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; | |
45fc3b11 | 64 | #ifdef CONFIG_PROC_FS |
1da177e4 | 65 | static struct file_operations neigh_stat_seq_fops; |
45fc3b11 | 66 | #endif |
1da177e4 LT |
67 | |
68 | /* | |
69 | Neighbour hash table buckets are protected with rwlock tbl->lock. | |
70 | ||
71 | - All the scans/updates to hash buckets MUST be made under this lock. | |
72 | - NOTHING clever should be made under this lock: no callbacks | |
73 | to protocol backends, no attempts to send something to network. | |
74 | It will result in deadlocks, if backend/driver wants to use neighbour | |
75 | cache. | |
76 | - If the entry requires some non-trivial actions, increase | |
77 | its reference count and release table lock. | |
78 | ||
79 | Neighbour entries are protected: | |
80 | - with reference count. | |
81 | - with rwlock neigh->lock | |
82 | ||
83 | Reference count prevents destruction. | |
84 | ||
85 | neigh->lock mainly serializes ll address data and its validity state. | |
86 | However, the same lock is used to protect another entry fields: | |
87 | - timer | |
88 | - resolution queue | |
89 | ||
90 | Again, nothing clever shall be made under neigh->lock, | |
91 | the most complicated procedure, which we allow is dev->hard_header. | |
92 | It is supposed, that dev->hard_header is simplistic and does | |
93 | not make callbacks to neighbour tables. | |
94 | ||
95 | The last lock is neigh_tbl_lock. It is pure SMP lock, protecting | |
96 | list of neighbour tables. This list is used only in process context, | |
97 | */ | |
98 | ||
99 | static DEFINE_RWLOCK(neigh_tbl_lock); | |
100 | ||
101 | static int neigh_blackhole(struct sk_buff *skb) | |
102 | { | |
103 | kfree_skb(skb); | |
104 | return -ENETDOWN; | |
105 | } | |
106 | ||
107 | /* | |
108 | * It is random distribution in the interval (1/2)*base...(3/2)*base. | |
109 | * It corresponds to default IPv6 settings and is not overridable, | |
110 | * because it is really reasonable choice. | |
111 | */ | |
112 | ||
113 | unsigned long neigh_rand_reach_time(unsigned long base) | |
114 | { | |
115 | return (base ? (net_random() % base) + (base >> 1) : 0); | |
116 | } | |
117 | ||
118 | ||
119 | static int neigh_forced_gc(struct neigh_table *tbl) | |
120 | { | |
121 | int shrunk = 0; | |
122 | int i; | |
123 | ||
124 | NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs); | |
125 | ||
126 | write_lock_bh(&tbl->lock); | |
127 | for (i = 0; i <= tbl->hash_mask; i++) { | |
128 | struct neighbour *n, **np; | |
129 | ||
130 | np = &tbl->hash_buckets[i]; | |
131 | while ((n = *np) != NULL) { | |
132 | /* Neighbour record may be discarded if: | |
133 | * - nobody refers to it. | |
134 | * - it is not permanent | |
135 | */ | |
136 | write_lock(&n->lock); | |
137 | if (atomic_read(&n->refcnt) == 1 && | |
138 | !(n->nud_state & NUD_PERMANENT)) { | |
139 | *np = n->next; | |
140 | n->dead = 1; | |
141 | shrunk = 1; | |
142 | write_unlock(&n->lock); | |
143 | neigh_release(n); | |
144 | continue; | |
145 | } | |
146 | write_unlock(&n->lock); | |
147 | np = &n->next; | |
148 | } | |
149 | } | |
150 | ||
151 | tbl->last_flush = jiffies; | |
152 | ||
153 | write_unlock_bh(&tbl->lock); | |
154 | ||
155 | return shrunk; | |
156 | } | |
157 | ||
158 | static int neigh_del_timer(struct neighbour *n) | |
159 | { | |
160 | if ((n->nud_state & NUD_IN_TIMER) && | |
161 | del_timer(&n->timer)) { | |
162 | neigh_release(n); | |
163 | return 1; | |
164 | } | |
165 | return 0; | |
166 | } | |
167 | ||
168 | static void pneigh_queue_purge(struct sk_buff_head *list) | |
169 | { | |
170 | struct sk_buff *skb; | |
171 | ||
172 | while ((skb = skb_dequeue(list)) != NULL) { | |
173 | dev_put(skb->dev); | |
174 | kfree_skb(skb); | |
175 | } | |
176 | } | |
177 | ||
49636bb1 | 178 | static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev) |
1da177e4 LT |
179 | { |
180 | int i; | |
181 | ||
1da177e4 LT |
182 | for (i = 0; i <= tbl->hash_mask; i++) { |
183 | struct neighbour *n, **np = &tbl->hash_buckets[i]; | |
184 | ||
185 | while ((n = *np) != NULL) { | |
186 | if (dev && n->dev != dev) { | |
187 | np = &n->next; | |
188 | continue; | |
189 | } | |
190 | *np = n->next; | |
191 | write_lock(&n->lock); | |
192 | neigh_del_timer(n); | |
193 | n->dead = 1; | |
194 | ||
195 | if (atomic_read(&n->refcnt) != 1) { | |
196 | /* The most unpleasant situation. | |
197 | We must destroy neighbour entry, | |
198 | but someone still uses it. | |
199 | ||
200 | The destroy will be delayed until | |
201 | the last user releases us, but | |
202 | we must kill timers etc. and move | |
203 | it to safe state. | |
204 | */ | |
205 | skb_queue_purge(&n->arp_queue); | |
206 | n->output = neigh_blackhole; | |
207 | if (n->nud_state & NUD_VALID) | |
208 | n->nud_state = NUD_NOARP; | |
209 | else | |
210 | n->nud_state = NUD_NONE; | |
211 | NEIGH_PRINTK2("neigh %p is stray.\n", n); | |
212 | } | |
213 | write_unlock(&n->lock); | |
214 | neigh_release(n); | |
215 | } | |
216 | } | |
49636bb1 | 217 | } |
1da177e4 | 218 | |
49636bb1 HX |
219 | void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev) |
220 | { | |
221 | write_lock_bh(&tbl->lock); | |
222 | neigh_flush_dev(tbl, dev); | |
223 | write_unlock_bh(&tbl->lock); | |
224 | } | |
225 | ||
226 | int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev) | |
227 | { | |
228 | write_lock_bh(&tbl->lock); | |
229 | neigh_flush_dev(tbl, dev); | |
1da177e4 LT |
230 | pneigh_ifdown(tbl, dev); |
231 | write_unlock_bh(&tbl->lock); | |
232 | ||
233 | del_timer_sync(&tbl->proxy_timer); | |
234 | pneigh_queue_purge(&tbl->proxy_queue); | |
235 | return 0; | |
236 | } | |
237 | ||
238 | static struct neighbour *neigh_alloc(struct neigh_table *tbl) | |
239 | { | |
240 | struct neighbour *n = NULL; | |
241 | unsigned long now = jiffies; | |
242 | int entries; | |
243 | ||
244 | entries = atomic_inc_return(&tbl->entries) - 1; | |
245 | if (entries >= tbl->gc_thresh3 || | |
246 | (entries >= tbl->gc_thresh2 && | |
247 | time_after(now, tbl->last_flush + 5 * HZ))) { | |
248 | if (!neigh_forced_gc(tbl) && | |
249 | entries >= tbl->gc_thresh3) | |
250 | goto out_entries; | |
251 | } | |
252 | ||
253 | n = kmem_cache_alloc(tbl->kmem_cachep, SLAB_ATOMIC); | |
254 | if (!n) | |
255 | goto out_entries; | |
256 | ||
257 | memset(n, 0, tbl->entry_size); | |
258 | ||
259 | skb_queue_head_init(&n->arp_queue); | |
260 | rwlock_init(&n->lock); | |
261 | n->updated = n->used = now; | |
262 | n->nud_state = NUD_NONE; | |
263 | n->output = neigh_blackhole; | |
264 | n->parms = neigh_parms_clone(&tbl->parms); | |
265 | init_timer(&n->timer); | |
266 | n->timer.function = neigh_timer_handler; | |
267 | n->timer.data = (unsigned long)n; | |
268 | ||
269 | NEIGH_CACHE_STAT_INC(tbl, allocs); | |
270 | n->tbl = tbl; | |
271 | atomic_set(&n->refcnt, 1); | |
272 | n->dead = 1; | |
273 | out: | |
274 | return n; | |
275 | ||
276 | out_entries: | |
277 | atomic_dec(&tbl->entries); | |
278 | goto out; | |
279 | } | |
280 | ||
281 | static struct neighbour **neigh_hash_alloc(unsigned int entries) | |
282 | { | |
283 | unsigned long size = entries * sizeof(struct neighbour *); | |
284 | struct neighbour **ret; | |
285 | ||
286 | if (size <= PAGE_SIZE) { | |
287 | ret = kmalloc(size, GFP_ATOMIC); | |
288 | } else { | |
289 | ret = (struct neighbour **) | |
290 | __get_free_pages(GFP_ATOMIC, get_order(size)); | |
291 | } | |
292 | if (ret) | |
293 | memset(ret, 0, size); | |
294 | ||
295 | return ret; | |
296 | } | |
297 | ||
298 | static void neigh_hash_free(struct neighbour **hash, unsigned int entries) | |
299 | { | |
300 | unsigned long size = entries * sizeof(struct neighbour *); | |
301 | ||
302 | if (size <= PAGE_SIZE) | |
303 | kfree(hash); | |
304 | else | |
305 | free_pages((unsigned long)hash, get_order(size)); | |
306 | } | |
307 | ||
308 | static void neigh_hash_grow(struct neigh_table *tbl, unsigned long new_entries) | |
309 | { | |
310 | struct neighbour **new_hash, **old_hash; | |
311 | unsigned int i, new_hash_mask, old_entries; | |
312 | ||
313 | NEIGH_CACHE_STAT_INC(tbl, hash_grows); | |
314 | ||
315 | BUG_ON(new_entries & (new_entries - 1)); | |
316 | new_hash = neigh_hash_alloc(new_entries); | |
317 | if (!new_hash) | |
318 | return; | |
319 | ||
320 | old_entries = tbl->hash_mask + 1; | |
321 | new_hash_mask = new_entries - 1; | |
322 | old_hash = tbl->hash_buckets; | |
323 | ||
324 | get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd)); | |
325 | for (i = 0; i < old_entries; i++) { | |
326 | struct neighbour *n, *next; | |
327 | ||
328 | for (n = old_hash[i]; n; n = next) { | |
329 | unsigned int hash_val = tbl->hash(n->primary_key, n->dev); | |
330 | ||
331 | hash_val &= new_hash_mask; | |
332 | next = n->next; | |
333 | ||
334 | n->next = new_hash[hash_val]; | |
335 | new_hash[hash_val] = n; | |
336 | } | |
337 | } | |
338 | tbl->hash_buckets = new_hash; | |
339 | tbl->hash_mask = new_hash_mask; | |
340 | ||
341 | neigh_hash_free(old_hash, old_entries); | |
342 | } | |
343 | ||
344 | struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey, | |
345 | struct net_device *dev) | |
346 | { | |
347 | struct neighbour *n; | |
348 | int key_len = tbl->key_len; | |
349 | u32 hash_val = tbl->hash(pkey, dev) & tbl->hash_mask; | |
350 | ||
351 | NEIGH_CACHE_STAT_INC(tbl, lookups); | |
352 | ||
353 | read_lock_bh(&tbl->lock); | |
354 | for (n = tbl->hash_buckets[hash_val]; n; n = n->next) { | |
355 | if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) { | |
356 | neigh_hold(n); | |
357 | NEIGH_CACHE_STAT_INC(tbl, hits); | |
358 | break; | |
359 | } | |
360 | } | |
361 | read_unlock_bh(&tbl->lock); | |
362 | return n; | |
363 | } | |
364 | ||
365 | struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, const void *pkey) | |
366 | { | |
367 | struct neighbour *n; | |
368 | int key_len = tbl->key_len; | |
369 | u32 hash_val = tbl->hash(pkey, NULL) & tbl->hash_mask; | |
370 | ||
371 | NEIGH_CACHE_STAT_INC(tbl, lookups); | |
372 | ||
373 | read_lock_bh(&tbl->lock); | |
374 | for (n = tbl->hash_buckets[hash_val]; n; n = n->next) { | |
375 | if (!memcmp(n->primary_key, pkey, key_len)) { | |
376 | neigh_hold(n); | |
377 | NEIGH_CACHE_STAT_INC(tbl, hits); | |
378 | break; | |
379 | } | |
380 | } | |
381 | read_unlock_bh(&tbl->lock); | |
382 | return n; | |
383 | } | |
384 | ||
385 | struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey, | |
386 | struct net_device *dev) | |
387 | { | |
388 | u32 hash_val; | |
389 | int key_len = tbl->key_len; | |
390 | int error; | |
391 | struct neighbour *n1, *rc, *n = neigh_alloc(tbl); | |
392 | ||
393 | if (!n) { | |
394 | rc = ERR_PTR(-ENOBUFS); | |
395 | goto out; | |
