Merge git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
[wrt350n-kernel.git] / net / ipv4 / fib_hash.c
blob58057dc8c2e22ccfcf814e7135a0920c1738c2fa
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.
6 * IPv4 FIB: lookup engine and maintenance routines.
8 * Version: $Id: fib_hash.c,v 1.13 2001/10/31 21:55:54 davem Exp $
10 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
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.
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>
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>
30 #include <linux/inetdevice.h>
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>
38 #include <net/net_namespace.h>
39 #include <net/ip.h>
40 #include <net/protocol.h>
41 #include <net/route.h>
42 #include <net/tcp.h>
43 #include <net/sock.h>
44 #include <net/ip_fib.h>
46 #include "fib_lookup.h"
48 static struct kmem_cache *fn_hash_kmem __read_mostly;
49 static struct kmem_cache *fn_alias_kmem __read_mostly;
51 struct fib_node {
52 struct hlist_node fn_hash;
53 struct list_head fn_alias;
54 __be32 fn_key;
55 struct fib_alias fn_embedded_alias;
58 struct fn_zone {
59 struct fn_zone *fz_next; /* Next not empty zone */
60 struct hlist_head *fz_hash; /* Hash table pointer */
61 int fz_nent; /* Number of entries */
63 int fz_divisor; /* Hash divisor */
64 u32 fz_hashmask; /* (fz_divisor - 1) */
65 #define FZ_HASHMASK(fz) ((fz)->fz_hashmask)
67 int fz_order; /* Zone order */
68 __be32 fz_mask;
69 #define FZ_MASK(fz) ((fz)->fz_mask)
72 /* NOTE. On fast computers evaluation of fz_hashmask and fz_mask
73 * can be cheaper than memory lookup, so that FZ_* macros are used.
76 struct fn_hash {
77 struct fn_zone *fn_zones[33];
78 struct fn_zone *fn_zone_list;
81 static inline u32 fn_hash(__be32 key, struct fn_zone *fz)
83 u32 h = ntohl(key)>>(32 - fz->fz_order);
84 h ^= (h>>20);
85 h ^= (h>>10);
86 h ^= (h>>5);
87 h &= FZ_HASHMASK(fz);
88 return h;
91 static inline __be32 fz_key(__be32 dst, struct fn_zone *fz)
93 return dst & FZ_MASK(fz);
96 static DEFINE_RWLOCK(fib_hash_lock);
97 static unsigned int fib_hash_genid;
99 #define FZ_MAX_DIVISOR ((PAGE_SIZE<<MAX_ORDER) / sizeof(struct hlist_head))
101 static struct hlist_head *fz_hash_alloc(int divisor)
103 unsigned long size = divisor * sizeof(struct hlist_head);
105 if (size <= PAGE_SIZE) {
106 return kzalloc(size, GFP_KERNEL);
107 } else {
108 return (struct hlist_head *)
109 __get_free_pages(GFP_KERNEL | __GFP_ZERO, get_order(size));
113 /* The fib hash lock must be held when this is called. */
114 static inline void fn_rebuild_zone(struct fn_zone *fz,
115 struct hlist_head *old_ht,
116 int old_divisor)
118 int i;
120 for (i = 0; i < old_divisor; i++) {
121 struct hlist_node *node, *n;
122 struct fib_node *f;
124 hlist_for_each_entry_safe(f, node, n, &old_ht[i], fn_hash) {
125 struct hlist_head *new_head;
127 hlist_del(&f->fn_hash);
129 new_head = &fz->fz_hash[fn_hash(f->fn_key, fz)];
130 hlist_add_head(&f->fn_hash, new_head);
135 static void fz_hash_free(struct hlist_head *hash, int divisor)
137 unsigned long size = divisor * sizeof(struct hlist_head);
139 if (size <= PAGE_SIZE)
140 kfree(hash);
141 else
142 free_pages((unsigned long)hash, get_order(size));
145 static void fn_rehash_zone(struct fn_zone *fz)
