anycast.c 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557
  1. /*
  2. * Anycast support for IPv6
  3. * Linux INET6 implementation
  4. *
  5. * Authors:
  6. * David L Stevens (dlstevens@us.ibm.com)
  7. *
  8. * based heavily on net/ipv6/mcast.c
  9. *
  10. * This program is free software; you can redistribute it and/or
  11. * modify it under the terms of the GNU General Public License
  12. * as published by the Free Software Foundation; either version
  13. * 2 of the License, or (at your option) any later version.
  14. */
  15. #include <linux/capability.h>
  16. #include <linux/module.h>
  17. #include <linux/errno.h>
  18. #include <linux/types.h>
  19. #include <linux/random.h>
  20. #include <linux/string.h>
  21. #include <linux/socket.h>
  22. #include <linux/sockios.h>
  23. #include <linux/net.h>
  24. #include <linux/in6.h>
  25. #include <linux/netdevice.h>
  26. #include <linux/if_arp.h>
  27. #include <linux/route.h>
  28. #include <linux/init.h>
  29. #include <linux/proc_fs.h>
  30. #include <linux/seq_file.h>
  31. #include <linux/slab.h>
  32. #include <net/net_namespace.h>
  33. #include <net/sock.h>
  34. #include <net/snmp.h>
  35. #include <net/ipv6.h>
  36. #include <net/protocol.h>
  37. #include <net/if_inet6.h>
  38. #include <net/ndisc.h>
  39. #include <net/addrconf.h>
  40. #include <net/ip6_route.h>
  41. #include <net/checksum.h>
  42. static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr);
  43. /*
  44. * socket join an anycast group
  45. */
  46. int ipv6_sock_ac_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
  47. {
  48. struct ipv6_pinfo *np = inet6_sk(sk);
  49. struct net_device *dev = NULL;
  50. struct inet6_dev *idev;
  51. struct ipv6_ac_socklist *pac;
  52. struct net *net = sock_net(sk);
  53. int ishost = !net->ipv6.devconf_all->forwarding;
  54. int err = 0;
  55. ASSERT_RTNL();
  56. if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
  57. return -EPERM;
  58. if (ipv6_addr_is_multicast(addr))
  59. return -EINVAL;
  60. if (ipv6_chk_addr(net, addr, NULL, 0))
  61. return -EINVAL;
  62. pac = sock_kmalloc(sk, sizeof(struct ipv6_ac_socklist), GFP_KERNEL);
  63. if (!pac)
  64. return -ENOMEM;
  65. pac->acl_next = NULL;
  66. pac->acl_addr = *addr;
  67. if (ifindex == 0) {
  68. struct rt6_info *rt;
  69. rt = rt6_lookup(net, addr, NULL, 0, 0);
  70. if (rt) {
  71. dev = rt->dst.dev;
  72. ip6_rt_put(rt);
  73. } else if (ishost) {
  74. err = -EADDRNOTAVAIL;
  75. goto error;
  76. } else {
  77. /* router, no matching interface: just pick one */
  78. dev = __dev_get_by_flags(net, IFF_UP,
  79. IFF_UP | IFF_LOOPBACK);
  80. }
  81. } else
  82. dev = __dev_get_by_index(net, ifindex);
  83. if (!dev) {
  84. err = -ENODEV;
  85. goto error;
  86. }
  87. idev = __in6_dev_get(dev);
  88. if (!idev) {
  89. if (ifindex)
  90. err = -ENODEV;
  91. else
  92. err = -EADDRNOTAVAIL;
  93. goto error;
  94. }
  95. /* reset ishost, now that we have a specific device */
  96. ishost = !idev->cnf.forwarding;
  97. pac->acl_ifindex = dev->ifindex;
  98. /* XXX
  99. * For hosts, allow link-local or matching prefix anycasts.
  100. * This obviates the need for propagating anycast routes while
  101. * still allowing some non-router anycast participation.
