Merge gregkh@master.kernel.org:/home/rmk/linux-2.6-arm
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv4 / fib_trie.c
blob23fb9d9768e369ecc0093805add00cf3f251a158
1 /*
2 * This program is free software; you can redistribute it and/or
3 * modify it under the terms of the GNU General Public License
4 * as published by the Free Software Foundation; either version
5 * 2 of the License, or (at your option) any later version.
7 * Robert Olsson <robert.olsson@its.uu.se> Uppsala Universitet
8 * & Swedish University of Agricultural Sciences.
10 * Jens Laas <jens.laas@data.slu.se> Swedish University of
11 * Agricultural Sciences.
13 * Hans Liss <hans.liss@its.uu.se> Uppsala Universitet
15 * This work is based on the LPC-trie which is originally descibed in:
17 * An experimental study of compression methods for dynamic tries
18 * Stefan Nilsson and Matti Tikkanen. Algorithmica, 33(1):19-33, 2002.
19 * http://www.nada.kth.se/~snilsson/public/papers/dyntrie2/
22 * IP-address lookup using LC-tries. Stefan Nilsson and Gunnar Karlsson
23 * IEEE Journal on Selected Areas in Communications, 17(6):1083-1092, June 1999
25 * Version: $Id: fib_trie.c,v 1.3 2005/06/08 14:20:01 robert Exp $
28 * Code from fib_hash has been reused which includes the following header:
31 * INET An implementation of the TCP/IP protocol suite for the LINUX
32 * operating system. INET is implemented using the BSD Socket
33 * interface as the means of communication with the user level.
35 * IPv4 FIB: lookup engine and maintenance routines.
38 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
40 * This program is free software; you can redistribute it and/or
41 * modify it under the terms of the GNU General Public License
42 * as published by the Free Software Foundation; either version
43 * 2 of the License, or (at your option) any later version.
45 * Substantial contributions to this work comes from:
47 * David S. Miller, <davem@davemloft.net>
48 * Stephen Hemminger <shemminger@osdl.org>
49 * Paul E. McKenney <paulmck@us.ibm.com>
50 * Patrick McHardy <kaber@trash.net>
53 #define VERSION "0.407"
55 #include <asm/uaccess.h>
56 #include <asm/system.h>
57 #include <asm/bitops.h>
58 #include <linux/types.h>
59 #include <linux/kernel.h>
60 #include <linux/sched.h>
61 #include <linux/mm.h>
62 #include <linux/string.h>
63 #include <linux/socket.h>
64 #include <linux/sockios.h>
65 #include <linux/errno.h>
66 #include <linux/in.h>
67 #include <linux/inet.h>
68 #include <linux/inetdevice.h>
69 #include <linux/netdevice.h>
70 #include <linux/if_arp.h>
71 #include <linux/proc_fs.h>
72 #include <linux/rcupdate.h>
73 #include <linux/skbuff.h>
74 #include <linux/netlink.h>
75 #include <linux/init.h>
76 #include <linux/list.h>
77 #include <net/ip.h>
78 #include <net/protocol.h>
79 #include <net/route.h>
80 #include <net/tcp.h>
81 #include <net/sock.h>
82 #include <net/ip_fib.h>
83 #include "fib_lookup.h"
85 #undef CONFIG_IP_FIB_TRIE_STATS
86 #define MAX_STAT_DEPTH 32
88 #define KEYLENGTH (8*sizeof(t_key))
89 #define MASK_PFX(k, l) (((l)==0)?0:(k >> (KEYLENGTH-l)) << (KEYLENGTH-l))
90 #define TKEY_GET_MASK(offset, bits) (((bits)==0)?0:((t_key)(-1) << (KEYLENGTH - bits) >> offset))
92 typedef unsigned int t_key;
94 #define T_TNODE 0
95 #define T_LEAF 1
96 #define NODE_TYPE_MASK 0x1UL
97 #define NODE_PARENT(node) \
98 ((struct tnode *)rcu_dereference(((node)->parent & ~NODE_TYPE_MASK)))
100 #define NODE_TYPE(node) ((node)->parent & NODE_TYPE_MASK)
102 #define NODE_SET_PARENT(node, ptr) \
103 rcu_assign_pointer((node)->parent, \
104 ((unsigned long)(ptr)) | NODE_TYPE(node))
106 #define IS_TNODE(n) (!(n->parent & T_LEAF))
107 #define IS_LEAF(n) (n->parent & T_LEAF)
109 struct node {
110 t_key key;
111 unsigned long parent;
114 struct leaf {
115 t_key key;
116 unsigned long parent;
117 struct hlist_head list;
118 struct rcu_head rcu;
121 struct leaf_info {
122 struct hlist_node hlist;
123 struct rcu_head rcu;
124 int plen;
125 struct list_head falh;
128 struct tnode {
129 t_key key;
130 unsigned long parent;
131 unsigned short pos:5; /* 2log(KEYLENGTH) bits needed */
132 unsigned short bits:5; /* 2log(KEYLENGTH) bits needed */
133 unsigned short full_children; /* KEYLENGTH bits needed */
134 unsigned short empty_children; /* KEYLENGTH bits needed */
135 struct rcu_head rcu;
136 struct node *child[0];
139 #ifdef CONFIG_IP_FIB_TRIE_STATS
140 struct trie_use_stats {
141 unsigned int gets;
142 unsigned int backtrack;
143 unsigned int semantic_match_passed;
144 unsigned int semantic_match_miss;
145 unsigned int null_node_hit;
146 unsigned int resize_node_skipped;
148 #endif
150 struct trie_stat {
151 unsigned int totdepth;
152 unsigned int maxdepth;
153 unsigned int tnodes;
154 unsigned int leaves;
155 unsigned int nullpointers;
156 unsigned int nodesizes[MAX_STAT_DEPTH];
159 struct trie {
160 struct node *trie;
161 #ifdef CONFIG_IP_FIB_TRIE_STATS
162 struct trie_use_stats stats;
163 #endif
164 int size;
165 unsigned int revision;
168 static void put_child(struct trie *t, struct tnode *tn, int i, struct node *n);
169 static void tnode_put_child_reorg(struct tnode *tn, int i, struct node *n, int wasfull);
170 static struct node *resize(struct trie *t, struct tnode *tn);
171 static struct tnode *inflate(struct trie *t, struct tnode *tn);
172 static struct tnode *halve(struct trie *t, struct tnode *tn);
173 static void tnode_free(struct tnode *tn);
175 static kmem_cache_t *fn_alias_kmem __read_mostly;
176 static struct trie *trie_local = NULL, *trie_main = NULL;
179 /* rcu_read_lock needs to be hold by caller from readside */
181 static inline struct node *tnode_get_child(struct tnode *tn, int i)
183 BUG_ON(i >= 1 << tn->bits);
185 return rcu_dereference(tn->child[i]);
188 static inline int tnode_child_length(const struct tnode *tn)
190 return 1 << tn->bits;
193 static inline t_key tkey_extract_bits(t_key a, int offset, int bits)
195 if (offset < KEYLENGTH)
196 return ((t_key)(a << offset)) >> (KEYLENGTH - bits);
197 else
198 return 0;
201 static inline int tkey_equals(t_key a, t_key b)
203 return a == b;
206 static inline int tkey_sub_equals(t_key a, int offset, int bits, t_key b)
208 if (bits == 0 || offset >= KEYLENGTH)
209 return 1;
210 bits = bits > KEYLENGTH ? KEYLENGTH : bits;
211 return ((a ^ b) << offset) >> (KEYLENGTH - bits) == 0;
214 static inline int tkey_mismatch(t_key a, int offset, t_key b)
216 t_key diff = a ^ b;
217 int i = offset;
219 if (!diff)
220 return 0;
221 while ((diff << i) >> (KEYLENGTH-1) == 0)
222 i++;
223 return i;
227 To understand this stuff, an understanding of keys and all their bits is
228 necessary. Every node in the trie has a key associated with it, but not
229 all of the bits in that key are significant.
231 Consider a node 'n' and its parent 'tp'.
233 If n is a leaf, every bit in its key is significant. Its presence is
234 necessitated by path compression, since during a tree traversal (when
235 searching for a leaf - unless we are doing an insertion) we will completely
236 ignore all skipped bits we encounter. Thus we need to verify, at the end of
237 a potentially successful search, that we have indeed been walking the
238 correct key path.
240 Note that we can never "miss" the correct key in the tree if present by
241 following the wrong path. Path compression ensures that segments of the key
242 that are the same for all keys with a given prefix are skipped, but the
243 skipped part *is* identical for each node in the subtrie below the skipped
244 bit! trie_insert() in this implementation takes care of that - note the
245 call to tkey_sub_equals() in trie_insert().
247 if n is an internal node - a 'tnode' here, the various parts of its key
248 have many different meanings.
250 Example:
251 _________________________________________________________________
252 | i | i | i | i | i | i | i | N | N | N | S | S | S | S | S | C |
253 -----------------------------------------------------------------
254 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
256 _________________________________________________________________
257 | C | C | C | u | u | u | u | u | u | u | u | u | u | u | u | u |
258 -----------------------------------------------------------------
259 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
261 tp->pos = 7
262 tp->bits = 3
263 n->pos = 15
264 n->bits = 4
266 First, let's just ignore the bits that come before the parent tp, that is
267 the bits from 0 to (tp->pos-1). They are *known* but at this point we do
268 not use them for anything.
270 The bits from (tp->pos) to (tp->pos + tp->bits - 1) - "N", above - are the
271 index into the parent's child array. That is, they will be used to find
272 'n' among tp's children.
274 The bits from (tp->pos + tp->bits) to (n->pos - 1) - "S" - are skipped bits
275 for the node n.
277 All the bits we have seen so far are significant to the node n. The rest
278 of the bits are really not needed or indeed known in n->key.
