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 33 /* 34 * NUD stands for "neighbor unreachability detection" 35 */ 36 37 #define NUD_IN_TIMER (NUD_INCOMPLETE|NUD_REACHABLE|NUD_DELAY|NUD_PROBE) 38 #define NUD_VALID (NUD_PERMANENT|NUD_NOARP|NUD_REACHABLE|NUD_PROBE|NUD_STALE|NUD_DELAY) 39 #define NUD_CONNECTED (NUD_PERMANENT|NUD_NOARP|NUD_REACHABLE) 40 41 struct neighbour; 42 43 enum { 44 NEIGH_VAR_MCAST_PROBES, 45 NEIGH_VAR_UCAST_PROBES, 46 NEIGH_VAR_APP_PROBES, 47 NEIGH_VAR_MCAST_REPROBES, 48 NEIGH_VAR_RETRANS_TIME, 49 NEIGH_VAR_BASE_REACHABLE_TIME, 50 NEIGH_VAR_DELAY_PROBE_TIME, 51 NEIGH_VAR_GC_STALETIME, 52 NEIGH_VAR_QUEUE_LEN_BYTES, 53 NEIGH_VAR_PROXY_QLEN, 54 NEIGH_VAR_ANYCAST_DELAY, 55 NEIGH_VAR_PROXY_DELAY, 56 NEIGH_VAR_LOCKTIME, 57 #define NEIGH_VAR_DATA_MAX (NEIGH_VAR_LOCKTIME + 1) 58 /* Following are used as a second way to access one of the above */ 59 NEIGH_VAR_QUEUE_LEN, /* same data as NEIGH_VAR_QUEUE_LEN_BYTES */ 60 NEIGH_VAR_RETRANS_TIME_MS, /* same data as NEIGH_VAR_RETRANS_TIME */ 61 NEIGH_VAR_BASE_REACHABLE_TIME_MS, /* same data as NEIGH_VAR_BASE_REACHABLE_TIME */ 62 /* Following are used by "default" only */ 63 NEIGH_VAR_GC_INTERVAL, 64 NEIGH_VAR_GC_THRESH1, 65 NEIGH_VAR_GC_THRESH2, 66 NEIGH_VAR_GC_THRESH3, 67 NEIGH_VAR_MAX 68 }; 69 70 struct neigh_parms { 71 possible_net_t net; 72 struct net_device *dev; 73 netdevice_tracker dev_tracker; 74 struct list_head list; 75 int (*neigh_setup)(struct neighbour *); 76 struct neigh_table *tbl; 77 78 void *sysctl_table; 79 80 int dead; 81 refcount_t refcnt; 82 struct rcu_head rcu_head; 83 84 int reachable_time; 85 int data[NEIGH_VAR_DATA_MAX]; 86 DECLARE_BITMAP(data_state, NEIGH_VAR_DATA_MAX); 87 }; 88 89 static inline void neigh_var_set(struct neigh_parms *p, int index, int val) 90 { 91 set_bit(index, p->data_state); 92 p->data[index] = val; 93 } 94 95 #define NEIGH_VAR(p, attr) ((p)->data[NEIGH_VAR_ ## attr]) 96 97 /* In ndo_neigh_setup, NEIGH_VAR_INIT should be used. 98 * In other cases, NEIGH_VAR_SET should be used. 99 */ 100 #define NEIGH_VAR_INIT(p, attr, val) (NEIGH_VAR(p, attr) = val) 101 #define NEIGH_VAR_SET(p, attr, val) neigh_var_set(p, NEIGH_VAR_ ## attr, val) 102 103 static inline void neigh_parms_data_state_setall(struct neigh_parms *p) 104 { 105 bitmap_fill(p->data_state, NEIGH_VAR_DATA_MAX); 106 } 107 108 static inline void neigh_parms_data_state_cleanall(struct neigh_parms *p) 109 { 110 bitmap_zero(p->data_state, NEIGH_VAR_DATA_MAX); 111 } 112 113 struct neigh_statistics { 114 unsigned long allocs; /* number of allocated neighs */ 115 unsigned long destroys; /* number of destroyed neighs */ 116 unsigned long hash_grows; /* number of hash resizes */ 117 118 unsigned long res_failed; /* number of failed resolutions */ 119 120 unsigned long lookups; /* number of lookups */ 121 unsigned long hits; /* number of hits (among lookups) */ 122 123 unsigned long rcv_probes_mcast; /* number of received mcast ipv6 */ 124 unsigned long rcv_probes_ucast; /* number of received ucast ipv6 */ 125 126 unsigned long periodic_gc_runs; /* number of periodic GC runs */ 127 unsigned long forced_gc_runs; /* number of forced GC runs */ 128 129 unsigned long unres_discards; /* number of unresolved drops */ 130 unsigned long table_fulls; /* times even gc couldn't help */ 131 }; 132 133 #define NEIGH_CACHE_STAT_INC(tbl, field) this_cpu_inc((tbl)->stats->field) 134 135 struct neighbour { 136 struct neighbour __rcu *next; 137 struct neigh_table *tbl; 138 struct neigh_parms *parms; 139 unsigned long confirmed; 140 unsigned long updated; 