Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Generic address resolution entity | |
3 | * | |
4 | * Authors: | |
5 | * Pedro Roque <roque@di.fc.ul.pt> | |
6 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> | |
7 | * | |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
12 | * | |
13 | * Fixes: | |
14 | * Vitaly E. Lavrov releasing NULL neighbor in neigh_add. | |
15 | * Harald Welte Add neighbour cache statistics like rtstat | |
16 | */ | |
17 | ||
18 | #include <linux/config.h> | |
19 | #include <linux/types.h> | |
20 | #include <linux/kernel.h> | |
21 | #include <linux/module.h> | |
22 | #include <linux/socket.h> | |
23 | #include <linux/sched.h> | |
24 | #include <linux/netdevice.h> | |
25 | #include <linux/proc_fs.h> | |
26 | #ifdef CONFIG_SYSCTL | |
27 | #include <linux/sysctl.h> | |
28 | #endif | |
29 | #include <linux/times.h> | |
30 | #include <net/neighbour.h> | |
31 | #include <net/dst.h> | |
32 | #include <net/sock.h> | |
33 | #include <linux/rtnetlink.h> | |
34 | #include <linux/random.h> | |
543537bd | 35 | #include <linux/string.h> |
1da177e4 LT |
36 | |
37 | #define NEIGH_DEBUG 1 | |
38 | ||
39 | #define NEIGH_PRINTK(x...) printk(x) | |
40 | #define NEIGH_NOPRINTK(x...) do { ; } while(0) | |
41 | #define NEIGH_PRINTK0 NEIGH_PRINTK | |
42 | #define NEIGH_PRINTK1 NEIGH_NOPRINTK | |
43 | #define NEIGH_PRINTK2 NEIGH_NOPRINTK | |
44 | ||
45 | #if NEIGH_DEBUG >= 1 | |
46 | #undef NEIGH_PRINTK1 | |
47 | #define NEIGH_PRINTK1 NEIGH_PRINTK | |
48 | #endif | |
49 | #if NEIGH_DEBUG >= 2 | |
50 | #undef NEIGH_PRINTK2 | |
51 | #define NEIGH_PRINTK2 NEIGH_PRINTK | |
52 | #endif | |
53 | ||
54 | #define PNEIGH_HASHMASK 0xF | |
55 | ||
56 | static void neigh_timer_handler(unsigned long arg); | |
57 | #ifdef CONFIG_ARPD | |
58 | static void neigh_app_notify(struct neighbour *n); | |
59 | #endif | |
60 | static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev); | |
61 | void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev); | |
62 | ||
63 | static struct neigh_table *neigh_tables; | |
64 | static struct file_operations neigh_stat_seq_fops; | |
65 | ||
66 | /* | |
67 | Neighbour hash table buckets are protected with rwlock tbl->lock. | |
68 | ||
69 | - All the scans/updates to hash buckets MUST be made under this lock. | |
70 | - NOTHING clever should be made under this lock: no callbacks | |
71 | to protocol backends, no attempts to send something to network. | |
72 | It will result in deadlocks, if backend/driver wants to use neighbour | |
73 | cache. | |
74 | - If the entry requires some non-trivial actions, increase | |
75 | its reference count and release table lock. | |
76 | ||
77 | Neighbour entries are protected: | |
78 | - with reference count. | |
79 | - with rwlock neigh->lock | |
80 | ||
81 | Reference count prevents destruction. | |
82 | ||
83 | neigh->lock mainly serializes ll address data and its validity state. | |
84 | However, the same lock is used to protect another entry fields: | |
85 | - timer | |
86 | - resolution queue | |
87 | ||
88 | Again, nothing clever shall be made under neigh->lock, | |
89 | the most complicated procedure, which we allow is dev->hard_header. | |
90 | It is supposed, that dev->hard_header is simplistic and does | |
91 | not make callbacks to neighbour tables. | |
92 | ||
93 | The last lock is neigh_tbl_lock. It is pure SMP lock, protecting | |
94 | list of neighbour tables. This list is used only in process context, | |
95 | */ | |
96 | ||
97 | static DEFINE_RWLOCK(neigh_tbl_lock); | |
98 | ||
99 | static int neigh_blackhole(struct sk_buff *skb) | |
100 | { | |
101 | kfree_skb(skb); | |
102 | return -ENETDOWN; | |
103 | } | |
104 | ||
105 | /* | |
106 | * It is random distribution in the interval (1/2)*base...(3/2)*base. | |
107 | * It corresponds to default IPv6 settings and is not overridable, | |
108 | * because it is really reasonable choice. | |
109 | */ | |
110 | ||
111 | unsigned long neigh_rand_reach_time(unsigned long base) | |
112 | { | |
113 | return (base ? (net_random() % base) + (base >> 1) : 0); | |
114 | } | |
115 | ||
116 | ||
117 | static int neigh_forced_gc(struct neigh_table *tbl) | |
118 | { | |
119 | int shrunk = 0; | |
120 | int i; | |
121 | ||
122 | NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs); | |
123 | ||
124 | write_lock_bh(&tbl->lock); | |
125 | for (i = 0; i <= tbl->hash_mask; i++) { | |
126 | struct neighbour *n, **np; | |
127 | ||
128 | np = &tbl->hash_buckets[i]; | |
129 | while ((n = *np) != NULL) { | |
130 | /* Neighbour record may be discarded if: | |
131 | * - nobody refers to it. | |
132 | * - it is not permanent | |
133 | */ | |
134 | write_lock(&n->lock); | |
135 | if (atomic_read(&n->refcnt) == 1 && | |
136 | !(n->nud_state & NUD_PERMANENT)) { | |
137 | *np = n->next; | |
138 | n->dead = 1; | |
139 | shrunk = 1; | |
140 | write_unlock(&n->lock); | |
141 | neigh_release(n); | |
142 | continue; | |
143 | } | |
144 | write_unlock(&n->lock); | |
145 | np = &n->next; | |
146 | } | |
147 | } | |
148 | ||
149 | tbl->last_flush = jiffies; | |
150 | ||
151 | write_unlock_bh(&tbl->lock); | |
152 | ||
153 | return shrunk; | |
154 | } | |
155 | ||
156 | static int neigh_del_timer(struct neighbour *n) | |
157 | { | |
158 | if ((n->nud_state & NUD_IN_TIMER) && | |
159 | del_timer(&n->timer)) { | |
160 | neigh_release(n); | |
161 | return 1; | |
162 | } | |
163 | return 0; | |
164 | } | |
165 | ||
166 | static void pneigh_queue_purge(struct sk_buff_head *list) | |
167 | { | |
168 | struct sk_buff *skb; | |
169 | ||
170 | while ((skb = skb_dequeue(list)) != NULL) { | |
171 | dev_put(skb->dev); | |
172 | kfree_skb(skb); | |
173 | } | |
174 | } | |
175 | ||
176 | void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev) | |
177 | { | |
178 | int i; | |
179 | ||
180 | write_lock_bh(&tbl->lock); | |
181 | ||
182 | for (i=0; i <= tbl->hash_mask; i++) { | |
183 | struct neighbour *n, **np; | |
184 | ||
185 | np = &tbl->hash_buckets[i]; | |
186 | while ((n = *np) != NULL) { | |
187 | if (dev && n->dev != dev) { | |
188 | np = &n->next; | |
189 | continue; | |
190 | } | |
191 | *np = n->next; | |
192 | write_lock_bh(&n->lock); | |
193 | n->dead = 1; | |
194 | neigh_del_timer(n); | |
195 | write_unlock_bh(&n->lock); | |
196 | neigh_release(n); | |
197 | } | |
198 | } | |
199 | ||
200 | write_unlock_bh(&tbl->lock); | |
201 | } | |
202 | ||
203 | int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev) | |
204 | { | |
205 | int i; | |
206 | ||
207 | write_lock_bh(&tbl->lock); | |
208 | ||
209 | for (i = 0; i <= tbl->hash_mask; i++) { | |
210 | struct neighbour *n, **np = &tbl->hash_buckets[i]; | |
211 | ||
212 | while ((n = *np) != NULL) { | |
213 | if (dev && n->dev != dev) { | |
214 | np = &n->next; | |
215 | continue; | |
216 | } | |
217 | *np = n->next; | |
218 | write_lock(&n->lock); | |
219 | neigh_del_timer(n); | |
220 | n->dead = 1; | |
221 | ||
222 | if (atomic_read(&n->refcnt) != 1) { | |
223 | /* The most unpleasant situation. | |
224 | We must destroy neighbour entry, | |
225 | but someone still uses it. | |
226 | ||
227 | The destroy will be delayed until | |
228 | the last user releases us, but | |
229 | we must kill timers etc. and move | |
230 | it to safe state. | |
231 | */ | |
232 | skb_queue_purge(&n->arp_queue); | |
233 | n->output = neigh_blackhole; | |
234 | if (n->nud_state & NUD_VALID) | |
235 | n->nud_state = NUD_NOARP; | |
236 | else | |
237 | n->nud_state = NUD_NONE; | |
238 | NEIGH_PRINTK2("neigh %p is stray.\n", n); | |
239 | } | |
240 | write_unlock(&n->lock); | |
241 | neigh_release(n); | |
242 | } | |
243 | } | |
244 | ||
245 | pneigh_ifdown(tbl, dev); | |
246 | write_unlock_bh(&tbl->lock); | |
247 | ||
248 | del_timer_sync(&tbl->proxy_timer); | |
249 | pneigh_queue_purge(&tbl->proxy_queue); | |
250 | return 0; | |
251 | } | |
252 | ||
253 | static struct neighbour *neigh_alloc(struct neigh_table *tbl) | |
254 | { | |
255 | struct neighbour *n = NULL; | |
256 | unsigned long now = jiffies; | |
257 | int entries; | |
258 | ||
259 | entries = atomic_inc_return(&tbl->entries) - 1; | |
260 | if (entries >= tbl->gc_thresh3 || | |
261 | (entries >= tbl->gc_thresh2 && | |
262 | time_after(now, tbl->last_flush + 5 * HZ))) { | |
263 | if (!neigh_forced_gc(tbl) && | |
264 | entries >= tbl->gc_thresh3) | |
265 | goto out_entries; | |
266 | } | |
267 | ||
268 | n = kmem_cache_alloc(tbl->kmem_cachep, SLAB_ATOMIC); | |
269 | if (!n) | |
270 | goto out_entries; | |
271 | ||
272 | memset(n, 0, tbl->entry_size); | |
273 | ||
274 | skb_queue_head_init(&n->arp_queue); | |
275 | rwlock_init(&n->lock); | |
276 | n->updated = n->used = now; | |
277 | n->nud_state = NUD_NONE; | |
278 | n->output = neigh_blackhole; | |
279 | n->parms = neigh_parms_clone(&tbl->parms); | |
280 | init_timer(&n->timer); | |
281 | n->timer.function = neigh_timer_handler; | |
282 | n->timer.data = (unsigned long)n; | |
283 | ||
284 | NEIGH_CACHE_STAT_INC(tbl, allocs); | |
285 | n->tbl = tbl; | |
286 | atomic_set(&n->refcnt, 1); | |
287 | n->dead = 1; | |
288 | out: | |
289 | return n; | |
290 | ||
291 | out_entries: | |
292 | atomic_dec(&tbl->entries); | |
293 | goto out; | |
294 | } | |
295 | ||
296 | static struct neighbour **neigh_hash_alloc(unsigned int entries) | |
297 | { | |
298 | unsigned long size = entries * sizeof(struct neighbour *); | |
299 | struct neighbour **ret; | |
300 | ||
301 | if (size <= PAGE_SIZE) { | |
302 | ret = kmalloc(size, GFP_ATOMIC); | |
303 | } else { | |
304 | ret = (struct neighbour **) | |
305 | __get_free_pages(GFP_ATOMIC, get_order(size)); | |
306 | } | |
307 | if (ret) | |
308 | memset(ret, 0, size); | |
309 | ||
310 | return ret; | |
311 | } | |
312 | ||
313 | static void neigh_hash_free(struct neighbour **hash, unsigned int entries) | |
314 | { | |
315 | unsigned long size = entries * sizeof(struct neighbour *); | |
316 | ||
317 | if (size <= PAGE_SIZE) | |
318 | kfree(hash); | |
319 | else | |
320 | free_pages((unsigned long)hash, get_order(size)); | |
321 | } | |
322 | ||
323 | static void neigh_hash_grow(struct neigh_table *tbl, unsigned long new_entries) | |
324 | { | |
325 | struct neighbour **new_hash, **old_hash; | |
326 | unsigned int i, new_hash_mask, old_entries; | |
327 | ||
328 | NEIGH_CACHE_STAT_INC(tbl, hash_grows); | |
329 | ||
330 | BUG_ON(new_entries & (new_entries - 1)); | |
331 | new_hash = neigh_hash_alloc(new_entries); | |
332 | if (!new_hash) | |
333 | return; | |
334 | ||
335 | old_entries = tbl->hash_mask + 1; | |
336 | new_hash_mask = new_entries - 1; | |
337 | old_hash = tbl->hash_buckets; | |
338 | ||
339 | get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd)); | |
340 | for (i = 0; i < old_entries; i++) { | |
341 | struct neighbour *n, *next; | |
342 | ||
343 | for (n = old_hash[i]; n; n = next) { | |
344 | unsigned int hash_val = tbl->hash(n->primary_key, n->dev); | |
345 | ||
346 | hash_val &= new_hash_mask; | |
347 | next = n->next; | |