147 struct hlist_head *ht, *old_ht;
148 int old_divisor, new_divisor;
149 u32 new_hashmask;
151 old_divisor = fz->fz_divisor;
153 switch (old_divisor) {
154 case 16:
155 new_divisor = 256;
156 break;
157 case 256:
158 new_divisor = 1024;
159 break;
160 default:
161 if ((old_divisor << 1) > FZ_MAX_DIVISOR) {
162 printk(KERN_CRIT "route.c: bad divisor %d!\n", old_divisor);
163 return;
165 new_divisor = (old_divisor << 1);
166 break;
169 new_hashmask = (new_divisor - 1);
171 #if RT_CACHE_DEBUG >= 2
172 printk(KERN_DEBUG "fn_rehash_zone: hash for zone %d grows from %d\n",
173 fz->fz_order, old_divisor);
174 #endif
176 ht = fz_hash_alloc(new_divisor);
178 if (ht) {
179 write_lock_bh(&fib_hash_lock);
180 old_ht = fz->fz_hash;
181 fz->fz_hash = ht;
182 fz->fz_hashmask = new_hashmask;
183 fz->fz_divisor = new_divisor;
184 fn_rebuild_zone(fz, old_ht, old_divisor);
185 fib_hash_genid++;
186 write_unlock_bh(&fib_hash_lock);
188 fz_hash_free(old_ht, old_divisor);
192 static inline void fn_free_node(struct fib_node * f)
194 kmem_cache_free(fn_hash_kmem, f);
197 static inline void fn_free_alias(struct fib_alias *fa, struct fib_node *f)
199 fib_release_info(fa->fa_info);
200 if (fa == &f->fn_embedded_alias)
201 fa->fa_info = NULL;
202 else
203 kmem_cache_free(fn_alias_kmem, fa);
206 static struct fn_zone *
207 fn_new_zone(struct fn_hash *table, int z)
209 int i;
210 struct fn_zone *fz = kzalloc(sizeof(struct fn_zone), GFP_KERNEL);
211 if (!fz)
212 return NULL;
214 if (z) {
215 fz->fz_divisor = 16;
216 } else {
217 fz->fz_divisor = 1;
219 fz->fz_hashmask = (fz->fz_divisor - 1);
220 fz->fz_hash = fz_hash_alloc(fz->fz_divisor);
221 if (!fz->fz_hash) {
222 kfree(fz);
223 return NULL;
225 fz->fz_order = z;
226 fz->fz_mask = inet_make_mask(z);
228 /* Find the first not empty zone with more specific mask */
229 for (i=z+1; i<=32; i++)
230 if (table->fn_zones[i])
231 break;
232 write_lock_bh(&fib_hash_lock);
233 if (i>32) {
234 /* No more specific masks, we are the first. */
235 fz->fz_next = table->fn_zone_list;
236 table->fn_zone_list = fz;
237 } else {
238 fz->fz_next = table->fn_zones[i]->fz_next;
239 table->fn_zones[i]->fz_next = fz;
241 table->fn_zones[z] = fz;
242 fib_hash_genid++;
243 write_unlock_bh(&fib_hash_lock);
244 return fz;
247 static int
248 fn_hash_lookup(struct fib_table *tb, const struct flowi *flp, struct fib_result *res)
250 int err;
251 struct fn_zone *fz;
252 struct fn_hash *t = (struct fn_hash*)tb->tb_data;
254 read_lock(&fib_hash_lock);
255 for (fz = t->fn_zone_list; fz; fz = fz->fz_next) {
256 struct hlist_head *head;
257 struct hlist_node *node;
258 struct fib_node *f;
259 __be32 k = fz_key(flp->fl4_dst, fz);
261 head = &fz->fz_hash[fn_hash(k, fz)];
262 hlist_for_each_entry(f, node, head, fn_hash) {
263 if (f->fn_key != k)
264 continue;
266 err = fib_semantic_match(&f->fn_alias,
267 flp, res,
268 f->fn_key, fz->fz_mask,
269 fz->fz_order);
270 if (err <= 0)
271 goto out;
274 err = 1;
275 out:
276 read_unlock(&fib_hash_lock);
277 return err;
280 static void
281 fn_hash_select_default(struct fib_table *tb, const struct flowi *flp, struct fib_result *res)
283 int order, last_idx;
284 struct hlist_node *node;
285 struct fib_node *f;
286 struct fib_info *fi = NULL;
287 struct fib_info *last_resort;
