projects/linux·blob master · refs

621 lines · 17.3 KiB

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 <[email protected]>
12 * Alexey Kuznetsov <[email protected]>
13 *
14 * Changes:
15 *
16 * Harald Welte: <[email protected]>
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#include <net/neighbour_tables.h>
33
34/*
35 * NUD stands for "neighbor unreachability detection"
36 */
37
38#define NUD_IN_TIMER (NUD_INCOMPLETE|NUD_REACHABLE|NUD_DELAY|NUD_PROBE)
39#define NUD_VALID (NUD_PERMANENT|NUD_NOARP|NUD_REACHABLE|NUD_PROBE|NUD_STALE|NUD_DELAY)
40#define NUD_CONNECTED (NUD_PERMANENT|NUD_NOARP|NUD_REACHABLE)
41
42struct neighbour;
43
44enum {
45 NEIGH_VAR_MCAST_PROBES,
46 NEIGH_VAR_UCAST_PROBES,
47 NEIGH_VAR_APP_PROBES,
48 NEIGH_VAR_MCAST_REPROBES,
49 NEIGH_VAR_RETRANS_TIME,
50 NEIGH_VAR_BASE_REACHABLE_TIME,
51 NEIGH_VAR_DELAY_PROBE_TIME,
52 NEIGH_VAR_INTERVAL_PROBE_TIME_MS,
53 NEIGH_VAR_GC_STALETIME,
54 NEIGH_VAR_QUEUE_LEN_BYTES,
55 NEIGH_VAR_PROXY_QLEN,
56 NEIGH_VAR_ANYCAST_DELAY,
57 NEIGH_VAR_PROXY_DELAY,
58 NEIGH_VAR_LOCKTIME,
59#define NEIGH_VAR_DATA_MAX (NEIGH_VAR_LOCKTIME + 1)
60 /* Following are used as a second way to access one of the above */
61 NEIGH_VAR_QUEUE_LEN, /* same data as NEIGH_VAR_QUEUE_LEN_BYTES */
62 NEIGH_VAR_RETRANS_TIME_MS, /* same data as NEIGH_VAR_RETRANS_TIME */
63 NEIGH_VAR_BASE_REACHABLE_TIME_MS, /* same data as NEIGH_VAR_BASE_REACHABLE_TIME */
64 /* Following are used by "default" only */
65 NEIGH_VAR_GC_INTERVAL,
66 NEIGH_VAR_GC_THRESH1,
67 NEIGH_VAR_GC_THRESH2,
68 NEIGH_VAR_GC_THRESH3,
69 NEIGH_VAR_MAX
70};
71
72struct neigh_parms {
73 possible_net_t net;
74 struct net_device *dev;
75 netdevice_tracker dev_tracker;
76 struct list_head list;
77 int (*neigh_setup)(struct neighbour *);
78 struct neigh_table *tbl;
79
80 void *sysctl_table;
81
82 int dead;
83 refcount_t refcnt;
84 struct rcu_head rcu_head;
85
86 int reachable_time;
87 u32 qlen;
88 int data[NEIGH_VAR_DATA_MAX];
89 DECLARE_BITMAP(data_state, NEIGH_VAR_DATA_MAX);
90};
91
92static inline void neigh_var_set(struct neigh_parms *p, int index, int val)
93{
94 set_bit(index, p->data_state);
95 WRITE_ONCE(p->data[index], val);
96}
97
98#define __NEIGH_VAR(p, attr) ((p)->data[NEIGH_VAR_ ## attr])
99#define NEIGH_VAR(p, attr) READ_ONCE(__NEIGH_VAR(p, attr))
100#define NEIGH_VAR_PTR(p, attr) (&(__NEIGH_VAR(p, attr)))
101
102/* In ndo_neigh_setup, NEIGH_VAR_INIT should be used.
103 * In other cases, NEIGH_VAR_SET should be used.
