Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * IPv4 FIB: lookup engine and maintenance routines. | |
7 | * | |
8 | * Version: $Id: fib_hash.c,v 1.13 2001/10/31 21:55:54 davem Exp $ | |
9 | * | |
10 | * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> | |
11 | * | |
12 | * This program is free software; you can redistribute it and/or | |
13 | * modify it under the terms of the GNU General Public License | |
14 | * as published by the Free Software Foundation; either version | |
15 | * 2 of the License, or (at your option) any later version. | |
16 | */ | |
17 | ||
1da177e4 LT |
18 | #include <asm/uaccess.h> |
19 | #include <asm/system.h> | |
20 | #include <linux/bitops.h> | |
21 | #include <linux/types.h> | |
22 | #include <linux/kernel.h> | |
1da177e4 LT |
23 | #include <linux/mm.h> |
24 | #include <linux/string.h> | |
25 | #include <linux/socket.h> | |
26 | #include <linux/sockios.h> | |
27 | #include <linux/errno.h> | |
28 | #include <linux/in.h> | |
29 | #include <linux/inet.h> | |
14c85021 | 30 | #include <linux/inetdevice.h> |
1da177e4 LT |
31 | #include <linux/netdevice.h> |
32 | #include <linux/if_arp.h> | |
33 | #include <linux/proc_fs.h> | |
34 | #include <linux/skbuff.h> | |
35 | #include <linux/netlink.h> | |
36 | #include <linux/init.h> | |
37 | ||
38 | #include <net/ip.h> | |
39 | #include <net/protocol.h> | |
40 | #include <net/route.h> | |
41 | #include <net/tcp.h> | |
42 | #include <net/sock.h> | |
43 | #include <net/ip_fib.h> | |
44 | ||
45 | #include "fib_lookup.h" | |
46 | ||
e18b890b CL |
47 | static struct kmem_cache *fn_hash_kmem __read_mostly; |
48 | static struct kmem_cache *fn_alias_kmem __read_mostly; | |
1da177e4 LT |
49 | |
50 | struct fib_node { | |
51 | struct hlist_node fn_hash; | |
52 | struct list_head fn_alias; | |
b6e80c6c | 53 | __be32 fn_key; |
1da177e4 LT |
54 | }; |
55 | ||
56 | struct fn_zone { | |
57 | struct fn_zone *fz_next; /* Next not empty zone */ | |
58 | struct hlist_head *fz_hash; /* Hash table pointer */ | |
59 | int fz_nent; /* Number of entries */ | |
60 | ||
61 | int fz_divisor; /* Hash divisor */ | |
62 | u32 fz_hashmask; /* (fz_divisor - 1) */ | |
63 | #define FZ_HASHMASK(fz) ((fz)->fz_hashmask) | |
64 | ||
65 | int fz_order; /* Zone order */ | |
b6e80c6c | 66 | __be32 fz_mask; |
1da177e4 LT |
67 | #define FZ_MASK(fz) ((fz)->fz_mask) |
68 | }; | |
69 | ||
70 | /* NOTE. On fast computers evaluation of fz_hashmask and fz_mask | |
71 | * can be cheaper than memory lookup, so that FZ_* macros are used. | |
72 | */ | |
73 | ||
74 | struct fn_hash { | |
75 | struct fn_zone *fn_zones[33]; | |
76 | struct fn_zone *fn_zone_list; | |
77 | }; | |
78 | ||
b6e80c6c | 79 | static inline u32 fn_hash(__be32 key, struct fn_zone *fz) |
1da177e4 LT |
80 | { |
81 | u32 h = ntohl(key)>>(32 - fz->fz_order); | |
82 | h ^= (h>>20); | |
83 | h ^= (h>>10); | |
84 | h ^= (h>>5); | |
85 | h &= FZ_HASHMASK(fz); | |
86 | return h; | |
87 | } | |
88 | ||
b6e80c6c | 89 | static inline __be32 fz_key(__be32 dst, struct fn_zone *fz) |
1da177e4 LT |
90 | { |
91 | return dst & FZ_MASK(fz); | |
92 | } | |
93 | ||
94 | static DEFINE_RWLOCK(fib_hash_lock); | |
95 | static unsigned int fib_hash_genid; | |
96 | ||
97 | #define FZ_MAX_DIVISOR ((PAGE_SIZE<<MAX_ORDER) / sizeof(struct hlist_head)) | |
98 | ||
99 | static struct hlist_head *fz_hash_alloc(int divisor) | |
100 | { | |
101 | unsigned long size = divisor * sizeof(struct hlist_head); | |
102 | ||
103 | if (size <= PAGE_SIZE) { | |
104 | return kmalloc(size, GFP_KERNEL); | |
105 | } else { | |
106 | return (struct hlist_head *) | |
107 | __get_free_pages(GFP_KERNEL, get_order(size)); | |
108 | } | |
109 | } | |
110 | ||
111 | /* The fib hash lock must be held when this is called. */ | |
112 | static inline void fn_rebuild_zone(struct fn_zone *fz, | |
113 | struct hlist_head *old_ht, | |
114 | int old_divisor) | |
115 | { | |
116 | int i; | |
117 | ||
118 | for (i = 0; i < old_divisor; i++) { | |
119 | struct hlist_node *node, *n; | |
120 | struct fib_node *f; | |
121 | ||
122 | hlist_for_each_entry_safe(f, node, n, &old_ht[i], fn_hash) { | |
123 | struct hlist_head *new_head; | |
124 | ||
125 | hlist_del(&f->fn_hash); | |
126 | ||
127 | new_head = &fz->fz_hash[fn_hash(f->fn_key, fz)]; | |