288 struct fn_hash *t = (struct fn_hash*)tb->tb_data;
289 struct fn_zone *fz = t->fn_zones[0];
291 if (fz == NULL)
292 return;
294 last_idx = -1;
295 last_resort = NULL;
296 order = -1;
298 read_lock(&fib_hash_lock);
299 hlist_for_each_entry(f, node, &fz->fz_hash[0], fn_hash) {
300 struct fib_alias *fa;
302 list_for_each_entry(fa, &f->fn_alias, fa_list) {
303 struct fib_info *next_fi = fa->fa_info;
305 if (fa->fa_scope != res->scope ||
306 fa->fa_type != RTN_UNICAST)
307 continue;
309 if (next_fi->fib_priority > res->fi->fib_priority)
310 break;
311 if (!next_fi->fib_nh[0].nh_gw ||
312 next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK)
313 continue;
314 fa->fa_state |= FA_S_ACCESSED;
316 if (fi == NULL) {
317 if (next_fi != res->fi)
318 break;
319 } else if (!fib_detect_death(fi, order, &last_resort,
320 &last_idx, tb->tb_default)) {
321 fib_result_assign(res, fi);
322 tb->tb_default = order;
323 goto out;
325 fi = next_fi;
326 order++;
330 if (order <= 0 || fi == NULL) {
331 tb->tb_default = -1;
332 goto out;
335 if (!fib_detect_death(fi, order, &last_resort, &last_idx,
336 tb->tb_default)) {
337 fib_result_assign(res, fi);
338 tb->tb_default = order;
339 goto out;
342 if (last_idx >= 0)
343 fib_result_assign(res, last_resort);
344 tb->tb_default = last_idx;
345 out:
346 read_unlock(&fib_hash_lock);
349 /* Insert node F to FZ. */
350 static inline void fib_insert_node(struct fn_zone *fz, struct fib_node *f)
352 struct hlist_head *head = &fz->fz_hash[fn_hash(f->fn_key, fz)];
354 hlist_add_head(&f->fn_hash, head);
357 /* Return the node in FZ matching KEY. */
358 static struct fib_node *fib_find_node(struct fn_zone *fz, __be32 key)
360 struct hlist_head *head = &fz->fz_hash[fn_hash(key, fz)];
361 struct hlist_node *node;
362 struct fib_node *f;
364 hlist_for_each_entry(f, node, head, fn_hash) {
365 if (f->fn_key == key)
366 return f;
369 return NULL;
372 static int fn_hash_insert(struct fib_table *tb, struct fib_config *cfg)
374 struct fn_hash *table = (struct fn_hash *) tb->tb_data;
375 <<<<<<< HEAD:net/ipv4/fib_hash.c
376 struct fib_node *new_f, *f;
377 =======
378 struct fib_node *new_f = NULL;
379 struct fib_node *f;
380 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/fib_hash.c
381 struct fib_alias *fa, *new_fa;
382 struct fn_zone *fz;
383 struct fib_info *fi;
384 u8 tos = cfg->fc_tos;
385 __be32 key;
386 int err;
388 if (cfg->fc_dst_len > 32)
389 return -EINVAL;
391 fz = table->fn_zones[cfg->fc_dst_len];
392 if (!fz && !(fz = fn_new_zone(table, cfg->fc_dst_len)))
393 return -ENOBUFS;
395 key = 0;
396 if (cfg->fc_dst) {
397 if (cfg->fc_dst & ~FZ_MASK(fz))
398 return -EINVAL;
399 key = fz_key(cfg->fc_dst, fz);
402 fi = fib_create_info(cfg);
403 if (IS_ERR(fi))
404 return PTR_ERR(fi);
406 if (fz->fz_nent > (fz->fz_divisor<<1) &&
407 fz->fz_divisor < FZ_MAX_DIVISOR &&
408 (cfg->fc_dst_len == 32 ||
409 (1 << cfg->fc_dst_len) > fz->fz_divisor))
410 fn_rehash_zone(fz);
412 f = fib_find_node(fz, key);
414 if (!f)
415 fa = NULL;
416 else
417 fa = fib_find_alias(&f->fn_alias, tos, fi->fib_priority);
419 /* Now fa, if non-NULL, points to the first fib alias
420 * with the same keys [prefix,tos,priority], if such key already
421 * exists or to the node before which we will insert new one.