  102. */
  103. if (!ipv6_chk_prefix(addr, dev)) {
  104. if (ishost)
  105. err = -EADDRNOTAVAIL;
  106. if (err)
  107. goto error;
  108. }
  109. err = __ipv6_dev_ac_inc(idev, addr);
  110. if (!err) {
  111. pac->acl_next = np->ipv6_ac_list;
  112. np->ipv6_ac_list = pac;
  113. pac = NULL;
  114. }
  115. error:
  116. if (pac)
  117. sock_kfree_s(sk, pac, sizeof(*pac));
  118. return err;
  119. }
  120. /*
  121. * socket leave an anycast group
  122. */
  123. int ipv6_sock_ac_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
  124. {
  125. struct ipv6_pinfo *np = inet6_sk(sk);
  126. struct net_device *dev;
  127. struct ipv6_ac_socklist *pac, *prev_pac;
  128. struct net *net = sock_net(sk);
  129. ASSERT_RTNL();
  130. prev_pac = NULL;
  131. for (pac = np->ipv6_ac_list; pac; pac = pac->acl_next) {
  132. if ((ifindex == 0 || pac->acl_ifindex == ifindex) &&
  133. ipv6_addr_equal(&pac->acl_addr, addr))
  134. break;
  135. prev_pac = pac;
  136. }
  137. if (!pac)
  138. return -ENOENT;
  139. if (prev_pac)
  140. prev_pac->acl_next = pac->acl_next;
  141. else
  142. np->ipv6_ac_list = pac->acl_next;
  143. dev = __dev_get_by_index(net, pac->acl_ifindex);
  144. if (dev)
  145. ipv6_dev_ac_dec(dev, &pac->acl_addr);
  146. sock_kfree_s(sk, pac, sizeof(*pac));
  147. return 0;
  148. }
  149. void ipv6_sock_ac_close(struct sock *sk)
  150. {
  151. struct ipv6_pinfo *np = inet6_sk(sk);
  152. struct net_device *dev = NULL;
  153. struct ipv6_ac_socklist *pac;
  154. struct net *net = sock_net(sk);
  155. int prev_index;
  156. if (!np->ipv6_ac_list)
  157. return;
  158. rtnl_lock();
  159. pac = np->ipv6_ac_list;
  160. np->ipv6_ac_list = NULL;
  161. prev_index = 0;
  162. while (pac) {
  163. struct ipv6_ac_socklist *next = pac->acl_next;
  164. if (pac->acl_ifindex != prev_index) {
  165. dev = __dev_get_by_index(net, pac->acl_ifindex);
  166. prev_index = pac->acl_ifindex;
  167. }
  168. if (dev)
  169. ipv6_dev_ac_dec(dev, &pac->acl_addr);
  170. sock_kfree_s(sk, pac, sizeof(*pac));
  171. pac = next;
  172. }
  173. rtnl_unlock();
  174. }
  175. static void aca_get(struct ifacaddr6 *aca)
  176. {
  177. atomic_inc(&aca->aca_refcnt);
  178. }
  179. static void aca_put(struct ifacaddr6 *ac)
  180. {
  181. if (atomic_dec_and_test(&ac->aca_refcnt)) {
  182. in6_dev_put(ac->aca_idev);
  183. dst_release(&ac->aca_rt->dst);
  184. kfree(ac);
  185. }
  186. }
  187. static struct ifacaddr6 *aca_alloc(struct rt6_info *rt,
  188. const struct in6_addr *addr)
  189. {
  190. struct inet6_dev *idev = rt->rt6i_idev;
  191. struct ifacaddr6 *aca;
  192. aca = kzalloc(sizeof(*aca), GFP_ATOMIC);
  193. if (!aca)
  194. return NULL;
  195. aca->aca_addr = *addr;
  196. in6_dev_hold(idev);
  197. aca->aca_idev = idev;
  198. aca->aca_rt = rt;
  199. aca->aca_users = 1;
  200. /* aca_tstamp should be updated upon changes */
  201. aca->aca_cstamp = aca->aca_tstamp = jiffies;
  202. atomic_set(&aca->aca_refcnt, 1);
  203. return aca;
  204. }
  205. /*
  206. * device anycast group inc (add if not found)
  207. */
  208. int __ipv6_dev_ac_inc(struct inet6_dev *idev, const struct in6_addr *addr)
  209. {
  210. struct ifacaddr6 *aca;
  211. struct rt6_info *rt;
  212. int err;
  213. ASSERT_RTNL();
  214. write_lock_bh(&idev->lock);
  215. if (idev->dead) {
  216. err = -ENODEV;
  217. goto out;
  218. }
  219. for (aca = idev->ac_list; aca; aca = aca->aca_next) {
  220. if (ipv6_addr_equal(&aca->aca_addr, addr)) {
  221. aca->aca_users++;
  222. err = 0;
  223. goto out;
  224. }
  225. }
  226. rt = addrconf_dst_alloc(idev, addr, true);
  227. if (IS_ERR(rt)) {
  228. err = PTR_ERR(rt);
  229. goto out;
  230. }
  231. aca = aca_alloc(rt, addr);
  232. if (!aca) {
  233. ip6_rt_put(rt);
  234. err = -ENOMEM;
  235. goto out;
  236. }
  237. aca->aca_next = idev->ac_list;
  238. idev->ac_list = aca;
  239. /* Hold this for addrconf_join_solict() below before we unlock,
  240. * it is already exposed via idev->ac_list.