396 | } | |
397 | ||
398 | memcpy(n->primary_key, pkey, key_len); | |
399 | n->dev = dev; | |
400 | dev_hold(dev); | |
401 | ||
402 | /* Protocol specific setup. */ | |
403 | if (tbl->constructor && (error = tbl->constructor(n)) < 0) { | |
404 | rc = ERR_PTR(error); | |
405 | goto out_neigh_release; | |
406 | } | |
407 | ||
408 | /* Device specific setup. */ | |
409 | if (n->parms->neigh_setup && | |
410 | (error = n->parms->neigh_setup(n)) < 0) { | |
411 | rc = ERR_PTR(error); | |
412 | goto out_neigh_release; | |
413 | } | |
414 | ||
415 | n->confirmed = jiffies - (n->parms->base_reachable_time << 1); | |
416 | ||
417 | write_lock_bh(&tbl->lock); | |
418 | ||
419 | if (atomic_read(&tbl->entries) > (tbl->hash_mask + 1)) | |
420 | neigh_hash_grow(tbl, (tbl->hash_mask + 1) << 1); | |
421 | ||
422 | hash_val = tbl->hash(pkey, dev) & tbl->hash_mask; | |
423 | ||
424 | if (n->parms->dead) { | |
425 | rc = ERR_PTR(-EINVAL); | |
426 | goto out_tbl_unlock; | |
427 | } | |
428 | ||
429 | for (n1 = tbl->hash_buckets[hash_val]; n1; n1 = n1->next) { | |
430 | if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) { | |
431 | neigh_hold(n1); | |
432 | rc = n1; | |
433 | goto out_tbl_unlock; | |
434 | } | |
435 | } | |
436 | ||
437 | n->next = tbl->hash_buckets[hash_val]; | |
438 | tbl->hash_buckets[hash_val] = n; | |
439 | n->dead = 0; | |
440 | neigh_hold(n); | |
441 | write_unlock_bh(&tbl->lock); | |
442 | NEIGH_PRINTK2("neigh %p is created.\n", n); | |
443 | rc = n; | |
444 | out: | |
445 | return rc; | |
446 | out_tbl_unlock: | |
447 | write_unlock_bh(&tbl->lock); | |
448 | out_neigh_release: | |
449 | neigh_release(n); | |
450 | goto out; | |
451 | } | |
452 | ||
453 | struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl, const void *pkey, | |
454 | struct net_device *dev, int creat) | |
455 | { | |
456 | struct pneigh_entry *n; | |
457 | int key_len = tbl->key_len; | |
458 | u32 hash_val = *(u32 *)(pkey + key_len - 4); | |
459 | ||
460 | hash_val ^= (hash_val >> 16); | |
461 | hash_val ^= hash_val >> 8; | |
462 | hash_val ^= hash_val >> 4; | |
463 | hash_val &= PNEIGH_HASHMASK; | |
464 | ||
465 | read_lock_bh(&tbl->lock); | |
466 | ||
467 | for (n = tbl->phash_buckets[hash_val]; n; n = n->next) { | |
468 | if (!memcmp(n->key, pkey, key_len) && | |
469 | (n->dev == dev || !n->dev)) { | |
470 | read_unlock_bh(&tbl->lock); | |
471 | goto out; | |
472 | } | |
473 | } | |
474 | read_unlock_bh(&tbl->lock); | |
475 | n = NULL; | |
476 | if (!creat) | |
477 | goto out; | |
478 | ||
479 | n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL); | |
480 | if (!n) | |
481 | goto out; | |
482 | ||
483 | memcpy(n->key, pkey, key_len); | |
484 | n->dev = dev; | |
485 | if (dev) | |
486 | dev_hold(dev); | |
487 | ||
488 | if (tbl->pconstructor && tbl->pconstructor(n)) { | |
489 | if (dev) | |
490 | dev_put(dev); | |
491 | kfree(n); | |
492 | n = NULL; | |
493 | goto out; | |
494 | } | |
495 | ||
496 | write_lock_bh(&tbl->lock); | |
497 | n->next = tbl->phash_buckets[hash_val]; | |
498 | tbl->phash_buckets[hash_val] = n; | |
499 | write_unlock_bh(&tbl->lock); | |
500 | out: | |
501 | return n; | |
502 | } | |
503 | ||
504 | ||
505 | int pneigh_delete(struct neigh_table *tbl, const void *pkey, | |
506 | struct net_device *dev) | |
507 | { | |
508 | struct pneigh_entry *n, **np; | |
509 | int key_len = tbl->key_len; | |
510 | u32 hash_val = *(u32 *)(pkey + key_len - 4); | |
511 | ||
512 | hash_val ^= (hash_val >> 16); | |
513 | hash_val ^= hash_val >> 8; | |
514 | hash_val ^= hash_val >> 4; | |
515 | hash_val &= PNEIGH_HASHMASK; | |
516 | ||
517 | write_lock_bh(&tbl->lock); | |
518 | for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL; | |
519 | np = &n->next) { | |
520 | if (!memcmp(n->key, pkey, key_len) && n->dev == dev) { | |
521 | *np = n->next; | |
522 | write_unlock_bh(&tbl->lock); | |
523 | if (tbl->pdestructor) | |
524 | tbl->pdestructor(n); | |
525 | if (n->dev) | |
526 | dev_put(n->dev); | |
527 | kfree(n); | |
528 | return 0; | |
529 | } | |
530 | } | |
531 | write_unlock_bh(&tbl->lock); | |
532 | return -ENOENT; | |
533 | } | |
534 | ||
535 | static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev) | |
536 | { | |
537 | struct pneigh_entry *n, **np; | |
538 | u32 h; | |
539 | ||
540 | for (h = 0; h <= PNEIGH_HASHMASK; h++) { | |
541 | np = &tbl->phash_buckets[h]; | |
542 | while ((n = *np) != NULL) { | |
543 | if (!dev || n->dev == dev) { | |
544 | *np = n->next; | |
545 | if (tbl->pdestructor) | |
546 | tbl->pdestructor(n); | |
547 | if (n->dev) | |
548 | dev_put(n->dev); | |
549 | kfree(n); | |
550 | continue; | |
551 | } | |
552 | np = &n->next; | |
553 | } | |
554 | } | |
555 | return -ENOENT; | |
556 | } | |
557 | ||
558 | ||
559 | /* | |
560 | * neighbour must already be out of the table; | |
561 | * | |
562 | */ | |
563 | void neigh_destroy(struct neighbour *neigh) | |
564 | { | |
565 | struct hh_cache *hh; | |
566 | ||
567 | NEIGH_CACHE_STAT_INC(neigh->tbl, destroys); | |
568 | ||
569 | if (!neigh->dead) { | |
570 | printk(KERN_WARNING | |
571 | "Destroying alive neighbour %p\n", neigh); | |
572 | dump_stack(); | |
573 | return; | |
574 | } | |
575 | ||
576 | if (neigh_del_timer(neigh)) | |
577 | printk(KERN_WARNING "Impossible event.\n"); | |
578 | ||
579 | while ((hh = neigh->hh) != NULL) { | |
580 | neigh->hh = hh->hh_next; | |
581 | hh->hh_next = NULL; | |
582 | write_lock_bh(&hh->hh_lock); | |
583 | hh->hh_output = neigh_blackhole; | |
584 | write_unlock_bh(&hh->hh_lock); | |
585 | if (atomic_dec_and_test(&hh->hh_refcnt)) | |
586 | kfree(hh); | |
587 | } | |
588 | ||
589 | if (neigh->ops && neigh->ops->destructor) | |
590 | (neigh->ops->destructor)(neigh); | |
591 | ||
592 | skb_queue_purge(&neigh->arp_queue); | |
593 | ||
594 | dev_put(neigh->dev); | |
595 | neigh_parms_put(neigh->parms); | |
596 | ||
597 | NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh); | |
598 | ||
599 | atomic_dec(&neigh->tbl->entries); | |
600 | kmem_cache_free(neigh->tbl->kmem_cachep, neigh); | |
601 | } | |
602 | ||
603 | /* Neighbour state is suspicious; | |
604 | disable fast path. | |
605 | ||
606 | Called with write_locked neigh. | |
607 | */ | |
608 | static void neigh_suspect(struct neighbour *neigh) | |
609 | { | |
610 | struct hh_cache *hh; | |
611 | ||
612 | NEIGH_PRINTK2("neigh %p is suspected.\n", neigh); | |
613 | ||
614 | neigh->output = neigh->ops->output; | |
615 | ||
616 | for (hh = neigh->hh; hh; hh = hh->hh_next) | |
617 | hh->hh_output = neigh->ops->output; | |
618 | } | |
619 | ||
620 | /* Neighbour state is OK; | |
621 | enable fast path. | |
622 | ||
623 | Called with write_locked neigh. | |
624 | */ | |
625 | static void neigh_connect(struct neighbour *neigh) | |
626 | { | |
627 | struct hh_cache *hh; | |
628 | ||
629 | NEIGH_PRINTK2("neigh %p is connected.\n", neigh); | |
630 | ||
631 | neigh->output = neigh->ops->connected_output; | |
632 | ||
633 | for (hh = neigh->hh; hh; hh = hh->hh_next) | |
634 | hh->hh_output = neigh->ops->hh_output; | |
635 | } | |
636 | ||
637 | static void neigh_periodic_timer(unsigned long arg) | |
638 | { | |
639 | struct neigh_table *tbl = (struct neigh_table *)arg; | |
640 | struct neighbour *n, **np; | |
641 | unsigned long expire, now = jiffies; | |
642 | ||
643 | NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs); | |
644 | ||
645 | write_lock(&tbl->lock); | |
646 | ||
647 | /* | |
648 | * periodically recompute ReachableTime from random function | |
649 | */ | |
650 | ||
651 | if (time_after(now, tbl->last_rand + 300 * HZ)) { | |
652 | struct neigh_parms *p; | |
653 | tbl->last_rand = now; | |
654 | for (p = &tbl->parms; p; p = p->next) | |
655 | p->reachable_time = | |
656 | neigh_rand_reach_time(p->base_reachable_time); | |
657 | } | |
658 | ||
659 | np = &tbl->hash_buckets[tbl->hash_chain_gc]; | |
660 | tbl->hash_chain_gc = ((tbl->hash_chain_gc + 1) & tbl->hash_mask); | |
661 | ||
662 | while ((n = *np) != NULL) { | |
663 | unsigned int state; | |
664 | ||
665 | write_lock(&n->lock); | |
666 | ||
667 | state = n->nud_state; | |
668 | if (state & (NUD_PERMANENT | NUD_IN_TIMER)) { | |
669 | write_unlock(&n->lock); | |
670 | goto next_elt; | |
671 | } | |
672 | ||
673 | if (time_before(n->used, n->confirmed)) | |
674 | n->used = n->confirmed; | |
675 | ||
676 | if (atomic_read(&n->refcnt) == 1 && | |
677 | (state == NUD_FAILED || | |
678 | time_after(now, n->used + n->parms->gc_staletime))) { | |
679 | *np = n->next; | |
680 | n->dead = 1; | |
681 | write_unlock(&n->lock); | |
682 | neigh_release(n); | |
683 | continue; | |
684 | } | |
685 | write_unlock(&n->lock); | |
686 | ||
687 | next_elt: | |
688 | np = &n->next; | |
689 | } | |
690 | ||
691 | /* Cycle through all hash buckets every base_reachable_time/2 ticks. | |
692 | * ARP entry timeouts range from 1/2 base_reachable_time to 3/2 | |
693 | * base_reachable_time. | |
694 | */ | |
695 | expire = tbl->parms.base_reachable_time >> 1; | |
696 | expire /= (tbl->hash_mask + 1); | |
697 | if (!expire) | |
698 | expire = 1; | |
699 | ||
700 | mod_timer(&tbl->gc_timer, now + expire); | |
701 | ||
702 | write_unlock(&tbl->lock); | |
703 | } | |
704 | ||
705 | static __inline__ int neigh_max_probes(struct neighbour *n) | |
706 | { | |
707 | struct neigh_parms *p = n->parms; | |
708 | return (n->nud_state & NUD_PROBE ? | |
709 | p->ucast_probes : | |
710 | p->ucast_probes + p->app_probes + p->mcast_probes); | |
711 | } | |
712 | ||
667347f1 DM |
713 | static inline void neigh_add_timer(struct neighbour *n, unsigned long when) |
714 | { | |
715 | if (unlikely(mod_timer(&n->timer, when))) { | |
716 | printk("NEIGH: BUG, double timer add, state is %x\n", | |
717 | n->nud_state); | |
20375502 | 718 | dump_stack(); |
667347f1 DM |
719 | } |
720 | } | |
1da177e4 LT |
721 | |
722 | /* Called when a timer expires for a neighbour entry. */ | |
723 | ||
724 | static void neigh_timer_handler(unsigned long arg) | |
725 | { | |
726 | unsigned long now, next; | |
727 | struct neighbour *neigh = (struct neighbour *)arg; | |
728 | unsigned state; | |
729 | int notify = 0; | |
730 | ||
731 | write_lock(&neigh->lock); | |
732 | ||
733 | state = neigh->nud_state; | |
734 | now = jiffies; | |
735 | next = now + HZ; | |
736 | ||
737 | if (!(state & NUD_IN_TIMER)) { | |
738 | #ifndef CONFIG_SMP | |
739 | printk(KERN_WARNING "neigh: timer & !nud_in_timer\n"); | |
740 | #endif | |
741 | goto out; | |
742 | } | |
743 | ||
744 | if (state & NUD_REACHABLE) { | |
745 | if (time_before_eq(now, | |
746 | neigh->confirmed + neigh->parms->reachable_time)) { | |
747 | NEIGH_PRINTK2("neigh %p is still alive.\n", neigh); | |
748 | next = neigh->confirmed + neigh->parms->reachable_time; | |