423 * If fa is NULL, we will need to allocate a new one and
424 * insert to the head of f.
426 * If f is NULL, no fib node matched the destination key
427 * and we need to allocate a new one of those as well.
430 if (fa && fa->fa_tos == tos &&
431 fa->fa_info->fib_priority == fi->fib_priority) {
432 struct fib_alias *fa_first, *fa_match;
434 err = -EEXIST;
435 if (cfg->fc_nlflags & NLM_F_EXCL)
436 goto out;
438 /* We have 2 goals:
439 * 1. Find exact match for type, scope, fib_info to avoid
440 * duplicate routes
441 * 2. Find next 'fa' (or head), NLM_F_APPEND inserts before it
443 fa_match = NULL;
444 fa_first = fa;
445 fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
446 list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
447 if (fa->fa_tos != tos)
448 break;
449 if (fa->fa_info->fib_priority != fi->fib_priority)
450 break;
451 if (fa->fa_type == cfg->fc_type &&
452 fa->fa_scope == cfg->fc_scope &&
453 fa->fa_info == fi) {
454 fa_match = fa;
455 break;
459 if (cfg->fc_nlflags & NLM_F_REPLACE) {
460 struct fib_info *fi_drop;
461 u8 state;
463 fa = fa_first;
464 if (fa_match) {
465 if (fa == fa_match)
466 err = 0;
467 goto out;
469 write_lock_bh(&fib_hash_lock);
470 fi_drop = fa->fa_info;
471 fa->fa_info = fi;
472 fa->fa_type = cfg->fc_type;
473 fa->fa_scope = cfg->fc_scope;
474 state = fa->fa_state;
475 fa->fa_state &= ~FA_S_ACCESSED;
476 fib_hash_genid++;
477 write_unlock_bh(&fib_hash_lock);
479 fib_release_info(fi_drop);
480 if (state & FA_S_ACCESSED)
481 rt_cache_flush(-1);
482 rtmsg_fib(RTM_NEWROUTE, key, fa, cfg->fc_dst_len, tb->tb_id,
483 &cfg->fc_nlinfo, NLM_F_REPLACE);
484 return 0;
487 /* Error if we find a perfect match which
488 * uses the same scope, type, and nexthop
489 * information.
491 if (fa_match)
492 goto out;
494 if (!(cfg->fc_nlflags & NLM_F_APPEND))
495 fa = fa_first;
498 err = -ENOENT;
499 if (!(cfg->fc_nlflags & NLM_F_CREATE))
500 goto out;
502 err = -ENOBUFS;
504 <<<<<<< HEAD:net/ipv4/fib_hash.c
505 new_f = NULL;
506 =======
507 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/fib_hash.c
508 if (!f) {
509 new_f = kmem_cache_zalloc(fn_hash_kmem, GFP_KERNEL);
510 if (new_f == NULL)
511 goto out;
513 INIT_HLIST_NODE(&new_f->fn_hash);
514 INIT_LIST_HEAD(&new_f->fn_alias);
515 new_f->fn_key = key;
516 f = new_f;
519 new_fa = &f->fn_embedded_alias;
520 if (new_fa->fa_info != NULL) {
521 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
522 if (new_fa == NULL)
523 <<<<<<< HEAD:net/ipv4/fib_hash.c
524 goto out_free_new_f;
525 =======
526 goto out;
527 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/fib_hash.c
529 new_fa->fa_info = fi;
530 new_fa->fa_tos = tos;
531 new_fa->fa_type = cfg->fc_type;
532 new_fa->fa_scope = cfg->fc_scope;
533 new_fa->fa_state = 0;
536 * Insert new entry to the list.
