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) { | |
77d04bd9 | 287 | ret = kzalloc(size, GFP_ATOMIC); |
1da177e4 LT |
288 | } else { |
289 | ret = (struct neighbour **) | |
77d04bd9 | 290 | __get_free_pages(GFP_ATOMIC|__GFP_ZERO, get_order(size)); |
1da177e4 | 291 | } |
1da177e4 LT |
292 | return ret; |
293 | } | |
294 | ||
295 | static void neigh_hash_free(struct neighbour **hash, unsigned int entries) | |
296 | { | |
297 | unsigned long size = entries * sizeof(struct neighbour *); | |
298 | ||
299 | if (size <= PAGE_SIZE) | |
300 | kfree(hash); | |
301 | else | |
302 | free_pages((unsigned long)hash, get_order(size)); | |
303 | } | |
304 | ||
305 | static void neigh_hash_grow(struct neigh_table *tbl, unsigned long new_entries) | |
306 | { | |
307 | struct neighbour **new_hash, **old_hash; | |
308 | unsigned int i, new_hash_mask, old_entries; | |
309 | ||
310 | NEIGH_CACHE_STAT_INC(tbl, hash_grows); | |
311 | ||
312 | BUG_ON(new_entries & (new_entries - 1)); | |
313 | new_hash = neigh_hash_alloc(new_entries); | |
314 | if (!new_hash) | |
315 | return; | |
316 | ||
317 | old_entries = tbl->hash_mask + 1; | |
318 | new_hash_mask = new_entries - 1; | |
319 | old_hash = tbl->hash_buckets; | |
320 | ||
321 | get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd)); | |
322 | for (i = 0; i < old_entries; i++) { | |
323 | struct neighbour *n, *next; | |
324 | ||
325 | for (n = old_hash[i]; n; n = next) { | |
326 | unsigned int hash_val = tbl->hash(n->primary_key, n->dev); | |
327 | ||
328 | hash_val &= new_hash_mask; | |
329 | next = n->next; | |
330 | ||
331 | n->next = new_hash[hash_val]; | |
332 | new_hash[hash_val] = n; | |
333 | } | |
334 | } | |
335 | tbl->hash_buckets = new_hash; | |
336 | tbl->hash_mask = new_hash_mask; | |
337 | ||
338 | neigh_hash_free(old_hash, old_entries); | |
339 | } | |
340 | ||
341 | struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey, | |
342 | struct net_device *dev) | |
343 | { | |
344 | struct neighbour *n; | |
345 | int key_len = tbl->key_len; | |
346 | u32 hash_val = tbl->hash(pkey, dev) & tbl->hash_mask; | |
347 | ||
348 | NEIGH_CACHE_STAT_INC(tbl, lookups); | |
349 | ||
350 | read_lock_bh(&tbl->lock); | |
351 | for (n = tbl->hash_buckets[hash_val]; n; n = n->next) { | |
352 | if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) { | |
353 | neigh_hold(n); | |
354 | NEIGH_CACHE_STAT_INC(tbl, hits); | |
355 | break; | |
356 | } | |
357 | } | |
358 | read_unlock_bh(&tbl->lock); | |
359 | return n; | |
360 | } | |
361 | ||
362 | struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, const void *pkey) | |
363 | { | |
364 | struct neighbour *n; | |
365 | int key_len = tbl->key_len; | |
366 | u32 hash_val = tbl->hash(pkey, NULL) & tbl->hash_mask; | |
367 | ||
368 | NEIGH_CACHE_STAT_INC(tbl, lookups); | |
369 | ||
370 | read_lock_bh(&tbl->lock); | |
371 | for (n = tbl->hash_buckets[hash_val]; n; n = n->next) { | |
372 | if (!memcmp(n->primary_key, pkey, key_len)) { | |
373 | neigh_hold(n); | |
374 | NEIGH_CACHE_STAT_INC(tbl, hits); | |
375 | break; | |
376 | } | |
377 | } | |
378 | read_unlock_bh(&tbl->lock); | |
379 | return n; | |
380 | } | |
381 | ||
382 | struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey, | |
383 | struct net_device *dev) | |
384 | { | |
385 | u32 hash_val; | |
386 | int key_len = tbl->key_len; | |
387 | int error; | |
388 | struct neighbour *n1, *rc, *n = neigh_alloc(tbl); | |
389 | ||
390 | if (!n) { | |
391 | rc = ERR_PTR(-ENOBUFS); | |
392 | goto out; | |
393 | } | |
394 | ||
395 | memcpy(n->primary_key, pkey, key_len); | |
396 | n->dev = dev; | |
397 | dev_hold(dev); | |
398 | ||
399 | /* Protocol specific setup. */ | |
400 | if (tbl->constructor && (error = tbl->constructor(n)) < 0) { | |
401 | rc = ERR_PTR(error); | |
402 | goto out_neigh_release; | |
403 | } | |
404 | ||
405 | /* Device specific setup. */ | |
406 | if (n->parms->neigh_setup && | |
407 | (error = n->parms->neigh_setup(n)) < 0) { | |
408 | rc = ERR_PTR(error); | |
409 | goto out_neigh_release; | |
410 | } | |
411 | ||
412 | n->confirmed = jiffies - (n->parms->base_reachable_time << 1); | |
413 | ||
414 | write_lock_bh(&tbl->lock); | |
415 | ||
416 | if (atomic_read(&tbl->entries) > (tbl->hash_mask + 1)) | |
417 | neigh_hash_grow(tbl, (tbl->hash_mask + 1) << 1); | |
418 | ||
419 | hash_val = tbl->hash(pkey, dev) & tbl->hash_mask; | |
420 | ||
421 | if (n->parms->dead) { | |
422 | rc = ERR_PTR(-EINVAL); | |
423 | goto out_tbl_unlock; | |
424 | } | |
425 | ||
426 | for (n1 = tbl->hash_buckets[hash_val]; n1; n1 = n1->next) { | |
427 | if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) { | |
428 | neigh_hold(n1); | |
429 | rc = n1; | |
430 | goto out_tbl_unlock; | |
431 | } | |
432 | } | |
433 | ||
434 | n->next = tbl->hash_buckets[hash_val]; | |
435 | tbl->hash_buckets[hash_val] = n; | |
436 | n->dead = 0; | |
437 | neigh_hold(n); | |
438 | write_unlock_bh(&tbl->lock); | |
439 | NEIGH_PRINTK2("neigh %p is created.\n", n); | |
440 | rc = n; | |
441 | out: | |
442 | return rc; | |
443 | out_tbl_unlock: | |
444 | write_unlock_bh(&tbl->lock); | |
445 | out_neigh_release: | |
446 | neigh_release(n); | |
447 | goto out; | |
448 | } | |
449 | ||
450 | struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl, const void *pkey, | |
451 | struct net_device *dev, int creat) | |
452 | { | |
453 | struct pneigh_entry *n; | |
454 | int key_len = tbl->key_len; | |
455 | u32 hash_val = *(u32 *)(pkey + key_len - 4); | |
456 | ||
457 | hash_val ^= (hash_val >> 16); | |
458 | hash_val ^= hash_val >> 8; | |
459 | hash_val ^= hash_val >> 4; | |
460 | hash_val &= PNEIGH_HASHMASK; | |
461 | ||
462 | read_lock_bh(&tbl->lock); | |
463 | ||
464 | for (n = tbl->phash_buckets[hash_val]; n; n = n->next) { | |
465 | if (!memcmp(n->key, pkey, key_len) && | |
466 | (n->dev == dev || !n->dev)) { | |
467 | read_unlock_bh(&tbl->lock); | |
468 | goto out; | |
469 | } | |
470 | } | |
471 | read_unlock_bh(&tbl->lock); | |
472 | n = NULL; | |
473 | if (!creat) | |
474 | goto out; | |
475 | ||
476 | n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL); | |
477 | if (!n) | |
478 | goto out; | |
479 | ||
480 | memcpy(n->key, pkey, key_len); | |
481 | n->dev = dev; | |
482 | if (dev) | |
483 | dev_hold(dev); | |
484 | ||
485 | if (tbl->pconstructor && tbl->pconstructor(n)) { | |
486 | if (dev) | |
487 | dev_put(dev); | |
488 | kfree(n); | |
489 | n = NULL; | |
490 | goto out; | |
491 | } | |
492 | ||
493 | write_lock_bh(&tbl->lock); | |
494 | n->next = tbl->phash_buckets[hash_val]; | |
495 | tbl->phash_buckets[hash_val] = n; | |
496 | write_unlock_bh(&tbl->lock); | |
497 | out: | |
498 | return n; | |
499 | } | |
500 | ||
501 | ||
502 | int pneigh_delete(struct neigh_table *tbl, const void *pkey, | |
503 | struct net_device *dev) | |
504 | { | |
505 | struct pneigh_entry *n, **np; | |
506 | int key_len = tbl->key_len; | |
507 | u32 hash_val = *(u32 *)(pkey + key_len - 4); | |
508 | ||
509 | hash_val ^= (hash_val >> 16); | |
510 | hash_val ^= hash_val >> 8; | |
511 | hash_val ^= hash_val >> 4; | |
512 | hash_val &= PNEIGH_HASHMASK; | |
513 | ||
514 | write_lock_bh(&tbl->lock); | |
515 | for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL; | |
516 | np = &n->next) { | |
517 | if (!memcmp(n->key, pkey, key_len) && n->dev == dev) { | |
518 | *np = n->next; | |
519 | write_unlock_bh(&tbl->lock); | |
520 | if (tbl->pdestructor) | |
521 | tbl->pdestructor(n); | |
522 | if (n->dev) | |
523 | dev_put(n->dev); | |
524 | kfree(n); | |
525 | return 0; | |
526 | } | |
527 | } | |
528 | write_unlock_bh(&tbl->lock); | |
529 | return -ENOENT; | |
530 | } | |
531 | ||
532 | static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev) | |
533 | { | |
534 | struct pneigh_entry *n, **np; | |
535 | u32 h; | |
536 | ||
537 | for (h = 0; h <= PNEIGH_HASHMASK; h++) { | |
538 | np = &tbl->phash_buckets[h]; | |
539 | while ((n = *np) != NULL) { | |
540 | if (!dev || n->dev == dev) { | |
541 | *np = n->next; | |
542 | if (tbl->pdestructor) | |
543 | tbl->pdestructor(n); | |
544 | if (n->dev) | |
545 | dev_put(n->dev); | |
546 | kfree(n); | |
547 | continue; | |
548 | } | |
549 | np = &n->next; | |
550 | } | |
551 | } | |
552 | return -ENOENT; | |
553 | } | |
554 | ||
555 | ||
556 | /* | |
557 | * neighbour must already be out of the table; | |
558 | * | |
559 | */ | |
560 | void neigh_destroy(struct neighbour *neigh) | |
561 | { | |
562 | struct hh_cache *hh; | |
563 | ||
564 | NEIGH_CACHE_STAT_INC(neigh->tbl, destroys); | |
565 | ||
566 | if (!neigh->dead) { | |
567 | printk(KERN_WARNING | |
568 | "Destroying alive neighbour %p\n", neigh); | |
569 | dump_stack(); | |
570 | return; | |
571 | } | |
572 | ||
573 | if (neigh_del_timer(neigh)) | |
574 | printk(KERN_WARNING "Impossible event.\n"); | |
575 | ||
576 | while ((hh = neigh->hh) != NULL) { | |
577 | neigh->hh = hh->hh_next; | |
578 | hh->hh_next = NULL; | |
579 | write_lock_bh(&hh->hh_lock); | |
580 | hh->hh_output = neigh_blackhole; | |
581 | write_unlock_bh(&hh->hh_lock); | |
582 | if (atomic_dec_and_test(&hh->hh_refcnt)) | |
583 | kfree(hh); | |
584 | } | |
585 | ||
c5ecd62c MT |
586 | if (neigh->parms->neigh_destructor) |
587 | (neigh->parms->neigh_destructor)(neigh); | |
1da177e4 LT |
588 | |
589 | skb_queue_purge(&neigh->arp_queue); | |
590 | ||
591 | dev_put(neigh->dev); | |
592 | neigh_parms_put(neigh->parms); | |
593 | ||
594 | NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh); | |
595 | ||
596 | atomic_dec(&neigh->tbl->entries); | |
597 | kmem_cache_free(neigh->tbl->kmem_cachep, neigh); | |
598 | } | |
599 | ||
600 | /* Neighbour state is suspicious; | |
601 | disable fast path. | |
602 | ||
603 | Called with write_locked neigh. | |
604 | */ | |
605 | static void neigh_suspect(struct neighbour *neigh) | |
606 | { | |
607 | struct hh_cache *hh; | |