749 | } else if (time_before_eq(now, | |
750 | neigh->used + neigh->parms->delay_probe_time)) { | |
751 | NEIGH_PRINTK2("neigh %p is delayed.\n", neigh); | |
752 | neigh->nud_state = NUD_DELAY; | |
753 | neigh_suspect(neigh); | |
754 | next = now + neigh->parms->delay_probe_time; | |
755 | } else { | |
756 | NEIGH_PRINTK2("neigh %p is suspected.\n", neigh); | |
757 | neigh->nud_state = NUD_STALE; | |
758 | neigh_suspect(neigh); | |
759 | } | |
760 | } else if (state & NUD_DELAY) { | |
761 | if (time_before_eq(now, | |
762 | neigh->confirmed + neigh->parms->delay_probe_time)) { | |
763 | NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh); | |
764 | neigh->nud_state = NUD_REACHABLE; | |
765 | neigh_connect(neigh); | |
766 | next = neigh->confirmed + neigh->parms->reachable_time; | |
767 | } else { | |
768 | NEIGH_PRINTK2("neigh %p is probed.\n", neigh); | |
769 | neigh->nud_state = NUD_PROBE; | |
770 | atomic_set(&neigh->probes, 0); | |
771 | next = now + neigh->parms->retrans_time; | |
772 | } | |
773 | } else { | |
774 | /* NUD_PROBE|NUD_INCOMPLETE */ | |
775 | next = now + neigh->parms->retrans_time; | |
776 | } | |
777 | ||
778 | if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) && | |
779 | atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) { | |
780 | struct sk_buff *skb; | |
781 | ||
782 | neigh->nud_state = NUD_FAILED; | |
783 | notify = 1; | |
784 | NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed); | |
785 | NEIGH_PRINTK2("neigh %p is failed.\n", neigh); | |
786 | ||
787 | /* It is very thin place. report_unreachable is very complicated | |
788 | routine. Particularly, it can hit the same neighbour entry! | |
789 | ||
790 | So that, we try to be accurate and avoid dead loop. --ANK | |
791 | */ | |
792 | while (neigh->nud_state == NUD_FAILED && | |
793 | (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) { | |
794 | write_unlock(&neigh->lock); | |
795 | neigh->ops->error_report(neigh, skb); | |
796 | write_lock(&neigh->lock); | |
797 | } | |
798 | skb_queue_purge(&neigh->arp_queue); | |
799 | } | |
800 | ||
801 | if (neigh->nud_state & NUD_IN_TIMER) { | |
1da177e4 LT |
802 | if (time_before(next, jiffies + HZ/2)) |
803 | next = jiffies + HZ/2; | |
6fb9974f HX |
804 | if (!mod_timer(&neigh->timer, next)) |
805 | neigh_hold(neigh); | |
1da177e4 LT |
806 | } |
807 | if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) { | |
808 | struct sk_buff *skb = skb_peek(&neigh->arp_queue); | |
809 | /* keep skb alive even if arp_queue overflows */ | |
810 | if (skb) | |
811 | skb_get(skb); | |
812 | write_unlock(&neigh->lock); | |
813 | neigh->ops->solicit(neigh, skb); | |
814 | atomic_inc(&neigh->probes); | |
815 | if (skb) | |
816 | kfree_skb(skb); | |
817 | } else { | |
818 | out: | |
819 | write_unlock(&neigh->lock); | |
820 | } | |
821 | ||
822 | #ifdef CONFIG_ARPD | |
823 | if (notify && neigh->parms->app_probes) | |
824 | neigh_app_notify(neigh); | |
825 | #endif | |
826 | neigh_release(neigh); | |
827 | } | |
828 | ||
829 | int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb) | |
830 | { | |
831 | int rc; | |
832 | unsigned long now; | |
833 | ||
834 | write_lock_bh(&neigh->lock); | |
835 | ||
836 | rc = 0; | |
837 | if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE)) | |
838 | goto out_unlock_bh; | |
839 | ||
840 | now = jiffies; | |
841 | ||
842 | if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) { | |
843 | if (neigh->parms->mcast_probes + neigh->parms->app_probes) { | |
844 | atomic_set(&neigh->probes, neigh->parms->ucast_probes); | |
845 | neigh->nud_state = NUD_INCOMPLETE; | |
846 | neigh_hold(neigh); | |
667347f1 | 847 | neigh_add_timer(neigh, now + 1); |
1da177e4 LT |
848 | } else { |
849 | neigh->nud_state = NUD_FAILED; | |
850 | write_unlock_bh(&neigh->lock); | |
851 | ||
852 | if (skb) | |
853 | kfree_skb(skb); | |
854 | return 1; | |
855 | } | |
856 | } else if (neigh->nud_state & NUD_STALE) { | |
857 | NEIGH_PRINTK2("neigh %p is delayed.\n", neigh); | |
858 | neigh_hold(neigh); | |
859 | neigh->nud_state = NUD_DELAY; | |
667347f1 DM |
860 | neigh_add_timer(neigh, |
861 | jiffies + neigh->parms->delay_probe_time); | |
1da177e4 LT |
862 | } |
863 | ||
864 | if (neigh->nud_state == NUD_INCOMPLETE) { | |
865 | if (skb) { | |
866 | if (skb_queue_len(&neigh->arp_queue) >= | |
867 | neigh->parms->queue_len) { | |
868 | struct sk_buff *buff; | |
869 | buff = neigh->arp_queue.next; | |
870 | __skb_unlink(buff, &neigh->arp_queue); | |
871 | kfree_skb(buff); | |
872 | } | |
873 | __skb_queue_tail(&neigh->arp_queue, skb); | |
874 | } | |
875 | rc = 1; | |
876 | } | |
877 | out_unlock_bh: | |
878 | write_unlock_bh(&neigh->lock); | |
879 | return rc; | |
880 | } | |
881 | ||
882 | static __inline__ void neigh_update_hhs(struct neighbour *neigh) | |
883 | { | |
884 | struct hh_cache *hh; | |
885 | void (*update)(struct hh_cache*, struct net_device*, unsigned char *) = | |
886 | neigh->dev->header_cache_update; | |
887 | ||
888 | if (update) { | |
889 | for (hh = neigh->hh; hh; hh = hh->hh_next) { | |
890 | write_lock_bh(&hh->hh_lock); | |
891 | update(hh, neigh->dev, neigh->ha); | |
892 | write_unlock_bh(&hh->hh_lock); | |
893 | } | |
894 | } | |
895 | } | |
896 | ||
897 | ||
898 | ||
899 | /* Generic update routine. | |
900 | -- lladdr is new lladdr or NULL, if it is not supplied. | |
901 | -- new is new state. | |
902 | -- flags | |
903 | NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr, | |
904 | if it is different. | |
905 | NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected" | |
906 | lladdr instead of overriding it | |
907 | if it is different. | |
908 | It also allows to retain current state | |
909 | if lladdr is unchanged. | |
910 | NEIGH_UPDATE_F_ADMIN means that the change is administrative. | |
911 | ||
912 | NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing | |
913 | NTF_ROUTER flag. | |
914 | NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as | |
915 | a router. | |
916 | ||
917 | Caller MUST hold reference count on the entry. | |
918 | */ | |
919 | ||
920 | int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, | |
921 | u32 flags) | |
922 | { | |
923 | u8 old; | |
924 | int err; | |
925 | #ifdef CONFIG_ARPD | |
926 | int notify = 0; | |
927 | #endif | |
928 | struct net_device *dev; | |
929 | int update_isrouter = 0; | |
930 | ||
931 | write_lock_bh(&neigh->lock); | |
932 | ||
933 | dev = neigh->dev; | |
934 | old = neigh->nud_state; | |
935 | err = -EPERM; | |
936 | ||
937 | if (!(flags & NEIGH_UPDATE_F_ADMIN) && | |
938 | (old & (NUD_NOARP | NUD_PERMANENT))) | |
939 | goto out; | |
940 | ||
941 | if (!(new & NUD_VALID)) { | |
942 | neigh_del_timer(neigh); | |
943 | if (old & NUD_CONNECTED) | |
944 | neigh_suspect(neigh); | |
945 | neigh->nud_state = new; | |
946 | err = 0; | |
947 | #ifdef CONFIG_ARPD | |
948 | notify = old & NUD_VALID; | |
949 | #endif | |
950 | goto out; | |
951 | } | |
952 | ||
953 | /* Compare new lladdr with cached one */ | |
954 | if (!dev->addr_len) { | |
955 | /* First case: device needs no address. */ | |
956 | lladdr = neigh->ha; | |
957 | } else if (lladdr) { | |
958 | /* The second case: if something is already cached | |
959 | and a new address is proposed: | |
960 | - compare new & old | |
961 | - if they are different, check override flag | |
962 | */ | |
963 | if ((old & NUD_VALID) && | |
964 | !memcmp(lladdr, neigh->ha, dev->addr_len)) | |
965 | lladdr = neigh->ha; | |
966 | } else { | |
967 | /* No address is supplied; if we know something, | |
968 | use it, otherwise discard the request. | |
969 | */ | |
970 | err = -EINVAL; | |
971 | if (!(old & NUD_VALID)) | |
972 | goto out; | |
973 | lladdr = neigh->ha; | |
974 | } | |
975 | ||
976 | if (new & NUD_CONNECTED) | |
977 | neigh->confirmed = jiffies; | |
978 | neigh->updated = jiffies; | |
979 | ||
980 | /* If entry was valid and address is not changed, | |
981 | do not change entry state, if new one is STALE. | |
982 | */ | |
983 | err = 0; | |
984 | update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER; | |
985 | if (old & NUD_VALID) { | |
986 | if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) { | |
987 | update_isrouter = 0; | |
988 | if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) && | |
989 | (old & NUD_CONNECTED)) { | |
990 | lladdr = neigh->ha; | |
991 | new = NUD_STALE; | |
992 | } else | |
993 | goto out; | |
994 | } else { | |
995 | if (lladdr == neigh->ha && new == NUD_STALE && | |
996 | ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) || | |
997 | (old & NUD_CONNECTED)) | |
998 | ) | |
999 | new = old; | |
1000 | } | |
1001 | } | |
1002 | ||
1003 | if (new != old) { | |
1004 | neigh_del_timer(neigh); | |
1005 | if (new & NUD_IN_TIMER) { | |
1006 | neigh_hold(neigh); | |
667347f1 | 1007 | neigh_add_timer(neigh, (jiffies + |
1da177e4 | 1008 | ((new & NUD_REACHABLE) ? |
667347f1 DM |
1009 | neigh->parms->reachable_time : |
1010 | 0))); | |
1da177e4 LT |
1011 | } |
1012 | neigh->nud_state = new; | |
1013 | } | |
1014 | ||
1015 | if (lladdr != neigh->ha) { | |
1016 | memcpy(&neigh->ha, lladdr, dev->addr_len); | |
1017 | neigh_update_hhs(neigh); | |
1018 | if (!(new & NUD_CONNECTED)) | |
1019 | neigh->confirmed = jiffies - | |
1020 | (neigh->parms->base_reachable_time << 1); | |
1021 | #ifdef CONFIG_ARPD | |
1022 | notify = 1; | |
1023 | #endif | |
1024 | } | |
1025 | if (new == old) | |
1026 | goto out; | |
1027 | if (new & NUD_CONNECTED) | |
1028 | neigh_connect(neigh); | |
1029 | else | |
1030 | neigh_suspect(neigh); | |
1031 | if (!(old & NUD_VALID)) { | |
1032 | struct sk_buff *skb; | |
1033 | ||
1034 | /* Again: avoid dead loop if something went wrong */ | |
1035 | ||
1036 | while (neigh->nud_state & NUD_VALID && | |
1037 | (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) { | |
1038 | struct neighbour *n1 = neigh; | |
1039 | write_unlock_bh(&neigh->lock); | |
1040 | /* On shaper/eql skb->dst->neighbour != neigh :( */ | |
1041 | if (skb->dst && skb->dst->neighbour) | |
1042 | n1 = skb->dst->neighbour; | |
1043 | n1->output(skb); | |
1044 | write_lock_bh(&neigh->lock); | |
1045 | } | |
1046 | skb_queue_purge(&neigh->arp_queue); | |
1047 | } | |
1048 | out: | |
1049 | if (update_isrouter) { | |
1050 | neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ? | |
1051 | (neigh->flags | NTF_ROUTER) : | |
1052 | (neigh->flags & ~NTF_ROUTER); | |
1053 | } | |
1054 | write_unlock_bh(&neigh->lock); | |
1055 | #ifdef CONFIG_ARPD | |
1056 | if (notify && neigh->parms->app_probes) | |
1057 | neigh_app_notify(neigh); | |
1058 | #endif | |
1059 | return err; | |
1060 | } | |
1061 | ||