539 write_lock_bh(&fib_hash_lock);
540 if (new_f)
541 fib_insert_node(fz, new_f);
542 list_add_tail(&new_fa->fa_list,
543 (fa ? &fa->fa_list : &f->fn_alias));
544 fib_hash_genid++;
545 write_unlock_bh(&fib_hash_lock);
547 if (new_f)
548 fz->fz_nent++;
549 rt_cache_flush(-1);
551 rtmsg_fib(RTM_NEWROUTE, key, new_fa, cfg->fc_dst_len, tb->tb_id,
552 &cfg->fc_nlinfo, 0);
553 return 0;
555 <<<<<<< HEAD:net/ipv4/fib_hash.c
556 out_free_new_f:
557 kmem_cache_free(fn_hash_kmem, new_f);
558 =======
559 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/fib_hash.c
560 out:
561 <<<<<<< HEAD:net/ipv4/fib_hash.c
562 =======
563 if (new_f)
564 kmem_cache_free(fn_hash_kmem, new_f);
565 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/fib_hash.c
566 fib_release_info(fi);
567 return err;
571 static int fn_hash_delete(struct fib_table *tb, struct fib_config *cfg)
573 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
574 struct fib_node *f;
575 struct fib_alias *fa, *fa_to_delete;
576 struct fn_zone *fz;
577 __be32 key;
579 if (cfg->fc_dst_len > 32)
580 return -EINVAL;
582 if ((fz = table->fn_zones[cfg->fc_dst_len]) == NULL)
583 return -ESRCH;
585 key = 0;
586 if (cfg->fc_dst) {
587 if (cfg->fc_dst & ~FZ_MASK(fz))
588 return -EINVAL;
589 key = fz_key(cfg->fc_dst, fz);
592 f = fib_find_node(fz, key);
594 if (!f)
595 fa = NULL;
596 else
597 fa = fib_find_alias(&f->fn_alias, cfg->fc_tos, 0);
598 if (!fa)
599 return -ESRCH;
601 fa_to_delete = NULL;
602 fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
603 list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
604 struct fib_info *fi = fa->fa_info;
606 if (fa->fa_tos != cfg->fc_tos)
607 break;
609 if ((!cfg->fc_type ||
610 fa->fa_type == cfg->fc_type) &&
611 (cfg->fc_scope == RT_SCOPE_NOWHERE ||
612 fa->fa_scope == cfg->fc_scope) &&
613 (!cfg->fc_protocol ||
614 fi->fib_protocol == cfg->fc_protocol) &&
615 fib_nh_match(cfg, fi) == 0) {
616 fa_to_delete = fa;
617 break;
621 if (fa_to_delete) {
622 int kill_fn;
624 fa = fa_to_delete;
625 rtmsg_fib(RTM_DELROUTE, key, fa, cfg->fc_dst_len,
626 tb->tb_id, &cfg->fc_nlinfo, 0);
628 kill_fn = 0;
629 write_lock_bh(&fib_hash_lock);
630 list_del(&fa->fa_list);
631 if (list_empty(&f->fn_alias)) {
632 hlist_del(&f->fn_hash);
633 kill_fn = 1;
635 fib_hash_genid++;
636 write_unlock_bh(&fib_hash_lock);
638 if (fa->fa_state & FA_S_ACCESSED)
639 rt_cache_flush(-1);
640 fn_free_alias(fa, f);
641 if (kill_fn) {
642 fn_free_node(f);
643 fz->fz_nent--;
646 return 0;
648 return -ESRCH;
651 static int fn_flush_list(struct fn_zone *fz, int idx)
653 struct hlist_head *head = &fz->fz_hash[idx];
654 struct hlist_node *node, *n;
655 struct fib_node *f;
656 int found = 0;
658 hlist_for_each_entry_safe(f, node, n, head, fn_hash) {
659 struct fib_alias *fa, *fa_node;
660 int kill_f;
662 kill_f = 0;
663 list_for_each_entry_safe(fa, fa_node, &f->fn_alias, fa_list) {
664 struct fib_info *fi = fa->fa_info;
666 if (fi && (fi->fib_flags&RTNH_F_DEAD)) {
667 write_lock_bh(&fib_hash_lock);
668 list_del(&fa->fa_list);
669 if (list_empty(&f->fn_alias)) {
670 hlist_del(&f->fn_hash);
671 kill_f = 1;
673 fib_hash_genid++;
674 write_unlock_bh(&fib_hash_lock);