608 | ||
609 | NEIGH_PRINTK2("neigh %p is suspected.\n", neigh); | |
610 | ||
611 | neigh->output = neigh->ops->output; | |
612 | ||
613 | for (hh = neigh->hh; hh; hh = hh->hh_next) | |
614 | hh->hh_output = neigh->ops->output; | |
615 | } | |
616 | ||
617 | /* Neighbour state is OK; | |
618 | enable fast path. | |
619 | ||
620 | Called with write_locked neigh. | |
621 | */ | |
622 | static void neigh_connect(struct neighbour *neigh) | |
623 | { | |
624 | struct hh_cache *hh; | |
625 | ||
626 | NEIGH_PRINTK2("neigh %p is connected.\n", neigh); | |
627 | ||
628 | neigh->output = neigh->ops->connected_output; | |
629 | ||
630 | for (hh = neigh->hh; hh; hh = hh->hh_next) | |
631 | hh->hh_output = neigh->ops->hh_output; | |
632 | } | |
633 | ||
634 | static void neigh_periodic_timer(unsigned long arg) | |
635 | { | |
636 | struct neigh_table *tbl = (struct neigh_table *)arg; | |
637 | struct neighbour *n, **np; | |
638 | unsigned long expire, now = jiffies; | |
639 | ||
640 | NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs); | |
641 | ||
642 | write_lock(&tbl->lock); | |
643 | ||
644 | /* | |
645 | * periodically recompute ReachableTime from random function | |
646 | */ | |
647 | ||
648 | if (time_after(now, tbl->last_rand + 300 * HZ)) { | |
649 | struct neigh_parms *p; | |
650 | tbl->last_rand = now; | |
651 | for (p = &tbl->parms; p; p = p->next) | |
652 | p->reachable_time = | |
653 | neigh_rand_reach_time(p->base_reachable_time); | |
654 | } | |
655 | ||
656 | np = &tbl->hash_buckets[tbl->hash_chain_gc]; | |
657 | tbl->hash_chain_gc = ((tbl->hash_chain_gc + 1) & tbl->hash_mask); | |
658 | ||
659 | while ((n = *np) != NULL) { | |
660 | unsigned int state; | |
661 | ||
662 | write_lock(&n->lock); | |
663 | ||
664 | state = n->nud_state; | |
665 | if (state & (NUD_PERMANENT | NUD_IN_TIMER)) { | |
666 | write_unlock(&n->lock); | |
667 | goto next_elt; | |
668 | } | |
669 | ||
670 | if (time_before(n->used, n->confirmed)) | |
671 | n->used = n->confirmed; | |
672 | ||
673 | if (atomic_read(&n->refcnt) == 1 && | |
674 | (state == NUD_FAILED || | |
675 | time_after(now, n->used + n->parms->gc_staletime))) { | |
676 | *np = n->next; | |
677 | n->dead = 1; | |
678 | write_unlock(&n->lock); | |
679 | neigh_release(n); | |
680 | continue; | |
681 | } | |
682 | write_unlock(&n->lock); | |
683 | ||
684 | next_elt: | |
685 | np = &n->next; | |
686 | } | |
687 | ||
688 | /* Cycle through all hash buckets every base_reachable_time/2 ticks. | |
689 | * ARP entry timeouts range from 1/2 base_reachable_time to 3/2 | |
690 | * base_reachable_time. | |
691 | */ | |
692 | expire = tbl->parms.base_reachable_time >> 1; | |
693 | expire /= (tbl->hash_mask + 1); | |
694 | if (!expire) | |
695 | expire = 1; | |
696 | ||
697 | mod_timer(&tbl->gc_timer, now + expire); | |
698 | ||
699 | write_unlock(&tbl->lock); | |
700 | } | |
701 | ||
702 | static __inline__ int neigh_max_probes(struct neighbour *n) | |
703 | { | |
704 | struct neigh_parms *p = n->parms; | |
705 | return (n->nud_state & NUD_PROBE ? | |
706 | p->ucast_probes : | |
707 | p->ucast_probes + p->app_probes + p->mcast_probes); | |
708 | } | |
709 | ||
667347f1 DM |
710 | static inline void neigh_add_timer(struct neighbour *n, unsigned long when) |
711 | { | |
712 | if (unlikely(mod_timer(&n->timer, when))) { | |
713 | printk("NEIGH: BUG, double timer add, state is %x\n", | |
714 | n->nud_state); | |
20375502 | 715 | dump_stack(); |
667347f1 DM |
716 | } |
717 | } | |
1da177e4 LT |
718 | |
719 | /* Called when a timer expires for a neighbour entry. */ | |
720 | ||
721 | static void neigh_timer_handler(unsigned long arg) | |
722 | { | |
723 | unsigned long now, next; | |
724 | struct neighbour *neigh = (struct neighbour *)arg; | |
725 | unsigned state; | |
726 | int notify = 0; | |
727 | ||
728 | write_lock(&neigh->lock); | |
729 | ||
730 | state = neigh->nud_state; | |
731 | now = jiffies; | |
732 | next = now + HZ; | |
733 | ||
734 | if (!(state & NUD_IN_TIMER)) { | |
735 | #ifndef CONFIG_SMP | |
736 | printk(KERN_WARNING "neigh: timer & !nud_in_timer\n"); | |
737 | #endif | |
738 | goto out; | |
739 | } | |
740 | ||
741 | if (state & NUD_REACHABLE) { | |
742 | if (time_before_eq(now, | |
743 | neigh->confirmed + neigh->parms->reachable_time)) { | |
744 | NEIGH_PRINTK2("neigh %p is still alive.\n", neigh); | |
745 | next = neigh->confirmed + neigh->parms->reachable_time; | |
746 | } else if (time_before_eq(now, | |
747 | neigh->used + neigh->parms->delay_probe_time)) { | |
748 | NEIGH_PRINTK2("neigh %p is delayed.\n", neigh); | |
749 | neigh->nud_state = NUD_DELAY; | |
955aaa2f | 750 | neigh->updated = jiffies; |
1da177e4 LT |
751 | neigh_suspect(neigh); |
752 | next = now + neigh->parms->delay_probe_time; | |
753 | } else { | |
754 | NEIGH_PRINTK2("neigh %p is suspected.\n", neigh); | |
755 | neigh->nud_state = NUD_STALE; | |
955aaa2f | 756 | neigh->updated = jiffies; |
1da177e4 LT |
757 | neigh_suspect(neigh); |
758 | } | |
759 | } else if (state & NUD_DELAY) { | |
760 | if (time_before_eq(now, | |
761 | neigh->confirmed + neigh->parms->delay_probe_time)) { | |
762 | NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh); | |
763 | neigh->nud_state = NUD_REACHABLE; | |
955aaa2f | 764 | neigh->updated = jiffies; |
1da177e4 LT |
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; | |
955aaa2f | 770 | neigh->updated = jiffies; |
1da177e4 LT |
771 | atomic_set(&neigh->probes, 0); |
772 | next = now + neigh->parms->retrans_time; | |
773 | } | |
774 | } else { | |
775 | /* NUD_PROBE|NUD_INCOMPLETE */ | |
776 | next = now + neigh->parms->retrans_time; | |
777 | } | |
778 | ||
779 | if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) && | |
780 | atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) { | |
781 | struct sk_buff *skb; | |
782 | ||
783 | neigh->nud_state = NUD_FAILED; | |
955aaa2f | 784 | neigh->updated = jiffies; |
1da177e4 LT |
785 | notify = 1; |
786 | NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed); | |
787 | NEIGH_PRINTK2("neigh %p is failed.\n", neigh); | |
788 | ||
789 | /* It is very thin place. report_unreachable is very complicated | |
790 | routine. Particularly, it can hit the same neighbour entry! | |
791 | ||
792 | So that, we try to be accurate and avoid dead loop. --ANK | |
793 | */ | |
794 | while (neigh->nud_state == NUD_FAILED && | |
795 | (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) { | |
796 | write_unlock(&neigh->lock); | |
797 | neigh->ops->error_report(neigh, skb); | |
798 | write_lock(&neigh->lock); | |
799 | } | |
800 | skb_queue_purge(&neigh->arp_queue); | |
801 | } | |
802 | ||
803 | if (neigh->nud_state & NUD_IN_TIMER) { | |
1da177e4 LT |
804 | if (time_before(next, jiffies + HZ/2)) |
805 | next = jiffies + HZ/2; | |
6fb9974f HX |
806 | if (!mod_timer(&neigh->timer, next)) |
807 | neigh_hold(neigh); | |
1da177e4 LT |
808 | } |
809 | if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) { | |
810 | struct sk_buff *skb = skb_peek(&neigh->arp_queue); | |
811 | /* keep skb alive even if arp_queue overflows */ | |
812 | if (skb) | |
813 | skb_get(skb); | |
814 | write_unlock(&neigh->lock); | |
815 | neigh->ops->solicit(neigh, skb); | |
816 | atomic_inc(&neigh->probes); | |
817 | if (skb) | |
818 | kfree_skb(skb); | |
819 | } else { | |
820 | out: | |
821 | write_unlock(&neigh->lock); | |
822 | } | |
823 | ||
824 | #ifdef CONFIG_ARPD | |
825 | if (notify && neigh->parms->app_probes) | |
826 | neigh_app_notify(neigh); | |
827 | #endif | |
828 | neigh_release(neigh); | |
829 | } | |
830 | ||
831 | int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb) | |
832 | { | |
833 | int rc; | |
834 | unsigned long now; | |
835 | ||
836 | write_lock_bh(&neigh->lock); | |
837 | ||
838 | rc = 0; | |
839 | if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE)) | |
840 | goto out_unlock_bh; | |
841 | ||
842 | now = jiffies; | |
843 | ||
844 | if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) { | |
845 | if (neigh->parms->mcast_probes + neigh->parms->app_probes) { | |
846 | atomic_set(&neigh->probes, neigh->parms->ucast_probes); | |
847 | neigh->nud_state = NUD_INCOMPLETE; | |
955aaa2f | 848 | neigh->updated = jiffies; |
1da177e4 | 849 | neigh_hold(neigh); |
667347f1 | 850 | neigh_add_timer(neigh, now + 1); |
1da177e4 LT |
851 | } else { |
852 | neigh->nud_state = NUD_FAILED; | |
955aaa2f | 853 | neigh->updated = jiffies; |
1da177e4 LT |
854 | write_unlock_bh(&neigh->lock); |
855 | ||
856 | if (skb) | |
857 | kfree_skb(skb); | |
858 | return 1; | |
859 | } | |
860 | } else if (neigh->nud_state & NUD_STALE) { | |
861 | NEIGH_PRINTK2("neigh %p is delayed.\n", neigh); | |
862 | neigh_hold(neigh); | |
863 | neigh->nud_state = NUD_DELAY; | |
955aaa2f | 864 | neigh->updated = jiffies; |
667347f1 DM |
865 | neigh_add_timer(neigh, |
866 | jiffies + neigh->parms->delay_probe_time); | |
1da177e4 LT |
867 | } |
868 | ||
869 | if (neigh->nud_state == NUD_INCOMPLETE) { | |
870 | if (skb) { | |
871 | if (skb_queue_len(&neigh->arp_queue) >= | |
872 | neigh->parms->queue_len) { | |
873 | struct sk_buff *buff; | |
874 | buff = neigh->arp_queue.next; | |
875 | __skb_unlink(buff, &neigh->arp_queue); | |
876 | kfree_skb(buff); | |
877 | } | |
878 | __skb_queue_tail(&neigh->arp_queue, skb); | |
879 | } | |
880 | rc = 1; | |
881 | } | |
882 | out_unlock_bh: | |
883 | write_unlock_bh(&neigh->lock); | |
884 | return rc; | |
885 | } | |
886 | ||
887 | static __inline__ void neigh_update_hhs(struct neighbour *neigh) | |
888 | { | |
889 | struct hh_cache *hh; | |
890 | void (*update)(struct hh_cache*, struct net_device*, unsigned char *) = | |
891 | neigh->dev->header_cache_update; | |
892 | ||
893 | if (update) { | |
894 | for (hh = neigh->hh; hh; hh = hh->hh_next) { | |
895 | write_lock_bh(&hh->hh_lock); | |