1062 | struct neighbour *neigh_event_ns(struct neigh_table *tbl, | |
1063 | u8 *lladdr, void *saddr, | |
1064 | struct net_device *dev) | |
1065 | { | |
1066 | struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev, | |
1067 | lladdr || !dev->addr_len); | |
1068 | if (neigh) | |
1069 | neigh_update(neigh, lladdr, NUD_STALE, | |
1070 | NEIGH_UPDATE_F_OVERRIDE); | |
1071 | return neigh; | |
1072 | } | |
1073 | ||
1074 | static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst, | |
1075 | u16 protocol) | |
1076 | { | |
1077 | struct hh_cache *hh; | |
1078 | struct net_device *dev = dst->dev; | |
1079 | ||
1080 | for (hh = n->hh; hh; hh = hh->hh_next) | |
1081 | if (hh->hh_type == protocol) | |
1082 | break; | |
1083 | ||
1084 | if (!hh && (hh = kmalloc(sizeof(*hh), GFP_ATOMIC)) != NULL) { | |
1085 | memset(hh, 0, sizeof(struct hh_cache)); | |
1086 | rwlock_init(&hh->hh_lock); | |
1087 | hh->hh_type = protocol; | |
1088 | atomic_set(&hh->hh_refcnt, 0); | |
1089 | hh->hh_next = NULL; | |
1090 | if (dev->hard_header_cache(n, hh)) { | |
1091 | kfree(hh); | |
1092 | hh = NULL; | |
1093 | } else { | |
1094 | atomic_inc(&hh->hh_refcnt); | |
1095 | hh->hh_next = n->hh; | |
1096 | n->hh = hh; | |
1097 | if (n->nud_state & NUD_CONNECTED) | |
1098 | hh->hh_output = n->ops->hh_output; | |
1099 | else | |
1100 | hh->hh_output = n->ops->output; | |
1101 | } | |
1102 | } | |
1103 | if (hh) { | |
1104 | atomic_inc(&hh->hh_refcnt); | |
1105 | dst->hh = hh; | |
1106 | } | |
1107 | } | |
1108 | ||
1109 | /* This function can be used in contexts, where only old dev_queue_xmit | |
1110 | worked, f.e. if you want to override normal output path (eql, shaper), | |
1111 | but resolution is not made yet. | |
1112 | */ | |
1113 | ||
1114 | int neigh_compat_output(struct sk_buff *skb) | |
1115 | { | |
1116 | struct net_device *dev = skb->dev; | |
1117 | ||
1118 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1119 | ||
1120 | if (dev->hard_header && | |
1121 | dev->hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL, | |
1122 | skb->len) < 0 && | |
1123 | dev->rebuild_header(skb)) | |
1124 | return 0; | |
1125 | ||
1126 | return dev_queue_xmit(skb); | |
1127 | } | |
1128 | ||
1129 | /* Slow and careful. */ | |
1130 | ||
1131 | int neigh_resolve_output(struct sk_buff *skb) | |
1132 | { | |
1133 | struct dst_entry *dst = skb->dst; | |
1134 | struct neighbour *neigh; | |
1135 | int rc = 0; | |
1136 | ||
1137 | if (!dst || !(neigh = dst->neighbour)) | |
1138 | goto discard; | |
1139 | ||
1140 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1141 | ||
1142 | if (!neigh_event_send(neigh, skb)) { | |
1143 | int err; | |
1144 | struct net_device *dev = neigh->dev; | |
1145 | if (dev->hard_header_cache && !dst->hh) { | |
1146 | write_lock_bh(&neigh->lock); | |
1147 | if (!dst->hh) | |
1148 | neigh_hh_init(neigh, dst, dst->ops->protocol); | |
1149 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1150 | neigh->ha, NULL, skb->len); | |
1151 | write_unlock_bh(&neigh->lock); | |
1152 | } else { | |
1153 | read_lock_bh(&neigh->lock); | |
1154 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1155 | neigh->ha, NULL, skb->len); | |
1156 | read_unlock_bh(&neigh->lock); | |
1157 | } | |
1158 | if (err >= 0) | |
1159 | rc = neigh->ops->queue_xmit(skb); | |
1160 | else | |
1161 | goto out_kfree_skb; | |
1162 | } | |
1163 | out: | |
1164 | return rc; | |
1165 | discard: | |
1166 | NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n", | |
1167 | dst, dst ? dst->neighbour : NULL); | |
1168 | out_kfree_skb: | |
1169 | rc = -EINVAL; | |
1170 | kfree_skb(skb); | |
1171 | goto out; | |
1172 | } | |
1173 | ||
1174 | /* As fast as possible without hh cache */ | |
1175 | ||
1176 | int neigh_connected_output(struct sk_buff *skb) | |
1177 | { | |
1178 | int err; | |
1179 | struct dst_entry *dst = skb->dst; | |
1180 | struct neighbour *neigh = dst->neighbour; | |
1181 | struct net_device *dev = neigh->dev; | |
1182 | ||
1183 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1184 | ||
1185 | read_lock_bh(&neigh->lock); | |
1186 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1187 | neigh->ha, NULL, skb->len); | |
1188 | read_unlock_bh(&neigh->lock); | |
1189 | if (err >= 0) | |
1190 | err = neigh->ops->queue_xmit(skb); | |
1191 | else { | |
1192 | err = -EINVAL; | |
1193 | kfree_skb(skb); | |
1194 | } | |
1195 | return err; | |
1196 | } | |
1197 | ||
1198 | static void neigh_proxy_process(unsigned long arg) | |
1199 | { | |
1200 | struct neigh_table *tbl = (struct neigh_table *)arg; | |
1201 | long sched_next = 0; | |
1202 | unsigned long now = jiffies; | |
1203 | struct sk_buff *skb; | |
1204 | ||
1205 | spin_lock(&tbl->proxy_queue.lock); | |
1206 | ||
1207 | skb = tbl->proxy_queue.next; | |
1208 | ||
1209 | while (skb != (struct sk_buff *)&tbl->proxy_queue) { | |
1210 | struct sk_buff *back = skb; | |
a61bbcf2 | 1211 | long tdif = NEIGH_CB(back)->sched_next - now; |
1da177e4 LT |
1212 | |
1213 | skb = skb->next; | |
1214 | if (tdif <= 0) { | |
1215 | struct net_device *dev = back->dev; | |
1216 | __skb_unlink(back, &tbl->proxy_queue); | |
1217 | if (tbl->proxy_redo && netif_running(dev)) | |
1218 | tbl->proxy_redo(back); | |
1219 | else | |
1220 | kfree_skb(back); | |
1221 | ||
1222 | dev_put(dev); | |
1223 | } else if (!sched_next || tdif < sched_next) | |
1224 | sched_next = tdif; | |
1225 | } | |
1226 | del_timer(&tbl->proxy_timer); | |
1227 | if (sched_next) | |
1228 | mod_timer(&tbl->proxy_timer, jiffies + sched_next); | |
1229 | spin_unlock(&tbl->proxy_queue.lock); | |
1230 | } | |
1231 | ||
1232 | void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p, | |
1233 | struct sk_buff *skb) | |
1234 | { | |
1235 | unsigned long now = jiffies; | |
1236 | unsigned long sched_next = now + (net_random() % p->proxy_delay); | |
1237 | ||
1238 | if (tbl->proxy_queue.qlen > p->proxy_qlen) { | |
1239 | kfree_skb(skb); | |
1240 | return; | |
1241 | } | |
a61bbcf2 PM |
1242 | |
1243 | NEIGH_CB(skb)->sched_next = sched_next; | |
1244 | NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED; | |
1da177e4 LT |
1245 | |
1246 | spin_lock(&tbl->proxy_queue.lock); | |
1247 | if (del_timer(&tbl->proxy_timer)) { | |
1248 | if (time_before(tbl->proxy_timer.expires, sched_next)) | |
1249 | sched_next = tbl->proxy_timer.expires; | |
1250 | } | |
1251 | dst_release(skb->dst); | |
1252 | skb->dst = NULL; | |
1253 | dev_hold(skb->dev); | |
1254 | __skb_queue_tail(&tbl->proxy_queue, skb); | |
1255 | mod_timer(&tbl->proxy_timer, sched_next); | |
1256 | spin_unlock(&tbl->proxy_queue.lock); | |
1257 | } | |
1258 | ||
1259 | ||
1260 | struct neigh_parms *neigh_parms_alloc(struct net_device *dev, | |
1261 | struct neigh_table *tbl) | |
1262 | { | |
1263 | struct neigh_parms *p = kmalloc(sizeof(*p), GFP_KERNEL); | |
1264 | ||
1265 | if (p) { | |
1266 | memcpy(p, &tbl->parms, sizeof(*p)); | |
1267 | p->tbl = tbl; | |
1268 | atomic_set(&p->refcnt, 1); | |
1269 | INIT_RCU_HEAD(&p->rcu_head); | |
1270 | p->reachable_time = | |
1271 | neigh_rand_reach_time(p->base_reachable_time); | |
c7fb64db TG |
1272 | if (dev) { |
1273 | if (dev->neigh_setup && dev->neigh_setup(dev, p)) { | |
1274 | kfree(p); | |
1275 | return NULL; | |
1276 | } | |
1277 | ||
1278 | dev_hold(dev); | |
1279 | p->dev = dev; | |
1da177e4 LT |
1280 | } |
1281 | p->sysctl_table = NULL; | |
1282 | write_lock_bh(&tbl->lock); | |
1283 | p->next = tbl->parms.next; | |
1284 | tbl->parms.next = p; | |
1285 | write_unlock_bh(&tbl->lock); | |
1286 | } | |
1287 | return p; | |
1288 | } | |
1289 | ||
1290 | static void neigh_rcu_free_parms(struct rcu_head *head) | |
1291 | { | |
1292 | struct neigh_parms *parms = | |
1293 | container_of(head, struct neigh_parms, rcu_head); | |
1294 | ||
1295 | neigh_parms_put(parms); | |
1296 | } | |
1297 | ||
1298 | void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms) | |
1299 | { | |
1300 | struct neigh_parms **p; | |
1301 | ||
1302 | if (!parms || parms == &tbl->parms) | |
1303 | return; | |
1304 | write_lock_bh(&tbl->lock); | |
1305 | for (p = &tbl->parms.next; *p; p = &(*p)->next) { | |
1306 | if (*p == parms) { | |
1307 | *p = parms->next; | |
1308 | parms->dead = 1; | |
1309 | write_unlock_bh(&tbl->lock); | |
c7fb64db TG |
1310 | if (parms->dev) |
1311 | dev_put(parms->dev); | |
1da177e4 LT |
1312 | call_rcu(&parms->rcu_head, neigh_rcu_free_parms); |
1313 | return; | |
1314 | } | |
1315 | } | |
1316 | write_unlock_bh(&tbl->lock); | |
1317 | NEIGH_PRINTK1("neigh_parms_release: not found\n"); | |
1318 | } | |
1319 | ||
1320 | void neigh_parms_destroy(struct neigh_parms *parms) | |
1321 | { | |
1322 | kfree(parms); | |
1323 | } | |
1324 | ||
1325 | ||
1326 | void neigh_table_init(struct neigh_table *tbl) | |
1327 | { | |
1328 | unsigned long now = jiffies; | |
1329 | unsigned long phsize; | |
1330 | ||
1331 | atomic_set(&tbl->parms.refcnt, 1); | |
1332 | INIT_RCU_HEAD(&tbl->parms.rcu_head); | |
1333 | tbl->parms.reachable_time = | |
1334 | neigh_rand_reach_time(tbl->parms.base_reachable_time); | |
1335 | ||
1336 | if (!tbl->kmem_cachep) | |
1337 | tbl->kmem_cachep = kmem_cache_create(tbl->id, | |
1338 | tbl->entry_size, | |
1339 | 0, SLAB_HWCACHE_ALIGN, | |
1340 | NULL, NULL); | |
1341 | ||
1342 | if (!tbl->kmem_cachep) | |
1343 | panic("cannot create neighbour cache"); | |
1344 | ||
1345 | tbl->stats = alloc_percpu(struct neigh_statistics); | |
1346 | if (!tbl->stats) | |
1347 | panic("cannot create neighbour cache statistics"); | |
1348 | ||
1349 | #ifdef CONFIG_PROC_FS | |
1350 | tbl->pde = create_proc_entry(tbl->id, 0, proc_net_stat); | |
1351 | if (!tbl->pde) | |
1352 | panic("cannot create neighbour proc dir entry"); | |
1353 | tbl->pde->proc_fops = &neigh_stat_seq_fops; | |
1354 | tbl->pde->data = tbl; | |
1355 | #endif | |
1356 | ||
1357 | tbl->hash_mask = 1; | |
1358 | tbl->hash_buckets = neigh_hash_alloc(tbl->hash_mask + 1); | |
1359 | ||
1360 | phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *); | |
1361 | tbl->phash_buckets = kmalloc(phsize, GFP_KERNEL); | |
1362 | ||
1363 | if (!tbl->hash_buckets || !tbl->phash_buckets) | |
1364 | panic("cannot allocate neighbour cache hashes"); | |
1365 | ||
1366 | memset(tbl->phash_buckets, 0, phsize); | |
1367 | ||
1368 | get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd)); | |
1369 | ||
1370 | rwlock_init(&tbl->lock); | |
1371 | init_timer(&tbl->gc_timer); | |
1372 | tbl->gc_timer.data = (unsigned long)tbl; | |
1373 | tbl->gc_timer.function = neigh_periodic_timer; | |
1374 | tbl->gc_timer.expires = now + 1; | |
1375 | add_timer(&tbl->gc_timer); | |
1376 | ||
1377 | init_timer(&tbl->proxy_timer); | |
1378 | tbl->proxy_timer.data = (unsigned long)tbl; | |
