intel-agp: Switch to wbinvd_on_all_cpus
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv4 / fib_hash.c
blob14972017b9c2ad727369a124f95116ff95af430d
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 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 #include <asm/uaccess.h>
17 #include <asm/system.h>
18 #include <linux/bitops.h>
19 #include <linux/types.h>
20 #include <linux/kernel.h>
21 #include <linux/mm.h>
22 #include <linux/string.h>
23 #include <linux/socket.h>
24 #include <linux/sockios.h>
25 #include <linux/errno.h>
26 #include <linux/in.h>
27 #include <linux/inet.h>
28 #include <linux/inetdevice.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/proc_fs.h>
32 #include <linux/skbuff.h>
33 #include <linux/netlink.h>
34 #include <linux/init.h>
36 #include <net/net_namespace.h>
37 #include <net/ip.h>
38 #include <net/protocol.h>
39 #include <net/route.h>
40 #include <net/tcp.h>
41 #include <net/sock.h>
42 #include <net/ip_fib.h>
44 #include "fib_lookup.h"
46 static struct kmem_cache *fn_hash_kmem __read_mostly;
47 static struct kmem_cache *fn_alias_kmem __read_mostly;
49 struct fib_node {
50 struct hlist_node fn_hash;
51 struct list_head fn_alias;
52 __be32 fn_key;
53 struct fib_alias fn_embedded_alias;
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 */
61 int fz_divisor; /* Hash divisor */
62 u32 fz_hashmask; /* (fz_divisor - 1) */
63 #define FZ_HASHMASK(fz) ((fz)->fz_hashmask)
65 int fz_order; /* Zone order */
66 __be32 fz_mask;
67 #define FZ_MASK(fz) ((fz)->fz_mask)
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.
74 struct fn_hash {
75 struct fn_zone *fn_zones[33];
76 struct fn_zone *fn_zone_list;
79 static inline u32 fn_hash(__be32 key, struct fn_zone *fz)
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;
89 static inline __be32 fz_key(__be32 dst, struct fn_zone *fz)
91 return dst & FZ_MASK(fz);
94 static DEFINE_RWLOCK(fib_hash_lock);
95 static unsigned int fib_hash_genid;
97 #define FZ_MAX_DIVISOR ((PAGE_SIZE<<MAX_ORDER) / sizeof(struct hlist_head))
99 static struct hlist_head *fz_hash_alloc(int divisor)
101 unsigned long size = divisor * sizeof(struct hlist_head);
103 if (size <= PAGE_SIZE) {
104 return kzalloc(size, GFP_KERNEL);
105 } else {
106 return (struct hlist_head *)
107 __get_free_pages(GFP_KERNEL | __GFP_ZERO, get_order(size));
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)
116 int i;
118 for (i = 0; i < old_divisor; i++) {
119 struct hlist_node *node, *n;
120 struct fib_node *f;
122 hlist_for_each_entry_safe(f, node, n, &old_ht[i], fn_hash) {
123 struct hlist_head *new_head;
125 hlist_del(&f->fn_hash);
127 new_head = &fz->fz_hash[fn_hash(f->fn_key, fz)];
128 hlist_add_head(&f->fn_hash, new_head);
133 static void fz_hash_free(struct hlist_head *hash, int divisor)
135 unsigned long size = divisor * sizeof(struct hlist_head);
137 if (size <= PAGE_SIZE)
138 kfree(hash);
139 else
140 free_pages((unsigned long)hash, get_order(size));
143 static void fn_rehash_zone(struct fn_zone *fz)