280 The bits from (n->pos) to (n->pos + n->bits - 1) - "C" - are the index into
281 n's child array, and will of course be different for each child.
284 The rest of the bits, from (n->pos + n->bits) onward, are completely unknown
285 at this point.
289 static inline void check_tnode(const struct tnode *tn)
291 WARN_ON(tn && tn->pos+tn->bits > 32);
294 static int halve_threshold = 25;
295 static int inflate_threshold = 50;
296 static int halve_threshold_root = 15;
297 static int inflate_threshold_root = 25;
300 static void __alias_free_mem(struct rcu_head *head)
302 struct fib_alias *fa = container_of(head, struct fib_alias, rcu);
303 kmem_cache_free(fn_alias_kmem, fa);
306 static inline void alias_free_mem_rcu(struct fib_alias *fa)
308 call_rcu(&fa->rcu, __alias_free_mem);
311 static void __leaf_free_rcu(struct rcu_head *head)
313 kfree(container_of(head, struct leaf, rcu));
316 static void __leaf_info_free_rcu(struct rcu_head *head)
318 kfree(container_of(head, struct leaf_info, rcu));
321 static inline void free_leaf_info(struct leaf_info *leaf)
323 call_rcu(&leaf->rcu, __leaf_info_free_rcu);
326 static struct tnode *tnode_alloc(unsigned int size)
328 struct page *pages;
330 if (size <= PAGE_SIZE)
331 return kcalloc(size, 1, GFP_KERNEL);
333 pages = alloc_pages(GFP_KERNEL|__GFP_ZERO, get_order(size));
334 if (!pages)
335 return NULL;
337 return page_address(pages);
340 static void __tnode_free_rcu(struct rcu_head *head)
342 struct tnode *tn = container_of(head, struct tnode, rcu);
343 unsigned int size = sizeof(struct tnode) +
344 (1 << tn->bits) * sizeof(struct node *);
346 if (size <= PAGE_SIZE)
347 kfree(tn);
348 else
349 free_pages((unsigned long)tn, get_order(size));
352 static inline void tnode_free(struct tnode *tn)
354 if(IS_LEAF(tn)) {
355 struct leaf *l = (struct leaf *) tn;
356 call_rcu_bh(&l->rcu, __leaf_free_rcu);
358 else
359 call_rcu(&tn->rcu, __tnode_free_rcu);
362 static struct leaf *leaf_new(void)
364 struct leaf *l = kmalloc(sizeof(struct leaf), GFP_KERNEL);
365 if (l) {
366 l->parent = T_LEAF;
367 INIT_HLIST_HEAD(&l->list);
369 return l;
372 static struct leaf_info *leaf_info_new(int plen)
374 struct leaf_info *li = kmalloc(sizeof(struct leaf_info), GFP_KERNEL);
375 if (li) {
376 li->plen = plen;
377 INIT_LIST_HEAD(&li->falh);
379 return li;
382 static struct tnode* tnode_new(t_key key, int pos, int bits)
384 int nchildren = 1<<bits;
385 int sz = sizeof(struct tnode) + nchildren * sizeof(struct node *);
386 struct tnode *tn = tnode_alloc(sz);
388 if (tn) {
389 memset(tn, 0, sz);
390 tn->parent = T_TNODE;
391 tn->pos = pos;
392 tn->bits = bits;
393 tn->key = key;
394 tn->full_children = 0;
395 tn->empty_children = 1<<bits;
398 pr_debug("AT %p s=%u %u\n", tn, (unsigned int) sizeof(struct tnode),
399 (unsigned int) (sizeof(struct node) * 1<<bits));
400 return tn;
404 * Check whether a tnode 'n' is "full", i.e. it is an internal node
405 * and no bits are skipped. See discussion in dyntree paper p. 6
408 static inline int tnode_full(const struct tnode *tn, const struct node *n)
410 if (n == NULL || IS_LEAF(n))
411 return 0;
413 return ((struct tnode *) n)->pos == tn->pos + tn->bits;
416 static inline void put_child(struct trie *t, struct tnode *tn, int i, struct node *n)
418 tnode_put_child_reorg(tn, i, n, -1);
422 * Add a child at position i overwriting the old value.
423 * Update the value of full_children and empty_children.
426 static void tnode_put_child_reorg(struct tnode *tn, int i, struct node *n, int wasfull)
428 struct node *chi = tn->child[i];
429 int isfull;
431 BUG_ON(i >= 1<<tn->bits);
434 /* update emptyChildren */
435 if (n == NULL && chi != NULL)
436 tn->empty_children++;
437 else if (n != NULL && chi == NULL)
438 tn->empty_children--;
440 /* update fullChildren */
441 if (wasfull == -1)
442 wasfull = tnode_full(tn, chi);
444 isfull = tnode_full(tn, n);
445 if (wasfull && !isfull)
446 tn->full_children--;
447 else if (!wasfull && isfull)
448 tn->full_children++;
450 if (n)
451 NODE_SET_PARENT(n, tn);
453 rcu_assign_pointer(tn->child[i], n);
456 static struct node *resize(struct trie *t, struct tnode *tn)
458 int i;
459 int err = 0;
460 struct tnode *old_tn;
461 int inflate_threshold_use;
462 int halve_threshold_use;
464 if (!tn)
465 return NULL;
467 pr_debug("In tnode_resize %p inflate_threshold=%d threshold=%d\n",
468 tn, inflate_threshold, halve_threshold);
470 /* No children */
471 if (tn->empty_children == tnode_child_length(tn)) {
472 tnode_free(tn);
473 return NULL;
475 /* One child */
476 if (tn->empty_children == tnode_child_length(tn) - 1)
477 for (i = 0; i < tnode_child_length(tn); i++) {
478 struct node *n;
480 n = tn->child[i];
481 if (!n)
482 continue;
484 /* compress one level */
485 NODE_SET_PARENT(n, NULL);
486 tnode_free(tn);
487 return n;
490 * Double as long as the resulting node has a number of
491 * nonempty nodes that are above the threshold.
495 * From "Implementing a dynamic compressed trie" by Stefan Nilsson of
496 * the Helsinki University of Technology and Matti Tikkanen of Nokia
497 * Telecommunications, page 6:
498 * "A node is doubled if the ratio of non-empty children to all
499 * children in the *doubled* node is at least 'high'."
501 * 'high' in this instance is the variable 'inflate_threshold'. It
502 * is expressed as a percentage, so we multiply it with
503 * tnode_child_length() and instead of multiplying by 2 (since the
504 * child array will be doubled by inflate()) and multiplying
505 * the left-hand side by 100 (to handle the percentage thing) we
506 * multiply the left-hand side by 50.
508 * The left-hand side may look a bit weird: tnode_child_length(tn)
509 * - tn->empty_children is of course the number of non-null children
510 * in the current node. tn->full_children is the number of "full"
511 * children, that is non-null tnodes with a skip value of 0.
512 * All of those will be doubled in the resulting inflated tnode, so
513 * we just count them one extra time here.
515 * A clearer way to write this would be:
517 * to_be_doubled = tn->full_children;
518 * not_to_be_doubled = tnode_child_length(tn) - tn->empty_children -
519 * tn->full_children;
521 * new_child_length = tnode_child_length(tn) * 2;
523 * new_fill_factor = 100 * (not_to_be_doubled + 2*to_be_doubled) /
524 * new_child_length;
525 * if (new_fill_factor >= inflate_threshold)
527 * ...and so on, tho it would mess up the while () loop.
529 * anyway,
530 * 100 * (not_to_be_doubled + 2*to_be_doubled) / new_child_length >=
531 * inflate_threshold
533 * avoid a division:
534 * 100 * (not_to_be_doubled + 2*to_be_doubled) >=
535 * inflate_threshold * new_child_length
537 * expand not_to_be_doubled and to_be_doubled, and shorten:
538 * 100 * (tnode_child_length(tn) - tn->empty_children +
539 * tn->full_children) >= inflate_threshold * new_child_length
541 * expand new_child_length:
542 * 100 * (tnode_child_length(tn) - tn->empty_children +
543 * tn->full_children) >=
544 * inflate_threshold * tnode_child_length(tn) * 2
546 * shorten again:
547 * 50 * (tn->full_children + tnode_child_length(tn) -
548 * tn->empty_children) >= inflate_threshold *
549 * tnode_child_length(tn)
553 check_tnode(tn);
555 /* Keep root node larger */
557 if(!tn->parent)
558 inflate_threshold_use = inflate_threshold_root;
559 else
560 inflate_threshold_use = inflate_threshold;
562 err = 0;
563 while ((tn->full_children > 0 &&
564 50 * (tn->full_children + tnode_child_length(tn) - tn->empty_children) >=
565 inflate_threshold_use * tnode_child_length(tn))) {
567 old_tn = tn;
568 tn = inflate(t, tn);
569 if (IS_ERR(tn)) {
570 tn = old_tn;
571 #ifdef CONFIG_IP_FIB_TRIE_STATS
572 t->stats.resize_node_skipped++;
573 #endif
574 break;
578 check_tnode(tn);
581 * Halve as long as the number of empty children in this
582 * node is above threshold.