104 */
105#define NEIGH_VAR_INIT(p, attr, val) (__NEIGH_VAR(p, attr) = val)
106#define NEIGH_VAR_SET(p, attr, val) neigh_var_set(p, NEIGH_VAR_ ## attr, val)
107
108static inline void neigh_parms_data_state_setall(struct neigh_parms *p)
109{
110 bitmap_fill(p->data_state, NEIGH_VAR_DATA_MAX);
111}
112
113static inline void neigh_parms_data_state_cleanall(struct neigh_parms *p)
114{
115 bitmap_zero(p->data_state, NEIGH_VAR_DATA_MAX);
116}
117
118struct neigh_statistics {
119 unsigned long allocs; /* number of allocated neighs */
120 unsigned long destroys; /* number of destroyed neighs */
121 unsigned long hash_grows; /* number of hash resizes */
122
123 unsigned long res_failed; /* number of failed resolutions */
124
125 unsigned long lookups; /* number of lookups */
126 unsigned long hits; /* number of hits (among lookups) */
127
128 unsigned long rcv_probes_mcast; /* number of received mcast ipv6 */
129 unsigned long rcv_probes_ucast; /* number of received ucast ipv6 */
130
131 unsigned long periodic_gc_runs; /* number of periodic GC runs */
132 unsigned long forced_gc_runs; /* number of forced GC runs */
133
134 unsigned long unres_discards; /* number of unresolved drops */
135 unsigned long table_fulls; /* times even gc couldn't help */
136};
137
138#define NEIGH_CACHE_STAT_INC(tbl, field) this_cpu_inc((tbl)->stats->field)
139
140struct neighbour {
141 struct hlist_node hash;
142 struct hlist_node dev_list;
143 struct neigh_table *tbl;
144 struct neigh_parms *parms;
145 unsigned long confirmed;
146 unsigned long updated;
147 rwlock_t lock;
148 refcount_t refcnt;
149 unsigned int arp_queue_len_bytes;
150 struct sk_buff_head arp_queue;
151 struct timer_list timer;
152 unsigned long used;
153 atomic_t probes;
154 u8 nud_state;
155 u8 type;
156 u8 dead;
157 u8 protocol;
158 u32 flags;
159 seqlock_t ha_lock;
160 unsigned char ha[ALIGN(MAX_ADDR_LEN, sizeof(unsigned long))] __aligned(8);
161 struct hh_cache hh;
162 int (*output)(struct neighbour *, struct sk_buff *);
163 const struct neigh_ops *ops;
164 struct list_head gc_list;
165 struct list_head managed_list;
166 struct rcu_head rcu;
167 struct net_device *dev;
168 netdevice_tracker dev_tracker;
169 u8 primary_key[];
170} __randomize_layout;
171
172struct neigh_ops {
173 int family;
174 void (*solicit)(struct neighbour *, struct sk_buff *);
175 void (*error_report)(struct neighbour *, struct sk_buff *);
176 int (*output)(struct neighbour *, struct sk_buff *);
177 int (*connected_output)(struct neighbour *, struct sk_buff *);
178};
179
180struct pneigh_entry {
181 struct pneigh_entry __rcu *next;
182 possible_net_t net;
183 struct net_device *dev;
184 netdevice_tracker dev_tracker;
185 union {
186 struct list_head free_node;
187 struct rcu_head rcu;
188 };
189 u32 flags;
190 u8 protocol;
191 bool permanent;
192 u32 key[];
193};
194
195/*
196 * neighbour table manipulation
197 */
198
199#define NEIGH_NUM_HASH_RND 4
200
201struct neigh_hash_table {
202 struct hlist_head *hash_heads;
203 unsigned int hash_shift;
204 __u32 hash_rnd[NEIGH_NUM_HASH_RND];
205 struct rcu_head rcu;
206};
207
208
209struct neigh_table {
210 int family;
211 unsigned int entry_size;
212 unsigned int key_len;
213 __be16 protocol;
214 __u32 (*hash)(const void *pkey,
215 const struct net_device *dev,
216 __u32 *hash_rnd);
217 bool (*key_eq)(const struct neighbour *, const void *pkey);
218 int (*constructor)(struct neighbour *);
219 int (*pconstructor)(struct pneigh_entry *);
220 void (*pdestructor)(struct pneigh_entry *);
221 void (*proxy_redo)(struct sk_buff *skb);
222 int (*is_multicast)(const void *pkey);
223 bool (*allow_add)(const struct net_device *dev,
224 struct netlink_ext_ack *extack);
225 char *id;
226 struct neigh_parms parms;
227 struct list_head parms_list;
228 int gc_interval;
229 int gc_thresh1;
230 int gc_thresh2;
231 int gc_thresh3;
232 unsigned long last_flush;
233 struct delayed_work gc_work;
234 struct delayed_work managed_work;
235 struct timer_list proxy_timer;
236 struct sk_buff_head proxy_queue;
237 atomic_t entries;
238 atomic_t gc_entries;
239 struct list_head gc_list;
240 struct list_head managed_list;
241 spinlock_t lock;
242 unsigned long last_rand;
243 struct neigh_statistics __percpu *stats;
244 struct neigh_hash_table __rcu *nht;
245 struct mutex phash_lock;
246 struct pneigh_entry __rcu **phash_buckets;
247};
248
249static inline int neigh_parms_family(struct neigh_parms *p)
250{
251 return p->tbl->family;
252}
253
254#define NEIGH_PRIV_ALIGN sizeof(long long)
