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