allow coexistance of N build and AC build.
[tomato.git] / release / src-rt-6.x / linux / linux-2.6 / net / ipv4 / fib_hash.c
blobd8e98023282e402b4daacdc56112670baf9d7d1d
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/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>
45 #include "fib_lookup.h"
47 static struct kmem_cache *fn_hash_kmem __read_mostly;
48 static struct kmem_cache *fn_alias_kmem __read_mostly;
50 struct fib_node {
51 struct hlist_node fn_hash;
52 struct list_head fn_alias;
53 __be32 fn_key;
54 struct fib_alias fn_embedded_alias;
57 #define EMBEDDED_HASH_SIZE (L1_CACHE_BYTES / sizeof(struct hlist_head))
59 struct fn_zone {
60 struct fn_zone *fz_next; /* Next not empty zone */
61 struct hlist_head *fz_hash; /* Hash table pointer */
62 u32 fz_hashmask; /* (fz_divisor - 1) */
64 u8 fz_order; /* Zone order (0..32) */
65 u8 fz_revorder; /* 32 - fz_order */
66 __be32 fz_mask; /* inet_make_mask(order) */
67 #define FZ_MASK(fz) ((fz)->fz_mask)
69 struct hlist_head fz_embedded_hash[EMBEDDED_HASH_SIZE];
71 int fz_nent; /* Number of entries */
72 int fz_divisor; /* Hash size (mask+1) */
75 struct fn_hash {
76 struct fn_zone *fn_zones[33];
77 struct fn_zone *fn_zone_list;
80 static inline u32 fn_hash(__be32 key, struct fn_zone *fz)
82 u32 h = ntohl(key) >> fz->fz_revorder;
83 h ^= (h>>20);
84 h ^= (h>>10);
85 h ^= (h>>5);
86 h &= fz->fz_hashmask;
87 return h;
90 static inline __be32 fz_key(__be32 dst, struct fn_zone *fz)
92 return dst & FZ_MASK(fz);
95 static DEFINE_RWLOCK(fib_hash_lock);
96 static unsigned int fib_hash_genid;
98 #define FZ_MAX_DIVISOR ((PAGE_SIZE<<MAX_ORDER) / sizeof(struct hlist_head))
100 static struct hlist_head *fz_hash_alloc(int divisor)
102 unsigned long size = divisor * sizeof(struct hlist_head);
104 if (size <= PAGE_SIZE) {
105 return kzalloc(size, GFP_KERNEL);
106 } else {
107 return (struct hlist_head *)
108 __get_free_pages(GFP_KERNEL | __GFP_ZERO, get_order(size));
112 /* The fib hash lock must be held when this is called. */
113 static inline void fn_rebuild_zone(struct fn_zone *fz,
114 struct hlist_head *old_ht,
115 int old_divisor)
117 int i;
119 for (i = 0; i < old_divisor; i++) {
120 struct hlist_node *node, *n;
121 struct fib_node *f;
123 hlist_for_each_entry_safe(f, node, n, &old_ht[i], fn_hash) {
124 struct hlist_head *new_head;
126 hlist_del(&f->fn_hash);
128 new_head = &fz->fz_hash[fn_hash(f->fn_key, fz)];
129 hlist_add_head(&f->fn_hash, new_head);
134 static void fz_hash_free(struct hlist_head *hash, int divisor)
136 unsigned long size = divisor * sizeof(struct hlist_head);
138 if (size <= PAGE_SIZE)
139 kfree(hash);
140 else
141 free_pages((unsigned long)hash, get_order(size));
144 static void fn_rehash_zone(struct fn_zone *fz)
146 struct hlist_head *ht, *old_ht;
147 int old_divisor, new_divisor;
148 u32 new_hashmask;
150 new_divisor = old_divisor = fz->fz_divisor;
152 switch (old_divisor) {
153 case EMBEDDED_HASH_SIZE:
154 new_divisor *= EMBEDDED_HASH_SIZE;
155 break;
156 case EMBEDDED_HASH_SIZE*EMBEDDED_HASH_SIZE:
157 new_divisor *= (EMBEDDED_HASH_SIZE/2);