128 | hlist_add_head(&f->fn_hash, new_head); | |
129 | } | |
130 | } | |
131 | } | |
132 | ||
133 | static void fz_hash_free(struct hlist_head *hash, int divisor) | |
134 | { | |
135 | unsigned long size = divisor * sizeof(struct hlist_head); | |
136 | ||
137 | if (size <= PAGE_SIZE) | |
138 | kfree(hash); | |
139 | else | |
140 | free_pages((unsigned long)hash, get_order(size)); | |
141 | } | |
142 | ||
143 | static void fn_rehash_zone(struct fn_zone *fz) | |
144 | { | |
145 | struct hlist_head *ht, *old_ht; | |
146 | int old_divisor, new_divisor; | |
147 | u32 new_hashmask; | |
e905a9ed | 148 | |
1da177e4 LT |
149 | old_divisor = fz->fz_divisor; |
150 | ||
151 | switch (old_divisor) { | |
152 | case 16: | |
153 | new_divisor = 256; | |
154 | break; | |
155 | case 256: | |
156 | new_divisor = 1024; | |
157 | break; | |
158 | default: | |
159 | if ((old_divisor << 1) > FZ_MAX_DIVISOR) { | |
160 | printk(KERN_CRIT "route.c: bad divisor %d!\n", old_divisor); | |
161 | return; | |
162 | } | |
163 | new_divisor = (old_divisor << 1); | |
164 | break; | |
165 | } | |
166 | ||
167 | new_hashmask = (new_divisor - 1); | |
168 | ||
169 | #if RT_CACHE_DEBUG >= 2 | |
170 | printk("fn_rehash_zone: hash for zone %d grows from %d\n", fz->fz_order, old_divisor); | |
171 | #endif | |
172 | ||
173 | ht = fz_hash_alloc(new_divisor); | |
174 | ||
175 | if (ht) { | |
176 | memset(ht, 0, new_divisor * sizeof(struct hlist_head)); | |
177 | ||
178 | write_lock_bh(&fib_hash_lock); | |
179 | old_ht = fz->fz_hash; | |
180 | fz->fz_hash = ht; | |
181 | fz->fz_hashmask = new_hashmask; | |
182 | fz->fz_divisor = new_divisor; | |
183 | fn_rebuild_zone(fz, old_ht, old_divisor); | |
184 | fib_hash_genid++; | |
185 | write_unlock_bh(&fib_hash_lock); | |
186 | ||
187 | fz_hash_free(old_ht, old_divisor); | |
188 | } | |
189 | } | |
190 | ||
191 | static inline void fn_free_node(struct fib_node * f) | |
192 | { | |
193 | kmem_cache_free(fn_hash_kmem, f); | |
194 | } | |
195 | ||
196 | static inline void fn_free_alias(struct fib_alias *fa) | |
197 | { | |
198 | fib_release_info(fa->fa_info); | |
199 | kmem_cache_free(fn_alias_kmem, fa); | |
200 | } | |
201 | ||
202 | static struct fn_zone * | |
203 | fn_new_zone(struct fn_hash *table, int z) | |
204 | { | |
205 | int i; | |
0da974f4 | 206 | struct fn_zone *fz = kzalloc(sizeof(struct fn_zone), GFP_KERNEL); |
1da177e4 LT |
207 | if (!fz) |
208 | return NULL; | |
209 | ||
1da177e4 LT |
210 | if (z) { |
211 | fz->fz_divisor = 16; | |
212 | } else { | |
213 | fz->fz_divisor = 1; | |
214 | } | |
215 | fz->fz_hashmask = (fz->fz_divisor - 1); | |
216 | fz->fz_hash = fz_hash_alloc(fz->fz_divisor); | |
217 | if (!fz->fz_hash) { | |
218 | kfree(fz); | |
219 | return NULL; | |
220 | } | |
221 | memset(fz->fz_hash, 0, fz->fz_divisor * sizeof(struct hlist_head *)); | |
222 | fz->fz_order = z; | |
223 | fz->fz_mask = inet_make_mask(z); | |
224 | ||
225 | /* Find the first not empty zone with more specific mask */ | |
226 | for (i=z+1; i<=32; i++) | |
227 | if (table->fn_zones[i]) | |
228 | break; | |
229 | write_lock_bh(&fib_hash_lock); | |
230 | if (i>32) { | |
231 | /* No more specific masks, we are the first. */ | |
232 | fz->fz_next = table->fn_zone_list; | |
233 | table->fn_zone_list = fz; | |
234 | } else { | |
235 | fz->fz_next = table->fn_zones[i]->fz_next; | |
236 | table->fn_zones[i]->fz_next = fz; | |
237 | } | |
238 | table->fn_zones[z] = fz; | |
239 | fib_hash_genid++; | |
240 | write_unlock_bh(&fib_hash_lock); | |
241 | return fz; | |
242 | } | |
243 | ||
244 | static int | |
245 | fn_hash_lookup(struct fib_table *tb, const struct flowi *flp, struct fib_result *res) | |
246 | { | |
247 | int err; | |
248 | struct fn_zone *fz; | |
249 | struct fn_hash *t = (struct fn_hash*)tb->tb_data; | |
250 | ||
251 | read_lock(&fib_hash_lock); | |
252 | for (fz = t->fn_zone_list; fz; fz = fz->fz_next) { | |
253 | struct hlist_head *head; | |
254 | struct hlist_node *node; | |
255 | struct fib_node *f; | |
b6e80c6c | 256 | __be32 k = fz_key(flp->fl4_dst, fz); |
1da177e4 LT |
257 | |
258 | head = &fz->fz_hash[fn_hash(k, fz)]; | |
259 | hlist_for_each_entry(f, node, head, fn_hash) { | |
260 | if (f->fn_key != k) | |
261 | continue; | |
262 | ||
263 | err = fib_semantic_match(&f->fn_alias, | |
264 | flp, res, | |