141 rwlock_t lock; 142 refcount_t refcnt; 143 unsigned int arp_queue_len_bytes; 144 struct sk_buff_head arp_queue; 145 struct timer_list timer; 146 unsigned long used; 147 atomic_t probes; 148 u8 nud_state; 149 u8 type; 150 u8 dead; 151 u8 protocol; 152 u32 flags; 153 seqlock_t ha_lock; 154 unsigned char ha[ALIGN(MAX_ADDR_LEN, sizeof(unsigned long))] __aligned(8); 155 struct hh_cache hh; 156 int (*output)(struct neighbour *, struct sk_buff *); 157 const struct neigh_ops *ops; 158 struct list_head gc_list; 159 struct list_head managed_list; 160 struct rcu_head rcu; 161 struct net_device *dev; 162 netdevice_tracker dev_tracker; 163 u8 primary_key[0]; 164 } __randomize_layout; 165 166 struct neigh_ops { 167 int family; 168 void (*solicit)(struct neighbour *, struct sk_buff *); 169 void (*error_report)(struct neighbour *, struct sk_buff *); 170 int (*output)(struct neighbour *, struct sk_buff *); 171 int (*connected_output)(struct neighbour *, struct sk_buff *); 172 }; 173 174 struct pneigh_entry { 175 struct pneigh_entry *next; 176 possible_net_t net; 177 struct net_device *dev; 178 netdevice_tracker dev_tracker; 179 u32 flags; 180 u8 protocol; 181 u8 key[]; 182 }; 183 184 /* 185 * neighbour table manipulation 186 */ 187 188 #define NEIGH_NUM_HASH_RND 4 189 190 struct neigh_hash_table { 191 struct neighbour __rcu **hash_buckets; 192 unsigned int hash_shift; 193 __u32 hash_rnd[NEIGH_NUM_HASH_RND]; 194 struct rcu_head rcu; 195 }; 196 197 198 struct neigh_table { 199 int family; 200 unsigned int entry_size; 201 unsigned int key_len; 202 __be16 protocol; 203 __u32 (*hash)(const void *pkey, 204 const struct net_device *dev, 205 __u32 *hash_rnd); 206 bool (*key_eq)(const struct neighbour *, const void *pkey); 207 int (*constructor)(struct neighbour *); 208 int (*pconstructor)(struct pneigh_entry *); 209 void (*pdestructor)(struct pneigh_entry *); 210 void (*proxy_redo)(struct sk_buff *skb); 211 int (*is_multicast)(const void *pkey); 212 bool (*allow_add)(const struct net_device *dev, 213 struct netlink_ext_ack *extack); 214 char *id; 215 struct neigh_parms parms; 216 struct list_head parms_list; 217 int gc_interval; 218 int gc_thresh1; 219 int gc_thresh2; 220 int gc_thresh3; 221 unsigned long last_flush; 222 struct delayed_work gc_work; 223 struct delayed_work managed_work; 224 struct timer_list proxy_timer; 225 struct sk_buff_head proxy_queue; 226 atomic_t entries; 227 atomic_t gc_entries; 228 struct list_head gc_list; 229 struct list_head managed_list; 230 rwlock_t lock; 231 unsigned long last_rand; 232 struct neigh_statistics __percpu *stats; 233 struct neigh_hash_table __rcu *nht; 234 struct pneigh_entry **phash_buckets; 235 }; 236 237 enum { 238 NEIGH_ARP_TABLE = 0, 239 NEIGH_ND_TABLE = 1, 240 NEIGH_DN_TABLE = 2, 241 NEIGH_NR_TABLES, 242 NEIGH_LINK_TABLE = NEIGH_NR_TABLES /* Pseudo table for neigh_xmit */ 243 }; 244 245 static inline int neigh_parms_family(struct neigh_parms *p) 246 { 247 return p->tbl->family; 248 } 249 250 #define NEIGH_PRIV_ALIGN sizeof(long long) 251 #define NEIGH_ENTRY_SIZE(size) ALIGN((size), NEIGH_PRIV_ALIGN) 252 253 static inline void *neighbour_priv(const struct neighbour *n) 254 { 255 return (char *)n + n->tbl->entry_size; 256 } 257 258 /* flags for neigh_update() */ 259 #define NEIGH_UPDATE_F_OVERRIDE BIT(0) 260 #define NEIGH_UPDATE_F_WEAK_OVERRIDE BIT(1) 261 #define NEIGH_UPDATE_F_OVERRIDE_ISROUTER BIT(2) 262 #define NEIGH_UPDATE_F_USE BIT(3) 263 #define NEIGH_UPDATE_F_MANAGED BIT(4) 264 #define NEIGH_UPDATE_F_EXT_LEARNED BIT(5) 265 #define NEIGH_UPDATE_F_ISROUTER