neighbour.h 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566
  1. /* SPDX-License-Identifier: GPL-2.0 */
  2. #ifndef _NET_NEIGHBOUR_H
  3. #define _NET_NEIGHBOUR_H
  4. #include <linux/neighbour.h>
  5. /*
  6. * Generic neighbour manipulation
  7. *
  8. * Authors:
  9. * Pedro Roque <roque@di.fc.ul.pt>
  10. * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
  11. *
  12. * Changes:
  13. *
  14. * Harald Welte: <laforge@gnumonks.org>
  15. * - Add neighbour cache statistics like rtstat
  16. */
  17. #include <linux/atomic.h>
  18. #include <linux/refcount.h>
  19. #include <linux/netdevice.h>
  20. #include <linux/skbuff.h>
  21. #include <linux/rcupdate.h>
  22. #include <linux/seq_file.h>
  23. #include <linux/bitmap.h>
  24. #include <linux/err.h>
  25. #include <linux/sysctl.h>
  26. #include <linux/workqueue.h>
  27. #include <net/rtnetlink.h>
  28. /*
  29. * NUD stands for "neighbor unreachability detection"
  30. */
  31. #define NUD_IN_TIMER (NUD_INCOMPLETE|NUD_REACHABLE|NUD_DELAY|NUD_PROBE)
  32. #define NUD_VALID (NUD_PERMANENT|NUD_NOARP|NUD_REACHABLE|NUD_PROBE|NUD_STALE|NUD_DELAY)
  33. #define NUD_CONNECTED (NUD_PERMANENT|NUD_NOARP|NUD_REACHABLE)
  34. struct neighbour;
  35. enum {
  36. NEIGH_VAR_MCAST_PROBES,
  37. NEIGH_VAR_UCAST_PROBES,
  38. NEIGH_VAR_APP_PROBES,
  39. NEIGH_VAR_MCAST_REPROBES,
  40. NEIGH_VAR_RETRANS_TIME,
  41. NEIGH_VAR_BASE_REACHABLE_TIME,
  42. NEIGH_VAR_DELAY_PROBE_TIME,
  43. NEIGH_VAR_GC_STALETIME,
  44. NEIGH_VAR_QUEUE_LEN_BYTES,
  45. NEIGH_VAR_PROXY_QLEN,
  46. NEIGH_VAR_ANYCAST_DELAY,
  47. NEIGH_VAR_PROXY_DELAY,
  48. NEIGH_VAR_LOCKTIME,
  49. #define NEIGH_VAR_DATA_MAX (NEIGH_VAR_LOCKTIME + 1)
  50. /* Following are used as a second way to access one of the above */
  51. NEIGH_VAR_QUEUE_LEN, /* same data as NEIGH_VAR_QUEUE_LEN_BYTES */
  52. NEIGH_VAR_RETRANS_TIME_MS, /* same data as NEIGH_VAR_RETRANS_TIME */
  53. NEIGH_VAR_BASE_REACHABLE_TIME_MS, /* same data as NEIGH_VAR_BASE_REACHABLE_TIME */
  54. /* Following are used by "default" only */
  55. NEIGH_VAR_GC_INTERVAL,
  56. NEIGH_VAR_GC_THRESH1,
  57. NEIGH_VAR_GC_THRESH2,
  58. NEIGH_VAR_GC_THRESH3,
  59. NEIGH_VAR_MAX
  60. };
  61. struct neigh_parms {
  62. possible_net_t net;
  63. struct net_device *dev;
  64. struct list_head list;
  65. int (*neigh_setup)(struct neighbour *);
  66. void (*neigh_cleanup)(struct neighbour *);
  67. struct neigh_table *tbl;
  68. void *sysctl_table;
  69. int dead;
  70. refcount_t refcnt;
  71. struct rcu_head rcu_head;
  72. int reachable_time;
  73. int data[NEIGH_VAR_DATA_MAX];
  74. DECLARE_BITMAP(data_state, NEIGH_VAR_DATA_MAX);
  75. };
  76. static inline void neigh_var_set(struct neigh_parms *p, int index, int val)
  77. {
  78. set_bit(index, p->data_state);
  79. p->data[index] = val;
  80. }
  81. #define NEIGH_VAR(p, attr) ((p)->data[NEIGH_VAR_ ## attr])
  82. /* In ndo_neigh_setup, NEIGH_VAR_INIT should be used.
  83. * In other cases, NEIGH_VAR_SET should be used.