348 | ||
349 | n->next = new_hash[hash_val]; | |
350 | new_hash[hash_val] = n; | |
351 | } | |
352 | } | |
353 | tbl->hash_buckets = new_hash; | |
354 | tbl->hash_mask = new_hash_mask; | |
355 | ||
356 | neigh_hash_free(old_hash, old_entries); | |
357 | } | |
358 | ||
359 | struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey, | |
360 | struct net_device *dev) | |
361 | { | |
362 | struct neighbour *n; | |
363 | int key_len = tbl->key_len; | |
364 | u32 hash_val = tbl->hash(pkey, dev) & tbl->hash_mask; | |
365 | ||
366 | NEIGH_CACHE_STAT_INC(tbl, lookups); | |
367 | ||
368 | read_lock_bh(&tbl->lock); | |
369 | for (n = tbl->hash_buckets[hash_val]; n; n = n->next) { | |
370 | if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) { | |
371 | neigh_hold(n); | |
372 | NEIGH_CACHE_STAT_INC(tbl, hits); | |
373 | break; | |
374 | } | |
375 | } | |
376 | read_unlock_bh(&tbl->lock); | |
377 | return n; | |
378 | } | |
379 | ||
380 | struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, const void *pkey) | |
381 | { | |
382 | struct neighbour *n; | |
383 | int key_len = tbl->key_len; | |
384 | u32 hash_val = tbl->hash(pkey, NULL) & tbl->hash_mask; | |
385 | ||
386 | NEIGH_CACHE_STAT_INC(tbl, lookups); | |
387 | ||
388 | read_lock_bh(&tbl->lock); | |
389 | for (n = tbl->hash_buckets[hash_val]; n; n = n->next) { | |
390 | if (!memcmp(n->primary_key, pkey, key_len)) { | |
391 | neigh_hold(n); | |
392 | NEIGH_CACHE_STAT_INC(tbl, hits); | |
393 | break; | |
394 | } | |
395 | } | |
396 | read_unlock_bh(&tbl->lock); | |
397 | return n; | |
398 | } | |
399 | ||
400 | struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey, | |
401 | struct net_device *dev) | |
402 | { | |
403 | u32 hash_val; | |
404 | int key_len = tbl->key_len; | |
405 | int error; | |
406 | struct neighbour *n1, *rc, *n = neigh_alloc(tbl); | |
407 | ||
408 | if (!n) { | |
409 | rc = ERR_PTR(-ENOBUFS); | |
410 | goto out; | |
411 | } | |
412 | ||
413 | memcpy(n->primary_key, pkey, key_len); | |
414 | n->dev = dev; | |
415 | dev_hold(dev); | |
416 | ||
417 | /* Protocol specific setup. */ | |
418 | if (tbl->constructor && (error = tbl->constructor(n)) < 0) { | |
419 | rc = ERR_PTR(error); | |
420 | goto out_neigh_release; | |
421 | } | |
422 | ||
423 | /* Device specific setup. */ | |
424 | if (n->parms->neigh_setup && | |
425 | (error = n->parms->neigh_setup(n)) < 0) { | |
426 | rc = ERR_PTR(error); | |
427 | goto out_neigh_release; | |
428 | } | |
429 | ||
430 | n->confirmed = jiffies - (n->parms->base_reachable_time << 1); | |
431 | ||
432 | write_lock_bh(&tbl->lock); | |
433 | ||
434 | if (atomic_read(&tbl->entries) > (tbl->hash_mask + 1)) | |
435 | neigh_hash_grow(tbl, (tbl->hash_mask + 1) << 1); | |
436 | ||
437 | hash_val = tbl->hash(pkey, dev) & tbl->hash_mask; | |
438 | ||
439 | if (n->parms->dead) { | |
440 | rc = ERR_PTR(-EINVAL); | |
441 | goto out_tbl_unlock; | |
442 | } | |
443 | ||
444 | for (n1 = tbl->hash_buckets[hash_val]; n1; n1 = n1->next) { | |
445 | if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) { | |
446 | neigh_hold(n1); | |
447 | rc = n1; | |
448 | goto out_tbl_unlock; | |
449 | } | |
450 | } | |
451 | ||
452 | n->next = tbl->hash_buckets[hash_val]; | |
453 | tbl->hash_buckets[hash_val] = n; | |
454 | n->dead = 0; | |
455 | neigh_hold(n); | |
456 | write_unlock_bh(&tbl->lock); | |
457 | NEIGH_PRINTK2("neigh %p is created.\n", n); | |
458 | rc = n; | |
459 | out: | |
460 | return rc; | |
461 | out_tbl_unlock: | |
462 | write_unlock_bh(&tbl->lock); | |
463 | out_neigh_release: | |
464 | neigh_release(n); | |
465 | goto out; | |
466 | } | |
467 | ||
468 | struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl, const void *pkey, | |
469 | struct net_device *dev, int creat) | |
470 | { | |
471 | struct pneigh_entry *n; | |
472 | int key_len = tbl->key_len; | |
473 | u32 hash_val = *(u32 *)(pkey + key_len - 4); | |
474 | ||
475 | hash_val ^= (hash_val >> 16); | |
476 | hash_val ^= hash_val >> 8; | |
477 | hash_val ^= hash_val >> 4; | |
478 | hash_val &= PNEIGH_HASHMASK; | |
479 | ||
480 | read_lock_bh(&tbl->lock); | |
481 | ||
482 | for (n = tbl->phash_buckets[hash_val]; n; n = n->next) { | |
483 | if (!memcmp(n->key, pkey, key_len) && | |
484 | (n->dev == dev || !n->dev)) { | |
485 | read_unlock_bh(&tbl->lock); | |
486 | goto out; | |
487 | } | |
488 | } | |
489 | read_unlock_bh(&tbl->lock); | |
490 | n = NULL; | |
491 | if (!creat) | |
492 | goto out; | |
493 | ||
494 | n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL); | |
495 | if (!n) | |
496 | goto out; | |
497 | ||
498 | memcpy(n->key, pkey, key_len); | |
499 | n->dev = dev; | |
500 | if (dev) | |
501 | dev_hold(dev); | |
502 | ||
503 | if (tbl->pconstructor && tbl->pconstructor(n)) { | |
504 | if (dev) | |
505 | dev_put(dev); | |
506 | kfree(n); | |
507 | n = NULL; | |
508 | goto out; | |
509 | } | |
510 | ||
511 | write_lock_bh(&tbl->lock); | |
512 | n->next = tbl->phash_buckets[hash_val]; | |
513 | tbl->phash_buckets[hash_val] = n; | |
514 | write_unlock_bh(&tbl->lock); | |
515 | out: | |
516 | return n; | |
517 | } | |
518 | ||
519 | ||
520 | int pneigh_delete(struct neigh_table *tbl, const void *pkey, | |
521 | struct net_device *dev) | |
522 | { | |
523 | struct pneigh_entry *n, **np; | |
524 | int key_len = tbl->key_len; | |
525 | u32 hash_val = *(u32 *)(pkey + key_len - 4); | |
526 | ||
527 | hash_val ^= (hash_val >> 16); | |
528 | hash_val ^= hash_val >> 8; | |
529 | hash_val ^= hash_val >> 4; | |
530 | hash_val &= PNEIGH_HASHMASK; | |
531 | ||
532 | write_lock_bh(&tbl->lock); | |
533 | for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL; | |
534 | np = &n->next) { | |
535 | if (!memcmp(n->key, pkey, key_len) && n->dev == dev) { | |
536 | *np = n->next; | |
537 | write_unlock_bh(&tbl->lock); | |
538 | if (tbl->pdestructor) | |
539 | tbl->pdestructor(n); | |
540 | if (n->dev) | |
541 | dev_put(n->dev); | |
542 | kfree(n); | |
543 | return 0; | |
544 | } | |
545 | } | |
546 | write_unlock_bh(&tbl->lock); | |
547 | return -ENOENT; | |
548 | } | |
549 | ||
550 | static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev) | |
551 | { | |
552 | struct pneigh_entry *n, **np; | |
553 | u32 h; | |
554 | ||
555 | for (h = 0; h <= PNEIGH_HASHMASK; h++) { | |
556 | np = &tbl->phash_buckets[h]; | |
557 | while ((n = *np) != NULL) { | |
558 | if (!dev || n->dev == dev) { | |
559 | *np = n->next; | |
560 | if (tbl->pdestructor) | |
561 | tbl->pdestructor(n); | |
562 | if (n->dev) | |
563 | dev_put(n->dev); | |
564 | kfree(n); | |
565 | continue; | |
566 | } | |
567 | np = &n->next; | |
568 | } | |
569 | } | |
570 | return -ENOENT; | |
571 | } | |
572 | ||
573 | ||
574 | /* | |
575 | * neighbour must already be out of the table; | |
576 | * | |
577 | */ | |
578 | void neigh_destroy(struct neighbour *neigh) | |
579 | { | |
580 | struct hh_cache *hh; | |
581 | ||
582 | NEIGH_CACHE_STAT_INC(neigh->tbl, destroys); | |
583 | ||
584 | if (!neigh->dead) { | |
585 | printk(KERN_WARNING | |
586 | "Destroying alive neighbour %p\n", neigh); | |
587 | dump_stack(); | |
588 | return; | |
589 | } | |
590 | ||
591 | if (neigh_del_timer(neigh)) | |
592 | printk(KERN_WARNING "Impossible event.\n"); | |
593 | ||
594 | while ((hh = neigh->hh) != NULL) { | |
595 | neigh->hh = hh->hh_next; | |
596 | hh->hh_next = NULL; | |
597 | write_lock_bh(&hh->hh_lock); | |
598 | hh->hh_output = neigh_blackhole; | |
599 | write_unlock_bh(&hh->hh_lock); | |
600 | if (atomic_dec_and_test(&hh->hh_refcnt)) | |
601 | kfree(hh); | |
602 | } | |
603 | ||
604 | if (neigh->ops && neigh->ops->destructor) | |
605 | (neigh->ops->destructor)(neigh); | |
606 | ||
607 | skb_queue_purge(&neigh->arp_queue); | |
608 | ||
609 | dev_put(neigh->dev); | |
610 | neigh_parms_put(neigh->parms); | |
611 | ||
612 | NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh); | |
613 | ||
614 | atomic_dec(&neigh->tbl->entries); | |
615 | kmem_cache_free(neigh->tbl->kmem_cachep, neigh); | |
616 | } | |
617 | ||
618 | /* Neighbour state is suspicious; | |
619 | disable fast path. | |
620 | ||
621 | Called with write_locked neigh. | |
622 | */ | |
623 | static void neigh_suspect(struct neighbour *neigh) | |
624 | { | |
625 | struct hh_cache *hh; | |
626 | ||
627 | NEIGH_PRINTK2("neigh %p is suspected.\n", neigh); | |
628 | ||
629 | neigh->output = neigh->ops->output; | |
630 | ||
631 | for (hh = neigh->hh; hh; hh = hh->hh_next) | |
632 | hh->hh_output = neigh->ops->output; | |
633 | } | |
634 | ||
635 | /* Neighbour state is OK; | |
636 | enable fast path. | |
637 | ||
638 | Called with write_locked neigh. | |
639 | */ | |
640 | static void neigh_connect(struct neighbour *neigh) | |
641 | { | |
642 | struct hh_cache *hh; | |
643 | ||
644 | NEIGH_PRINTK2("neigh %p is connected.\n", neigh); | |
645 | ||
646 | neigh->output = neigh->ops->connected_output; | |
647 | ||
648 | for (hh = neigh->hh; hh; hh = hh->hh_next) | |
649 | hh->hh_output = neigh->ops->hh_output; | |
650 | } | |
651 | ||
652 | static void neigh_periodic_timer(unsigned long arg) | |
653 | { | |
654 | struct neigh_table *tbl = (struct neigh_table *)arg; | |
655 | struct neighbour *n, **np; | |
656 | unsigned long expire, now = jiffies; | |
657 | ||
658 | NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs); | |
659 | ||
660 | write_lock(&tbl->lock); | |
661 | ||
662 | /* | |
663 | * periodically recompute ReachableTime from random function | |
664 | */ | |
665 | ||
666 | if (time_after(now, tbl->last_rand + 300 * HZ)) { | |
667 | struct neigh_parms *p; | |
668 | tbl->last_rand = now; | |
669 | for (p = &tbl->parms; p; p = p->next) | |
670 | p->reachable_time = | |
671 | neigh_rand_reach_time(p->base_reachable_time); | |
672 | } | |
673 | ||
674 | np = &tbl->hash_buckets[tbl->hash_chain_gc]; | |
675 | tbl->hash_chain_gc = ((tbl->hash_chain_gc + 1) & tbl->hash_mask); | |
676 | ||
677 | while ((n = *np) != NULL) { | |
678 | unsigned int state; | |
679 | ||
680 | write_lock(&n->lock); | |
681 | ||
682 | state = n->nud_state; | |
683 | if (state & (NUD_PERMANENT | NUD_IN_TIMER)) { | |
684 | write_unlock(&n->lock); | |
685 | goto next_elt; | |
686 | } | |
687 | ||
688 | if (time_before(n->used, n->confirmed)) | |
689 | n->used = n->confirmed; | |
690 | ||
691 | if (atomic_read(&n->refcnt) == 1 && | |
692 | (state == NUD_FAILED || | |
693 | time_after(now, n->used + n->parms->gc_staletime))) { | |
694 | *np = n->next; | |
695 | n->dead = 1; | |
696 | write_unlock(&n->lock); | |
697 | neigh_release(n); | |
698 | continue; | |
699 | } | |
700 | write_unlock(&n->lock); | |
701 | ||
702 | next_elt: | |
703 | np = &n->next; | |
704 | } | |
705 | ||
706 | /* Cycle through all hash buckets every base_reachable_time/2 ticks. | |
707 | * ARP entry timeouts range from 1/2 base_reachable_time to 3/2 | |
708 | * base_reachable_time. | |
709 | */ | |
710 | expire = tbl->parms.base_reachable_time >> 1; | |
711 | expire /= (tbl->hash_mask + 1); | |
