in_skb, nlh, tb, extack);
595 if (err < 0)
596 return err;
597
598 ifal = nlmsg_data(nlh);
599
600 if (ifal->ifal_family != AF_INET6 ||
601 ifal->ifal_prefixlen != 128)
602 return -EINVAL;
603
604 if (ifal->ifal_index &&
605 !addrlbl_ifindex_exists(net, ifal->ifal_index))
606 return -EINVAL;
607
608 if (!tb[IFAL_ADDRESS])
609 return -EINVAL;
610 addr = nla_data(tb[IFAL_ADDRESS]);
611
612 skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL);
613 if (!skb)
614 return -ENOBUFS;
615
616 err = -ESRCH;
617
618 rcu_read_lock();
619 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
620 lseq = READ_ONCE(net->ipv6.ip6addrlbl_table.seq);
621 if (p)
622 err = ip6addrlbl_fill(skb, p, lseq,
623 NETLINK_CB(in_skb).portid,
624 nlh->nlmsg_seq,
625 RTM_NEWADDRLABEL, 0);
626 rcu_read_unlock();
627
628 if (err < 0) {
629 WARN_ON(err == -EMSGSIZE);
630 kfree_skb(skb);
631 } else {
632 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
633 }
634 return err;
635}
636
637static const struct rtnl_msg_handler ipv6_adddr_label_rtnl_msg_handlers[] __initconst_or_module = {
638 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDRLABEL,
639 .doit = ip6addrlbl_newdel, .flags = RTNL_FLAG_DOIT_UNLOCKED},
640 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDRLABEL,
641 .doit = ip6addrlbl_newdel, .flags = RTNL_FLAG_DOIT_UNLOCKED},
642 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDRLABEL,
643 .doit = ip6addrlbl_get, .dumpit = ip6addrlbl_dump,
644 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
645};
646
647int __init ipv6_addr_label_rtnl_register(void)
648{
649 return rtnl_register_many(ipv6_adddr_label_rtnl_msg_handlers);
650}
1/*
2 * IPv6 Address Label subsystem
3 * for the IPv6 "Default" Source Address Selection
4 *
5 * Copyright (C)2007 USAGI/WIDE Project
6 */
7/*
8 * Author:
9 * YOSHIFUJI Hideaki @ USAGI/WIDE Project <yoshfuji@linux-ipv6.org>
10 */
11
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
22#if 0
23#define ADDRLABEL(x...) printk(x)
24#else
25#define ADDRLABEL(x...) do { ; } while(0)
26#endif
27
28/*
29 * Policy Table
30 */
31struct ip6addrlbl_entry
32{
33#ifdef CONFIG_NET_NS
34 struct net *lbl_net;
35#endif
36 struct in6_addr prefix;
37 int prefixlen;
38 int ifindex;
39 int addrtype;
40 u32 label;
41 struct hlist_node list;
42 atomic_t refcnt;
43 struct rcu_head rcu;
44};
45
46static struct ip6addrlbl_table
47{
48 struct hlist_head head;
49 spinlock_t lock;
50 u32 seq;
51} ip6addrlbl_table;
52
53static inline
54struct net *ip6addrlbl_net(const struct ip6addrlbl_entry *lbl)
55{
56 return read_pnet(&lbl->lbl_net);
57}
58
59/*
60 * Default policy table (RFC3484 + extensions)
61 *
62 * prefix addr_type label
63 * -------------------------------------------------------------------------
64 * ::1/128 LOOPBACK 0
65 * ::/0 N/A 1
66 * 2002::/16 N/A 2
67 * ::/96 COMPATv4 3
68 * ::ffff:0:0/96 V4MAPPED 4
69 * fc00::/7 N/A 5 ULA (RFC 4193)
70 * 2001::/32 N/A 6 Teredo (RFC 4380)
71 * 2001:10::/28 N/A 7 ORCHID (RFC 4843)
72 *
73 * Note: 0xffffffff is used if we do not have any policies.