  84. */
  85. #define NEIGH_VAR_INIT(p, attr, val) (NEIGH_VAR(p, attr) = val)
  86. #define NEIGH_VAR_SET(p, attr, val) neigh_var_set(p, NEIGH_VAR_ ## attr, val)
  87. static inline void neigh_parms_data_state_setall(struct neigh_parms *p)
  88. {
  89. bitmap_fill(p->data_state, NEIGH_VAR_DATA_MAX);
  90. }
  91. static inline void neigh_parms_data_state_cleanall(struct neigh_parms *p)
  92. {
  93. bitmap_zero(p->data_state, NEIGH_VAR_DATA_MAX);
  94. }
  95. struct neigh_statistics {
  96. unsigned long allocs; /* number of allocated neighs */
  97. unsigned long destroys; /* number of destroyed neighs */
  98. unsigned long hash_grows; /* number of hash resizes */
  99. unsigned long res_failed; /* number of failed resolutions */
  100. unsigned long lookups; /* number of lookups */
  101. unsigned long hits; /* number of hits (among lookups) */
  102. unsigned long rcv_probes_mcast; /* number of received mcast ipv6 */
  103. unsigned long rcv_probes_ucast; /* number of received ucast ipv6 */
  104. unsigned long periodic_gc_runs; /* number of periodic GC runs */
  105. unsigned long forced_gc_runs; /* number of forced GC runs */
  106. unsigned long unres_discards; /* number of unresolved drops */
  107. unsigned long table_fulls; /* times even gc couldn't help */
  108. };
  109. #define NEIGH_CACHE_STAT_INC(tbl, field) this_cpu_inc((tbl)->stats->field)
  110. struct neighbour {
  111. struct neighbour __rcu *next;
  112. struct neigh_table *tbl;
  113. struct neigh_parms *parms;
  114. unsigned long confirmed;
  115. unsigned long updated;
  116. rwlock_t lock;
  117. refcount_t refcnt;
  118. struct sk_buff_head arp_queue;
  119. unsigned int arp_queue_len_bytes;
  120. struct timer_list timer;
  121. unsigned long used;
  122. atomic_t probes;
  123. __u8 flags;
  124. __u8 nud_state;
  125. __u8 type;
  126. __u8 dead;
  127. seqlock_t ha_lock;
  128. unsigned char ha[ALIGN(MAX_ADDR_LEN, sizeof(unsigned long))];
  129. struct hh_cache hh;
  130. int (*output)(struct neighbour *, struct sk_buff *);
  131. const struct neigh_ops *ops;
  132. struct rcu_head rcu;
  133. struct net_device *dev;
  134. u8 primary_key[0];
  135. } __randomize_layout;
  136. struct neigh_ops {
  137. int family;
  138. void (*solicit)(struct neighbour *, struct sk_buff *);
  139. void (*error_report)(struct neighbour *, struct sk_buff *);
  140. int (*output)(struct neighbour *, struct sk_buff *);
  141. int (*connected_output)(struct neighbour *, struct sk_buff *);
  142. };
  143. struct pneigh_entry {
  144. struct pneigh_entry *next;
  145. possible_net_t net;
  146. struct net_device *dev;
  147. u8 flags;
  148. u8 key[0];
  149. };
  150. /*
  151. * neighbour table manipulation
  152. */
  153. #define NEIGH_NUM_HASH_RND 4
  154. struct neigh_hash_table {
  155. struct neighbour __rcu **hash_buckets;
  156. unsigned int hash_shift;
  157. __u32 hash_rnd[NEIGH_NUM_HASH_RND];
  158. struct rcu_head rcu;
  159. };
  160. struct neigh_table {
  161. int family;
  162. unsigned int entry_size;
  163. unsigned int key_len;
  164. __be16 protocol;
  165. __u32 (*hash)(const void *pkey,
  166. const struct net_device *dev,
  167. __u32 *hash_rnd);
  168. bool (*key_eq)(const struct neighbour *, const void *pkey);
  169. int (*constructor)(struct neighbour *);
  170. int (*pconstructor)(struct pneigh_entry *);
  171. void (*pdestructor)(struct pneigh_entry *);
  172. void (*proxy_redo)(struct sk_buff *skb);
  173. char *id;
  174. struct neigh_parms parms;
  175. struct list_head parms_list;
  176. int gc_interval;
  177. int gc_thresh1;
  178. int gc_thresh2;
  179. int gc_thresh3;
  180. unsigned long last_flush;
  181. struct delayed_work gc_work;
  182. struct timer_list proxy_timer;
  183. struct sk_buff_head proxy_queue;
  184. atomic_t entries;
  185. rwlock_t lock;
  186. unsigned long last_rand;
  187. struct neigh_statistics __percpu *stats;
  188. struct neigh_hash_table __rcu *nht;
  189. struct pneigh_entry **phash_buckets;
  190. };
  191. enum {
  192. NEIGH_ARP_TABLE = 0,
  193. NEIGH_ND_TABLE = 1,
  194. NEIGH_DN_TABLE = 2,
  195. NEIGH_NR_TABLES,
  196. NEIGH_LINK_TABLE = NEIGH_NR_TABLES /* Pseudo table for neigh_xmit */
  197. };
