Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[cris-mirror.git] / net / ipv6 / addrlabel.c
blob1d6ced37ad718398c947cf49b3b486d4b88f3f6f
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * IPv6 Address Label subsystem
4 * for the IPv6 "Default" Source Address Selection
6 * Copyright (C)2007 USAGI/WIDE Project
7 */
8 /*
9 * Author:
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>
23 #if 0
24 #define ADDRLABEL(x...) printk(x)
25 #else
26 #define ADDRLABEL(x...) do { ; } while (0)
27 #endif
30 * Policy Table
32 struct 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;
43 * Default policy table (RFC6724 + extensions)
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
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.
63 #define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL
65 static const __net_initconst struct ip6addrlbl_init_table
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,
113 /* Find label */
114 static bool __ip6addrlbl_match(const struct ip6addrlbl_entry *p,
115 const struct in6_addr *addr,
116 int addrtype, int ifindex)
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;
127 static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net,
128 const struct in6_addr *addr,
129 int type, int ifindex)
131 struct ip6addrlbl_entry *p;
133 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
134 if (__ip6addrlbl_match(p, addr, type, ifindex))
135 return p;
137 return NULL;
140 u32 ipv6_addr_label(struct net *net,
141 const struct in6_addr *addr, int type, int ifindex)
143 u32 label;
144 struct ip6addrlbl_entry *p;
146 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK;
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();
153 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n",
154 __func__, addr, type, ifindex, label);
156 return label;
159 /* allocate one entry */
160 static struct ip6addrlbl_entry *ip6addrlbl_alloc(const struct in6_addr *prefix,
161 int prefixlen, int ifindex,
162 u32 label)
164 struct ip6addrlbl_entry *newp;
165 int addrtype;
167 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n",
168 __func__, prefix, prefixlen, ifindex, (unsigned int)label);
170 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK);
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;
189 newp = kmalloc(sizeof(*newp), GFP_KERNEL);
190 if (!newp)
191 return ERR_PTR(-ENOMEM);
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;
202 /* add a label */
203 static int __ip6addrlbl_add(struct net *net, struct ip6addrlbl_entry *newp,
204 int replace)
206 struct ip6addrlbl_entry *last = NULL, *p = NULL;
207 struct hlist_node *n;
208 int ret = 0;
210 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", __func__, newp,
211 replace);
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;
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;
229 last = p;
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);
235 out:
236 if (!ret)
237 net->ipv6.ip6addrlbl_table.seq++;
238 return ret;
241 /* add a label */
242 static int ip6addrlbl_add(struct net *net,
243 const struct in6_addr *prefix, int prefixlen,
244 int ifindex, u32 label, int replace)
246 struct ip6addrlbl_entry *newp;
247 int ret = 0;
249 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n",
250 __func__, prefix, prefixlen, ifindex, (unsigned int)label,
251 replace);
253 newp = ip6addrlbl_alloc(prefix, prefixlen, ifindex, label);
254 if (IS_ERR(newp))
255 return PTR_ERR(newp);
256 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
257 ret = __ip6addrlbl_add(net, newp, replace);
258 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
259 if (ret)
260 kfree(newp);
261 return ret;
264 /* remove a label */
265 static int __ip6addrlbl_del(struct net *net,