676 fn_free_alias(fa, f);
677 found++;
680 if (kill_f) {
681 fn_free_node(f);
682 fz->fz_nent--;
685 return found;
688 static int fn_hash_flush(struct fib_table *tb)
690 struct fn_hash *table = (struct fn_hash *) tb->tb_data;
691 struct fn_zone *fz;
692 int found = 0;
694 for (fz = table->fn_zone_list; fz; fz = fz->fz_next) {
695 int i;
697 for (i = fz->fz_divisor - 1; i >= 0; i--)
698 found += fn_flush_list(fz, i);
700 return found;
704 static inline int
705 fn_hash_dump_bucket(struct sk_buff *skb, struct netlink_callback *cb,
706 struct fib_table *tb,
707 struct fn_zone *fz,
708 struct hlist_head *head)
710 struct hlist_node *node;
711 struct fib_node *f;
712 int i, s_i;
714 s_i = cb->args[4];
715 i = 0;
716 hlist_for_each_entry(f, node, head, fn_hash) {
717 struct fib_alias *fa;
719 list_for_each_entry(fa, &f->fn_alias, fa_list) {
720 if (i < s_i)
721 goto next;
723 if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
724 cb->nlh->nlmsg_seq,
725 RTM_NEWROUTE,
726 tb->tb_id,
727 fa->fa_type,
728 fa->fa_scope,
729 f->fn_key,
730 fz->fz_order,
731 fa->fa_tos,
732 fa->fa_info,
733 NLM_F_MULTI) < 0) {
734 cb->args[4] = i;
735 return -1;
737 next:
738 i++;
741 cb->args[4] = i;
742 return skb->len;
745 static inline int
746 fn_hash_dump_zone(struct sk_buff *skb, struct netlink_callback *cb,
747 struct fib_table *tb,
748 struct fn_zone *fz)
750 int h, s_h;
752 if (fz->fz_hash == NULL)
753 return skb->len;
754 s_h = cb->args[3];
755 for (h = s_h; h < fz->fz_divisor; h++) {
756 if (hlist_empty(&fz->fz_hash[h]))
757 continue;
758 if (fn_hash_dump_bucket(skb, cb, tb, fz, &fz->fz_hash[h]) < 0) {
759 cb->args[3] = h;
760 return -1;
762 memset(&cb->args[4], 0,
763 sizeof(cb->args) - 4*sizeof(cb->args[0]));
765 cb->args[3] = h;
766 return skb->len;
769 static int fn_hash_dump(struct fib_table *tb, struct sk_buff *skb, struct netlink_callback *cb)
771 int m, s_m;
772 struct fn_zone *fz;
773 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
775 s_m = cb->args[2];
776 read_lock(&fib_hash_lock);
777 for (fz = table->fn_zone_list, m=0; fz; fz = fz->fz_next, m++) {
778 if (m < s_m) continue;
779 if (fn_hash_dump_zone(skb, cb, tb, fz) < 0) {
780 cb->args[2] = m;
781 read_unlock(&fib_hash_lock);
782 return -1;
784 memset(&cb->args[3], 0,
785 sizeof(cb->args) - 3*sizeof(cb->args[0]));
787 read_unlock(&fib_hash_lock);
788 cb->args[2] = m;
789 return skb->len;
792 void __init fib_hash_init(void)
794 fn_hash_kmem = kmem_cache_create("ip_fib_hash", sizeof(struct fib_node),
795 0, SLAB_PANIC, NULL);
797 fn_alias_kmem = kmem_cache_create("ip_fib_alias", sizeof(struct fib_alias),
798 0, SLAB_PANIC, NULL);
802 struct fib_table *fib_hash_table(u32 id)
804 struct fib_table *tb;
806 tb = kmalloc(sizeof(struct fib_table) + sizeof(struct fn_hash),
807 GFP_KERNEL);
808 if (tb == NULL)
809 return NULL;
811 tb->tb_id = id;
812 tb->tb_default = -1;
813 tb->tb_lookup = fn_hash_lookup;
814 tb->tb_insert = fn_hash_insert;
815 tb->tb_delete = fn_hash_delete;
816 tb->tb_flush = fn_hash_flush;
817 tb->tb_select_default = fn_hash_select_default;
818 tb->tb_dump = fn_hash_dump;
819 memset(tb->tb_data, 0, sizeof(struct fn_hash));