712 | if (!expire) | |
713 | expire = 1; | |
714 | ||
715 | mod_timer(&tbl->gc_timer, now + expire); | |
716 | ||
717 | write_unlock(&tbl->lock); | |
718 | } | |
719 | ||
720 | static __inline__ int neigh_max_probes(struct neighbour *n) | |
721 | { | |
722 | struct neigh_parms *p = n->parms; | |
723 | return (n->nud_state & NUD_PROBE ? | |
724 | p->ucast_probes : | |
725 | p->ucast_probes + p->app_probes + p->mcast_probes); | |
726 | } | |
727 | ||
728 | ||
729 | /* Called when a timer expires for a neighbour entry. */ | |
730 | ||
731 | static void neigh_timer_handler(unsigned long arg) | |
732 | { | |
733 | unsigned long now, next; | |
734 | struct neighbour *neigh = (struct neighbour *)arg; | |
735 | unsigned state; | |
736 | int notify = 0; | |
737 | ||
738 | write_lock(&neigh->lock); | |
739 | ||
740 | state = neigh->nud_state; | |
741 | now = jiffies; | |
742 | next = now + HZ; | |
743 | ||
744 | if (!(state & NUD_IN_TIMER)) { | |
745 | #ifndef CONFIG_SMP | |
746 | printk(KERN_WARNING "neigh: timer & !nud_in_timer\n"); | |
747 | #endif | |
748 | goto out; | |
749 | } | |
750 | ||
751 | if (state & NUD_REACHABLE) { | |
752 | if (time_before_eq(now, | |
753 | neigh->confirmed + neigh->parms->reachable_time)) { | |
754 | NEIGH_PRINTK2("neigh %p is still alive.\n", neigh); | |
755 | next = neigh->confirmed + neigh->parms->reachable_time; | |
756 | } else if (time_before_eq(now, | |
757 | neigh->used + neigh->parms->delay_probe_time)) { | |
758 | NEIGH_PRINTK2("neigh %p is delayed.\n", neigh); | |
759 | neigh->nud_state = NUD_DELAY; | |
760 | neigh_suspect(neigh); | |
761 | next = now + neigh->parms->delay_probe_time; | |
762 | } else { | |
763 | NEIGH_PRINTK2("neigh %p is suspected.\n", neigh); | |
764 | neigh->nud_state = NUD_STALE; | |
765 | neigh_suspect(neigh); | |
766 | } | |
767 | } else if (state & NUD_DELAY) { | |
768 | if (time_before_eq(now, | |
769 | neigh->confirmed + neigh->parms->delay_probe_time)) { | |
770 | NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh); | |
771 | neigh->nud_state = NUD_REACHABLE; | |
772 | neigh_connect(neigh); | |
773 | next = neigh->confirmed + neigh->parms->reachable_time; | |
774 | } else { | |
775 | NEIGH_PRINTK2("neigh %p is probed.\n", neigh); | |
776 | neigh->nud_state = NUD_PROBE; | |
777 | atomic_set(&neigh->probes, 0); | |
778 | next = now + neigh->parms->retrans_time; | |
779 | } | |
780 | } else { | |
781 | /* NUD_PROBE|NUD_INCOMPLETE */ | |
782 | next = now + neigh->parms->retrans_time; | |
783 | } | |
784 | ||
785 | if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) && | |
786 | atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) { | |
787 | struct sk_buff *skb; | |
788 | ||
789 | neigh->nud_state = NUD_FAILED; | |
790 | notify = 1; | |
791 | NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed); | |
792 | NEIGH_PRINTK2("neigh %p is failed.\n", neigh); | |
793 | ||
794 | /* It is very thin place. report_unreachable is very complicated | |
795 | routine. Particularly, it can hit the same neighbour entry! | |
796 | ||
797 | So that, we try to be accurate and avoid dead loop. --ANK | |
798 | */ | |
799 | while (neigh->nud_state == NUD_FAILED && | |
800 | (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) { | |
801 | write_unlock(&neigh->lock); | |
802 | neigh->ops->error_report(neigh, skb); | |
803 | write_lock(&neigh->lock); | |
804 | } | |
805 | skb_queue_purge(&neigh->arp_queue); | |
806 | } | |
807 | ||
808 | if (neigh->nud_state & NUD_IN_TIMER) { | |
809 | neigh_hold(neigh); | |
810 | if (time_before(next, jiffies + HZ/2)) | |
811 | next = jiffies + HZ/2; | |
812 | neigh->timer.expires = next; | |
813 | add_timer(&neigh->timer); | |
814 | } | |
815 | if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) { | |
816 | struct sk_buff *skb = skb_peek(&neigh->arp_queue); | |
817 | /* keep skb alive even if arp_queue overflows */ | |
818 | if (skb) | |
819 | skb_get(skb); | |
820 | write_unlock(&neigh->lock); | |
821 | neigh->ops->solicit(neigh, skb); | |
822 | atomic_inc(&neigh->probes); | |
823 | if (skb) | |
824 | kfree_skb(skb); | |
825 | } else { | |
826 | out: | |
827 | write_unlock(&neigh->lock); | |
828 | } | |
829 | ||
830 | #ifdef CONFIG_ARPD | |
831 | if (notify && neigh->parms->app_probes) | |
832 | neigh_app_notify(neigh); | |
833 | #endif | |
834 | neigh_release(neigh); | |
835 | } | |
836 | ||
837 | int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb) | |
838 | { | |
839 | int rc; | |
840 | unsigned long now; | |
841 | ||
842 | write_lock_bh(&neigh->lock); | |
843 | ||
844 | rc = 0; | |
845 | if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE)) | |
846 | goto out_unlock_bh; | |
847 | ||
848 | now = jiffies; | |
849 | ||
850 | if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) { | |
851 | if (neigh->parms->mcast_probes + neigh->parms->app_probes) { | |
852 | atomic_set(&neigh->probes, neigh->parms->ucast_probes); | |
853 | neigh->nud_state = NUD_INCOMPLETE; | |
854 | neigh_hold(neigh); | |
855 | neigh->timer.expires = now + 1; | |
856 | add_timer(&neigh->timer); | |
857 | } else { | |
858 | neigh->nud_state = NUD_FAILED; | |
859 | write_unlock_bh(&neigh->lock); | |
860 | ||
861 | if (skb) | |
862 | kfree_skb(skb); | |
863 | return 1; | |
864 | } | |
865 | } else if (neigh->nud_state & NUD_STALE) { | |
866 | NEIGH_PRINTK2("neigh %p is delayed.\n", neigh); | |
867 | neigh_hold(neigh); | |
868 | neigh->nud_state = NUD_DELAY; | |
869 | neigh->timer.expires = jiffies + neigh->parms->delay_probe_time; | |
870 | add_timer(&neigh->timer); | |
871 | } | |
872 | ||
873 | if (neigh->nud_state == NUD_INCOMPLETE) { | |
874 | if (skb) { | |
875 | if (skb_queue_len(&neigh->arp_queue) >= | |
876 | neigh->parms->queue_len) { | |
877 | struct sk_buff *buff; | |
878 | buff = neigh->arp_queue.next; | |
879 | __skb_unlink(buff, &neigh->arp_queue); | |
880 | kfree_skb(buff); | |
881 | } | |
882 | __skb_queue_tail(&neigh->arp_queue, skb); | |
883 | } | |
884 | rc = 1; | |
885 | } | |
886 | out_unlock_bh: | |
887 | write_unlock_bh(&neigh->lock); | |
888 | return rc; | |
889 | } | |
890 | ||
891 | static __inline__ void neigh_update_hhs(struct neighbour *neigh) | |
892 | { | |
893 | struct hh_cache *hh; | |
894 | void (*update)(struct hh_cache*, struct net_device*, unsigned char *) = | |
895 | neigh->dev->header_cache_update; | |
896 | ||
897 | if (update) { | |
898 | for (hh = neigh->hh; hh; hh = hh->hh_next) { | |
899 | write_lock_bh(&hh->hh_lock); | |
900 | update(hh, neigh->dev, neigh->ha); | |
901 | write_unlock_bh(&hh->hh_lock); | |
902 | } | |
903 | } | |
904 | } | |
905 | ||
906 | ||
907 | ||
908 | /* Generic update routine. | |
909 | -- lladdr is new lladdr or NULL, if it is not supplied. | |
910 | -- new is new state. | |
911 | -- flags | |
912 | NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr, | |
913 | if it is different. | |
914 | NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected" | |
915 | lladdr instead of overriding it | |
916 | if it is different. | |
917 | It also allows to retain current state | |
918 | if lladdr is unchanged. | |
919 | NEIGH_UPDATE_F_ADMIN means that the change is administrative. | |
920 | ||
921 | NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing | |
922 | NTF_ROUTER flag. | |
923 | NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as | |
924 | a router. | |
925 | ||
926 | Caller MUST hold reference count on the entry. | |
927 | */ | |
928 | ||
929 | int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, | |
930 | u32 flags) | |
931 | { | |
932 | u8 old; | |
933 | int err; | |
934 | #ifdef CONFIG_ARPD | |
935 | int notify = 0; | |
936 | #endif | |
937 | struct net_device *dev; | |
938 | int update_isrouter = 0; | |
939 | ||
940 | write_lock_bh(&neigh->lock); | |
941 | ||
942 | dev = neigh->dev; | |
943 | old = neigh->nud_state; | |
944 | err = -EPERM; | |
945 | ||
946 | if (!(flags & NEIGH_UPDATE_F_ADMIN) && | |
947 | (old & (NUD_NOARP | NUD_PERMANENT))) | |
948 | goto out; | |
949 | ||
950 | if (!(new & NUD_VALID)) { | |
951 | neigh_del_timer(neigh); | |
952 | if (old & NUD_CONNECTED) | |
953 | neigh_suspect(neigh); | |
954 | neigh->nud_state = new; | |
955 | err = 0; | |
956 | #ifdef CONFIG_ARPD | |
957 | notify = old & NUD_VALID; | |
958 | #endif | |
959 | goto out; | |
960 | } | |
961 | ||
962 | /* Compare new lladdr with cached one */ | |
963 | if (!dev->addr_len) { | |
964 | /* First case: device needs no address. */ | |
965 | lladdr = neigh->ha; | |
966 | } else if (lladdr) { | |
967 | /* The second case: if something is already cached | |
968 | and a new address is proposed: | |
969 | - compare new & old | |
970 | - if they are different, check override flag | |
971 | */ | |
972 | if ((old & NUD_VALID) && | |
973 | !memcmp(lladdr, neigh->ha, dev->addr_len)) | |
974 | lladdr = neigh->ha; | |
975 | } else { | |
976 | /* No address is supplied; if we know something, | |
977 | use it, otherwise discard the request. | |
978 | */ | |
979 | err = -EINVAL; | |
980 | if (!(old & NUD_VALID)) | |
981 | goto out; | |
982 | lladdr = neigh->ha; | |
983 | } | |
984 | ||
985 | if (new & NUD_CONNECTED) | |
986 | neigh->confirmed = jiffies; | |
987 | neigh->updated = jiffies; | |
988 | ||
989 | /* If entry was valid and address is not changed, | |
990 | do not change entry state, if new one is STALE. | |
991 | */ | |
992 | err = 0; | |
993 | update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER; | |
994 | if (old & NUD_VALID) { | |
995 | if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) { | |
996 | update_isrouter = 0; | |
997 | if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) && | |
998 | (old & NUD_CONNECTED)) { | |
999 | lladdr = neigh->ha; | |
1000 | new = NUD_STALE; | |
1001 | } else | |
1002 | goto out; | |
1003 | } else { | |
1004 | if (lladdr == neigh->ha && new == NUD_STALE && | |
1005 | ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) || | |
1006 | (old & NUD_CONNECTED)) | |
1007 | ) | |
1008 | new = old; | |
1009 | } | |
1010 | } | |
1011 | ||
1012 | if (new != old) { | |
1013 | neigh_del_timer(neigh); | |
1014 | if (new & NUD_IN_TIMER) { | |
1015 | neigh_hold(neigh); | |
1016 | neigh->timer.expires = jiffies + | |
1017 | ((new & NUD_REACHABLE) ? | |
1018 | neigh->parms->reachable_time : 0); | |
1019 | add_timer(&neigh->timer); | |
1020 | } | |
1021 | neigh->nud_state = new; | |
1022 | } | |
1023 | ||
1024 | if (lladdr != neigh->ha) { | |
1025 | memcpy(&neigh->ha, lladdr, dev->addr_len); | |
1026 | neigh_update_hhs(neigh); | |
1027 | if (!(new & NUD_CONNECTED)) | |
1028 | neigh->confirmed = jiffies - | |
1029 | (neigh->parms->base_reachable_time << 1); | |
1030 | #ifdef CONFIG_ARPD | |
1031 | notify = 1; | |
1032 | #endif | |
1033 | } | |
1034 | if (new == old) | |
1035 | goto out; | |
1036 | if (new & NUD_CONNECTED) | |
1037 | neigh_connect(neigh); | |
1038 | else | |
1039 | neigh_suspect(neigh); | |