145 struct hlist_head *ht, *old_ht;
146 int old_divisor, new_divisor;
147 u32 new_hashmask;
149 old_divisor = fz->fz_divisor;
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;
163 new_divisor = (old_divisor << 1);
164 break;
167 new_hashmask = (new_divisor - 1);
169 #if RT_CACHE_DEBUG >= 2
170 printk(KERN_DEBUG "fn_rehash_zone: hash for zone %d grows from %d\n",
171 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 fz_hash_free(old_ht, old_divisor);
190 static inline void fn_free_node(struct fib_node * f)
192 kmem_cache_free(fn_hash_kmem, f);
195 static inline void fn_free_alias(struct fib_alias *fa, struct fib_node *f)
197 fib_release_info(fa->fa_info);
198 if (fa == &f->fn_embedded_alias)
199 fa->fa_info = NULL;
200 else
201 kmem_cache_free(fn_alias_kmem, fa);
204 static struct fn_zone *
205 fn_new_zone(struct fn_hash *table, int z)
207 int i;
208 struct fn_zone *fz = kzalloc(sizeof(struct fn_zone), GFP_KERNEL);
209 if (!fz)
210 return NULL;
212 if (z) {
213 fz->fz_divisor = 16;
214 } else {
215 fz->fz_divisor = 1;
217 fz->fz_hashmask = (fz->fz_divisor - 1);
218 fz->fz_hash = fz_hash_alloc(fz->fz_divisor);
219 if (!fz->fz_hash) {
220 kfree(fz);
221 return NULL;
223 fz->fz_order = z;
224 fz->fz_mask = inet_make_mask(z);
226 /* Find the first not empty zone with more specific mask */
227 for (i=z+1; i<=32; i++)
228 if (table->fn_zones[i])
229 break;
230 write_lock_bh(&fib_hash_lock);
231 if (i>32) {
232 /* No more specific masks, we are the first. */
233 fz->fz_next = table->fn_zone_list;
234 table->fn_zone_list = fz;
235 } else {
236 fz->fz_next = table->fn_zones[i]->fz_next;
237 table->fn_zones[i]->fz_next = fz;
239 table->fn_zones[z] = fz;
240 fib_hash_genid++;
241 write_unlock_bh(&fib_hash_lock);
242 return fz;
245 int fib_table_lookup(struct fib_table *tb,
246 const struct flowi *flp, struct fib_result *res)
248 int err;
249 struct fn_zone *fz;
250 struct fn_hash *t = (struct fn_hash *)tb->tb_data;
252 read_lock(&fib_hash_lock);
253 for (fz = t->fn_zone_list; fz; fz = fz->fz_next) {
254 struct hlist_head *head;
255 struct hlist_node *node;
256 struct fib_node *f;
257 __be32 k = fz_key(flp->fl4_dst, fz);
259 head = &fz->fz_hash[fn_hash(k, fz)];
260 hlist_for_each_entry(f, node, head, fn_hash) {
261 if (f->fn_key != k)
262 continue;
264 err = fib_semantic_match(&f->fn_alias,
265 flp, res,
266 fz->fz_order);
267 if (err <= 0)
268 goto out;
271 err = 1;
272 out:
273 read_unlock(&fib_hash_lock);
274 return err;
277 void fib_table_select_default(struct fib_table *tb,
278 const struct flowi *flp, struct fib_result *res)
280 int order, last_idx;
281 struct hlist_node *node;
282 struct fib_node *f;
283 struct fib_info *fi = NULL;
284 struct fib_info *last_resort;
285 struct fn_hash *t = (struct fn_hash *)tb->tb_data;
286 struct fn_zone *fz = t->fn_zones[0];
288 if (fz == NULL)
289 return;
291 last_idx = -1;
292 last_resort = NULL;
293 order = -1;
295 read_lock(&fib_hash_lock);
296 hlist_for_each_entry(f, node, &fz->fz_hash[0], fn_hash) {
297 struct fib_alias *fa;