586 /* Keep root node larger */
588 if(!tn->parent)
589 halve_threshold_use = halve_threshold_root;
590 else
591 halve_threshold_use = halve_threshold;
593 err = 0;
594 while (tn->bits > 1 &&
595 100 * (tnode_child_length(tn) - tn->empty_children) <
596 halve_threshold_use * tnode_child_length(tn)) {
598 old_tn = tn;
599 tn = halve(t, tn);
600 if (IS_ERR(tn)) {
601 tn = old_tn;
602 #ifdef CONFIG_IP_FIB_TRIE_STATS
603 t->stats.resize_node_skipped++;
604 #endif
605 break;
610 /* Only one child remains */
611 if (tn->empty_children == tnode_child_length(tn) - 1)
612 for (i = 0; i < tnode_child_length(tn); i++) {
613 struct node *n;
615 n = tn->child[i];
616 if (!n)
617 continue;
619 /* compress one level */
621 NODE_SET_PARENT(n, NULL);
622 tnode_free(tn);
623 return n;
626 return (struct node *) tn;
629 static struct tnode *inflate(struct trie *t, struct tnode *tn)
631 struct tnode *inode;
632 struct tnode *oldtnode = tn;
633 int olen = tnode_child_length(tn);
634 int i;
636 pr_debug("In inflate\n");
638 tn = tnode_new(oldtnode->key, oldtnode->pos, oldtnode->bits + 1);
640 if (!tn)
641 return ERR_PTR(-ENOMEM);
644 * Preallocate and store tnodes before the actual work so we
645 * don't get into an inconsistent state if memory allocation
646 * fails. In case of failure we return the oldnode and inflate
647 * of tnode is ignored.
650 for (i = 0; i < olen; i++) {
651 struct tnode *inode = (struct tnode *) tnode_get_child(oldtnode, i);
653 if (inode &&
654 IS_TNODE(inode) &&
655 inode->pos == oldtnode->pos + oldtnode->bits &&
656 inode->bits > 1) {
657 struct tnode *left, *right;
658 t_key m = TKEY_GET_MASK(inode->pos, 1);
660 left = tnode_new(inode->key&(~m), inode->pos + 1,
661 inode->bits - 1);
662 if (!left)
663 goto nomem;
665 right = tnode_new(inode->key|m, inode->pos + 1,
666 inode->bits - 1);
668 if (!right) {
669 tnode_free(left);
670 goto nomem;
673 put_child(t, tn, 2*i, (struct node *) left);
674 put_child(t, tn, 2*i+1, (struct node *) right);
678 for (i = 0; i < olen; i++) {
679 struct node *node = tnode_get_child(oldtnode, i);
680 struct tnode *left, *right;
681 int size, j;
683 /* An empty child */
684 if (node == NULL)
685 continue;
687 /* A leaf or an internal node with skipped bits */
689 if (IS_LEAF(node) || ((struct tnode *) node)->pos >
690 tn->pos + tn->bits - 1) {
691 if (tkey_extract_bits(node->key, oldtnode->pos + oldtnode->bits,
692 1) == 0)
693 put_child(t, tn, 2*i, node);
694 else
695 put_child(t, tn, 2*i+1, node);
696 continue;
699 /* An internal node with two children */
700 inode = (struct tnode *) node;
702 if (inode->bits == 1) {
703 put_child(t, tn, 2*i, inode->child[0]);
704 put_child(t, tn, 2*i+1, inode->child[1]);
706 tnode_free(inode);
707 continue;
710 /* An internal node with more than two children */
712 /* We will replace this node 'inode' with two new
713 * ones, 'left' and 'right', each with half of the
714 * original children. The two new nodes will have
715 * a position one bit further down the key and this
716 * means that the "significant" part of their keys
717 * (see the discussion near the top of this file)
718 * will differ by one bit, which will be "0" in
719 * left's key and "1" in right's key. Since we are
720 * moving the key position by one step, the bit that
721 * we are moving away from - the bit at position
722 * (inode->pos) - is the one that will differ between
723 * left and right. So... we synthesize that bit in the
724 * two new keys.
725 * The mask 'm' below will be a single "one" bit at
726 * the position (inode->pos)
729 /* Use the old key, but set the new significant
730 * bit to zero.
733 left = (struct tnode *) tnode_get_child(tn, 2*i);
734 put_child(t, tn, 2*i, NULL);
736 BUG_ON(!left);
738 right = (struct tnode *) tnode_get_child(tn, 2*i+1);
739 put_child(t, tn, 2*i+1, NULL);
741 BUG_ON(!right);
743 size = tnode_child_length(left);
744 for (j = 0; j < size; j++) {
745 put_child(t, left, j, inode->child[j]);
746 put_child(t, right, j, inode->child[j + size]);
748 put_child(t, tn, 2*i, resize(t, left));
749 put_child(t, tn, 2*i+1, resize(t, right));
751 tnode_free(inode);
753 tnode_free(oldtnode);
754 return tn;
755 nomem:
757 int size = tnode_child_length(tn);
758 int j;
760 for (j = 0; j < size; j++)
761 if (tn->child[j])
762 tnode_free((struct tnode *)tn->child[j]);
764 tnode_free(tn);
766 return ERR_PTR(-ENOMEM);
770 static struct tnode *halve(struct trie *t, struct tnode *tn)
772 struct tnode *oldtnode = tn;
773 struct node *left, *right;
774 int i;
775 int olen = tnode_child_length(tn);
777 pr_debug("In halve\n");
779 tn = tnode_new(oldtnode->key, oldtnode->pos, oldtnode->bits - 1);
781 if (!tn)
782 return ERR_PTR(-ENOMEM);
785 * Preallocate and store tnodes before the actual work so we
786 * don't get into an inconsistent state if memory allocation
787 * fails. In case of failure we return the oldnode and halve
788 * of tnode is ignored.
791 for (i = 0; i < olen; i += 2) {
792 left = tnode_get_child(oldtnode, i);
793 right = tnode_get_child(oldtnode, i+1);
795 /* Two nonempty children */
796 if (left && right) {
797 struct tnode *newn;
799 newn = tnode_new(left->key, tn->pos + tn->bits, 1);
801 if (!newn)
802 goto nomem;
804 put_child(t, tn, i/2, (struct node *)newn);
809 for (i = 0; i < olen; i += 2) {
810 struct tnode *newBinNode;
812 left = tnode_get_child(oldtnode, i);
813 right = tnode_get_child(oldtnode, i+1);
815 /* At least one of the children is empty */
816 if (left == NULL) {
817 if (right == NULL) /* Both are empty */
818 continue;
819 put_child(t, tn, i/2, right);
820 continue;
823 if (right == NULL) {
824 put_child(t, tn, i/2, left);
825 continue;
828 /* Two nonempty children */
829 newBinNode = (struct tnode *) tnode_get_child(tn, i/2);
830 put_child(t, tn, i/2, NULL);
831 put_child(t, newBinNode, 0, left);
832 put_child(t, newBinNode, 1, right);
833 put_child(t, tn, i/2, resize(t, newBinNode));
835 tnode_free(oldtnode);
836 return tn;
837 nomem:
839 int size = tnode_child_length(tn);
840 int j;
842 for (j = 0; j < size; j++)
843 if (tn->child[j])
844 tnode_free((struct tnode *)tn->child[j]);
846 tnode_free(tn);
848 return ERR_PTR(-ENOMEM);
852 static void trie_init(struct trie *t)
854 if (!t)
855 return;
857 t->size = 0;
858 rcu_assign_pointer(t->trie, NULL);
859 t->revision = 0;
860 #ifdef CONFIG_IP_FIB_TRIE_STATS
861 memset(&t->stats, 0, sizeof(struct trie_use_stats));
862 #endif
865 /* readside must use rcu_read_lock currently dump routines
866 via get_fa_head and dump */
868 static struct leaf_info *find_leaf_info(struct leaf *l, int plen)
870 struct hlist_head *head = &l->list;
871 struct hlist_node *node;
872 struct leaf_info *li;
874 hlist_for_each_entry_rcu(li, node, head, hlist)
875 if (li->plen == plen)
876 return li;
878 return NULL;
881 static inline struct list_head * get_fa_head(struct leaf *l, int plen)
883 struct leaf_info *li = find_leaf_info(l, plen);
885 if (!li)
886 return NULL;
888 return &li->falh;
891 static void insert_leaf_info(struct hlist_head *head, struct leaf_info *new)
893 struct leaf_info *li = NULL, *last = NULL;
894 struct hlist_node *node;
896 if (hlist_empty(head)) {
897 hlist_add_head_rcu(&new->hlist, head);
898 } else {
899 hlist_for_each_entry(li, node, head, hlist) {
900 if (new->plen > li->plen)
901 break;
903 last = li;
905 if (last)
906 hlist_add_after_rcu(&last->hlist, &new->hlist);
907 else
908 hlist_add_before_rcu(&new->hlist, &li->hlist);
912 /* rcu_read_lock needs to be hold by caller from readside */
914 static struct leaf *
915 fib_find_node(struct trie *t, u32 key)
917 int pos;
918 struct tnode *tn;
919 struct node *n;
921 pos = 0;
922 n = rcu_dereference(t->trie);
924 while (n != NULL && NODE_TYPE(n) == T_TNODE) {
925 tn = (struct tnode *) n;
927 check_tnode(tn);
929 if (tkey_sub_equals(tn->key, pos, tn->pos-pos, key)) {
930 pos = tn->pos + tn->bits;
931 n = tnode_get_child(tn, tkey_extract_bits(key, tn->pos, tn->bits));
932 } else
933 break;
935 /* Case we have found a leaf. Compare prefixes */
937 if (n != NULL && IS_LEAF(n) && tkey_equals(key, n->key))
938 return (struct leaf *)n;
940 return NULL;
943 static struct node *trie_rebalance(struct trie *t, struct tnode *tn)
945 int wasfull;
946 t_key cindex, key;
947 struct tnode *tp = NULL;
949 key = tn->key;
951 while (tn != NULL && NODE_PARENT(tn) != NULL) {
953 tp = NODE_PARENT(tn);
954 cindex = tkey_extract_bits(key, tp->pos, tp->bits);
955 wasfull = tnode_full(tp, tnode_get_child(tp, cindex));
956 tn = (struct tnode *) resize (t, (struct tnode *)tn);
957 tnode_put_child_reorg((struct tnode *)tp, cindex,(struct node*)tn, wasfull);
959 if (!NODE_PARENT(tn))
960 break;
962 tn = NODE_PARENT(tn);
964 /* Handle last (top) tnode */
965 if (IS_TNODE(tn))
966 tn = (struct tnode*) resize(t, (struct tnode *)tn);
968 return (struct node*) tn;
971 /* only used from updater-side */
973 static struct list_head *
974 fib_insert_node(struct trie *t, int *err, u32 key, int plen)
976 int pos, newpos;
977 struct tnode *tp = NULL, *tn = NULL;
978 struct node *n;
979 struct leaf *l;
980 int missbit;
981 struct list_head *fa_head = NULL;
982 struct leaf_info *li;
983 t_key cindex;
985 pos = 0;
986 n = t->trie;
988 /* If we point to NULL, stop. Either the tree is empty and we should
989 * just put a new leaf in if, or we have reached an empty child slot,
990 * and we should just put our new leaf in that.