1040 | if (!(old & NUD_VALID)) { | |
1041 | struct sk_buff *skb; | |
1042 | ||
1043 | /* Again: avoid dead loop if something went wrong */ | |
1044 | ||
1045 | while (neigh->nud_state & NUD_VALID && | |
1046 | (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) { | |
1047 | struct neighbour *n1 = neigh; | |
1048 | write_unlock_bh(&neigh->lock); | |
1049 | /* On shaper/eql skb->dst->neighbour != neigh :( */ | |
1050 | if (skb->dst && skb->dst->neighbour) | |
1051 | n1 = skb->dst->neighbour; | |
1052 | n1->output(skb); | |
1053 | write_lock_bh(&neigh->lock); | |
1054 | } | |
1055 | skb_queue_purge(&neigh->arp_queue); | |
1056 | } | |
1057 | out: | |
1058 | if (update_isrouter) { | |
1059 | neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ? | |
1060 | (neigh->flags | NTF_ROUTER) : | |
1061 | (neigh->flags & ~NTF_ROUTER); | |
1062 | } | |
1063 | write_unlock_bh(&neigh->lock); | |
1064 | #ifdef CONFIG_ARPD | |
1065 | if (notify && neigh->parms->app_probes) | |
1066 | neigh_app_notify(neigh); | |
1067 | #endif | |
1068 | return err; | |
1069 | } | |
1070 | ||
1071 | struct neighbour *neigh_event_ns(struct neigh_table *tbl, | |
1072 | u8 *lladdr, void *saddr, | |
1073 | struct net_device *dev) | |
1074 | { | |
1075 | struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev, | |
1076 | lladdr || !dev->addr_len); | |
1077 | if (neigh) | |
1078 | neigh_update(neigh, lladdr, NUD_STALE, | |
1079 | NEIGH_UPDATE_F_OVERRIDE); | |
1080 | return neigh; | |
1081 | } | |
1082 | ||
1083 | static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst, | |
1084 | u16 protocol) | |
1085 | { | |
1086 | struct hh_cache *hh; | |
1087 | struct net_device *dev = dst->dev; | |
1088 | ||
1089 | for (hh = n->hh; hh; hh = hh->hh_next) | |
1090 | if (hh->hh_type == protocol) | |
1091 | break; | |
1092 | ||
1093 | if (!hh && (hh = kmalloc(sizeof(*hh), GFP_ATOMIC)) != NULL) { | |
1094 | memset(hh, 0, sizeof(struct hh_cache)); | |
1095 | rwlock_init(&hh->hh_lock); | |
1096 | hh->hh_type = protocol; | |
1097 | atomic_set(&hh->hh_refcnt, 0); | |
1098 | hh->hh_next = NULL; | |
1099 | if (dev->hard_header_cache(n, hh)) { | |
1100 | kfree(hh); | |
1101 | hh = NULL; | |
1102 | } else { | |
1103 | atomic_inc(&hh->hh_refcnt); | |
1104 | hh->hh_next = n->hh; | |
1105 | n->hh = hh; | |
1106 | if (n->nud_state & NUD_CONNECTED) | |
1107 | hh->hh_output = n->ops->hh_output; | |
1108 | else | |
1109 | hh->hh_output = n->ops->output; | |
1110 | } | |
1111 | } | |
1112 | if (hh) { | |
1113 | atomic_inc(&hh->hh_refcnt); | |
1114 | dst->hh = hh; | |
1115 | } | |
1116 | } | |
1117 | ||
1118 | /* This function can be used in contexts, where only old dev_queue_xmit | |
1119 | worked, f.e. if you want to override normal output path (eql, shaper), | |
1120 | but resolution is not made yet. | |
1121 | */ | |
1122 | ||
1123 | int neigh_compat_output(struct sk_buff *skb) | |
1124 | { | |
1125 | struct net_device *dev = skb->dev; | |
1126 | ||
1127 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1128 | ||
1129 | if (dev->hard_header && | |
1130 | dev->hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL, | |
1131 | skb->len) < 0 && | |
1132 | dev->rebuild_header(skb)) | |
1133 | return 0; | |
1134 | ||
1135 | return dev_queue_xmit(skb); | |
1136 | } | |
1137 | ||
1138 | /* Slow and careful. */ | |
1139 | ||
1140 | int neigh_resolve_output(struct sk_buff *skb) | |
1141 | { | |
1142 | struct dst_entry *dst = skb->dst; | |
1143 | struct neighbour *neigh; | |
1144 | int rc = 0; | |
1145 | ||
1146 | if (!dst || !(neigh = dst->neighbour)) | |
1147 | goto discard; | |
1148 | ||
1149 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1150 | ||
1151 | if (!neigh_event_send(neigh, skb)) { | |
1152 | int err; | |
1153 | struct net_device *dev = neigh->dev; | |
1154 | if (dev->hard_header_cache && !dst->hh) { | |
1155 | write_lock_bh(&neigh->lock); | |
1156 | if (!dst->hh) | |
1157 | neigh_hh_init(neigh, dst, dst->ops->protocol); | |
1158 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1159 | neigh->ha, NULL, skb->len); | |
1160 | write_unlock_bh(&neigh->lock); | |
1161 | } else { | |
1162 | read_lock_bh(&neigh->lock); | |
1163 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1164 | neigh->ha, NULL, skb->len); | |
1165 | read_unlock_bh(&neigh->lock); | |
1166 | } | |
1167 | if (err >= 0) | |
1168 | rc = neigh->ops->queue_xmit(skb); | |
1169 | else | |
1170 | goto out_kfree_skb; | |
1171 | } | |
1172 | out: | |
1173 | return rc; | |
1174 | discard: | |
1175 | NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n", | |
1176 | dst, dst ? dst->neighbour : NULL); | |
1177 | out_kfree_skb: | |
1178 | rc = -EINVAL; | |
1179 | kfree_skb(skb); | |
1180 | goto out; | |
1181 | } | |
1182 | ||
1183 | /* As fast as possible without hh cache */ | |
1184 | ||
1185 | int neigh_connected_output(struct sk_buff *skb) | |
1186 | { | |
1187 | int err; | |
1188 | struct dst_entry *dst = skb->dst; | |
1189 | struct neighbour *neigh = dst->neighbour; | |
1190 | struct net_device *dev = neigh->dev; | |
1191 | ||
1192 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1193 | ||
1194 | read_lock_bh(&neigh->lock); | |
1195 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1196 | neigh->ha, NULL, skb->len); | |
1197 | read_unlock_bh(&neigh->lock); | |
1198 | if (err >= 0) | |
1199 | err = neigh->ops->queue_xmit(skb); | |
1200 | else { | |
1201 | err = -EINVAL; | |
1202 | kfree_skb(skb); | |
1203 | } | |
1204 | return err; | |
1205 | } | |
1206 | ||
1207 | static void neigh_proxy_process(unsigned long arg) | |
1208 | { | |
1209 | struct neigh_table *tbl = (struct neigh_table *)arg; | |
1210 | long sched_next = 0; | |
1211 | unsigned long now = jiffies; | |
1212 | struct sk_buff *skb; | |
1213 | ||
1214 | spin_lock(&tbl->proxy_queue.lock); | |
1215 | ||
1216 | skb = tbl->proxy_queue.next; | |
1217 | ||
1218 | while (skb != (struct sk_buff *)&tbl->proxy_queue) { | |
1219 | struct sk_buff *back = skb; | |
a61bbcf2 | 1220 | long tdif = NEIGH_CB(back)->sched_next - now; |
1da177e4 LT |
1221 | |
1222 | skb = skb->next; | |
1223 | if (tdif <= 0) { | |
1224 | struct net_device *dev = back->dev; | |
1225 | __skb_unlink(back, &tbl->proxy_queue); | |
1226 | if (tbl->proxy_redo && netif_running(dev)) | |
1227 | tbl->proxy_redo(back); | |
1228 | else | |
1229 | kfree_skb(back); | |
1230 | ||
1231 | dev_put(dev); | |
1232 | } else if (!sched_next || tdif < sched_next) | |
1233 | sched_next = tdif; | |
1234 | } | |
1235 | del_timer(&tbl->proxy_timer); | |
1236 | if (sched_next) | |
1237 | mod_timer(&tbl->proxy_timer, jiffies + sched_next); | |
1238 | spin_unlock(&tbl->proxy_queue.lock); | |
1239 | } | |
1240 | ||
1241 | void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p, | |
1242 | struct sk_buff *skb) | |
1243 | { | |
1244 | unsigned long now = jiffies; | |
1245 | unsigned long sched_next = now + (net_random() % p->proxy_delay); | |
1246 | ||
1247 | if (tbl->proxy_queue.qlen > p->proxy_qlen) { | |
1248 | kfree_skb(skb); | |
1249 | return; | |
1250 | } | |
a61bbcf2 PM |
1251 | |
1252 | NEIGH_CB(skb)->sched_next = sched_next; | |
1253 | NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED; | |
1da177e4 LT |
1254 | |
1255 | spin_lock(&tbl->proxy_queue.lock); | |
1256 | if (del_timer(&tbl->proxy_timer)) { | |
1257 | if (time_before(tbl->proxy_timer.expires, sched_next)) | |
1258 | sched_next = tbl->proxy_timer.expires; | |
1259 | } | |
1260 | dst_release(skb->dst); | |
1261 | skb->dst = NULL; | |
1262 | dev_hold(skb->dev); | |
1263 | __skb_queue_tail(&tbl->proxy_queue, skb); | |
1264 | mod_timer(&tbl->proxy_timer, sched_next); | |
1265 | spin_unlock(&tbl->proxy_queue.lock); | |
1266 | } | |
1267 | ||
1268 | ||
1269 | struct neigh_parms *neigh_parms_alloc(struct net_device *dev, | |
1270 | struct neigh_table *tbl) | |
1271 | { | |
1272 | struct neigh_parms *p = kmalloc(sizeof(*p), GFP_KERNEL); | |
1273 | ||
1274 | if (p) { | |
1275 | memcpy(p, &tbl->parms, sizeof(*p)); | |
1276 | p->tbl = tbl; | |
1277 | atomic_set(&p->refcnt, 1); | |
1278 | INIT_RCU_HEAD(&p->rcu_head); | |
1279 | p->reachable_time = | |
1280 | neigh_rand_reach_time(p->base_reachable_time); | |
c7fb64db TG |
1281 | if (dev) { |
1282 | if (dev->neigh_setup && dev->neigh_setup(dev, p)) { | |
1283 | kfree(p); | |
1284 | return NULL; | |
1285 | } | |
1286 | ||
1287 | dev_hold(dev); | |
1288 | p->dev = dev; | |
1da177e4 LT |
1289 | } |
1290 | p->sysctl_table = NULL; | |
1291 | write_lock_bh(&tbl->lock); | |
1292 | p->next = tbl->parms.next; | |
1293 | tbl->parms.next = p; | |
1294 | write_unlock_bh(&tbl->lock); | |
1295 | } | |
1296 | return p; | |
1297 | } | |
1298 | ||
1299 | static void neigh_rcu_free_parms(struct rcu_head *head) | |
1300 | { | |
1301 | struct neigh_parms *parms = | |
1302 | container_of(head, struct neigh_parms, rcu_head); | |
1303 | ||
1304 | neigh_parms_put(parms); | |
1305 | } | |
1306 | ||
1307 | void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms) | |
1308 | { | |
1309 | struct neigh_parms **p; | |
1310 | ||
1311 | if (!parms || parms == &tbl->parms) | |
1312 | return; | |
1313 | write_lock_bh(&tbl->lock); | |
1314 | for (p = &tbl->parms.next; *p; p = &(*p)->next) { | |
1315 | if (*p == parms) { | |
1316 | *p = parms->next; | |
1317 | parms->dead = 1; | |
1318 | write_unlock_bh(&tbl->lock); | |
c7fb64db TG |
1319 | if (parms->dev) |
1320 | dev_put(parms->dev); | |
1da177e4 LT |
1321 | call_rcu(&parms->rcu_head, neigh_rcu_free_parms); |
1322 | return; | |
1323 | } | |
1324 | } | |
1325 | write_unlock_bh(&tbl->lock); | |
1326 | NEIGH_PRINTK1("neigh_parms_release: not found\n"); | |
1327 | } | |
1328 | ||
1329 | void neigh_parms_destroy(struct neigh_parms *parms) | |
1330 | { | |
1331 | kfree(parms); | |
1332 | } | |
1333 | ||
1334 | ||
1335 | void neigh_table_init(struct neigh_table *tbl) | |
1336 | { | |
1337 | unsigned long now = jiffies; | |
1338 | unsigned long phsize; | |
1339 | ||
1340 | atomic_set(&tbl->parms.refcnt, 1); | |
1341 | INIT_RCU_HEAD(&tbl->parms.rcu_head); | |
1342 | tbl->parms.reachable_time = | |
1343 | neigh_rand_reach_time(tbl->parms.base_reachable_time); | |
1344 | ||
1345 | if (!tbl->kmem_cachep) | |
1346 | tbl->kmem_cachep = kmem_cache_create(tbl->id, | |
1347 | tbl->entry_size, | |
1348 | 0, SLAB_HWCACHE_ALIGN, | |
1349 | NULL, NULL); | |
1350 | ||
1351 | if (!tbl->kmem_cachep) | |
1352 | panic("cannot create neighbour cache"); | |
1353 | ||
1354 | tbl->stats = alloc_percpu(struct neigh_statistics); | |
1355 | if (!tbl->stats) | |
1356 | panic("cannot create neighbour cache statistics"); | |
1357 | ||
1358 | #ifdef CONFIG_PROC_FS | |
1359 | tbl->pde = create_proc_entry(tbl->id, 0, proc_net_stat); | |
1360 | if (!tbl->pde) | |
1361 | panic("cannot create neighbour proc dir entry"); | |
1362 | tbl->pde->proc_fops = &neigh_stat_seq_fops; | |
1363 | tbl->pde->data = tbl; | |
1364 | #endif | |
1365 | ||
1366 | tbl->hash_mask = 1; | |
1367 | tbl->hash_buckets = neigh_hash_alloc(tbl->hash_mask + 1); | |