158 break;
159 default:
160 if ((old_divisor << 1) > FZ_MAX_DIVISOR) {
161 printk(KERN_CRIT "route.c: bad divisor %d!\n", old_divisor);
162 return;
164 new_divisor = (old_divisor << 1);
165 break;
168 new_hashmask = (new_divisor - 1);
170 #if RT_CACHE_DEBUG >= 2
171 printk("fn_rehash_zone: hash for zone %d grows from %d\n", fz->fz_order, old_divisor);
172 #endif
174 ht = fz_hash_alloc(new_divisor);
176 if (ht) {
177 write_lock_bh(&fib_hash_lock);
178 old_ht = fz->fz_hash;
179 fz->fz_hash = ht;
180 fz->fz_hashmask = new_hashmask;
181 fz->fz_divisor = new_divisor;
182 fn_rebuild_zone(fz, old_ht, old_divisor);
183 fib_hash_genid++;
184 write_unlock_bh(&fib_hash_lock);
186 if (old_ht != fz->fz_embedded_hash)
187 fz_hash_free(old_ht, old_divisor);
191 static inline void fn_free_node(struct fib_node * f)
193 kmem_cache_free(fn_hash_kmem, f);
196 static inline void fn_free_alias(struct fib_alias *fa, struct fib_node *f)
198 fib_release_info(fa->fa_info);
199 if (fa == &f->fn_embedded_alias)
200 fa->fa_info = NULL;
201 else
202 kmem_cache_free(fn_alias_kmem, fa);
205 static struct fn_zone *
206 fn_new_zone(struct fn_hash *table, int z)
208 int i;
209 struct fn_zone *fz = kzalloc(sizeof(struct fn_zone), GFP_KERNEL);
210 if (!fz)
211 return NULL;
213 fz->fz_divisor = z ? EMBEDDED_HASH_SIZE : 1;
214 fz->fz_hashmask = fz->fz_divisor - 1;
215 fz->fz_hash = fz->fz_embedded_hash;
216 fz->fz_order = z;
217 fz->fz_revorder = 32 - z;
218 fz->fz_mask = inet_make_mask(z);
220 /* Find the first not empty zone with more specific mask */
221 for (i=z+1; i<=32; i++)
222 if (table->fn_zones[i])
223 break;
224 write_lock_bh(&fib_hash_lock);
225 if (i>32) {
226 /* No more specific masks, we are the first. */
227 fz->fz_next = table->fn_zone_list;
228 table->fn_zone_list = fz;
229 } else {
230 fz->fz_next = table->fn_zones[i]->fz_next;
231 table->fn_zones[i]->fz_next = fz;
233 table->fn_zones[z] = fz;
234 fib_hash_genid++;
235 write_unlock_bh(&fib_hash_lock);
236 return fz;
239 static int
240 fn_hash_lookup(struct fib_table *tb, const struct flowi *flp, struct fib_result *res)
242 int err;
243 struct fn_zone *fz;
244 struct fn_hash *t = (struct fn_hash*)tb->tb_data;
246 read_lock(&fib_hash_lock);
247 for (fz = t->fn_zone_list; fz; fz = fz->fz_next) {
248 struct hlist_head *head;
249 struct hlist_node *node;
250 struct fib_node *f;
251 __be32 k = fz_key(flp->fl4_dst, fz);
253 head = &fz->fz_hash[fn_hash(k, fz)];
254 hlist_for_each_entry(f, node, head, fn_hash) {
255 if (f->fn_key != k)
256 continue;
258 err = fib_semantic_match(&f->fn_alias,
259 flp, res,
260 f->fn_key, fz->fz_mask,
261 fz->fz_order);
262 if (err <= 0)
263 goto out;
266 err = 1;
267 out:
268 read_unlock(&fib_hash_lock);
269 return err;
272 static void
273 fn_hash_select_default(struct fib_table *tb, const struct flowi *flp, struct fib_result *res)
275 int order, last_idx;
276 struct hlist_node *node;
277 struct fib_node *f;
278 struct fib_info *fi = NULL;
279 struct fib_info *last_resort;
280 struct fn_hash *t = (struct fn_hash*)tb->tb_data;