265 | f->fn_key, fz->fz_mask, | |
266 | fz->fz_order); | |
267 | if (err <= 0) | |
268 | goto out; | |
269 | } | |
270 | } | |
271 | err = 1; | |
272 | out: | |
273 | read_unlock(&fib_hash_lock); | |
274 | return err; | |
275 | } | |
276 | ||
277 | static int fn_hash_last_dflt=-1; | |
278 | ||
279 | static void | |
280 | fn_hash_select_default(struct fib_table *tb, const struct flowi *flp, struct fib_result *res) | |
281 | { | |
282 | int order, last_idx; | |
283 | struct hlist_node *node; | |
284 | struct fib_node *f; | |
285 | struct fib_info *fi = NULL; | |
286 | struct fib_info *last_resort; | |
287 | struct fn_hash *t = (struct fn_hash*)tb->tb_data; | |
288 | struct fn_zone *fz = t->fn_zones[0]; | |
289 | ||
290 | if (fz == NULL) | |
291 | return; | |
292 | ||
293 | last_idx = -1; | |
294 | last_resort = NULL; | |
295 | order = -1; | |
296 | ||
297 | read_lock(&fib_hash_lock); | |
298 | hlist_for_each_entry(f, node, &fz->fz_hash[0], fn_hash) { | |
299 | struct fib_alias *fa; | |
300 | ||
301 | list_for_each_entry(fa, &f->fn_alias, fa_list) { | |
302 | struct fib_info *next_fi = fa->fa_info; | |
303 | ||
304 | if (fa->fa_scope != res->scope || | |
305 | fa->fa_type != RTN_UNICAST) | |
306 | continue; | |
307 | ||
308 | if (next_fi->fib_priority > res->fi->fib_priority) | |
309 | break; | |
310 | if (!next_fi->fib_nh[0].nh_gw || | |
311 | next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK) | |
312 | continue; | |
313 | fa->fa_state |= FA_S_ACCESSED; | |
314 | ||
315 | if (fi == NULL) { | |
316 | if (next_fi != res->fi) | |
317 | break; | |
318 | } else if (!fib_detect_death(fi, order, &last_resort, | |
319 | &last_idx, &fn_hash_last_dflt)) { | |
320 | if (res->fi) | |
321 | fib_info_put(res->fi); | |
322 | res->fi = fi; | |
323 | atomic_inc(&fi->fib_clntref); | |
324 | fn_hash_last_dflt = order; | |
325 | goto out; | |
326 | } | |
327 | fi = next_fi; | |
328 | order++; | |
329 | } | |
330 | } | |
331 | ||
332 | if (order <= 0 || fi == NULL) { | |
333 | fn_hash_last_dflt = -1; | |
334 | goto out; | |
335 | } | |
336 | ||
337 | if (!fib_detect_death(fi, order, &last_resort, &last_idx, &fn_hash_last_dflt)) { | |
338 | if (res->fi) | |
339 | fib_info_put(res->fi); | |
340 | res->fi = fi; | |
341 | atomic_inc(&fi->fib_clntref); | |
342 | fn_hash_last_dflt = order; | |
343 | goto out; | |
344 | } | |
345 | ||
346 | if (last_idx >= 0) { | |
347 | if (res->fi) | |
348 | fib_info_put(res->fi); | |
349 | res->fi = last_resort; | |
350 | if (last_resort) | |
351 | atomic_inc(&last_resort->fib_clntref); | |
352 | } | |
353 | fn_hash_last_dflt = last_idx; | |
354 | out: | |
355 | read_unlock(&fib_hash_lock); | |
356 | } | |
357 | ||
358 | /* Insert node F to FZ. */ | |
359 | static inline void fib_insert_node(struct fn_zone *fz, struct fib_node *f) | |
360 | { | |
361 | struct hlist_head *head = &fz->fz_hash[fn_hash(f->fn_key, fz)]; | |
362 | ||
363 | hlist_add_head(&f->fn_hash, head); | |
364 | } | |
365 | ||
366 | /* Return the node in FZ matching KEY. */ | |
b6e80c6c | 367 | static struct fib_node *fib_find_node(struct fn_zone *fz, __be32 key) |
1da177e4 LT |
368 | { |
369 | struct hlist_head *head = &fz->fz_hash[fn_hash(key, fz)]; | |
370 | struct hlist_node *node; | |
371 | struct fib_node *f; | |
372 | ||
373 | hlist_for_each_entry(f, node, head, fn_hash) { | |
374 | if (f->fn_key == key) | |
375 | return f; | |
376 | } | |
377 | ||
378 | return NULL; | |
379 | } | |
380 | ||
4e902c57 | 381 | static int fn_hash_insert(struct fib_table *tb, struct fib_config *cfg) |
1da177e4 LT |
382 | { |
383 | struct fn_hash *table = (struct fn_hash *) tb->tb_data; | |
384 | struct fib_node *new_f, *f; | |
385 | struct fib_alias *fa, *new_fa; | |
386 | struct fn_zone *fz; | |
387 | struct fib_info *fi; | |
4e902c57 | 388 | u8 tos = cfg->fc_tos; |
b6e80c6c | 389 | __be32 key; |
1da177e4 LT |
390 | int err; |
391 | ||
4e902c57 | 392 | if (cfg->fc_dst_len > 32) |
1da177e4 | 393 | return -EINVAL; |
4e902c57 TG |
394 | |
395 | fz = table->fn_zones[cfg->fc_dst_len]; | |
396 | if (!fz && !(fz = fn_new_zone(table, cfg->fc_dst_len))) | |
1da177e4 LT |
397 | return -ENOBUFS; |
398 | ||
399 | key = 0; | |
4e902c57 TG |
400 | if (cfg->fc_dst) { |
401 | if (cfg->fc_dst & ~FZ_MASK(fz)) | |