896 | update(hh, neigh->dev, neigh->ha); | |
897 | write_unlock_bh(&hh->hh_lock); | |
898 | } | |
899 | } | |
900 | } | |
901 | ||
902 | ||
903 | ||
904 | /* Generic update routine. | |
905 | -- lladdr is new lladdr or NULL, if it is not supplied. | |
906 | -- new is new state. | |
907 | -- flags | |
908 | NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr, | |
909 | if it is different. | |
910 | NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected" | |
911 | lladdr instead of overriding it | |
912 | if it is different. | |
913 | It also allows to retain current state | |
914 | if lladdr is unchanged. | |
915 | NEIGH_UPDATE_F_ADMIN means that the change is administrative. | |
916 | ||
917 | NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing | |
918 | NTF_ROUTER flag. | |
919 | NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as | |
920 | a router. | |
921 | ||
922 | Caller MUST hold reference count on the entry. | |
923 | */ | |
924 | ||
925 | int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new, | |
926 | u32 flags) | |
927 | { | |
928 | u8 old; | |
929 | int err; | |
930 | #ifdef CONFIG_ARPD | |
931 | int notify = 0; | |
932 | #endif | |
933 | struct net_device *dev; | |
934 | int update_isrouter = 0; | |
935 | ||
936 | write_lock_bh(&neigh->lock); | |
937 | ||
938 | dev = neigh->dev; | |
939 | old = neigh->nud_state; | |
940 | err = -EPERM; | |
941 | ||
942 | if (!(flags & NEIGH_UPDATE_F_ADMIN) && | |
943 | (old & (NUD_NOARP | NUD_PERMANENT))) | |
944 | goto out; | |
945 | ||
946 | if (!(new & NUD_VALID)) { | |
947 | neigh_del_timer(neigh); | |
948 | if (old & NUD_CONNECTED) | |
949 | neigh_suspect(neigh); | |
950 | neigh->nud_state = new; | |
951 | err = 0; | |
952 | #ifdef CONFIG_ARPD | |
953 | notify = old & NUD_VALID; | |
954 | #endif | |
955 | goto out; | |
956 | } | |
957 | ||
958 | /* Compare new lladdr with cached one */ | |
959 | if (!dev->addr_len) { | |
960 | /* First case: device needs no address. */ | |
961 | lladdr = neigh->ha; | |
962 | } else if (lladdr) { | |
963 | /* The second case: if something is already cached | |
964 | and a new address is proposed: | |
965 | - compare new & old | |
966 | - if they are different, check override flag | |
967 | */ | |
968 | if ((old & NUD_VALID) && | |
969 | !memcmp(lladdr, neigh->ha, dev->addr_len)) | |
970 | lladdr = neigh->ha; | |
971 | } else { | |
972 | /* No address is supplied; if we know something, | |
973 | use it, otherwise discard the request. | |
974 | */ | |
975 | err = -EINVAL; | |
976 | if (!(old & NUD_VALID)) | |
977 | goto out; | |
978 | lladdr = neigh->ha; | |
979 | } | |
980 | ||
981 | if (new & NUD_CONNECTED) | |
982 | neigh->confirmed = jiffies; | |
983 | neigh->updated = jiffies; | |
984 | ||
985 | /* If entry was valid and address is not changed, | |
986 | do not change entry state, if new one is STALE. | |
987 | */ | |
988 | err = 0; | |
989 | update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER; | |
990 | if (old & NUD_VALID) { | |
991 | if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) { | |
992 | update_isrouter = 0; | |
993 | if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) && | |
994 | (old & NUD_CONNECTED)) { | |
995 | lladdr = neigh->ha; | |
996 | new = NUD_STALE; | |
997 | } else | |
998 | goto out; | |
999 | } else { | |
1000 | if (lladdr == neigh->ha && new == NUD_STALE && | |
1001 | ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) || | |
1002 | (old & NUD_CONNECTED)) | |
1003 | ) | |
1004 | new = old; | |
1005 | } | |
1006 | } | |
1007 | ||
1008 | if (new != old) { | |
1009 | neigh_del_timer(neigh); | |
1010 | if (new & NUD_IN_TIMER) { | |
1011 | neigh_hold(neigh); | |
667347f1 | 1012 | neigh_add_timer(neigh, (jiffies + |
1da177e4 | 1013 | ((new & NUD_REACHABLE) ? |
667347f1 DM |
1014 | neigh->parms->reachable_time : |
1015 | 0))); | |
1da177e4 LT |
1016 | } |
1017 | neigh->nud_state = new; | |
1018 | } | |
1019 | ||
1020 | if (lladdr != neigh->ha) { | |
1021 | memcpy(&neigh->ha, lladdr, dev->addr_len); | |
1022 | neigh_update_hhs(neigh); | |
1023 | if (!(new & NUD_CONNECTED)) | |
1024 | neigh->confirmed = jiffies - | |
1025 | (neigh->parms->base_reachable_time << 1); | |
1026 | #ifdef CONFIG_ARPD | |
1027 | notify = 1; | |
1028 | #endif | |
1029 | } | |
1030 | if (new == old) | |
1031 | goto out; | |
1032 | if (new & NUD_CONNECTED) | |
1033 | neigh_connect(neigh); | |
1034 | else | |
1035 | neigh_suspect(neigh); | |
1036 | if (!(old & NUD_VALID)) { | |
1037 | struct sk_buff *skb; | |
1038 | ||
1039 | /* Again: avoid dead loop if something went wrong */ | |
1040 | ||
1041 | while (neigh->nud_state & NUD_VALID && | |
1042 | (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) { | |
1043 | struct neighbour *n1 = neigh; | |
1044 | write_unlock_bh(&neigh->lock); | |
1045 | /* On shaper/eql skb->dst->neighbour != neigh :( */ | |
1046 | if (skb->dst && skb->dst->neighbour) | |
1047 | n1 = skb->dst->neighbour; | |
1048 | n1->output(skb); | |
1049 | write_lock_bh(&neigh->lock); | |
1050 | } | |
1051 | skb_queue_purge(&neigh->arp_queue); | |
1052 | } | |
1053 | out: | |
1054 | if (update_isrouter) { | |
1055 | neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ? | |
1056 | (neigh->flags | NTF_ROUTER) : | |
1057 | (neigh->flags & ~NTF_ROUTER); | |
1058 | } | |
1059 | write_unlock_bh(&neigh->lock); | |
1060 | #ifdef CONFIG_ARPD | |
1061 | if (notify && neigh->parms->app_probes) | |
1062 | neigh_app_notify(neigh); | |
1063 | #endif | |
1064 | return err; | |
1065 | } | |
1066 | ||
1067 | struct neighbour *neigh_event_ns(struct neigh_table *tbl, | |
1068 | u8 *lladdr, void *saddr, | |
1069 | struct net_device *dev) | |
1070 | { | |
1071 | struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev, | |
1072 | lladdr || !dev->addr_len); | |
1073 | if (neigh) | |
1074 | neigh_update(neigh, lladdr, NUD_STALE, | |
1075 | NEIGH_UPDATE_F_OVERRIDE); | |
1076 | return neigh; | |
1077 | } | |
1078 | ||
1079 | static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst, | |
1080 | u16 protocol) | |
1081 | { | |
1082 | struct hh_cache *hh; | |
1083 | struct net_device *dev = dst->dev; | |
1084 | ||
1085 | for (hh = n->hh; hh; hh = hh->hh_next) | |
1086 | if (hh->hh_type == protocol) | |
1087 | break; | |
1088 | ||
77d04bd9 | 1089 | if (!hh && (hh = kzalloc(sizeof(*hh), GFP_ATOMIC)) != NULL) { |
1da177e4 LT |
1090 | rwlock_init(&hh->hh_lock); |
1091 | hh->hh_type = protocol; | |
1092 | atomic_set(&hh->hh_refcnt, 0); | |
1093 | hh->hh_next = NULL; | |
1094 | if (dev->hard_header_cache(n, hh)) { | |
1095 | kfree(hh); | |
1096 | hh = NULL; | |
1097 | } else { | |
1098 | atomic_inc(&hh->hh_refcnt); | |
1099 | hh->hh_next = n->hh; | |
1100 | n->hh = hh; | |
1101 | if (n->nud_state & NUD_CONNECTED) | |
1102 | hh->hh_output = n->ops->hh_output; | |
1103 | else | |
1104 | hh->hh_output = n->ops->output; | |
1105 | } | |
1106 | } | |
1107 | if (hh) { | |
1108 | atomic_inc(&hh->hh_refcnt); | |
1109 | dst->hh = hh; | |
1110 | } | |
1111 | } | |
1112 | ||
1113 | /* This function can be used in contexts, where only old dev_queue_xmit | |
1114 | worked, f.e. if you want to override normal output path (eql, shaper), | |
1115 | but resolution is not made yet. | |
1116 | */ | |
1117 | ||
1118 | int neigh_compat_output(struct sk_buff *skb) | |
1119 | { | |
1120 | struct net_device *dev = skb->dev; | |
1121 | ||
1122 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1123 | ||
1124 | if (dev->hard_header && | |
1125 | dev->hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL, | |
1126 | skb->len) < 0 && | |
1127 | dev->rebuild_header(skb)) | |
1128 | return 0; | |
1129 | ||
1130 | return dev_queue_xmit(skb); | |
1131 | } | |
1132 | ||
1133 | /* Slow and careful. */ | |
1134 | ||
1135 | int neigh_resolve_output(struct sk_buff *skb) | |
1136 | { | |
1137 | struct dst_entry *dst = skb->dst; | |
1138 | struct neighbour *neigh; | |
1139 | int rc = 0; | |
1140 | ||
1141 | if (!dst || !(neigh = dst->neighbour)) | |
1142 | goto discard; | |
1143 | ||
1144 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1145 | ||
1146 | if (!neigh_event_send(neigh, skb)) { | |
1147 | int err; | |
1148 | struct net_device *dev = neigh->dev; | |
1149 | if (dev->hard_header_cache && !dst->hh) { | |
1150 | write_lock_bh(&neigh->lock); | |
1151 | if (!dst->hh) | |
1152 | neigh_hh_init(neigh, dst, dst->ops->protocol); | |
1153 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1154 | neigh->ha, NULL, skb->len); | |
1155 | write_unlock_bh(&neigh->lock); | |
1156 | } else { | |
1157 | read_lock_bh(&neigh->lock); | |
1158 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1159 | neigh->ha, NULL, skb->len); | |
1160 | read_unlock_bh(&neigh->lock); | |
1161 | } | |
1162 | if (err >= 0) | |
1163 | rc = neigh->ops->queue_xmit(skb); | |
1164 | else | |
1165 | goto out_kfree_skb; | |
1166 | } | |
1167 | out: | |
1168 | return rc; | |
1169 | discard: | |
1170 | NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n", | |
1171 | dst, dst ? dst->neighbour : NULL); | |
1172 | out_kfree_skb: | |
1173 | rc = -EINVAL; | |
1174 | kfree_skb(skb); | |
1175 | goto out; | |
1176 | } | |
1177 | ||
1178 | /* As fast as possible without hh cache */ | |
1179 | ||
1180 | int neigh_connected_output(struct sk_buff *skb) | |
1181 | { | |
1182 | int err; | |
1183 | struct dst_entry *dst = skb->dst; | |
1184 | struct neighbour *neigh = dst->neighbour; | |
1185 | struct net_device *dev = neigh->dev; | |
1186 | ||
1187 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1188 | ||
1189 | read_lock_bh(&neigh->lock); | |
1190 | err = dev->hard_header(skb, dev, ntohs(skb->protocol), | |
1191 | neigh->ha, NULL, skb->len); | |
1192 | read_unlock_bh(&neigh->lock); | |
1193 | if (err >= 0) | |
1194 | err = neigh->ops->queue_xmit(skb); | |
1195 | else { | |