281 struct fn_zone *fz = t->fn_zones[0];
283 if (fz == NULL)
284 return;
286 last_idx = -1;
287 last_resort = NULL;
288 order = -1;
290 read_lock(&fib_hash_lock);
291 hlist_for_each_entry(f, node, &fz->fz_hash[0], fn_hash) {
292 struct fib_alias *fa;
294 list_for_each_entry(fa, &f->fn_alias, fa_list) {
295 struct fib_info *next_fi = fa->fa_info;
297 if (fa->fa_scope != res->scope ||
298 fa->fa_type != RTN_UNICAST)
299 continue;
301 if (next_fi->fib_priority > res->fi->fib_priority)
302 break;
303 if (!next_fi->fib_nh[0].nh_gw ||
304 next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK)
305 continue;
306 fa->fa_state |= FA_S_ACCESSED;
308 if (fi == NULL) {
309 if (next_fi != res->fi)
310 break;
311 } else if (!fib_detect_death(fi, order, &last_resort,
312 &last_idx, &tb->tb_default)) {
313 fib_result_assign(res, fi);
314 tb->tb_default = order;
315 goto out;
317 fi = next_fi;
318 order++;
322 if (order <= 0 || fi == NULL) {
323 tb->tb_default = -1;
324 goto out;
327 if (!fib_detect_death(fi, order, &last_resort, &last_idx, &tb->tb_default)) {
328 fib_result_assign(res, fi);
329 tb->tb_default = order;
330 goto out;
333 if (last_idx >= 0)
334 fib_result_assign(res, last_resort);
335 tb->tb_default = last_idx;
336 out:
337 read_unlock(&fib_hash_lock);
340 /* Insert node F to FZ. */
341 static inline void fib_insert_node(struct fn_zone *fz, struct fib_node *f)
343 struct hlist_head *head = &fz->fz_hash[fn_hash(f->fn_key, fz)];
345 hlist_add_head(&f->fn_hash, head);
348 /* Return the node in FZ matching KEY. */
349 static struct fib_node *fib_find_node(struct fn_zone *fz, __be32 key)
351 struct hlist_head *head = &fz->fz_hash[fn_hash(key, fz)];
352 struct hlist_node *node;
353 struct fib_node *f;
355 hlist_for_each_entry(f, node, head, fn_hash) {
356 if (f->fn_key == key)
357 return f;
360 return NULL;
363 static int fn_hash_insert(struct fib_table *tb, struct fib_config *cfg)
365 struct fn_hash *table = (struct fn_hash *) tb->tb_data;
366 struct fib_node *new_f = NULL;
367 struct fib_node *f;
368 struct fib_alias *fa, *new_fa;
369 struct fn_zone *fz;
370 struct fib_info *fi;
371 u8 tos = cfg->fc_tos;
372 __be32 key;
373 int err;
375 if (cfg->fc_dst_len > 32)
376 return -EINVAL;
378 fz = table->fn_zones[cfg->fc_dst_len];
379 if (!fz && !(fz = fn_new_zone(table, cfg->fc_dst_len)))
380 return -ENOBUFS;
382 key = 0;
383 if (cfg->fc_dst) {
384 if (cfg->fc_dst & ~FZ_MASK(fz))
385 return -EINVAL;
386 key = fz_key(cfg->fc_dst, fz);
389 fi = fib_create_info(cfg);
390 if (IS_ERR(fi))
391 return PTR_ERR(fi);
393 if (fz->fz_nent > (fz->fz_divisor<<1) &&
394 fz->fz_divisor < FZ_MAX_DIVISOR &&
395 (cfg->fc_dst_len == 32 ||
396 (1 << cfg->fc_dst_len) > fz->fz_divisor))
397 fn_rehash_zone(fz);
399 f = fib_find_node(fz, key);
401 if (!f)
402 fa = NULL;
403 else
404 fa = fib_find_alias(&f->fn_alias, tos, fi->fib_priority);
406 /* Now fa, if non-NULL, points to the first fib alias
407 * with the same keys [prefix,tos,priority], if such key already
408 * exists or to the node before which we will insert new one.