74 */
75
76#define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
77
78static const __net_initdata struct ip6addrlbl_init_table
79{
80 const struct in6_addr *prefix;
81 int prefixlen;
82 u32 label;
83} ip6addrlbl_init_table[] = {
84 { /* ::/0 */
85 .prefix = &in6addr_any,
86 .label = 1,
87 },{ /* fc00::/7 */
88 .prefix = &(struct in6_addr){{{ 0xfc }}},
89 .prefixlen = 7,
90 .label = 5,
91 },{ /* 2002::/16 */
92 .prefix = &(struct in6_addr){{{ 0x20, 0x02 }}},
93 .prefixlen = 16,
94 .label = 2,
95 },{ /* 2001::/32 */
96 .prefix = &(struct in6_addr){{{ 0x20, 0x01 }}},
97 .prefixlen = 32,
98 .label = 6,
99 },{ /* 2001:10::/28 */
100 .prefix = &(struct in6_addr){{{ 0x20, 0x01, 0x00, 0x10 }}},
101 .prefixlen = 28,
102 .label = 7,
103 },{ /* ::ffff:0:0 */
104 .prefix = &(struct in6_addr){{{ [10] = 0xff, [11] = 0xff }}},
105 .prefixlen = 96,
106 .label = 4,
107 },{ /* ::/96 */
108 .prefix = &in6addr_any,
109 .prefixlen = 96,
110 .label = 3,
111 },{ /* ::1/128 */
112 .prefix = &in6addr_loopback,
113 .prefixlen = 128,
114 .label = 0,
115 }
116};
117
118/* Object management */
119static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p)
120{
121#ifdef CONFIG_NET_NS
122 release_net(p->lbl_net);
123#endif
124 kfree(p);
125}
126
127static void ip6addrlbl_free_rcu(struct rcu_head *h)
128{
129 ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu));
130}
131
132static bool ip6addrlbl_hold(struct ip6addrlbl_entry *p)
133{
134 return atomic_inc_not_zero(&p->refcnt);
135}
136
137static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p)
138{
139 if (atomic_dec_and_test(&p->refcnt))
140 call_rcu(&p->rcu, ip6addrlbl_free_rcu);
141}
142
143/* Find label */
144static bool __ip6addrlbl_match(struct net *net,
145 const struct ip6addrlbl_entry *p,
146 const struct in6_addr *addr,
147 int addrtype, int ifindex)
148{
149 if (!net_eq(ip6addrlbl_net(p), net))
150 return false;
151 if (p->ifindex && p->ifindex != ifindex)
152 return false;
153 if (p->addrtype && p->addrtype != addrtype)
154 return false;
155 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen))
156 return false;
157 return true;
158}
159
160static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
161 const struct in6_addr *addr,
162 int type, int ifindex)
163{
164 struct hlist_node *pos;
165 struct ip6addrlbl_entry *p;
166 hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) {
167 if (__ip6addrlbl_match(net, p, addr, type, ifindex))
168 return p;
169 }
170 return NULL;
171}
172
173u32 ipv6_addr_label(struct net *net,
174 const struct in6_addr *addr, int type, int ifindex)
175{
176 u32 label;
177 struct ip6addrlbl_entry *p;
178
179 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
180
181 rcu_read_lock();
182 p = __ipv6_addr_label(net, addr, type, ifindex);
183 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT;
184 rcu_read_unlock();
185
186 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
187 __func__, addr, type, ifindex, label);
188
189 return label;
190}
191
192/* allocate one entry */
193static struct ip6addrlbl_entry *ip6addrlbl_alloc(struct net *net,
194 const struct in6_addr *prefix,
195 int prefixlen, int ifindex,
196 u32 label)
197{
198 struct ip6addrlbl_entry *newp;
199 int addrtype;
200
201 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
202 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
203
204 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
205
206 switch (addrtype) {
207 case IPV6_ADDR_MAPPED:
208 if (prefixlen > 96)
209 return ERR_PTR(-EINVAL);
210 if (prefixlen < 96)
211 addrtype = 0;
212 break;
213 case IPV6_ADDR_COMPATv4:
214 if (prefixlen != 96)
215 addrtype = 0;
216 break;
217 case IPV6_ADDR_LOOPBACK:
218 if (prefixlen != 128)
219 addrtype = 0;
220 break;
221 }
222
223 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
224 if (!newp)
225 return ERR_PTR(-ENOMEM);
226
227 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen);
228 newp->prefixlen = prefixlen;
229 newp->ifindex = ifindex;
230 newp->addrtype = addrtype;
231 newp->label = label;
232 INIT_HLIST_NODE(&newp->list);
233#ifdef CONFIG_NET_NS
234 newp->lbl_net = hold_net(net);
235#endif
236 atomic_set(&newp->refcnt, 1);
237 return newp;
238}
239
240/* add a label */
241static int __ip6addrlbl_add(struct ip6addrlbl_entry *newp, int replace)
242{
243 int ret = 0;
244
245 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n",
246 __func__,
247 newp, replace);
248
