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_DN_TABLE = 2,
243 NEIGH_NR_TABLES,
244 NEIGH_LINK_TABLE = NEIGH_NR_TABLES /* Pseudo table for neigh_xmit */
245 };
246
247 static inline int neigh_parms_family(struct neigh_parms *p)
248 {
249 return p->tbl->family;
250 }
251
252 #define NEIGH_PRIV_ALIGN sizeof(long long)
253 #define NEIGH_ENTRY_SIZE(size) ALIGN((size), NEIGH_PRIV_ALIGN)
254
255 static inline void *neighbour_priv(const struct neighbour *n)
256 {
257 return (char *)n + n->tbl->entry_size;
258 }
259
260 /* flags for neigh_update() */
261 #define NEIGH_UPDATE_F_OVERRIDE BIT(0)
262 #define NEIGH_UPDATE_F_WEAK_OVERRIDE BIT(1)
263 #define NEIGH_UPDATE_F_OVERRIDE_ISROUTER BIT(2)
264 #define NEIGH_UPDATE_F_USE BIT(3)
265 #define NEIGH_UPDATE_F_MANAGED BIT(4)
266 #define NEIGH_UPDATE_F_EXT_LEARNED BIT(5)
267 #define NEIGH_UPDATE_F_ISROUTER BIT(6)
268 #define NEIGH_UPDATE_F_ADMIN BIT(7)
269
270 /* In-kernel representation for NDA_FLAGS_EXT flags: */
271 #define NTF_OLD_MASK 0xff
272 #define NTF_EXT_SHIFT 8
273 #define NTF_EXT_MASK (NTF_EXT_MANAGED)
274
275 #define NTF_MANAGED (NTF_EXT_MANAGED << NTF_EXT_SHIFT)
276
277 extern const struct nla_policy nda_policy[];
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(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(nht->hash_buckets[hash_val]);
308 n != NULL;
309 n = rcu_dereference(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 static inline void neigh_confirm(struct neighbour *n)
325 {
326 if (n) {
327 unsigned long now = jiffies;
328
329 /* avoid dirtying neighbour */
330 if (READ_ONCE(n->confirmed) != now)
331 WRITE_ONCE(n->confirmed, now);
332 }
333 }
334
335 void neigh_table_init(int index, struct neigh_table *tbl);
336 int neigh_table_clear(int index, struct neigh_table *tbl);
337 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
338 struct net_device *dev);
339 struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey,
340 struct net_device *dev, bool want_ref);
341 static inline struct neighbour *neigh_create(struct neigh_table *tbl,
342 const void *pkey,
343 struct net_device *dev)
344 {
345 return __neigh_create(tbl, pkey, dev, true);
346 }
347 void neigh_destroy(struct neighbour *neigh);
348 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb,
349 const bool immediate_ok);
350 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, u32 flags,
351 u32 nlmsg_pid);
352 void __neigh_set_probe_once(struct neighbour *neigh);
353 bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl);
354 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev);
355 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
356 int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev);
357 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb);
358 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb);
359 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb);
360 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
361 u8 *lladdr, void *saddr,
362 struct net_device *dev);
363
364 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
365 struct neigh_table *tbl);
366 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms);
367
368 static inline
369 struct net *neigh_parms_net(const struct neigh_parms *parms)
370 {
371 return read_pnet(&parms->net);
372 }
373
374 unsigned long neigh_rand_reach_time(unsigned long base);
375
376 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
377 struct sk_buff *skb);
378 struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net,
379 const void *key, struct net_device *dev,
380 int creat);
381 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl, struct net *net,
382 const void *key, struct net_device *dev);
383 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *key,
384 struct net_device *dev);