410 * If fa is NULL, we will need to allocate a new one and
411 * insert to the head of f.
413 * If f is NULL, no fib node matched the destination key
414 * and we need to allocate a new one of those as well.
417 if (fa && fa->fa_tos == tos &&
418 fa->fa_info->fib_priority == fi->fib_priority) {
419 struct fib_alias *fa_first, *fa_match;
421 err = -EEXIST;
422 if (cfg->fc_nlflags & NLM_F_EXCL)
423 goto out;
425 /* We have 2 goals:
426 * 1. Find exact match for type, scope, fib_info to avoid
427 * duplicate routes
428 * 2. Find next 'fa' (or head), NLM_F_APPEND inserts before it
430 fa_match = NULL;
431 fa_first = fa;
432 fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
433 list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
434 if (fa->fa_tos != tos)
435 break;
436 if (fa->fa_info->fib_priority != fi->fib_priority)
437 break;
438 if (fa->fa_type == cfg->fc_type &&
439 fa->fa_scope == cfg->fc_scope &&
440 fa->fa_info == fi) {
441 fa_match = fa;
442 break;
446 if (cfg->fc_nlflags & NLM_F_REPLACE) {
447 struct fib_info *fi_drop;
448 u8 state;
450 fa = fa_first;
451 if (fa_match) {
452 if (fa == fa_match)
453 err = 0;
454 goto out;
456 write_lock_bh(&fib_hash_lock);
457 fi_drop = fa->fa_info;
458 fa->fa_info = fi;
459 fa->fa_type = cfg->fc_type;
460 fa->fa_scope = cfg->fc_scope;
461 state = fa->fa_state;
462 fa->fa_state &= ~FA_S_ACCESSED;
463 fib_hash_genid++;
464 write_unlock_bh(&fib_hash_lock);
466 fib_release_info(fi_drop);
467 if (state & FA_S_ACCESSED)
468 rt_cache_flush(-1);
469 rtmsg_fib(RTM_NEWROUTE, key, fa, cfg->fc_dst_len, tb->tb_id,
470 &cfg->fc_nlinfo, NLM_F_REPLACE);
471 return 0;
474 /* Error if we find a perfect match which
475 * uses the same scope, type, and nexthop
476 * information.
478 if (fa_match)
479 goto out;
481 if (!(cfg->fc_nlflags & NLM_F_APPEND))
482 fa = fa_first;
485 err = -ENOENT;
486 if (!(cfg->fc_nlflags & NLM_F_CREATE))
487 goto out;
489 err = -ENOBUFS;
491 if (!f) {
492 new_f = kmem_cache_zalloc(fn_hash_kmem, GFP_KERNEL);
493 if (new_f == NULL)
494 goto out;
496 INIT_HLIST_NODE(&new_f->fn_hash);
497 INIT_LIST_HEAD(&new_f->fn_alias);
498 new_f->fn_key = key;
499 f = new_f;
502 new_fa = &f->fn_embedded_alias;
503 if (new_fa->fa_info != NULL) {
504 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
505 if (new_fa == NULL)
506 goto out;
508 new_fa->fa_info = fi;
509 new_fa->fa_tos = tos;
510 new_fa->fa_type = cfg->fc_type;
511 new_fa->fa_scope = cfg->fc_scope;
512 new_fa->fa_state = 0;
515 * Insert new entry to the list.