991 * If we point to a T_TNODE, check if it matches our key. Note that
992 * a T_TNODE might be skipping any number of bits - its 'pos' need
993 * not be the parent's 'pos'+'bits'!
995 * If it does match the current key, get pos/bits from it, extract
996 * the index from our key, push the T_TNODE and walk the tree.
998 * If it doesn't, we have to replace it with a new T_TNODE.
1000 * If we point to a T_LEAF, it might or might not have the same key
1001 * as we do. If it does, just change the value, update the T_LEAF's
1002 * value, and return it.
1003 * If it doesn't, we need to replace it with a T_TNODE.
1006 while (n != NULL && NODE_TYPE(n) == T_TNODE) {
1007 tn = (struct tnode *) n;
1009 check_tnode(tn);
1011 if (tkey_sub_equals(tn->key, pos, tn->pos-pos, key)) {
1012 tp = tn;
1013 pos = tn->pos + tn->bits;
1014 n = tnode_get_child(tn, tkey_extract_bits(key, tn->pos, tn->bits));
1016 BUG_ON(n && NODE_PARENT(n) != tn);
1017 } else
1018 break;
1022 * n ----> NULL, LEAF or TNODE
1024 * tp is n's (parent) ----> NULL or TNODE
1027 BUG_ON(tp && IS_LEAF(tp));
1029 /* Case 1: n is a leaf. Compare prefixes */
1031 if (n != NULL && IS_LEAF(n) && tkey_equals(key, n->key)) {
1032 struct leaf *l = (struct leaf *) n;
1034 li = leaf_info_new(plen);
1036 if (!li) {
1037 *err = -ENOMEM;
1038 goto err;
1041 fa_head = &li->falh;
1042 insert_leaf_info(&l->list, li);
1043 goto done;
1045 t->size++;
1046 l = leaf_new();
1048 if (!l) {
1049 *err = -ENOMEM;
1050 goto err;
1053 l->key = key;
1054 li = leaf_info_new(plen);
1056 if (!li) {
1057 tnode_free((struct tnode *) l);
1058 *err = -ENOMEM;
1059 goto err;
1062 fa_head = &li->falh;
1063 insert_leaf_info(&l->list, li);
1065 if (t->trie && n == NULL) {
1066 /* Case 2: n is NULL, and will just insert a new leaf */
1068 NODE_SET_PARENT(l, tp);
1070 cindex = tkey_extract_bits(key, tp->pos, tp->bits);
1071 put_child(t, (struct tnode *)tp, cindex, (struct node *)l);
1072 } else {
1073 /* Case 3: n is a LEAF or a TNODE and the key doesn't match. */
1075 * Add a new tnode here
1076 * first tnode need some special handling
1079 if (tp)
1080 pos = tp->pos+tp->bits;
1081 else
1082 pos = 0;
1084 if (n) {
1085 newpos = tkey_mismatch(key, pos, n->key);
1086 tn = tnode_new(n->key, newpos, 1);
1087 } else {
1088 newpos = 0;
1089 tn = tnode_new(key, newpos, 1); /* First tnode */
1092 if (!tn) {
1093 free_leaf_info(li);
1094 tnode_free((struct tnode *) l);
1095 *err = -ENOMEM;
1096 goto err;
1099 NODE_SET_PARENT(tn, tp);
1101 missbit = tkey_extract_bits(key, newpos, 1);
1102 put_child(t, tn, missbit, (struct node *)l);
1103 put_child(t, tn, 1-missbit, n);
1105 if (tp) {
1106 cindex = tkey_extract_bits(key, tp->pos, tp->bits);
1107 put_child(t, (struct tnode *)tp, cindex, (struct node *)tn);
1108 } else {
1109 rcu_assign_pointer(t->trie, (struct node *)tn); /* First tnode */
1110 tp = tn;
1114 if (tp && tp->pos + tp->bits > 32)
1115 printk(KERN_WARNING "fib_trie tp=%p pos=%d, bits=%d, key=%0x plen=%d\n",
1116 tp, tp->pos, tp->bits, key, plen);
1118 /* Rebalance the trie */
1120 rcu_assign_pointer(t->trie, trie_rebalance(t, tp));
1121 done:
1122 t->revision++;
1123 err:
1124 return fa_head;
1127 static int
1128 fn_trie_insert(struct fib_table *tb, struct rtmsg *r, struct kern_rta *rta,
1129 struct nlmsghdr *nlhdr, struct netlink_skb_parms *req)
1131 struct trie *t = (struct trie *) tb->tb_data;
1132 struct fib_alias *fa, *new_fa;
1133 struct list_head *fa_head = NULL;
1134 struct fib_info *fi;
1135 int plen = r->rtm_dst_len;
1136 int type = r->rtm_type;
1137 u8 tos = r->rtm_tos;
1138 u32 key, mask;
1139 int err;
1140 struct leaf *l;
1142 if (plen > 32)
1143 return -EINVAL;
1145 key = 0;
1146 if (rta->rta_dst)
1147 memcpy(&key, rta->rta_dst, 4);
1149 key = ntohl(key);
1151 pr_debug("Insert table=%d %08x/%d\n", tb->tb_id, key, plen);
1153 mask = ntohl(inet_make_mask(plen));
1155 if (key & ~mask)
1156 return -EINVAL;
1158 key = key & mask;
1160 fi = fib_create_info(r, rta, nlhdr, &err);
1162 if (!fi)
1163 goto err;
1165 l = fib_find_node(t, key);
1166 fa = NULL;
1168 if (l) {
1169 fa_head = get_fa_head(l, plen);
1170 fa = fib_find_alias(fa_head, tos, fi->fib_priority);
1173 /* Now fa, if non-NULL, points to the first fib alias
1174 * with the same keys [prefix,tos,priority], if such key already
1175 * exists or to the node before which we will insert new one.
1177 * If fa is NULL, we will need to allocate a new one and
1178 * insert to the head of f.
1180 * If f is NULL, no fib node matched the destination key
1181 * and we need to allocate a new one of those as well.
1184 if (fa && fa->fa_info->fib_priority == fi->fib_priority) {
1185 struct fib_alias *fa_orig;
1187 err = -EEXIST;
1188 if (nlhdr->nlmsg_flags & NLM_F_EXCL)
1189 goto out;
1191 if (nlhdr->nlmsg_flags & NLM_F_REPLACE) {
1192 struct fib_info *fi_drop;
1193 u8 state;
1195 err = -ENOBUFS;
1196 new_fa = kmem_cache_alloc(fn_alias_kmem, SLAB_KERNEL);
1197 if (new_fa == NULL)
1198 goto out;
1200 fi_drop = fa->fa_info;
1201 new_fa->fa_tos = fa->fa_tos;
1202 new_fa->fa_info = fi;
1203 new_fa->fa_type = type;
1204 new_fa->fa_scope = r->rtm_scope;
1205 state = fa->fa_state;
1206 new_fa->fa_state &= ~FA_S_ACCESSED;
1208 list_replace_rcu(&fa->fa_list, &new_fa->fa_list);
1209 alias_free_mem_rcu(fa);
1211 fib_release_info(fi_drop);
1212 if (state & FA_S_ACCESSED)
1213 rt_cache_flush(-1);
1215 goto succeeded;
1217 /* Error if we find a perfect match which
1218 * uses the same scope, type, and nexthop
1219 * information.
1221 fa_orig = fa;
1222 list_for_each_entry(fa, fa_orig->fa_list.prev, fa_list) {
1223 if (fa->fa_tos != tos)
1224 break;
1225 if (fa->fa_info->fib_priority != fi->fib_priority)
1226 break;
1227 if (fa->fa_type == type &&
1228 fa->fa_scope == r->rtm_scope &&
1229 fa->fa_info == fi) {
1230 goto out;
1233 if (!(nlhdr->nlmsg_flags & NLM_F_APPEND))
1234 fa = fa_orig;
1236 err = -ENOENT;
1237 if (!(nlhdr->nlmsg_flags & NLM_F_CREATE))
1238 goto out;
1240 err = -ENOBUFS;
1241 new_fa = kmem_cache_alloc(fn_alias_kmem, SLAB_KERNEL);
1242 if (new_fa == NULL)
1243 goto out;
1245 new_fa->fa_info = fi;
1246 new_fa->fa_tos = tos;
1247 new_fa->fa_type = type;
1248 new_fa->fa_scope = r->rtm_scope;
1249 new_fa->fa_state = 0;
1251 * Insert new entry to the list.
