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 int data[NEIGH_VAR_DATA_MAX]; 87 DECLARE_BITMAP(data_state, NEIGH_VAR_DATA_MAX); 88 }; 89 90 static inline void neigh_var_set(struct neigh_parms *p, int index, int val) 91 { 92 set_bit(index, p->data_state); 93 p->data[index] = val; 94 } 95 96 #define NEIGH_VAR(p, attr) ((p)->data[NEIGH_VAR_ ## attr]) 97 98 /* In ndo_neigh_setup, NEIGH_VAR_INIT should be used. 99 * In other cases, NEIGH_VAR_SET should be used. 100 */ 101 #define NEIGH_VAR_INIT(p, attr, val) (NEIGH_VAR(p, attr) = val) 102 #define NEIGH_VAR_SET(p, attr, val) neigh_var_set(p, NEIGH_VAR_ ## attr, val) 103 104 static inline void neigh_parms_data_state_setall(struct neigh_parms *p) 105 { 106 bitmap_fill(p->data_state, NEIGH_VAR_DATA_MAX); 107 } 108 109 static inline void neigh_parms_data_state_cleanall(struct neigh_parms *p) 110 { 111 bitmap_zero(p->data_state, NEIGH_VAR_DATA_MAX); 112 } 113 114 struct neigh_statistics { 115 unsigned long allocs; /* number of allocated neighs */ 116 unsigned long destroys; /* number of destroyed neighs */ 117 unsigned long hash_grows; /* number of hash resizes */ 118 119 unsigned long res_failed; /* number of failed resolutions */ 120 121 unsigned long lookups; /* number of lookups */ 122 unsigned long hits; /* number of hits (among lookups) */ 123 124 unsigned long rcv_probes_mcast; /* number of received mcast ipv6 */ 125 unsigned long rcv_probes_ucast; /* number of received ucast ipv6 */ 126 127 unsigned long periodic_gc_runs; /* number of periodic GC runs */ 128 unsigned long forced_gc_runs; /* number of forced GC runs */ 129 130 unsigned long unres_discards; /* number of unresolved drops */ 131 unsigned long table_fulls; /* times even gc couldn't help */ 132 }; 133 134 #define NEIGH_CACHE_STAT_INC(tbl, field) this_cpu_inc((tbl)->stats->field) 135 136 struct neighbour { 137 struct neighbour __rcu *next; 138 struct neigh_table *tbl; 139 struct neigh_parms *parms; 140 unsigned long confirmed; 141 unsigned long updated; 142 rwlock_t lock; 143 refcount_t refcnt; 144 unsigned int arp_queue_len_bytes; 145 struct sk_buff_head arp_queue; 146 struct timer_list timer; 147 unsigned long used; 148 atomic_t probes; 149 u8 nud_state; 150 u8 type; 151 u8 dead; 152 u8 protocol; 153 u32 flags; 154 seqlock_t ha_lock; 155 unsigned char ha[ALIGN(MAX_ADDR_LEN, sizeof(unsigned long))] __aligned(8); 156 struct hh_cache hh; 157 int (*output)(struct neighbour *, struct sk_buff *); 158 const struct neigh_ops *ops; 159 struct list_head gc_list; 160 struct list_head managed_list; 161 struct rcu_head rcu; 162 struct net_device *dev; 163 netdevice_tracker dev_tracker; 164 u8 primary_key[0]; 165 } __randomize_layout; 166 167 struct neigh_ops { 168 int family; 169 void (*solicit)(struct neighbour *, struct sk_buff *); 170 void (*error_report)(struct neighbour *, struct sk_buff *); 171 int (*output)(struct neighbour *, struct sk_buff *); 172 int (*connected_output)(struct neighbour *, struct sk_buff *); 173 }; 174 175 struct pneigh_entry { 176 struct pneigh_entry *next; 177 possible_net_t net; 178 struct net_device *dev; 179 netdevice_tracker dev_tracker; 180 u32 flags; 181 u8 protocol; 182 u8 key[]; 183 }; 184 185 /* 186 * neighbour table manipulation 187 */ 188 189 #define NEIGH_NUM_HASH_RND 4 190 191 struct neigh_hash_table { 192 struct neighbour __rcu **hash_buckets; 193 unsigned int hash_shift; 194 __u32 hash_rnd[NEIGH_NUM_HASH_RND]; 195 struct rcu_head rcu; 196 }; 197 198 199 struct neigh_table { 200 int family; 201 unsigned int entry_size; 202 unsigned int key_len; 203 __be16 protocol; 204 __u32 (*hash)(const void *pkey, 205 const struct net_device *dev, 206 __u32 *hash_rnd); 207 bool (*key_eq)(const struct neighbour *, const void *pkey); 208 int (*constructor)(struct neighbour *); 