299 list_for_each_entry(fa, &f->fn_alias, fa_list) {
300 struct fib_info *next_fi = fa->fa_info;
302 if (fa->fa_scope != res->scope ||
303 fa->fa_type != RTN_UNICAST)
304 continue;
306 if (next_fi->fib_priority > res->fi->fib_priority)
307 break;
308 if (!next_fi->fib_nh[0].nh_gw ||
309 next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK)
310 continue;
311 fa->fa_state |= FA_S_ACCESSED;
313 if (fi == NULL) {
314 if (next_fi != res->fi)
315 break;
316 } else if (!fib_detect_death(fi, order, &last_resort,
317 &last_idx, tb->tb_default)) {
318 fib_result_assign(res, fi);
319 tb->tb_default = order;
320 goto out;
322 fi = next_fi;
323 order++;
327 if (order <= 0 || fi == NULL) {
328 tb->tb_default = -1;
329 goto out;
332 if (!fib_detect_death(fi, order, &last_resort, &last_idx,
333 tb->tb_default)) {
334 fib_result_assign(res, fi);
335 tb->tb_default = order;
336 goto out;
339 if (last_idx >= 0)
340 fib_result_assign(res, last_resort);
341 tb->tb_default = last_idx;
342 out:
343 read_unlock(&fib_hash_lock);
346 /* Insert node F to FZ. */
347 static inline void fib_insert_node(struct fn_zone *fz, struct fib_node *f)
349 struct hlist_head *head = &fz->fz_hash[fn_hash(f->fn_key, fz)];
351 hlist_add_head(&f->fn_hash, head);
354 /* Return the node in FZ matching KEY. */
355 static struct fib_node *fib_find_node(struct fn_zone *fz, __be32 key)
357 struct hlist_head *head = &fz->fz_hash[fn_hash(key, fz)];
358 struct hlist_node *node;
359 struct fib_node *f;
361 hlist_for_each_entry(f, node, head, fn_hash) {
362 if (f->fn_key == key)
363 return f;
366 return NULL;
369 int fib_table_insert(struct fib_table *tb, struct fib_config *cfg)
371 struct fn_hash *table = (struct fn_hash *) tb->tb_data;
372 struct fib_node *new_f = NULL;
373 struct fib_node *f;
374 struct fib_alias *fa, *new_fa;
375 struct fn_zone *fz;
376 struct fib_info *fi;
377 u8 tos = cfg->fc_tos;
378 __be32 key;
379 int err;
381 if (cfg->fc_dst_len > 32)
382 return -EINVAL;
384 fz = table->fn_zones[cfg->fc_dst_len];
385 if (!fz && !(fz = fn_new_zone(table, cfg->fc_dst_len)))
386 return -ENOBUFS;
388 key = 0;
389 if (cfg->fc_dst) {
390 if (cfg->fc_dst & ~FZ_MASK(fz))
391 return -EINVAL;
392 key = fz_key(cfg->fc_dst, fz);
395 fi = fib_create_info(cfg);
396 if (IS_ERR(fi))
397 return PTR_ERR(fi);
399 if (fz->fz_nent > (fz->fz_divisor<<1) &&
400 fz->fz_divisor < FZ_MAX_DIVISOR &&
401 (cfg->fc_dst_len == 32 ||
402 (1 << cfg->fc_dst_len) > fz->fz_divisor))
403 fn_rehash_zone(fz);
405 f = fib_find_node(fz, key);
407 if (!f)
408 fa = NULL;
409 else
410 fa = fib_find_alias(&f->fn_alias, tos, fi->fib_priority);
412 /* Now fa, if non-NULL, points to the first fib alias
413 * with the same keys [prefix,tos,priority], if such key already
414 * exists or to the node before which we will insert new one.
416 * If fa is NULL, we will need to allocate a new one and
417 * insert to the head of f.
419 * If f is NULL, no fib node matched the destination key
420 * and we need to allocate a new one of those as well.