  198. static inline int neigh_parms_family(struct neigh_parms *p)
  199. {
  200. return p->tbl->family;
  201. }
  202. #define NEIGH_PRIV_ALIGN sizeof(long long)
  203. #define NEIGH_ENTRY_SIZE(size) ALIGN((size), NEIGH_PRIV_ALIGN)
  204. static inline void *neighbour_priv(const struct neighbour *n)
  205. {
  206. return (char *)n + n->tbl->entry_size;
  207. }
  208. /* flags for neigh_update() */
  209. #define NEIGH_UPDATE_F_OVERRIDE 0x00000001
  210. #define NEIGH_UPDATE_F_WEAK_OVERRIDE 0x00000002
  211. #define NEIGH_UPDATE_F_OVERRIDE_ISROUTER 0x00000004
  212. #define NEIGH_UPDATE_F_EXT_LEARNED 0x20000000
  213. #define NEIGH_UPDATE_F_ISROUTER 0x40000000
  214. #define NEIGH_UPDATE_F_ADMIN 0x80000000
  215. static inline bool neigh_key_eq16(const struct neighbour *n, const void *pkey)
  216. {
  217. return *(const u16 *)n->primary_key == *(const u16 *)pkey;
  218. }
  219. static inline bool neigh_key_eq32(const struct neighbour *n, const void *pkey)
  220. {
  221. return *(const u32 *)n->primary_key == *(const u32 *)pkey;
  222. }
  223. static inline bool neigh_key_eq128(const struct neighbour *n, const void *pkey)
  224. {
  225. const u32 *n32 = (const u32 *)n->primary_key;
  226. const u32 *p32 = pkey;
  227. return ((n32[0] ^ p32[0]) | (n32[1] ^ p32[1]) |
  228. (n32[2] ^ p32[2]) | (n32[3] ^ p32[3])) == 0;
  229. }
  230. static inline struct neighbour *___neigh_lookup_noref(
  231. struct neigh_table *tbl,
  232. bool (*key_eq)(const struct neighbour *n, const void *pkey),
  233. __u32 (*hash)(const void *pkey,
  234. const struct net_device *dev,
  235. __u32 *hash_rnd),
  236. const void *pkey,
  237. struct net_device *dev)
  238. {
  239. struct neigh_hash_table *nht = rcu_dereference_bh(tbl->nht);
  240. struct neighbour *n;
  241. u32 hash_val;
  242. hash_val = hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
  243. for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
  244. n != NULL;
  245. n = rcu_dereference_bh(n->next)) {
  246. if (n->dev == dev && key_eq(n, pkey))
  247. return n;
  248. }
  249. return NULL;
  250. }
  251. static inline struct neighbour *__neigh_lookup_noref(struct neigh_table *tbl,
  252. const void *pkey,
  253. struct net_device *dev)
  254. {
  255. return ___neigh_lookup_noref(tbl, tbl->key_eq, tbl->hash, pkey, dev);
  256. }
  257. void neigh_table_init(int index, struct neigh_table *tbl);
  258. int neigh_table_clear(int index, struct neigh_table *tbl);
  259. struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
  260. struct net_device *dev);
  261. struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
  262. const void *pkey);
  263. struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey,
  264. struct net_device *dev, bool want_ref);
  265. static inline struct neighbour *neigh_create(struct neigh_table *tbl,
  266. const void *pkey,
  267. struct net_device *dev)
  268. {
  269. return __neigh_create(tbl, pkey, dev, true);
  270. }
  271. void neigh_destroy(struct neighbour *neigh);
  272. int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb);
  273. int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, u32 flags,
  274. u32 nlmsg_pid);
  275. void __neigh_set_probe_once(struct neighbour *neigh);
  276. bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl);
  277. void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev);
  278. int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
  279. int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb);
  280. int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb);
  281. int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb);
  282. struct neighbour *neigh_event_ns(struct neigh_table *tbl,
  283. u8 *lladdr, void *saddr,
  284. struct net_device *dev);
  285. struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
  286. struct neigh_table *tbl);
  287. void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms);
  288. static inline
  289. struct net *neigh_parms_net(const struct neigh_parms *parms)
  290. {
  291. return read_pnet(&parms->net);
  292. }
  293. unsigned long neigh_rand_reach_time(unsigned long base);
  294. void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