1196 | err = -EINVAL; | |
1197 | kfree_skb(skb); | |
1198 | } | |
1199 | return err; | |
1200 | } | |
1201 | ||
1202 | static void neigh_proxy_process(unsigned long arg) | |
1203 | { | |
1204 | struct neigh_table *tbl = (struct neigh_table *)arg; | |
1205 | long sched_next = 0; | |
1206 | unsigned long now = jiffies; | |
1207 | struct sk_buff *skb; | |
1208 | ||
1209 | spin_lock(&tbl->proxy_queue.lock); | |
1210 | ||
1211 | skb = tbl->proxy_queue.next; | |
1212 | ||
1213 | while (skb != (struct sk_buff *)&tbl->proxy_queue) { | |
1214 | struct sk_buff *back = skb; | |
a61bbcf2 | 1215 | long tdif = NEIGH_CB(back)->sched_next - now; |
1da177e4 LT |
1216 | |
1217 | skb = skb->next; | |
1218 | if (tdif <= 0) { | |
1219 | struct net_device *dev = back->dev; | |
1220 | __skb_unlink(back, &tbl->proxy_queue); | |
1221 | if (tbl->proxy_redo && netif_running(dev)) | |
1222 | tbl->proxy_redo(back); | |
1223 | else | |
1224 | kfree_skb(back); | |
1225 | ||
1226 | dev_put(dev); | |
1227 | } else if (!sched_next || tdif < sched_next) | |
1228 | sched_next = tdif; | |
1229 | } | |
1230 | del_timer(&tbl->proxy_timer); | |
1231 | if (sched_next) | |
1232 | mod_timer(&tbl->proxy_timer, jiffies + sched_next); | |
1233 | spin_unlock(&tbl->proxy_queue.lock); | |
1234 | } | |
1235 | ||
1236 | void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p, | |
1237 | struct sk_buff *skb) | |
1238 | { | |
1239 | unsigned long now = jiffies; | |
1240 | unsigned long sched_next = now + (net_random() % p->proxy_delay); | |
1241 | ||
1242 | if (tbl->proxy_queue.qlen > p->proxy_qlen) { | |
1243 | kfree_skb(skb); | |
1244 | return; | |
1245 | } | |
a61bbcf2 PM |
1246 | |
1247 | NEIGH_CB(skb)->sched_next = sched_next; | |
1248 | NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED; | |
1da177e4 LT |
1249 | |
1250 | spin_lock(&tbl->proxy_queue.lock); | |
1251 | if (del_timer(&tbl->proxy_timer)) { | |
1252 | if (time_before(tbl->proxy_timer.expires, sched_next)) | |
1253 | sched_next = tbl->proxy_timer.expires; | |
1254 | } | |
1255 | dst_release(skb->dst); | |
1256 | skb->dst = NULL; | |
1257 | dev_hold(skb->dev); | |
1258 | __skb_queue_tail(&tbl->proxy_queue, skb); | |
1259 | mod_timer(&tbl->proxy_timer, sched_next); | |
1260 | spin_unlock(&tbl->proxy_queue.lock); | |
1261 | } | |
1262 | ||
1263 | ||
1264 | struct neigh_parms *neigh_parms_alloc(struct net_device *dev, | |
1265 | struct neigh_table *tbl) | |
1266 | { | |
1267 | struct neigh_parms *p = kmalloc(sizeof(*p), GFP_KERNEL); | |
1268 | ||
1269 | if (p) { | |
1270 | memcpy(p, &tbl->parms, sizeof(*p)); | |
1271 | p->tbl = tbl; | |
1272 | atomic_set(&p->refcnt, 1); | |
1273 | INIT_RCU_HEAD(&p->rcu_head); | |
1274 | p->reachable_time = | |
1275 | neigh_rand_reach_time(p->base_reachable_time); | |
c7fb64db TG |
1276 | if (dev) { |
1277 | if (dev->neigh_setup && dev->neigh_setup(dev, p)) { | |
1278 | kfree(p); | |
1279 | return NULL; | |
1280 | } | |
1281 | ||
1282 | dev_hold(dev); | |
1283 | p->dev = dev; | |
1da177e4 LT |
1284 | } |
1285 | p->sysctl_table = NULL; | |
1286 | write_lock_bh(&tbl->lock); | |
1287 | p->next = tbl->parms.next; | |
1288 | tbl->parms.next = p; | |
1289 | write_unlock_bh(&tbl->lock); | |
1290 | } | |
1291 | return p; | |
1292 | } | |
1293 | ||
1294 | static void neigh_rcu_free_parms(struct rcu_head *head) | |
1295 | { | |
1296 | struct neigh_parms *parms = | |
1297 | container_of(head, struct neigh_parms, rcu_head); | |
1298 | ||
1299 | neigh_parms_put(parms); | |
1300 | } | |
1301 | ||
1302 | void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms) | |
1303 | { | |
1304 | struct neigh_parms **p; | |
1305 | ||
1306 | if (!parms || parms == &tbl->parms) | |
1307 | return; | |
1308 | write_lock_bh(&tbl->lock); | |
1309 | for (p = &tbl->parms.next; *p; p = &(*p)->next) { | |
1310 | if (*p == parms) { | |
1311 | *p = parms->next; | |
1312 | parms->dead = 1; | |
1313 | write_unlock_bh(&tbl->lock); | |
c7fb64db TG |
1314 | if (parms->dev) |
1315 | dev_put(parms->dev); | |
1da177e4 LT |
1316 | call_rcu(&parms->rcu_head, neigh_rcu_free_parms); |
1317 | return; | |
1318 | } | |
1319 | } | |
1320 | write_unlock_bh(&tbl->lock); | |
1321 | NEIGH_PRINTK1("neigh_parms_release: not found\n"); | |
1322 | } | |
1323 | ||
1324 | void neigh_parms_destroy(struct neigh_parms *parms) | |
1325 | { | |
1326 | kfree(parms); | |
1327 | } | |
1328 | ||
1329 | ||
1330 | void neigh_table_init(struct neigh_table *tbl) | |
1331 | { | |
1332 | unsigned long now = jiffies; | |
1333 | unsigned long phsize; | |
1334 | ||
1335 | atomic_set(&tbl->parms.refcnt, 1); | |
1336 | INIT_RCU_HEAD(&tbl->parms.rcu_head); | |
1337 | tbl->parms.reachable_time = | |
1338 | neigh_rand_reach_time(tbl->parms.base_reachable_time); | |
1339 | ||
1340 | if (!tbl->kmem_cachep) | |
1341 | tbl->kmem_cachep = kmem_cache_create(tbl->id, | |
1342 | tbl->entry_size, | |
1343 | 0, SLAB_HWCACHE_ALIGN, | |
1344 | NULL, NULL); | |
1345 | ||
1346 | if (!tbl->kmem_cachep) | |
1347 | panic("cannot create neighbour cache"); | |
1348 | ||
1349 | tbl->stats = alloc_percpu(struct neigh_statistics); | |
1350 | if (!tbl->stats) | |
1351 | panic("cannot create neighbour cache statistics"); | |
1352 | ||
1353 | #ifdef CONFIG_PROC_FS | |
1354 | tbl->pde = create_proc_entry(tbl->id, 0, proc_net_stat); | |
1355 | if (!tbl->pde) | |
1356 | panic("cannot create neighbour proc dir entry"); | |
1357 | tbl->pde->proc_fops = &neigh_stat_seq_fops; | |
1358 | tbl->pde->data = tbl; | |
1359 | #endif | |
1360 | ||
1361 | tbl->hash_mask = 1; | |
1362 | tbl->hash_buckets = neigh_hash_alloc(tbl->hash_mask + 1); | |
1363 | ||
1364 | phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *); | |
77d04bd9 | 1365 | tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL); |
1da177e4 LT |
1366 | |
1367 | if (!tbl->hash_buckets || !tbl->phash_buckets) | |
1368 | panic("cannot allocate neighbour cache hashes"); | |
1369 | ||
1da177e4 LT |
1370 | get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd)); |
1371 | ||
1372 | rwlock_init(&tbl->lock); | |
1373 | init_timer(&tbl->gc_timer); | |
1374 | tbl->gc_timer.data = (unsigned long)tbl; | |
1375 | tbl->gc_timer.function = neigh_periodic_timer; | |
1376 | tbl->gc_timer.expires = now + 1; | |
1377 | add_timer(&tbl->gc_timer); | |
1378 | ||
1379 | init_timer(&tbl->proxy_timer); | |
1380 | tbl->proxy_timer.data = (unsigned long)tbl; | |
1381 | tbl->proxy_timer.function = neigh_proxy_process; | |
1382 | skb_queue_head_init(&tbl->proxy_queue); | |
1383 | ||
1384 | tbl->last_flush = now; | |
1385 | tbl->last_rand = now + tbl->parms.reachable_time * 20; | |
1386 | write_lock(&neigh_tbl_lock); | |
1387 | tbl->next = neigh_tables; | |
1388 | neigh_tables = tbl; | |
1389 | write_unlock(&neigh_tbl_lock); | |
1390 | } | |
1391 | ||
1392 | int neigh_table_clear(struct neigh_table *tbl) | |
1393 | { | |
1394 | struct neigh_table **tp; | |
1395 | ||
1396 | /* It is not clean... Fix it to unload IPv6 module safely */ | |
1397 | del_timer_sync(&tbl->gc_timer); | |
1398 | del_timer_sync(&tbl->proxy_timer); | |
1399 | pneigh_queue_purge(&tbl->proxy_queue); | |
1400 | neigh_ifdown(tbl, NULL); | |
1401 | if (atomic_read(&tbl->entries)) | |
1402 | printk(KERN_CRIT "neighbour leakage\n"); | |
1403 | write_lock(&neigh_tbl_lock); | |
1404 | for (tp = &neigh_tables; *tp; tp = &(*tp)->next) { | |
1405 | if (*tp == tbl) { | |
1406 | *tp = tbl->next; | |
1407 | break; | |
1408 | } | |
1409 | } | |
1410 | write_unlock(&neigh_tbl_lock); | |
1411 | ||
1412 | neigh_hash_free(tbl->hash_buckets, tbl->hash_mask + 1); | |
1413 | tbl->hash_buckets = NULL; | |
1414 | ||
1415 | kfree(tbl->phash_buckets); | |
1416 | tbl->phash_buckets = NULL; | |
1417 | ||
1418 | return 0; | |
1419 | } | |
1420 | ||
1421 | int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1422 | { | |
1423 | struct ndmsg *ndm = NLMSG_DATA(nlh); | |
1424 | struct rtattr **nda = arg; | |
1425 | struct neigh_table *tbl; | |
1426 | struct net_device *dev = NULL; | |
1427 | int err = -ENODEV; | |
1428 | ||
1429 | if (ndm->ndm_ifindex && | |
1430 | (dev = dev_get_by_index(ndm->ndm_ifindex)) == NULL) | |
1431 | goto out; | |
1432 | ||
1433 | read_lock(&neigh_tbl_lock); | |
1434 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1435 | struct rtattr *dst_attr = nda[NDA_DST - 1]; | |
1436 | struct neighbour *n; | |
1437 | ||
1438 | if (tbl->family != ndm->ndm_family) | |
1439 | continue; | |
1440 | read_unlock(&neigh_tbl_lock); | |
1441 | ||
1442 | err = -EINVAL; | |
1443 | if (!dst_attr || RTA_PAYLOAD(dst_attr) < tbl->key_len) | |
1444 | goto out_dev_put; | |
1445 | ||
1446 | if (ndm->ndm_flags & NTF_PROXY) { | |
1447 | err = pneigh_delete(tbl, RTA_DATA(dst_attr), dev); | |
1448 | goto out_dev_put; | |
1449 | } | |
1450 | ||
1451 | if (!dev) | |
1452 | goto out; | |
1453 | ||
1454 | n = neigh_lookup(tbl, RTA_DATA(dst_attr), dev); | |
1455 | if (n) { | |
1456 | err = neigh_update(n, NULL, NUD_FAILED, | |
1457 | NEIGH_UPDATE_F_OVERRIDE| | |
1458 | NEIGH_UPDATE_F_ADMIN); | |
1459 | neigh_release(n); | |
1460 | } | |
1461 | goto out_dev_put; | |
1462 | } | |
1463 | read_unlock(&neigh_tbl_lock); | |
1464 | err = -EADDRNOTAVAIL; | |
1465 | out_dev_put: | |
1466 | if (dev) | |
1467 | dev_put(dev); | |
1468 | out: | |
1469 | return err; | |
1470 | } | |
1471 | ||
1472 | int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1473 | { | |
1474 | struct ndmsg *ndm = NLMSG_DATA(nlh); | |
1475 | struct rtattr **nda = arg; | |
1476 | struct neigh_table *tbl; | |
1477 | struct net_device *dev = NULL; | |
1478 | int err = -ENODEV; | |
1479 | ||
1480 | if (ndm->ndm_ifindex && | |
1481 | (dev = dev_get_by_index(ndm->ndm_ifindex)) == NULL) | |
1482 | goto out; | |
1483 | ||
1484 | read_lock(&neigh_tbl_lock); | |
1485 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1486 | struct rtattr *lladdr_attr = nda[NDA_LLADDR - 1]; | |