1379 | tbl->proxy_timer.function = neigh_proxy_process; | |
1380 | skb_queue_head_init(&tbl->proxy_queue); | |
1381 | ||
1382 | tbl->last_flush = now; | |
1383 | tbl->last_rand = now + tbl->parms.reachable_time * 20; | |
1384 | write_lock(&neigh_tbl_lock); | |
1385 | tbl->next = neigh_tables; | |
1386 | neigh_tables = tbl; | |
1387 | write_unlock(&neigh_tbl_lock); | |
1388 | } | |
1389 | ||
1390 | int neigh_table_clear(struct neigh_table *tbl) | |
1391 | { | |
1392 | struct neigh_table **tp; | |
1393 | ||
1394 | /* It is not clean... Fix it to unload IPv6 module safely */ | |
1395 | del_timer_sync(&tbl->gc_timer); | |
1396 | del_timer_sync(&tbl->proxy_timer); | |
1397 | pneigh_queue_purge(&tbl->proxy_queue); | |
1398 | neigh_ifdown(tbl, NULL); | |
1399 | if (atomic_read(&tbl->entries)) | |
1400 | printk(KERN_CRIT "neighbour leakage\n"); | |
1401 | write_lock(&neigh_tbl_lock); | |
1402 | for (tp = &neigh_tables; *tp; tp = &(*tp)->next) { | |
1403 | if (*tp == tbl) { | |
1404 | *tp = tbl->next; | |
1405 | break; | |
1406 | } | |
1407 | } | |
1408 | write_unlock(&neigh_tbl_lock); | |
1409 | ||
1410 | neigh_hash_free(tbl->hash_buckets, tbl->hash_mask + 1); | |
1411 | tbl->hash_buckets = NULL; | |
1412 | ||
1413 | kfree(tbl->phash_buckets); | |
1414 | tbl->phash_buckets = NULL; | |
1415 | ||
1416 | return 0; | |
1417 | } | |
1418 | ||
1419 | int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1420 | { | |
1421 | struct ndmsg *ndm = NLMSG_DATA(nlh); | |
1422 | struct rtattr **nda = arg; | |
1423 | struct neigh_table *tbl; | |
1424 | struct net_device *dev = NULL; | |
1425 | int err = -ENODEV; | |
1426 | ||
1427 | if (ndm->ndm_ifindex && | |
1428 | (dev = dev_get_by_index(ndm->ndm_ifindex)) == NULL) | |
1429 | goto out; | |
1430 | ||
1431 | read_lock(&neigh_tbl_lock); | |
1432 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1433 | struct rtattr *dst_attr = nda[NDA_DST - 1]; | |
1434 | struct neighbour *n; | |
1435 | ||
1436 | if (tbl->family != ndm->ndm_family) | |
1437 | continue; | |
1438 | read_unlock(&neigh_tbl_lock); | |
1439 | ||
1440 | err = -EINVAL; | |
1441 | if (!dst_attr || RTA_PAYLOAD(dst_attr) < tbl->key_len) | |
1442 | goto out_dev_put; | |
1443 | ||
1444 | if (ndm->ndm_flags & NTF_PROXY) { | |
1445 | err = pneigh_delete(tbl, RTA_DATA(dst_attr), dev); | |
1446 | goto out_dev_put; | |
1447 | } | |
1448 | ||
1449 | if (!dev) | |
1450 | goto out; | |
1451 | ||
1452 | n = neigh_lookup(tbl, RTA_DATA(dst_attr), dev); | |
1453 | if (n) { | |
1454 | err = neigh_update(n, NULL, NUD_FAILED, | |
1455 | NEIGH_UPDATE_F_OVERRIDE| | |
1456 | NEIGH_UPDATE_F_ADMIN); | |
1457 | neigh_release(n); | |
1458 | } | |
1459 | goto out_dev_put; | |
1460 | } | |
1461 | read_unlock(&neigh_tbl_lock); | |
1462 | err = -EADDRNOTAVAIL; | |
1463 | out_dev_put: | |
1464 | if (dev) | |
1465 | dev_put(dev); | |
1466 | out: | |
1467 | return err; | |
1468 | } | |
1469 | ||
1470 | int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1471 | { | |
1472 | struct ndmsg *ndm = NLMSG_DATA(nlh); | |
1473 | struct rtattr **nda = arg; | |
1474 | struct neigh_table *tbl; | |
1475 | struct net_device *dev = NULL; | |
1476 | int err = -ENODEV; | |
1477 | ||
1478 | if (ndm->ndm_ifindex && | |
1479 | (dev = dev_get_by_index(ndm->ndm_ifindex)) == NULL) | |
1480 | goto out; | |
1481 | ||
1482 | read_lock(&neigh_tbl_lock); | |
1483 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1484 | struct rtattr *lladdr_attr = nda[NDA_LLADDR - 1]; | |
1485 | struct rtattr *dst_attr = nda[NDA_DST - 1]; | |
1486 | int override = 1; | |
1487 | struct neighbour *n; | |
1488 | ||
1489 | if (tbl->family != ndm->ndm_family) | |
1490 | continue; | |
1491 | read_unlock(&neigh_tbl_lock); | |
1492 | ||
1493 | err = -EINVAL; | |
1494 | if (!dst_attr || RTA_PAYLOAD(dst_attr) < tbl->key_len) | |
1495 | goto out_dev_put; | |
1496 | ||
1497 | if (ndm->ndm_flags & NTF_PROXY) { | |
1498 | err = -ENOBUFS; | |
1499 | if (pneigh_lookup(tbl, RTA_DATA(dst_attr), dev, 1)) | |
1500 | err = 0; | |
1501 | goto out_dev_put; | |
1502 | } | |
1503 | ||
1504 | err = -EINVAL; | |
1505 | if (!dev) | |
1506 | goto out; | |
1507 | if (lladdr_attr && RTA_PAYLOAD(lladdr_attr) < dev->addr_len) | |
1508 | goto out_dev_put; | |
1509 | ||
1510 | n = neigh_lookup(tbl, RTA_DATA(dst_attr), dev); | |
1511 | if (n) { | |
1512 | if (nlh->nlmsg_flags & NLM_F_EXCL) { | |
1513 | err = -EEXIST; | |
1514 | neigh_release(n); | |
1515 | goto out_dev_put; | |
1516 | } | |
1517 | ||
1518 | override = nlh->nlmsg_flags & NLM_F_REPLACE; | |
1519 | } else if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { | |
1520 | err = -ENOENT; | |
1521 | goto out_dev_put; | |
1522 | } else { | |
1523 | n = __neigh_lookup_errno(tbl, RTA_DATA(dst_attr), dev); | |
1524 | if (IS_ERR(n)) { | |
1525 | err = PTR_ERR(n); | |
1526 | goto out_dev_put; | |
1527 | } | |
1528 | } | |
1529 | ||
1530 | err = neigh_update(n, | |
1531 | lladdr_attr ? RTA_DATA(lladdr_attr) : NULL, | |
1532 | ndm->ndm_state, | |
1533 | (override ? NEIGH_UPDATE_F_OVERRIDE : 0) | | |
1534 | NEIGH_UPDATE_F_ADMIN); | |
1535 | ||
1536 | neigh_release(n); | |
1537 | goto out_dev_put; | |
1538 | } | |
1539 | ||
1540 | read_unlock(&neigh_tbl_lock); | |
1541 | err = -EADDRNOTAVAIL; | |
1542 | out_dev_put: | |
1543 | if (dev) | |
1544 | dev_put(dev); | |
1545 | out: | |
1546 | return err; | |
1547 | } | |
1548 | ||
c7fb64db TG |
1549 | static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms) |
1550 | { | |
e386c6eb TG |
1551 | struct rtattr *nest = NULL; |
1552 | ||
1553 | nest = RTA_NEST(skb, NDTA_PARMS); | |
c7fb64db TG |
1554 | |
1555 | if (parms->dev) | |
1556 | RTA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex); | |
1557 | ||
1558 | RTA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt)); | |
1559 | RTA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len); | |
1560 | RTA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen); | |
1561 | RTA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes); | |
1562 | RTA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes); | |
1563 | RTA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes); | |
1564 | RTA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time); | |
1565 | RTA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME, | |
1566 | parms->base_reachable_time); | |
1567 | RTA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime); | |
1568 | RTA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time); | |
1569 | RTA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time); | |
1570 | RTA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay); | |
1571 | RTA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay); | |
1572 | RTA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime); | |
1573 | ||
1574 | return RTA_NEST_END(skb, nest); | |
1575 | ||
1576 | rtattr_failure: | |
1577 | return RTA_NEST_CANCEL(skb, nest); | |
1578 | } | |
1579 | ||
1580 | static int neightbl_fill_info(struct neigh_table *tbl, struct sk_buff *skb, | |
1581 | struct netlink_callback *cb) | |
1582 | { | |
1583 | struct nlmsghdr *nlh; | |
1584 | struct ndtmsg *ndtmsg; | |
1585 | ||
1797754e TG |
1586 | nlh = NLMSG_NEW_ANSWER(skb, cb, RTM_NEWNEIGHTBL, sizeof(struct ndtmsg), |
1587 | NLM_F_MULTI); | |
c7fb64db | 1588 | |
4b6ea82d | 1589 | ndtmsg = NLMSG_DATA(nlh); |
c7fb64db TG |
1590 | |
1591 | read_lock_bh(&tbl->lock); | |
1592 | ndtmsg->ndtm_family = tbl->family; | |
9ef1d4c7 PM |
1593 | ndtmsg->ndtm_pad1 = 0; |
1594 | ndtmsg->ndtm_pad2 = 0; | |
c7fb64db TG |
1595 | |
1596 | RTA_PUT_STRING(skb, NDTA_NAME, tbl->id); | |
1597 | RTA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval); | |
1598 | RTA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1); | |
1599 | RTA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2); | |
1600 | RTA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3); | |
1601 | ||
1602 | { | |
1603 | unsigned long now = jiffies; | |
1604 | unsigned int flush_delta = now - tbl->last_flush; | |
1605 | unsigned int rand_delta = now - tbl->last_rand; | |
1606 | ||
1607 | struct ndt_config ndc = { | |
1608 | .ndtc_key_len = tbl->key_len, | |
1609 | .ndtc_entry_size = tbl->entry_size, | |
1610 | .ndtc_entries = atomic_read(&tbl->entries), | |
1611 | .ndtc_last_flush = jiffies_to_msecs(flush_delta), | |
1612 | .ndtc_last_rand = jiffies_to_msecs(rand_delta), | |
1613 | .ndtc_hash_rnd = tbl->hash_rnd, | |
1614 | .ndtc_hash_mask = tbl->hash_mask, | |
1615 | .ndtc_hash_chain_gc = tbl->hash_chain_gc, | |
1616 | .ndtc_proxy_qlen = tbl->proxy_queue.qlen, | |
1617 | }; | |
1618 | ||
1619 | RTA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc); | |
1620 | } | |
1621 | ||
1622 | { | |
1623 | int cpu; | |
1624 | struct ndt_stats ndst; | |
1625 | ||
1626 | memset(&ndst, 0, sizeof(ndst)); | |
1627 | ||
670c02c2 | 1628 | for_each_cpu(cpu) { |
c7fb64db TG |
1629 | struct neigh_statistics *st; |
1630 | ||
c7fb64db TG |
1631 | st = per_cpu_ptr(tbl->stats, cpu); |
1632 | ndst.ndts_allocs += st->allocs; | |
1633 | ndst.ndts_destroys += st->destroys; | |
1634 | ndst.ndts_hash_grows += st->hash_grows; | |
1635 | ndst.ndts_res_failed += st->res_failed; | |
1636 | ndst.ndts_lookups += st->lookups; | |
1637 | ndst.ndts_hits += st->hits; | |
1638 | ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast; | |
1639 | ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast; | |
1640 | ndst.ndts_periodic_gc_runs += st->periodic_gc_runs; | |
1641 | ndst.ndts_forced_gc_runs += st->forced_gc_runs; | |
1642 | } | |
1643 | ||
1644 | RTA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst); | |
1645 | } | |
1646 | ||
1647 | BUG_ON(tbl->parms.dev); | |
1648 | if (neightbl_fill_parms(skb, &tbl->parms) < 0) | |
1649 | goto rtattr_failure; | |
1650 | ||
1651 | read_unlock_bh(&tbl->lock); | |
1652 | return NLMSG_END(skb, nlh); | |
1653 | ||
1654 | rtattr_failure: | |
1655 | read_unlock_bh(&tbl->lock); | |
1656 | return NLMSG_CANCEL(skb, nlh); | |
1657 | ||
1658 | nlmsg_failure: | |
1659 | return -1; | |
1660 | } | |
1661 | ||
1662 | static int neightbl_fill_param_info(struct neigh_table *tbl, | |
1663 | struct neigh_parms *parms, | |
1664 | struct sk_buff *skb, | |
1665 | struct netlink_callback *cb) | |
1666 | { | |
1667 | struct ndtmsg *ndtmsg; | |
1668 | struct nlmsghdr *nlh; | |
1669 | ||
1797754e TG |
1670 | nlh = NLMSG_NEW_ANSWER(skb, cb, RTM_NEWNEIGHTBL, sizeof(struct ndtmsg), |
1671 | NLM_F_MULTI); | |
c7fb64db | 1672 | |
4b6ea82d | 1673 | ndtmsg = NLMSG_DATA(nlh); |
c7fb64db TG |
1674 | |
1675 | read_lock_bh(&tbl->lock); | |
1676 | ndtmsg->ndtm_family = tbl->family; | |
9ef1d4c7 PM |
1677 | ndtmsg->ndtm_pad1 = 0; |
1678 | ndtmsg->ndtm_pad2 = 0; | |