209 int (*pconstructor)(struct pneigh_entry *); 210 void (*pdestructor)(struct pneigh_entry *); 211 void (*proxy_redo)(struct sk_buff *skb); 212 int (*is_multicast)(const void *pkey); 213 bool (*allow_add)(const struct net_device *dev, 214 struct netlink_ext_ack *extack); 215 char *id; 216 struct neigh_parms parms; 217 struct list_head parms_list; 218 int gc_interval; 219 int gc_thresh1; 220 int gc_thresh2; 221 int gc_thresh3; 222 unsigned long last_flush; 223 struct delayed_work gc_work; 224 struct delayed_work managed_work; 225 struct timer_list proxy_timer; 226 struct sk_buff_head proxy_queue; 227 atomic_t entries; 228 atomic_t gc_entries; 229 struct list_head gc_list; 230 struct list_head managed_list; 231 rwlock_t lock; 232 unsigned long last_rand; 233 struct neigh_statistics __percpu *stats; 234 struct neigh_hash_table __rcu *nht; 235 struct pneigh_entry **phash_buckets; 236 }; 237 238 enum { 239 NEIGH_ARP_TABLE = 0, 240 NEIGH_ND_TABLE = 1, 241 NEIGH_DN_TABLE = 2, 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_eq16(const struct neighbour *n, const void *pkey) 279 { 280 return *(const u16 *)n->primary_key == *(const u16 *)pkey; 281 } 282 283 static inline bool neigh_key_eq32(const struct neighbour *n, const void *pkey) 284 { 285 return *(const u32 *)n->primary_key == *(const u32 *)pkey; 286 } 287 288 static inline bool neigh_key_eq128(const struct neighbour *n, const void *pkey) 289 { 290 const u32 *n32 = (const u32 *)n->primary_key; 291 const u32 *p32 = pkey; 292 293 return ((n32[0] ^ p32[0]) | (n32[1] ^ p32[1]) | 294 (n32[2] ^ p32[2]) | (n32[3] ^ p32[3])) == 0; 295 } 296 297 static inline struct neighbour *___neigh_lookup_noref( 298 struct neigh_table *tbl, 299 bool (*key_eq)(const struct neighbour *n, const void *pkey), 300 __u32 (*hash)(const void *pkey, 301 const struct net_device *dev, 302 __u32 *hash_rnd), 303 const void *pkey, 304 struct net_device *dev) 305 { 306 struct neigh_hash_table *nht = rcu_dereference_bh(tbl->nht); 307 struct neighbour *n; 308 u32 hash_val; 309 310 hash_val = hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift); 311 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]); 312 n != NULL; 313 n = rcu_dereference_bh(n->next)) { 314 if (n->dev == dev && key_eq(n, pkey)) 315 return n; 316 } 317 318 return NULL; 319 } 320 321 static inline struct neighbour *__neigh_lookup_noref(struct neigh_table *tbl, 322 const void *pkey, 323 struct net_device *dev) 324 { 325 return ___neigh_lookup_noref(tbl, tbl->key_eq, tbl->hash, pkey, dev); 326 } 327 328 static inline void neigh_confirm(struct neighbour *n) 329 { 330 if (n) { 331 unsigned long now = jiffies; 332 333 /* avoid dirtying neighbour */ 334 if (READ_ONCE(n->confirmed) != now) 335 WRITE_ONCE(n->confirmed, now); 336 } 337 } 338 339 void neigh_table_init(int index, struct neigh_table *tbl); 340 int neigh_table_clear(int index, struct neigh_table *tbl); 341 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey, 342 struct net_device *dev); 343 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net, 344 const void *pkey); 345 struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey, 346 struct net_device *dev, bool want_ref); 347 static inline struct neighbour *neigh_create(struct neigh_table *tbl, 348 const void *pkey, 349 struct net_device *dev) 350 { 351 return __neigh_create(tbl, pkey, dev, true); 352 } 353 void neigh_destroy(struct neighbour *neigh); 354 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb, 355 const bool immediate_ok); 356 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, u32 flags, 357 u32 nlmsg_pid); 358 void __neigh_set_probe_once(struct neighbour *neigh); 359 bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl); 360 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev); 