ipv4: Allow amount of dirty memory from fib resizing to be controllable
[linux-2.6-block.git] / include / net / neighbour.h
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         void    (*neigh_cleanup)(struct neighbour *);
76         struct neigh_table *tbl;
77
78         void    *sysctl_table;
79
80         int dead;
81         refcount_t refcnt;
82         struct rcu_head rcu_head;
83
84         int     reachable_time;
85         int     data[NEIGH_VAR_DATA_MAX];
86         DECLARE_BITMAP(data_state, NEIGH_VAR_DATA_MAX);
87 };
88
89 static inline void neigh_var_set(struct neigh_parms *p, int index, int val)
90 {
91         set_bit(index, p->data_state);
92         p->data[index] = val;
93 }
94
95 #define NEIGH_VAR(p, attr) ((p)->data[NEIGH_VAR_ ## attr])
96
97 /* In ndo_neigh_setup, NEIGH_VAR_INIT should be used.
98  * In other cases, NEIGH_VAR_SET should be used.
99  */
100 #define NEIGH_VAR_INIT(p, attr, val) (NEIGH_VAR(p, attr) = val)
101 #define NEIGH_VAR_SET(p, attr, val) neigh_var_set(p, NEIGH_VAR_ ## attr, val)
102
103 static inline void neigh_parms_data_state_setall(struct neigh_parms *p)
104 {
105         bitmap_fill(p->data_state, NEIGH_VAR_DATA_MAX);
106 }
107
108 static inline void neigh_parms_data_state_cleanall(struct neigh_parms *p)
109 {
110         bitmap_zero(p->data_state, NEIGH_VAR_DATA_MAX);
111 }
112
113 struct neigh_statistics {
114         unsigned long allocs;           /* number of allocated neighs */
115         unsigned long destroys;         /* number of destroyed neighs */
116         unsigned long hash_grows;       /* number of hash resizes */
117
118         unsigned long res_failed;       /* number of failed resolutions */
119
120         unsigned long lookups;          /* number of lookups */
121         unsigned long hits;             /* number of hits (among lookups) */
122
123         unsigned long rcv_probes_mcast; /* number of received mcast ipv6 */
124         unsigned long rcv_probes_ucast; /* number of received ucast ipv6 */
125
126         unsigned long periodic_gc_runs; /* number of periodic GC runs */
127         unsigned long forced_gc_runs;   /* number of forced GC runs */
128
129         unsigned long unres_discards;   /* number of unresolved drops */
130         unsigned long table_fulls;      /* times even gc couldn't help */
131 };
132
133 #define NEIGH_CACHE_STAT_INC(tbl, field) this_cpu_inc((tbl)->stats->field)
134
135 struct neighbour {
136         struct neighbour __rcu  *next;
137         struct neigh_table      *tbl;
138         struct neigh_parms      *parms;
139         unsigned long           confirmed;
140         unsigned long           updated;
141         rwlock_t                lock;
142         refcount_t              refcnt;
143         unsigned int            arp_queue_len_bytes;
144         struct sk_buff_head     arp_queue;
145         struct timer_list       timer;
146         unsigned long           used;
147         atomic_t                probes;
148         __u8                    flags;
149         __u8                    nud_state;
150         __u8                    type;
151         __u8                    dead;
152         u8                      protocol;
153         seqlock_t               ha_lock;
154         unsigned char           ha[ALIGN(MAX_ADDR_LEN, sizeof(unsigned long))] __aligned(8);
155         struct hh_cache         hh;
156         int                     (*output)(struct neighbour *, struct sk_buff *);
157         const struct neigh_ops  *ops;
158         struct list_head        gc_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         u8                      flags;
177         u8                      protocol;
178         u8                      key[0];
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         char                    *id;
209         struct neigh_parms      parms;
210         struct list_head        parms_list;
211         int                     gc_interval;
212         int                     gc_thresh1;
213         int                     gc_thresh2;
214         int                     gc_thresh3;
215         unsigned long           last_flush;
216         struct delayed_work     gc_work;
217         struct timer_list       proxy_timer;
218         struct sk_buff_head     proxy_queue;
219         atomic_t                entries;
220         atomic_t                gc_entries;
221         struct list_head        gc_list;
222         rwlock_t                lock;
223         unsigned long           last_rand;
224         struct neigh_statistics __percpu *stats;
225         struct neigh_hash_table __rcu *nht;
226         struct pneigh_entry     **phash_buckets;
227 };
228
229 enum {
230         NEIGH_ARP_TABLE = 0,
231         NEIGH_ND_TABLE = 1,
232         NEIGH_DN_TABLE = 2,
233         NEIGH_NR_TABLES,
234         NEIGH_LINK_TABLE = NEIGH_NR_TABLES /* Pseudo table for neigh_xmit */
