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 u8 primary_key[0]; 162 } __randomize_layout; 163 164 struct neigh_ops { 165 int family; 166 void (*solicit)(struct neighbour *, struct sk_buff *); 167 void (*error_report)(struct neighbour *, struct sk_buff *); 168 int (*output)(struct neighbour *, struct sk_buff *); 169 int (*connected_output)(struct neighbour *, struct sk_buff *); 170 }; 171 172 struct pneigh_entry { 173 struct pneigh_entry *next; 174 possible_net_t net; 175 struct net_device *dev; 176 u32 flags; 177 u8 protocol; 178 u8 key[]; 179 }; 180 181 /* 182 * neighbour table manipulation 183 */ 184 185 #define NEIGH_NUM_HASH_RND 4 186 187 struct neigh_hash_table { 188 struct neighbour __rcu **hash_buckets; 189 unsigned int hash_shift; 190 __u32 hash_rnd[NEIGH_NUM_HASH_RND]; 191 struct rcu_head rcu; 192 }; 193 194 195 struct neigh_table { 196 int family; 197 unsigned int entry_size; 198 unsigned int key_len; 199 __be16 protocol; 200 __u32 (*hash)(const void *pkey, 201 const struct net_device *dev, 202 __u32 *hash_rnd); 203 bool (*key_eq)(const struct neighbour *, const void *pkey); 204 int (*constructor)(struct neighbour *); 205 int (*pconstructor)(struct pneigh_entry *); 206 void (*pdestructor)(struct pneigh_entry *); 207 void (*proxy_redo)(struct sk_buff *skb); 208 int (*is_multicast)(const void *pkey); 209 bool (*allow_add)(const struct net_device *dev, 210 struct netlink_ext_ack *extack); 211 char *id; 212 struct neigh_parms parms; 213 struct list_head parms_list; 214 int gc_interval; 215 int gc_thresh1; 216 int gc_thresh2; 217 int gc_thresh3; 218 unsigned long last_flush; 219 struct delayed_work gc_work; 220 struct delayed_work managed_work; 221 struct timer_list proxy_timer; 222 struct sk_buff_head proxy_queue; 223 atomic_t entries; 224 atomic_t gc_entries; 225 struct list_head gc_list; 226 struct list_head managed_list; 227 rwlock_t lock; 228 unsigned long last_rand; 229 struct neigh_statistics __percpu *stats; 230 struct neigh_hash_table __rcu *nht; 231 struct pneigh_entry **phash_buckets; 232 }; 233 234 enum { 235 NEIGH_ARP_TABLE = 0, 236 NEIGH_ND_TABLE = 1, 237 NEIGH_DN_TABLE = 2, 238 NEIGH_NR_TABLES, 239 NEIGH_LINK_TABLE = NEIGH_NR_TABLES /* Pseudo table for neigh_xmit */ 240 }; 241 242 static inline int neigh_parms_family(struct neigh_parms *p) 243 { 244 return p->tbl->family; 245 } 246 247 #define NEIGH_PRIV_ALIGN sizeof(long long) 248 #define NEIGH_ENTRY_SIZE(size) ALIGN((size), NEIGH_PRIV_ALIGN) 249 250 static inline void *neighbour_priv(const struct neighbour *n) 251 { 252 return (char *)n + n->tbl->entry_size; 253 } 254 255 /* flags for neigh_update() */ 256 #define NEIGH_UPDATE_F_OVERRIDE BIT(0) 257 #define NEIGH_UPDATE_F_WEAK_OVERRIDE BIT(1) 258 #define NEIGH_UPDATE_F_OVERRIDE_ISROUTER BIT(2) 259 #define NEIGH_UPDATE_F_USE BIT(3) 260 #define NEIGH_UPDATE_F_MANAGED BIT(4) 261 #define NEIGH_UPDATE_F_EXT_LEARNED BIT(5) 262 #define NEIGH_UPDATE_F_ISROUTER BIT(6) 263 #define NEIGH_UPDATE_F_ADMIN BIT(7) 264 265 /* In-kernel representation for NDA_FLAGS_EXT flags: */ 266 #define NTF_OLD_MASK 0xff 267 #define NTF_EXT_SHIFT 8 268 #define NTF_EXT_MASK (NTF_EXT_MANAGED) 269 270 #define NTF_MANAGED (NTF_EXT_MANAGED << NTF_EXT_SHIFT) 271 272 extern const struct nla_policy nda_policy[]; 273 274 static inline bool neigh_key_eq16(const struct neighbour *n, const void *pkey) 275 { 276 return *(const u16 *)n->primary_key == *(const u16 *)pkey; 277 } 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_bh(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 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]); 308 n != NULL; 309 n = rcu_dereference_bh(n->next)) { 310 if (n->dev == dev && key_eq(n, pkey)) 311 return n; 312 } 313 314 return NULL; 315 } 316 317 static inline struct neighbour *__neigh_lookup_noref(struct neigh_table *tbl, 318 const void *pkey, 319 struct net_device *dev) 320 { 321 return ___neigh_lookup_noref(tbl, tbl->key_eq, tbl->hash, pkey, dev); 322 } 323 324 void neigh_table_init(int index, struct neigh_table *tbl); 325 int neigh_table_clear(int index, struct neigh_table *tbl); 