  241. */
  242. aca_get(aca);
  243. write_unlock_bh(&idev->lock);
  244. ip6_ins_rt(rt);
  245. addrconf_join_solict(idev->dev, &aca->aca_addr);
  246. aca_put(aca);
  247. return 0;
  248. out:
  249. write_unlock_bh(&idev->lock);
  250. return err;
  251. }
  252. /*
  253. * device anycast group decrement
  254. */
  255. int __ipv6_dev_ac_dec(struct inet6_dev *idev, const struct in6_addr *addr)
  256. {
  257. struct ifacaddr6 *aca, *prev_aca;
  258. ASSERT_RTNL();
  259. write_lock_bh(&idev->lock);
  260. prev_aca = NULL;
  261. for (aca = idev->ac_list; aca; aca = aca->aca_next) {
  262. if (ipv6_addr_equal(&aca->aca_addr, addr))
  263. break;
  264. prev_aca = aca;
  265. }
  266. if (!aca) {
  267. write_unlock_bh(&idev->lock);
  268. return -ENOENT;
  269. }
  270. if (--aca->aca_users > 0) {
  271. write_unlock_bh(&idev->lock);
  272. return 0;
  273. }
  274. if (prev_aca)
  275. prev_aca->aca_next = aca->aca_next;
  276. else
  277. idev->ac_list = aca->aca_next;
  278. write_unlock_bh(&idev->lock);
  279. addrconf_leave_solict(idev, &aca->aca_addr);
  280. dst_hold(&aca->aca_rt->dst);
  281. ip6_del_rt(aca->aca_rt);
  282. aca_put(aca);
  283. return 0;
  284. }
  285. /* called with rtnl_lock() */
  286. static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr)
  287. {
  288. struct inet6_dev *idev = __in6_dev_get(dev);
  289. if (!idev)
  290. return -ENODEV;
  291. return __ipv6_dev_ac_dec(idev, addr);
  292. }
  293. void ipv6_ac_destroy_dev(struct inet6_dev *idev)
  294. {
  295. struct ifacaddr6 *aca;
  296. write_lock_bh(&idev->lock);
  297. while ((aca = idev->ac_list) != NULL) {
  298. idev->ac_list = aca->aca_next;
  299. write_unlock_bh(&idev->lock);
  300. addrconf_leave_solict(idev, &aca->aca_addr);
  301. dst_hold(&aca->aca_rt->dst);
  302. ip6_del_rt(aca->aca_rt);
  303. aca_put(aca);
  304. write_lock_bh(&idev->lock);
  305. }
  306. write_unlock_bh(&idev->lock);
  307. }
  308. /*
  309. * check if the interface has this anycast address
  310. * called with rcu_read_lock()
  311. */
  312. static bool ipv6_chk_acast_dev(struct net_device *dev, const struct in6_addr *addr)
  313. {
  314. struct inet6_dev *idev;
  315. struct ifacaddr6 *aca;
  316. idev = __in6_dev_get(dev);
  317. if (idev) {
  318. read_lock_bh(&idev->lock);
  319. for (aca = idev->ac_list; aca; aca = aca->aca_next)
  320. if (ipv6_addr_equal(&aca->aca_addr, addr))
  321. break;
  322. read_unlock_bh(&idev->lock);
  323. return aca != NULL;
  324. }
  325. return false;
  326. }
  327. /*
  328. * check if given interface (or any, if dev==0) has this anycast address
  329. */
  330. bool ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
  331. const struct in6_addr *addr)
  332. {
  333. bool found = false;
  334. rcu_read_lock();
  335. if (dev)
  336. found = ipv6_chk_acast_dev(dev, addr);
  337. else
  338. for_each_netdev_rcu(net, dev)
  339. if (ipv6_chk_acast_dev(dev, addr)) {
  340. found = true;
  341. break;
  342. }
  343. rcu_read_unlock();
  344. return found;
  345. }
  346. /* check if this anycast address is link-local on given interface or
  347. * is global
  348. */
  349. bool ipv6_chk_acast_addr_src(struct net *net, struct net_device *dev,
  350. const struct in6_addr *addr)
  351. {
  352. return ipv6_chk_acast_addr(net,
  353. (ipv6_addr_type(addr) & IPV6_ADDR_LINKLOCAL ?