BIT(6) 266 #define NEIGH_UPDATE_F_ADMIN BIT(7) 267 268 /* In-kernel representation for NDA_FLAGS_EXT flags: */ 269 #define NTF_OLD_MASK 0xff 270 #define NTF_EXT_SHIFT 8 271 #define NTF_EXT_MASK (NTF_EXT_MANAGED) 272 273 #define NTF_MANAGED (NTF_EXT_MANAGED << NTF_EXT_SHIFT) 274 275 extern const struct nla_policy nda_policy[]; 276 277 static inline bool neigh_key_eq16(const struct neighbour *n, const void *pkey) 278 { 279 return *(const u16 *)n->primary_key == *(const u16 *)pkey; 280 } 281 282 static inline bool neigh_key_eq32(const struct neighbour *n, const void *pkey) 283 { 284 return *(const u32 *)n->primary_key == *(const u32 *)pkey; 285 } 286 287 static inline bool neigh_key_eq128(const struct neighbour *n, const void *pkey) 288 { 289 const u32 *n32 = (const u32 *)n->primary_key; 290 const u32 *p32 = pkey; 291 292 return ((n32[0] ^ p32[0]) | (n32[1] ^ p32[1]) | 293 (n32[2] ^ p32[2]) | (n32[3] ^ p32[3])) == 0; 294 } 295 296 static inline struct neighbour *___neigh_lookup_noref( 297 struct neigh_table *tbl, 298 bool (*key_eq)(const struct neighbour *n, const void *pkey), 299 __u32 (*hash)(const void *pkey, 300 const struct net_device *dev, 301 __u32 *hash_rnd), 302 const void *pkey, 303 struct net_device *dev) 304 { 305 struct neigh_hash_table *nht = rcu_dereference_bh(tbl->nht); 306 struct neighbour *n; 307 u32 hash_val; 308 309 hash_val = hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift); 310 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]); 311 n != NULL; 312 n = rcu_dereference_bh(n->next)) { 313 if (n->dev == dev && key_eq(n, pkey)) 314 return n; 315 } 316 317 return NULL; 318 } 319 320 static inline struct neighbour *__neigh_lookup_noref(struct neigh_table *tbl, 321 const void *pkey, 322 struct net_device *dev) 323 { 324 return ___neigh_lookup_noref(tbl, tbl->key_eq, tbl->hash, pkey, dev); 325 } 326 327 static inline void neigh_confirm(struct neighbour *n) 328 { 329 if (n) { 330 unsigned long now = jiffies; 331 332 /* avoid dirtying neighbour */ 333 if (READ_ONCE(n->confirmed) != now) 334 WRITE_ONCE(n->confirmed, now); 335 } 336 } 337 338 void neigh_table_init(int index, struct neigh_table *tbl); 339 int neigh_table_clear(int index, struct neigh_table *tbl); 340 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey, 341 struct net_device *dev); 342 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net, 343 const void *pkey); 344 struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey, 345 struct net_device *dev, bool want_ref); 346 static inline struct neighbour *neigh_create(struct neigh_table *tbl, 347 const void *pkey, 348 struct net_device *dev) 349 { 350 return __neigh_create(tbl, pkey, dev, true); 351 } 352 void neigh_destroy(struct neighbour *neigh); 353 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb, 354 const bool immediate_ok); 355 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, u32 flags, 356 u32 nlmsg_pid); 357 void __neigh_set_probe_once(struct neighbour *neigh); 358 bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl); 359 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev); 360 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev); 361 int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev); 362 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb); 363 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb); 364 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb); 365 struct neighbour *neigh_event_ns(struct neigh_table *tbl, 366 u8 *lladdr, void *saddr, 367 struct net_device *dev); 368 369 struct