1da177e4 | 402 | return -EINVAL; |
4e902c57 | 403 | key = fz_key(cfg->fc_dst, fz); |
1da177e4 LT |
404 | } |
405 | ||
4e902c57 TG |
406 | fi = fib_create_info(cfg); |
407 | if (IS_ERR(fi)) | |
408 | return PTR_ERR(fi); | |
1da177e4 LT |
409 | |
410 | if (fz->fz_nent > (fz->fz_divisor<<1) && | |
411 | fz->fz_divisor < FZ_MAX_DIVISOR && | |
4e902c57 TG |
412 | (cfg->fc_dst_len == 32 || |
413 | (1 << cfg->fc_dst_len) > fz->fz_divisor)) | |
1da177e4 LT |
414 | fn_rehash_zone(fz); |
415 | ||
416 | f = fib_find_node(fz, key); | |
417 | ||
418 | if (!f) | |
419 | fa = NULL; | |
420 | else | |
421 | fa = fib_find_alias(&f->fn_alias, tos, fi->fib_priority); | |
422 | ||
423 | /* Now fa, if non-NULL, points to the first fib alias | |
424 | * with the same keys [prefix,tos,priority], if such key already | |
425 | * exists or to the node before which we will insert new one. | |
426 | * | |
427 | * If fa is NULL, we will need to allocate a new one and | |
428 | * insert to the head of f. | |
429 | * | |
430 | * If f is NULL, no fib node matched the destination key | |
431 | * and we need to allocate a new one of those as well. | |
432 | */ | |
433 | ||
434 | if (fa && fa->fa_tos == tos && | |
435 | fa->fa_info->fib_priority == fi->fib_priority) { | |
436 | struct fib_alias *fa_orig; | |
437 | ||
438 | err = -EEXIST; | |
4e902c57 | 439 | if (cfg->fc_nlflags & NLM_F_EXCL) |
1da177e4 LT |
440 | goto out; |
441 | ||
4e902c57 | 442 | if (cfg->fc_nlflags & NLM_F_REPLACE) { |
1da177e4 LT |
443 | struct fib_info *fi_drop; |
444 | u8 state; | |
445 | ||
446 | write_lock_bh(&fib_hash_lock); | |
447 | fi_drop = fa->fa_info; | |
448 | fa->fa_info = fi; | |
4e902c57 TG |
449 | fa->fa_type = cfg->fc_type; |
450 | fa->fa_scope = cfg->fc_scope; | |
1da177e4 LT |
451 | state = fa->fa_state; |
452 | fa->fa_state &= ~FA_S_ACCESSED; | |
453 | fib_hash_genid++; | |
454 | write_unlock_bh(&fib_hash_lock); | |
455 | ||
456 | fib_release_info(fi_drop); | |
457 | if (state & FA_S_ACCESSED) | |
458 | rt_cache_flush(-1); | |
b8f55831 MK |
459 | rtmsg_fib(RTM_NEWROUTE, key, fa, cfg->fc_dst_len, tb->tb_id, |
460 | &cfg->fc_nlinfo, NLM_F_REPLACE); | |
1da177e4 LT |
461 | return 0; |
462 | } | |
463 | ||
464 | /* Error if we find a perfect match which | |
465 | * uses the same scope, type, and nexthop | |
466 | * information. | |
467 | */ | |
468 | fa_orig = fa; | |
469 | fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list); | |
470 | list_for_each_entry_continue(fa, &f->fn_alias, fa_list) { | |
471 | if (fa->fa_tos != tos) | |
472 | break; | |
473 | if (fa->fa_info->fib_priority != fi->fib_priority) | |
474 | break; | |
4e902c57 TG |
475 | if (fa->fa_type == cfg->fc_type && |
476 | fa->fa_scope == cfg->fc_scope && | |
1da177e4 LT |
477 | fa->fa_info == fi) |
478 | goto out; | |
479 | } | |
4e902c57 | 480 | if (!(cfg->fc_nlflags & NLM_F_APPEND)) |
1da177e4 LT |
481 | fa = fa_orig; |
482 | } | |
483 | ||
484 | err = -ENOENT; | |
4e902c57 | 485 | if (!(cfg->fc_nlflags & NLM_F_CREATE)) |
1da177e4 LT |
486 | goto out; |
487 | ||
488 | err = -ENOBUFS; | |
e94b1766 | 489 | new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL); |
1da177e4 LT |
490 | if (new_fa == NULL) |
491 | goto out; | |
492 | ||
493 | new_f = NULL; | |
494 | if (!f) { | |
e94b1766 | 495 | new_f = kmem_cache_alloc(fn_hash_kmem, GFP_KERNEL); |
1da177e4 LT |
496 | if (new_f == NULL) |
497 | goto out_free_new_fa; | |
498 | ||
499 | INIT_HLIST_NODE(&new_f->fn_hash); | |
500 | INIT_LIST_HEAD(&new_f->fn_alias); | |
501 | new_f->fn_key = key; | |
502 | f = new_f; | |
503 | } | |
504 | ||
505 | new_fa->fa_info = fi; | |
506 | new_fa->fa_tos = tos; | |
4e902c57 TG |
507 | new_fa->fa_type = cfg->fc_type; |
508 | new_fa->fa_scope = cfg->fc_scope; | |
1da177e4 LT |
509 | new_fa->fa_state = 0; |
510 | ||
511 | /* | |
512 | * Insert new entry to the list. | |
513 | */ | |
514 | ||
515 | write_lock_bh(&fib_hash_lock); | |
516 | if (new_f) | |
517 | fib_insert_node(fz, new_f); | |
518 | list_add_tail(&new_fa->fa_list, | |
519 | (fa ? &fa->fa_list : &f->fn_alias)); | |
520 | fib_hash_genid++; | |
521 | write_unlock_bh(&fib_hash_lock); | |
522 | ||
523 | if (new_f) | |
524 | fz->fz_nent++; | |
525 | rt_cache_flush(-1); | |
526 | ||
4e902c57 | 527 | rtmsg_fib(RTM_NEWROUTE, key, new_fa, cfg->fc_dst_len, tb->tb_id, |