235 };
236
237 static inline int neigh_parms_family(struct neigh_parms *p)
238 {
239         return p->tbl->family;
240 }
241
242 #define NEIGH_PRIV_ALIGN        sizeof(long long)
243 #define NEIGH_ENTRY_SIZE(size)  ALIGN((size), NEIGH_PRIV_ALIGN)
244
245 static inline void *neighbour_priv(const struct neighbour *n)
246 {
247         return (char *)n + n->tbl->entry_size;
248 }
249
250 /* flags for neigh_update() */
251 #define NEIGH_UPDATE_F_OVERRIDE                 0x00000001
252 #define NEIGH_UPDATE_F_WEAK_OVERRIDE            0x00000002
253 #define NEIGH_UPDATE_F_OVERRIDE_ISROUTER        0x00000004
254 #define NEIGH_UPDATE_F_EXT_LEARNED              0x20000000
255 #define NEIGH_UPDATE_F_ISROUTER                 0x40000000
256 #define NEIGH_UPDATE_F_ADMIN                    0x80000000
257
258 extern const struct nla_policy nda_policy[];
259
260 static inline bool neigh_key_eq16(const struct neighbour *n, const void *pkey)
261 {
262         return *(const u16 *)n->primary_key == *(const u16 *)pkey;
263 }
264
265 static inline bool neigh_key_eq32(const struct neighbour *n, const void *pkey)
266 {
267         return *(const u32 *)n->primary_key == *(const u32 *)pkey;
268 }
269
270 static inline bool neigh_key_eq128(const struct neighbour *n, const void *pkey)
271 {
272         const u32 *n32 = (const u32 *)n->primary_key;
273         const u32 *p32 = pkey;
274
275         return ((n32[0] ^ p32[0]) | (n32[1] ^ p32[1]) |
276                 (n32[2] ^ p32[2]) | (n32[3] ^ p32[3])) == 0;
277 }
278
279 static inline struct neighbour *___neigh_lookup_noref(
280         struct neigh_table *tbl,
281         bool (*key_eq)(const struct neighbour *n, const void *pkey),
282         __u32 (*hash)(const void *pkey,
283                       const struct net_device *dev,
284                       __u32 *hash_rnd),
285         const void *pkey,
286         struct net_device *dev)
287 {
288         struct neigh_hash_table *nht = rcu_dereference_bh(tbl->nht);
289         struct neighbour *n;
290         u32 hash_val;
291
292         hash_val = hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
293         for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
294              n != NULL;
295              n = rcu_dereference_bh(n->next)) {
296                 if (n->dev == dev && key_eq(n, pkey))
297                         return n;
298         }
299
300         return NULL;
301 }
302
303 static inline struct neighbour *__neigh_lookup_noref(struct neigh_table *tbl,
304                                                      const void *pkey,
305                                                      struct net_device *dev)
306 {
307         return ___neigh_lookup_noref(tbl, tbl->key_eq, tbl->hash, pkey, dev);
308 }
309
310 void neigh_table_init(int index, struct neigh_table *tbl);
311 int neigh_table_clear(int index, struct neigh_table *tbl);
312 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
313                                struct net_device *dev);
314 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
315                                      const void *pkey);
316 struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey,
317                                  struct net_device *dev, bool want_ref);
318 static inline struct neighbour *neigh_create(struct neigh_table *tbl,
319                                              const void *pkey,
320                                              struct net_device *dev)
321 {
322         return __neigh_create(tbl, pkey, dev, true);
323 }
324 void neigh_destroy(struct neighbour *neigh);
325 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb);
326 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, u32 flags,
327                  u32 nlmsg_pid);
328 void __neigh_set_probe_once(struct neighbour *neigh);
329 bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl);
330 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev);
331 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
332 int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev);
333 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb);
334 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb);
335 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb);
336 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
337                                                 u8 *lladdr, void *saddr,
338                                                 struct net_device *dev);
339
340 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
341                                       struct neigh_table *tbl);
342 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms);
343
344 static inline
345 struct net *neigh_parms_net(const struct neigh_parms *parms)
346 {
347         return read_pnet(&parms->net);
348 }
349
350 unsigned long neigh_rand_reach_time(unsigned long base);
351
352 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