266 const struct in6_addr *prefix, int prefixlen,
267 int ifindex)
269 struct ip6addrlbl_entry *p = NULL;
270 struct hlist_node *n;
271 int ret = -ESRCH;
273 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
274 __func__, prefix, prefixlen, ifindex);
276 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
277 if (p->prefixlen == prefixlen &&
278 p->ifindex == ifindex &&
279 ipv6_addr_equal(&p->prefix, prefix)) {
280 hlist_del_rcu(&p->list);
281 kfree_rcu(p, rcu);
282 ret = 0;
283 break;
286 return ret;
289 static int ip6addrlbl_del(struct net *net,
290 const struct in6_addr *prefix, int prefixlen,
291 int ifindex)
293 struct in6_addr prefix_buf;
294 int ret;
296 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n",
297 __func__, prefix, prefixlen, ifindex);
299 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen);
300 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
301 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex);
302 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
303 return ret;
306 /* add default label */
307 static int __net_init ip6addrlbl_net_init(struct net *net)
309 int err = 0;
310 int i;
312 ADDRLABEL(KERN_DEBUG "%s\n", __func__);
314 spin_lock_init(&net->ipv6.ip6addrlbl_table.lock);
315 INIT_HLIST_HEAD(&net->ipv6.ip6addrlbl_table.head);
317 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) {
318 int ret = ip6addrlbl_add(net,
319 ip6addrlbl_init_table[i].prefix,
320 ip6addrlbl_init_table[i].prefixlen,
322 ip6addrlbl_init_table[i].label, 0);
323 /* XXX: should we free all rules when we catch an error? */
324 if (ret && (!err || err != -ENOMEM))
325 err = ret;
327 return err;
330 static void __net_exit ip6addrlbl_net_exit(struct net *net)
332 struct ip6addrlbl_entry *p = NULL;
333 struct hlist_node *n;
335 /* Remove all labels belonging to the exiting net */
336 spin_lock(&net->ipv6.ip6addrlbl_table.lock);
337 hlist_for_each_entry_safe(p, n, &net->ipv6.ip6addrlbl_table.head, list) {
338 hlist_del_rcu(&p->list);
339 kfree_rcu(p, rcu);
341 spin_unlock(&net->ipv6.ip6addrlbl_table.lock);
344 static struct pernet_operations ipv6_addr_label_ops = {
345 .init = ip6addrlbl_net_init,
346 .exit = ip6addrlbl_net_exit,
349 int __init ipv6_addr_label_init(void)
351 return register_pernet_subsys(&ipv6_addr_label_ops);
354 void ipv6_addr_label_cleanup(void)
356 unregister_pernet_subsys(&ipv6_addr_label_ops);
359 static const struct nla_policy ifal_policy[IFAL_MAX+1] = {
360 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), },
361 [IFAL_LABEL] = { .len = sizeof(u32), },
364 static bool addrlbl_ifindex_exists(struct net *net, int ifindex)
367 struct net_device *dev;
369 rcu_read_lock();
370 dev = dev_get_by_index_rcu(net, ifindex);
371 rcu_read_unlock();
373 return dev != NULL;
376 static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh,
377 struct netlink_ext_ack *extack)
379 struct net *net = sock_net(skb->sk);
380 struct ifaddrlblmsg *ifal;
381 struct nlattr *tb[IFAL_MAX+1];
382 struct in6_addr *pfx;
383 u32 label;
384 int err = 0;
386 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
387 extack);
388 if (err < 0)
389 return err;
391 ifal = nlmsg_data(nlh);
393 if (ifal->ifal_family != AF_INET6 ||
394 ifal->ifal_prefixlen > 128)
395 return -EINVAL;
397 if (!tb[IFAL_ADDRESS])
398 return -EINVAL;
399 pfx = nla_data(tb[IFAL_ADDRESS]);
401 if (!tb[IFAL_LABEL])
402 return -EINVAL;
403 label = nla_get_u32(tb[IFAL_LABEL]);
404 if (label == IPV6_ADDR_LABEL_DEFAULT)
405 return -EINVAL;
407 switch (nlh->nlmsg_type) {
408 case RTM_NEWADDRLABEL:
409 if (ifal->ifal_index &&
410 !addrlbl_ifindex_exists(net, ifal->ifal_index))
411 return -EINVAL;
413 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen,