249 if (hlist_empty(&ip6addrlbl_table.head)) {
250 hlist_add_head_rcu(&newp->list, &ip6addrlbl_table.head);
251 } else {
252 struct hlist_node *pos, *n;
253 struct ip6addrlbl_entry *p = NULL;
254 hlist_for_each_entry_safe(p, pos, n,
255 &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)) {
260 if (!replace) {
261 ret = -EEXIST;
262 goto out;
263 }
264 hlist_replace_rcu(&p->list, &newp->list);
265 ip6addrlbl_put(p);
266 goto out;
267 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) ||
268 (p->prefixlen < newp->prefixlen)) {
269 hlist_add_before_rcu(&newp->list, &p->list);
270 goto out;
271 }
272 }
273 hlist_add_after_rcu(&p->list, &newp->list);
274 }
275out:
276 if (!ret)
277 ip6addrlbl_table.seq++;
278 return ret;
279}
280
281/* add a label */
282static int ip6addrlbl_add(struct net *net,
283 const struct in6_addr *prefix, int prefixlen,
284 int ifindex, u32 label, int replace)
285{
286 struct ip6addrlbl_entry *newp;
287 int ret = 0;
288
289 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
290 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
291 replace);
292
293 newp = ip6addrlbl_alloc(net, prefix, prefixlen, ifindex, label);
294 if (IS_ERR(newp))
295 return PTR_ERR(newp);
296 spin_lock(&ip6addrlbl_table.lock);
297 ret = __ip6addrlbl_add(newp, replace);
298 spin_unlock(&ip6addrlbl_table.lock);
299 if (ret)
300 ip6addrlbl_free(newp);
301 return ret;
302}
303
304/* remove a label */
305static int __ip6addrlbl_del(struct net *net,
306 const struct in6_addr *prefix, int prefixlen,
307 int ifindex)
308{
309 struct ip6addrlbl_entry *p = NULL;
310 struct hlist_node *pos, *n;
311 int ret = -ESRCH;
312
313 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
314 __func__, prefix, prefixlen, ifindex);
315
316 hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) {
317 if (p->prefixlen == prefixlen &&
318 net_eq(ip6addrlbl_net(p), net) &&
319 p->ifindex == ifindex &&
320 ipv6_addr_equal(&p->prefix, prefix)) {
321 hlist_del_rcu(&p->list);
322 ip6addrlbl_put(p);
323 ret = 0;
324 break;
325 }
326 }
327 return ret;
328}
329
330static int ip6addrlbl_del(struct net *net,
331 const struct in6_addr *prefix, int prefixlen,
332 int ifindex)
333{
334 struct in6_addr prefix_buf;
335 int ret;
336
337 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
338 __func__, prefix, prefixlen, ifindex);
339
340 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
341 spin_lock(&ip6addrlbl_table.lock);
342 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
343 spin_unlock(&ip6addrlbl_table.lock);
344 return ret;
345}
346
347/* add default label */
348static int __net_init ip6addrlbl_net_init(struct net *net)
349{
350 int err = 0;
351 int i;
352
353 ADDRLABEL(KERN_DEBUG "%s\n", __func__);
354
355 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
356 int ret = ip6addrlbl_add(net,
357 ip6addrlbl_init_table[i].prefix,
358 ip6addrlbl_init_table[i].prefixlen,
359 0,
360 ip6addrlbl_init_table[i].label, 0);
361 /* XXX: should we free all rules when we catch an error? */
362 if (ret && (!err || err != -ENOMEM))
363 err = ret;
364 }
365 return err;
366}
367
368static void __net_exit ip6addrlbl_net_exit(struct net *net)
369{
370 struct ip6addrlbl_entry *p = NULL;
371 struct hlist_node *pos, *n;
372
373 /* Remove all labels belonging to the exiting net */
374 spin_lock(&ip6addrlbl_table.lock);
375 hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) {
376 if (net_eq(ip6addrlbl_net(p), net)) {
377 hlist_del_rcu(&p->list);
378 ip6addrlbl_put(p);
379 }
380 }
381 spin_unlock(&ip6addrlbl_table.lock);
382}
383
384static struct pernet_operations ipv6_addr_label_ops = {
385 .init = ip6addrlbl_net_init,
386 .exit = ip6addrlbl_net_exit,
387};
388
389int __init ipv6_addr_label_init(void)
390{
391 spin_lock_init(&ip6addrlbl_table.lock);
392
393 return register_pernet_subsys(&ipv6_addr_label_ops);
394}
395
396void ipv6_addr_label_cleanup(void)
397{
398 unregister_pernet_subsys(&ipv6_addr_label_ops);
399}
400
401static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
402 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
403 [IFAL_LABEL] = { .len = sizeof(u32), },
404};
405
406static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
407 void *arg)
408{
409 struct net *net = sock_net(skb->sk);
410 struct ifaddrlblmsg *ifal;
411 struct nlattr *tb[IFAL_MAX+1];
412 struct in6_addr *pfx;
413 u32 label;
414 int err = 0;
415
416 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy);