b8f55831 | 528 | &cfg->fc_nlinfo, 0); |
1da177e4 LT |
529 | return 0; |
530 | ||
531 | out_free_new_fa: | |
532 | kmem_cache_free(fn_alias_kmem, new_fa); | |
533 | out: | |
534 | fib_release_info(fi); | |
535 | return err; | |
536 | } | |
537 | ||
538 | ||
4e902c57 | 539 | static int fn_hash_delete(struct fib_table *tb, struct fib_config *cfg) |
1da177e4 LT |
540 | { |
541 | struct fn_hash *table = (struct fn_hash*)tb->tb_data; | |
542 | struct fib_node *f; | |
543 | struct fib_alias *fa, *fa_to_delete; | |
1da177e4 | 544 | struct fn_zone *fz; |
b6e80c6c | 545 | __be32 key; |
1da177e4 | 546 | |
4e902c57 | 547 | if (cfg->fc_dst_len > 32) |
1da177e4 | 548 | return -EINVAL; |
4e902c57 TG |
549 | |
550 | if ((fz = table->fn_zones[cfg->fc_dst_len]) == NULL) | |
1da177e4 LT |
551 | return -ESRCH; |
552 | ||
553 | key = 0; | |
4e902c57 TG |
554 | if (cfg->fc_dst) { |
555 | if (cfg->fc_dst & ~FZ_MASK(fz)) | |
1da177e4 | 556 | return -EINVAL; |
4e902c57 | 557 | key = fz_key(cfg->fc_dst, fz); |
1da177e4 LT |
558 | } |
559 | ||
560 | f = fib_find_node(fz, key); | |
561 | ||
562 | if (!f) | |
563 | fa = NULL; | |
564 | else | |
4e902c57 | 565 | fa = fib_find_alias(&f->fn_alias, cfg->fc_tos, 0); |
1da177e4 LT |
566 | if (!fa) |
567 | return -ESRCH; | |
568 | ||
569 | fa_to_delete = NULL; | |
570 | fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list); | |
571 | list_for_each_entry_continue(fa, &f->fn_alias, fa_list) { | |
572 | struct fib_info *fi = fa->fa_info; | |
573 | ||
4e902c57 | 574 | if (fa->fa_tos != cfg->fc_tos) |
1da177e4 LT |
575 | break; |
576 | ||
4e902c57 TG |
577 | if ((!cfg->fc_type || |
578 | fa->fa_type == cfg->fc_type) && | |
579 | (cfg->fc_scope == RT_SCOPE_NOWHERE || | |
580 | fa->fa_scope == cfg->fc_scope) && | |
581 | (!cfg->fc_protocol || | |
582 | fi->fib_protocol == cfg->fc_protocol) && | |
583 | fib_nh_match(cfg, fi) == 0) { | |
1da177e4 LT |
584 | fa_to_delete = fa; |
585 | break; | |
586 | } | |
587 | } | |
588 | ||
589 | if (fa_to_delete) { | |
590 | int kill_fn; | |
591 | ||
592 | fa = fa_to_delete; | |
4e902c57 | 593 | rtmsg_fib(RTM_DELROUTE, key, fa, cfg->fc_dst_len, |
b8f55831 | 594 | tb->tb_id, &cfg->fc_nlinfo, 0); |
1da177e4 LT |
595 | |
596 | kill_fn = 0; | |
597 | write_lock_bh(&fib_hash_lock); | |
598 | list_del(&fa->fa_list); | |
599 | if (list_empty(&f->fn_alias)) { | |
600 | hlist_del(&f->fn_hash); | |
601 | kill_fn = 1; | |
602 | } | |
603 | fib_hash_genid++; | |
604 | write_unlock_bh(&fib_hash_lock); | |
605 | ||
606 | if (fa->fa_state & FA_S_ACCESSED) | |
607 | rt_cache_flush(-1); | |
608 | fn_free_alias(fa); | |
609 | if (kill_fn) { | |
610 | fn_free_node(f); | |
611 | fz->fz_nent--; | |
612 | } | |
613 | ||
614 | return 0; | |
615 | } | |
616 | return -ESRCH; | |
617 | } | |
618 | ||
619 | static int fn_flush_list(struct fn_zone *fz, int idx) | |
620 | { | |
621 | struct hlist_head *head = &fz->fz_hash[idx]; | |
622 | struct hlist_node *node, *n; | |
623 | struct fib_node *f; | |
624 | int found = 0; | |
625 | ||
626 | hlist_for_each_entry_safe(f, node, n, head, fn_hash) { | |
627 | struct fib_alias *fa, *fa_node; | |
628 | int kill_f; | |
629 | ||
630 | kill_f = 0; | |
631 | list_for_each_entry_safe(fa, fa_node, &f->fn_alias, fa_list) { | |
632 | struct fib_info *fi = fa->fa_info; | |
633 | ||
634 | if (fi && (fi->fib_flags&RTNH_F_DEAD)) { | |
635 | write_lock_bh(&fib_hash_lock); | |
636 | list_del(&fa->fa_list); | |
637 | if (list_empty(&f->fn_alias)) { | |
638 | hlist_del(&f->fn_hash); | |
639 | kill_f = 1; | |
640 | } | |
641 | fib_hash_genid++; | |
642 | write_unlock_bh(&fib_hash_lock); | |
643 | ||
644 | fn_free_alias(fa); | |
645 | found++; | |
646 | } | |
647 | } | |
648 | if (kill_f) { | |
649 | fn_free_node(f); | |
650 | fz->fz_nent--; | |
651 | } | |
652 | } | |
653 | return found; | |
654 | } | |
655 | ||
656 | static int fn_hash_flush(struct fib_table *tb) | |
657 | { | |
658 | struct fn_hash *table = (struct fn_hash *) tb->tb_data; | |
659 | struct fn_zone *fz; | |
660 | int found = 0; | |
661 | ||
662 | for (fz = table->fn_zone_list; fz; fz = fz->fz_next) { | |
663 | int i; | |
664 | ||
665 | for (i = fz->fz_divisor - 1; i >= 0; i--) | |
666 | found += fn_flush_list(fz, i); | |
667 | } | |
668 | return found; | |