c7fb64db TG |
1679 | RTA_PUT_STRING(skb, NDTA_NAME, tbl->id); |
1680 | ||
1681 | if (neightbl_fill_parms(skb, parms) < 0) | |
1682 | goto rtattr_failure; | |
1683 | ||
1684 | read_unlock_bh(&tbl->lock); | |
1685 | return NLMSG_END(skb, nlh); | |
1686 | ||
1687 | rtattr_failure: | |
1688 | read_unlock_bh(&tbl->lock); | |
1689 | return NLMSG_CANCEL(skb, nlh); | |
1690 | ||
1691 | nlmsg_failure: | |
1692 | return -1; | |
1693 | } | |
1694 | ||
1695 | static inline struct neigh_parms *lookup_neigh_params(struct neigh_table *tbl, | |
1696 | int ifindex) | |
1697 | { | |
1698 | struct neigh_parms *p; | |
1699 | ||
1700 | for (p = &tbl->parms; p; p = p->next) | |
1701 | if ((p->dev && p->dev->ifindex == ifindex) || | |
1702 | (!p->dev && !ifindex)) | |
1703 | return p; | |
1704 | ||
1705 | return NULL; | |
1706 | } | |
1707 | ||
1708 | int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1709 | { | |
1710 | struct neigh_table *tbl; | |
1711 | struct ndtmsg *ndtmsg = NLMSG_DATA(nlh); | |
1712 | struct rtattr **tb = arg; | |
1713 | int err = -EINVAL; | |
1714 | ||
1715 | if (!tb[NDTA_NAME - 1] || !RTA_PAYLOAD(tb[NDTA_NAME - 1])) | |
1716 | return -EINVAL; | |
1717 | ||
1718 | read_lock(&neigh_tbl_lock); | |
1719 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1720 | if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family) | |
1721 | continue; | |
1722 | ||
1723 | if (!rtattr_strcmp(tb[NDTA_NAME - 1], tbl->id)) | |
1724 | break; | |
1725 | } | |
1726 | ||
1727 | if (tbl == NULL) { | |
1728 | err = -ENOENT; | |
1729 | goto errout; | |
1730 | } | |
1731 | ||
1732 | /* | |
1733 | * We acquire tbl->lock to be nice to the periodic timers and | |
1734 | * make sure they always see a consistent set of values. | |
1735 | */ | |
1736 | write_lock_bh(&tbl->lock); | |
1737 | ||
1738 | if (tb[NDTA_THRESH1 - 1]) | |
1739 | tbl->gc_thresh1 = RTA_GET_U32(tb[NDTA_THRESH1 - 1]); | |
1740 | ||
1741 | if (tb[NDTA_THRESH2 - 1]) | |
1742 | tbl->gc_thresh2 = RTA_GET_U32(tb[NDTA_THRESH2 - 1]); | |
1743 | ||
1744 | if (tb[NDTA_THRESH3 - 1]) | |
1745 | tbl->gc_thresh3 = RTA_GET_U32(tb[NDTA_THRESH3 - 1]); | |
1746 | ||
1747 | if (tb[NDTA_GC_INTERVAL - 1]) | |
1748 | tbl->gc_interval = RTA_GET_MSECS(tb[NDTA_GC_INTERVAL - 1]); | |
1749 | ||
1750 | if (tb[NDTA_PARMS - 1]) { | |
1751 | struct rtattr *tbp[NDTPA_MAX]; | |
1752 | struct neigh_parms *p; | |
1753 | u32 ifindex = 0; | |
1754 | ||
1755 | if (rtattr_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS - 1]) < 0) | |
1756 | goto rtattr_failure; | |
1757 | ||
1758 | if (tbp[NDTPA_IFINDEX - 1]) | |
1759 | ifindex = RTA_GET_U32(tbp[NDTPA_IFINDEX - 1]); | |
1760 | ||
1761 | p = lookup_neigh_params(tbl, ifindex); | |
1762 | if (p == NULL) { | |
1763 | err = -ENOENT; | |
1764 | goto rtattr_failure; | |
1765 | } | |
1766 | ||
1767 | if (tbp[NDTPA_QUEUE_LEN - 1]) | |
1768 | p->queue_len = RTA_GET_U32(tbp[NDTPA_QUEUE_LEN - 1]); | |
1769 | ||
1770 | if (tbp[NDTPA_PROXY_QLEN - 1]) | |
1771 | p->proxy_qlen = RTA_GET_U32(tbp[NDTPA_PROXY_QLEN - 1]); | |
1772 | ||
1773 | if (tbp[NDTPA_APP_PROBES - 1]) | |
1774 | p->app_probes = RTA_GET_U32(tbp[NDTPA_APP_PROBES - 1]); | |
1775 | ||
1776 | if (tbp[NDTPA_UCAST_PROBES - 1]) | |
1777 | p->ucast_probes = | |
1778 | RTA_GET_U32(tbp[NDTPA_UCAST_PROBES - 1]); | |
1779 | ||
1780 | if (tbp[NDTPA_MCAST_PROBES - 1]) | |
1781 | p->mcast_probes = | |
1782 | RTA_GET_U32(tbp[NDTPA_MCAST_PROBES - 1]); | |
1783 | ||
1784 | if (tbp[NDTPA_BASE_REACHABLE_TIME - 1]) | |
1785 | p->base_reachable_time = | |
1786 | RTA_GET_MSECS(tbp[NDTPA_BASE_REACHABLE_TIME - 1]); | |
1787 | ||
1788 | if (tbp[NDTPA_GC_STALETIME - 1]) | |
1789 | p->gc_staletime = | |
1790 | RTA_GET_MSECS(tbp[NDTPA_GC_STALETIME - 1]); | |
1791 | ||
1792 | if (tbp[NDTPA_DELAY_PROBE_TIME - 1]) | |
1793 | p->delay_probe_time = | |
1794 | RTA_GET_MSECS(tbp[NDTPA_DELAY_PROBE_TIME - 1]); | |
1795 | ||
1796 | if (tbp[NDTPA_RETRANS_TIME - 1]) | |
1797 | p->retrans_time = | |
1798 | RTA_GET_MSECS(tbp[NDTPA_RETRANS_TIME - 1]); | |
1799 | ||
1800 | if (tbp[NDTPA_ANYCAST_DELAY - 1]) | |
1801 | p->anycast_delay = | |
1802 | RTA_GET_MSECS(tbp[NDTPA_ANYCAST_DELAY - 1]); | |
1803 | ||
1804 | if (tbp[NDTPA_PROXY_DELAY - 1]) | |
1805 | p->proxy_delay = | |
1806 | RTA_GET_MSECS(tbp[NDTPA_PROXY_DELAY - 1]); | |
1807 | ||
1808 | if (tbp[NDTPA_LOCKTIME - 1]) | |
1809 | p->locktime = RTA_GET_MSECS(tbp[NDTPA_LOCKTIME - 1]); | |
1810 | } | |
1811 | ||
1812 | err = 0; | |
1813 | ||
1814 | rtattr_failure: | |
1815 | write_unlock_bh(&tbl->lock); | |
1816 | errout: | |
1817 | read_unlock(&neigh_tbl_lock); | |
1818 | return err; | |
1819 | } | |
1820 | ||
1821 | int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb) | |
1822 | { | |
1823 | int idx, family; | |
1824 | int s_idx = cb->args[0]; | |
1825 | struct neigh_table *tbl; | |
1826 | ||
1827 | family = ((struct rtgenmsg *)NLMSG_DATA(cb->nlh))->rtgen_family; | |
1828 | ||
1829 | read_lock(&neigh_tbl_lock); | |
1830 | for (tbl = neigh_tables, idx = 0; tbl; tbl = tbl->next) { | |
1831 | struct neigh_parms *p; | |
1832 | ||
1833 | if (idx < s_idx || (family && tbl->family != family)) | |
1834 | continue; | |
1835 | ||
1836 | if (neightbl_fill_info(tbl, skb, cb) <= 0) | |
1837 | break; | |
1838 | ||
1839 | for (++idx, p = tbl->parms.next; p; p = p->next, idx++) { | |
1840 | if (idx < s_idx) | |
1841 | continue; | |
1842 | ||
1843 | if (neightbl_fill_param_info(tbl, p, skb, cb) <= 0) | |
1844 | goto out; | |
1845 | } | |
1846 | ||
1847 | } | |
1848 | out: | |
1849 | read_unlock(&neigh_tbl_lock); | |
1850 | cb->args[0] = idx; | |
1851 | ||
1852 | return skb->len; | |
1853 | } | |
1da177e4 LT |
1854 | |
1855 | static int neigh_fill_info(struct sk_buff *skb, struct neighbour *n, | |
b6544c0b | 1856 | u32 pid, u32 seq, int event, unsigned int flags) |
1da177e4 LT |
1857 | { |
1858 | unsigned long now = jiffies; | |
1859 | unsigned char *b = skb->tail; | |
1860 | struct nda_cacheinfo ci; | |
1861 | int locked = 0; | |
1862 | u32 probes; | |
b6544c0b JHS |
1863 | struct nlmsghdr *nlh = NLMSG_NEW(skb, pid, seq, event, |
1864 | sizeof(struct ndmsg), flags); | |
1da177e4 LT |
1865 | struct ndmsg *ndm = NLMSG_DATA(nlh); |
1866 | ||
1da177e4 | 1867 | ndm->ndm_family = n->ops->family; |
9ef1d4c7 PM |
1868 | ndm->ndm_pad1 = 0; |
1869 | ndm->ndm_pad2 = 0; | |
1da177e4 LT |
1870 | ndm->ndm_flags = n->flags; |
1871 | ndm->ndm_type = n->type; | |
1872 | ndm->ndm_ifindex = n->dev->ifindex; | |
1873 | RTA_PUT(skb, NDA_DST, n->tbl->key_len, n->primary_key); | |
1874 | read_lock_bh(&n->lock); | |
1875 | locked = 1; | |
1876 | ndm->ndm_state = n->nud_state; | |
1877 | if (n->nud_state & NUD_VALID) | |
1878 | RTA_PUT(skb, NDA_LLADDR, n->dev->addr_len, n->ha); | |
1879 | ci.ndm_used = now - n->used; | |
1880 | ci.ndm_confirmed = now - n->confirmed; | |
1881 | ci.ndm_updated = now - n->updated; | |
1882 | ci.ndm_refcnt = atomic_read(&n->refcnt) - 1; | |
1883 | probes = atomic_read(&n->probes); | |
1884 | read_unlock_bh(&n->lock); | |
1885 | locked = 0; | |
1886 | RTA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci); | |
1887 | RTA_PUT(skb, NDA_PROBES, sizeof(probes), &probes); | |
1888 | nlh->nlmsg_len = skb->tail - b; | |
1889 | return skb->len; | |
1890 | ||
1891 | nlmsg_failure: | |
1892 | rtattr_failure: | |
1893 | if (locked) | |
1894 | read_unlock_bh(&n->lock); | |
1895 | skb_trim(skb, b - skb->data); | |
1896 | return -1; | |
1897 | } | |
1898 | ||
1899 | ||
1900 | static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb, | |
1901 | struct netlink_callback *cb) | |
1902 | { | |
1903 | struct neighbour *n; | |
1904 | int rc, h, s_h = cb->args[1]; | |
1905 | int idx, s_idx = idx = cb->args[2]; | |
1906 | ||
1907 | for (h = 0; h <= tbl->hash_mask; h++) { | |
1908 | if (h < s_h) | |
1909 | continue; | |
1910 | if (h > s_h) | |
1911 | s_idx = 0; | |
1912 | read_lock_bh(&tbl->lock); | |
1913 | for (n = tbl->hash_buckets[h], idx = 0; n; n = n->next, idx++) { | |
1914 | if (idx < s_idx) | |
1915 | continue; | |
1916 | if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid, | |
1917 | cb->nlh->nlmsg_seq, | |
b6544c0b JHS |
1918 | RTM_NEWNEIGH, |
1919 | NLM_F_MULTI) <= 0) { | |
1da177e4 LT |
1920 | read_unlock_bh(&tbl->lock); |
1921 | rc = -1; | |
1922 | goto out; | |
1923 | } | |
1924 | } | |
1925 | read_unlock_bh(&tbl->lock); | |
1926 | } | |
1927 | rc = skb->len; | |
1928 | out: | |
1929 | cb->args[1] = h; | |
1930 | cb->args[2] = idx; | |
1931 | return rc; | |
1932 | } | |
1933 | ||
1934 | int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb) | |
1935 | { | |
1936 | struct neigh_table *tbl; | |
1937 | int t, family, s_t; | |
1938 | ||
1939 | read_lock(&neigh_tbl_lock); | |
1940 | family = ((struct rtgenmsg *)NLMSG_DATA(cb->nlh))->rtgen_family; | |
1941 | s_t = cb->args[0]; | |
1942 | ||
1943 | for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) { | |
1944 | if (t < s_t || (family && tbl->family != family)) | |
1945 | continue; | |
1946 | if (t > s_t) | |
1947 | memset(&cb->args[1], 0, sizeof(cb->args) - | |
1948 | sizeof(cb->args[0])); | |
1949 | if (neigh_dump_table(tbl, skb, cb) < 0) | |
1950 | break; | |
1951 | } | |
1952 | read_unlock(&neigh_tbl_lock); | |
1953 | ||
1954 | cb->args[0] = t; | |
1955 | return skb->len; | |
1956 | } | |
1957 | ||
1958 | void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie) | |
1959 | { | |
1960 | int chain; | |
1961 | ||
1962 | read_lock_bh(&tbl->lock); | |
1963 | for (chain = 0; chain <= tbl->hash_mask; chain++) { | |
1964 | struct neighbour *n; | |
1965 | ||
1966 | for (n = tbl->hash_buckets[chain]; n; n = n->next) | |
1967 | cb(n, cookie); | |
1968 | } | |
1969 | read_unlock_bh(&tbl->lock); | |
1970 | } | |
1971 | EXPORT_SYMBOL(neigh_for_each); | |
1972 | ||
1973 | /* The tbl->lock must be held as a writer and BH disabled. */ | |
1974 | void __neigh_for_each_release(struct neigh_table *tbl, | |
1975 | int (*cb)(struct neighbour *)) | |
1976 | { | |
1977 | int chain; | |
1978 | ||
1979 | for (chain = 0; chain <= tbl->hash_mask; chain++) { | |
1980 | struct neighbour *n, **np; | |
1981 | ||
1982 | np = &tbl->hash_buckets[chain]; | |
1983 | while ((n = *np) != NULL) { | |
1984 | int release; | |
1985 | ||
1986 | write_lock(&n->lock); | |
1987 | release = cb(n); | |
1988 | if (release) { | |
1989 | *np = n->next; | |
1990 | n->dead = 1; | |
1991 | } else | |
1992 | np = &n->next; | |
1993 | write_unlock(&n->lock); | |
1994 | if (release) | |
1995 | neigh_release(n); | |
1996 | } | |
1997 | } | |
1998 | } | |
1999 | EXPORT_SYMBOL(__neigh_for_each_release); | |
2000 | ||
2001 | #ifdef CONFIG_PROC_FS | |
2002 | ||
2003 | static struct neighbour *neigh_get_first(struct seq_file *seq) | |