423 if (fa && fa->fa_tos == tos &&
424 fa->fa_info->fib_priority == fi->fib_priority) {
425 struct fib_alias *fa_first, *fa_match;
427 err = -EEXIST;
428 if (cfg->fc_nlflags & NLM_F_EXCL)
429 goto out;
431 /* We have 2 goals:
432 * 1. Find exact match for type, scope, fib_info to avoid
433 * duplicate routes
434 * 2. Find next 'fa' (or head), NLM_F_APPEND inserts before it
436 fa_match = NULL;
437 fa_first = fa;
438 fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
439 list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
440 if (fa->fa_tos != tos)
441 break;
442 if (fa->fa_info->fib_priority != fi->fib_priority)
443 break;
444 if (fa->fa_type == cfg->fc_type &&
445 fa->fa_scope == cfg->fc_scope &&
446 fa->fa_info == fi) {
447 fa_match = fa;
448 break;
452 if (cfg->fc_nlflags & NLM_F_REPLACE) {
453 struct fib_info *fi_drop;
454 u8 state;
456 fa = fa_first;
457 if (fa_match) {
458 if (fa == fa_match)
459 err = 0;
460 goto out;
462 write_lock_bh(&fib_hash_lock);
463 fi_drop = fa->fa_info;
464 fa->fa_info = fi;
465 fa->fa_type = cfg->fc_type;
466 fa->fa_scope = cfg->fc_scope;
467 state = fa->fa_state;
468 fa->fa_state &= ~FA_S_ACCESSED;
469 fib_hash_genid++;
470 write_unlock_bh(&fib_hash_lock);
472 fib_release_info(fi_drop);
473 if (state & FA_S_ACCESSED)
474 rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
475 rtmsg_fib(RTM_NEWROUTE, key, fa, cfg->fc_dst_len, tb->tb_id,
476 &cfg->fc_nlinfo, NLM_F_REPLACE);
477 return 0;
480 /* Error if we find a perfect match which
481 * uses the same scope, type, and nexthop
482 * information.
484 if (fa_match)
485 goto out;
487 if (!(cfg->fc_nlflags & NLM_F_APPEND))
488 fa = fa_first;
491 err = -ENOENT;
492 if (!(cfg->fc_nlflags & NLM_F_CREATE))
493 goto out;
495 err = -ENOBUFS;
497 if (!f) {
498 new_f = kmem_cache_zalloc(fn_hash_kmem, GFP_KERNEL);
499 if (new_f == NULL)
500 goto out;
502 INIT_HLIST_NODE(&new_f->fn_hash);
503 INIT_LIST_HEAD(&new_f->fn_alias);
504 new_f->fn_key = key;
505 f = new_f;
508 new_fa = &f->fn_embedded_alias;
509 if (new_fa->fa_info != NULL) {
510 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
511 if (new_fa == NULL)
512 goto out;
514 new_fa->fa_info = fi;
515 new_fa->fa_tos = tos;
516 new_fa->fa_type = cfg->fc_type;
517 new_fa->fa_scope = cfg->fc_scope;
518 new_fa->fa_state = 0;
521 * Insert new entry to the list.
524 write_lock_bh(&fib_hash_lock);
525 if (new_f)
526 fib_insert_node(fz, new_f);
527 list_add_tail(&new_fa->fa_list,
528 (fa ? &fa->fa_list : &f->fn_alias));
529 fib_hash_genid++;
530 write_unlock_bh(&fib_hash_lock);
532 if (new_f)
533 fz->fz_nent++;
534 rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
536 rtmsg_fib(RTM_NEWROUTE, key, new_fa, cfg->fc_dst_len, tb->tb_id,
537 &cfg->fc_nlinfo, 0);
538 return 0;
540 out:
541 if (new_f)
542 kmem_cache_free(fn_hash_kmem, new_f);
543 fib_release_info(fi);
544 return err;
547 int fib_table_delete(struct fib_table *tb, struct fib_config *cfg)
549 struct fn_hash *table = (struct fn_hash *)tb->tb_data;
550 struct fib_node *f;
551 struct fib_alias *fa, *fa_to_delete;
552 struct fn_zone *fz;