1254 if (!fa_head) {
1255 err = 0;
1256 fa_head = fib_insert_node(t, &err, key, plen);
1257 if (err)
1258 goto out_free_new_fa;
1261 list_add_tail_rcu(&new_fa->fa_list,
1262 (fa ? &fa->fa_list : fa_head));
1264 rt_cache_flush(-1);
1265 rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen, tb->tb_id, nlhdr, req);
1266 succeeded:
1267 return 0;
1269 out_free_new_fa:
1270 kmem_cache_free(fn_alias_kmem, new_fa);
1271 out:
1272 fib_release_info(fi);
1273 err:
1274 return err;
1278 /* should be called with rcu_read_lock */
1279 static inline int check_leaf(struct trie *t, struct leaf *l,
1280 t_key key, int *plen, const struct flowi *flp,
1281 struct fib_result *res)
1283 int err, i;
1284 t_key mask;
1285 struct leaf_info *li;
1286 struct hlist_head *hhead = &l->list;
1287 struct hlist_node *node;
1289 hlist_for_each_entry_rcu(li, node, hhead, hlist) {
1290 i = li->plen;
1291 mask = ntohl(inet_make_mask(i));
1292 if (l->key != (key & mask))
1293 continue;
1295 if ((err = fib_semantic_match(&li->falh, flp, res, l->key, mask, i)) <= 0) {
1296 *plen = i;
1297 #ifdef CONFIG_IP_FIB_TRIE_STATS
1298 t->stats.semantic_match_passed++;
1299 #endif
1300 return err;
1302 #ifdef CONFIG_IP_FIB_TRIE_STATS
1303 t->stats.semantic_match_miss++;
1304 #endif
1306 return 1;
1309 static int
1310 fn_trie_lookup(struct fib_table *tb, const struct flowi *flp, struct fib_result *res)
1312 struct trie *t = (struct trie *) tb->tb_data;
1313 int plen, ret = 0;
1314 struct node *n;
1315 struct tnode *pn;
1316 int pos, bits;
1317 t_key key = ntohl(flp->fl4_dst);
1318 int chopped_off;
1319 t_key cindex = 0;
1320 int current_prefix_length = KEYLENGTH;
1321 struct tnode *cn;
1322 t_key node_prefix, key_prefix, pref_mismatch;
1323 int mp;
1325 rcu_read_lock();
1327 n = rcu_dereference(t->trie);
1328 if (!n)
1329 goto failed;
1331 #ifdef CONFIG_IP_FIB_TRIE_STATS
1332 t->stats.gets++;
1333 #endif
1335 /* Just a leaf? */
1336 if (IS_LEAF(n)) {
1337 if ((ret = check_leaf(t, (struct leaf *)n, key, &plen, flp, res)) <= 0)
1338 goto found;
1339 goto failed;
1341 pn = (struct tnode *) n;
1342 chopped_off = 0;
1344 while (pn) {
1345 pos = pn->pos;
1346 bits = pn->bits;
1348 if (!chopped_off)
1349 cindex = tkey_extract_bits(MASK_PFX(key, current_prefix_length), pos, bits);
1351 n = tnode_get_child(pn, cindex);
1353 if (n == NULL) {
1354 #ifdef CONFIG_IP_FIB_TRIE_STATS
1355 t->stats.null_node_hit++;
1356 #endif
1357 goto backtrace;
1360 if (IS_LEAF(n)) {
1361 if ((ret = check_leaf(t, (struct leaf *)n, key, &plen, flp, res)) <= 0)
1362 goto found;
1363 else
1364 goto backtrace;
1367 #define HL_OPTIMIZE
1368 #ifdef HL_OPTIMIZE
1369 cn = (struct tnode *)n;
1372 * It's a tnode, and we can do some extra checks here if we
1373 * like, to avoid descending into a dead-end branch.
1374 * This tnode is in the parent's child array at index
1375 * key[p_pos..p_pos+p_bits] but potentially with some bits
1376 * chopped off, so in reality the index may be just a
1377 * subprefix, padded with zero at the end.
1378 * We can also take a look at any skipped bits in this
1379 * tnode - everything up to p_pos is supposed to be ok,
1380 * and the non-chopped bits of the index (se previous
1381 * paragraph) are also guaranteed ok, but the rest is
1382 * considered unknown.
1384 * The skipped bits are key[pos+bits..cn->pos].
1387 /* If current_prefix_length < pos+bits, we are already doing
1388 * actual prefix matching, which means everything from
1389 * pos+(bits-chopped_off) onward must be zero along some
1390 * branch of this subtree - otherwise there is *no* valid
1391 * prefix present. Here we can only check the skipped
1392 * bits. Remember, since we have already indexed into the
1393 * parent's child array, we know that the bits we chopped of
1394 * *are* zero.
1397 /* NOTA BENE: CHECKING ONLY SKIPPED BITS FOR THE NEW NODE HERE */
1399 if (current_prefix_length < pos+bits) {
1400 if (tkey_extract_bits(cn->key, current_prefix_length,
1401 cn->pos - current_prefix_length) != 0 ||
1402 !(cn->child[0]))
1403 goto backtrace;
1407 * If chopped_off=0, the index is fully validated and we
1408 * only need to look at the skipped bits for this, the new,
1409 * tnode. What we actually want to do is to find out if
1410 * these skipped bits match our key perfectly, or if we will
1411 * have to count on finding a matching prefix further down,
1412 * because if we do, we would like to have some way of
1413 * verifying the existence of such a prefix at this point.
1416 /* The only thing we can do at this point is to verify that
1417 * any such matching prefix can indeed be a prefix to our
1418 * key, and if the bits in the node we are inspecting that
1419 * do not match our key are not ZERO, this cannot be true.
1420 * Thus, find out where there is a mismatch (before cn->pos)
1421 * and verify that all the mismatching bits are zero in the
1422 * new tnode's key.
1425 /* Note: We aren't very concerned about the piece of the key
1426 * that precede pn->pos+pn->bits, since these have already been
1427 * checked. The bits after cn->pos aren't checked since these are
1428 * by definition "unknown" at this point. Thus, what we want to
1429 * see is if we are about to enter the "prefix matching" state,
1430 * and in that case verify that the skipped bits that will prevail
1431 * throughout this subtree are zero, as they have to be if we are
1432 * to find a matching prefix.
1435 node_prefix = MASK_PFX(cn->key, cn->pos);
1436 key_prefix = MASK_PFX(key, cn->pos);
1437 pref_mismatch = key_prefix^node_prefix;
1438 mp = 0;
1440 /* In short: If skipped bits in this node do not match the search
1441 * key, enter the "prefix matching" state.directly.
1443 if (pref_mismatch) {
1444 while (!(pref_mismatch & (1<<(KEYLENGTH-1)))) {
1445 mp++;
1446 pref_mismatch = pref_mismatch <<1;
1448 key_prefix = tkey_extract_bits(cn->key, mp, cn->pos-mp);
1450 if (key_prefix != 0)
1451 goto backtrace;
1453 if (current_prefix_length >= cn->pos)
1454 current_prefix_length = mp;
1456 #endif
1457 pn = (struct tnode *)n; /* Descend */
1458 chopped_off = 0;
1459 continue;
1461 backtrace:
1462 chopped_off++;
1464 /* As zero don't change the child key (cindex) */
1465 while ((chopped_off <= pn->bits) && !(cindex & (1<<(chopped_off-1))))
1466 chopped_off++;
1468 /* Decrease current_... with bits chopped off */
1469 if (current_prefix_length > pn->pos + pn->bits - chopped_off)
1470 current_prefix_length = pn->pos + pn->bits - chopped_off;
1473 * Either we do the actual chop off according or if we have
1474 * chopped off all bits in this tnode walk up to our parent.
1477 if (chopped_off <= pn->bits) {
1478 cindex &= ~(1 << (chopped_off-1));
1479 } else {
1480 if (NODE_PARENT(pn) == NULL)
1481 goto failed;
1483 /* Get Child's index */
1484 cindex = tkey_extract_bits(pn->key, NODE_PARENT(pn)->pos, NODE_PARENT(pn)->bits);
1485 pn = NODE_PARENT(pn);
1486 chopped_off = 0;
1488 #ifdef CONFIG_IP_FIB_TRIE_STATS
1489 t->stats.backtrack++;
1490 #endif
1491 goto backtrace;
1494 failed:
1495 ret = 1;
1496 found:
1497 rcu_read_unlock();
1498 return ret;
1501 /* only called from updater side */
1502 static int trie_leaf_remove(struct trie *t, t_key key)
1504 t_key cindex;
1505 struct tnode *tp = NULL;
1506 struct node *n = t->trie;
1507 struct leaf *l;
1509 pr_debug("entering trie_leaf_remove(%p)\n", n);
1511 /* Note that in the case skipped bits, those bits are *not* checked!
1512 * When we finish this, we will have NULL or a T_LEAF, and the
1513 * T_LEAF may or may not match our key.
1516 while (n != NULL && IS_TNODE(n)) {
1517 struct tnode *tn = (struct tnode *) n;
1518 check_tnode(tn);
1519 n = tnode_get_child(tn ,tkey_extract_bits(key, tn->pos, tn->bits));
1521 BUG_ON(n && NODE_PARENT(n) != tn);
1523 l = (struct leaf *) n;
1525 if (!n || !tkey_equals(l->key, key))
1526 return 0;
1529 * Key found.