2004 | { | |
2005 | struct neigh_seq_state *state = seq->private; | |
2006 | struct neigh_table *tbl = state->tbl; | |
2007 | struct neighbour *n = NULL; | |
2008 | int bucket = state->bucket; | |
2009 | ||
2010 | state->flags &= ~NEIGH_SEQ_IS_PNEIGH; | |
2011 | for (bucket = 0; bucket <= tbl->hash_mask; bucket++) { | |
2012 | n = tbl->hash_buckets[bucket]; | |
2013 | ||
2014 | while (n) { | |
2015 | if (state->neigh_sub_iter) { | |
2016 | loff_t fakep = 0; | |
2017 | void *v; | |
2018 | ||
2019 | v = state->neigh_sub_iter(state, n, &fakep); | |
2020 | if (!v) | |
2021 | goto next; | |
2022 | } | |
2023 | if (!(state->flags & NEIGH_SEQ_SKIP_NOARP)) | |
2024 | break; | |
2025 | if (n->nud_state & ~NUD_NOARP) | |
2026 | break; | |
2027 | next: | |
2028 | n = n->next; | |
2029 | } | |
2030 | ||
2031 | if (n) | |
2032 | break; | |
2033 | } | |
2034 | state->bucket = bucket; | |
2035 | ||
2036 | return n; | |
2037 | } | |
2038 | ||
2039 | static struct neighbour *neigh_get_next(struct seq_file *seq, | |
2040 | struct neighbour *n, | |
2041 | loff_t *pos) | |
2042 | { | |
2043 | struct neigh_seq_state *state = seq->private; | |
2044 | struct neigh_table *tbl = state->tbl; | |
2045 | ||
2046 | if (state->neigh_sub_iter) { | |
2047 | void *v = state->neigh_sub_iter(state, n, pos); | |
2048 | if (v) | |
2049 | return n; | |
2050 | } | |
2051 | n = n->next; | |
2052 | ||
2053 | while (1) { | |
2054 | while (n) { | |
2055 | if (state->neigh_sub_iter) { | |
2056 | void *v = state->neigh_sub_iter(state, n, pos); | |
2057 | if (v) | |
2058 | return n; | |
2059 | goto next; | |
2060 | } | |
2061 | if (!(state->flags & NEIGH_SEQ_SKIP_NOARP)) | |
2062 | break; | |
2063 | ||
2064 | if (n->nud_state & ~NUD_NOARP) | |
2065 | break; | |
2066 | next: | |
2067 | n = n->next; | |
2068 | } | |
2069 | ||
2070 | if (n) | |
2071 | break; | |
2072 | ||
2073 | if (++state->bucket > tbl->hash_mask) | |
2074 | break; | |
2075 | ||
2076 | n = tbl->hash_buckets[state->bucket]; | |
2077 | } | |
2078 | ||
2079 | if (n && pos) | |
2080 | --(*pos); | |
2081 | return n; | |
2082 | } | |
2083 | ||
2084 | static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos) | |
2085 | { | |
2086 | struct neighbour *n = neigh_get_first(seq); | |
2087 | ||
2088 | if (n) { | |
2089 | while (*pos) { | |
2090 | n = neigh_get_next(seq, n, pos); | |
2091 | if (!n) | |
2092 | break; | |
2093 | } | |
2094 | } | |
2095 | return *pos ? NULL : n; | |
2096 | } | |
2097 | ||
2098 | static struct pneigh_entry *pneigh_get_first(struct seq_file *seq) | |
2099 | { | |
2100 | struct neigh_seq_state *state = seq->private; | |
2101 | struct neigh_table *tbl = state->tbl; | |
2102 | struct pneigh_entry *pn = NULL; | |
2103 | int bucket = state->bucket; | |
2104 | ||
2105 | state->flags |= NEIGH_SEQ_IS_PNEIGH; | |
2106 | for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) { | |
2107 | pn = tbl->phash_buckets[bucket]; | |
2108 | if (pn) | |
2109 | break; | |
2110 | } | |
2111 | state->bucket = bucket; | |
2112 | ||
2113 | return pn; | |
2114 | } | |
2115 | ||
2116 | static struct pneigh_entry *pneigh_get_next(struct seq_file *seq, | |
2117 | struct pneigh_entry *pn, | |
2118 | loff_t *pos) | |
2119 | { | |
2120 | struct neigh_seq_state *state = seq->private; | |
2121 | struct neigh_table *tbl = state->tbl; | |
2122 | ||
2123 | pn = pn->next; | |
2124 | while (!pn) { | |
2125 | if (++state->bucket > PNEIGH_HASHMASK) | |
2126 | break; | |
2127 | pn = tbl->phash_buckets[state->bucket]; | |
2128 | if (pn) | |
2129 | break; | |
2130 | } | |
2131 | ||
2132 | if (pn && pos) | |
2133 | --(*pos); | |
2134 | ||
2135 | return pn; | |
2136 | } | |
2137 | ||
2138 | static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos) | |
2139 | { | |
2140 | struct pneigh_entry *pn = pneigh_get_first(seq); | |
2141 | ||
2142 | if (pn) { | |
2143 | while (*pos) { | |
2144 | pn = pneigh_get_next(seq, pn, pos); | |
2145 | if (!pn) | |
2146 | break; | |
2147 | } | |
2148 | } | |
2149 | return *pos ? NULL : pn; | |
2150 | } | |
2151 | ||
2152 | static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos) | |
2153 | { | |
2154 | struct neigh_seq_state *state = seq->private; | |
2155 | void *rc; | |
2156 | ||
2157 | rc = neigh_get_idx(seq, pos); | |
2158 | if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY)) | |
2159 | rc = pneigh_get_idx(seq, pos); | |
2160 | ||
2161 | return rc; | |
2162 | } | |
2163 | ||
2164 | void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags) | |
2165 | { | |
2166 | struct neigh_seq_state *state = seq->private; | |
2167 | loff_t pos_minus_one; | |
2168 | ||
2169 | state->tbl = tbl; | |
2170 | state->bucket = 0; | |
2171 | state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH); | |
2172 | ||
2173 | read_lock_bh(&tbl->lock); | |
2174 | ||
2175 | pos_minus_one = *pos - 1; | |
2176 | return *pos ? neigh_get_idx_any(seq, &pos_minus_one) : SEQ_START_TOKEN; | |
2177 | } | |
2178 | EXPORT_SYMBOL(neigh_seq_start); | |
2179 | ||
2180 | void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2181 | { | |
2182 | struct neigh_seq_state *state; | |
2183 | void *rc; | |
2184 | ||
2185 | if (v == SEQ_START_TOKEN) { | |
2186 | rc = neigh_get_idx(seq, pos); | |
2187 | goto out; | |
2188 | } | |
2189 | ||
2190 | state = seq->private; | |
2191 | if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) { | |
2192 | rc = neigh_get_next(seq, v, NULL); | |
2193 | if (rc) | |
2194 | goto out; | |
2195 | if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY)) | |
2196 | rc = pneigh_get_first(seq); | |
2197 | } else { | |
2198 | BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY); | |
2199 | rc = pneigh_get_next(seq, v, NULL); | |
2200 | } | |
2201 | out: | |
2202 | ++(*pos); | |
2203 | return rc; | |
2204 | } | |
2205 | EXPORT_SYMBOL(neigh_seq_next); | |
2206 | ||
2207 | void neigh_seq_stop(struct seq_file *seq, void *v) | |
2208 | { | |
2209 | struct neigh_seq_state *state = seq->private; | |
2210 | struct neigh_table *tbl = state->tbl; | |
2211 | ||
2212 | read_unlock_bh(&tbl->lock); | |
2213 | } | |
2214 | EXPORT_SYMBOL(neigh_seq_stop); | |
2215 | ||
2216 | /* statistics via seq_file */ | |
2217 | ||
2218 | static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos) | |
2219 | { | |
2220 | struct proc_dir_entry *pde = seq->private; | |
2221 | struct neigh_table *tbl = pde->data; | |
2222 | int cpu; | |
2223 | ||
2224 | if (*pos == 0) | |
2225 | return SEQ_START_TOKEN; | |
2226 | ||
2227 | for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) { | |
2228 | if (!cpu_possible(cpu)) | |
2229 | continue; | |
2230 | *pos = cpu+1; | |
2231 | return per_cpu_ptr(tbl->stats, cpu); | |
2232 | } | |
2233 | return NULL; | |
2234 | } | |
2235 | ||
2236 | static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2237 | { | |
2238 | struct proc_dir_entry *pde = seq->private; | |
2239 | struct neigh_table *tbl = pde->data; | |
2240 | int cpu; | |
2241 | ||
2242 | for (cpu = *pos; cpu < NR_CPUS; ++cpu) { | |
2243 | if (!cpu_possible(cpu)) | |
2244 | continue; | |
2245 | *pos = cpu+1; | |
2246 | return per_cpu_ptr(tbl->stats, cpu); | |
2247 | } | |
2248 | return NULL; | |
2249 | } | |
2250 | ||
2251 | static void neigh_stat_seq_stop(struct seq_file *seq, void *v) | |
2252 | { | |
2253 | ||
2254 | } | |
2255 | ||
2256 | static int neigh_stat_seq_show(struct seq_file *seq, void *v) | |
2257 | { | |
2258 | struct proc_dir_entry *pde = seq->private; | |
2259 | struct neigh_table *tbl = pde->data; | |
2260 | struct neigh_statistics *st = v; | |
2261 | ||
2262 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 2263 | 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 |
2264 | return 0; |
2265 | } | |
2266 | ||
2267 | seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx " | |
2268 | "%08lx %08lx %08lx %08lx\n", | |
2269 | atomic_read(&tbl->entries), | |
2270 | ||
2271 | st->allocs, | |
2272 | st->destroys, | |
2273 | st->hash_grows, | |
2274 | ||
2275 | st->lookups, | |
2276 | st->hits, | |
2277 | ||
2278 | st->res_failed, | |
2279 | ||
2280 | st->rcv_probes_mcast, | |
2281 | st->rcv_probes_ucast, | |
2282 | ||
2283 | st->periodic_gc_runs, | |
2284 | st->forced_gc_runs | |
2285 | ); | |
2286 | ||
2287 | return 0; | |
2288 | } | |
2289 | ||
2290 | static struct seq_operations neigh_stat_seq_ops = { | |
2291 | .start = neigh_stat_seq_start, | |
2292 | .next = neigh_stat_seq_next, | |
2293 | .stop = neigh_stat_seq_stop, | |
2294 | .show = neigh_stat_seq_show, | |
2295 | }; | |
2296 | ||
2297 | static int neigh_stat_seq_open(struct inode *inode, struct file *file) | |
2298 | { | |
2299 | int ret = seq_open(file, &neigh_stat_seq_ops); | |
2300 | ||
2301 | if (!ret) { | |
2302 | struct seq_file *sf = file->private_data; | |
2303 | sf->private = PDE(inode); | |
2304 | } | |
2305 | return ret; | |
2306 | }; | |
2307 | ||
2308 | static struct file_operations neigh_stat_seq_fops = { | |
2309 | .owner = THIS_MODULE, | |
2310 | .open = neigh_stat_seq_open, | |
2311 | .read = seq_read, | |
2312 | .llseek = seq_lseek, | |
2313 | .release = seq_release, | |
2314 | }; | |
2315 | ||
2316 | #endif /* CONFIG_PROC_FS */ | |
2317 | ||
2318 | #ifdef CONFIG_ARPD | |
2319 | void neigh_app_ns(struct neighbour *n) | |
2320 | { | |
2321 | struct nlmsghdr *nlh; | |
2322 | int size = NLMSG_SPACE(sizeof(struct ndmsg) + 256); | |
2323 | struct sk_buff *skb = alloc_skb(size, GFP_ATOMIC); | |
2324 | ||
2325 | if (!skb) | |
2326 | return; | |
2327 | ||
b6544c0b | 2328 | if (neigh_fill_info(skb, n, 0, 0, RTM_GETNEIGH, 0) < 0) { |
1da177e4 LT |
2329 | kfree_skb(skb); |
2330 | return; | |
2331 | } | |
2332 | nlh = (struct nlmsghdr *)skb->data; | |
2333 | nlh->nlmsg_flags = NLM_F_REQUEST; | |
ac6d439d PM |
2334 | NETLINK_CB(skb).dst_group = RTNLGRP_NEIGH; |
2335 | netlink_broadcast(rtnl, skb, 0, RTNLGRP_NEIGH, GFP_ATOMIC); | |
1da177e4 LT |
2336 | } |
2337 | ||
2338 | static void neigh_app_notify(struct neighbour *n) | |
2339 | { | |
2340 | struct nlmsghdr *nlh; | |
2341 | int size = NLMSG_SPACE(sizeof(struct ndmsg) + 256); | |
2342 | struct sk_buff *skb = alloc_skb(size, GFP_ATOMIC); | |
2343 | ||
2344 | if (!skb) | |
2345 | return; | |
2346 | ||
b6544c0b | 2347 | if (neigh_fill_info(skb, n, 0, 0, RTM_NEWNEIGH, 0) < 0) { |
1da177e4 LT |
2348 | kfree_skb(skb); |
2349 | return; | |
2350 | } | |
2351 | nlh = (struct nlmsghdr *)skb->data; | |
ac6d439d PM |
2352 | NETLINK_CB(skb).dst_group = RTNLGRP_NEIGH; |
2353 | netlink_broadcast(rtnl, skb, 0, RTNLGRP_NEIGH, GFP_ATOMIC); | |
1da177e4 LT |
2354 | } |
2355 | ||
2356 | #endif /* CONFIG_ARPD */ | |
2357 | ||
2358 | #ifdef CONFIG_SYSCTL | |
2359 | ||
2360 | static struct neigh_sysctl_table { | |