  295. struct sk_buff *skb);
  296. struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net,
  297. const void *key, struct net_device *dev,
  298. int creat);
  299. struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl, struct net *net,
  300. const void *key, struct net_device *dev);
  301. int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *key,
  302. struct net_device *dev);
  303. static inline struct net *pneigh_net(const struct pneigh_entry *pneigh)
  304. {
  305. return read_pnet(&pneigh->net);
  306. }
  307. void neigh_app_ns(struct neighbour *n);
  308. void neigh_for_each(struct neigh_table *tbl,
  309. void (*cb)(struct neighbour *, void *), void *cookie);
  310. void __neigh_for_each_release(struct neigh_table *tbl,
  311. int (*cb)(struct neighbour *));
  312. int neigh_xmit(int fam, struct net_device *, const void *, struct sk_buff *);
  313. void pneigh_for_each(struct neigh_table *tbl,
  314. void (*cb)(struct pneigh_entry *));
  315. struct neigh_seq_state {
  316. struct seq_net_private p;
  317. struct neigh_table *tbl;
  318. struct neigh_hash_table *nht;
  319. void *(*neigh_sub_iter)(struct neigh_seq_state *state,
  320. struct neighbour *n, loff_t *pos);
  321. unsigned int bucket;
  322. unsigned int flags;
  323. #define NEIGH_SEQ_NEIGH_ONLY 0x00000001
  324. #define NEIGH_SEQ_IS_PNEIGH 0x00000002
  325. #define NEIGH_SEQ_SKIP_NOARP 0x00000004
  326. };
  327. void *neigh_seq_start(struct seq_file *, loff_t *, struct neigh_table *,
  328. unsigned int);
  329. void *neigh_seq_next(struct seq_file *, void *, loff_t *);
  330. void neigh_seq_stop(struct seq_file *, void *);
  331. int neigh_proc_dointvec(struct ctl_table *ctl, int write,
  332. void __user *buffer, size_t *lenp, loff_t *ppos);
  333. int neigh_proc_dointvec_jiffies(struct ctl_table *ctl, int write,
  334. void __user *buffer,
  335. size_t *lenp, loff_t *ppos);
  336. int neigh_proc_dointvec_ms_jiffies(struct ctl_table *ctl, int write,
  337. void __user *buffer,
  338. size_t *lenp, loff_t *ppos);
  339. int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
  340. proc_handler *proc_handler);
  341. void neigh_sysctl_unregister(struct neigh_parms *p);
  342. static inline void __neigh_parms_put(struct neigh_parms *parms)
  343. {
  344. refcount_dec(&parms->refcnt);
  345. }
  346. static inline struct neigh_parms *neigh_parms_clone(struct neigh_parms *parms)
  347. {
  348. refcount_inc(&parms->refcnt);
  349. return parms;
  350. }
  351. /*
  352. * Neighbour references
  353. */
  354. static inline void neigh_release(struct neighbour *neigh)
  355. {
  356. if (refcount_dec_and_test(&neigh->refcnt))
  357. neigh_destroy(neigh);
  358. }
  359. static inline struct neighbour * neigh_clone(struct neighbour *neigh)
  360. {
  361. if (neigh)
  362. refcount_inc(&neigh->refcnt);
  363. return neigh;
  364. }
  365. #define neigh_hold(n) refcount_inc(&(n)->refcnt)
  366. static inline int neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
  367. {
  368. unsigned long now = jiffies;
  369. if (READ_ONCE(neigh->used) != now)
  370. WRITE_ONCE(neigh->used, now);
  371. if (!(neigh->nud_state&(NUD_CONNECTED|NUD_DELAY|NUD_PROBE)))
  372. return __neigh_event_send(neigh, skb);
  373. return 0;
  374. }
  375. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  376. static inline int neigh_hh_bridge(struct hh_cache *hh, struct sk_buff *skb)
  377. {
  378. unsigned int seq, hh_alen;
  379. do {
  380. seq = read_seqbegin(&hh->hh_lock);
  381. hh_alen = HH_DATA_ALIGN(ETH_HLEN);
  382. memcpy(skb->data - hh_alen, hh->hh_data, ETH_ALEN + hh_alen - ETH_HLEN);
  383. } while (read_seqretry(&hh->hh_lock, seq));
  384. return 0;
  385. }
  386. #endif
  387. static inline int neigh_hh_output(const struct hh_cache *hh, struct sk_buff *skb)
  388. {
  389. unsigned int hh_alen = 0;
  390. unsigned int seq;
  391. unsigned int hh_len;
  392. do {
  393. seq = read_seqbegin(&hh->hh_lock);
  394. hh_len = READ_ONCE(hh->hh_len);
  395. if (likely(hh_len <= HH_DATA_MOD)) {
  396. hh_alen = HH_DATA_MOD;
  397. /* skb_push() would proceed silently if we have room for
  398. * the unaligned size but not for the aligned size:
  399. * check headroom explicitly.