553 __be32 key;
555 if (cfg->fc_dst_len > 32)
556 return -EINVAL;
558 if ((fz = table->fn_zones[cfg->fc_dst_len]) == NULL)
559 return -ESRCH;
561 key = 0;
562 if (cfg->fc_dst) {
563 if (cfg->fc_dst & ~FZ_MASK(fz))
564 return -EINVAL;
565 key = fz_key(cfg->fc_dst, fz);
568 f = fib_find_node(fz, key);
570 if (!f)
571 fa = NULL;
572 else
573 fa = fib_find_alias(&f->fn_alias, cfg->fc_tos, 0);
574 if (!fa)
575 return -ESRCH;
577 fa_to_delete = NULL;
578 fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
579 list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
580 struct fib_info *fi = fa->fa_info;
582 if (fa->fa_tos != cfg->fc_tos)
583 break;
585 if ((!cfg->fc_type ||
586 fa->fa_type == cfg->fc_type) &&
587 (cfg->fc_scope == RT_SCOPE_NOWHERE ||
588 fa->fa_scope == cfg->fc_scope) &&
589 (!cfg->fc_protocol ||
590 fi->fib_protocol == cfg->fc_protocol) &&
591 fib_nh_match(cfg, fi) == 0) {
592 fa_to_delete = fa;
593 break;
597 if (fa_to_delete) {
598 int kill_fn;
600 fa = fa_to_delete;
601 rtmsg_fib(RTM_DELROUTE, key, fa, cfg->fc_dst_len,
602 tb->tb_id, &cfg->fc_nlinfo, 0);
604 kill_fn = 0;
605 write_lock_bh(&fib_hash_lock);
606 list_del(&fa->fa_list);
607 if (list_empty(&f->fn_alias)) {
608 hlist_del(&f->fn_hash);
609 kill_fn = 1;
611 fib_hash_genid++;
612 write_unlock_bh(&fib_hash_lock);
614 if (fa->fa_state & FA_S_ACCESSED)
615 rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
616 fn_free_alias(fa, f);
617 if (kill_fn) {
618 fn_free_node(f);
619 fz->fz_nent--;
622 return 0;
624 return -ESRCH;
627 static int fn_flush_list(struct fn_zone *fz, int idx)
629 struct hlist_head *head = &fz->fz_hash[idx];
630 struct hlist_node *node, *n;
631 struct fib_node *f;
632 int found = 0;
634 hlist_for_each_entry_safe(f, node, n, head, fn_hash) {
635 struct fib_alias *fa, *fa_node;
636 int kill_f;
638 kill_f = 0;
639 list_for_each_entry_safe(fa, fa_node, &f->fn_alias, fa_list) {
640 struct fib_info *fi = fa->fa_info;
642 if (fi && (fi->fib_flags&RTNH_F_DEAD)) {
643 write_lock_bh(&fib_hash_lock);
644 list_del(&fa->fa_list);
645 if (list_empty(&f->fn_alias)) {
646 hlist_del(&f->fn_hash);
647 kill_f = 1;
649 fib_hash_genid++;
650 write_unlock_bh(&fib_hash_lock);
652 fn_free_alias(fa, f);
653 found++;
656 if (kill_f) {
657 fn_free_node(f);
658 fz->fz_nent--;
661 return found;
664 int fib_table_flush(struct fib_table *tb)
666 struct fn_hash *table = (struct fn_hash *) tb->tb_data;
667 struct fn_zone *fz;
668 int found = 0;
670 for (fz = table->fn_zone_list; fz; fz = fz->fz_next) {
671 int i;
673 for (i = fz->fz_divisor - 1; i >= 0; i--)
674 found += fn_flush_list(fz, i);
676 return found;
680 static inline int
681 fn_hash_dump_bucket(struct sk_buff *skb, struct netlink_callback *cb,
682 struct fib_table *tb,
683 struct fn_zone *fz,
684 struct hlist_head *head)
686 struct hlist_node *node;
687 struct fib_node *f;
688 int i, s_i;
690 s_i = cb->args[4];
691 i = 0;
692 hlist_for_each_entry(f, node, head, fn_hash) {
693 struct fib_alias *fa;
695 list_for_each_entry(fa, &f->fn_alias, fa_list) {
696 if (i < s_i)
697 goto next;
699 if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