518 write_lock_bh(&fib_hash_lock);
519 if (new_f)
520 fib_insert_node(fz, new_f);
521 list_add_tail(&new_fa->fa_list,
522 (fa ? &fa->fa_list : &f->fn_alias));
523 fib_hash_genid++;
524 write_unlock_bh(&fib_hash_lock);
526 if (new_f)
527 fz->fz_nent++;
528 rt_cache_flush(-1);
530 rtmsg_fib(RTM_NEWROUTE, key, new_fa, cfg->fc_dst_len, tb->tb_id,
531 &cfg->fc_nlinfo, 0);
532 return 0;
534 out:
535 if (new_f)
536 kmem_cache_free(fn_hash_kmem, new_f);
537 fib_release_info(fi);
538 return err;
542 static int fn_hash_delete(struct fib_table *tb, struct fib_config *cfg)
544 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
545 struct fib_node *f;
546 struct fib_alias *fa, *fa_to_delete;
547 struct fn_zone *fz;
548 __be32 key;
550 if (cfg->fc_dst_len > 32)
551 return -EINVAL;
553 if ((fz = table->fn_zones[cfg->fc_dst_len]) == NULL)
554 return -ESRCH;
556 key = 0;
557 if (cfg->fc_dst) {
558 if (cfg->fc_dst & ~FZ_MASK(fz))
559 return -EINVAL;
560 key = fz_key(cfg->fc_dst, fz);
563 f = fib_find_node(fz, key);
565 if (!f)
566 fa = NULL;
567 else
568 fa = fib_find_alias(&f->fn_alias, cfg->fc_tos, 0);
569 if (!fa)
570 return -ESRCH;
572 fa_to_delete = NULL;
573 fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
574 list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
575 struct fib_info *fi = fa->fa_info;
577 if (fa->fa_tos != cfg->fc_tos)
578 break;
580 if ((!cfg->fc_type ||
581 fa->fa_type == cfg->fc_type) &&
582 (cfg->fc_scope == RT_SCOPE_NOWHERE ||
583 fa->fa_scope == cfg->fc_scope) &&
584 (!cfg->fc_protocol ||
585 fi->fib_protocol == cfg->fc_protocol) &&
586 fib_nh_match(cfg, fi) == 0) {
587 fa_to_delete = fa;
588 break;
592 if (fa_to_delete) {
593 int kill_fn;
595 fa = fa_to_delete;
596 rtmsg_fib(RTM_DELROUTE, key, fa, cfg->fc_dst_len,
597 tb->tb_id, &cfg->fc_nlinfo, 0);
599 kill_fn = 0;
600 write_lock_bh(&fib_hash_lock);
601 list_del(&fa->fa_list);
602 if (list_empty(&f->fn_alias)) {
603 hlist_del(&f->fn_hash);
604 kill_fn = 1;
606 fib_hash_genid++;
607 write_unlock_bh(&fib_hash_lock);
609 if (fa->fa_state & FA_S_ACCESSED)
610 rt_cache_flush(-1);
611 fn_free_alias(fa, f);
612 if (kill_fn) {
613 fn_free_node(f);
614 fz->fz_nent--;
617 return 0;
619 return -ESRCH;
622 static int fn_flush_list(struct fn_zone *fz, int idx)
624 struct hlist_head *head = &fz->fz_hash[idx];
625 struct hlist_node *node, *n;
626 struct fib_node *f;
627 int found = 0;
629 hlist_for_each_entry_safe(f, node, n, head, fn_hash) {
630 struct fib_alias *fa, *fa_node;
631 int kill_f;
633 kill_f = 0;
634 list_for_each_entry_safe(fa, fa_node, &f->fn_alias, fa_list) {
635 struct fib_info *fi = fa->fa_info;
637 if (fi && (fi->fib_flags&RTNH_F_DEAD)) {
638 write_lock_bh(&fib_hash_lock);
639 list_del(&fa->fa_list);
640 if (list_empty(&f->fn_alias)) {
641 hlist_del(&f->fn_hash);
642 kill_f = 1;
644 fib_hash_genid++;
645 write_unlock_bh(&fib_hash_lock);
647 fn_free_alias(fa, f);
648 found++;
651 if (kill_f) {
652 fn_free_node(f);
653 fz->fz_nent--;
656 return found;
659 static int fn_hash_flush(struct fib_table *tb)