1487 | struct rtattr *dst_attr = nda[NDA_DST - 1]; | |
1488 | int override = 1; | |
1489 | struct neighbour *n; | |
1490 | ||
1491 | if (tbl->family != ndm->ndm_family) | |
1492 | continue; | |
1493 | read_unlock(&neigh_tbl_lock); | |
1494 | ||
1495 | err = -EINVAL; | |
1496 | if (!dst_attr || RTA_PAYLOAD(dst_attr) < tbl->key_len) | |
1497 | goto out_dev_put; | |
1498 | ||
1499 | if (ndm->ndm_flags & NTF_PROXY) { | |
1500 | err = -ENOBUFS; | |
1501 | if (pneigh_lookup(tbl, RTA_DATA(dst_attr), dev, 1)) | |
1502 | err = 0; | |
1503 | goto out_dev_put; | |
1504 | } | |
1505 | ||
1506 | err = -EINVAL; | |
1507 | if (!dev) | |
1508 | goto out; | |
1509 | if (lladdr_attr && RTA_PAYLOAD(lladdr_attr) < dev->addr_len) | |
1510 | goto out_dev_put; | |
1511 | ||
1512 | n = neigh_lookup(tbl, RTA_DATA(dst_attr), dev); | |
1513 | if (n) { | |
1514 | if (nlh->nlmsg_flags & NLM_F_EXCL) { | |
1515 | err = -EEXIST; | |
1516 | neigh_release(n); | |
1517 | goto out_dev_put; | |
1518 | } | |
1519 | ||
1520 | override = nlh->nlmsg_flags & NLM_F_REPLACE; | |
1521 | } else if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { | |
1522 | err = -ENOENT; | |
1523 | goto out_dev_put; | |
1524 | } else { | |
1525 | n = __neigh_lookup_errno(tbl, RTA_DATA(dst_attr), dev); | |
1526 | if (IS_ERR(n)) { | |
1527 | err = PTR_ERR(n); | |
1528 | goto out_dev_put; | |
1529 | } | |
1530 | } | |
1531 | ||
1532 | err = neigh_update(n, | |
1533 | lladdr_attr ? RTA_DATA(lladdr_attr) : NULL, | |
1534 | ndm->ndm_state, | |
1535 | (override ? NEIGH_UPDATE_F_OVERRIDE : 0) | | |
1536 | NEIGH_UPDATE_F_ADMIN); | |
1537 | ||
1538 | neigh_release(n); | |
1539 | goto out_dev_put; | |
1540 | } | |
1541 | ||
1542 | read_unlock(&neigh_tbl_lock); | |
1543 | err = -EADDRNOTAVAIL; | |
1544 | out_dev_put: | |
1545 | if (dev) | |
1546 | dev_put(dev); | |
1547 | out: | |
1548 | return err; | |
1549 | } | |
1550 | ||
c7fb64db TG |
1551 | static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms) |
1552 | { | |
e386c6eb TG |
1553 | struct rtattr *nest = NULL; |
1554 | ||
1555 | nest = RTA_NEST(skb, NDTA_PARMS); | |
c7fb64db TG |
1556 | |
1557 | if (parms->dev) | |
1558 | RTA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex); | |
1559 | ||
1560 | RTA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt)); | |
1561 | RTA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len); | |
1562 | RTA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen); | |
1563 | RTA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes); | |
1564 | RTA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes); | |
1565 | RTA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes); | |
1566 | RTA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time); | |
1567 | RTA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME, | |
1568 | parms->base_reachable_time); | |
1569 | RTA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime); | |
1570 | RTA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time); | |
1571 | RTA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time); | |
1572 | RTA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay); | |
1573 | RTA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay); | |
1574 | RTA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime); | |
1575 | ||
1576 | return RTA_NEST_END(skb, nest); | |
1577 | ||
1578 | rtattr_failure: | |
1579 | return RTA_NEST_CANCEL(skb, nest); | |
1580 | } | |
1581 | ||
1582 | static int neightbl_fill_info(struct neigh_table *tbl, struct sk_buff *skb, | |
1583 | struct netlink_callback *cb) | |
1584 | { | |
1585 | struct nlmsghdr *nlh; | |
1586 | struct ndtmsg *ndtmsg; | |
1587 | ||
1797754e TG |
1588 | nlh = NLMSG_NEW_ANSWER(skb, cb, RTM_NEWNEIGHTBL, sizeof(struct ndtmsg), |
1589 | NLM_F_MULTI); | |
c7fb64db | 1590 | |
4b6ea82d | 1591 | ndtmsg = NLMSG_DATA(nlh); |
c7fb64db TG |
1592 | |
1593 | read_lock_bh(&tbl->lock); | |
1594 | ndtmsg->ndtm_family = tbl->family; | |
9ef1d4c7 PM |
1595 | ndtmsg->ndtm_pad1 = 0; |
1596 | ndtmsg->ndtm_pad2 = 0; | |
c7fb64db TG |
1597 | |
1598 | RTA_PUT_STRING(skb, NDTA_NAME, tbl->id); | |
1599 | RTA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval); | |
1600 | RTA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1); | |
1601 | RTA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2); | |
1602 | RTA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3); | |
1603 | ||
1604 | { | |
1605 | unsigned long now = jiffies; | |
1606 | unsigned int flush_delta = now - tbl->last_flush; | |
1607 | unsigned int rand_delta = now - tbl->last_rand; | |
1608 | ||
1609 | struct ndt_config ndc = { | |
1610 | .ndtc_key_len = tbl->key_len, | |
1611 | .ndtc_entry_size = tbl->entry_size, | |
1612 | .ndtc_entries = atomic_read(&tbl->entries), | |
1613 | .ndtc_last_flush = jiffies_to_msecs(flush_delta), | |
1614 | .ndtc_last_rand = jiffies_to_msecs(rand_delta), | |
1615 | .ndtc_hash_rnd = tbl->hash_rnd, | |
1616 | .ndtc_hash_mask = tbl->hash_mask, | |
1617 | .ndtc_hash_chain_gc = tbl->hash_chain_gc, | |
1618 | .ndtc_proxy_qlen = tbl->proxy_queue.qlen, | |
1619 | }; | |
1620 | ||
1621 | RTA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc); | |
1622 | } | |
1623 | ||
1624 | { | |
1625 | int cpu; | |
1626 | struct ndt_stats ndst; | |
1627 | ||
1628 | memset(&ndst, 0, sizeof(ndst)); | |
1629 | ||
670c02c2 | 1630 | for_each_cpu(cpu) { |
c7fb64db TG |
1631 | struct neigh_statistics *st; |
1632 | ||
c7fb64db TG |
1633 | st = per_cpu_ptr(tbl->stats, cpu); |
1634 | ndst.ndts_allocs += st->allocs; | |
1635 | ndst.ndts_destroys += st->destroys; | |
1636 | ndst.ndts_hash_grows += st->hash_grows; | |
1637 | ndst.ndts_res_failed += st->res_failed; | |
1638 | ndst.ndts_lookups += st->lookups; | |
1639 | ndst.ndts_hits += st->hits; | |
1640 | ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast; | |
1641 | ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast; | |
1642 | ndst.ndts_periodic_gc_runs += st->periodic_gc_runs; | |
1643 | ndst.ndts_forced_gc_runs += st->forced_gc_runs; | |
1644 | } | |
1645 | ||
1646 | RTA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst); | |
1647 | } | |
1648 | ||
1649 | BUG_ON(tbl->parms.dev); | |
1650 | if (neightbl_fill_parms(skb, &tbl->parms) < 0) | |
1651 | goto rtattr_failure; | |
1652 | ||
1653 | read_unlock_bh(&tbl->lock); | |
1654 | return NLMSG_END(skb, nlh); | |
1655 | ||
1656 | rtattr_failure: | |
1657 | read_unlock_bh(&tbl->lock); | |
1658 | return NLMSG_CANCEL(skb, nlh); | |
1659 | ||
1660 | nlmsg_failure: | |
1661 | return -1; | |
1662 | } | |
1663 | ||
1664 | static int neightbl_fill_param_info(struct neigh_table *tbl, | |
1665 | struct neigh_parms *parms, | |
1666 | struct sk_buff *skb, | |
1667 | struct netlink_callback *cb) | |
1668 | { | |
1669 | struct ndtmsg *ndtmsg; | |
1670 | struct nlmsghdr *nlh; | |
1671 | ||
1797754e TG |
1672 | nlh = NLMSG_NEW_ANSWER(skb, cb, RTM_NEWNEIGHTBL, sizeof(struct ndtmsg), |
1673 | NLM_F_MULTI); | |
c7fb64db | 1674 | |
4b6ea82d | 1675 | ndtmsg = NLMSG_DATA(nlh); |
c7fb64db TG |
1676 | |
1677 | read_lock_bh(&tbl->lock); | |
1678 | ndtmsg->ndtm_family = tbl->family; | |
9ef1d4c7 PM |
1679 | ndtmsg->ndtm_pad1 = 0; |
1680 | ndtmsg->ndtm_pad2 = 0; | |
c7fb64db TG |
1681 | RTA_PUT_STRING(skb, NDTA_NAME, tbl->id); |
1682 | ||
1683 | if (neightbl_fill_parms(skb, parms) < 0) | |
1684 | goto rtattr_failure; | |
1685 | ||
1686 | read_unlock_bh(&tbl->lock); | |
1687 | return NLMSG_END(skb, nlh); | |
1688 | ||
1689 | rtattr_failure: | |
1690 | read_unlock_bh(&tbl->lock); | |
1691 | return NLMSG_CANCEL(skb, nlh); | |
1692 | ||
1693 | nlmsg_failure: | |
1694 | return -1; | |
1695 | } | |
1696 | ||
1697 | static inline struct neigh_parms *lookup_neigh_params(struct neigh_table *tbl, | |
1698 | int ifindex) | |
1699 | { | |
1700 | struct neigh_parms *p; | |
1701 | ||
1702 | for (p = &tbl->parms; p; p = p->next) | |
1703 | if ((p->dev && p->dev->ifindex == ifindex) || | |
1704 | (!p->dev && !ifindex)) | |
1705 | return p; | |
1706 | ||
1707 | return NULL; | |
1708 | } | |
1709 | ||
1710 | int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
1711 | { | |
1712 | struct neigh_table *tbl; | |
1713 | struct ndtmsg *ndtmsg = NLMSG_DATA(nlh); | |
1714 | struct rtattr **tb = arg; | |
1715 | int err = -EINVAL; | |
1716 | ||
1717 | if (!tb[NDTA_NAME - 1] || !RTA_PAYLOAD(tb[NDTA_NAME - 1])) | |
1718 | return -EINVAL; | |
1719 | ||
1720 | read_lock(&neigh_tbl_lock); | |
1721 | for (tbl = neigh_tables; tbl; tbl = tbl->next) { | |
1722 | if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family) | |
1723 | continue; | |
1724 | ||
1725 | if (!rtattr_strcmp(tb[NDTA_NAME - 1], tbl->id)) | |
1726 | break; | |
1727 | } | |
1728 | ||
1729 | if (tbl == NULL) { | |
1730 | err = -ENOENT; | |
1731 | goto errout; | |
1732 | } | |
1733 | ||
1734 | /* | |
1735 | * We acquire tbl->lock to be nice to the periodic timers and | |
1736 | * make sure they always see a consistent set of values. | |
1737 | */ | |
1738 | write_lock_bh(&tbl->lock); | |
1739 | ||
1740 | if (tb[NDTA_THRESH1 - 1]) | |
1741 | tbl->gc_thresh1 = RTA_GET_U32(tb[NDTA_THRESH1 - 1]); | |
1742 | ||
1743 | if (tb[NDTA_THRESH2 - 1]) | |
1744 | tbl->gc_thresh2 = RTA_GET_U32(tb[NDTA_THRESH2 - 1]); | |
1745 | ||
1746 | if (tb[NDTA_THRESH3 - 1]) | |
1747 | tbl->gc_thresh3 = RTA_GET_U32(tb[NDTA_THRESH3 - 1]); | |
1748 | ||
1749 | if (tb[NDTA_GC_INTERVAL - 1]) | |
1750 | tbl->gc_interval = RTA_GET_MSECS(tb[NDTA_GC_INTERVAL - 1]); | |
1751 | ||
1752 | if (tb[NDTA_PARMS - 1]) { | |
1753 | struct rtattr *tbp[NDTPA_MAX]; | |
1754 | struct neigh_parms *p; | |
1755 | u32 ifindex = 0; | |
1756 | ||
1757 | if (rtattr_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS - 1]) < 0) | |
1758 | goto rtattr_failure; | |
1759 | ||
1760 | if (tbp[NDTPA_IFINDEX - 1]) | |