353                     struct sk_buff *skb);
354 struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net,
355                                    const void *key, struct net_device *dev,
356                                    int creat);
357 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl, struct net *net,
358                                      const void *key, struct net_device *dev);
359 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *key,
360                   struct net_device *dev);
361
362 static inline struct net *pneigh_net(const struct pneigh_entry *pneigh)
363 {
364         return read_pnet(&pneigh->net);
365 }
366
367 void neigh_app_ns(struct neighbour *n);
368 void neigh_for_each(struct neigh_table *tbl,
369                     void (*cb)(struct neighbour *, void *), void *cookie);
370 void __neigh_for_each_release(struct neigh_table *tbl,
371                               int (*cb)(struct neighbour *));
372 int neigh_xmit(int fam, struct net_device *, const void *, struct sk_buff *);
373 void pneigh_for_each(struct neigh_table *tbl,
374                      void (*cb)(struct pneigh_entry *));
375
376 struct neigh_seq_state {
377         struct seq_net_private p;
378         struct neigh_table *tbl;
379         struct neigh_hash_table *nht;
380         void *(*neigh_sub_iter)(struct neigh_seq_state *state,
381                                 struct neighbour *n, loff_t *pos);
382         unsigned int bucket;
383         unsigned int flags;
384 #define NEIGH_SEQ_NEIGH_ONLY    0x00000001
385 #define NEIGH_SEQ_IS_PNEIGH     0x00000002
386 #define NEIGH_SEQ_SKIP_NOARP    0x00000004
387 };
388 void *neigh_seq_start(struct seq_file *, loff_t *, struct neigh_table *,
389                       unsigned int);
390 void *neigh_seq_next(struct seq_file *, void *, loff_t *);
391 void neigh_seq_stop(struct seq_file *, void *);
392
393 int neigh_proc_dointvec(struct ctl_table *ctl, int write,
394                         void __user *buffer, size_t *lenp, loff_t *ppos);
395 int neigh_proc_dointvec_jiffies(struct ctl_table *ctl, int write,
396                                 void __user *buffer,
397                                 size_t *lenp, loff_t *ppos);
398 int neigh_proc_dointvec_ms_jiffies(struct ctl_table *ctl, int write,
399                                    void __user *buffer,
400                                    size_t *lenp, loff_t *ppos);
401
402 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
403                           proc_handler *proc_handler);
404 void neigh_sysctl_unregister(struct neigh_parms *p);
405
406 static inline void __neigh_parms_put(struct neigh_parms *parms)
407 {
408         refcount_dec(&parms->refcnt);
409 }
410
411 static inline struct neigh_parms *neigh_parms_clone(struct neigh_parms *parms)
412 {
413         refcount_inc(&parms->refcnt);
414         return parms;
415 }
416
417 /*
418  *      Neighbour references
419  */
420
421 static inline void neigh_release(struct neighbour *neigh)
422 {
423         if (refcount_dec_and_test(&neigh->refcnt))
424                 neigh_destroy(neigh);
425 }
426
427 static inline struct neighbour * neigh_clone(struct neighbour *neigh)
428 {
429         if (neigh)
430                 refcount_inc(&neigh->refcnt);
431         return neigh;
432 }
433
434 #define neigh_hold(n)   refcount_inc(&(n)->refcnt)
435
436 static inline int neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
437 {
438         unsigned long now = jiffies;
439         
440         if (neigh->used != now)
441                 neigh->used = now;
442         if (!(neigh->nud_state&(NUD_CONNECTED|NUD_DELAY|NUD_PROBE)))
443                 return __neigh_event_send(neigh, skb);
444         return 0;
445 }
446
447 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
448 static inline int neigh_hh_bridge(struct hh_cache *hh, struct sk_buff *skb)
449 {
450         unsigned int seq, hh_alen;
451
452         do {
453                 seq = read_seqbegin(&hh->hh_lock);
454                 hh_alen = HH_DATA_ALIGN(ETH_HLEN);
455                 memcpy(skb->data - hh_alen, hh->hh_data, ETH_ALEN + hh_alen - ETH_HLEN);
456         } while (read_seqretry(&hh->hh_lock, seq));
457         return 0;
458 }
459 #endif
460
461 static inline int neigh_hh_output(const struct hh_cache *hh, struct sk_buff *skb)
462 {
463         unsigned int hh_alen = 0;
464         unsigned int seq;
465         unsigned int hh_len;
466
467         do {
468                 seq = read_seqbegin(&hh->hh_lock);
469                 hh_len = hh->hh_len;
470                 if (likely(hh_len <= HH_DATA_MOD)) {
471                         hh_alen = HH_DATA_MOD;
472
473                         /* skb_push() would proceed silently if we have room for
474                          * the unaligned size but not for the aligned size:
475                          * check headroom explicitly.