1530 * Remove the leaf and rebalance the tree
1533 t->revision++;
1534 t->size--;
1536 preempt_disable();
1537 tp = NODE_PARENT(n);
1538 tnode_free((struct tnode *) n);
1540 if (tp) {
1541 cindex = tkey_extract_bits(key, tp->pos, tp->bits);
1542 put_child(t, (struct tnode *)tp, cindex, NULL);
1543 rcu_assign_pointer(t->trie, trie_rebalance(t, tp));
1544 } else
1545 rcu_assign_pointer(t->trie, NULL);
1546 preempt_enable();
1548 return 1;
1551 static int
1552 fn_trie_delete(struct fib_table *tb, struct rtmsg *r, struct kern_rta *rta,
1553 struct nlmsghdr *nlhdr, struct netlink_skb_parms *req)
1555 struct trie *t = (struct trie *) tb->tb_data;
1556 u32 key, mask;
1557 int plen = r->rtm_dst_len;
1558 u8 tos = r->rtm_tos;
1559 struct fib_alias *fa, *fa_to_delete;
1560 struct list_head *fa_head;
1561 struct leaf *l;
1562 struct leaf_info *li;
1565 if (plen > 32)
1566 return -EINVAL;
1568 key = 0;
1569 if (rta->rta_dst)
1570 memcpy(&key, rta->rta_dst, 4);
1572 key = ntohl(key);
1573 mask = ntohl(inet_make_mask(plen));
1575 if (key & ~mask)
1576 return -EINVAL;
1578 key = key & mask;
1579 l = fib_find_node(t, key);
1581 if (!l)
1582 return -ESRCH;
1584 fa_head = get_fa_head(l, plen);
1585 fa = fib_find_alias(fa_head, tos, 0);
1587 if (!fa)
1588 return -ESRCH;
1590 pr_debug("Deleting %08x/%d tos=%d t=%p\n", key, plen, tos, t);
1592 fa_to_delete = NULL;
1593 fa_head = fa->fa_list.prev;
1595 list_for_each_entry(fa, fa_head, fa_list) {
1596 struct fib_info *fi = fa->fa_info;
1598 if (fa->fa_tos != tos)
1599 break;
1601 if ((!r->rtm_type ||
1602 fa->fa_type == r->rtm_type) &&
1603 (r->rtm_scope == RT_SCOPE_NOWHERE ||
1604 fa->fa_scope == r->rtm_scope) &&
1605 (!r->rtm_protocol ||
1606 fi->fib_protocol == r->rtm_protocol) &&
1607 fib_nh_match(r, nlhdr, rta, fi) == 0) {
1608 fa_to_delete = fa;
1609 break;
1613 if (!fa_to_delete)
1614 return -ESRCH;
1616 fa = fa_to_delete;
1617 rtmsg_fib(RTM_DELROUTE, htonl(key), fa, plen, tb->tb_id, nlhdr, req);
1619 l = fib_find_node(t, key);
1620 li = find_leaf_info(l, plen);
1622 list_del_rcu(&fa->fa_list);
1624 if (list_empty(fa_head)) {
1625 hlist_del_rcu(&li->hlist);
1626 free_leaf_info(li);
1629 if (hlist_empty(&l->list))
1630 trie_leaf_remove(t, key);
1632 if (fa->fa_state & FA_S_ACCESSED)
1633 rt_cache_flush(-1);
1635 fib_release_info(fa->fa_info);
1636 alias_free_mem_rcu(fa);
1637 return 0;
1640 static int trie_flush_list(struct trie *t, struct list_head *head)
1642 struct fib_alias *fa, *fa_node;
1643 int found = 0;
1645 list_for_each_entry_safe(fa, fa_node, head, fa_list) {
1646 struct fib_info *fi = fa->fa_info;
1648 if (fi && (fi->fib_flags & RTNH_F_DEAD)) {
1649 list_del_rcu(&fa->fa_list);
1650 fib_release_info(fa->fa_info);
1651 alias_free_mem_rcu(fa);
1652 found++;
1655 return found;
1658 static int trie_flush_leaf(struct trie *t, struct leaf *l)
1660 int found = 0;
1661 struct hlist_head *lih = &l->list;
1662 struct hlist_node *node, *tmp;
1663 struct leaf_info *li = NULL;
1665 hlist_for_each_entry_safe(li, node, tmp, lih, hlist) {
1666 found += trie_flush_list(t, &li->falh);
1668 if (list_empty(&li->falh)) {
1669 hlist_del_rcu(&li->hlist);
1670 free_leaf_info(li);
1673 return found;
1676 /* rcu_read_lock needs to be hold by caller from readside */
1678 static struct leaf *nextleaf(struct trie *t, struct leaf *thisleaf)
1680 struct node *c = (struct node *) thisleaf;
1681 struct tnode *p;
1682 int idx;
1683 struct node *trie = rcu_dereference(t->trie);
1685 if (c == NULL) {
1686 if (trie == NULL)
1687 return NULL;
1689 if (IS_LEAF(trie)) /* trie w. just a leaf */
1690 return (struct leaf *) trie;
1692 p = (struct tnode*) trie; /* Start */
1693 } else
1694 p = (struct tnode *) NODE_PARENT(c);
1696 while (p) {
1697 int pos, last;
1699 /* Find the next child of the parent */
1700 if (c)
1701 pos = 1 + tkey_extract_bits(c->key, p->pos, p->bits);
1702 else
1703 pos = 0;
1705 last = 1 << p->bits;
1706 for (idx = pos; idx < last ; idx++) {
1707 c = rcu_dereference(p->child[idx]);
1709 if (!c)
1710 continue;
1712 /* Decend if tnode */
1713 while (IS_TNODE(c)) {
1714 p = (struct tnode *) c;
1715 idx = 0;
1717 /* Rightmost non-NULL branch */
1718 if (p && IS_TNODE(p))
1719 while (!(c = rcu_dereference(p->child[idx]))
1720 && idx < (1<<p->bits)) idx++;
1722 /* Done with this tnode? */
1723 if (idx >= (1 << p->bits) || !c)
1724 goto up;
1726 return (struct leaf *) c;
1729 /* No more children go up one step */
1730 c = (struct node *) p;
1731 p = (struct tnode *) NODE_PARENT(p);
1733 return NULL; /* Ready. Root of trie */
1736 static int fn_trie_flush(struct fib_table *tb)
1738 struct trie *t = (struct trie *) tb->tb_data;
1739 struct leaf *ll = NULL, *l = NULL;
1740 int found = 0, h;
1742 t->revision++;
1744 for (h = 0; (l = nextleaf(t, l)) != NULL; h++) {
1745 found += trie_flush_leaf(t, l);
1747 if (ll && hlist_empty(&ll->list))
1748 trie_leaf_remove(t, ll->key);
1749 ll = l;
1752 if (ll && hlist_empty(&ll->list))
1753 trie_leaf_remove(t, ll->key);
1755 pr_debug("trie_flush found=%d\n", found);
1756 return found;
1759 static int trie_last_dflt = -1;
1761 static void
1762 fn_trie_select_default(struct fib_table *tb, const struct flowi *flp, struct fib_result *res)
1764 struct trie *t = (struct trie *) tb->tb_data;
1765 int order, last_idx;
1766 struct fib_info *fi = NULL;
1767 struct fib_info *last_resort;
1768 struct fib_alias *fa = NULL;
1769 struct list_head *fa_head;
1770 struct leaf *l;
1772 last_idx = -1;
1773 last_resort = NULL;
1774 order = -1;
1776 rcu_read_lock();
1778 l = fib_find_node(t, 0);
1779 if (!l)
1780 goto out;
1782 fa_head = get_fa_head(l, 0);
1783 if (!fa_head)
1784 goto out;
1786 if (list_empty(fa_head))
1787 goto out;
1789 list_for_each_entry_rcu(fa, fa_head, fa_list) {
1790 struct fib_info *next_fi = fa->fa_info;
1792 if (fa->fa_scope != res->scope ||
1793 fa->fa_type != RTN_UNICAST)
1794 continue;
1796 if (next_fi->fib_priority > res->fi->fib_priority)
1797 break;
1798 if (!next_fi->fib_nh[0].nh_gw ||
1799 next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK)
1800 continue;
1801 fa->fa_state |= FA_S_ACCESSED;
1803 if (fi == NULL) {
1804 if (next_fi != res->fi)
1805 break;
1806 } else if (!fib_detect_death(fi, order, &last_resort,
1807 &last_idx, &trie_last_dflt)) {
1808 if (res->fi)
1809 fib_info_put(res->fi);
1810 res->fi = fi;
1811 atomic_inc(&fi->fib_clntref);
1812 trie_last_dflt = order;
1813 goto out;
1815 fi = next_fi;
1816 order++;
1818 if (order <= 0 || fi == NULL) {
1819 trie_last_dflt = -1;
1820 goto out;
1823 if (!fib_detect_death(fi, order, &last_resort, &last_idx, &trie_last_dflt)) {
1824 if (res->fi)
1825 fib_info_put(res->fi);
1826 res->fi = fi;
1827 atomic_inc(&fi->fib_clntref);
1828 trie_last_dflt = order;
1829 goto out;
1831 if (last_idx >= 0) {
1832 if (res->fi)
1833 fib_info_put(res->fi);
1834 res->fi = last_resort;
1835 if (last_resort)
1836 atomic_inc(&last_resort->fib_clntref);
1838 trie_last_dflt = last_idx;
1839 out:;
1840 rcu_read_unlock();
1843 static int fn_trie_dump_fa(t_key key, int plen, struct list_head *fah, struct fib_table *tb,
1844 struct sk_buff *skb, struct netlink_callback *cb)
1846 int i, s_i;
1847 struct fib_alias *fa;
1849 u32 xkey = htonl(key);
1851 s_i = cb->args[3];
1852 i = 0;
1854 /* rcu_read_lock is hold by caller */
1856 list_for_each_entry_rcu(fa, fah, fa_list) {
1857 if (i < s_i) {
1858 i++;
1859 continue;
1861 BUG_ON(!fa->fa_info);
1863 if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
1864 cb->nlh->nlmsg_seq,
1865 RTM_NEWROUTE,