414 ifal->ifal_index, label,
415 nlh->nlmsg_flags & NLM_F_REPLACE);
416 break;
417 case RTM_DELADDRLABEL:
418 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen,
419 ifal->ifal_index);
420 break;
421 default:
422 err = -EOPNOTSUPP;
424 return err;
427 static void ip6addrlbl_putmsg(struct nlmsghdr *nlh,
428 int prefixlen, int ifindex, u32 lseq)
430 struct ifaddrlblmsg *ifal = nlmsg_data(nlh);
431 ifal->ifal_family = AF_INET6;
432 ifal->ifal_prefixlen = prefixlen;
433 ifal->ifal_flags = 0;
434 ifal->ifal_index = ifindex;
435 ifal->ifal_seq = lseq;
438 static int ip6addrlbl_fill(struct sk_buff *skb,
439 struct ip6addrlbl_entry *p,
440 u32 lseq,
441 u32 portid, u32 seq, int event,
442 unsigned int flags)
444 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, event,
445 sizeof(struct ifaddrlblmsg), flags);
446 if (!nlh)
447 return -EMSGSIZE;
449 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq);
451 if (nla_put_in6_addr(skb, IFAL_ADDRESS, &p->prefix) < 0 ||
452 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) {
453 nlmsg_cancel(skb, nlh);
454 return -EMSGSIZE;
457 nlmsg_end(skb, nlh);
458 return 0;
461 static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb)
463 struct net *net = sock_net(skb->sk);
464 struct ip6addrlbl_entry *p;
465 int idx = 0, s_idx = cb->args[0];
466 int err;
468 rcu_read_lock();
469 hlist_for_each_entry_rcu(p, &net->ipv6.ip6addrlbl_table.head, list) {
470 if (idx >= s_idx) {
471 err = ip6addrlbl_fill(skb, p,
472 net->ipv6.ip6addrlbl_table.seq,
473 NETLINK_CB(cb->skb).portid,
474 cb->nlh->nlmsg_seq,
475 RTM_NEWADDRLABEL,
476 NLM_F_MULTI);
477 if (err < 0)
478 break;
480 idx++;
482 rcu_read_unlock();
483 cb->args[0] = idx;
484 return skb->len;
487 static inline int ip6addrlbl_msgsize(void)
489 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg))
490 + nla_total_size(16) /* IFAL_ADDRESS */
491 + nla_total_size(4); /* IFAL_LABEL */
494 static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
495 struct netlink_ext_ack *extack)
497 struct net *net = sock_net(in_skb->sk);
498 struct ifaddrlblmsg *ifal;
499 struct nlattr *tb[IFAL_MAX+1];
500 struct in6_addr *addr;
501 u32 lseq;
502 int err = 0;
503 struct ip6addrlbl_entry *p;
504 struct sk_buff *skb;
506 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy,
507 extack);
508 if (err < 0)
509 return err;
511 ifal = nlmsg_data(nlh);
513 if (ifal->ifal_family != AF_INET6 ||
514 ifal->ifal_prefixlen != 128)
515 return -EINVAL;
517 if (ifal->ifal_index &&
518 !addrlbl_ifindex_exists(net, ifal->ifal_index))
519 return -EINVAL;
521 if (!tb[IFAL_ADDRESS])
522 return -EINVAL;
523 addr = nla_data(tb[IFAL_ADDRESS]);
525 skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL);
526 if (!skb)
527 return -ENOBUFS;
529 err = -ESRCH;
531 rcu_read_lock();
532 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index);
533 lseq = net->ipv6.ip6addrlbl_table.seq;
534 if (p)
535 err = ip6addrlbl_fill(skb, p, lseq,
536 NETLINK_CB(in_skb).portid,
537 nlh->nlmsg_seq,
538 RTM_NEWADDRLABEL, 0);
539 rcu_read_unlock();
541 if (err < 0) {
542 WARN_ON(err == -EMSGSIZE);
543 kfree_skb(skb);
544 } else {
545 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
547 return err;
550 int __init ipv6_addr_label_rtnl_register(void)
552 int ret;
554 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDRLABEL,
555 ip6addrlbl_newdel,
556 NULL, RTNL_FLAG_DOIT_UNLOCKED);
557 if (ret < 0)
558 return ret;
559 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDRLABEL,
560 ip6addrlbl_newdel,
561 NULL, RTNL_FLAG_DOIT_UNLOCKED);
562 if (ret < 0)
563 return ret;
564 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDRLABEL,
565 ip6addrlbl_get,
566 ip6addrlbl_dump, RTNL_FLAG_DOIT_UNLOCKED);
567 return ret;