255#define NEIGH_ENTRY_SIZE(size) ALIGN((size), NEIGH_PRIV_ALIGN)
256
257static inline void *neighbour_priv(const struct neighbour *n)
258{
259 return (char *)n + n->tbl->entry_size;
260}
261
262/* flags for neigh_update() */
263#define NEIGH_UPDATE_F_OVERRIDE BIT(0)
264#define NEIGH_UPDATE_F_WEAK_OVERRIDE BIT(1)
265#define NEIGH_UPDATE_F_OVERRIDE_ISROUTER BIT(2)
266#define NEIGH_UPDATE_F_USE BIT(3)
267#define NEIGH_UPDATE_F_MANAGED BIT(4)
268#define NEIGH_UPDATE_F_EXT_LEARNED BIT(5)
269#define NEIGH_UPDATE_F_ISROUTER BIT(6)
270#define NEIGH_UPDATE_F_ADMIN BIT(7)
271#define NEIGH_UPDATE_F_EXT_VALIDATED BIT(8)
272
273/* In-kernel representation for NDA_FLAGS_EXT flags: */
274#define NTF_OLD_MASK 0xff
275#define NTF_EXT_SHIFT 8
276#define NTF_EXT_MASK (NTF_EXT_MANAGED | NTF_EXT_EXT_VALIDATED)
277
278#define NTF_MANAGED (NTF_EXT_MANAGED << NTF_EXT_SHIFT)
279#define NTF_EXT_VALIDATED (NTF_EXT_EXT_VALIDATED << NTF_EXT_SHIFT)
280
281extern const struct nla_policy nda_policy[];
282
283#define neigh_for_each_in_bucket(pos, head) hlist_for_each_entry(pos, head, hash)
284#define neigh_for_each_in_bucket_rcu(pos, head) \
285 hlist_for_each_entry_rcu(pos, head, hash)
286#define neigh_for_each_in_bucket_safe(pos, tmp, head) \
287 hlist_for_each_entry_safe(pos, tmp, head, hash)
288
289static inline bool neigh_key_eq32(const struct neighbour *n, const void *pkey)
290{
291 return *(const u32 *)n->primary_key == *(const u32 *)pkey;
292}
293
294static inline bool neigh_key_eq128(const struct neighbour *n, const void *pkey)
295{
296 const u32 *n32 = (const u32 *)n->primary_key;
297 const u32 *p32 = pkey;
298
299 return ((n32[0] ^ p32[0]) | (n32[1] ^ p32[1]) |
300 (n32[2] ^ p32[2]) | (n32[3] ^ p32[3])) == 0;
301}
302
303static inline struct neighbour *___neigh_lookup_noref(
304 struct neigh_table *tbl,
305 bool (*key_eq)(const struct neighbour *n, const void *pkey),
306 __u32 (*hash)(const void *pkey,
307 const struct net_device *dev,
308 __u32 *hash_rnd),
309 const void *pkey,
310 struct net_device *dev)
311{
312 struct neigh_hash_table *nht = rcu_dereference(tbl->nht);
313 struct neighbour *n;
314 u32 hash_val;
315
316 hash_val = hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
317 neigh_for_each_in_bucket_rcu(n, &nht->hash_heads[hash_val])
318 if (n->dev == dev && key_eq(n, pkey))
319 return n;
320
321 return NULL;
322}
323
324static inline struct neighbour *__neigh_lookup_noref(struct neigh_table *tbl,
325 const void *pkey,
326 struct net_device *dev)
327{
328 return ___neigh_lookup_noref(tbl, tbl->key_eq, tbl->hash, pkey, dev);
329}
330
331static inline void neigh_confirm(struct neighbour *n)
332{
333 if (n) {
334 unsigned long now = jiffies;
335
336 /* avoid dirtying neighbour */
337 if (READ_ONCE(n->confirmed) != now)
338 WRITE_ONCE(n->confirmed, now);
339 }
340}
341
342void neigh_table_init(int index, struct neigh_table *tbl);
343int neigh_table_clear(int index, struct neigh_table *tbl);
344struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
345 struct net_device *dev);
346struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey,
347 struct net_device *dev, bool want_ref);
348static inline struct neighbour *neigh_create(struct neigh_table *tbl,
349 const void *pkey,
350 struct net_device *dev)
351{
352 return __neigh_create(tbl, pkey, dev, true);
353}
354void neigh_destroy(struct neighbour *neigh);
355int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb,
356 const bool immediate_ok);
357int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, u32 flags,
358 u32 nlmsg_pid);
359void __neigh_set_probe_once(struct neighbour *neigh);
360bool neigh_remove_one(struct neighbour *ndel);
361void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev);
362int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
363int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev);
364int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb);
365int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb);
366int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb);
367struct neighbour *neigh_event_ns(struct neigh_table *tbl,
368 u8 *lladdr, void *saddr,
369 struct net_device *dev);
370
371struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
372 struct neigh_table *tbl);
373void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms);
374
375static inline
376struct net *neigh_parms_net(const struct neigh_parms *parms)
377{
378 return read_pnet(&parms->net);
379}
380