669 | } | |
670 | ||
671 | ||
672 | static inline int | |
673 | fn_hash_dump_bucket(struct sk_buff *skb, struct netlink_callback *cb, | |
674 | struct fib_table *tb, | |
675 | struct fn_zone *fz, | |
676 | struct hlist_head *head) | |
677 | { | |
678 | struct hlist_node *node; | |
679 | struct fib_node *f; | |
680 | int i, s_i; | |
681 | ||
1af5a8c4 | 682 | s_i = cb->args[4]; |
1da177e4 LT |
683 | i = 0; |
684 | hlist_for_each_entry(f, node, head, fn_hash) { | |
685 | struct fib_alias *fa; | |
686 | ||
687 | list_for_each_entry(fa, &f->fn_alias, fa_list) { | |
688 | if (i < s_i) | |
689 | goto next; | |
690 | ||
691 | if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid, | |
692 | cb->nlh->nlmsg_seq, | |
693 | RTM_NEWROUTE, | |
694 | tb->tb_id, | |
695 | fa->fa_type, | |
696 | fa->fa_scope, | |
be403ea1 | 697 | f->fn_key, |
1da177e4 LT |
698 | fz->fz_order, |
699 | fa->fa_tos, | |
b6544c0b JHS |
700 | fa->fa_info, |
701 | NLM_F_MULTI) < 0) { | |
1af5a8c4 | 702 | cb->args[4] = i; |
1da177e4 LT |
703 | return -1; |
704 | } | |
705 | next: | |
706 | i++; | |
707 | } | |
708 | } | |
1af5a8c4 | 709 | cb->args[4] = i; |
1da177e4 LT |
710 | return skb->len; |
711 | } | |
712 | ||
713 | static inline int | |
714 | fn_hash_dump_zone(struct sk_buff *skb, struct netlink_callback *cb, | |
715 | struct fib_table *tb, | |
716 | struct fn_zone *fz) | |
717 | { | |
718 | int h, s_h; | |
719 | ||
1af5a8c4 | 720 | s_h = cb->args[3]; |
1da177e4 LT |
721 | for (h=0; h < fz->fz_divisor; h++) { |
722 | if (h < s_h) continue; | |
723 | if (h > s_h) | |
1af5a8c4 PM |
724 | memset(&cb->args[4], 0, |
725 | sizeof(cb->args) - 4*sizeof(cb->args[0])); | |
1da177e4 LT |
726 | if (fz->fz_hash == NULL || |
727 | hlist_empty(&fz->fz_hash[h])) | |
728 | continue; | |
729 | if (fn_hash_dump_bucket(skb, cb, tb, fz, &fz->fz_hash[h])<0) { | |
1af5a8c4 | 730 | cb->args[3] = h; |
1da177e4 LT |
731 | return -1; |
732 | } | |
733 | } | |
1af5a8c4 | 734 | cb->args[3] = h; |
1da177e4 LT |
735 | return skb->len; |
736 | } | |
737 | ||
738 | static int fn_hash_dump(struct fib_table *tb, struct sk_buff *skb, struct netlink_callback *cb) | |
739 | { | |
740 | int m, s_m; | |
741 | struct fn_zone *fz; | |
742 | struct fn_hash *table = (struct fn_hash*)tb->tb_data; | |
743 | ||
1af5a8c4 | 744 | s_m = cb->args[2]; |
1da177e4 LT |
745 | read_lock(&fib_hash_lock); |
746 | for (fz = table->fn_zone_list, m=0; fz; fz = fz->fz_next, m++) { | |
747 | if (m < s_m) continue; | |
748 | if (m > s_m) | |
1af5a8c4 PM |
749 | memset(&cb->args[3], 0, |
750 | sizeof(cb->args) - 3*sizeof(cb->args[0])); | |
1da177e4 | 751 | if (fn_hash_dump_zone(skb, cb, tb, fz) < 0) { |
1af5a8c4 | 752 | cb->args[2] = m; |
1da177e4 LT |
753 | read_unlock(&fib_hash_lock); |
754 | return -1; | |
755 | } | |
756 | } | |
757 | read_unlock(&fib_hash_lock); | |
1af5a8c4 | 758 | cb->args[2] = m; |
1da177e4 LT |
759 | return skb->len; |
760 | } | |
761 | ||
762 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
2dfe55b4 | 763 | struct fib_table * fib_hash_init(u32 id) |
1da177e4 | 764 | #else |
2dfe55b4 | 765 | struct fib_table * __init fib_hash_init(u32 id) |
1da177e4 LT |
766 | #endif |
767 | { | |
768 | struct fib_table *tb; | |
769 | ||
770 | if (fn_hash_kmem == NULL) | |
771 | fn_hash_kmem = kmem_cache_create("ip_fib_hash", | |
772 | sizeof(struct fib_node), | |
773 | 0, SLAB_HWCACHE_ALIGN, | |
20c2df83 | 774 | NULL); |
1da177e4 LT |
775 | |
776 | if (fn_alias_kmem == NULL) | |
777 | fn_alias_kmem = kmem_cache_create("ip_fib_alias", | |
778 | sizeof(struct fib_alias), | |
779 | 0, SLAB_HWCACHE_ALIGN, | |
20c2df83 | 780 | NULL); |
1da177e4 LT |
781 | |
782 | tb = kmalloc(sizeof(struct fib_table) + sizeof(struct fn_hash), | |
783 | GFP_KERNEL); | |
784 | if (tb == NULL) | |
785 | return NULL; | |
786 | ||
787 | tb->tb_id = id; | |
788 | tb->tb_lookup = fn_hash_lookup; | |
789 | tb->tb_insert = fn_hash_insert; | |
790 | tb->tb_delete = fn_hash_delete; | |
791 | tb->tb_flush = fn_hash_flush; | |
792 | tb->tb_select_default = fn_hash_select_default; | |
793 | tb->tb_dump = fn_hash_dump; | |
794 | memset(tb->tb_data, 0, sizeof(struct fn_hash)); | |
795 | return tb; | |
796 | } | |
797 | ||
798 | /* ------------------------------------------------------------------------ */ | |