1368 | ||
1369 | phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *); | |
1370 | tbl->phash_buckets = kmalloc(phsize, GFP_KERNEL); | |
1371 | ||
1372 | if (!tbl->hash_buckets || !tbl->phash_buckets) | |
1373 | panic("cannot allocate neighbour cache hashes"); | |
1374 | ||
1375 | memset(tbl->phash_buckets, 0, phsize); | |
1376 | ||
1377 | get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd)); | |
1378 | ||
1379 | rwlock_init(&tbl->lock); | |
1380 | init_timer(&tbl->gc_timer); | |
1381 | tbl->gc_timer.data = (unsigned long)tbl; | |
1382 | tbl->gc_timer.function = neigh_periodic_timer; | |
1383 | tbl->gc_timer.expires = now + 1; | |
1384 | add_timer(&tbl->gc_timer); | |
1385 | ||
1386 | init_timer(&tbl->proxy_timer); | |
1387 | tbl->proxy_timer.data = (unsigned long)tbl; | |
1388 | tbl->proxy_timer.function = neigh_proxy_process; | |
1389 | skb_queue_head_init(&tbl->proxy_queue); | |
1390 | ||
1391 | tbl->last_flush = now; | |
1392 | tbl->last_rand = now + tbl->parms.reachable_time * 20; | |
1393 | write_lock(&neigh_tbl_lock); | |
1394 | tbl->next = neigh_tables; | |
1395 | neigh_tables = tbl; | |
1396 | write_unlock(&neigh_tbl_lock); | |
1397 | } | |
1398 | ||
1399 | int neigh_table_clear(struct neigh_table *tbl) | |
1400 | { | |
1401 | struct neigh_table **tp; | |
1402 | ||
1403 | /* It is not clean... Fix it to unload IPv6 module safely */ | |
1404 | del_timer_sync(&tbl->gc_timer); | |
1405 | del_timer_sync(&tbl->proxy_timer); | |
1406 | pneigh_queue_purge(&tbl->proxy_queue); | |
1407 | neigh_ifdown(tbl, NULL); | |
1408 | if (atomic_read(&tbl->entries)) | |
1409 | printk(KERN_CRIT "neighbour leakage\n"); | |
1410 | write_lock(&neigh_tbl_lock); | |
1411 | for (tp = &neigh_tables; *tp; tp = &(*tp)->next) { | |
1412 | if (*tp == tbl) { | |
1413 | *tp = tbl->next; | |
1414 | break; | |
1415 | } | |
1416 | } | |
1417 | write_unlock(&neigh_tbl_lock); | |
1418 | ||
1419 | neigh_hash_free(tbl->hash_buckets, tbl->hash_mask + 1); | |
1420 | tbl->hash_buckets = NULL; | |
1421 | ||
1422 | kfree(tbl->phash_buckets); | |
1423 | tbl->phash_buckets = NULL; | |
1424 | ||
1425 | return 0; | |
1426 | } | |
1427 | ||
1428 | int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1429 | { | |
1430 | struct ndmsg *ndm = NLMSG_DATA(nlh); | |
1431 | struct rtattr **nda = arg; | |
1432 | struct neigh_table *tbl; | |
1433 | struct net_device *dev = NULL; | |
1434 | int err = -ENODEV; | |
1435 | ||
1436 | if (ndm->ndm_ifindex && | |
1437 | (dev = dev_get_by_index(ndm->ndm_ifindex)) == NULL) | |
1438 | goto out; | |
1439 | ||
1440 | read_lock(&neigh_tbl_lock); | |
1441 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1442 | struct rtattr *dst_attr = nda[NDA_DST - 1]; | |
1443 | struct neighbour *n; | |
1444 | ||
1445 | if (tbl->family != ndm->ndm_family) | |
1446 | continue; | |
1447 | read_unlock(&neigh_tbl_lock); | |
1448 | ||
1449 | err = -EINVAL; | |
1450 | if (!dst_attr || RTA_PAYLOAD(dst_attr) < tbl->key_len) | |
1451 | goto out_dev_put; | |
1452 | ||
1453 | if (ndm->ndm_flags & NTF_PROXY) { | |
1454 | err = pneigh_delete(tbl, RTA_DATA(dst_attr), dev); | |
1455 | goto out_dev_put; | |
1456 | } | |
1457 | ||
1458 | if (!dev) | |
1459 | goto out; | |
1460 | ||
1461 | n = neigh_lookup(tbl, RTA_DATA(dst_attr), dev); | |
1462 | if (n) { | |
1463 | err = neigh_update(n, NULL, NUD_FAILED, | |
1464 | NEIGH_UPDATE_F_OVERRIDE| | |
1465 | NEIGH_UPDATE_F_ADMIN); | |
1466 | neigh_release(n); | |
1467 | } | |
1468 | goto out_dev_put; | |
1469 | } | |
1470 | read_unlock(&neigh_tbl_lock); | |
1471 | err = -EADDRNOTAVAIL; | |
1472 | out_dev_put: | |
1473 | if (dev) | |
1474 | dev_put(dev); | |
1475 | out: | |
1476 | return err; | |
1477 | } | |
1478 | ||
1479 | int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1480 | { | |
1481 | struct ndmsg *ndm = NLMSG_DATA(nlh); | |
1482 | struct rtattr **nda = arg; | |
1483 | struct neigh_table *tbl; | |
1484 | struct net_device *dev = NULL; | |
1485 | int err = -ENODEV; | |
1486 | ||
1487 | if (ndm->ndm_ifindex && | |
1488 | (dev = dev_get_by_index(ndm->ndm_ifindex)) == NULL) | |
1489 | goto out; | |
1490 | ||
1491 | read_lock(&neigh_tbl_lock); | |
1492 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1493 | struct rtattr *lladdr_attr = nda[NDA_LLADDR - 1]; | |
1494 | struct rtattr *dst_attr = nda[NDA_DST - 1]; | |
1495 | int override = 1; | |
1496 | struct neighbour *n; | |
1497 | ||
1498 | if (tbl->family != ndm->ndm_family) | |
1499 | continue; | |
1500 | read_unlock(&neigh_tbl_lock); | |
1501 | ||
1502 | err = -EINVAL; | |
1503 | if (!dst_attr || RTA_PAYLOAD(dst_attr) < tbl->key_len) | |
1504 | goto out_dev_put; | |
1505 | ||
1506 | if (ndm->ndm_flags & NTF_PROXY) { | |
1507 | err = -ENOBUFS; | |
1508 | if (pneigh_lookup(tbl, RTA_DATA(dst_attr), dev, 1)) | |
1509 | err = 0; | |
1510 | goto out_dev_put; | |
1511 | } | |
1512 | ||
1513 | err = -EINVAL; | |
1514 | if (!dev) | |
1515 | goto out; | |
1516 | if (lladdr_attr && RTA_PAYLOAD(lladdr_attr) < dev->addr_len) | |
1517 | goto out_dev_put; | |
1518 | ||
1519 | n = neigh_lookup(tbl, RTA_DATA(dst_attr), dev); | |
1520 | if (n) { | |
1521 | if (nlh->nlmsg_flags & NLM_F_EXCL) { | |
1522 | err = -EEXIST; | |
1523 | neigh_release(n); | |
1524 | goto out_dev_put; | |
1525 | } | |
1526 | ||
1527 | override = nlh->nlmsg_flags & NLM_F_REPLACE; | |
1528 | } else if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { | |
1529 | err = -ENOENT; | |
1530 | goto out_dev_put; | |
1531 | } else { | |
1532 | n = __neigh_lookup_errno(tbl, RTA_DATA(dst_attr), dev); | |
1533 | if (IS_ERR(n)) { | |
1534 | err = PTR_ERR(n); | |
1535 | goto out_dev_put; | |
1536 | } | |
1537 | } | |
1538 | ||
1539 | err = neigh_update(n, | |
1540 | lladdr_attr ? RTA_DATA(lladdr_attr) : NULL, | |
1541 | ndm->ndm_state, | |
1542 | (override ? NEIGH_UPDATE_F_OVERRIDE : 0) | | |
1543 | NEIGH_UPDATE_F_ADMIN); | |
1544 | ||
1545 | neigh_release(n); | |
1546 | goto out_dev_put; | |
1547 | } | |
1548 | ||
1549 | read_unlock(&neigh_tbl_lock); | |
1550 | err = -EADDRNOTAVAIL; | |
1551 | out_dev_put: | |
1552 | if (dev) | |
1553 | dev_put(dev); | |
1554 | out: | |
1555 | return err; | |
1556 | } | |
1557 | ||
c7fb64db TG |
1558 | static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms) |
1559 | { | |
e386c6eb TG |
1560 | struct rtattr *nest = NULL; |
1561 | ||
1562 | nest = RTA_NEST(skb, NDTA_PARMS); | |
c7fb64db TG |
1563 | |
1564 | if (parms->dev) | |
1565 | RTA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex); | |
1566 | ||
1567 | RTA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt)); | |
1568 | RTA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len); | |
1569 | RTA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen); | |
1570 | RTA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes); | |
1571 | RTA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes); | |
1572 | RTA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes); | |
1573 | RTA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time); | |
1574 | RTA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME, | |
1575 | parms->base_reachable_time); | |
1576 | RTA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime); | |
1577 | RTA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time); | |
1578 | RTA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time); | |
1579 | RTA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay); | |
1580 | RTA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay); | |
1581 | RTA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime); | |
1582 | ||
1583 | return RTA_NEST_END(skb, nest); | |
1584 | ||
1585 | rtattr_failure: | |
1586 | return RTA_NEST_CANCEL(skb, nest); | |
1587 | } | |
1588 | ||
1589 | static int neightbl_fill_info(struct neigh_table *tbl, struct sk_buff *skb, | |
1590 | struct netlink_callback *cb) | |
1591 | { | |
1592 | struct nlmsghdr *nlh; | |
1593 | struct ndtmsg *ndtmsg; | |
1594 | ||
1797754e TG |
1595 | nlh = NLMSG_NEW_ANSWER(skb, cb, RTM_NEWNEIGHTBL, sizeof(struct ndtmsg), |
1596 | NLM_F_MULTI); | |
c7fb64db | 1597 | |
4b6ea82d | 1598 | ndtmsg = NLMSG_DATA(nlh); |
c7fb64db TG |
1599 | |
1600 | read_lock_bh(&tbl->lock); | |
1601 | ndtmsg->ndtm_family = tbl->family; | |
9ef1d4c7 PM |
1602 | ndtmsg->ndtm_pad1 = 0; |
1603 | ndtmsg->ndtm_pad2 = 0; | |
c7fb64db TG |
1604 | |
1605 | RTA_PUT_STRING(skb, NDTA_NAME, tbl->id); | |
1606 | RTA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval); | |
1607 | RTA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1); | |
1608 | RTA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2); | |
1609 | RTA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3); | |
1610 | ||
1611 | { | |
1612 | unsigned long now = jiffies; | |
1613 | unsigned int flush_delta = now - tbl->last_flush; | |
1614 | unsigned int rand_delta = now - tbl->last_rand; | |
1615 | ||
1616 | struct ndt_config ndc = { | |
1617 | .ndtc_key_len = tbl->key_len, | |
1618 | .ndtc_entry_size = tbl->entry_size, | |
1619 | .ndtc_entries = atomic_read(&tbl->entries), | |
1620 | .ndtc_last_flush = jiffies_to_msecs(flush_delta), | |
1621 | .ndtc_last_rand = jiffies_to_msecs(rand_delta), | |
1622 | .ndtc_hash_rnd = tbl->hash_rnd, | |
1623 | .ndtc_hash_mask = tbl->hash_mask, | |
1624 | .ndtc_hash_chain_gc = tbl->hash_chain_gc, | |
1625 | .ndtc_proxy_qlen = tbl->proxy_queue.qlen, | |
1626 | }; | |
1627 | ||
1628 | RTA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc); | |
1629 | } | |
1630 | ||
1631 | { | |
1632 | int cpu; | |
1633 | struct ndt_stats ndst; | |
1634 | ||
1635 | memset(&ndst, 0, sizeof(ndst)); | |
1636 | ||
1637 | for (cpu = 0; cpu < NR_CPUS; cpu++) { | |
1638 | struct neigh_statistics *st; | |
1639 | ||
1640 | if (!cpu_possible(cpu)) | |
1641 | continue; | |
1642 | ||
1643 | st = per_cpu_ptr(tbl->stats, cpu); | |
1644 | ndst.ndts_allocs += st->allocs; | |
1645 | ndst.ndts_destroys += st->destroys; | |
1646 | ndst.ndts_hash_grows += st->hash_grows; | |
1647 | ndst.ndts_res_failed += st->res_failed; | |
1648 | ndst.ndts_lookups += st->lookups; | |
1649 | ndst.ndts_hits += st->hits; | |
1650 | ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast; | |
1651 | ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast; | |
1652 | ndst.ndts_periodic_gc_runs += st->periodic_gc_runs; | |
1653 | ndst.ndts_forced_gc_runs += st->forced_gc_runs; | |
1654 | } | |
1655 | ||
1656 | RTA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst); | |
1657 | } | |
1658 | ||
1659 | BUG_ON(tbl->parms.dev); | |
1660 | if (neightbl_fill_parms(skb, &tbl->parms) < 0) | |
1661 | goto rtattr_failure; | |
1662 | ||
1663 | read_unlock_bh(&tbl->lock); | |
1664 | return NLMSG_END(skb, nlh); | |
1665 | ||
1666 | rtattr_failure: | |
1667 | read_unlock_bh(&tbl->lock); | |
1668 | return NLMSG_CANCEL(skb, nlh); | |
1669 | ||
1670 | nlmsg_failure: | |
1671 | return -1; | |
1672 | } | |
1673 | ||
