1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _NET_NEIGHBOUR_H 3 #define _NET_NEIGHBOUR_H 4 5 #include <linux/neighbour.h> 6 7 /* 8 * Generic neighbour manipulation 9 * 10 * Authors: 11 * Pedro Roque <roque@di.fc.ul.pt> 12 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 13 * 14 * Changes: 15 * 16 * Harald Welte: <laforge@gnumonks.org> 17 * - Add neighbour cache statistics like rtstat 18 */ 19 20 #include <linux/atomic.h> 21 #include <linux/refcount.h> 22 #include <linux/netdevice.h> 23 #include <linux/skbuff.h> 24 #include <linux/rcupdate.h> 25 #include <linux/seq_file.h> 26 #include <linux/bitmap.h> 27 28 #include <linux/err.h> 29 #include <linux/sysctl.h> 30 #include <linux/workqueue.h> 31 #include <net/rtnetlink.h> 32 #include <net/neighbour_tables.h> 33 34 /* 35 * NUD stands for "neighbor unreachability detection" 36 */ 37 38 #define NUD_IN_TIMER (NUD_INCOMPLETE|NUD_REACHABLE|NUD_DELAY|NUD_PROBE) 39 #define NUD_VALID (NUD_PERMANENT|NUD_NOARP|NUD_REACHABLE|NUD_PROBE|NUD_STALE|NUD_DELAY) 40 #define NUD_CONNECTED (NUD_PERMANENT|NUD_NOARP|NUD_REACHABLE) 41 42 struct neighbour; 43 44 enum { 45 NEIGH_VAR_MCAST_PROBES, 46 NEIGH_VAR_UCAST_PROBES, 47 NEIGH_VAR_APP_PROBES, 48 NEIGH_VAR_MCAST_REPROBES, 49 NEIGH_VAR_RETRANS_TIME, 50 NEIGH_VAR_BASE_REACHABLE_TIME, 51 NEIGH_VAR_DELAY_PROBE_TIME, 52 NEIGH_VAR_INTERVAL_PROBE_TIME_MS, 53 NEIGH_VAR_GC_STALETIME, 54 NEIGH_VAR_QUEUE_LEN_BYTES, 55 NEIGH_VAR_PROXY_QLEN, 56 NEIGH_VAR_ANYCAST_DELAY, 57 NEIGH_VAR_PROXY_DELAY, 58 NEIGH_VAR_LOCKTIME, 59 #define NEIGH_VAR_DATA_MAX (NEIGH_VAR_LOCKTIME + 1) 60 /* Following are used as a second way to access one of the above */ 61 NEIGH_VAR_QUEUE_LEN, /* same data as NEIGH_VAR_QUEUE_LEN_BYTES */ 62 NEIGH_VAR_RETRANS_TIME_MS, /* same data as NEIGH_VAR_RETRANS_TIME */ 63 NEIGH_VAR_BASE_REACHABLE_TIME_MS, /* same data as NEIGH_VAR_BASE_REACHABLE_TIME */ 64 /* Following are used by "default" only */ 65 NEIGH_VAR_GC_INTERVAL, 66 NEIGH_VAR_GC_THRESH1, 67 NEIGH_VAR_GC_THRESH2, 68 NEIGH_VAR_GC_THRESH3, 69 NEIGH_VAR_MAX 70 }; 71 72 struct neigh_parms { 73 possible_net_t net; 74 struct net_device *dev; 75 netdevice_tracker dev_tracker; 76 struct list_head list; 77 int (*neigh_setup)(struct neighbour *); 78 struct neigh_table *tbl; 79 80 void *sysctl_table; 81 82 int dead; 83 refcount_t refcnt; 84 struct rcu_head rcu_head; 85 86 int reachable_time; 87 u32 qlen; 88 int data[NEIGH_VAR_DATA_MAX]; 89 DECLARE_BITMAP(data_state, NEIGH_VAR_DATA_MAX); 90 }; 91 92 static inline void neigh_var_set(struct neigh_parms *p, int index, int val) 93 { 94 set_bit(index, p->data_state); 95 p->data[index] = val; 96 } 97 98 #define NEIGH_VAR(p, attr) ((p)->data[NEIGH_VAR_ ## attr]) 99 100 /* In ndo_neigh_setup, NEIGH_VAR_INIT should be used. 101 * In other cases, NEIGH_VAR_SET should be used. 102 */ 103 #define NEIGH_VAR_INIT(p, attr, val) (NEIGH_VAR(p, attr) = val) 104 #define NEIGH_VAR_SET(p, attr, val) neigh_var_set(p, NEIGH_VAR_ ## attr, val) 105 106 static inline void neigh_parms_data_state_setall(struct neigh_parms *p) 107 { 108 