661 struct fn_hash *table = (struct fn_hash *) tb->tb_data;
662 struct fn_zone *fz;
663 int found = 0;
665 for (fz = table->fn_zone_list; fz; fz = fz->fz_next) {
666 int i;
668 for (i = fz->fz_divisor - 1; i >= 0; i--)
669 found += fn_flush_list(fz, i);
671 return found;
675 static inline int
676 fn_hash_dump_bucket(struct sk_buff *skb, struct netlink_callback *cb,
677 struct fib_table *tb,
678 struct fn_zone *fz,
679 struct hlist_head *head)
681 struct hlist_node *node;
682 struct fib_node *f;
683 int i, s_i;
685 s_i = cb->args[4];
686 i = 0;
687 hlist_for_each_entry(f, node, head, fn_hash) {
688 struct fib_alias *fa;
690 list_for_each_entry(fa, &f->fn_alias, fa_list) {
691 if (i < s_i)
692 goto next;
694 if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
695 cb->nlh->nlmsg_seq,
696 RTM_NEWROUTE,
697 tb->tb_id,
698 fa->fa_type,
699 fa->fa_scope,
700 f->fn_key,
701 fz->fz_order,
702 fa->fa_tos,
703 fa->fa_info,
704 NLM_F_MULTI) < 0) {
705 cb->args[4] = i;
706 return -1;
708 next:
709 i++;
712 cb->args[4] = i;
713 return skb->len;
716 static inline int
717 fn_hash_dump_zone(struct sk_buff *skb, struct netlink_callback *cb,
718 struct fib_table *tb,
719 struct fn_zone *fz)
721 int h, s_h;
723 s_h = cb->args[3];
724 for (h=0; h < fz->fz_divisor; h++) {
725 if (h < s_h) continue;
726 if (h > s_h)
727 memset(&cb->args[4], 0,
728 sizeof(cb->args) - 4*sizeof(cb->args[0]));
729 if (fz->fz_hash == NULL ||
730 hlist_empty(&fz->fz_hash[h]))
731 continue;
732 if (fn_hash_dump_bucket(skb, cb, tb, fz, &fz->fz_hash[h])<0) {
733 cb->args[3] = h;
734 return -1;
737 cb->args[3] = h;
738 return skb->len;
741 static int fn_hash_dump(struct fib_table *tb, struct sk_buff *skb, struct netlink_callback *cb)
743 int m, s_m;
744 struct fn_zone *fz;
745 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
747 s_m = cb->args[2];
748 read_lock(&fib_hash_lock);
749 for (fz = table->fn_zone_list, m=0; fz; fz = fz->fz_next, m++) {
750 if (m < s_m) continue;
751 if (m > s_m)
752 memset(&cb->args[3], 0,
753 sizeof(cb->args) - 3*sizeof(cb->args[0]));
754 if (fn_hash_dump_zone(skb, cb, tb, fz) < 0) {
755 cb->args[2] = m;
756 read_unlock(&fib_hash_lock);
757 return -1;
760 read_unlock(&fib_hash_lock);
761 cb->args[2] = m;
762 return skb->len;
765 #ifdef CONFIG_IP_MULTIPLE_TABLES
766 struct fib_table * fib_hash_init(u32 id)
767 #else
768 struct fib_table * __init fib_hash_init(u32 id)
769 #endif
771 struct fib_table *tb;
773 if (fn_hash_kmem == NULL)
774 fn_hash_kmem = kmem_cache_create("ip_fib_hash",
775 sizeof(struct fib_node),
776 0, SLAB_PANIC,
777 NULL, NULL);
779 if (fn_alias_kmem == NULL)
780 fn_alias_kmem = kmem_cache_create("ip_fib_alias",
781 sizeof(struct fib_alias),
782 0, SLAB_PANIC,
783 NULL, NULL);
785 tb = kmalloc(sizeof(struct fib_table) + sizeof(struct fn_hash),
786 GFP_KERNEL);
787 if (tb == NULL)
788 return NULL;
790 tb->tb_id = id;
791 tb->tb_default = -1;
792 tb->tb_lookup = fn_hash_lookup;
793 tb->tb_insert = fn_hash_insert;
794 tb->tb_delete = fn_hash_delete;
795 tb->tb_flush = fn_hash_flush;
796 tb->tb_select_default = fn_hash_select_default;