361 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev); 362 int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev); 363 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb); 364 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb); 365 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb); 366 struct neighbour *neigh_event_ns(struct neigh_table *tbl, 367 u8 *lladdr, void *saddr, 368 struct net_device *dev); 369 370 struct neigh_parms *neigh_parms_alloc(struct net_device *dev, 371 struct neigh_table *tbl); 372 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms); 373 374 static inline 375 struct net *neigh_parms_net(const struct neigh_parms *parms) 376 { 377 return read_pnet(&parms->net); 378 } 379 380 unsigned long neigh_rand_reach_time(unsigned long base); 381 382 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p, 383 struct sk_buff *skb); 384 struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net, 385 const void *key, struct net_device *dev, 386 int creat); 387 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl, struct net *net, 388 const void *key, struct net_device *dev); 389 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *key, 390 struct net_device *dev); 391 392 static inline struct net *pneigh_net(const struct pneigh_entry *pneigh) 393 { 394 return read_pnet(&pneigh->net); 395 } 396 397 void neigh_app_ns(struct neighbour *n); 398 void neigh_for_each(struct neigh_table *tbl, 399 void (*cb)(struct neighbour *, void *), void *cookie); 400 void __neigh_for_each_release(struct neigh_table *tbl, 401 int (*cb)(struct neighbour *)); 402 int neigh_xmit(int fam, struct net_device *, const void *, struct sk_buff *); 403 void pneigh_for_each(struct neigh_table *tbl, 404 void (*cb)(struct pneigh_entry *)); 405 406 struct neigh_seq_state { 407 struct seq_net_private p; 408 struct neigh_table *tbl; 409 struct neigh_hash_table *nht; 410 void *(*neigh_sub_iter)(struct neigh_seq_state *state, 411 struct neighbour *n, loff_t *pos); 412 unsigned int bucket; 413 unsigned int flags; 414 #define NEIGH_SEQ_NEIGH_ONLY 0x00000001 415 #define NEIGH_SEQ_IS_PNEIGH 0x00000002 416 #define NEIGH_SEQ_SKIP_NOARP 0x00000004 417 }; 418 void *neigh_seq_start(struct seq_file *, loff_t *, struct neigh_table *, 419 unsigned int); 420 void *neigh_seq_next(struct seq_file *, void *, loff_t *); 421 void neigh_seq_stop(struct seq_file *, void *); 422 423 int neigh_proc_dointvec(struct ctl_table *ctl, int write, 424 void *buffer, size_t *lenp, loff_t *ppos); 425 int neigh_proc_dointvec_jiffies(struct ctl_table *ctl, int write, 426 void *buffer, 427 size_t *lenp, loff_t *ppos); 428 int neigh_proc_dointvec_ms_jiffies(struct ctl_table *ctl, int write, 429 void *buffer, size_t *lenp, loff_t *ppos); 430 431 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p, 432 proc_handler *proc_handler); 433 void neigh_sysctl_unregister(struct neigh_parms *p); 434 435 static inline void __neigh_parms_put(struct neigh_parms *parms) 436 { 437 refcount_dec(&parms->refcnt); 438 } 439 440 static inline struct neigh_parms *neigh_parms_clone(struct neigh_parms *parms) 441 { 442 refcount_inc(&parms->refcnt); 443 return parms; 444 } 445 446 /* 447 * Neighbour references 448 */ 449 450 static inline void neigh_release(struct neighbour *neigh) 451 { 452 if (refcount_dec_and_test(&neigh->refcnt)) 453 neigh_destroy(neigh); 454 } 455 456 static inline struct neighbour * neigh_clone(struct neighbour *neigh) 457 { 458 if (neigh) 459 refcount_inc(&neigh->refcnt); 460 return neigh; 461 } 462 463 #define neigh_hold(n) refcount_inc(&(n)->refcnt) 464 465 static __always_inline int neigh_event_send_probe(struct neighbour *neigh, 466 struct sk_buff *skb, 467 const bool immediate_ok) 468 { 469 unsigned long now = jiffies; 470 471 if (READ_ONCE(neigh->used) != now) 472 WRITE_ONCE(neigh->used, now); 473 if (!(neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))) 474 return __neigh_event_send(neigh, skb, immediate_ok); 475 return 0; 476 } 477 478 static inline int neigh_event_send(struct neighbour *neigh, struct sk_buff *skb) 479 { 480 return neigh_event_send_probe(neigh, skb, true); 481 } 482 483 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 484 static inline int neigh_hh_bridge(struct hh_cache *hh, struct sk_buff *skb) 485 { 486 unsigned int seq, hh_alen; 487 488 do { 489 seq = read_seqbegin(&hh->hh_lock); 490 hh_alen = HH_DATA_ALIGN(ETH_HLEN); 491 memcpy(skb->data - hh_alen, hh->hh_data, ETH_ALEN + hh_alen - ETH_HLEN); 492 } while (read_seqretry(&hh->hh_lock, seq)); 493 return 0; 494 } 495 #endif 496 497 static inline int neigh_hh_output(const struct hh_cache *hh, struct sk_buff *skb) 498 { 499 unsigned int hh_alen = 0; 500 unsigned int seq; 501 unsigned int hh_len; 502 503 do { 504 seq = read_seqbegin(&hh->hh_lock); 505 hh_len = READ_ONCE(hh->hh_len); 506 if (likely(hh_len <= HH_DATA_MOD)) { 507 hh_alen = HH_DATA_MOD; 508 509 /* skb_push() would proceed silently if we have room for 510 * the unaligned size but not for the aligned size: 511 * check headroom explicitly. 512 */ 513 if (likely(skb_headroom(skb) >= HH_DATA_MOD)) { 514 /* this is inlined by gcc */ 515 memcpy(skb->data - HH_DATA_MOD, hh->hh_data, 516 HH_DATA_MOD); 517 } 518 } else { 519 hh_alen = HH_DATA_ALIGN(hh_len); 520 521 if (likely(skb_headroom(skb) >= hh_alen)) { 522 memcpy(skb->data - hh_alen, hh->hh_data, 523 hh_alen); 524 } 525 } 526 } while (read_seqretry(&hh->hh_lock, seq)); 527 528 if (WARN_ON_ONCE(skb_headroom(skb) < hh_alen)) { 529 kfree_skb(skb); 530 return NET_XMIT_DROP; 531 } 532 533 __skb_push(skb, hh_len); 534 return dev_queue_xmit(skb); 535 } 536 537 static inline int neigh_output(struct neighbour *n, struct sk_buff *skb, 538 bool skip_cache) 539 { 540 const struct hh_cache *hh = &n->hh; 541 542 /* n->nud_state and hh->hh_len could be changed under us. 543 * neigh_hh_output() is taking care of the race later. 544 */ 545 if (!skip_cache && 546 (READ_ONCE(n->nud_state) & NUD_CONNECTED) && 547 READ_ONCE(hh->hh_len)) 548 return neigh_hh_output(hh, skb); 549 550 return n->output(n, skb); 551 } 552 553 static inline struct neighbour * 554 __neigh_lookup(struct neigh_table *tbl, const void *pkey, struct net_device *dev, int creat) 555 { 556 struct neighbour *n = neigh_lookup(tbl, pkey, dev); 557 558 if (n || !creat) 559 return n; 560 561 n = neigh_create(tbl, pkey, dev); 562 return IS_ERR(n) ? NULL : n; 563 } 564 565 static inline struct neighbour * 566 __neigh_lookup_errno(struct neigh_table *tbl, const void *pkey, 567 struct net_device *dev) 568 { 569 struct neighbour *n = neigh_lookup(tbl, pkey, dev); 570 571 if (n) 572 return n; 573 574 return neigh_create(tbl, pkey, dev); 575 } 576 577 struct neighbour_cb { 578 unsigned long sched_next; 579 unsigned int flags; 580 }; 581 582 #define LOCALLY_ENQUEUED 0x1 583 584 #define NEIGH_CB(skb) ((struct neighbour_cb *)(skb)->cb) 585 586 static inline void neigh_ha_snapshot(char *dst, const struct neighbour *n, 587 const struct net_device *dev) 588 { 589 unsigned int seq; 590 591 do { 592 seq = read_seqbegin(&n->ha_lock); 593 memcpy(dst, n->ha, dev->addr_len); 594 } while (read_seqretry(&n->ha_lock, seq)); 595 } 596 597 static inline void neigh_update_is_router(struct neighbour *neigh, u32 flags, 598 int *notify) 599 { 600 u8 ndm_flags = 0; 601 602 ndm_flags |= (flags & NEIGH_UPDATE_F_ISROUTER) ? NTF_ROUTER : 0; 603 if ((neigh->flags ^ ndm_flags) & NTF_ROUTER) { 604 if (ndm_flags & NTF_ROUTER) 605 neigh->flags |= NTF_ROUTER; 606 else 607 neigh->flags &= ~NTF_ROUTER; 608 *notify = 1; 609 } 610 } 611 #endif 612