bitmap_fill(p->data_state, NEIGH_VAR_DATA_MAX); 109 } 110 111 static inline void neigh_parms_data_state_cleanall(struct neigh_parms *p) 112 { 113 bitmap_zero(p->data_state, NEIGH_VAR_DATA_MAX); 114 } 115 116 struct neigh_statistics { 117 unsigned long allocs; /* number of allocated neighs */ 118 unsigned long destroys; /* number of destroyed neighs */ 119 unsigned long hash_grows; /* number of hash resizes */ 120 121 unsigned long res_failed; /* number of failed resolutions */ 122 123 unsigned long lookups; /* number of lookups */ 124 unsigned long hits; /* number of hits (among lookups) */ 125 126 unsigned long rcv_probes_mcast; /* number of received mcast ipv6 */ 127 unsigned long rcv_probes_ucast; /* number of received ucast ipv6 */ 128 129 unsigned long periodic_gc_runs; /* number of periodic GC runs */ 130 unsigned long forced_gc_runs; /* number of forced GC runs */ 131 132 unsigned long unres_discards; /* number of unresolved drops */ 133 unsigned long table_fulls; /* times even gc couldn't help */ 134 }; 135 136 #define NEIGH_CACHE_STAT_INC(tbl, field) this_cpu_inc((tbl)->stats->field) 137 138 struct neighbour { 139 struct hlist_node hash; 140 struct hlist_node dev_list; 141 struct neigh_table *tbl; 142 struct neigh_parms *parms; 143 unsigned long confirmed; 144 unsigned long updated; 145 rwlock_t lock; 146 refcount_t refcnt; 147 unsigned int arp_queue_len_bytes; 148 struct sk_buff_head arp_queue; 149 struct timer_list timer; 150 unsigned long used; 151 atomic_t probes; 152 u8 nud_state; 153 u8 type; 154 u8 dead; 155 u8 protocol; 156 u32 flags; 157 seqlock_t ha_lock; 158 unsigned char ha[ALIGN(MAX_ADDR_LEN, sizeof(unsigned long))] __aligned(8); 159 struct hh_cache hh; 160 int (*output)(struct neighbour *, struct sk_buff *); 161 const struct neigh_ops *ops; 162 struct list_head gc_list; 163 struct list_head managed_list; 164 struct rcu_head rcu; 165 struct net_device *dev; 166 netdevice_tracker dev_tracker; 167 u8 primary_key[]; 168 } __randomize_layout; 169 170 struct neigh_ops { 171 int family; 172 void (*solicit)(struct neighbour *, struct sk_buff *); 173 void (*error_report)(struct neighbour *, struct sk_buff *); 174 int (*output)(struct neighbour *, struct sk_buff *); 175 int (*connected_output)(struct neighbour *, struct sk_buff *); 176 }; 177 178 struct pneigh_entry { 179 struct pneigh_entry *next; 180 possible_net_t net; 181 struct net_device *dev; 182 netdevice_tracker dev_tracker; 183 u32 flags; 184 u8 protocol; 185 u32 key[]; 186 }; 187 188 /* 189 * neighbour table manipulation 190 */ 191 192 #define NEIGH_NUM_HASH_RND 4 193 194 struct neigh_hash_table { 195 struct hlist_head *hash_heads; 196 unsigned int hash_shift; 197 __u32 hash_rnd[NEIGH_NUM_HASH_RND]; 198 struct rcu_head rcu; 199 }; 200 201 202 struct neigh_table { 203 int family; 204 unsigned int entry_size; 205 unsigned int key_len; 206 __be16 protocol; 207 __u32 (*hash)(const void *pkey, 208 const struct net_device *dev, 209 __u32 *hash_rnd); 210 bool (*key_eq)(const struct neighbour *, const void *pkey); 211 int (*constructor)(struct