  354. dev : NULL),
  355. addr);
  356. }
  357. #ifdef CONFIG_PROC_FS
  358. struct ac6_iter_state {
  359. struct seq_net_private p;
  360. struct net_device *dev;
  361. struct inet6_dev *idev;
  362. };
  363. #define ac6_seq_private(seq) ((struct ac6_iter_state *)(seq)->private)
  364. static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
  365. {
  366. struct ifacaddr6 *im = NULL;
  367. struct ac6_iter_state *state = ac6_seq_private(seq);
  368. struct net *net = seq_file_net(seq);
  369. state->idev = NULL;
  370. for_each_netdev_rcu(net, state->dev) {
  371. struct inet6_dev *idev;
  372. idev = __in6_dev_get(state->dev);
  373. if (!idev)
  374. continue;
  375. read_lock_bh(&idev->lock);
  376. im = idev->ac_list;
  377. if (im) {
  378. state->idev = idev;
  379. break;
  380. }
  381. read_unlock_bh(&idev->lock);
  382. }
  383. return im;
  384. }
  385. static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
  386. {
  387. struct ac6_iter_state *state = ac6_seq_private(seq);
  388. im = im->aca_next;
  389. while (!im) {
  390. if (likely(state->idev != NULL))
  391. read_unlock_bh(&state->idev->lock);
  392. state->dev = next_net_device_rcu(state->dev);
  393. if (!state->dev) {
  394. state->idev = NULL;
  395. break;
  396. }
  397. state->idev = __in6_dev_get(state->dev);
  398. if (!state->idev)
  399. continue;
  400. read_lock_bh(&state->idev->lock);
  401. im = state->idev->ac_list;
  402. }
  403. return im;
  404. }
  405. static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
  406. {
  407. struct ifacaddr6 *im = ac6_get_first(seq);
  408. if (im)
  409. while (pos && (im = ac6_get_next(seq, im)) != NULL)
  410. --pos;
  411. return pos ? NULL : im;
  412. }
  413. static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
  414. __acquires(RCU)
  415. {
  416. rcu_read_lock();
  417. return ac6_get_idx(seq, *pos);
  418. }
  419. static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
  420. {
  421. struct ifacaddr6 *im = ac6_get_next(seq, v);
  422. ++*pos;
  423. return im;
  424. }
  425. static void ac6_seq_stop(struct seq_file *seq, void *v)
  426. __releases(RCU)
  427. {
  428. struct ac6_iter_state *state = ac6_seq_private(seq);
  429. if (likely(state->idev != NULL)) {
  430. read_unlock_bh(&state->idev->lock);
  431. state->idev = NULL;
  432. }
  433. rcu_read_unlock();
  434. }
  435. static int ac6_seq_show(struct seq_file *seq, void *v)
  436. {
  437. struct ifacaddr6 *im = (struct ifacaddr6 *)v;
  438. struct ac6_iter_state *state = ac6_seq_private(seq);
  439. seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
  440. state->dev->ifindex, state->dev->name,
  441. &im->aca_addr, im->aca_users);
  442. return 0;
  443. }
  444. static const struct seq_operations ac6_seq_ops = {
  445. .start = ac6_seq_start,
  446. .next = ac6_seq_next,
  447. .stop = ac6_seq_stop,
  448. .show = ac6_seq_show,
  449. };
  450. static int ac6_seq_open(struct inode *inode, struct file *file)
  451. {
  452. return seq_open_net(inode, file, &ac6_seq_ops,
  453. sizeof(struct ac6_iter_state));
  454. }
  455. static const struct file_operations ac6_seq_fops = {
  456. .owner = THIS_MODULE,
  457. .open = ac6_seq_open,
  458. .read = seq_read,
  459. .llseek = seq_lseek,
  460. .release = seq_release_net,
  461. };
  462. int __net_init ac6_proc_init(struct net *net)
  463. {
  464. if (!proc_create("anycast6", S_IRUGO, net->proc_net, &ac6_seq_fops))
  465. return -ENOMEM;
  466. return 0;
  467. }
  468. void ac6_proc_exit(struct net *net)
  469. {
  470. remove_proc_entry("anycast6", net->proc_net);
  471. }
  472. #endif