700 cb->nlh->nlmsg_seq,
701 RTM_NEWROUTE,
702 tb->tb_id,
703 fa->fa_type,
704 fa->fa_scope,
705 f->fn_key,
706 fz->fz_order,
707 fa->fa_tos,
708 fa->fa_info,
709 NLM_F_MULTI) < 0) {
710 cb->args[4] = i;
711 return -1;
713 next:
714 i++;
717 cb->args[4] = i;
718 return skb->len;
721 static inline int
722 fn_hash_dump_zone(struct sk_buff *skb, struct netlink_callback *cb,
723 struct fib_table *tb,
724 struct fn_zone *fz)
726 int h, s_h;
728 if (fz->fz_hash == NULL)
729 return skb->len;
730 s_h = cb->args[3];
731 for (h = s_h; h < fz->fz_divisor; h++) {
732 if (hlist_empty(&fz->fz_hash[h]))
733 continue;
734 if (fn_hash_dump_bucket(skb, cb, tb, fz, &fz->fz_hash[h]) < 0) {
735 cb->args[3] = h;
736 return -1;
738 memset(&cb->args[4], 0,
739 sizeof(cb->args) - 4*sizeof(cb->args[0]));
741 cb->args[3] = h;
742 return skb->len;
745 int fib_table_dump(struct fib_table *tb, struct sk_buff *skb,
746 struct netlink_callback *cb)
748 int m, s_m;
749 struct fn_zone *fz;
750 struct fn_hash *table = (struct fn_hash *)tb->tb_data;
752 s_m = cb->args[2];
753 read_lock(&fib_hash_lock);
754 for (fz = table->fn_zone_list, m=0; fz; fz = fz->fz_next, m++) {
755 if (m < s_m) continue;
756 if (fn_hash_dump_zone(skb, cb, tb, fz) < 0) {
757 cb->args[2] = m;
758 read_unlock(&fib_hash_lock);
759 return -1;
761 memset(&cb->args[3], 0,
762 sizeof(cb->args) - 3*sizeof(cb->args[0]));
764 read_unlock(&fib_hash_lock);
765 cb->args[2] = m;
766 return skb->len;
769 void __init fib_hash_init(void)
771 fn_hash_kmem = kmem_cache_create("ip_fib_hash", sizeof(struct fib_node),
772 0, SLAB_PANIC, NULL);
774 fn_alias_kmem = kmem_cache_create("ip_fib_alias", sizeof(struct fib_alias),
775 0, SLAB_PANIC, NULL);
779 struct fib_table *fib_hash_table(u32 id)
781 struct fib_table *tb;
783 tb = kmalloc(sizeof(struct fib_table) + sizeof(struct fn_hash),
784 GFP_KERNEL);
785 if (tb == NULL)
786 return NULL;
788 tb->tb_id = id;
789 tb->tb_default = -1;
791 memset(tb->tb_data, 0, sizeof(struct fn_hash));
792 return tb;
795 /* ------------------------------------------------------------------------ */
796 #ifdef CONFIG_PROC_FS
798 struct fib_iter_state {
799 struct seq_net_private p;
800 struct fn_zone *zone;
801 int bucket;
802 struct hlist_head *hash_head;
803 struct fib_node *fn;
804 struct fib_alias *fa;
805 loff_t pos;
806 unsigned int genid;
807 int valid;
810 static struct fib_alias *fib_get_first(struct seq_file *seq)
812 struct fib_iter_state *iter = seq->private;
813 struct fib_table *main_table;
814 struct fn_hash *table;
816 main_table = fib_get_table(seq_file_net(seq), RT_TABLE_MAIN);
817 table = (struct fn_hash *)main_table->tb_data;
819 iter->bucket = 0;
820 iter->hash_head = NULL;
821 iter->fn = NULL;
822 iter->fa = NULL;
823 iter->pos = 0;
824 iter->genid = fib_hash_genid;
825 iter->valid = 1;
827 for (iter->zone = table->fn_zone_list; iter->zone;
828 iter->zone = iter->zone->fz_next) {
829 int maxslot;
831 if (!iter->zone->fz_nent)
832 continue;
834 iter->hash_head = iter->zone->fz_hash;
835 maxslot = iter->zone->fz_divisor;