neighbour *); 212 int (*pconstructor)(struct pneigh_entry *); 213 void (*pdestructor)(struct pneigh_entry *); 214 void (*proxy_redo)(struct sk_buff *skb); 215 int (*is_multicast)(const void *pkey); 216 bool (*allow_add)(const struct net_device *dev, 217 struct netlink_ext_ack *extack); 218 char *id; 219 struct neigh_parms parms; 220 struct list_head parms_list; 221 int gc_interval; 222 int gc_thresh1; 223 int gc_thresh2; 224 int gc_thresh3; 225 unsigned long last_flush; 226 struct delayed_work gc_work; 227 struct delayed_work managed_work; 228 struct timer_list proxy_timer; 229 struct sk_buff_head proxy_queue; 230 atomic_t entries; 231 atomic_t gc_entries; 232 struct list_head gc_list; 233 struct list_head managed_list; 234 rwlock_t lock; 235 unsigned long last_rand; 236 struct neigh_statistics __percpu *stats; 237 struct neigh_hash_table __rcu *nht; 238 struct pneigh_entry **phash_buckets; 239 }; 240 241 static inline int neigh_parms_family(struct neigh_parms *p) 242 { 243 return p->tbl->family; 244 } 245 246 #define NEIGH_PRIV_ALIGN sizeof(long long) 247 #define NEIGH_ENTRY_SIZE(size) ALIGN((size), NEIGH_PRIV_ALIGN) 248 249 static inline void *neighbour_priv(const struct neighbour *n) 250 { 251 return (char *)n + n->tbl->entry_size; 252 } 253 254 /* flags for neigh_update() */ 255 #define NEIGH_UPDATE_F_OVERRIDE BIT(0) 256 #define NEIGH_UPDATE_F_WEAK_OVERRIDE BIT(1) 257 #define NEIGH_UPDATE_F_OVERRIDE_ISROUTER BIT(2) 258 #define NEIGH_UPDATE_F_USE BIT(3) 259 #define NEIGH_UPDATE_F_MANAGED BIT(4) 260 #define NEIGH_UPDATE_F_EXT_LEARNED BIT(5) 261 #define NEIGH_UPDATE_F_ISROUTER BIT(6) 262 #define NEIGH_UPDATE_F_ADMIN BIT(7) 263 264 /* In-kernel representation for NDA_FLAGS_EXT flags: */ 265 #define NTF_OLD_MASK 0xff 266 #define NTF_EXT_SHIFT 8 267 #define NTF_EXT_MASK (NTF_EXT_MANAGED) 268 269 #define NTF_MANAGED (NTF_EXT_MANAGED << NTF_EXT_SHIFT) 270 271 extern const struct nla_policy nda_policy[]; 272 273 #define neigh_for_each_in_bucket(pos, head) hlist_for_each_entry(pos, head, hash) 274 #define neigh_for_each_in_bucket_rcu(pos, head) \ 275 hlist_for_each_entry_rcu(pos, head, hash) 276 #define neigh_for_each_in_bucket_safe(pos, tmp, head) \ 277 hlist_for_each_entry_safe(pos, tmp, head, hash) 278 279 static inline bool neigh_key_eq32(const struct neighbour *n, const void *pkey) 280 { 281 return *(const u32 *)n->primary_key == *(const u32 *)pkey; 282 } 283 284 static inline bool neigh_key_eq128(const struct neighbour *n, const void *pkey) 285 { 286 const u32 *n32 = (const u32 *)n->primary_key; 287 const u32 *p32 = pkey; 288 289 return ((n32[0] ^ p32[0]) | (n32[1] ^ p32[1]) | 290 (n32[2] ^ p32[2]) | (n32[3] ^ p32[3])) == 0; 291 } 292 293 static inline struct neighbour *___neigh_lookup_noref( 294 struct neigh_table *tbl, 295 bool (*key_eq)(const struct neighbour *n, const void *pkey), 296 __u32 (*hash)(const void *pkey, 297 const struct net_device *dev, 298 __u32 *hash_rnd), 299 const void *pkey, 300 struct net_device *dev) 301 { 302 struct neigh_hash_table *nht = rcu_dereference(tbl->nht); 