1674 | static int neightbl_fill_param_info(struct neigh_table *tbl, | |
1675 | struct neigh_parms *parms, | |
1676 | struct sk_buff *skb, | |
1677 | struct netlink_callback *cb) | |
1678 | { | |
1679 | struct ndtmsg *ndtmsg; | |
1680 | struct nlmsghdr *nlh; | |
1681 | ||
1797754e TG |
1682 | nlh = NLMSG_NEW_ANSWER(skb, cb, RTM_NEWNEIGHTBL, sizeof(struct ndtmsg), |
1683 | NLM_F_MULTI); | |
c7fb64db | 1684 | |
4b6ea82d | 1685 | ndtmsg = NLMSG_DATA(nlh); |
c7fb64db TG |
1686 | |
1687 | read_lock_bh(&tbl->lock); | |
1688 | ndtmsg->ndtm_family = tbl->family; | |
9ef1d4c7 PM |
1689 | ndtmsg->ndtm_pad1 = 0; |
1690 | ndtmsg->ndtm_pad2 = 0; | |
c7fb64db TG |
1691 | RTA_PUT_STRING(skb, NDTA_NAME, tbl->id); |
1692 | ||
1693 | if (neightbl_fill_parms(skb, parms) < 0) | |
1694 | goto rtattr_failure; | |
1695 | ||
1696 | read_unlock_bh(&tbl->lock); | |
1697 | return NLMSG_END(skb, nlh); | |
1698 | ||
1699 | rtattr_failure: | |
1700 | read_unlock_bh(&tbl->lock); | |
1701 | return NLMSG_CANCEL(skb, nlh); | |
1702 | ||
1703 | nlmsg_failure: | |
1704 | return -1; | |
1705 | } | |
1706 | ||
1707 | static inline struct neigh_parms *lookup_neigh_params(struct neigh_table *tbl, | |
1708 | int ifindex) | |
1709 | { | |
1710 | struct neigh_parms *p; | |
1711 | ||
1712 | for (p = &tbl->parms; p; p = p->next) | |
1713 | if ((p->dev && p->dev->ifindex == ifindex) || | |
1714 | (!p->dev && !ifindex)) | |
1715 | return p; | |
1716 | ||
1717 | return NULL; | |
1718 | } | |
1719 | ||
1720 | int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1721 | { | |
1722 | struct neigh_table *tbl; | |
1723 | struct ndtmsg *ndtmsg = NLMSG_DATA(nlh); | |
1724 | struct rtattr **tb = arg; | |
1725 | int err = -EINVAL; | |
1726 | ||
1727 | if (!tb[NDTA_NAME - 1] || !RTA_PAYLOAD(tb[NDTA_NAME - 1])) | |
1728 | return -EINVAL; | |
1729 | ||
1730 | read_lock(&neigh_tbl_lock); | |
1731 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1732 | if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family) | |
1733 | continue; | |
1734 | ||
1735 | if (!rtattr_strcmp(tb[NDTA_NAME - 1], tbl->id)) | |
1736 | break; | |
1737 | } | |
1738 | ||
1739 | if (tbl == NULL) { | |
1740 | err = -ENOENT; | |
1741 | goto errout; | |
1742 | } | |
1743 | ||
1744 | /* | |
1745 | * We acquire tbl->lock to be nice to the periodic timers and | |
1746 | * make sure they always see a consistent set of values. | |
1747 | */ | |
1748 | write_lock_bh(&tbl->lock); | |
1749 | ||
1750 | if (tb[NDTA_THRESH1 - 1]) | |
1751 | tbl->gc_thresh1 = RTA_GET_U32(tb[NDTA_THRESH1 - 1]); | |
1752 | ||
1753 | if (tb[NDTA_THRESH2 - 1]) | |
1754 | tbl->gc_thresh2 = RTA_GET_U32(tb[NDTA_THRESH2 - 1]); | |
1755 | ||
1756 | if (tb[NDTA_THRESH3 - 1]) | |
1757 | tbl->gc_thresh3 = RTA_GET_U32(tb[NDTA_THRESH3 - 1]); | |
1758 | ||
1759 | if (tb[NDTA_GC_INTERVAL - 1]) | |
1760 | tbl->gc_interval = RTA_GET_MSECS(tb[NDTA_GC_INTERVAL - 1]); | |
1761 | ||
1762 | if (tb[NDTA_PARMS - 1]) { | |
1763 | struct rtattr *tbp[NDTPA_MAX]; | |
1764 | struct neigh_parms *p; | |
1765 | u32 ifindex = 0; | |
1766 | ||
1767 | if (rtattr_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS - 1]) < 0) | |
1768 | goto rtattr_failure; | |
1769 | ||
1770 | if (tbp[NDTPA_IFINDEX - 1]) | |
1771 | ifindex = RTA_GET_U32(tbp[NDTPA_IFINDEX - 1]); | |
1772 | ||
1773 | p = lookup_neigh_params(tbl, ifindex); | |
1774 | if (p == NULL) { | |
1775 | err = -ENOENT; | |
1776 | goto rtattr_failure; | |
1777 | } | |
1778 | ||
1779 | if (tbp[NDTPA_QUEUE_LEN - 1]) | |
1780 | p->queue_len = RTA_GET_U32(tbp[NDTPA_QUEUE_LEN - 1]); | |
1781 | ||
1782 | if (tbp[NDTPA_PROXY_QLEN - 1]) | |
1783 | p->proxy_qlen = RTA_GET_U32(tbp[NDTPA_PROXY_QLEN - 1]); | |
1784 | ||
1785 | if (tbp[NDTPA_APP_PROBES - 1]) | |
1786 | p->app_probes = RTA_GET_U32(tbp[NDTPA_APP_PROBES - 1]); | |
1787 | ||
1788 | if (tbp[NDTPA_UCAST_PROBES - 1]) | |
1789 | p->ucast_probes = | |
1790 | RTA_GET_U32(tbp[NDTPA_UCAST_PROBES - 1]); | |
1791 | ||
1792 | if (tbp[NDTPA_MCAST_PROBES - 1]) | |
1793 | p->mcast_probes = | |
1794 | RTA_GET_U32(tbp[NDTPA_MCAST_PROBES - 1]); | |
1795 | ||
1796 | if (tbp[NDTPA_BASE_REACHABLE_TIME - 1]) | |
1797 | p->base_reachable_time = | |
1798 | RTA_GET_MSECS(tbp[NDTPA_BASE_REACHABLE_TIME - 1]); | |
1799 | ||
1800 | if (tbp[NDTPA_GC_STALETIME - 1]) | |
1801 | p->gc_staletime = | |
1802 | RTA_GET_MSECS(tbp[NDTPA_GC_STALETIME - 1]); | |
1803 | ||
1804 | if (tbp[NDTPA_DELAY_PROBE_TIME - 1]) | |
1805 | p->delay_probe_time = | |
1806 | RTA_GET_MSECS(tbp[NDTPA_DELAY_PROBE_TIME - 1]); | |
1807 | ||
1808 | if (tbp[NDTPA_RETRANS_TIME - 1]) | |
1809 | p->retrans_time = | |
1810 | RTA_GET_MSECS(tbp[NDTPA_RETRANS_TIME - 1]); | |
1811 | ||
1812 | if (tbp[NDTPA_ANYCAST_DELAY - 1]) | |
1813 | p->anycast_delay = | |
1814 | RTA_GET_MSECS(tbp[NDTPA_ANYCAST_DELAY - 1]); | |
1815 | ||
1816 | if (tbp[NDTPA_PROXY_DELAY - 1]) | |
1817 | p->proxy_delay = | |
1818 | RTA_GET_MSECS(tbp[NDTPA_PROXY_DELAY - 1]); | |
1819 | ||
1820 | if (tbp[NDTPA_LOCKTIME - 1]) | |
1821 | p->locktime = RTA_GET_MSECS(tbp[NDTPA_LOCKTIME - 1]); | |
1822 | } | |
1823 | ||
1824 | err = 0; | |
1825 | ||
1826 | rtattr_failure: | |
1827 | write_unlock_bh(&tbl->lock); | |
1828 | errout: | |
1829 | read_unlock(&neigh_tbl_lock); | |
1830 | return err; | |
1831 | } | |
1832 | ||
1833 | int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb) | |
1834 | { | |
1835 | int idx, family; | |
1836 | int s_idx = cb->args[0]; | |
1837 | struct neigh_table *tbl; | |
1838 | ||
1839 | family = ((struct rtgenmsg *)NLMSG_DATA(cb->nlh))->rtgen_family; | |
1840 | ||
1841 | read_lock(&neigh_tbl_lock); | |
1842 | for (tbl = neigh_tables, idx = 0; tbl; tbl = tbl->next) { | |
1843 | struct neigh_parms *p; | |
1844 | ||
1845 | if (idx < s_idx || (family && tbl->family != family)) | |
1846 | continue; | |
1847 | ||
1848 | if (neightbl_fill_info(tbl, skb, cb) <= 0) | |
1849 | break; | |
1850 | ||
1851 | for (++idx, p = tbl->parms.next; p; p = p->next, idx++) { | |
1852 | if (idx < s_idx) | |
1853 | continue; | |
1854 | ||
1855 | if (neightbl_fill_param_info(tbl, p, skb, cb) <= 0) | |
1856 | goto out; | |
1857 | } | |
1858 | ||
1859 | } | |
1860 | out: | |
1861 | read_unlock(&neigh_tbl_lock); | |
1862 | cb->args[0] = idx; | |
1863 | ||
1864 | return skb->len; | |
1865 | } | |
1da177e4 LT |
1866 | |
1867 | static int neigh_fill_info(struct sk_buff *skb, struct neighbour *n, | |
b6544c0b | 1868 | u32 pid, u32 seq, int event, unsigned int flags) |
1da177e4 LT |
1869 | { |
1870 | unsigned long now = jiffies; | |
1871 | unsigned char *b = skb->tail; | |
1872 | struct nda_cacheinfo ci; | |
1873 | int locked = 0; | |
1874 | u32 probes; | |
b6544c0b JHS |
1875 | struct nlmsghdr *nlh = NLMSG_NEW(skb, pid, seq, event, |
1876 | sizeof(struct ndmsg), flags); | |
1da177e4 LT |
1877 | struct ndmsg *ndm = NLMSG_DATA(nlh); |
1878 | ||
1da177e4 | 1879 | ndm->ndm_family = n->ops->family; |
9ef1d4c7 PM |
1880 | ndm->ndm_pad1 = 0; |
1881 | ndm->ndm_pad2 = 0; | |
1da177e4 LT |
1882 | ndm->ndm_flags = n->flags; |
1883 | ndm->ndm_type = n->type; | |
1884 | ndm->ndm_ifindex = n->dev->ifindex; | |
1885 | RTA_PUT(skb, NDA_DST, n->tbl->key_len, n->primary_key); | |
1886 | read_lock_bh(&n->lock); | |
1887 | locked = 1; | |
1888 | ndm->ndm_state = n->nud_state; | |
1889 | if (n->nud_state & NUD_VALID) | |
1890 | RTA_PUT(skb, NDA_LLADDR, n->dev->addr_len, n->ha); | |
1891 | ci.ndm_used = now - n->used; | |
1892 | ci.ndm_confirmed = now - n->confirmed; | |
1893 | ci.ndm_updated = now - n->updated; | |
1894 | ci.ndm_refcnt = atomic_read(&n->refcnt) - 1; | |
1895 | probes = atomic_read(&n->probes); | |
1896 | read_unlock_bh(&n->lock); | |
1897 | locked = 0; | |
1898 | RTA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci); | |
1899 | RTA_PUT(skb, NDA_PROBES, sizeof(probes), &probes); | |
1900 | nlh->nlmsg_len = skb->tail - b; | |
1901 | return skb->len; | |
1902 | ||
1903 | nlmsg_failure: | |
1904 | rtattr_failure: | |
1905 | if (locked) | |
1906 | read_unlock_bh(&n->lock); | |
1907 | skb_trim(skb, b - skb->data); | |
1908 | return -1; | |
1909 | } | |
1910 | ||
1911 | ||
1912 | static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb, | |
1913 | struct netlink_callback *cb) | |
1914 | { | |
1915 | struct neighbour *n; | |
1916 | int rc, h, s_h = cb->args[1]; | |
1917 | int idx, s_idx = idx = cb->args[2]; | |
1918 | ||
1919 | for (h = 0; h <= tbl->hash_mask; h++) { | |
1920 | if (h < s_h) | |
1921 | continue; | |
1922 | if (h > s_h) | |
1923 | s_idx = 0; | |
1924 | read_lock_bh(&tbl->lock); | |
1925 | for (n = tbl->hash_buckets[h], idx = 0; n; n = n->next, idx++) { | |
1926 | if (idx < s_idx) | |
1927 | continue; | |
1928 | if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid, | |
1929 | cb->nlh->nlmsg_seq, | |
b6544c0b JHS |
1930 | RTM_NEWNEIGH, |
1931 | NLM_F_MULTI) <= 0) { | |
1da177e4 LT |
1932 | read_unlock_bh(&tbl->lock); |
1933 | rc = -1; | |
1934 | goto out; | |
1935 | } | |
1936 | } | |
1937 | read_unlock_bh(&tbl->lock); | |
1938 | } | |
1939 | rc = skb->len; | |
1940 | out: | |
1941 | cb->args[1] = h; | |
1942 | cb->args[2] = idx; | |
1943 | return rc; | |
1944 | } | |
1945 | ||
1946 | int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb) | |
1947 | { | |
1948 | struct neigh_table *tbl; | |
1949 | int t, family, s_t; | |
1950 | ||
1951 | read_lock(&neigh_tbl_lock); | |
1952 | family = ((struct rtgenmsg *)NLMSG_DATA(cb->nlh))->rtgen_family; | |
1953 | s_t = cb->args[0]; | |
1954 | ||
1955 | for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) { | |
1956 | if (t < s_t || (family && tbl->family != family)) | |
1957 | continue; | |
1958 | if (t > s_t) | |
1959 | memset(&cb->args[1], 0, sizeof(cb->args) - | |
1960 | sizeof(cb->args[0])); | |
1961 | if (neigh_dump_table(tbl, skb, cb) < 0) | |
1962 | break; | |
1963 | } | |
1964 | read_unlock(&neigh_tbl_lock); | |
1965 | ||
1966 | cb->args[0] = t; | |
1967 | return skb->len; | |
1968 | } | |
1969 | ||
1970 | void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie) | |
1971 | { | |
1972 | int chain; | |
1973 | ||
1974 | read_lock_bh(&tbl->lock); | |
1975 | for (chain = 0; chain <= tbl->hash_mask; chain++) { | |
1976 | struct neighbour *n; | |
1977 | ||
1978 | for (n = tbl->hash_buckets[chain]; n; n = n->next) | |
1979 | cb(n, cookie); | |
1980 | } | |
1981 | read_unlock_bh(&tbl->lock); | |
1982 | } | |
1983 | EXPORT_SYMBOL(neigh_for_each); | |
1984 | ||
1985 | /* The tbl->lock must be held as a writer and BH disabled. */ | |
1986 | void __neigh_for_each_release(struct neigh_table *tbl, | |
1987 | int (*cb)(struct neighbour *)) | |
1988 | { | |
1989 | int chain; | |
1990 | ||
1991 | for (chain = 0; chain <= tbl->hash_mask; chain++) { | |
1992 | struct neighbour *n, **np; | |
1993 | ||
1994 | np = &tbl->hash_buckets[chain]; | |
1995 | while ((n = *np) != NULL) { | |
1996 | int release; | |
1997 | ||
1998 | write_lock(&n->lock); | |
1999 | release = cb(n); | |
2000 | if (release) { | |