381unsigned long neigh_rand_reach_time(unsigned long base);
382
383static inline void neigh_set_reach_time(struct neigh_parms *p)
384{
385 unsigned long base = NEIGH_VAR(p, BASE_REACHABLE_TIME);
386
387 WRITE_ONCE(p->reachable_time, neigh_rand_reach_time(base));
388}
389
390void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
391 struct sk_buff *skb);
392struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net,
393 const void *key, struct net_device *dev);
394int pneigh_create(struct neigh_table *tbl, struct net *net, const void *key,
395 struct net_device *dev, u32 flags, u8 protocol,
396 bool permanent);
397int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *key,
398 struct net_device *dev);
399
400static inline struct net *pneigh_net(const struct pneigh_entry *pneigh)
401{
402 return read_pnet(&pneigh->net);
403}
404
405void neigh_app_ns(struct neighbour *n);
406void neigh_for_each(struct neigh_table *tbl,
407 void (*cb)(struct neighbour *, void *), void *cookie);
408void __neigh_for_each_release(struct neigh_table *tbl,
409 int (*cb)(struct neighbour *));
410int neigh_xmit(int fam, struct net_device *, const void *, struct sk_buff *);
411
412struct neigh_seq_state {
413 struct seq_net_private p;
414 struct neigh_table *tbl;
415 struct neigh_hash_table *nht;
416 void *(*neigh_sub_iter)(struct neigh_seq_state *state,
417 struct neighbour *n, loff_t *pos);
418 unsigned int bucket;
419 unsigned int flags;
420#define NEIGH_SEQ_NEIGH_ONLY 0x00000001
421#define NEIGH_SEQ_IS_PNEIGH 0x00000002
422#define NEIGH_SEQ_SKIP_NOARP 0x00000004
423};
424void *neigh_seq_start(struct seq_file *, loff_t *, struct neigh_table *,
425 unsigned int);
426void *neigh_seq_next(struct seq_file *, void *, loff_t *);
427void neigh_seq_stop(struct seq_file *, void *);
428
429int neigh_proc_dointvec(const struct ctl_table *ctl, int write,
430 void *buffer, size_t *lenp, loff_t *ppos);
431int neigh_proc_dointvec_jiffies(const struct ctl_table *ctl, int write,
432 void *buffer,
433 size_t *lenp, loff_t *ppos);
434int neigh_proc_dointvec_ms_jiffies(const struct ctl_table *ctl, int write,
435 void *buffer, size_t *lenp, loff_t *ppos);
436
437int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
438 proc_handler *proc_handler);
439void neigh_sysctl_unregister(struct neigh_parms *p);
440
441static inline void __neigh_parms_put(struct neigh_parms *parms)
442{
443 refcount_dec(&parms->refcnt);
444}
445
446static inline struct neigh_parms *neigh_parms_clone(struct neigh_parms *parms)
447{
448 refcount_inc(&parms->refcnt);
449 return parms;
450}
451
452/*
453 * Neighbour references
454 */
455
456static inline void neigh_release(struct neighbour *neigh)
457{
458 if (refcount_dec_and_test(&neigh->refcnt))
459 neigh_destroy(neigh);
460}
461
462static inline struct neighbour * neigh_clone(struct neighbour *neigh)
463{
464 if (neigh)
465 refcount_inc(&neigh->refcnt);
466 return neigh;
467}
468
469#define neigh_hold(n) refcount_inc(&(n)->refcnt)
470
471static __always_inline int neigh_event_send_probe(struct neighbour *neigh,
472 struct sk_buff *skb,
473 const bool immediate_ok)
474{
475 unsigned long now = jiffies;
476
477 if (READ_ONCE(neigh->used) != now)
478 WRITE_ONCE(neigh->used, now);
479 if (!(READ_ONCE(neigh->nud_state) & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE)))
480 return __neigh_event_send(neigh, skb, immediate_ok);
481 return 0;
482}
483
484static inline int neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
485{
486 return neigh_event_send_probe(neigh, skb, true);
487}
488
489#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
490static inline int neigh_hh_bridge(struct hh_cache *hh, struct sk_buff *skb)
491{
492 unsigned int seq, hh_alen = HH_DATA_ALIGN(ETH_HLEN);
493 int err;
494
495 err = skb_cow_head(skb, hh_alen);
496 if (err)
497 return err;
498
499 do {
500 seq = read_seqbegin(&hh->hh_lock);
501 memcpy(skb->data - hh_alen, hh->hh_data, ETH_ALEN + hh_alen - ETH_HLEN);
502 } while (read_seqretry(&hh->hh_lock, seq));
503 return 0;
504}
505#endif
506
507static inline int neigh_hh_output(const struct hh_cache *hh, struct sk_buff *skb)
508{
509 unsigned int hh_alen = 0;
510 unsigned int seq;
511 unsigned int hh_len;
512
513 do {
514 seq = read_seqbegin(&hh->hh_lock);
515 hh_len = READ_ONCE(hh->hh_len);
516 if (likely(hh_len <= HH_DATA_MOD)) {
517 hh_alen = HH_DATA_MOD;
518
519 /* skb_push() would proceed silently if we have room for
520 * the unaligned size but not for the aligned size:
521 * check headroom explicitly.