303 struct neighbour *n; 304 u32 hash_val; 305 306 hash_val = hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift); 307 neigh_for_each_in_bucket_rcu(n, &nht->hash_heads[hash_val]) 308 if (n->dev == dev && key_eq(n, pkey)) 309 return n; 310 311 return NULL; 312 } 313 314 static inline struct neighbour *__neigh_lookup_noref(struct neigh_table *tbl, 315 const void *pkey, 316 struct net_device *dev) 317 { 318 return ___neigh_lookup_noref(tbl, tbl->key_eq, tbl->hash, pkey, dev); 319 } 320 321 static inline void neigh_confirm(struct neighbour *n) 322 { 323 if (n) { 324 unsigned long now = jiffies; 325 326 /* avoid dirtying neighbour */ 327 if (READ_ONCE(n->confirmed) != now) 328 WRITE_ONCE(n->confirmed, now); 329 } 330 } 331 332 void neigh_table_init(int index, struct neigh_table *tbl); 333 int neigh_table_clear(int index, struct neigh_table *tbl); 334 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey, 335 struct net_device *dev); 336 struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey, 337 struct net_device *dev, bool want_ref); 338 static inline struct neighbour *neigh_create(struct neigh_table *tbl, 339 const void *pkey, 340 struct net_device *dev) 341 { 342 return __neigh_create(tbl, pkey, dev, true); 343 } 344 void neigh_destroy(struct neighbour *neigh); 345 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb, 346 const bool immediate_ok); 347 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, u32 flags, 348 u32 nlmsg_pid); 349 void __neigh_set_probe_once(struct neighbour *neigh); 350 bool neigh_remove_one(struct neighbour *ndel); 351 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev); 352 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev); 353 int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev); 354 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb); 355 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb); 356 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb); 357 struct neighbour *neigh_event_ns(struct neigh_table *tbl, 358 u8 *lladdr, void *saddr, 359 struct net_device *dev); 360 361 struct neigh_parms *neigh_parms_alloc(struct net_device *dev, 362 struct neigh_table *tbl); 363 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms); 364 365 static inline 366 struct net *neigh_parms_net(const struct neigh_parms *parms) 367 { 368 return read_pnet(&parms->net); 369 } 370 371 unsigned long neigh_rand_reach_time(unsigned long base); 372 373 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p, 374 struct sk_buff *skb); 375 struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net, 376 const void *key, struct net_device *dev, 377 int creat); 378 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl, struct net *net, 379 const void *key, struct net_device *dev); 380 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *key, 381 struct net_device *dev); 382 383 static inline struct net *pneigh_net(const struct pneigh_entry *pneigh) 384 { 385 return read_pnet(&pneigh->net); 386 } 387 388 void neigh_app_ns(struct neighbour *n); 389 void