2001 | *np = n->next; | |
2002 | n->dead = 1; | |
2003 | } else | |
2004 | np = &n->next; | |
2005 | write_unlock(&n->lock); | |
2006 | if (release) | |
2007 | neigh_release(n); | |
2008 | } | |
2009 | } | |
2010 | } | |
2011 | EXPORT_SYMBOL(__neigh_for_each_release); | |
2012 | ||
2013 | #ifdef CONFIG_PROC_FS | |
2014 | ||
2015 | static struct neighbour *neigh_get_first(struct seq_file *seq) | |
2016 | { | |
2017 | struct neigh_seq_state *state = seq->private; | |
2018 | struct neigh_table *tbl = state->tbl; | |
2019 | struct neighbour *n = NULL; | |
2020 | int bucket = state->bucket; | |
2021 | ||
2022 | state->flags &= ~NEIGH_SEQ_IS_PNEIGH; | |
2023 | for (bucket = 0; bucket <= tbl->hash_mask; bucket++) { | |
2024 | n = tbl->hash_buckets[bucket]; | |
2025 | ||
2026 | while (n) { | |
2027 | if (state->neigh_sub_iter) { | |
2028 | loff_t fakep = 0; | |
2029 | void *v; | |
2030 | ||
2031 | v = state->neigh_sub_iter(state, n, &fakep); | |
2032 | if (!v) | |
2033 | goto next; | |
2034 | } | |
2035 | if (!(state->flags & NEIGH_SEQ_SKIP_NOARP)) | |
2036 | break; | |
2037 | if (n->nud_state & ~NUD_NOARP) | |
2038 | break; | |
2039 | next: | |
2040 | n = n->next; | |
2041 | } | |
2042 | ||
2043 | if (n) | |
2044 | break; | |
2045 | } | |
2046 | state->bucket = bucket; | |
2047 | ||
2048 | return n; | |
2049 | } | |
2050 | ||
2051 | static struct neighbour *neigh_get_next(struct seq_file *seq, | |
2052 | struct neighbour *n, | |
2053 | loff_t *pos) | |
2054 | { | |
2055 | struct neigh_seq_state *state = seq->private; | |
2056 | struct neigh_table *tbl = state->tbl; | |
2057 | ||
2058 | if (state->neigh_sub_iter) { | |
2059 | void *v = state->neigh_sub_iter(state, n, pos); | |
2060 | if (v) | |
2061 | return n; | |
2062 | } | |
2063 | n = n->next; | |
2064 | ||
2065 | while (1) { | |
2066 | while (n) { | |
2067 | if (state->neigh_sub_iter) { | |
2068 | void *v = state->neigh_sub_iter(state, n, pos); | |
2069 | if (v) | |
2070 | return n; | |
2071 | goto next; | |
2072 | } | |
2073 | if (!(state->flags & NEIGH_SEQ_SKIP_NOARP)) | |
2074 | break; | |
2075 | ||
2076 | if (n->nud_state & ~NUD_NOARP) | |
2077 | break; | |
2078 | next: | |
2079 | n = n->next; | |
2080 | } | |
2081 | ||
2082 | if (n) | |
2083 | break; | |
2084 | ||
2085 | if (++state->bucket > tbl->hash_mask) | |
2086 | break; | |
2087 | ||
2088 | n = tbl->hash_buckets[state->bucket]; | |
2089 | } | |
2090 | ||
2091 | if (n && pos) | |
2092 | --(*pos); | |
2093 | return n; | |
2094 | } | |
2095 | ||
2096 | static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos) | |
2097 | { | |
2098 | struct neighbour *n = neigh_get_first(seq); | |
2099 | ||
2100 | if (n) { | |
2101 | while (*pos) { | |
2102 | n = neigh_get_next(seq, n, pos); | |
2103 | if (!n) | |
2104 | break; | |
2105 | } | |
2106 | } | |
2107 | return *pos ? NULL : n; | |
2108 | } | |
2109 | ||
2110 | static struct pneigh_entry *pneigh_get_first(struct seq_file *seq) | |
2111 | { | |
2112 | struct neigh_seq_state *state = seq->private; | |
2113 | struct neigh_table *tbl = state->tbl; | |
2114 | struct pneigh_entry *pn = NULL; | |
2115 | int bucket = state->bucket; | |
2116 | ||
2117 | state->flags |= NEIGH_SEQ_IS_PNEIGH; | |
2118 | for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) { | |
2119 | pn = tbl->phash_buckets[bucket]; | |
2120 | if (pn) | |
2121 | break; | |
2122 | } | |
2123 | state->bucket = bucket; | |
2124 | ||
2125 | return pn; | |
2126 | } | |
2127 | ||
2128 | static struct pneigh_entry *pneigh_get_next(struct seq_file *seq, | |
2129 | struct pneigh_entry *pn, | |
2130 | loff_t *pos) | |
2131 | { | |
2132 | struct neigh_seq_state *state = seq->private; | |
2133 | struct neigh_table *tbl = state->tbl; | |
2134 | ||
2135 | pn = pn->next; | |
2136 | while (!pn) { | |
2137 | if (++state->bucket > PNEIGH_HASHMASK) | |
2138 | break; | |
2139 | pn = tbl->phash_buckets[state->bucket]; | |
2140 | if (pn) | |
2141 | break; | |
2142 | } | |
2143 | ||
2144 | if (pn && pos) | |
2145 | --(*pos); | |
2146 | ||
2147 | return pn; | |
2148 | } | |
2149 | ||
2150 | static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos) | |
2151 | { | |
2152 | struct pneigh_entry *pn = pneigh_get_first(seq); | |
2153 | ||
2154 | if (pn) { | |
2155 | while (*pos) { | |
2156 | pn = pneigh_get_next(seq, pn, pos); | |
2157 | if (!pn) | |
2158 | break; | |
2159 | } | |
2160 | } | |
2161 | return *pos ? NULL : pn; | |
2162 | } | |
2163 | ||
2164 | static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos) | |
2165 | { | |
2166 | struct neigh_seq_state *state = seq->private; | |
2167 | void *rc; | |
2168 | ||
2169 | rc = neigh_get_idx(seq, pos); | |
2170 | if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY)) | |
2171 | rc = pneigh_get_idx(seq, pos); | |
2172 | ||
2173 | return rc; | |
2174 | } | |
2175 | ||
2176 | void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags) | |
2177 | { | |
2178 | struct neigh_seq_state *state = seq->private; | |
2179 | loff_t pos_minus_one; | |
2180 | ||
2181 | state->tbl = tbl; | |
2182 | state->bucket = 0; | |
2183 | state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH); | |
2184 | ||
2185 | read_lock_bh(&tbl->lock); | |
2186 | ||
2187 | pos_minus_one = *pos - 1; | |
2188 | return *pos ? neigh_get_idx_any(seq, &pos_minus_one) : SEQ_START_TOKEN; | |
2189 | } | |
2190 | EXPORT_SYMBOL(neigh_seq_start); | |
2191 | ||
2192 | void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2193 | { | |
2194 | struct neigh_seq_state *state; | |
2195 | void *rc; | |
2196 | ||
2197 | if (v == SEQ_START_TOKEN) { | |
2198 | rc = neigh_get_idx(seq, pos); | |
2199 | goto out; | |
2200 | } | |
2201 | ||
2202 | state = seq->private; | |
2203 | if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) { | |
2204 | rc = neigh_get_next(seq, v, NULL); | |
2205 | if (rc) | |
2206 | goto out; | |
2207 | if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY)) | |
2208 | rc = pneigh_get_first(seq); | |
2209 | } else { | |
2210 | BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY); | |
2211 | rc = pneigh_get_next(seq, v, NULL); | |
2212 | } | |
2213 | out: | |
2214 | ++(*pos); | |
2215 | return rc; | |
2216 | } | |
2217 | EXPORT_SYMBOL(neigh_seq_next); | |
2218 | ||
2219 | void neigh_seq_stop(struct seq_file *seq, void *v) | |
2220 | { | |
2221 | struct neigh_seq_state *state = seq->private; | |
2222 | struct neigh_table *tbl = state->tbl; | |
2223 | ||
2224 | read_unlock_bh(&tbl->lock); | |
2225 | } | |
2226 | EXPORT_SYMBOL(neigh_seq_stop); | |
2227 | ||
2228 | /* statistics via seq_file */ | |
2229 | ||
2230 | static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos) | |
2231 | { | |
2232 | struct proc_dir_entry *pde = seq->private; | |
2233 | struct neigh_table *tbl = pde->data; | |
2234 | int cpu; | |
2235 | ||
2236 | if (*pos == 0) | |
2237 | return SEQ_START_TOKEN; | |
2238 | ||
2239 | for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) { | |
2240 | if (!cpu_possible(cpu)) | |
2241 | continue; | |
2242 | *pos = cpu+1; | |
2243 | return per_cpu_ptr(tbl->stats, cpu); | |
2244 | } | |
2245 | return NULL; | |
2246 | } | |
2247 | ||
2248 | static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2249 | { | |
2250 | struct proc_dir_entry *pde = seq->private; | |
2251 | struct neigh_table *tbl = pde->data; | |
2252 | int cpu; | |
2253 | ||
2254 | for (cpu = *pos; cpu < NR_CPUS; ++cpu) { | |
2255 | if (!cpu_possible(cpu)) | |
2256 | continue; | |
2257 | *pos = cpu+1; | |
2258 | return per_cpu_ptr(tbl->stats, cpu); | |
2259 | } | |
2260 | return NULL; | |
2261 | } | |
2262 | ||
2263 | static void neigh_stat_seq_stop(struct seq_file *seq, void *v) | |
2264 | { | |
2265 | ||
2266 | } | |
2267 | ||
2268 | static int neigh_stat_seq_show(struct seq_file *seq, void *v) | |
2269 | { | |
2270 | struct proc_dir_entry *pde = seq->private; | |
2271 | struct neigh_table *tbl = pde->data; | |
2272 | struct neigh_statistics *st = v; | |
2273 | ||
2274 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 2275 | seq_printf(seq, "entries allocs destroys hash_grows lookups hits res_failed rcv_probes_mcast rcv_probes_ucast periodic_gc_runs forced_gc_runs\n"); |
1da177e4 LT |
2276 | return 0; |
2277 | } | |
2278 | ||
2279 | seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx " | |
2280 | "%08lx %08lx %08lx %08lx\n", | |
2281 | atomic_read(&tbl->entries), | |
2282 | ||
2283 | st->allocs, | |
2284 | st->destroys, | |
2285 | st->hash_grows, | |
2286 | ||
2287 | st->lookups, | |
2288 | st->hits, | |
2289 | ||
2290 | st->res_failed, | |
2291 | ||
2292 | st->rcv_probes_mcast, | |
2293 | st->rcv_probes_ucast, | |
2294 | ||
2295 | st->periodic_gc_runs, | |
2296 | st->forced_gc_runs | |
2297 | ); | |
2298 | ||
2299 | return 0; | |
2300 | } | |
2301 | ||
2302 | static struct seq_operations neigh_stat_seq_ops = { | |
2303 | .start = neigh_stat_seq_start, | |
2304 | .next = neigh_stat_seq_next, | |
2305 | .stop = neigh_stat_seq_stop, | |
2306 | .show = neigh_stat_seq_show, | |
2307 | }; | |
2308 | ||
2309 | static int neigh_stat_seq_open(struct inode *inode, struct file *file) | |
2310 | { | |
2311 | int ret = seq_open(file, &neigh_stat_seq_ops); | |
2312 | ||
2313 | if (!ret) { | |
2314 | struct seq_file *sf = file->private_data; | |
2315 | sf->private = PDE(inode); | |
2316 | } | |
2317 | return ret; | |
2318 | }; | |
2319 | ||
2320 | static struct file_operations neigh_stat_seq_fops = { | |
2321 | .owner = THIS_MODULE, | |
2322 | .open = neigh_stat_seq_open, | |
2323 | .read = seq_read, | |
2324 | .llseek = seq_lseek, | |
2325 | .release = seq_release, | |
2326 | }; | |
2327 | ||
2328 | #endif /* CONFIG_PROC_FS */ | |
2329 | ||
2330 | #ifdef CONFIG_ARPD | |
2331 | void neigh_app_ns(struct neighbour *n) | |
2332 | { | |
2333 | struct nlmsghdr *nlh; | |
2334 | int size = NLMSG_SPACE(sizeof(struct ndmsg) + 256); | |
2335 | struct sk_buff *skb = alloc_skb(size, GFP_ATOMIC); | |
2336 | ||
2337 | if (!skb) | |
2338 | return; | |
2339 | ||
b6544c0b | 2340 | if (neigh_fill_info(skb, n, 0, 0, RTM_GETNEIGH, 0) < 0) { |
1da177e4 LT |
2341 | kfree_skb(skb); |
2342 | return; | |
2343 | } | |
2344 | nlh = (struct nlmsghdr *)skb->data; | |
2345 | nlh->nlmsg_flags = NLM_F_REQUEST; | |
ac6d439d PM |
2346 | NETLINK_CB(skb).dst_group = RTNLGRP_NEIGH; |
2347 | netlink_broadcast(rtnl, skb, 0, RTNLGRP_NEIGH, GFP_ATOMIC); | |
1da177e4 LT |
2348 | } |
2349 | ||
2350 | static void neigh_app_notify(struct neighbour *n) | |
2351 | { | |
2352 | struct nlmsghdr *nlh; | |
2353 | int size = NLMSG_SPACE(sizeof(struct ndmsg) + 256); | |
2354 | struct sk_buff *skb = alloc_skb(size, GFP_ATOMIC); | |
2355 | ||
2356 | if (!skb) | |
2357 | return; | |
2358 | ||
b6544c0b | 2359 | if (neigh_fill_info(skb, n, 0, 0, RTM_NEWNEIGH, 0) < 0) { |
1da177e4 LT |
2360 | kfree_skb(skb); |
2361 | return; | |
2362 | } | |
2363 | nlh = (struct nlmsghdr *)skb->data; | |
ac6d439d PM |
2364 | NETLINK_CB(skb).dst_group = RTNLGRP_NEIGH; |