2361 | struct ctl_table_header *sysctl_header; | |
2362 | ctl_table neigh_vars[__NET_NEIGH_MAX]; | |
2363 | ctl_table neigh_dev[2]; | |
2364 | ctl_table neigh_neigh_dir[2]; | |
2365 | ctl_table neigh_proto_dir[2]; | |
2366 | ctl_table neigh_root_dir[2]; | |
2367 | } neigh_sysctl_template = { | |
2368 | .neigh_vars = { | |
2369 | { | |
2370 | .ctl_name = NET_NEIGH_MCAST_SOLICIT, | |
2371 | .procname = "mcast_solicit", | |
2372 | .maxlen = sizeof(int), | |
2373 | .mode = 0644, | |
2374 | .proc_handler = &proc_dointvec, | |
2375 | }, | |
2376 | { | |
2377 | .ctl_name = NET_NEIGH_UCAST_SOLICIT, | |
2378 | .procname = "ucast_solicit", | |
2379 | .maxlen = sizeof(int), | |
2380 | .mode = 0644, | |
2381 | .proc_handler = &proc_dointvec, | |
2382 | }, | |
2383 | { | |
2384 | .ctl_name = NET_NEIGH_APP_SOLICIT, | |
2385 | .procname = "app_solicit", | |
2386 | .maxlen = sizeof(int), | |
2387 | .mode = 0644, | |
2388 | .proc_handler = &proc_dointvec, | |
2389 | }, | |
2390 | { | |
2391 | .ctl_name = NET_NEIGH_RETRANS_TIME, | |
2392 | .procname = "retrans_time", | |
2393 | .maxlen = sizeof(int), | |
2394 | .mode = 0644, | |
2395 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2396 | }, | |
2397 | { | |
2398 | .ctl_name = NET_NEIGH_REACHABLE_TIME, | |
2399 | .procname = "base_reachable_time", | |
2400 | .maxlen = sizeof(int), | |
2401 | .mode = 0644, | |
2402 | .proc_handler = &proc_dointvec_jiffies, | |
2403 | .strategy = &sysctl_jiffies, | |
2404 | }, | |
2405 | { | |
2406 | .ctl_name = NET_NEIGH_DELAY_PROBE_TIME, | |
2407 | .procname = "delay_first_probe_time", | |
2408 | .maxlen = sizeof(int), | |
2409 | .mode = 0644, | |
2410 | .proc_handler = &proc_dointvec_jiffies, | |
2411 | .strategy = &sysctl_jiffies, | |
2412 | }, | |
2413 | { | |
2414 | .ctl_name = NET_NEIGH_GC_STALE_TIME, | |
2415 | .procname = "gc_stale_time", | |
2416 | .maxlen = sizeof(int), | |
2417 | .mode = 0644, | |
2418 | .proc_handler = &proc_dointvec_jiffies, | |
2419 | .strategy = &sysctl_jiffies, | |
2420 | }, | |
2421 | { | |
2422 | .ctl_name = NET_NEIGH_UNRES_QLEN, | |
2423 | .procname = "unres_qlen", | |
2424 | .maxlen = sizeof(int), | |
2425 | .mode = 0644, | |
2426 | .proc_handler = &proc_dointvec, | |
2427 | }, | |
2428 | { | |
2429 | .ctl_name = NET_NEIGH_PROXY_QLEN, | |
2430 | .procname = "proxy_qlen", | |
2431 | .maxlen = sizeof(int), | |
2432 | .mode = 0644, | |
2433 | .proc_handler = &proc_dointvec, | |
2434 | }, | |
2435 | { | |
2436 | .ctl_name = NET_NEIGH_ANYCAST_DELAY, | |
2437 | .procname = "anycast_delay", | |
2438 | .maxlen = sizeof(int), | |
2439 | .mode = 0644, | |
2440 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2441 | }, | |
2442 | { | |
2443 | .ctl_name = NET_NEIGH_PROXY_DELAY, | |
2444 | .procname = "proxy_delay", | |
2445 | .maxlen = sizeof(int), | |
2446 | .mode = 0644, | |
2447 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2448 | }, | |
2449 | { | |
2450 | .ctl_name = NET_NEIGH_LOCKTIME, | |
2451 | .procname = "locktime", | |
2452 | .maxlen = sizeof(int), | |
2453 | .mode = 0644, | |
2454 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2455 | }, | |
2456 | { | |
2457 | .ctl_name = NET_NEIGH_GC_INTERVAL, | |
2458 | .procname = "gc_interval", | |
2459 | .maxlen = sizeof(int), | |
2460 | .mode = 0644, | |
2461 | .proc_handler = &proc_dointvec_jiffies, | |
2462 | .strategy = &sysctl_jiffies, | |
2463 | }, | |
2464 | { | |
2465 | .ctl_name = NET_NEIGH_GC_THRESH1, | |
2466 | .procname = "gc_thresh1", | |
2467 | .maxlen = sizeof(int), | |
2468 | .mode = 0644, | |
2469 | .proc_handler = &proc_dointvec, | |
2470 | }, | |
2471 | { | |
2472 | .ctl_name = NET_NEIGH_GC_THRESH2, | |
2473 | .procname = "gc_thresh2", | |
2474 | .maxlen = sizeof(int), | |
2475 | .mode = 0644, | |
2476 | .proc_handler = &proc_dointvec, | |
2477 | }, | |
2478 | { | |
2479 | .ctl_name = NET_NEIGH_GC_THRESH3, | |
2480 | .procname = "gc_thresh3", | |
2481 | .maxlen = sizeof(int), | |
2482 | .mode = 0644, | |
2483 | .proc_handler = &proc_dointvec, | |
2484 | }, | |
2485 | { | |
2486 | .ctl_name = NET_NEIGH_RETRANS_TIME_MS, | |
2487 | .procname = "retrans_time_ms", | |
2488 | .maxlen = sizeof(int), | |
2489 | .mode = 0644, | |
2490 | .proc_handler = &proc_dointvec_ms_jiffies, | |
2491 | .strategy = &sysctl_ms_jiffies, | |
2492 | }, | |
2493 | { | |
2494 | .ctl_name = NET_NEIGH_REACHABLE_TIME_MS, | |
2495 | .procname = "base_reachable_time_ms", | |
2496 | .maxlen = sizeof(int), | |
2497 | .mode = 0644, | |
2498 | .proc_handler = &proc_dointvec_ms_jiffies, | |
2499 | .strategy = &sysctl_ms_jiffies, | |
2500 | }, | |
2501 | }, | |
2502 | .neigh_dev = { | |
2503 | { | |
2504 | .ctl_name = NET_PROTO_CONF_DEFAULT, | |
2505 | .procname = "default", | |
2506 | .mode = 0555, | |
2507 | }, | |
2508 | }, | |
2509 | .neigh_neigh_dir = { | |
2510 | { | |
2511 | .procname = "neigh", | |
2512 | .mode = 0555, | |
2513 | }, | |
2514 | }, | |
2515 | .neigh_proto_dir = { | |
2516 | { | |
2517 | .mode = 0555, | |
2518 | }, | |
2519 | }, | |
2520 | .neigh_root_dir = { | |
2521 | { | |
2522 | .ctl_name = CTL_NET, | |
2523 | .procname = "net", | |
2524 | .mode = 0555, | |
2525 | }, | |
2526 | }, | |
2527 | }; | |
2528 | ||
2529 | int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p, | |
2530 | int p_id, int pdev_id, char *p_name, | |
2531 | proc_handler *handler, ctl_handler *strategy) | |
2532 | { | |
2533 | struct neigh_sysctl_table *t = kmalloc(sizeof(*t), GFP_KERNEL); | |
2534 | const char *dev_name_source = NULL; | |
2535 | char *dev_name = NULL; | |
2536 | int err = 0; | |
2537 | ||
2538 | if (!t) | |
2539 | return -ENOBUFS; | |
2540 | memcpy(t, &neigh_sysctl_template, sizeof(*t)); | |
2541 | t->neigh_vars[0].data = &p->mcast_probes; | |
2542 | t->neigh_vars[1].data = &p->ucast_probes; | |
2543 | t->neigh_vars[2].data = &p->app_probes; | |
2544 | t->neigh_vars[3].data = &p->retrans_time; | |
2545 | t->neigh_vars[4].data = &p->base_reachable_time; | |
2546 | t->neigh_vars[5].data = &p->delay_probe_time; | |
2547 | t->neigh_vars[6].data = &p->gc_staletime; | |
2548 | t->neigh_vars[7].data = &p->queue_len; | |
2549 | t->neigh_vars[8].data = &p->proxy_qlen; | |
2550 | t->neigh_vars[9].data = &p->anycast_delay; | |
2551 | t->neigh_vars[10].data = &p->proxy_delay; | |
2552 | t->neigh_vars[11].data = &p->locktime; | |
2553 | ||
2554 | if (dev) { | |
2555 | dev_name_source = dev->name; | |
2556 | t->neigh_dev[0].ctl_name = dev->ifindex; | |
2557 | t->neigh_vars[12].procname = NULL; | |
2558 | t->neigh_vars[13].procname = NULL; | |
2559 | t->neigh_vars[14].procname = NULL; | |
2560 | t->neigh_vars[15].procname = NULL; | |
2561 | } else { | |
2562 | dev_name_source = t->neigh_dev[0].procname; | |
2563 | t->neigh_vars[12].data = (int *)(p + 1); | |
2564 | t->neigh_vars[13].data = (int *)(p + 1) + 1; | |
2565 | t->neigh_vars[14].data = (int *)(p + 1) + 2; | |
2566 | t->neigh_vars[15].data = (int *)(p + 1) + 3; | |
2567 | } | |
2568 | ||
2569 | t->neigh_vars[16].data = &p->retrans_time; | |
2570 | t->neigh_vars[17].data = &p->base_reachable_time; | |
2571 | ||
2572 | if (handler || strategy) { | |
2573 | /* RetransTime */ | |
2574 | t->neigh_vars[3].proc_handler = handler; | |
2575 | t->neigh_vars[3].strategy = strategy; | |
2576 | t->neigh_vars[3].extra1 = dev; | |
2577 | /* ReachableTime */ | |
2578 | t->neigh_vars[4].proc_handler = handler; | |
2579 | t->neigh_vars[4].strategy = strategy; | |
2580 | t->neigh_vars[4].extra1 = dev; | |
2581 | /* RetransTime (in milliseconds)*/ | |
2582 | t->neigh_vars[16].proc_handler = handler; | |
2583 | t->neigh_vars[16].strategy = strategy; | |
2584 | t->neigh_vars[16].extra1 = dev; | |
2585 | /* ReachableTime (in milliseconds) */ | |
2586 | t->neigh_vars[17].proc_handler = handler; | |
2587 | t->neigh_vars[17].strategy = strategy; | |
2588 | t->neigh_vars[17].extra1 = dev; | |
2589 | } | |
2590 | ||
543537bd | 2591 | dev_name = kstrdup(dev_name_source, GFP_KERNEL); |
1da177e4 LT |
2592 | if (!dev_name) { |
2593 | err = -ENOBUFS; | |
2594 | goto free; | |
2595 | } | |
2596 | ||
2597 | t->neigh_dev[0].procname = dev_name; | |
2598 | ||
2599 | t->neigh_neigh_dir[0].ctl_name = pdev_id; | |
2600 | ||
2601 | t->neigh_proto_dir[0].procname = p_name; | |
2602 | t->neigh_proto_dir[0].ctl_name = p_id; | |
2603 | ||
2604 | t->neigh_dev[0].child = t->neigh_vars; | |
2605 | t->neigh_neigh_dir[0].child = t->neigh_dev; | |
2606 | t->neigh_proto_dir[0].child = t->neigh_neigh_dir; | |
2607 | t->neigh_root_dir[0].child = t->neigh_proto_dir; | |
2608 | ||
2609 | t->sysctl_header = register_sysctl_table(t->neigh_root_dir, 0); | |
2610 | if (!t->sysctl_header) { | |
2611 | err = -ENOBUFS; | |
2612 | goto free_procname; | |
2613 | } | |
2614 | p->sysctl_table = t; | |
2615 | return 0; | |
2616 | ||
2617 | /* error path */ | |
2618 | free_procname: | |
2619 | kfree(dev_name); | |
2620 | free: | |
2621 | kfree(t); | |
2622 | ||
2623 | return err; | |
2624 | } | |
2625 | ||
2626 | void neigh_sysctl_unregister(struct neigh_parms *p) | |
2627 | { | |
2628 | if (p->sysctl_table) { | |
2629 | struct neigh_sysctl_table *t = p->sysctl_table; | |
2630 | p->sysctl_table = NULL; | |
2631 | unregister_sysctl_table(t->sysctl_header); | |
2632 | kfree(t->neigh_dev[0].procname); | |
2633 | kfree(t); | |
2634 | } | |
2635 | } | |
2636 | ||
2637 | #endif /* CONFIG_SYSCTL */ | |
2638 | ||
2639 | EXPORT_SYMBOL(__neigh_event_send); | |
2640 | EXPORT_SYMBOL(neigh_add); | |
2641 | EXPORT_SYMBOL(neigh_changeaddr); | |
2642 | EXPORT_SYMBOL(neigh_compat_output); | |
2643 | EXPORT_SYMBOL(neigh_connected_output); | |
2644 | EXPORT_SYMBOL(neigh_create); | |
2645 | EXPORT_SYMBOL(neigh_delete); | |
2646 | EXPORT_SYMBOL(neigh_destroy); | |
2647 | EXPORT_SYMBOL(neigh_dump_info); | |
2648 | EXPORT_SYMBOL(neigh_event_ns); | |
2649 | EXPORT_SYMBOL(neigh_ifdown); | |
2650 | EXPORT_SYMBOL(neigh_lookup); | |
2651 | EXPORT_SYMBOL(neigh_lookup_nodev); | |
2652 | EXPORT_SYMBOL(neigh_parms_alloc); | |
2653 | EXPORT_SYMBOL(neigh_parms_release); | |
2654 | EXPORT_SYMBOL(neigh_rand_reach_time); | |
2655 | EXPORT_SYMBOL(neigh_resolve_output); | |
2656 | EXPORT_SYMBOL(neigh_table_clear); | |
2657 | EXPORT_SYMBOL(neigh_table_init); | |
2658 | EXPORT_SYMBOL(neigh_update); | |
2659 | EXPORT_SYMBOL(neigh_update_hhs); | |
2660 | EXPORT_SYMBOL(pneigh_enqueue); | |
2661 | EXPORT_SYMBOL(pneigh_lookup); | |
c7fb64db TG |
2662 | EXPORT_SYMBOL(neightbl_dump_info); |
2663 | EXPORT_SYMBOL(neightbl_set); | |
1da177e4 LT |
2664 | |
2665 | #ifdef CONFIG_ARPD | |
2666 | EXPORT_SYMBOL(neigh_app_ns); | |
2667 | #endif | |
2668 | #ifdef CONFIG_SYSCTL | |
2669 | EXPORT_SYMBOL(neigh_sysctl_register); | |
2670 | EXPORT_SYMBOL(neigh_sysctl_unregister); | |
2671 | #endif |