837 for (iter->bucket = 0; iter->bucket < maxslot;
838 ++iter->bucket, ++iter->hash_head) {
839 struct hlist_node *node;
840 struct fib_node *fn;
842 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
843 struct fib_alias *fa;
845 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
846 iter->fn = fn;
847 iter->fa = fa;
848 goto out;
853 out:
854 return iter->fa;
857 static struct fib_alias *fib_get_next(struct seq_file *seq)
859 struct fib_iter_state *iter = seq->private;
860 struct fib_node *fn;
861 struct fib_alias *fa;
863 /* Advance FA, if any. */
864 fn = iter->fn;
865 fa = iter->fa;
866 if (fa) {
867 BUG_ON(!fn);
868 list_for_each_entry_continue(fa, &fn->fn_alias, fa_list) {
869 iter->fa = fa;
870 goto out;
874 fa = iter->fa = NULL;
876 /* Advance FN. */
877 if (fn) {
878 struct hlist_node *node = &fn->fn_hash;
879 hlist_for_each_entry_continue(fn, node, fn_hash) {
880 iter->fn = fn;
882 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
883 iter->fa = fa;
884 goto out;
889 fn = iter->fn = NULL;
891 /* Advance hash chain. */
892 if (!iter->zone)
893 goto out;
895 for (;;) {
896 struct hlist_node *node;
897 int maxslot;
899 maxslot = iter->zone->fz_divisor;
901 while (++iter->bucket < maxslot) {
902 iter->hash_head++;
904 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
905 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
906 iter->fn = fn;
907 iter->fa = fa;
908 goto out;
913 iter->zone = iter->zone->fz_next;
915 if (!iter->zone)
916 goto out;
918 iter->bucket = 0;
919 iter->hash_head = iter->zone->fz_hash;
921 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
922 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
923 iter->fn = fn;
924 iter->fa = fa;
925 goto out;
929 out:
930 iter->pos++;
931 return fa;
934 static struct fib_alias *fib_get_idx(struct seq_file *seq, loff_t pos)
936 struct fib_iter_state *iter = seq->private;
937 struct fib_alias *fa;
939 if (iter->valid && pos >= iter->pos && iter->genid == fib_hash_genid) {
940 fa = iter->fa;
941 pos -= iter->pos;
942 } else
943 fa = fib_get_first(seq);
945 if (fa)
946 while (pos && (fa = fib_get_next(seq)))
947 --pos;
948 return pos ? NULL : fa;
951 static void *fib_seq_start(struct seq_file *seq, loff_t *pos)
952 __acquires(fib_hash_lock)
954 void *v = NULL;
956 read_lock(&fib_hash_lock);
957 if (fib_get_table(seq_file_net(seq), RT_TABLE_MAIN))
958 v = *pos ? fib_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
959 return v;
962 static void *fib_seq_next(struct seq_file *seq, void *v, loff_t *pos)
964 ++*pos;
965 return v == SEQ_START_TOKEN ? fib_get_first(seq) : fib_get_next(seq);
968 static void fib_seq_stop(struct seq_file *seq, void *v)
969 __releases(fib_hash_lock)
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_net(inode, 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_net,
1058 int __net_init fib_proc_init(struct net *net)
1060 if (!proc_net_fops_create(net, "route", S_IRUGO, &fib_seq_fops))
1061 return -ENOMEM;
1062 return 0;
1065 void __net_exit fib_proc_exit(struct net *net)
1067 proc_net_remove(net, "route");
1069 #endif /* CONFIG_PROC_FS */