476                          */
477                         if (likely(skb_headroom(skb) >= HH_DATA_MOD)) {
478                                 /* this is inlined by gcc */
479                                 memcpy(skb->data - HH_DATA_MOD, hh->hh_data,
480                                        HH_DATA_MOD);
481                         }
482                 } else {
483                         hh_alen = HH_DATA_ALIGN(hh_len);
484
485                         if (likely(skb_headroom(skb) >= hh_alen)) {
486                                 memcpy(skb->data - hh_alen, hh->hh_data,
487                                        hh_alen);
488                         }
489                 }
490         } while (read_seqretry(&hh->hh_lock, seq));
491
492         if (WARN_ON_ONCE(skb_headroom(skb) < hh_alen)) {
493                 kfree_skb(skb);
494                 return NET_XMIT_DROP;
495         }
496
497         __skb_push(skb, hh_len);
498         return dev_queue_xmit(skb);
499 }
500
501 static inline int neigh_output(struct neighbour *n, struct sk_buff *skb)
502 {
503         const struct hh_cache *hh = &n->hh;
504
505         if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
506                 return neigh_hh_output(hh, skb);
507         else
508                 return n->output(n, skb);
509 }
510
511 static inline struct neighbour *
512 __neigh_lookup(struct neigh_table *tbl, const void *pkey, struct net_device *dev, int creat)
513 {
514         struct neighbour *n = neigh_lookup(tbl, pkey, dev);
515
516         if (n || !creat)
517                 return n;
518
519         n = neigh_create(tbl, pkey, dev);
520         return IS_ERR(n) ? NULL : n;
521 }
522
523 static inline struct neighbour *
524 __neigh_lookup_errno(struct neigh_table *tbl, const void *pkey,
525   struct net_device *dev)
526 {
527         struct neighbour *n = neigh_lookup(tbl, pkey, dev);
528
529         if (n)
530                 return n;
531
532         return neigh_create(tbl, pkey, dev);
533 }
534
535 struct neighbour_cb {
536         unsigned long sched_next;
537         unsigned int flags;
538 };
539
540 #define LOCALLY_ENQUEUED 0x1
541
542 #define NEIGH_CB(skb)   ((struct neighbour_cb *)(skb)->cb)
543
544 static inline void neigh_ha_snapshot(char *dst, const struct neighbour *n,
545                                      const struct net_device *dev)
546 {
547         unsigned int seq;
548
549         do {
550                 seq = read_seqbegin(&n->ha_lock);
551                 memcpy(dst, n->ha, dev->addr_len);
552         } while (read_seqretry(&n->ha_lock, seq));
553 }
554
555 static inline void neigh_update_is_router(struct neighbour *neigh, u32 flags,
556                                           int *notify)
557 {
558         u8 ndm_flags = 0;
559
560         ndm_flags |= (flags & NEIGH_UPDATE_F_ISROUTER) ? NTF_ROUTER : 0;
561         if ((neigh->flags ^ ndm_flags) & NTF_ROUTER) {
562                 if (ndm_flags & NTF_ROUTER)
563                         neigh->flags |= NTF_ROUTER;
564                 else
565                         neigh->flags &= ~NTF_ROUTER;
566                 *notify = 1;
567         }
568 }
569 #endif