  400. */
  401. if (likely(skb_headroom(skb) >= HH_DATA_MOD)) {
  402. /* this is inlined by gcc */
  403. memcpy(skb->data - HH_DATA_MOD, hh->hh_data,
  404. HH_DATA_MOD);
  405. }
  406. } else {
  407. hh_alen = HH_DATA_ALIGN(hh_len);
  408. if (likely(skb_headroom(skb) >= hh_alen)) {
  409. memcpy(skb->data - hh_alen, hh->hh_data,
  410. hh_alen);
  411. }
  412. }
  413. } while (read_seqretry(&hh->hh_lock, seq));
  414. if (WARN_ON_ONCE(skb_headroom(skb) < hh_alen)) {
  415. kfree_skb(skb);
  416. return NET_XMIT_DROP;
  417. }
  418. __skb_push(skb, hh_len);
  419. return dev_queue_xmit(skb);
  420. }
  421. static inline int neigh_output(struct neighbour *n, struct sk_buff *skb)
  422. {
  423. const struct hh_cache *hh = &n->hh;
  424. if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
  425. return neigh_hh_output(hh, skb);
  426. else
  427. return n->output(n, skb);
  428. }
  429. static inline struct neighbour *
  430. __neigh_lookup(struct neigh_table *tbl, const void *pkey, struct net_device *dev, int creat)
  431. {
  432. struct neighbour *n = neigh_lookup(tbl, pkey, dev);
  433. if (n || !creat)
  434. return n;
  435. n = neigh_create(tbl, pkey, dev);
  436. return IS_ERR(n) ? NULL : n;
  437. }
  438. static inline struct neighbour *
  439. __neigh_lookup_errno(struct neigh_table *tbl, const void *pkey,
  440. struct net_device *dev)
  441. {
  442. struct neighbour *n = neigh_lookup(tbl, pkey, dev);
  443. if (n)
  444. return n;
  445. return neigh_create(tbl, pkey, dev);
  446. }
  447. struct neighbour_cb {
  448. unsigned long sched_next;
  449. unsigned int flags;
  450. };
  451. #define LOCALLY_ENQUEUED 0x1
  452. #define NEIGH_CB(skb) ((struct neighbour_cb *)(skb)->cb)
  453. static inline void neigh_ha_snapshot(char *dst, const struct neighbour *n,
  454. const struct net_device *dev)
  455. {
  456. unsigned int seq;
  457. do {
  458. seq = read_seqbegin(&n->ha_lock);
  459. memcpy(dst, n->ha, dev->addr_len);
  460. } while (read_seqretry(&n->ha_lock, seq));
  461. }
  462. static inline void neigh_update_ext_learned(struct neighbour *neigh, u32 flags,
  463. int *notify)
  464. {
  465. u8 ndm_flags = 0;
  466. if (!(flags & NEIGH_UPDATE_F_ADMIN))
  467. return;
  468. ndm_flags |= (flags & NEIGH_UPDATE_F_EXT_LEARNED) ? NTF_EXT_LEARNED : 0;
  469. if ((neigh->flags ^ ndm_flags) & NTF_EXT_LEARNED) {
  470. if (ndm_flags & NTF_EXT_LEARNED)
  471. neigh->flags |= NTF_EXT_LEARNED;
  472. else
  473. neigh->flags &= ~NTF_EXT_LEARNED;
  474. *notify = 1;
  475. }
  476. }
  477. #endif