1761 | ifindex = RTA_GET_U32(tbp[NDTPA_IFINDEX - 1]); | |
1762 | ||
1763 | p = lookup_neigh_params(tbl, ifindex); | |
1764 | if (p == NULL) { | |
1765 | err = -ENOENT; | |
1766 | goto rtattr_failure; | |
1767 | } | |
1768 | ||
1769 | if (tbp[NDTPA_QUEUE_LEN - 1]) | |
1770 | p->queue_len = RTA_GET_U32(tbp[NDTPA_QUEUE_LEN - 1]); | |
1771 | ||
1772 | if (tbp[NDTPA_PROXY_QLEN - 1]) | |
1773 | p->proxy_qlen = RTA_GET_U32(tbp[NDTPA_PROXY_QLEN - 1]); | |
1774 | ||
1775 | if (tbp[NDTPA_APP_PROBES - 1]) | |
1776 | p->app_probes = RTA_GET_U32(tbp[NDTPA_APP_PROBES - 1]); | |
1777 | ||
1778 | if (tbp[NDTPA_UCAST_PROBES - 1]) | |
1779 | p->ucast_probes = | |
1780 | RTA_GET_U32(tbp[NDTPA_UCAST_PROBES - 1]); | |
1781 | ||
1782 | if (tbp[NDTPA_MCAST_PROBES - 1]) | |
1783 | p->mcast_probes = | |
1784 | RTA_GET_U32(tbp[NDTPA_MCAST_PROBES - 1]); | |
1785 | ||
1786 | if (tbp[NDTPA_BASE_REACHABLE_TIME - 1]) | |
1787 | p->base_reachable_time = | |
1788 | RTA_GET_MSECS(tbp[NDTPA_BASE_REACHABLE_TIME - 1]); | |
1789 | ||
1790 | if (tbp[NDTPA_GC_STALETIME - 1]) | |
1791 | p->gc_staletime = | |
1792 | RTA_GET_MSECS(tbp[NDTPA_GC_STALETIME - 1]); | |
1793 | ||
1794 | if (tbp[NDTPA_DELAY_PROBE_TIME - 1]) | |
1795 | p->delay_probe_time = | |
1796 | RTA_GET_MSECS(tbp[NDTPA_DELAY_PROBE_TIME - 1]); | |
1797 | ||
1798 | if (tbp[NDTPA_RETRANS_TIME - 1]) | |
1799 | p->retrans_time = | |
1800 | RTA_GET_MSECS(tbp[NDTPA_RETRANS_TIME - 1]); | |
1801 | ||
1802 | if (tbp[NDTPA_ANYCAST_DELAY - 1]) | |
1803 | p->anycast_delay = | |
1804 | RTA_GET_MSECS(tbp[NDTPA_ANYCAST_DELAY - 1]); | |
1805 | ||
1806 | if (tbp[NDTPA_PROXY_DELAY - 1]) | |
1807 | p->proxy_delay = | |
1808 | RTA_GET_MSECS(tbp[NDTPA_PROXY_DELAY - 1]); | |
1809 | ||
1810 | if (tbp[NDTPA_LOCKTIME - 1]) | |
1811 | p->locktime = RTA_GET_MSECS(tbp[NDTPA_LOCKTIME - 1]); | |
1812 | } | |
1813 | ||
1814 | err = 0; | |
1815 | ||
1816 | rtattr_failure: | |
1817 | write_unlock_bh(&tbl->lock); | |
1818 | errout: | |
1819 | read_unlock(&neigh_tbl_lock); | |
1820 | return err; | |
1821 | } | |
1822 | ||
1823 | int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb) | |
1824 | { | |
1825 | int idx, family; | |
1826 | int s_idx = cb->args[0]; | |
1827 | struct neigh_table *tbl; | |
1828 | ||
1829 | family = ((struct rtgenmsg *)NLMSG_DATA(cb->nlh))->rtgen_family; | |
1830 | ||
1831 | read_lock(&neigh_tbl_lock); | |
1832 | for (tbl = neigh_tables, idx = 0; tbl; tbl = tbl->next) { | |
1833 | struct neigh_parms *p; | |
1834 | ||
1835 | if (idx < s_idx || (family && tbl->family != family)) | |
1836 | continue; | |
1837 | ||
1838 | if (neightbl_fill_info(tbl, skb, cb) <= 0) | |
1839 | break; | |
1840 | ||
1841 | for (++idx, p = tbl->parms.next; p; p = p->next, idx++) { | |
1842 | if (idx < s_idx) | |
1843 | continue; | |
1844 | ||
1845 | if (neightbl_fill_param_info(tbl, p, skb, cb) <= 0) | |
1846 | goto out; | |
1847 | } | |
1848 | ||
1849 | } | |
1850 | out: | |
1851 | read_unlock(&neigh_tbl_lock); | |
1852 | cb->args[0] = idx; | |
1853 | ||
1854 | return skb->len; | |
1855 | } | |
1da177e4 LT |
1856 | |
1857 | static int neigh_fill_info(struct sk_buff *skb, struct neighbour *n, | |
b6544c0b | 1858 | u32 pid, u32 seq, int event, unsigned int flags) |
1da177e4 LT |
1859 | { |
1860 | unsigned long now = jiffies; | |
1861 | unsigned char *b = skb->tail; | |
1862 | struct nda_cacheinfo ci; | |
1863 | int locked = 0; | |
1864 | u32 probes; | |
b6544c0b JHS |
1865 | struct nlmsghdr *nlh = NLMSG_NEW(skb, pid, seq, event, |
1866 | sizeof(struct ndmsg), flags); | |
1da177e4 LT |
1867 | struct ndmsg *ndm = NLMSG_DATA(nlh); |
1868 | ||
1da177e4 | 1869 | ndm->ndm_family = n->ops->family; |
9ef1d4c7 PM |
1870 | ndm->ndm_pad1 = 0; |
1871 | ndm->ndm_pad2 = 0; | |
1da177e4 LT |
1872 | ndm->ndm_flags = n->flags; |
1873 | ndm->ndm_type = n->type; | |
1874 | ndm->ndm_ifindex = n->dev->ifindex; | |
1875 | RTA_PUT(skb, NDA_DST, n->tbl->key_len, n->primary_key); | |
1876 | read_lock_bh(&n->lock); | |
1877 | locked = 1; | |
1878 | ndm->ndm_state = n->nud_state; | |
1879 | if (n->nud_state & NUD_VALID) | |
1880 | RTA_PUT(skb, NDA_LLADDR, n->dev->addr_len, n->ha); | |
1881 | ci.ndm_used = now - n->used; | |
1882 | ci.ndm_confirmed = now - n->confirmed; | |
1883 | ci.ndm_updated = now - n->updated; | |
1884 | ci.ndm_refcnt = atomic_read(&n->refcnt) - 1; | |
1885 | probes = atomic_read(&n->probes); | |
1886 | read_unlock_bh(&n->lock); | |
1887 | locked = 0; | |
1888 | RTA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci); | |
1889 | RTA_PUT(skb, NDA_PROBES, sizeof(probes), &probes); | |
1890 | nlh->nlmsg_len = skb->tail - b; | |
1891 | return skb->len; | |
1892 | ||
1893 | nlmsg_failure: | |
1894 | rtattr_failure: | |
1895 | if (locked) | |
1896 | read_unlock_bh(&n->lock); | |
1897 | skb_trim(skb, b - skb->data); | |
1898 | return -1; | |
1899 | } | |
1900 | ||
1901 | ||
1902 | static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb, | |
1903 | struct netlink_callback *cb) | |
1904 | { | |
1905 | struct neighbour *n; | |
1906 | int rc, h, s_h = cb->args[1]; | |
1907 | int idx, s_idx = idx = cb->args[2]; | |
1908 | ||
1909 | for (h = 0; h <= tbl->hash_mask; h++) { | |
1910 | if (h < s_h) | |
1911 | continue; | |
1912 | if (h > s_h) | |
1913 | s_idx = 0; | |
1914 | read_lock_bh(&tbl->lock); | |
1915 | for (n = tbl->hash_buckets[h], idx = 0; n; n = n->next, idx++) { | |
1916 | if (idx < s_idx) | |
1917 | continue; | |
1918 | if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid, | |
1919 | cb->nlh->nlmsg_seq, | |
b6544c0b JHS |
1920 | RTM_NEWNEIGH, |
1921 | NLM_F_MULTI) <= 0) { | |
1da177e4 LT |
1922 | read_unlock_bh(&tbl->lock); |
1923 | rc = -1; | |
1924 | goto out; | |
1925 | } | |
1926 | } | |
1927 | read_unlock_bh(&tbl->lock); | |
1928 | } | |
1929 | rc = skb->len; | |
1930 | out: | |
1931 | cb->args[1] = h; | |
1932 | cb->args[2] = idx; | |
1933 | return rc; | |
1934 | } | |
1935 | ||
1936 | int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb) | |
1937 | { | |
1938 | struct neigh_table *tbl; | |
1939 | int t, family, s_t; | |
1940 | ||
1941 | read_lock(&neigh_tbl_lock); | |
1942 | family = ((struct rtgenmsg *)NLMSG_DATA(cb->nlh))->rtgen_family; | |
1943 | s_t = cb->args[0]; | |
1944 | ||
1945 | for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) { | |
1946 | if (t < s_t || (family && tbl->family != family)) | |
1947 | continue; | |
1948 | if (t > s_t) | |
1949 | memset(&cb->args[1], 0, sizeof(cb->args) - | |
1950 | sizeof(cb->args[0])); | |
1951 | if (neigh_dump_table(tbl, skb, cb) < 0) | |
1952 | break; | |
1953 | } | |
1954 | read_unlock(&neigh_tbl_lock); | |
1955 | ||
1956 | cb->args[0] = t; | |
1957 | return skb->len; | |
1958 | } | |
1959 | ||
1960 | void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie) | |
1961 | { | |
1962 | int chain; | |
1963 | ||
1964 | read_lock_bh(&tbl->lock); | |
1965 | for (chain = 0; chain <= tbl->hash_mask; chain++) { | |
1966 | struct neighbour *n; | |
1967 | ||
1968 | for (n = tbl->hash_buckets[chain]; n; n = n->next) | |
1969 | cb(n, cookie); | |
1970 | } | |
1971 | read_unlock_bh(&tbl->lock); | |
1972 | } | |
1973 | EXPORT_SYMBOL(neigh_for_each); | |
1974 | ||
1975 | /* The tbl->lock must be held as a writer and BH disabled. */ | |
1976 | void __neigh_for_each_release(struct neigh_table *tbl, | |
1977 | int (*cb)(struct neighbour *)) | |
1978 | { | |
1979 | int chain; | |
1980 | ||
1981 | for (chain = 0; chain <= tbl->hash_mask; chain++) { | |
1982 | struct neighbour *n, **np; | |
1983 | ||
1984 | np = &tbl->hash_buckets[chain]; | |
1985 | while ((n = *np) != NULL) { | |
1986 | int release; | |
1987 | ||
1988 | write_lock(&n->lock); | |
1989 | release = cb(n); | |
1990 | if (release) { | |
1991 | *np = n->next; | |
1992 | n->dead = 1; | |
1993 | } else | |
1994 | np = &n->next; | |
1995 | write_unlock(&n->lock); | |
1996 | if (release) | |
1997 | neigh_release(n); | |
1998 | } | |
1999 | } | |
2000 | } | |
2001 | EXPORT_SYMBOL(__neigh_for_each_release); | |
2002 | ||
2003 | #ifdef CONFIG_PROC_FS | |
2004 | ||
2005 | static struct neighbour *neigh_get_first(struct seq_file *seq) | |
2006 | { | |
2007 | struct neigh_seq_state *state = seq->private; | |
2008 | struct neigh_table *tbl = state->tbl; | |
2009 | struct neighbour *n = NULL; | |
2010 | int bucket = state->bucket; | |
2011 | ||
2012 | state->flags &= ~NEIGH_SEQ_IS_PNEIGH; | |
2013 | for (bucket = 0; bucket <= tbl->hash_mask; bucket++) { | |
2014 | n = tbl->hash_buckets[bucket]; | |
2015 | ||
2016 | while (n) { | |
2017 | if (state->neigh_sub_iter) { | |
2018 | loff_t fakep = 0; | |
2019 | void *v; | |
2020 | ||
2021 | v = state->neigh_sub_iter(state, n, &fakep); | |
2022 | if (!v) | |
2023 | goto next; | |
2024 | } | |
2025 | if (!(state->flags & NEIGH_SEQ_SKIP_NOARP)) | |
2026 | break; | |
2027 | if (n->nud_state & ~NUD_NOARP) | |
2028 | break; | |
2029 | next: | |
2030 | n = n->next; | |
2031 | } | |
2032 | ||
2033 | if (n) | |
2034 | break; | |
2035 | } | |
2036 | state->bucket = bucket; | |
2037 | ||
2038 | return n; | |
2039 | } | |
2040 | ||
2041 | static struct neighbour *neigh_get_next(struct seq_file *seq, | |
2042 | struct neighbour *n, | |
2043 | loff_t *pos) | |
2044 | { | |
2045 | struct neigh_seq_state *state = seq->private; | |
2046 | struct neigh_table *tbl = state->tbl; | |
2047 | ||
2048 | if (state->neigh_sub_iter) { | |
2049 | void *v = state->neigh_sub_iter(state, n, pos); | |
2050 | if (v) | |
2051 | return n; | |
2052 | } | |
2053 | n = n->next; | |
2054 | ||
2055 | while (1) { | |
2056 | while (n) { | |
2057 | if (state->neigh_sub_iter) { | |
2058 | void *v = state->neigh_sub_iter(state, n, pos); | |
2059 | if (v) | |
2060 | return n; | |
2061 | goto next; | |
2062 | } | |
2063 | if (!(state->flags & NEIGH_SEQ_SKIP_NOARP)) | |
2064 | break; | |
2065 | ||
2066 | if (n->nud_state & ~NUD_NOARP) | |