820 return tb;
823 /* ------------------------------------------------------------------------ */
824 #ifdef CONFIG_PROC_FS
826 struct fib_iter_state {
827 struct seq_net_private p;
828 struct fn_zone *zone;
829 int bucket;
830 struct hlist_head *hash_head;
831 struct fib_node *fn;
832 struct fib_alias *fa;
833 loff_t pos;
834 unsigned int genid;
835 int valid;
838 static struct fib_alias *fib_get_first(struct seq_file *seq)
840 struct fib_iter_state *iter = seq->private;
841 struct fib_table *main_table;
842 struct fn_hash *table;
844 main_table = fib_get_table(iter->p.net, RT_TABLE_MAIN);
845 table = (struct fn_hash *)main_table->tb_data;
847 iter->bucket = 0;
848 iter->hash_head = NULL;
849 iter->fn = NULL;
850 iter->fa = NULL;
851 iter->pos = 0;
852 iter->genid = fib_hash_genid;
853 iter->valid = 1;
855 for (iter->zone = table->fn_zone_list; iter->zone;
856 iter->zone = iter->zone->fz_next) {
857 int maxslot;
859 if (!iter->zone->fz_nent)
860 continue;
862 iter->hash_head = iter->zone->fz_hash;
863 maxslot = iter->zone->fz_divisor;
865 for (iter->bucket = 0; iter->bucket < maxslot;
866 ++iter->bucket, ++iter->hash_head) {
867 struct hlist_node *node;
868 struct fib_node *fn;
870 hlist_for_each_entry(fn,node,iter->hash_head,fn_hash) {
871 struct fib_alias *fa;
873 list_for_each_entry(fa,&fn->fn_alias,fa_list) {
874 iter->fn = fn;
875 iter->fa = fa;
876 goto out;
881 out:
882 return iter->fa;
885 static struct fib_alias *fib_get_next(struct seq_file *seq)
887 struct fib_iter_state *iter = seq->private;
888 struct fib_node *fn;
889 struct fib_alias *fa;
891 /* Advance FA, if any. */
892 fn = iter->fn;
893 fa = iter->fa;
894 if (fa) {
895 BUG_ON(!fn);
896 list_for_each_entry_continue(fa, &fn->fn_alias, fa_list) {
897 iter->fa = fa;
898 goto out;
902 fa = iter->fa = NULL;
904 /* Advance FN. */
905 if (fn) {
906 struct hlist_node *node = &fn->fn_hash;
907 hlist_for_each_entry_continue(fn, node, fn_hash) {
908 iter->fn = fn;
910 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
911 iter->fa = fa;
912 goto out;
917 fn = iter->fn = NULL;
919 /* Advance hash chain. */
920 if (!iter->zone)
921 goto out;
923 for (;;) {
924 struct hlist_node *node;
925 int maxslot;
927 maxslot = iter->zone->fz_divisor;
929 while (++iter->bucket < maxslot) {
930 iter->hash_head++;
932 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
933 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
934 iter->fn = fn;
935 iter->fa = fa;
936 goto out;
941 iter->zone = iter->zone->fz_next;
943 if (!iter->zone)
944 goto out;
946 iter->bucket = 0;
947 iter->hash_head = iter->zone->fz_hash;
949 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
950 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
951 iter->fn = fn;
952 iter->fa = fa;
953 goto out;
957 out:
958 iter->pos++;
959 return fa;
962 static struct fib_alias *fib_get_idx(struct seq_file *seq, loff_t pos)
964 struct fib_iter_state *iter = seq->private;
965 struct fib_alias *fa;
967 if (iter->valid && pos >= iter->pos && iter->genid == fib_hash_genid) {
968 fa = iter->fa;
969 pos -= iter->pos;
970 } else
971 fa = fib_get_first(seq);
973 if (fa)
974 while (pos && (fa = fib_get_next(seq)))