797 tb->tb_dump = fn_hash_dump;
798 memset(tb->tb_data, 0, sizeof(struct fn_hash));
799 return tb;
802 /* ------------------------------------------------------------------------ */
803 #ifdef CONFIG_PROC_FS
805 struct fib_iter_state {
806 struct fn_zone *zone;
807 int bucket;
808 struct hlist_head *hash_head;
809 struct fib_node *fn;
810 struct fib_alias *fa;
811 loff_t pos;
812 unsigned int genid;
813 int valid;
816 static struct fib_alias *fib_get_first(struct seq_file *seq)
818 struct fib_iter_state *iter = seq->private;
819 struct fn_hash *table = (struct fn_hash *) ip_fib_main_table->tb_data;
821 iter->bucket = 0;
822 iter->hash_head = NULL;
823 iter->fn = NULL;
824 iter->fa = NULL;
825 iter->pos = 0;
826 iter->genid = fib_hash_genid;
827 iter->valid = 1;
829 for (iter->zone = table->fn_zone_list; iter->zone;
830 iter->zone = iter->zone->fz_next) {
831 int maxslot;
833 if (!iter->zone->fz_nent)
834 continue;
836 iter->hash_head = iter->zone->fz_hash;
837 maxslot = iter->zone->fz_divisor;
839 for (iter->bucket = 0; iter->bucket < maxslot;
840 ++iter->bucket, ++iter->hash_head) {
841 struct hlist_node *node;
842 struct fib_node *fn;
844 hlist_for_each_entry(fn,node,iter->hash_head,fn_hash) {
845 struct fib_alias *fa;
847 list_for_each_entry(fa,&fn->fn_alias,fa_list) {
848 iter->fn = fn;
849 iter->fa = fa;
850 goto out;
855 out:
856 return iter->fa;
859 static struct fib_alias *fib_get_next(struct seq_file *seq)
861 struct fib_iter_state *iter = seq->private;
862 struct fib_node *fn;
863 struct fib_alias *fa;
865 /* Advance FA, if any. */
866 fn = iter->fn;
867 fa = iter->fa;
868 if (fa) {
869 BUG_ON(!fn);
870 list_for_each_entry_continue(fa, &fn->fn_alias, fa_list) {
871 iter->fa = fa;
872 goto out;
876 fa = iter->fa = NULL;
878 /* Advance FN. */
879 if (fn) {
880 struct hlist_node *node = &fn->fn_hash;
881 hlist_for_each_entry_continue(fn, node, fn_hash) {
882 iter->fn = fn;
884 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
885 iter->fa = fa;
886 goto out;
891 fn = iter->fn = NULL;
893 /* Advance hash chain. */
894 if (!iter->zone)
895 goto out;
897 for (;;) {
898 struct hlist_node *node;
899 int maxslot;
901 maxslot = iter->zone->fz_divisor;
903 while (++iter->bucket < maxslot) {
904 iter->hash_head++;
906 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
907 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
908 iter->fn = fn;
909 iter->fa = fa;
910 goto out;
915 iter->zone = iter->zone->fz_next;
917 if (!iter->zone)
918 goto out;
920 iter->bucket = 0;
921 iter->hash_head = iter->zone->fz_hash;
923 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
924 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
925 iter->fn = fn;
926 iter->fa = fa;
927 goto out;
931 out:
932 iter->pos++;
933 return fa;
936 static struct fib_alias *fib_get_idx(struct seq_file *seq, loff_t pos)
938 struct fib_iter_state *iter = seq->private;
939 struct fib_alias *fa;
941 if (iter->valid && pos >= iter->pos && iter->genid == fib_hash_genid) {
942 fa = iter->fa;
943 pos -= iter->pos;
