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