799 | #ifdef CONFIG_PROC_FS | |
800 | ||
801 | struct fib_iter_state { | |
802 | struct fn_zone *zone; | |
803 | int bucket; | |
804 | struct hlist_head *hash_head; | |
805 | struct fib_node *fn; | |
806 | struct fib_alias *fa; | |
807 | loff_t pos; | |
808 | unsigned int genid; | |
809 | int valid; | |
810 | }; | |
811 | ||
812 | static struct fib_alias *fib_get_first(struct seq_file *seq) | |
813 | { | |
814 | struct fib_iter_state *iter = seq->private; | |
815 | struct fn_hash *table = (struct fn_hash *) ip_fib_main_table->tb_data; | |
816 | ||
817 | iter->bucket = 0; | |
818 | iter->hash_head = NULL; | |
819 | iter->fn = NULL; | |
820 | iter->fa = NULL; | |
821 | iter->pos = 0; | |
822 | iter->genid = fib_hash_genid; | |
823 | iter->valid = 1; | |
824 | ||
825 | for (iter->zone = table->fn_zone_list; iter->zone; | |
826 | iter->zone = iter->zone->fz_next) { | |
827 | int maxslot; | |
828 | ||
829 | if (!iter->zone->fz_nent) | |
830 | continue; | |
831 | ||
832 | iter->hash_head = iter->zone->fz_hash; | |
833 | maxslot = iter->zone->fz_divisor; | |
834 | ||
835 | for (iter->bucket = 0; iter->bucket < maxslot; | |
836 | ++iter->bucket, ++iter->hash_head) { | |
837 | struct hlist_node *node; | |
838 | struct fib_node *fn; | |
839 | ||
840 | hlist_for_each_entry(fn,node,iter->hash_head,fn_hash) { | |
841 | struct fib_alias *fa; | |
842 | ||
843 | list_for_each_entry(fa,&fn->fn_alias,fa_list) { | |
844 | iter->fn = fn; | |
845 | iter->fa = fa; | |
846 | goto out; | |
847 | } | |
848 | } | |
849 | } | |
850 | } | |
851 | out: | |
852 | return iter->fa; | |
853 | } | |
854 | ||
855 | static struct fib_alias *fib_get_next(struct seq_file *seq) | |
856 | { | |
857 | struct fib_iter_state *iter = seq->private; | |
858 | struct fib_node *fn; | |
859 | struct fib_alias *fa; | |
860 | ||
861 | /* Advance FA, if any. */ | |
862 | fn = iter->fn; | |
863 | fa = iter->fa; | |
864 | if (fa) { | |
865 | BUG_ON(!fn); | |
866 | list_for_each_entry_continue(fa, &fn->fn_alias, fa_list) { | |
867 | iter->fa = fa; | |
868 | goto out; | |
869 | } | |
870 | } | |
871 | ||
872 | fa = iter->fa = NULL; | |
873 | ||
874 | /* Advance FN. */ | |
875 | if (fn) { | |
876 | struct hlist_node *node = &fn->fn_hash; | |
877 | hlist_for_each_entry_continue(fn, node, fn_hash) { | |
878 | iter->fn = fn; | |
879 | ||
880 | list_for_each_entry(fa, &fn->fn_alias, fa_list) { | |
881 | iter->fa = fa; | |
882 | goto out; | |
883 | } | |
884 | } | |
885 | } | |
886 | ||
887 | fn = iter->fn = NULL; | |
888 | ||
889 | /* Advance hash chain. */ | |
890 | if (!iter->zone) | |
891 | goto out; | |
892 | ||
893 | for (;;) { | |
894 | struct hlist_node *node; | |
895 | int maxslot; | |
896 | ||
897 | maxslot = iter->zone->fz_divisor; | |
898 | ||
899 | while (++iter->bucket < maxslot) { | |
900 | iter->hash_head++; | |
901 | ||
902 | hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) { | |
903 | list_for_each_entry(fa, &fn->fn_alias, fa_list) { | |
904 | iter->fn = fn; | |
905 | iter->fa = fa; | |
906 | goto out; | |
907 | } | |
908 | } | |
909 | } | |
910 | ||
911 | iter->zone = iter->zone->fz_next; | |
912 | ||
913 | if (!iter->zone) | |
914 | goto out; | |
e905a9ed | 915 | |
1da177e4 LT |
916 | iter->bucket = 0; |
917 | iter->hash_head = iter->zone->fz_hash; | |
918 | ||
919 | hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) { | |
920 | list_for_each_entry(fa, &fn->fn_alias, fa_list) { | |
921 | iter->fn = fn; | |
922 | iter->fa = fa; | |
923 | goto out; | |
924 | } | |
925 | } | |
926 | } | |
927 | out: | |
928 | iter->pos++; | |
929 | return fa; | |
930 | } | |
931 | ||
932 | static struct fib_alias *fib_get_idx(struct seq_file *seq, loff_t pos) | |
933 | { | |
934 | struct fib_iter_state *iter = seq->private; | |
935 | struct fib_alias *fa; | |
e905a9ed | 936 | |
1da177e4 LT |
937 | if (iter->valid && pos >= iter->pos && iter->genid == fib_hash_genid) { |
938 | fa = iter->fa; | |
939 | pos -= iter->pos; | |
940 | } else | |
941 | fa = fib_get_first(seq); | |
942 | ||
943 | if (fa) | |
944 | while (pos && (fa = fib_get_next(seq))) | |
945 | --pos; | |
946 | return pos ? NULL : fa; | |
947 | } | |
948 | ||
949 | static void *fib_seq_start(struct seq_file *seq, loff_t *pos) | |
950 | { | |
951 | void *v = NULL; | |