neigh_parms *neigh_parms_alloc(struct net_device *dev, 370 struct neigh_table *tbl); 371 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms); 372 373 static inline 374 struct net *neigh_parms_net(const struct neigh_parms *parms) 375 { 376 return read_pnet(&parms->net); 377 } 378 379 unsigned long neigh_rand_reach_time(unsigned long base); 380 381 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p, 382 struct sk_buff *skb); 383 struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net, 384 const void *key, struct net_device *dev, 385 int creat); 386 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl, struct net *net, 387 const void *key, struct net_device *dev); 388 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *key, 389 struct net_device *dev); 390 391 static inline struct net *pneigh_net(const struct pneigh_entry *pneigh) 392 { 393 return read_pnet(&pneigh->net); 394 } 395 396 void neigh_app_ns(struct neighbour *n); 397 void neigh_for_each(struct neigh_table *tbl, 398 void (*cb)(struct neighbour *, void *), void *cookie); 399 void __neigh_for_each_release(struct neigh_table *tbl, 400 int (*cb)(struct neighbour *)); 401 int neigh_xmit(int fam, struct net_device *, const void *, struct sk_buff *); 402 void pneigh_for_each(struct neigh_table *tbl, 403 void (*cb)(struct pneigh_entry *)); 404 405 struct neigh_seq_state { 406 struct seq_net_private p; 407 struct neigh_table *tbl; 408 struct neigh_hash_table *nht; 409 void *(*neigh_sub_iter)(struct neigh_seq_state *state, 410 struct neighbour *n, loff_t *pos); 411 unsigned int bucket; 412 unsigned int flags; 413 #define NEIGH_SEQ_NEIGH_ONLY 0x00000001 414 #define NEIGH_SEQ_IS_PNEIGH 0x00000002 415 #define NEIGH_SEQ_SKIP_NOARP 0x00000004 416 }; 417 void *neigh_seq_start(struct seq_file *, loff_t *, struct neigh_table *, 418 unsigned int); 419 void *neigh_seq_next(struct seq_file *, void *, loff_t *); 420 void neigh_seq_stop(struct seq_file *, void *); 421 422 int neigh_proc_dointvec(struct ctl_table *ctl, int write, 423 void *buffer, size_t *lenp, loff_t *ppos); 424 int neigh_proc_dointvec_jiffies(struct ctl_table *ctl, int write, 425 void *buffer, 426 size_t *lenp, loff_t *ppos); 427 int neigh_proc_dointvec_ms_jiffies(struct ctl_table *ctl, int write, 428 void *buffer, size_t *lenp, loff_t *ppos); 429 430 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p, 431 proc_handler *proc_handler); 432 void neigh_sysctl_unregister(struct neigh_parms *p); 433 434 static inline void __neigh_parms_put(struct neigh_parms *parms) 435 { 436 refcount_dec(&parms->refcnt); 437 } 438 439 static inline struct neigh_parms *neigh_parms_clone(struct neigh_parms *parms) 440 { 441 refcount_inc(&parms->refcnt); 442 return parms; 443 } 444 445 /* 446 * Neighbour references 447 */ 448 449 static inline void neigh_release(struct neighbour *neigh) 450 { 451 if (refcount_dec_and_test(&neigh->refcnt)) 452 neigh_destroy(neigh); 453 } 454 455 static inline struct neighbour * neigh_clone(struct neighbour *neigh) 456 { 457 if (neigh) 458 refcount_inc(&neigh->refcnt); 459 return neigh; 460 } 461 462 #define neigh_hold(n) refcount_inc(&(n)->refcnt) 463 464 static __always_inline int neigh_event_send_probe(struct neighbour *neigh, 465 struct sk_buff *skb, 466 const bool immediate_ok) 467 { 468 unsigned long now = jiffies; 469 470 if (READ_ONCE(neigh->used) != now) 471 WRITE_ONCE(neigh->used, now); 472 if (!(neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))) 473 return __neigh_event_send(neigh, skb, immediate_ok); 474 return 0; 475 } 476 477 static inline int neigh_event_send(struct neighbour *neigh, struct sk_buff *skb) 478 { 479 return neigh_event_send_probe(neigh, skb, true); 480 } 481 482 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 483 static inline int neigh_hh_bridge(struct hh_cache *hh, struct sk_buff *skb) 484 { 485 unsigned int seq, hh_alen; 486 487 do { 488 seq = read_seqbegin(&hh->hh_lock); 489 hh_alen = HH_DATA_ALIGN(ETH_HLEN); 490 memcpy(skb->data - hh_alen, hh->hh_data, ETH_ALEN + hh_alen - ETH_HLEN); 491 } while (read_seqretry(&hh->hh_lock, seq)); 492 return 0; 493 } 494 #endif 495 496 static inline int neigh_hh_output(const struct hh_cache *hh, struct sk_buff *skb) 497 { 498 unsigned int hh_alen = 0; 499 unsigned int seq; 500 unsigned int hh_len; 501 502 do { 503 seq = read_seqbegin(&hh->hh_lock); 504 hh_len = READ_ONCE(hh->hh_len); 505 if (likely(hh_len <= HH_DATA_MOD)) { 506 hh_alen = HH_DATA_MOD; 507 508 /* skb_push() would proceed silently if we have room for 509 * the unaligned size but not for the aligned size: 510 * check headroom explicitly. 511 */ 512 if (likely(skb_headroom(skb) >= HH_DATA_MOD)) { 513 /* this is inlined by gcc */ 514 memcpy(skb->data - HH_DATA_MOD, hh->hh_data, 515 HH_DATA_MOD); 516 } 517 } else { 518 hh_alen = HH_DATA_ALIGN(hh_len); 519 520 if (likely(skb_headroom(skb) >= hh_alen)) { 521 memcpy(skb->data - hh_alen, hh->hh_data, 522 hh_alen); 523 } 524 } 525 } while (read_seqretry(&hh->hh_lock, seq)); 526 527 if (WARN_ON_ONCE(skb_headroom(skb) < hh_alen)) { 528 kfree_skb(skb); 529 return NET_XMIT_DROP; 530 } 531 532 __skb_push(skb, hh_len); 533 return dev_queue_xmit(skb); 534 } 535 536 static inline int neigh_output(struct neighbour *n, struct sk_buff *skb, 537 bool skip_cache) 538 { 539 const struct hh_cache *hh = &n->hh; 540 541 /* n->nud_state and hh->hh_len could be changed under us. 542 * neigh_hh_output() is taking care of the race later. 543 */ 544 if (!skip_cache && 545 (READ_ONCE(n->nud_state) & NUD_CONNECTED) && 546 READ_ONCE(hh->hh_len)) 547 return neigh_hh_output(hh, skb); 548 549 return n->output(n, skb); 550 } 551 552 static inline struct neighbour * 553 __neigh_lookup(struct neigh_table *tbl, const void *pkey, struct net_device *dev, int creat) 554 { 555 struct neighbour *n = neigh_lookup(tbl, pkey, dev); 556 557 if (n || !creat) 558 return n; 559 560 n = neigh_create(tbl, pkey, dev); 561 return IS_ERR(n) ? NULL : n; 562 } 563 564 static inline struct neighbour * 565 __neigh_lookup_errno(struct neigh_table *tbl, const void *pkey, 566 struct net_device *dev) 567 { 568 struct neighbour *n = neigh_lookup(tbl, pkey, dev); 569 570 if (n) 571 return n; 572 573 return neigh_create(tbl, pkey, dev); 574 } 575 576 struct neighbour_cb { 577 unsigned long sched_next; 578 unsigned int flags; 579 }; 580 581 #define LOCALLY_ENQUEUED 0x1 582 583 #define NEIGH_CB(skb) ((struct neighbour_cb *)(skb)->cb) 584 585 static inline void neigh_ha_snapshot(char *dst, const struct neighbour *n, 586 const struct net_device *dev) 587 { 588 unsigned int seq; 589 590 do { 591 seq = read_seqbegin(&n->ha_lock); 592 memcpy(dst, n->ha, dev->addr_len); 593 } while (read_seqretry(&n->ha_lock, seq)); 594 } 595 596 static inline void neigh_update_is_router(struct neighbour *neigh, u32 flags, 597 int *notify) 598 { 599 u8 ndm_flags = 0; 600 601 ndm_flags |= (flags & NEIGH_UPDATE_F_ISROUTER) ? NTF_ROUTER : 0; 602 if ((neigh->flags ^ ndm_flags) & NTF_ROUTER) { 603 if (ndm_flags & NTF_ROUTER) 604 neigh->flags |= NTF_ROUTER; 605 else 606 neigh->flags &= ~NTF_ROUTER; 607 *notify = 1; 608 } 609 } 610 #endif 611