944 } else
945 fa = fib_get_first(seq);
947 if (fa)
948 while (pos && (fa = fib_get_next(seq)))
949 --pos;
950 return pos ? NULL : fa;
953 static void *fib_seq_start(struct seq_file *seq, loff_t *pos)
955 void *v = NULL;
957 read_lock(&fib_hash_lock);
958 if (ip_fib_main_table)
959 v = *pos ? fib_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
960 return v;
963 static void *fib_seq_next(struct seq_file *seq, void *v, loff_t *pos)
965 ++*pos;
966 return v == SEQ_START_TOKEN ? fib_get_first(seq) : fib_get_next(seq);
969 static void fib_seq_stop(struct seq_file *seq, void *v)
971 read_unlock(&fib_hash_lock);
974 static unsigned fib_flag_trans(int type, __be32 mask, struct fib_info *fi)
976 static const unsigned type2flags[RTN_MAX + 1] = {
977 [7] = RTF_REJECT, [8] = RTF_REJECT,
979 unsigned flags = type2flags[type];
981 if (fi && fi->fib_nh->nh_gw)
982 flags |= RTF_GATEWAY;
983 if (mask == htonl(0xFFFFFFFF))
984 flags |= RTF_HOST;
985 flags |= RTF_UP;
986 return flags;
990 * This outputs /proc/net/route.
992 * It always works in backward compatibility mode.
993 * The format of the file is not supposed to be changed.
995 static int fib_seq_show(struct seq_file *seq, void *v)
997 struct fib_iter_state *iter;
998 int len;
999 __be32 prefix, mask;
1000 unsigned flags;
1001 struct fib_node *f;
1002 struct fib_alias *fa;
1003 struct fib_info *fi;
1005 if (v == SEQ_START_TOKEN) {
1006 seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
1007 "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
1008 "\tWindow\tIRTT");
1009 goto out;
1012 iter = seq->private;
1013 f = iter->fn;
1014 fa = iter->fa;
1015 fi = fa->fa_info;
1016 prefix = f->fn_key;
1017 mask = FZ_MASK(iter->zone);
1018 flags = fib_flag_trans(fa->fa_type, mask, fi);
1019 if (fi)
1020 seq_printf(seq,
1021 "%s\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u%n",
1022 fi->fib_dev ? fi->fib_dev->name : "*", prefix,
1023 fi->fib_nh->nh_gw, flags, 0, 0, fi->fib_priority,
1024 mask, (fi->fib_advmss ? fi->fib_advmss + 40 : 0),
1025 fi->fib_window,
1026 fi->fib_rtt >> 3, &len);
1027 else
1028 seq_printf(seq,
1029 "*\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u%n",
1030 prefix, 0, flags, 0, 0, 0, mask, 0, 0, 0, &len);
1032 seq_printf(seq, "%*s\n", 127 - len, "");
1033 out:
1034 return 0;
1037 static const struct seq_operations fib_seq_ops = {
1038 .start = fib_seq_start,
1039 .next = fib_seq_next,
1040 .stop = fib_seq_stop,
1041 .show = fib_seq_show,
1044 static int fib_seq_open(struct inode *inode, struct file *file)
1046 return seq_open_private(file, &fib_seq_ops,
1047 sizeof(struct fib_iter_state));
1050 static const struct file_operations fib_seq_fops = {
1051 .owner = THIS_MODULE,
1052 .open = fib_seq_open,
1053 .read = seq_read,
1054 .llseek = seq_lseek,
1055 .release = seq_release_private,
1058 int __init fib_proc_init(void)
1060 if (!proc_net_fops_create("route", S_IRUGO, &fib_seq_fops))
1061 return -ENOMEM;
1062 return 0;
1065 void __init fib_proc_exit(void)
1067 proc_net_remove("route");
1069 #endif /* CONFIG_PROC_FS */