522 */
523 if (likely(skb_headroom(skb) >= HH_DATA_MOD)) {
524 /* this is inlined by gcc */
525 memcpy(skb->data - HH_DATA_MOD, hh->hh_data,
526 HH_DATA_MOD);
527 }
528 } else {
529 hh_alen = HH_DATA_ALIGN(hh_len);
530
531 if (likely(skb_headroom(skb) >= hh_alen)) {
532 memcpy(skb->data - hh_alen, hh->hh_data,
533 hh_alen);
534 }
535 }
536 } while (read_seqretry(&hh->hh_lock, seq));
537
538 if (WARN_ON_ONCE(skb_headroom(skb) < hh_alen)) {
539 kfree_skb(skb);
540 return NET_XMIT_DROP;
541 }
542
543 __skb_push(skb, hh_len);
544 return dev_queue_xmit(skb);
545}
546
547static inline int neigh_output(struct neighbour *n, struct sk_buff *skb,
548 bool skip_cache)
549{
550 const struct hh_cache *hh = &n->hh;
551
552 /* n->nud_state and hh->hh_len could be changed under us.
553 * neigh_hh_output() is taking care of the race later.
554 */
555 if (!skip_cache &&
556 (READ_ONCE(n->nud_state) & NUD_CONNECTED) &&
557 READ_ONCE(hh->hh_len))
558 return neigh_hh_output(hh, skb);
559
560 return READ_ONCE(n->output)(n, skb);
561}
562
563static inline struct neighbour *
564__neigh_lookup(struct neigh_table *tbl, const void *pkey, struct net_device *dev, int creat)
565{
566 struct neighbour *n = neigh_lookup(tbl, pkey, dev);
567
568 if (n || !creat)
569 return n;
570
571 n = neigh_create(tbl, pkey, dev);
572 return IS_ERR(n) ? NULL : n;
573}
574
575static inline struct neighbour *
576__neigh_lookup_errno(struct neigh_table *tbl, const void *pkey,
577 struct net_device *dev)
578{
579 struct neighbour *n = neigh_lookup(tbl, pkey, dev);
580
581 if (n)
582 return n;
583
584 return neigh_create(tbl, pkey, dev);
585}
586
587struct neighbour_cb {
588 unsigned long sched_next;
589 unsigned int flags;
590};
591
592#define LOCALLY_ENQUEUED 0x1
593
594#define NEIGH_CB(skb) ((struct neighbour_cb *)(skb)->cb)
595
596static inline void neigh_ha_snapshot(char *dst, const struct neighbour *n,
597 const struct net_device *dev)
598{
599 unsigned int seq;
600
601 do {
602 seq = read_seqbegin(&n->ha_lock);
603 memcpy(dst, n->ha, dev->addr_len);
604 } while (read_seqretry(&n->ha_lock, seq));
605}
606
607static inline void neigh_update_is_router(struct neighbour *neigh, u32 flags,
608 int *notify)
609{
610 u8 ndm_flags = 0;
611
612 ndm_flags |= (flags & NEIGH_UPDATE_F_ISROUTER) ? NTF_ROUTER : 0;
613 if ((neigh->flags ^ ndm_flags) & NTF_ROUTER) {
614 if (ndm_flags & NTF_ROUTER)
615 neigh->flags |= NTF_ROUTER;
616 else
617 neigh->flags &= ~NTF_ROUTER;
618 *notify = 1;
619 }
620}
621#endif