326 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey, 327 struct net_device *dev); 328 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net, 329 const void *pkey); 330 struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey, 331 struct net_device *dev, bool want_ref); 332 static inline struct neighbour *neigh_create(struct neigh_table *tbl, 333 const void *pkey, 334 struct net_device *dev) 335 { 336 return __neigh_create(tbl, pkey, dev, true); 337 } 338 void neigh_destroy(struct neighbour *neigh); 339 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb); 340 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, u32 flags, 341 u32 nlmsg_pid); 342 void __neigh_set_probe_once(struct neighbour *neigh); 343 bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl); 344 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev); 345 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev); 346 int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev); 347 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb); 348 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb); 349 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb); 350 struct neighbour *neigh_event_ns(struct neigh_table *tbl, 351 u8 *lladdr, void *saddr, 352 struct net_device *dev); 353 354 struct neigh_parms *neigh_parms_alloc(struct net_device *dev, 355 struct neigh_table *tbl); 356 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms); 357 358 static inline 359 struct net *neigh_parms_net(const struct neigh_parms *parms) 360 { 361 return read_pnet(&parms->net); 362 } 363 364 unsigned long neigh_rand_reach_time(unsigned long base); 365 366 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p, 367 struct sk_buff *skb); 368 struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net, 369 const void *key, struct net_device *dev, 370 int creat); 371 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl, struct net *net, 372 const void *key, struct net_device *dev); 373 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *key, 374 struct net_device *dev); 375 376 static inline struct net *pneigh_net(const struct pneigh_entry *pneigh) 377 { 378 return read_pnet(&pneigh->net); 379 } 380 381 void neigh_app_ns(struct neighbour *n); 382 void neigh_for_each(struct neigh_table *tbl, 383 void (*cb)(struct neighbour *, void *), void *cookie); 384 void __neigh_for_each_release(struct neigh_table *tbl, 385 int (*cb)(struct neighbour *)); 386 int neigh_xmit(int fam, struct net_device *, const void *, struct sk_buff *); 387 void pneigh_for_each(struct neigh_table *tbl, 388 void (*cb)(struct pneigh_entry *)); 389 390 struct neigh_seq_state { 391 struct seq_net_private p; 392 struct neigh_table *tbl; 393 struct neigh_hash_table *nht; 394 void *(*neigh_sub_iter)(struct neigh_seq_state *state, 395 struct neighbour *n, loff_t *pos); 396 unsigned int bucket; 397 unsigned int flags; 398 #define NEIGH_SEQ_NEIGH_ONLY 0x00000001 399 #define NEIGH_SEQ_IS_PNEIGH 0x00000002 400 #define NEIGH_SEQ_SKIP_NOARP 0x00000004 401 }; 402 void *neigh_seq_start(struct seq_file *, loff_t *, struct neigh_table *, 403 unsigned int); 404 void *neigh_seq_next(struct seq_file *, void *, loff_t *); 405 void neigh_seq_stop(struct seq_file *, void *); 406 407 int neigh_proc_dointvec(struct ctl_table *ctl, int write, 408 void *buffer, size_t *lenp, loff_t *ppos); 409 int neigh_proc_dointvec_jiffies(struct ctl_table *ctl, int write, 410 void *buffer, 411 size_t *lenp, loff_t *ppos); 412 int neigh_proc_dointvec_ms_jiffies(struct ctl_table *ctl, int write, 413 void *buffer, size_t *lenp, loff_t *ppos); 414 415 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p, 416 proc_handler *proc_handler); 417 void neigh_sysctl_unregister(struct neigh_parms *p); 418 419 static inline void __neigh_parms_put(struct neigh_parms *parms) 420 { 421 refcount_dec(&parms->refcnt); 422 } 423 424 static inline struct neigh_parms *neigh_parms_clone(struct neigh_parms *parms) 425 { 426 refcount_inc(&parms->refcnt); 427 return parms; 428 } 429 430 /* 431 * Neighbour references 432 */ 433 434 static inline void neigh_release(struct neighbour *neigh) 435 { 436 if (refcount_dec_and_test(&neigh->refcnt)) 437 neigh_destroy(neigh); 438 } 439 440 static inline struct neighbour * neigh_clone(struct neighbour *neigh) 441 { 442 if (neigh) 443 refcount_inc(&neigh->refcnt); 444 return neigh; 445 } 446 447 #define neigh_hold(n) refcount_inc(&(n)->refcnt) 448 449 static inline int neigh_event_send(struct neighbour *neigh, struct sk_buff *skb) 450 { 451 unsigned long now = jiffies; 452 453 if (READ_ONCE(neigh->used) != now) 454 WRITE_ONCE(neigh->used, now); 455 if (!