1866 tb->tb_id,
1867 fa->fa_type,
1868 fa->fa_scope,
1869 &xkey,
1870 plen,
1871 fa->fa_tos,
1872 fa->fa_info, 0) < 0) {
1873 cb->args[3] = i;
1874 return -1;
1876 i++;
1878 cb->args[3] = i;
1879 return skb->len;
1882 static int fn_trie_dump_plen(struct trie *t, int plen, struct fib_table *tb, struct sk_buff *skb,
1883 struct netlink_callback *cb)
1885 int h, s_h;
1886 struct list_head *fa_head;
1887 struct leaf *l = NULL;
1889 s_h = cb->args[2];
1891 for (h = 0; (l = nextleaf(t, l)) != NULL; h++) {
1892 if (h < s_h)
1893 continue;
1894 if (h > s_h)
1895 memset(&cb->args[3], 0,
1896 sizeof(cb->args) - 3*sizeof(cb->args[0]));
1898 fa_head = get_fa_head(l, plen);
1900 if (!fa_head)
1901 continue;
1903 if (list_empty(fa_head))
1904 continue;
1906 if (fn_trie_dump_fa(l->key, plen, fa_head, tb, skb, cb)<0) {
1907 cb->args[2] = h;
1908 return -1;
1911 cb->args[2] = h;
1912 return skb->len;
1915 static int fn_trie_dump(struct fib_table *tb, struct sk_buff *skb, struct netlink_callback *cb)
1917 int m, s_m;
1918 struct trie *t = (struct trie *) tb->tb_data;
1920 s_m = cb->args[1];
1922 rcu_read_lock();
1923 for (m = 0; m <= 32; m++) {
1924 if (m < s_m)
1925 continue;
1926 if (m > s_m)
1927 memset(&cb->args[2], 0,
1928 sizeof(cb->args) - 2*sizeof(cb->args[0]));
1930 if (fn_trie_dump_plen(t, 32-m, tb, skb, cb)<0) {
1931 cb->args[1] = m;
1932 goto out;
1935 rcu_read_unlock();
1936 cb->args[1] = m;
1937 return skb->len;
1938 out:
1939 rcu_read_unlock();
1940 return -1;
1943 /* Fix more generic FIB names for init later */
1945 #ifdef CONFIG_IP_MULTIPLE_TABLES
1946 struct fib_table * fib_hash_init(int id)
1947 #else
1948 struct fib_table * __init fib_hash_init(int id)
1949 #endif
1951 struct fib_table *tb;
1952 struct trie *t;
1954 if (fn_alias_kmem == NULL)
1955 fn_alias_kmem = kmem_cache_create("ip_fib_alias",
1956 sizeof(struct fib_alias),
1957 0, SLAB_HWCACHE_ALIGN,
1958 NULL, NULL);
1960 tb = kmalloc(sizeof(struct fib_table) + sizeof(struct trie),
1961 GFP_KERNEL);
1962 if (tb == NULL)
1963 return NULL;
1965 tb->tb_id = id;
1966 tb->tb_lookup = fn_trie_lookup;
1967 tb->tb_insert = fn_trie_insert;
1968 tb->tb_delete = fn_trie_delete;
1969 tb->tb_flush = fn_trie_flush;
1970 tb->tb_select_default = fn_trie_select_default;
1971 tb->tb_dump = fn_trie_dump;
1972 memset(tb->tb_data, 0, sizeof(struct trie));
1974 t = (struct trie *) tb->tb_data;
1976 trie_init(t);
1978 if (id == RT_TABLE_LOCAL)
1979 trie_local = t;
1980 else if (id == RT_TABLE_MAIN)
1981 trie_main = t;
1983 if (id == RT_TABLE_LOCAL)
1984 printk(KERN_INFO "IPv4 FIB: Using LC-trie version %s\n", VERSION);
1986 return tb;
1989 #ifdef CONFIG_PROC_FS
1990 /* Depth first Trie walk iterator */
1991 struct fib_trie_iter {
1992 struct tnode *tnode;
1993 struct trie *trie;
1994 unsigned index;
1995 unsigned depth;
1998 static struct node *fib_trie_get_next(struct fib_trie_iter *iter)
2000 struct tnode *tn = iter->tnode;
2001 unsigned cindex = iter->index;
2002 struct tnode *p;
2004 pr_debug("get_next iter={node=%p index=%d depth=%d}\n",
2005 iter->tnode, iter->index, iter->depth);
2006 rescan:
2007 while (cindex < (1<<tn->bits)) {
2008 struct node *n = tnode_get_child(tn, cindex);
2010 if (n) {
2011 if (IS_LEAF(n)) {
2012 iter->tnode = tn;
2013 iter->index = cindex + 1;
2014 } else {
2015 /* push down one level */
2016 iter->tnode = (struct tnode *) n;
2017 iter->index = 0;
2018 ++iter->depth;
2020 return n;
2023 ++cindex;
2026 /* Current node exhausted, pop back up */
2027 p = NODE_PARENT(tn);
2028 if (p) {
2029 cindex = tkey_extract_bits(tn->key, p->pos, p->bits)+1;
2030 tn = p;
2031 --iter->depth;
2032 goto rescan;
2035 /* got root? */
2036 return NULL;
2039 static struct node *fib_trie_get_first(struct fib_trie_iter *iter,
2040 struct trie *t)
2042 struct node *n ;
2044 if(!t)
2045 return NULL;
2047 n = rcu_dereference(t->trie);
2049 if(!iter)
2050 return NULL;
2052 if (n && IS_TNODE(n)) {
2053 iter->tnode = (struct tnode *) n;
2054 iter->trie = t;
2055 iter->index = 0;
2056 iter->depth = 1;
2057 return n;
2059 return NULL;
2062 static void trie_collect_stats(struct trie *t, struct trie_stat *s)
2064 struct node *n;
2065 struct fib_trie_iter iter;
2067 memset(s, 0, sizeof(*s));
2069 rcu_read_lock();
2070 for (n = fib_trie_get_first(&iter, t); n;
2071 n = fib_trie_get_next(&iter)) {
2072 if (IS_LEAF(n)) {
2073 s->leaves++;
2074 s->totdepth += iter.depth;
2075 if (iter.depth > s->maxdepth)
2076 s->maxdepth = iter.depth;
2077 } else {
2078 const struct tnode *tn = (const struct tnode *) n;
2079 int i;
2081 s->tnodes++;
2082 if(tn->bits < MAX_STAT_DEPTH)
2083 s->nodesizes[tn->bits]++;
2085 for (i = 0; i < (1<<tn->bits); i++)
2086 if (!tn->child[i])
2087 s->nullpointers++;
2090 rcu_read_unlock();
2094 * This outputs /proc/net/fib_triestats
2096 static void trie_show_stats(struct seq_file *seq, struct trie_stat *stat)
2098 unsigned i, max, pointers, bytes, avdepth;
2100 if (stat->leaves)
2101 avdepth = stat->totdepth*100 / stat->leaves;
2102 else
2103 avdepth = 0;
2105 seq_printf(seq, "\tAver depth: %d.%02d\n", avdepth / 100, avdepth % 100 );
2106 seq_printf(seq, "\tMax depth: %u\n", stat->maxdepth);
2108 seq_printf(seq, "\tLeaves: %u\n", stat->leaves);
2110 bytes = sizeof(struct leaf) * stat->leaves;
2111 seq_printf(seq, "\tInternal nodes: %d\n\t", stat->tnodes);
2112 bytes += sizeof(struct tnode) * stat->tnodes;
2114 max = MAX_STAT_DEPTH;
2115 while (max > 0 && stat->nodesizes[max-1] == 0)
2116 max--;
2118 pointers = 0;
2119 for (i = 1; i <= max; i++)
2120 if (stat->nodesizes[i] != 0) {
2121 seq_printf(seq, " %d: %d", i, stat->nodesizes[i]);
2122 pointers += (1<<i) * stat->nodesizes[i];
2124 seq_putc(seq, '\n');
2125 seq_printf(seq, "\tPointers: %d\n", pointers);
2127 bytes += sizeof(struct node *) * pointers;
2128 seq_printf(seq, "Null ptrs: %d\n", stat->nullpointers);
2129 seq_printf(seq, "Total size: %d kB\n", (bytes + 1023) / 1024);
2131 #ifdef CONFIG_IP_FIB_TRIE_STATS
2132 seq_printf(seq, "Counters:\n---------\n");
2133 seq_printf(seq,"gets = %d\n", t->stats.gets);
2134 seq_printf(seq,"backtracks = %d\n", t->stats.backtrack);
2135 seq_printf(seq,"semantic match passed = %d\n", t->stats.semantic_match_passed);
2136 seq_printf(seq,"semantic match miss = %d\n", t->stats.semantic_match_miss);
2137 seq_printf(seq,"null node hit= %d\n", t->stats.null_node_hit);
2138 seq_printf(seq,"skipped node resize = %d\n", t->stats.resize_node_skipped);
2139 #ifdef CLEAR_STATS
2140 memset(&(t->stats), 0, sizeof(t->stats));
2141 #endif
2142 #endif /* CONFIG_IP_FIB_TRIE_STATS */
2145 static int fib_triestat_seq_show(struct seq_file *seq, void *v)
2147 struct trie_stat *stat;
2149 stat = kmalloc(sizeof(*stat), GFP_KERNEL);
2150 if (!stat)
2151 return -ENOMEM;
2153 seq_printf(seq, "Basic info: size of leaf: %Zd bytes, size of tnode: %Zd bytes.\n",
2154 sizeof(struct leaf), sizeof(struct tnode));
2156 if (trie_local) {
2157 seq_printf(seq, "Local:\n");
2158 trie_collect_stats(trie_local, stat);
2159 trie_show_stats(seq, stat);
2162 if (trie_main) {
2163 seq_printf(seq, "Main:\n");
2164 trie_collect_stats(trie_main, stat);
2165 trie_show_stats(seq, stat);
2167 kfree(stat);
2169 return 0;
2172 static int fib_triestat_seq_open(struct inode *inode, struct file *file)
2174 return single_open(file, fib_triestat_seq_show, NULL);
2177 static struct file_operations fib_triestat_fops = {
2178 .owner = THIS_MODULE,
2179 .open = fib_triestat_seq_open,
2180 .read = seq_read,
2181 .llseek = seq_lseek,
2182 .release = single_release,