385
386 static inline struct net *pneigh_net(const struct pneigh_entry *pneigh)
387 {
388 return read_pnet(&pneigh->net);
389 }
390
391 void neigh_app_ns(struct neighbour *n);
392 void neigh_for_each(struct neigh_table *tbl,
393 void (*cb)(struct neighbour *, void *), void *cookie);
394 void __neigh_for_each_release(struct neigh_table *tbl,
395 int (*cb)(struct neighbour *));
396 int neigh_xmit(int fam, struct net_device *, const void *, struct sk_buff *);
397
398 struct neigh_seq_state {
399 struct seq_net_private p;
400 struct neigh_table *tbl;
401 struct neigh_hash_table *nht;
402 void *(*neigh_sub_iter)(struct neigh_seq_state *state,
403 struct neighbour *n, loff_t *pos);
404 unsigned int bucket;
405 unsigned int flags;
406 #define NEIGH_SEQ_NEIGH_ONLY 0x00000001
407 #define NEIGH_SEQ_IS_PNEIGH 0x00000002
408 #define NEIGH_SEQ_SKIP_NOARP 0x00000004
409 };
410 void *neigh_seq_start(struct seq_file *, loff_t *, struct neigh_table *,
411 unsigned int);
412 void *neigh_seq_next(struct seq_file *, void *, loff_t *);
413 void neigh_seq_stop(struct seq_file *, void *);
414
415 int neigh_proc_dointvec(const struct ctl_table *ctl, int write,
416 void *buffer, size_t *lenp, loff_t *ppos);
417 int neigh_proc_dointvec_jiffies(const struct ctl_table *ctl, int write,
418 void *buffer,
419 size_t *lenp, loff_t *ppos);
420 int neigh_proc_dointvec_ms_jiffies(const struct ctl_table *ctl, int write,
421 void *buffer, size_t *lenp, loff_t *ppos);
422
423 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
424 proc_handler *proc_handler);
425 void neigh_sysctl_unregister(struct neigh_parms *p);
426
427 static inline void __neigh_parms_put(struct neigh_parms *parms)
428 {
429 refcount_dec(&parms->refcnt);
430 }
431
432 static inline struct neigh_parms *neigh_parms_clone(struct neigh_parms *parms)
433 {
434 refcount_inc(&parms->refcnt);
435 return parms;
436 }
437
438 /*
439 * Neighbour references
440 */
441
442 static inline void neigh_release(struct neighbour *neigh)
443 {
444 if (refcount_dec_and_test(&neigh->refcnt))
445 neigh_destroy(neigh);
446 }
447
448 static inline struct neighbour * neigh_clone(struct neighbour *neigh)
449 {
450 if (neigh)
451 refcount_inc(&neigh->refcnt);
452 return neigh;
453 }
454
455 #define neigh_hold(n) refcount_inc(&(n)->refcnt)
456
457 static __always_inline int neigh_event_send_probe(struct neighbour *neigh,
458 struct sk_buff *skb,
459 const bool immediate_ok)
460 {
461 unsigned long now = jiffies;
462
463 if (READ_ONCE(neigh->used) != now)
464 WRITE_ONCE(neigh->used, now);
465 if (!(READ_ONCE(neigh->nud_state) & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE)))
466 return __neigh_event_send(neigh, skb, immediate_ok);
467 return 0;
468 }
469
470 static inline int neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
471 {
472 return neigh_event_send_probe(neigh, skb, true);
473 }
474
475 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
476 static inline int neigh_hh_bridge(struct hh_cache *hh, struct sk_buff *skb)
477 {
478 unsigned int seq, hh_alen;
479
480 do {
481 seq = read_seqbegin(&hh->hh_lock);
482 hh_alen = HH_DATA_ALIGN(ETH_HLEN);
483 memcpy(skb->data - hh_alen, hh->hh_data, ETH_ALEN + hh_alen - ETH_HLEN);
484 } while (read_seqretry(&hh->hh_lock, seq));
485 return 0;
486 }
487 #endif
488
489 static inline int neigh_hh_output(const struct hh_cache *hh, struct sk_buff *skb)
490 {
491 unsigned int hh_alen = 0;
492 unsigned int seq;
493 unsigned int hh_len;
494
495 do {
496 seq = read_seqbegin(&hh->hh_lock);
497 hh_len = READ_ONCE(hh->hh_len);
498 if (likely(hh_len <= HH_DATA_MOD)) {
499 hh_alen = HH_DATA_MOD;
500
501 /* skb_push() would proceed silently if we have room for
502 * the unaligned size but not for the aligned size:
503 * check headroom explicitly.