417 if (err < 0)
418 return err;
419
420 ifal = nlmsg_data(nlh);
421
422 if (ifal->ifal_family != AF_INET6 ||
423 ifal->ifal_prefixlen > 128)
424 return -EINVAL;
425
426 if (!tb[IFAL_ADDRESS])
427 return -EINVAL;
428
429 pfx = nla_data(tb[IFAL_ADDRESS]);
430 if (!pfx)
431 return -EINVAL;
432
433 if (!tb[IFAL_LABEL])
434 return -EINVAL;
435 label = nla_get_u32(tb[IFAL_LABEL]);
436 if (label == IPV6_ADDR_LABEL_DEFAULT)
437 return -EINVAL;
438
439 switch(nlh->nlmsg_type) {
440 case RTM_NEWADDRLABEL:
441 if (ifal->ifal_index &&
442 !__dev_get_by_index(net, ifal->ifal_index))
443 return -EINVAL;
444
445 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
446 ifal->ifal_index, label,
447 nlh->nlmsg_flags & NLM_F_REPLACE);
448 break;
449 case RTM_DELADDRLABEL:
450 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
451 ifal->ifal_index);
452 break;
453 default:
454 err = -EOPNOTSUPP;
455 }
456 return err;
457}
458
459static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
460 int prefixlen, int ifindex, u32 lseq)
461{
462 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
463 ifal->ifal_family = AF_INET6;
464 ifal->ifal_prefixlen = prefixlen;
465 ifal->ifal_flags = 0;
466 ifal->ifal_index = ifindex;
467 ifal->ifal_seq = lseq;
468};
469
470static int ip6addrlbl_fill(struct sk_buff *skb,
471 struct ip6addrlbl_entry *p,
472 u32 lseq,
473 u32 pid, u32 seq, int event,
474 unsigned int flags)
475{
476 struct nlmsghdr *nlh = nlmsg_put(skb, pid, seq, event,
477 sizeof(struct ifaddrlblmsg), flags);
478 if (!nlh)
479 return -EMSGSIZE;
480
481 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
482
483 if (nla_put(skb, IFAL_ADDRESS, 16, &p->prefix) < 0 ||
484 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
485 nlmsg_cancel(skb, nlh);
486 return -EMSGSIZE;
487 }
488
489 return nlmsg_end(skb, nlh);
490}
491
492static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
493{
494 struct net *net = sock_net(skb->sk);
495 struct ip6addrlbl_entry *p;
496 struct hlist_node *pos;
497 int idx = 0, s_idx = cb->args[0];
498 int err;
499
500 rcu_read_lock();
501 hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) {
502 if (idx >= s_idx &&
503 net_eq(ip6addrlbl_net(p), net)) {
504 if ((err = ip6addrlbl_fill(skb, p,
505 ip6addrlbl_table.seq,
506 NETLINK_CB(cb->skb).pid,
507 cb->nlh->nlmsg_seq,
508 RTM_NEWADDRLABEL,
509 NLM_F_MULTI)) <= 0)
510 break;
511 }
512 idx++;
513 }
514 rcu_read_unlock();
515 cb->args[0] = idx;
516 return skb->len;
517}
518
519static inline int ip6addrlbl_msgsize(void)
520{
521 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
522 + nla_total_size(16) /* IFAL_ADDRESS */
523 + nla_total_size(4); /* IFAL_LABEL */
524}
525
526static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr* nlh,
527 void *arg)
528{
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;
533 u32 lseq;
534 int err = 0;
535 struct ip6addrlbl_entry *p;
536 struct sk_buff *skb;
537
538 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy);
539 if (err < 0)
540 return err;
541
542 ifal = nlmsg_data(nlh);
543
544 if (ifal->ifal_family != AF_INET6 ||
545 ifal->ifal_prefixlen != 128)
546 return -EINVAL;
547
548 if (ifal->ifal_index &&
549 !__dev_get_by_index(net, ifal->ifal_index))
550 return -EINVAL;
551
552 if (!tb[IFAL_ADDRESS])
553 return -EINVAL;
554
555 addr = nla_data(tb[IFAL_ADDRESS]);
556 if (!addr)
557 return -EINVAL;
558
559 rcu_read_lock();
560 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
561 if (p && ip6addrlbl_hold(p))
562 p = NULL;
563 lseq = ip6addrlbl_table.seq;
564 rcu_read_unlock();
565
566 if (!p) {
567 err = -ESRCH;
568 goto out;
569 }
570
571 if (!(skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL))) {
572 ip6addrlbl_put(p);
573 return -ENOBUFS;
574 }
575
576 err = ip6addrlbl_fill(skb, p, lseq,
577 NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
578 RTM_NEWADDRLABEL, 0);
579
580 ip6addrlbl_put(p);
581
582 if (err < 0) {
583 WARN_ON(err == -EMSGSIZE);
584 kfree_skb(skb);
585 goto out;
586 }
587
588 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
589out:
590 return err;
591}
592
593void __init ipv6_addr_label_rtnl_register(void)
594{
595 __rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel,
596 NULL, NULL);
597 __rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel,
598 NULL, NULL);
599 __rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get,
600 ip6addrlbl_dump, NULL);
601}
602