2067 | break; | |
2068 | next: | |
2069 | n = n->next; | |
2070 | } | |
2071 | ||
2072 | if (n) | |
2073 | break; | |
2074 | ||
2075 | if (++state->bucket > tbl->hash_mask) | |
2076 | break; | |
2077 | ||
2078 | n = tbl->hash_buckets[state->bucket]; | |
2079 | } | |
2080 | ||
2081 | if (n && pos) | |
2082 | --(*pos); | |
2083 | return n; | |
2084 | } | |
2085 | ||
2086 | static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos) | |
2087 | { | |
2088 | struct neighbour *n = neigh_get_first(seq); | |
2089 | ||
2090 | if (n) { | |
2091 | while (*pos) { | |
2092 | n = neigh_get_next(seq, n, pos); | |
2093 | if (!n) | |
2094 | break; | |
2095 | } | |
2096 | } | |
2097 | return *pos ? NULL : n; | |
2098 | } | |
2099 | ||
2100 | static struct pneigh_entry *pneigh_get_first(struct seq_file *seq) | |
2101 | { | |
2102 | struct neigh_seq_state *state = seq->private; | |
2103 | struct neigh_table *tbl = state->tbl; | |
2104 | struct pneigh_entry *pn = NULL; | |
2105 | int bucket = state->bucket; | |
2106 | ||
2107 | state->flags |= NEIGH_SEQ_IS_PNEIGH; | |
2108 | for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) { | |
2109 | pn = tbl->phash_buckets[bucket]; | |
2110 | if (pn) | |
2111 | break; | |
2112 | } | |
2113 | state->bucket = bucket; | |
2114 | ||
2115 | return pn; | |
2116 | } | |
2117 | ||
2118 | static struct pneigh_entry *pneigh_get_next(struct seq_file *seq, | |
2119 | struct pneigh_entry *pn, | |
2120 | loff_t *pos) | |
2121 | { | |
2122 | struct neigh_seq_state *state = seq->private; | |
2123 | struct neigh_table *tbl = state->tbl; | |
2124 | ||
2125 | pn = pn->next; | |
2126 | while (!pn) { | |
2127 | if (++state->bucket > PNEIGH_HASHMASK) | |
2128 | break; | |
2129 | pn = tbl->phash_buckets[state->bucket]; | |
2130 | if (pn) | |
2131 | break; | |
2132 | } | |
2133 | ||
2134 | if (pn && pos) | |
2135 | --(*pos); | |
2136 | ||
2137 | return pn; | |
2138 | } | |
2139 | ||
2140 | static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos) | |
2141 | { | |
2142 | struct pneigh_entry *pn = pneigh_get_first(seq); | |
2143 | ||
2144 | if (pn) { | |
2145 | while (*pos) { | |
2146 | pn = pneigh_get_next(seq, pn, pos); | |
2147 | if (!pn) | |
2148 | break; | |
2149 | } | |
2150 | } | |
2151 | return *pos ? NULL : pn; | |
2152 | } | |
2153 | ||
2154 | static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos) | |
2155 | { | |
2156 | struct neigh_seq_state *state = seq->private; | |
2157 | void *rc; | |
2158 | ||
2159 | rc = neigh_get_idx(seq, pos); | |
2160 | if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY)) | |
2161 | rc = pneigh_get_idx(seq, pos); | |
2162 | ||
2163 | return rc; | |
2164 | } | |
2165 | ||
2166 | void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags) | |
2167 | { | |
2168 | struct neigh_seq_state *state = seq->private; | |
2169 | loff_t pos_minus_one; | |
2170 | ||
2171 | state->tbl = tbl; | |
2172 | state->bucket = 0; | |
2173 | state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH); | |
2174 | ||
2175 | read_lock_bh(&tbl->lock); | |
2176 | ||
2177 | pos_minus_one = *pos - 1; | |
2178 | return *pos ? neigh_get_idx_any(seq, &pos_minus_one) : SEQ_START_TOKEN; | |
2179 | } | |
2180 | EXPORT_SYMBOL(neigh_seq_start); | |
2181 | ||
2182 | void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2183 | { | |
2184 | struct neigh_seq_state *state; | |
2185 | void *rc; | |
2186 | ||
2187 | if (v == SEQ_START_TOKEN) { | |
2188 | rc = neigh_get_idx(seq, pos); | |
2189 | goto out; | |
2190 | } | |
2191 | ||
2192 | state = seq->private; | |
2193 | if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) { | |
2194 | rc = neigh_get_next(seq, v, NULL); | |
2195 | if (rc) | |
2196 | goto out; | |
2197 | if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY)) | |
2198 | rc = pneigh_get_first(seq); | |
2199 | } else { | |
2200 | BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY); | |
2201 | rc = pneigh_get_next(seq, v, NULL); | |
2202 | } | |
2203 | out: | |
2204 | ++(*pos); | |
2205 | return rc; | |
2206 | } | |
2207 | EXPORT_SYMBOL(neigh_seq_next); | |
2208 | ||
2209 | void neigh_seq_stop(struct seq_file *seq, void *v) | |
2210 | { | |
2211 | struct neigh_seq_state *state = seq->private; | |
2212 | struct neigh_table *tbl = state->tbl; | |
2213 | ||
2214 | read_unlock_bh(&tbl->lock); | |
2215 | } | |
2216 | EXPORT_SYMBOL(neigh_seq_stop); | |
2217 | ||
2218 | /* statistics via seq_file */ | |
2219 | ||
2220 | static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos) | |
2221 | { | |
2222 | struct proc_dir_entry *pde = seq->private; | |
2223 | struct neigh_table *tbl = pde->data; | |
2224 | int cpu; | |
2225 | ||
2226 | if (*pos == 0) | |
2227 | return SEQ_START_TOKEN; | |
2228 | ||
2229 | for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) { | |
2230 | if (!cpu_possible(cpu)) | |
2231 | continue; | |
2232 | *pos = cpu+1; | |
2233 | return per_cpu_ptr(tbl->stats, cpu); | |
2234 | } | |
2235 | return NULL; | |
2236 | } | |
2237 | ||
2238 | static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2239 | { | |
2240 | struct proc_dir_entry *pde = seq->private; | |
2241 | struct neigh_table *tbl = pde->data; | |
2242 | int cpu; | |
2243 | ||
2244 | for (cpu = *pos; cpu < NR_CPUS; ++cpu) { | |
2245 | if (!cpu_possible(cpu)) | |
2246 | continue; | |
2247 | *pos = cpu+1; | |
2248 | return per_cpu_ptr(tbl->stats, cpu); | |
2249 | } | |
2250 | return NULL; | |
2251 | } | |
2252 | ||
2253 | static void neigh_stat_seq_stop(struct seq_file *seq, void *v) | |
2254 | { | |
2255 | ||
2256 | } | |
2257 | ||
2258 | static int neigh_stat_seq_show(struct seq_file *seq, void *v) | |
2259 | { | |
2260 | struct proc_dir_entry *pde = seq->private; | |
2261 | struct neigh_table *tbl = pde->data; | |
2262 | struct neigh_statistics *st = v; | |
2263 | ||
2264 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 2265 | 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 |
2266 | return 0; |
2267 | } | |
2268 | ||
2269 | seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx " | |
2270 | "%08lx %08lx %08lx %08lx\n", | |
2271 | atomic_read(&tbl->entries), | |
2272 | ||
2273 | st->allocs, | |
2274 | st->destroys, | |
2275 | st->hash_grows, | |
2276 | ||
2277 | st->lookups, | |
2278 | st->hits, | |
2279 | ||
2280 | st->res_failed, | |
2281 | ||
2282 | st->rcv_probes_mcast, | |
2283 | st->rcv_probes_ucast, | |
2284 | ||
2285 | st->periodic_gc_runs, | |
2286 | st->forced_gc_runs | |
2287 | ); | |
2288 | ||
2289 | return 0; | |
2290 | } | |
2291 | ||
2292 | static struct seq_operations neigh_stat_seq_ops = { | |
2293 | .start = neigh_stat_seq_start, | |
2294 | .next = neigh_stat_seq_next, | |
2295 | .stop = neigh_stat_seq_stop, | |
2296 | .show = neigh_stat_seq_show, | |
2297 | }; | |
2298 | ||
2299 | static int neigh_stat_seq_open(struct inode *inode, struct file *file) | |
2300 | { | |
2301 | int ret = seq_open(file, &neigh_stat_seq_ops); | |
2302 | ||
2303 | if (!ret) { | |
2304 | struct seq_file *sf = file->private_data; | |
2305 | sf->private = PDE(inode); | |
2306 | } | |
2307 | return ret; | |
2308 | }; | |
2309 | ||
2310 | static struct file_operations neigh_stat_seq_fops = { | |
2311 | .owner = THIS_MODULE, | |
2312 | .open = neigh_stat_seq_open, | |
2313 | .read = seq_read, | |
2314 | .llseek = seq_lseek, | |
2315 | .release = seq_release, | |
2316 | }; | |
2317 | ||
2318 | #endif /* CONFIG_PROC_FS */ | |
2319 | ||
2320 | #ifdef CONFIG_ARPD | |
2321 | void neigh_app_ns(struct neighbour *n) | |
2322 | { | |
2323 | struct nlmsghdr *nlh; | |
2324 | int size = NLMSG_SPACE(sizeof(struct ndmsg) + 256); | |
2325 | struct sk_buff *skb = alloc_skb(size, GFP_ATOMIC); | |
2326 | ||
2327 | if (!skb) | |
2328 | return; | |
2329 | ||
b6544c0b | 2330 | if (neigh_fill_info(skb, n, 0, 0, RTM_GETNEIGH, 0) < 0) { |
1da177e4 LT |
2331 | kfree_skb(skb); |
2332 | return; | |
2333 | } | |
2334 | nlh = (struct nlmsghdr *)skb->data; | |
2335 | nlh->nlmsg_flags = NLM_F_REQUEST; | |
ac6d439d PM |
2336 | NETLINK_CB(skb).dst_group = RTNLGRP_NEIGH; |
2337 | netlink_broadcast(rtnl, skb, 0, RTNLGRP_NEIGH, GFP_ATOMIC); | |
1da177e4 LT |
2338 | } |
2339 | ||
2340 | static void neigh_app_notify(struct neighbour *n) | |
2341 | { | |
2342 | struct nlmsghdr *nlh; | |
2343 | int size = NLMSG_SPACE(sizeof(struct ndmsg) + 256); | |
2344 | struct sk_buff *skb = alloc_skb(size, GFP_ATOMIC); | |
2345 | ||
2346 | if (!skb) | |
2347 | return; | |
2348 | ||
b6544c0b | 2349 | if (neigh_fill_info(skb, n, 0, 0, RTM_NEWNEIGH, 0) < 0) { |
1da177e4 LT |
2350 | kfree_skb(skb); |
2351 | return; | |
2352 | } | |
2353 | nlh = (struct nlmsghdr *)skb->data; | |
ac6d439d PM |
2354 | NETLINK_CB(skb).dst_group = RTNLGRP_NEIGH; |
2355 | netlink_broadcast(rtnl, skb, 0, RTNLGRP_NEIGH, GFP_ATOMIC); | |
1da177e4 LT |
2356 | } |
2357 | ||
2358 | #endif /* CONFIG_ARPD */ | |
2359 | ||
2360 | #ifdef CONFIG_SYSCTL | |
2361 | ||
2362 | static struct neigh_sysctl_table { | |
2363 | struct ctl_table_header *sysctl_header; | |
2364 | ctl_table neigh_vars[__NET_NEIGH_MAX]; | |
2365 | ctl_table neigh_dev[2]; | |
2366 | ctl_table neigh_neigh_dir[2]; | |
2367 | ctl_table neigh_proto_dir[2]; | |
2368 | ctl_table neigh_root_dir[2]; | |
2369 | } neigh_sysctl_template = { | |
2370 | .neigh_vars = { | |
2371 | { | |
2372 | .ctl_name = NET_NEIGH_MCAST_SOLICIT, | |
2373 | .procname = "mcast_solicit", | |
2374 | .maxlen = sizeof(int), | |
2375 | .mode = 0644, | |
2376 | .proc_handler = &proc_dointvec, | |
2377 | }, | |
2378 | { | |
2379 | .ctl_name = NET_NEIGH_UCAST_SOLICIT, | |
2380 | .procname = "ucast_solicit", | |
2381 | .maxlen = sizeof(int), | |
2382 | .mode = 0644, | |
2383 | .proc_handler = &proc_dointvec, | |
2384 | }, | |
2385 | { | |
2386 | .ctl_name = NET_NEIGH_APP_SOLICIT, | |
2387 | .procname = "app_solicit", | |
2388 | .maxlen = sizeof(int), | |