(neigh->nud_state&(NUD_CONNECTED|NUD_DELAY|NUD_PROBE))) 456 return __neigh_event_send(neigh, skb); 457 return 0; 458 } 459 460 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 461 static inline int neigh_hh_bridge(struct hh_cache *hh, struct sk_buff *skb) 462 { 463 unsigned int seq, hh_alen; 464 465 do { 466 seq = read_seqbegin(&hh->hh_lock); 467 hh_alen = HH_DATA_ALIGN(ETH_HLEN); 468 memcpy(skb->data - hh_alen, hh->hh_data, ETH_ALEN + hh_alen - ETH_HLEN); 469 } while (read_seqretry(&hh->hh_lock, seq)); 470 return 0; 471 } 472 #endif 473 474 static inline int neigh_hh_output(const struct hh_cache *hh, struct sk_buff *skb) 475 { 476 unsigned int hh_alen = 0; 477 unsigned int seq; 478 unsigned int hh_len; 479 480 do { 481 seq = read_seqbegin(&hh->hh_lock); 482 hh_len = READ_ONCE(hh->hh_len); 483 if (likely(hh_len <= HH_DATA_MOD)) { 484 hh_alen = HH_DATA_MOD; 485 486 /* skb_push() would proceed silently if we have room for 487 * the unaligned size but not for the aligned size: 488 * check headroom explicitly. 489 */ 490 if (likely(skb_headroom(skb) >= HH_DATA_MOD)) { 491 /* this is inlined by gcc */ 492 memcpy(skb->data - HH_DATA_MOD, hh->hh_data, 493 HH_DATA_MOD); 494 } 495 } else { 496 hh_alen = HH_DATA_ALIGN(hh_len); 497 498 if (likely(skb_headroom(skb) >= hh_alen)) { 499 memcpy(skb->data - hh_alen, hh->hh_data, 500 hh_alen); 501 } 502 } 503 } while (read_seqretry(&hh->hh_lock, seq)); 504 505 if (WARN_ON_ONCE(skb_headroom(skb) < hh_alen)) { 506 kfree_skb(skb); 507 return NET_XMIT_DROP; 508 } 509 510 __skb_push(skb, hh_len); 511 return dev_queue_xmit(skb); 512 } 513 514 static inline int neigh_output(struct neighbour *n, struct sk_buff *skb, 515 bool skip_cache) 516 { 517 const struct hh_cache *hh = &n->hh; 518 519 /* n->nud_state and hh->hh_len could be changed under us. 520 * neigh_hh_output() is taking care of the race later. 521 */ 522 if (!skip_cache && 523 (READ_ONCE(n->nud_state) & NUD_CONNECTED) && 524 READ_ONCE(hh->hh_len)) 525 return neigh_hh_output(hh, skb); 526 527 return n->output(n, skb); 528 } 529 530 static inline struct neighbour * 531 __neigh_lookup(struct neigh_table *tbl, const void *pkey, struct net_device *dev, int creat) 532 { 533 struct neighbour *n = neigh_lookup(tbl, pkey, dev); 534 535 if (n || !creat) 536 return n; 537 538 n = neigh_create(tbl, pkey, dev); 539 return IS_ERR(n) ? NULL : n; 540 } 541 542 static inline struct neighbour * 543 __neigh_lookup_errno(struct neigh_table *tbl, const void *pkey, 544 struct net_device *dev) 545 { 546 struct neighbour *n = neigh_lookup(tbl, pkey, dev); 547 548 if (n) 549 return n; 550 551 return neigh_create(tbl, pkey, dev); 552 } 553 554 struct neighbour_cb { 555 unsigned long sched_next; 556 unsigned int flags; 557 }; 558 559 #define LOCALLY_ENQUEUED 0x1 560 561 #define NEIGH_CB(skb) ((struct neighbour_cb *)(skb)->cb) 562 563 static inline void neigh_ha_snapshot(char *dst, const struct neighbour *n, 564 const struct net_device *dev) 565 { 566 unsigned int seq; 567 568 do { 569 seq = read_seqbegin(&n->ha_lock); 570 memcpy(dst, n->ha, dev->addr_len); 571 } while (read_seqretry(&n->ha_lock, seq)); 572 } 573 574 static inline void neigh_update_is_router(struct neighbour *neigh, u32 flags, 575 int *notify) 576 { 577 u8 ndm_flags = 0; 578 579 ndm_flags |= (flags & NEIGH_UPDATE_F_ISROUTER) ? NTF_ROUTER : 0; 580 if ((neigh->flags ^ ndm_flags) & NTF_ROUTER) { 581 if (ndm_flags & NTF_ROUTER) 582 neigh->flags |= NTF_ROUTER; 583 else 584 neigh->flags &= ~NTF_ROUTER; 585 *notify = 1; 586 } 587 } 588 #endif 589