504 */
505 if (likely(skb_headroom(skb) >= HH_DATA_MOD)) {
506 /* this is inlined by gcc */
507 memcpy(skb->data - HH_DATA_MOD, hh->hh_data,
508 HH_DATA_MOD);
509 }
510 } else {
511 hh_alen = HH_DATA_ALIGN(hh_len);
512
513 if (likely(skb_headroom(skb) >= hh_alen)) {
514 memcpy(skb->data - hh_alen, hh->hh_data,
515 hh_alen);
516 }
517 }
518 } while (read_seqretry(&hh->hh_lock, seq));
519
520 if (WARN_ON_ONCE(skb_headroom(skb) < hh_alen)) {
521 kfree_skb(skb);
522 return NET_XMIT_DROP;
523 }
524
525 __skb_push(skb, hh_len);
526 return dev_queue_xmit(skb);
527 }
528
529 static inline int neigh_output(struct neighbour *n, struct sk_buff *skb,
530 bool skip_cache)
531 {
532 const struct hh_cache *hh = &n->hh;
533
534 /* n->nud_state and hh->hh_len could be changed under us.
535 * neigh_hh_output() is taking care of the race later.
536 */
537 if (!skip_cache &&
538 (READ_ONCE(n->nud_state) & NUD_CONNECTED) &&
539 READ_ONCE(hh->hh_len))
540 return neigh_hh_output(hh, skb);
541
542 return READ_ONCE(n->output)(n, skb);
543 }
544
545 static inline struct neighbour *
546 __neigh_lookup(struct neigh_table *tbl, const void *pkey, struct net_device *dev, int creat)
547 {
548 struct neighbour *n = neigh_lookup(tbl, pkey, dev);
549
550 if (n || !creat)
551 return n;
552
553 n = neigh_create(tbl, pkey, dev);
554 return IS_ERR(n) ? NULL : n;
555 }
556
557 static inline struct neighbour *
558 __neigh_lookup_errno(struct neigh_table *tbl, const void *pkey,
559 struct net_device *dev)
560 {
561 struct neighbour *n = neigh_lookup(tbl, pkey, dev);
562
563 if (n)
564 return n;
565
566 return neigh_create(tbl, pkey, dev);
567 }
568
569 struct neighbour_cb {
570 unsigned long sched_next;
571 unsigned int flags;
572 };
573
574 #define LOCALLY_ENQUEUED 0x1
575
576 #define NEIGH_CB(skb) ((struct neighbour_cb *)(skb)->cb)
577
578 static inline void neigh_ha_snapshot(char *dst, const struct neighbour *n,
579 const struct net_device *dev)
580 {
581 unsigned int seq;
582
583 do {
584 seq = read_seqbegin(&n->ha_lock);
585 memcpy(dst, n->ha, dev->addr_len);
586 } while (read_seqretry(&n->ha_lock, seq));
587 }
588
589 static inline void neigh_update_is_router(struct neighbour *neigh, u32 flags,
590 int *notify)
591 {
592 u8 ndm_flags = 0;
593
594 ndm_flags |= (flags & NEIGH_UPDATE_F_ISROUTER) ? NTF_ROUTER : 0;
595 if ((neigh->flags ^ ndm_flags) & NTF_ROUTER) {
596 if (ndm_flags & NTF_ROUTER)
597 neigh->flags |= NTF_ROUTER;
598 else
599 neigh->flags &= ~NTF_ROUTER;
600 *notify = 1;
601 }
602 }
603 #endif
604
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.