952 | ||
953 | read_lock(&fib_hash_lock); | |
954 | if (ip_fib_main_table) | |
955 | v = *pos ? fib_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
956 | return v; | |
957 | } | |
958 | ||
959 | static void *fib_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
960 | { | |
961 | ++*pos; | |
962 | return v == SEQ_START_TOKEN ? fib_get_first(seq) : fib_get_next(seq); | |
963 | } | |
964 | ||
965 | static void fib_seq_stop(struct seq_file *seq, void *v) | |
966 | { | |
967 | read_unlock(&fib_hash_lock); | |
968 | } | |
969 | ||
b6e80c6c | 970 | static unsigned fib_flag_trans(int type, __be32 mask, struct fib_info *fi) |
1da177e4 | 971 | { |
9b5b5cff | 972 | static const unsigned type2flags[RTN_MAX + 1] = { |
1da177e4 LT |
973 | [7] = RTF_REJECT, [8] = RTF_REJECT, |
974 | }; | |
975 | unsigned flags = type2flags[type]; | |
976 | ||
977 | if (fi && fi->fib_nh->nh_gw) | |
978 | flags |= RTF_GATEWAY; | |
b6e80c6c | 979 | if (mask == htonl(0xFFFFFFFF)) |
1da177e4 LT |
980 | flags |= RTF_HOST; |
981 | flags |= RTF_UP; | |
982 | return flags; | |
983 | } | |
984 | ||
e905a9ed | 985 | /* |
1da177e4 LT |
986 | * This outputs /proc/net/route. |
987 | * | |
988 | * It always works in backward compatibility mode. | |
989 | * The format of the file is not supposed to be changed. | |
990 | */ | |
991 | static int fib_seq_show(struct seq_file *seq, void *v) | |
992 | { | |
993 | struct fib_iter_state *iter; | |
994 | char bf[128]; | |
b6e80c6c | 995 | __be32 prefix, mask; |
1da177e4 LT |
996 | unsigned flags; |
997 | struct fib_node *f; | |
998 | struct fib_alias *fa; | |
999 | struct fib_info *fi; | |
1000 | ||
1001 | if (v == SEQ_START_TOKEN) { | |
1002 | seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway " | |
1003 | "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU" | |
1004 | "\tWindow\tIRTT"); | |
1005 | goto out; | |
1006 | } | |
1007 | ||
1008 | iter = seq->private; | |
1009 | f = iter->fn; | |
1010 | fa = iter->fa; | |
1011 | fi = fa->fa_info; | |
1012 | prefix = f->fn_key; | |
1013 | mask = FZ_MASK(iter->zone); | |
1014 | flags = fib_flag_trans(fa->fa_type, mask, fi); | |
1015 | if (fi) | |
1016 | snprintf(bf, sizeof(bf), | |
1017 | "%s\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u", | |
1018 | fi->fib_dev ? fi->fib_dev->name : "*", prefix, | |
1019 | fi->fib_nh->nh_gw, flags, 0, 0, fi->fib_priority, | |
1020 | mask, (fi->fib_advmss ? fi->fib_advmss + 40 : 0), | |
1021 | fi->fib_window, | |
1022 | fi->fib_rtt >> 3); | |
1023 | else | |
1024 | snprintf(bf, sizeof(bf), | |
1025 | "*\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u", | |
1026 | prefix, 0, flags, 0, 0, 0, mask, 0, 0, 0); | |
1027 | seq_printf(seq, "%-127s\n", bf); | |
1028 | out: | |
1029 | return 0; | |
1030 | } | |
1031 | ||
f690808e | 1032 | static const struct seq_operations fib_seq_ops = { |
1da177e4 LT |
1033 | .start = fib_seq_start, |
1034 | .next = fib_seq_next, | |
1035 | .stop = fib_seq_stop, | |
1036 | .show = fib_seq_show, | |
1037 | }; | |
1038 | ||
1039 | static int fib_seq_open(struct inode *inode, struct file *file) | |
1040 | { | |
1041 | struct seq_file *seq; | |
1042 | int rc = -ENOMEM; | |
0da974f4 | 1043 | struct fib_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL); |
e905a9ed | 1044 | |
1da177e4 LT |
1045 | if (!s) |
1046 | goto out; | |
1047 | ||
1048 | rc = seq_open(file, &fib_seq_ops); | |
1049 | if (rc) | |
1050 | goto out_kfree; | |
1051 | ||
1052 | seq = file->private_data; | |
1053 | seq->private = s; | |
1da177e4 LT |
1054 | out: |
1055 | return rc; | |
1056 | out_kfree: | |
1057 | kfree(s); | |
1058 | goto out; | |
1059 | } | |
1060 | ||
9a32144e | 1061 | static const struct file_operations fib_seq_fops = { |
1da177e4 LT |
1062 | .owner = THIS_MODULE, |
1063 | .open = fib_seq_open, | |
1064 | .read = seq_read, | |
1065 | .llseek = seq_lseek, | |
1066 | .release = seq_release_private, | |
1067 | }; | |
1068 | ||
1069 | int __init fib_proc_init(void) | |
1070 | { | |
1071 | if (!proc_net_fops_create("route", S_IRUGO, &fib_seq_fops)) | |
1072 | return -ENOMEM; | |
1073 | return 0; | |
1074 | } | |
1075 | ||
1076 | void __init fib_proc_exit(void) | |
1077 | { | |
1078 | proc_net_remove("route"); | |
1079 | } | |
1080 | #endif /* CONFIG_PROC_FS */ |