975 --pos;
976 return pos ? NULL : fa;
979 static void *fib_seq_start(struct seq_file *seq, loff_t *pos)
980 __acquires(fib_hash_lock)
982 struct fib_iter_state *iter = seq->private;
983 void *v = NULL;
985 read_lock(&fib_hash_lock);
986 if (fib_get_table(iter->p.net, RT_TABLE_MAIN))
987 v = *pos ? fib_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
988 return v;
991 static void *fib_seq_next(struct seq_file *seq, void *v, loff_t *pos)
993 ++*pos;
994 return v == SEQ_START_TOKEN ? fib_get_first(seq) : fib_get_next(seq);
997 static void fib_seq_stop(struct seq_file *seq, void *v)
998 __releases(fib_hash_lock)
1000 read_unlock(&fib_hash_lock);
1003 static unsigned fib_flag_trans(int type, __be32 mask, struct fib_info *fi)
1005 static const unsigned type2flags[RTN_MAX + 1] = {
1006 [7] = RTF_REJECT, [8] = RTF_REJECT,
1008 unsigned flags = type2flags[type];
1010 if (fi && fi->fib_nh->nh_gw)
1011 flags |= RTF_GATEWAY;
1012 if (mask == htonl(0xFFFFFFFF))
1013 flags |= RTF_HOST;
1014 flags |= RTF_UP;
1015 return flags;
1019 * This outputs /proc/net/route.
1021 * It always works in backward compatibility mode.
1022 * The format of the file is not supposed to be changed.
1024 static int fib_seq_show(struct seq_file *seq, void *v)
1026 struct fib_iter_state *iter;
1027 char bf[128];
1028 __be32 prefix, mask;
1029 unsigned flags;
1030 struct fib_node *f;
1031 struct fib_alias *fa;
1032 struct fib_info *fi;
1034 if (v == SEQ_START_TOKEN) {
1035 seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
1036 "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
1037 "\tWindow\tIRTT");
1038 goto out;
1041 iter = seq->private;
1042 f = iter->fn;
1043 fa = iter->fa;
1044 fi = fa->fa_info;
1045 prefix = f->fn_key;
1046 mask = FZ_MASK(iter->zone);
1047 flags = fib_flag_trans(fa->fa_type, mask, fi);
1048 if (fi)
1049 snprintf(bf, sizeof(bf),
1050 "%s\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u",
1051 fi->fib_dev ? fi->fib_dev->name : "*", prefix,
1052 fi->fib_nh->nh_gw, flags, 0, 0, fi->fib_priority,
1053 mask, (fi->fib_advmss ? fi->fib_advmss + 40 : 0),
1054 fi->fib_window,
1055 fi->fib_rtt >> 3);
1056 else
1057 snprintf(bf, sizeof(bf),
1058 "*\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u",
1059 prefix, 0, flags, 0, 0, 0, mask, 0, 0, 0);
1060 seq_printf(seq, "%-127s\n", bf);
1061 out:
1062 return 0;
1065 static const struct seq_operations fib_seq_ops = {
1066 .start = fib_seq_start,
1067 .next = fib_seq_next,
1068 .stop = fib_seq_stop,
1069 .show = fib_seq_show,
1072 static int fib_seq_open(struct inode *inode, struct file *file)
1074 return seq_open_net(inode, file, &fib_seq_ops,
1075 sizeof(struct fib_iter_state));
1078 static const struct file_operations fib_seq_fops = {
1079 .owner = THIS_MODULE,
1080 .open = fib_seq_open,
1081 .read = seq_read,
1082 .llseek = seq_lseek,
1083 .release = seq_release_net,
1086 int __net_init fib_proc_init(struct net *net)
1088 if (!proc_net_fops_create(net, "route", S_IRUGO, &fib_seq_fops))
1089 return -ENOMEM;
1090 return 0;
1093 void __net_exit fib_proc_exit(struct net *net)
1095 proc_net_remove(net, "route");
1097 #endif /* CONFIG_PROC_FS */