2389 | .mode = 0644, | |
2390 | .proc_handler = &proc_dointvec, | |
2391 | }, | |
2392 | { | |
2393 | .ctl_name = NET_NEIGH_RETRANS_TIME, | |
2394 | .procname = "retrans_time", | |
2395 | .maxlen = sizeof(int), | |
2396 | .mode = 0644, | |
2397 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2398 | }, | |
2399 | { | |
2400 | .ctl_name = NET_NEIGH_REACHABLE_TIME, | |
2401 | .procname = "base_reachable_time", | |
2402 | .maxlen = sizeof(int), | |
2403 | .mode = 0644, | |
2404 | .proc_handler = &proc_dointvec_jiffies, | |
2405 | .strategy = &sysctl_jiffies, | |
2406 | }, | |
2407 | { | |
2408 | .ctl_name = NET_NEIGH_DELAY_PROBE_TIME, | |
2409 | .procname = "delay_first_probe_time", | |
2410 | .maxlen = sizeof(int), | |
2411 | .mode = 0644, | |
2412 | .proc_handler = &proc_dointvec_jiffies, | |
2413 | .strategy = &sysctl_jiffies, | |
2414 | }, | |
2415 | { | |
2416 | .ctl_name = NET_NEIGH_GC_STALE_TIME, | |
2417 | .procname = "gc_stale_time", | |
2418 | .maxlen = sizeof(int), | |
2419 | .mode = 0644, | |
2420 | .proc_handler = &proc_dointvec_jiffies, | |
2421 | .strategy = &sysctl_jiffies, | |
2422 | }, | |
2423 | { | |
2424 | .ctl_name = NET_NEIGH_UNRES_QLEN, | |
2425 | .procname = "unres_qlen", | |
2426 | .maxlen = sizeof(int), | |
2427 | .mode = 0644, | |
2428 | .proc_handler = &proc_dointvec, | |
2429 | }, | |
2430 | { | |
2431 | .ctl_name = NET_NEIGH_PROXY_QLEN, | |
2432 | .procname = "proxy_qlen", | |
2433 | .maxlen = sizeof(int), | |
2434 | .mode = 0644, | |
2435 | .proc_handler = &proc_dointvec, | |
2436 | }, | |
2437 | { | |
2438 | .ctl_name = NET_NEIGH_ANYCAST_DELAY, | |
2439 | .procname = "anycast_delay", | |
2440 | .maxlen = sizeof(int), | |
2441 | .mode = 0644, | |
2442 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2443 | }, | |
2444 | { | |
2445 | .ctl_name = NET_NEIGH_PROXY_DELAY, | |
2446 | .procname = "proxy_delay", | |
2447 | .maxlen = sizeof(int), | |
2448 | .mode = 0644, | |
2449 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2450 | }, | |
2451 | { | |
2452 | .ctl_name = NET_NEIGH_LOCKTIME, | |
2453 | .procname = "locktime", | |
2454 | .maxlen = sizeof(int), | |
2455 | .mode = 0644, | |
2456 | .proc_handler = &proc_dointvec_userhz_jiffies, | |
2457 | }, | |
2458 | { | |
2459 | .ctl_name = NET_NEIGH_GC_INTERVAL, | |
2460 | .procname = "gc_interval", | |
2461 | .maxlen = sizeof(int), | |
2462 | .mode = 0644, | |
2463 | .proc_handler = &proc_dointvec_jiffies, | |
2464 | .strategy = &sysctl_jiffies, | |
2465 | }, | |
2466 | { | |
2467 | .ctl_name = NET_NEIGH_GC_THRESH1, | |
2468 | .procname = "gc_thresh1", | |
2469 | .maxlen = sizeof(int), | |
2470 | .mode = 0644, | |
2471 | .proc_handler = &proc_dointvec, | |
2472 | }, | |
2473 | { | |
2474 | .ctl_name = NET_NEIGH_GC_THRESH2, | |
2475 | .procname = "gc_thresh2", | |
2476 | .maxlen = sizeof(int), | |
2477 | .mode = 0644, | |
2478 | .proc_handler = &proc_dointvec, | |
2479 | }, | |
2480 | { | |
2481 | .ctl_name = NET_NEIGH_GC_THRESH3, | |
2482 | .procname = "gc_thresh3", | |
2483 | .maxlen = sizeof(int), | |
2484 | .mode = 0644, | |
2485 | .proc_handler = &proc_dointvec, | |
2486 | }, | |
2487 | { | |
2488 | .ctl_name = NET_NEIGH_RETRANS_TIME_MS, | |
2489 | .procname = "retrans_time_ms", | |
2490 | .maxlen = sizeof(int), | |
2491 | .mode = 0644, | |
2492 | .proc_handler = &proc_dointvec_ms_jiffies, | |
2493 | .strategy = &sysctl_ms_jiffies, | |
2494 | }, | |
2495 | { | |
2496 | .ctl_name = NET_NEIGH_REACHABLE_TIME_MS, | |
2497 | .procname = "base_reachable_time_ms", | |
2498 | .maxlen = sizeof(int), | |
2499 | .mode = 0644, | |
2500 | .proc_handler = &proc_dointvec_ms_jiffies, | |
2501 | .strategy = &sysctl_ms_jiffies, | |
2502 | }, | |
2503 | }, | |
2504 | .neigh_dev = { | |
2505 | { | |
2506 | .ctl_name = NET_PROTO_CONF_DEFAULT, | |
2507 | .procname = "default", | |
2508 | .mode = 0555, | |
2509 | }, | |
2510 | }, | |
2511 | .neigh_neigh_dir = { | |
2512 | { | |
2513 | .procname = "neigh", | |
2514 | .mode = 0555, | |
2515 | }, | |
2516 | }, | |
2517 | .neigh_proto_dir = { | |
2518 | { | |
2519 | .mode = 0555, | |
2520 | }, | |
2521 | }, | |
2522 | .neigh_root_dir = { | |
2523 | { | |
2524 | .ctl_name = CTL_NET, | |
2525 | .procname = "net", | |
2526 | .mode = 0555, | |
2527 | }, | |
2528 | }, | |
2529 | }; | |
2530 | ||
2531 | int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p, | |
2532 | int p_id, int pdev_id, char *p_name, | |
2533 | proc_handler *handler, ctl_handler *strategy) | |
2534 | { | |
2535 | struct neigh_sysctl_table *t = kmalloc(sizeof(*t), GFP_KERNEL); | |
2536 | const char *dev_name_source = NULL; | |
2537 | char *dev_name = NULL; | |
2538 | int err = 0; | |
2539 | ||
2540 | if (!t) | |
2541 | return -ENOBUFS; | |
2542 | memcpy(t, &neigh_sysctl_template, sizeof(*t)); | |
2543 | t->neigh_vars[0].data = &p->mcast_probes; | |
2544 | t->neigh_vars[1].data = &p->ucast_probes; | |
2545 | t->neigh_vars[2].data = &p->app_probes; | |
2546 | t->neigh_vars[3].data = &p->retrans_time; | |
2547 | t->neigh_vars[4].data = &p->base_reachable_time; | |
2548 | t->neigh_vars[5].data = &p->delay_probe_time; | |
2549 | t->neigh_vars[6].data = &p->gc_staletime; | |
2550 | t->neigh_vars[7].data = &p->queue_len; | |
2551 | t->neigh_vars[8].data = &p->proxy_qlen; | |
2552 | t->neigh_vars[9].data = &p->anycast_delay; | |
2553 | t->neigh_vars[10].data = &p->proxy_delay; | |
2554 | t->neigh_vars[11].data = &p->locktime; | |
2555 | ||
2556 | if (dev) { | |
2557 | dev_name_source = dev->name; | |
2558 | t->neigh_dev[0].ctl_name = dev->ifindex; | |
2559 | t->neigh_vars[12].procname = NULL; | |
2560 | t->neigh_vars[13].procname = NULL; | |
2561 | t->neigh_vars[14].procname = NULL; | |
2562 | t->neigh_vars[15].procname = NULL; | |
2563 | } else { | |
2564 | dev_name_source = t->neigh_dev[0].procname; | |
2565 | t->neigh_vars[12].data = (int *)(p + 1); | |
2566 | t->neigh_vars[13].data = (int *)(p + 1) + 1; | |
2567 | t->neigh_vars[14].data = (int *)(p + 1) + 2; | |
2568 | t->neigh_vars[15].data = (int *)(p + 1) + 3; | |
2569 | } | |
2570 | ||
2571 | t->neigh_vars[16].data = &p->retrans_time; | |
2572 | t->neigh_vars[17].data = &p->base_reachable_time; | |
2573 | ||
2574 | if (handler || strategy) { | |
2575 | /* RetransTime */ | |
2576 | t->neigh_vars[3].proc_handler = handler; | |
2577 | t->neigh_vars[3].strategy = strategy; | |
2578 | t->neigh_vars[3].extra1 = dev; | |
2579 | /* ReachableTime */ | |
2580 | t->neigh_vars[4].proc_handler = handler; | |
2581 | t->neigh_vars[4].strategy = strategy; | |
2582 | t->neigh_vars[4].extra1 = dev; | |
2583 | /* RetransTime (in milliseconds)*/ | |
2584 | t->neigh_vars[16].proc_handler = handler; | |
2585 | t->neigh_vars[16].strategy = strategy; | |
2586 | t->neigh_vars[16].extra1 = dev; | |
2587 | /* ReachableTime (in milliseconds) */ | |
2588 | t->neigh_vars[17].proc_handler = handler; | |
2589 | t->neigh_vars[17].strategy = strategy; | |
2590 | t->neigh_vars[17].extra1 = dev; | |
2591 | } | |
2592 | ||
543537bd | 2593 | dev_name = kstrdup(dev_name_source, GFP_KERNEL); |
1da177e4 LT |
2594 | if (!dev_name) { |
2595 | err = -ENOBUFS; | |
2596 | goto free; | |
2597 | } | |
2598 | ||
2599 | t->neigh_dev[0].procname = dev_name; | |
2600 | ||
2601 | t->neigh_neigh_dir[0].ctl_name = pdev_id; | |
2602 | ||
2603 | t->neigh_proto_dir[0].procname = p_name; | |
2604 | t->neigh_proto_dir[0].ctl_name = p_id; | |
2605 | ||
2606 | t->neigh_dev[0].child = t->neigh_vars; | |
2607 | t->neigh_neigh_dir[0].child = t->neigh_dev; | |
2608 | t->neigh_proto_dir[0].child = t->neigh_neigh_dir; | |
2609 | t->neigh_root_dir[0].child = t->neigh_proto_dir; | |
2610 | ||
2611 | t->sysctl_header = register_sysctl_table(t->neigh_root_dir, 0); | |
2612 | if (!t->sysctl_header) { | |
2613 | err = -ENOBUFS; | |
2614 | goto free_procname; | |
2615 | } | |
2616 | p->sysctl_table = t; | |
2617 | return 0; | |
2618 | ||
2619 | /* error path */ | |
2620 | free_procname: | |
2621 | kfree(dev_name); | |
2622 | free: | |
2623 | kfree(t); | |
2624 | ||
2625 | return err; | |
2626 | } | |
2627 | ||
2628 | void neigh_sysctl_unregister(struct neigh_parms *p) | |
2629 | { | |
2630 | if (p->sysctl_table) { | |
2631 | struct neigh_sysctl_table *t = p->sysctl_table; | |
2632 | p->sysctl_table = NULL; | |
2633 | unregister_sysctl_table(t->sysctl_header); | |
2634 | kfree(t->neigh_dev[0].procname); | |
2635 | kfree(t); | |
2636 | } | |
2637 | } | |
2638 | ||
2639 | #endif /* CONFIG_SYSCTL */ | |
2640 | ||
2641 | EXPORT_SYMBOL(__neigh_event_send); | |
2642 | EXPORT_SYMBOL(neigh_add); | |
2643 | EXPORT_SYMBOL(neigh_changeaddr); | |
2644 | EXPORT_SYMBOL(neigh_compat_output); | |
2645 | EXPORT_SYMBOL(neigh_connected_output); | |
2646 | EXPORT_SYMBOL(neigh_create); | |
2647 | EXPORT_SYMBOL(neigh_delete); | |
2648 | EXPORT_SYMBOL(neigh_destroy); | |
2649 | EXPORT_SYMBOL(neigh_dump_info); | |
2650 | EXPORT_SYMBOL(neigh_event_ns); | |
2651 | EXPORT_SYMBOL(neigh_ifdown); | |
2652 | EXPORT_SYMBOL(neigh_lookup); | |
2653 | EXPORT_SYMBOL(neigh_lookup_nodev); | |
2654 | EXPORT_SYMBOL(neigh_parms_alloc); | |
2655 | EXPORT_SYMBOL(neigh_parms_release); | |
2656 | EXPORT_SYMBOL(neigh_rand_reach_time); | |
2657 | EXPORT_SYMBOL(neigh_resolve_output); | |
2658 | EXPORT_SYMBOL(neigh_table_clear); | |
2659 | EXPORT_SYMBOL(neigh_table_init); | |
2660 | EXPORT_SYMBOL(neigh_update); | |
2661 | EXPORT_SYMBOL(neigh_update_hhs); | |
2662 | EXPORT_SYMBOL(pneigh_enqueue); | |
2663 | EXPORT_SYMBOL(pneigh_lookup); | |
c7fb64db TG |
2664 | EXPORT_SYMBOL(neightbl_dump_info); |
2665 | EXPORT_SYMBOL(neightbl_set); | |
1da177e4 LT |
2666 | |
2667 | #ifdef CONFIG_ARPD | |
2668 | EXPORT_SYMBOL(neigh_app_ns); | |
2669 | #endif | |
2670 | #ifdef CONFIG_SYSCTL | |
2671 | EXPORT_SYMBOL(neigh_sysctl_register); | |
2672 | EXPORT_SYMBOL(neigh_sysctl_unregister); | |
2673 | #endif |