neigh_for_each(struct neigh_table *tbl, 390 void (*cb)(struct neighbour *, void *), void *cookie); 391 void __neigh_for_each_release(struct neigh_table *tbl, 392 int (*cb)(struct neighbour *)); 393 int neigh_xmit(int fam, struct net_device *, const void *, struct sk_buff *); 394 395 struct neigh_seq_state { 396 struct seq_net_private p; 397 struct neigh_table *tbl; 398 struct neigh_hash_table *nht; 399 void *(*neigh_sub_iter)(struct neigh_seq_state *state, 400 struct neighbour *n, loff_t *pos); 401 unsigned int bucket; 402 unsigned int flags; 403 #define NEIGH_SEQ_NEIGH_ONLY 0x00000001 404 #define NEIGH_SEQ_IS_PNEIGH 0x00000002 405 #define NEIGH_SEQ_SKIP_NOARP 0x00000004 406 }; 407 void *neigh_seq_start(struct seq_file *, loff_t *, struct neigh_table *, 408 unsigned int); 409 void *neigh_seq_next(struct seq_file *, void *, loff_t *); 410 void neigh_seq_stop(struct seq_file *, void *); 411 412 int neigh_proc_dointvec(const struct ctl_table *ctl, int write, 413 void *buffer, size_t *lenp, loff_t *ppos); 414 int neigh_proc_dointvec_jiffies(const struct ctl_table *ctl, int write, 415 void *buffer, 416 size_t *lenp, loff_t *ppos); 417 int neigh_proc_dointvec_ms_jiffies(const struct ctl_table *ctl, int write, 418 void *buffer, size_t *lenp, loff_t *ppos); 419 420 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p, 421 proc_handler *proc_handler); 422 void neigh_sysctl_unregister(struct neigh_parms *p); 423 424 static inline void __neigh_parms_put(struct neigh_parms *parms) 425 { 426 refcount_dec(&parms->refcnt); 427 } 428 429 static inline struct neigh_parms *neigh_parms_clone(struct neigh_parms *parms) 430 { 431 refcount_inc(&parms->refcnt); 432 return parms; 433 } 434 435 /* 436 * Neighbour references 437 */ 438 439 static inline void neigh_release(struct neighbour *neigh) 440 { 441 if (refcount_dec_and_test(&neigh->refcnt)) 442 neigh_destroy(neigh); 443 } 444 445 static inline struct neighbour * neigh_clone(struct neighbour *neigh) 446 { 447 if (neigh) 448 refcount_inc(&neigh->refcnt); 449 return neigh; 450 } 451 452 #define neigh_hold(n) refcount_inc(&(n)->refcnt) 453 454 static __always_inline int neigh_event_send_probe(struct neighbour *neigh, 455 struct sk_buff *skb, 456 const bool immediate_ok) 457 { 458 unsigned long now = jiffies; 459 460 if (READ_ONCE(neigh->used) != now) 461 WRITE_ONCE(neigh->used, now); 462 if (!(READ_ONCE(neigh->nud_state) & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))) 463 return __neigh_event_send(neigh, skb, immediate_ok); 464 return 0; 465 } 466 467 static inline int neigh_event_send(struct neighbour *neigh, struct sk_buff *skb) 468 { 469 return neigh_event_send_probe(neigh, skb, true); 470 } 471 472 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 473 static inline int neigh_hh_bridge(struct hh_cache *hh, struct sk_buff *skb) 474 { 475 unsigned int seq, hh_alen; 476 477 do { 478 seq = read_seqbegin(&hh->hh_lock); 479 hh_alen = HH_DATA_ALIGN(ETH_HLEN); 480 memcpy(skb->data - hh_alen, hh->hh_data, ETH_ALEN + hh_alen - ETH_HLEN); 481 } while (read_seqretry(&hh->hh_lock, seq)); 482 return 0; 483 } 484 #endif 485 486 static inline int neigh_hh_output(const struct hh_cache *hh, struct sk_buff *skb) 487 { 488 unsigned int hh_alen = 0; 489 unsigned int seq; 490 unsigned int hh_len; 491 492 do { 493 seq = read_seqbegin(&hh->hh_lock); 494 hh_len = READ_ONCE(hh->hh_len); 495 if (likely(hh_len <= HH_DATA_MOD)) { 496 hh_alen = HH_DATA_MOD; 497 498 /* skb_push() would proceed silently if we have room for 499 * the unaligned size but not for the aligned size: 500 * check headroom explicitly. 501 */ 502 if (likely(skb_headroom(skb) >= HH_DATA_MOD)) { 503 /* this is inlined by gcc */ 504 memcpy(skb->data - HH_DATA_MOD, hh->hh_data, 505 HH_DATA_MOD); 506 } 507 } else { 508 hh_alen = HH_DATA_ALIGN(hh_len); 509 510 if (likely(skb_headroom(skb) >= hh_alen)) { 511 memcpy(skb->data - hh_alen, hh->hh_data, 512 hh_alen); 513 } 514 } 515 } while (read_seqretry(&hh->hh_lock, seq)); 516 517 if (WARN_ON_ONCE(skb_headroom(skb) < hh_alen)) { 518 kfree_skb(skb); 519 return NET_XMIT_DROP; 520 } 521 522 __skb_push(skb, hh_len); 523 return dev_queue_xmit(skb); 524 } 525 526 static inline int neigh_output(struct neighbour *n, struct sk_buff *skb, 527 bool skip_cache) 528 { 529 const struct hh_cache *hh = &n->hh; 530 531 /* n->nud_state and hh->hh_len could be changed under us. 532 * neigh_hh_output() is taking care of the race later. 533 */ 534 if (!skip_cache && 535 (READ_ONCE(n->nud_state) & NUD_CONNECTED) && 536 READ_ONCE(hh->hh_len)) 537 return neigh_hh_output(hh, skb); 538 539 return READ_ONCE(n->output)(n, skb); 540 } 541 542 static inline struct neighbour * 543 __neigh_lookup(struct neigh_table *tbl, const void *pkey, struct net_device *dev, int creat) 544 { 545 struct neighbour *n = neigh_lookup(tbl, pkey, dev); 546 547 if (n || !creat) 548 return n; 549 550 n = neigh_create(tbl, pkey, dev); 551 return IS_ERR(n) ? NULL : n; 552 } 553 554 static inline struct neighbour * 555 __neigh_lookup_errno(struct neigh_table *tbl, const void *pkey, 556 struct net_device *dev) 557 { 558 struct neighbour *n = neigh_lookup(tbl, pkey, dev); 559 560 if (n) 561 return n; 562 563 return neigh_create(tbl, pkey, dev); 564 } 565 566 struct neighbour_cb { 567 unsigned long sched_next; 568 unsigned int flags; 569 }; 570 571 #define LOCALLY_ENQUEUED 0x1 572 573 #define NEIGH_CB(skb) ((struct neighbour_cb *)(skb)->cb) 574 575 static inline void neigh_ha_snapshot(char *dst, const struct neighbour *n, 576 const struct net_device *dev) 577 { 578 unsigned int seq; 579 580 do { 581 seq = read_seqbegin(&n->ha_lock); 582 memcpy(dst, n->ha, dev->addr_len); 583 } while (read_seqretry(&n->ha_lock, seq)); 584 } 585 586 static inline void neigh_update_is_router(struct neighbour *neigh, u32 flags, 587 int *notify) 588 { 589 u8 ndm_flags = 0; 590 591 ndm_flags |= (flags & NEIGH_UPDATE_F_ISROUTER) ? NTF_ROUTER : 0; 592 if ((neigh->flags ^ ndm_flags) & NTF_ROUTER) { 593 if (ndm_flags & NTF_ROUTER) 594 neigh->flags |= NTF_ROUTER; 595 else 596 neigh->flags &= ~NTF_ROUTER; 597 *notify = 1; 598 } 599 } 600 #endif 601