2365 | netlink_broadcast(rtnl, skb, 0, RTNLGRP_NEIGH, GFP_ATOMIC); | |
1da177e4 LT |
2366 | } |
2367 | ||
2368 | #endif /* CONFIG_ARPD */ | |
2369 | ||
2370 | #ifdef CONFIG_SYSCTL | |
2371 | ||
2372 | static struct neigh_sysctl_table { | |
2373 | struct ctl_table_header *sysctl_header; | |
2374 | ctl_table neigh_vars[__NET_NEIGH_MAX]; | |
2375 | ctl_table neigh_dev[2]; | |
2376 | ctl_table neigh_neigh_dir[2]; | |
2377 | ctl_table neigh_proto_dir[2]; | |
2378 | ctl_table neigh_root_dir[2]; | |
2379 | } neigh_sysctl_template = { | |
2380 | .neigh_vars = { | |
2381 | { | |
2382 | .ctl_name = NET_NEIGH_MCAST_SOLICIT, | |
2383 | .procname = "mcast_solicit", | |
2384 | .maxlen = sizeof(int), | |
2385 | .mode = 0644, | |
2386 | .proc_handler = &proc_dointvec, | |
2387 | }, | |
2388 | { | |
2389 | .ctl_name = NET_NEIGH_UCAST_SOLICIT, | |
2390 | .procname = "ucast_solicit", | |
2391 | .maxlen = sizeof(int), | |
2392 | .mode = 0644, | |
2393 | .proc_handler = &proc_dointvec, | |
2394 | }, | |
2395 | { | |
2396 | .ctl_name = NET_NEIGH_APP_SOLICIT, | |
2397 | .procname = "app_solicit", | |
2398 | .maxlen = sizeof(int), | |
2399 | .mode = 0644, | |
2400 | .proc_handler = &proc_dointvec, | |
2401 | }, | |
2402 | { | |
2403 | .ctl_name = NET_NEIGH_RETRANS_TIME, | |
2404 | .procname = "retrans_time", | |
2405 | .maxlen = sizeof(int), | |
2406 | .mode = 0644, | |
2407 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2408 | }, | |
2409 | { | |
2410 | .ctl_name = NET_NEIGH_REACHABLE_TIME, | |
2411 | .procname = "base_reachable_time", | |
2412 | .maxlen = sizeof(int), | |
2413 | .mode = 0644, | |
2414 | .proc_handler = &proc_dointvec_jiffies, | |
2415 | .strategy = &sysctl_jiffies, | |
2416 | }, | |
2417 | { | |
2418 | .ctl_name = NET_NEIGH_DELAY_PROBE_TIME, | |
2419 | .procname = "delay_first_probe_time", | |
2420 | .maxlen = sizeof(int), | |
2421 | .mode = 0644, | |
2422 | .proc_handler = &proc_dointvec_jiffies, | |
2423 | .strategy = &sysctl_jiffies, | |
2424 | }, | |
2425 | { | |
2426 | .ctl_name = NET_NEIGH_GC_STALE_TIME, | |
2427 | .procname = "gc_stale_time", | |
2428 | .maxlen = sizeof(int), | |
2429 | .mode = 0644, | |
2430 | .proc_handler = &proc_dointvec_jiffies, | |
2431 | .strategy = &sysctl_jiffies, | |
2432 | }, | |
2433 | { | |
2434 | .ctl_name = NET_NEIGH_UNRES_QLEN, | |
2435 | .procname = "unres_qlen", | |
2436 | .maxlen = sizeof(int), | |
2437 | .mode = 0644, | |
2438 | .proc_handler = &proc_dointvec, | |
2439 | }, | |
2440 | { | |
2441 | .ctl_name = NET_NEIGH_PROXY_QLEN, | |
2442 | .procname = "proxy_qlen", | |
2443 | .maxlen = sizeof(int), | |
2444 | .mode = 0644, | |
2445 | .proc_handler = &proc_dointvec, | |
2446 | }, | |
2447 | { | |
2448 | .ctl_name = NET_NEIGH_ANYCAST_DELAY, | |
2449 | .procname = "anycast_delay", | |
2450 | .maxlen = sizeof(int), | |
2451 | .mode = 0644, | |
2452 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2453 | }, | |
2454 | { | |
2455 | .ctl_name = NET_NEIGH_PROXY_DELAY, | |
2456 | .procname = "proxy_delay", | |
2457 | .maxlen = sizeof(int), | |
2458 | .mode = 0644, | |
2459 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2460 | }, | |
2461 | { | |
2462 | .ctl_name = NET_NEIGH_LOCKTIME, | |
2463 | .procname = "locktime", | |
2464 | .maxlen = sizeof(int), | |
2465 | .mode = 0644, | |
2466 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2467 | }, | |
2468 | { | |
2469 | .ctl_name = NET_NEIGH_GC_INTERVAL, | |
2470 | .procname = "gc_interval", | |
2471 | .maxlen = sizeof(int), | |
2472 | .mode = 0644, | |
2473 | .proc_handler = &proc_dointvec_jiffies, | |
2474 | .strategy = &sysctl_jiffies, | |
2475 | }, | |
2476 | { | |
2477 | .ctl_name = NET_NEIGH_GC_THRESH1, | |
2478 | .procname = "gc_thresh1", | |
2479 | .maxlen = sizeof(int), | |
2480 | .mode = 0644, | |
2481 | .proc_handler = &proc_dointvec, | |
2482 | }, | |
2483 | { | |
2484 | .ctl_name = NET_NEIGH_GC_THRESH2, | |
2485 | .procname = "gc_thresh2", | |
2486 | .maxlen = sizeof(int), | |
2487 | .mode = 0644, | |
2488 | .proc_handler = &proc_dointvec, | |
2489 | }, | |
2490 | { | |
2491 | .ctl_name = NET_NEIGH_GC_THRESH3, | |
2492 | .procname = "gc_thresh3", | |
2493 | .maxlen = sizeof(int), | |
2494 | .mode = 0644, | |
2495 | .proc_handler = &proc_dointvec, | |
2496 | }, | |
2497 | { | |
2498 | .ctl_name = NET_NEIGH_RETRANS_TIME_MS, | |
2499 | .procname = "retrans_time_ms", | |
2500 | .maxlen = sizeof(int), | |
2501 | .mode = 0644, | |
2502 | .proc_handler = &proc_dointvec_ms_jiffies, | |
2503 | .strategy = &sysctl_ms_jiffies, | |
2504 | }, | |
2505 | { | |
2506 | .ctl_name = NET_NEIGH_REACHABLE_TIME_MS, | |
2507 | .procname = "base_reachable_time_ms", | |
2508 | .maxlen = sizeof(int), | |
2509 | .mode = 0644, | |
2510 | .proc_handler = &proc_dointvec_ms_jiffies, | |
2511 | .strategy = &sysctl_ms_jiffies, | |
2512 | }, | |
2513 | }, | |
2514 | .neigh_dev = { | |
2515 | { | |
2516 | .ctl_name = NET_PROTO_CONF_DEFAULT, | |
2517 | .procname = "default", | |
2518 | .mode = 0555, | |
2519 | }, | |
2520 | }, | |
2521 | .neigh_neigh_dir = { | |
2522 | { | |
2523 | .procname = "neigh", | |
2524 | .mode = 0555, | |
2525 | }, | |
2526 | }, | |
2527 | .neigh_proto_dir = { | |
2528 | { | |
2529 | .mode = 0555, | |
2530 | }, | |
2531 | }, | |
2532 | .neigh_root_dir = { | |
2533 | { | |
2534 | .ctl_name = CTL_NET, | |
2535 | .procname = "net", | |
2536 | .mode = 0555, | |
2537 | }, | |
2538 | }, | |
2539 | }; | |
2540 | ||
2541 | int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p, | |
2542 | int p_id, int pdev_id, char *p_name, | |
2543 | proc_handler *handler, ctl_handler *strategy) | |
2544 | { | |
2545 | struct neigh_sysctl_table *t = kmalloc(sizeof(*t), GFP_KERNEL); | |
2546 | const char *dev_name_source = NULL; | |
2547 | char *dev_name = NULL; | |
2548 | int err = 0; | |
2549 | ||
2550 | if (!t) | |
2551 | return -ENOBUFS; | |
2552 | memcpy(t, &neigh_sysctl_template, sizeof(*t)); | |
2553 | t->neigh_vars[0].data = &p->mcast_probes; | |
2554 | t->neigh_vars[1].data = &p->ucast_probes; | |
2555 | t->neigh_vars[2].data = &p->app_probes; | |
2556 | t->neigh_vars[3].data = &p->retrans_time; | |
2557 | t->neigh_vars[4].data = &p->base_reachable_time; | |
2558 | t->neigh_vars[5].data = &p->delay_probe_time; | |
2559 | t->neigh_vars[6].data = &p->gc_staletime; | |
2560 | t->neigh_vars[7].data = &p->queue_len; | |
2561 | t->neigh_vars[8].data = &p->proxy_qlen; | |
2562 | t->neigh_vars[9].data = &p->anycast_delay; | |
2563 | t->neigh_vars[10].data = &p->proxy_delay; | |
2564 | t->neigh_vars[11].data = &p->locktime; | |
2565 | ||
2566 | if (dev) { | |
2567 | dev_name_source = dev->name; | |
2568 | t->neigh_dev[0].ctl_name = dev->ifindex; | |
2569 | t->neigh_vars[12].procname = NULL; | |
2570 | t->neigh_vars[13].procname = NULL; | |
2571 | t->neigh_vars[14].procname = NULL; | |
2572 | t->neigh_vars[15].procname = NULL; | |
2573 | } else { | |
2574 | dev_name_source = t->neigh_dev[0].procname; | |
2575 | t->neigh_vars[12].data = (int *)(p + 1); | |
2576 | t->neigh_vars[13].data = (int *)(p + 1) + 1; | |
2577 | t->neigh_vars[14].data = (int *)(p + 1) + 2; | |
2578 | t->neigh_vars[15].data = (int *)(p + 1) + 3; | |
2579 | } | |
2580 | ||
2581 | t->neigh_vars[16].data = &p->retrans_time; | |
2582 | t->neigh_vars[17].data = &p->base_reachable_time; | |
2583 | ||
2584 | if (handler || strategy) { | |
2585 | /* RetransTime */ | |
2586 | t->neigh_vars[3].proc_handler = handler; | |
2587 | t->neigh_vars[3].strategy = strategy; | |
2588 | t->neigh_vars[3].extra1 = dev; | |
2589 | /* ReachableTime */ | |
2590 | t->neigh_vars[4].proc_handler = handler; | |
2591 | t->neigh_vars[4].strategy = strategy; | |
2592 | t->neigh_vars[4].extra1 = dev; | |
2593 | /* RetransTime (in milliseconds)*/ | |
2594 | t->neigh_vars[16].proc_handler = handler; | |
2595 | t->neigh_vars[16].strategy = strategy; | |
2596 | t->neigh_vars[16].extra1 = dev; | |
2597 | /* ReachableTime (in milliseconds) */ | |
2598 | t->neigh_vars[17].proc_handler = handler; | |
2599 | t->neigh_vars[17].strategy = strategy; | |
2600 | t->neigh_vars[17].extra1 = dev; | |
2601 | } | |
2602 | ||
543537bd | 2603 | dev_name = kstrdup(dev_name_source, GFP_KERNEL); |
1da177e4 LT |
2604 | if (!dev_name) { |
2605 | err = -ENOBUFS; | |
2606 | goto free; | |
2607 | } | |
2608 | ||
2609 | t->neigh_dev[0].procname = dev_name; | |
2610 | ||
2611 | t->neigh_neigh_dir[0].ctl_name = pdev_id; | |
2612 | ||
2613 | t->neigh_proto_dir[0].procname = p_name; | |
2614 | t->neigh_proto_dir[0].ctl_name = p_id; | |
2615 | ||
2616 | t->neigh_dev[0].child = t->neigh_vars; | |
2617 | t->neigh_neigh_dir[0].child = t->neigh_dev; | |
2618 | t->neigh_proto_dir[0].child = t->neigh_neigh_dir; | |
2619 | t->neigh_root_dir[0].child = t->neigh_proto_dir; | |
2620 | ||
2621 | t->sysctl_header = register_sysctl_table(t->neigh_root_dir, 0); | |
2622 | if (!t->sysctl_header) { | |
2623 | err = -ENOBUFS; | |
2624 | goto free_procname; | |
2625 | } | |
2626 | p->sysctl_table = t; | |
2627 | return 0; | |
2628 | ||
2629 | /* error path */ | |
2630 | free_procname: | |
2631 | kfree(dev_name); | |
2632 | free: | |
2633 | kfree(t); | |
2634 | ||
2635 | return err; | |
2636 | } | |
2637 | ||
2638 | void neigh_sysctl_unregister(struct neigh_parms *p) | |
2639 | { | |
2640 | if (p->sysctl_table) { | |
2641 | struct neigh_sysctl_table *t = p->sysctl_table; | |
2642 | p->sysctl_table = NULL; | |
2643 | unregister_sysctl_table(t->sysctl_header); | |
2644 | kfree(t->neigh_dev[0].procname); | |
2645 | kfree(t); | |
2646 | } | |
2647 | } | |
2648 | ||
2649 | #endif /* CONFIG_SYSCTL */ | |
2650 | ||
2651 | EXPORT_SYMBOL(__neigh_event_send); | |
2652 | EXPORT_SYMBOL(neigh_add); | |
2653 | EXPORT_SYMBOL(neigh_changeaddr); | |
2654 | EXPORT_SYMBOL(neigh_compat_output); | |
2655 | EXPORT_SYMBOL(neigh_connected_output); | |
2656 | EXPORT_SYMBOL(neigh_create); | |
2657 | EXPORT_SYMBOL(neigh_delete); | |
2658 | EXPORT_SYMBOL(neigh_destroy); | |
2659 | EXPORT_SYMBOL(neigh_dump_info); | |
2660 | EXPORT_SYMBOL(neigh_event_ns); | |
2661 | EXPORT_SYMBOL(neigh_ifdown); | |
2662 | EXPORT_SYMBOL(neigh_lookup); | |
2663 | EXPORT_SYMBOL(neigh_lookup_nodev); | |
2664 | EXPORT_SYMBOL(neigh_parms_alloc); | |
2665 | EXPORT_SYMBOL(neigh_parms_release); | |
2666 | EXPORT_SYMBOL(neigh_rand_reach_time); | |
2667 | EXPORT_SYMBOL(neigh_resolve_output); | |
2668 | EXPORT_SYMBOL(neigh_table_clear); | |
2669 | EXPORT_SYMBOL(neigh_table_init); | |
2670 | EXPORT_SYMBOL(neigh_update); | |
2671 | EXPORT_SYMBOL(neigh_update_hhs); | |
2672 | EXPORT_SYMBOL(pneigh_enqueue); | |
2673 | EXPORT_SYMBOL(pneigh_lookup); | |
c7fb64db TG |
2674 | EXPORT_SYMBOL(neightbl_dump_info); |
2675 | EXPORT_SYMBOL(neightbl_set); | |
1da177e4 LT |
2676 | |
2677 | #ifdef CONFIG_ARPD | |
2678 | EXPORT_SYMBOL(neigh_app_ns); | |
2679 | #endif | |
2680 | #ifdef CONFIG_SYSCTL | |
2681 | EXPORT_SYMBOL(neigh_sysctl_register); | |
2682 | EXPORT_SYMBOL(neigh_sysctl_unregister); | |
2683 | #endif |