2185 static struct node *fib_trie_get_idx(struct fib_trie_iter *iter,
2186 loff_t pos)
2188 loff_t idx = 0;
2189 struct node *n;
2191 for (n = fib_trie_get_first(iter, trie_local);
2192 n; ++idx, n = fib_trie_get_next(iter)) {
2193 if (pos == idx)
2194 return n;
2197 for (n = fib_trie_get_first(iter, trie_main);
2198 n; ++idx, n = fib_trie_get_next(iter)) {
2199 if (pos == idx)
2200 return n;
2202 return NULL;
2205 static void *fib_trie_seq_start(struct seq_file *seq, loff_t *pos)
2207 rcu_read_lock();
2208 if (*pos == 0)
2209 return SEQ_START_TOKEN;
2210 return fib_trie_get_idx(seq->private, *pos - 1);
2213 static void *fib_trie_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2215 struct fib_trie_iter *iter = seq->private;
2216 void *l = v;
2218 ++*pos;
2219 if (v == SEQ_START_TOKEN)
2220 return fib_trie_get_idx(iter, 0);
2222 v = fib_trie_get_next(iter);
2223 BUG_ON(v == l);
2224 if (v)
2225 return v;
2227 /* continue scan in next trie */
2228 if (iter->trie == trie_local)
2229 return fib_trie_get_first(iter, trie_main);
2231 return NULL;
2234 static void fib_trie_seq_stop(struct seq_file *seq, void *v)
2236 rcu_read_unlock();
2239 static void seq_indent(struct seq_file *seq, int n)
2241 while (n-- > 0) seq_puts(seq, " ");
2244 static inline const char *rtn_scope(enum rt_scope_t s)
2246 static char buf[32];
2248 switch(s) {
2249 case RT_SCOPE_UNIVERSE: return "universe";
2250 case RT_SCOPE_SITE: return "site";
2251 case RT_SCOPE_LINK: return "link";
2252 case RT_SCOPE_HOST: return "host";
2253 case RT_SCOPE_NOWHERE: return "nowhere";
2254 default:
2255 snprintf(buf, sizeof(buf), "scope=%d", s);
2256 return buf;
2260 static const char *rtn_type_names[__RTN_MAX] = {
2261 [RTN_UNSPEC] = "UNSPEC",
2262 [RTN_UNICAST] = "UNICAST",
2263 [RTN_LOCAL] = "LOCAL",
2264 [RTN_BROADCAST] = "BROADCAST",
2265 [RTN_ANYCAST] = "ANYCAST",
2266 [RTN_MULTICAST] = "MULTICAST",
2267 [RTN_BLACKHOLE] = "BLACKHOLE",
2268 [RTN_UNREACHABLE] = "UNREACHABLE",
2269 [RTN_PROHIBIT] = "PROHIBIT",
2270 [RTN_THROW] = "THROW",
2271 [RTN_NAT] = "NAT",
2272 [RTN_XRESOLVE] = "XRESOLVE",
2275 static inline const char *rtn_type(unsigned t)
2277 static char buf[32];
2279 if (t < __RTN_MAX && rtn_type_names[t])
2280 return rtn_type_names[t];
2281 snprintf(buf, sizeof(buf), "type %d", t);
2282 return buf;
2285 /* Pretty print the trie */
2286 static int fib_trie_seq_show(struct seq_file *seq, void *v)
2288 const struct fib_trie_iter *iter = seq->private;
2289 struct node *n = v;
2291 if (v == SEQ_START_TOKEN)
2292 return 0;
2294 if (IS_TNODE(n)) {
2295 struct tnode *tn = (struct tnode *) n;
2296 t_key prf = ntohl(MASK_PFX(tn->key, tn->pos));
2298 if (!NODE_PARENT(n)) {
2299 if (iter->trie == trie_local)
2300 seq_puts(seq, "<local>:\n");
2301 else
2302 seq_puts(seq, "<main>:\n");
2304 seq_indent(seq, iter->depth-1);
2305 seq_printf(seq, " +-- %d.%d.%d.%d/%d %d %d %d\n",
2306 NIPQUAD(prf), tn->pos, tn->bits, tn->full_children,
2307 tn->empty_children);
2309 } else {
2310 struct leaf *l = (struct leaf *) n;
2311 int i;
2312 u32 val = ntohl(l->key);
2314 seq_indent(seq, iter->depth);
2315 seq_printf(seq, " |-- %d.%d.%d.%d\n", NIPQUAD(val));
2316 for (i = 32; i >= 0; i--) {
2317 struct leaf_info *li = find_leaf_info(l, i);
2318 if (li) {
2319 struct fib_alias *fa;
2320 list_for_each_entry_rcu(fa, &li->falh, fa_list) {
2321 seq_indent(seq, iter->depth+1);
2322 seq_printf(seq, " /%d %s %s", i,
2323 rtn_scope(fa->fa_scope),
2324 rtn_type(fa->fa_type));
2325 if (fa->fa_tos)
2326 seq_printf(seq, "tos =%d\n",
2327 fa->fa_tos);
2328 seq_putc(seq, '\n');
2334 return 0;
2337 static struct seq_operations fib_trie_seq_ops = {
2338 .start = fib_trie_seq_start,
2339 .next = fib_trie_seq_next,
2340 .stop = fib_trie_seq_stop,
2341 .show = fib_trie_seq_show,
2344 static int fib_trie_seq_open(struct inode *inode, struct file *file)
2346 struct seq_file *seq;
2347 int rc = -ENOMEM;
2348 struct fib_trie_iter *s = kmalloc(sizeof(*s), GFP_KERNEL);
2350 if (!s)
2351 goto out;
2353 rc = seq_open(file, &fib_trie_seq_ops);
2354 if (rc)
2355 goto out_kfree;
2357 seq = file->private_data;
2358 seq->private = s;
2359 memset(s, 0, sizeof(*s));
2360 out:
2361 return rc;
2362 out_kfree:
2363 kfree(s);
2364 goto out;
2367 static struct file_operations fib_trie_fops = {
2368 .owner = THIS_MODULE,
2369 .open = fib_trie_seq_open,
2370 .read = seq_read,
2371 .llseek = seq_lseek,
2372 .release = seq_release_private,
2375 static unsigned fib_flag_trans(int type, u32 mask, const struct fib_info *fi)
2377 static unsigned type2flags[RTN_MAX + 1] = {
2378 [7] = RTF_REJECT, [8] = RTF_REJECT,
2380 unsigned flags = type2flags[type];
2382 if (fi && fi->fib_nh->nh_gw)
2383 flags |= RTF_GATEWAY;
2384 if (mask == 0xFFFFFFFF)
2385 flags |= RTF_HOST;
2386 flags |= RTF_UP;
2387 return flags;
2391 * This outputs /proc/net/route.
2392 * The format of the file is not supposed to be changed
2393 * and needs to be same as fib_hash output to avoid breaking
2394 * legacy utilities
2396 static int fib_route_seq_show(struct seq_file *seq, void *v)
2398 const struct fib_trie_iter *iter = seq->private;
2399 struct leaf *l = v;
2400 int i;
2401 char bf[128];
2403 if (v == SEQ_START_TOKEN) {
2404 seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
2405 "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
2406 "\tWindow\tIRTT");
2407 return 0;
2410 if (iter->trie == trie_local)
2411 return 0;
2412 if (IS_TNODE(l))
2413 return 0;
2415 for (i=32; i>=0; i--) {
2416 struct leaf_info *li = find_leaf_info(l, i);
2417 struct fib_alias *fa;
2418 u32 mask, prefix;
2420 if (!li)
2421 continue;
2423 mask = inet_make_mask(li->plen);
2424 prefix = htonl(l->key);
2426 list_for_each_entry_rcu(fa, &li->falh, fa_list) {
2427 const struct fib_info *fi = fa->fa_info;
2428 unsigned flags = fib_flag_trans(fa->fa_type, mask, fi);
2430 if (fa->fa_type == RTN_BROADCAST
2431 || fa->fa_type == RTN_MULTICAST)
2432 continue;
2434 if (fi)
2435 snprintf(bf, sizeof(bf),
2436 "%s\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u",
2437 fi->fib_dev ? fi->fib_dev->name : "*",
2438 prefix,
2439 fi->fib_nh->nh_gw, flags, 0, 0,
2440 fi->fib_priority,
2441 mask,
2442 (fi->fib_advmss ? fi->fib_advmss + 40 : 0),
2443 fi->fib_window,
2444 fi->fib_rtt >> 3);
2445 else
2446 snprintf(bf, sizeof(bf),
2447 "*\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u",
2448 prefix, 0, flags, 0, 0, 0,
2449 mask, 0, 0, 0);
2451 seq_printf(seq, "%-127s\n", bf);
2455 return 0;
2458 static struct seq_operations fib_route_seq_ops = {
2459 .start = fib_trie_seq_start,
2460 .next = fib_trie_seq_next,
2461 .stop = fib_trie_seq_stop,
2462 .show = fib_route_seq_show,
2465 static int fib_route_seq_open(struct inode *inode, struct file *file)
2467 struct seq_file *seq;
2468 int rc = -ENOMEM;
2469 struct fib_trie_iter *s = kmalloc(sizeof(*s), GFP_KERNEL);
2471 if (!s)
2472 goto out;
2474 rc = seq_open(file, &fib_route_seq_ops);
2475 if (rc)
2476 goto out_kfree;
2478 seq = file->private_data;
2479 seq->private = s;
2480 memset(s, 0, sizeof(*s));
2481 out:
2482 return rc;
2483 out_kfree:
2484 kfree(s);
2485 goto out;
2488 static struct file_operations fib_route_fops = {
2489 .owner = THIS_MODULE,
2490 .open = fib_route_seq_open,
2491 .read = seq_read,
2492 .llseek = seq_lseek,
2493 .release = seq_release_private,
2496 int __init fib_proc_init(void)
2498 if (!proc_net_fops_create("fib_trie", S_IRUGO, &fib_trie_fops))
2499 goto out1;
2501 if (!proc_net_fops_create("fib_triestat", S_IRUGO, &fib_triestat_fops))
2502 goto out2;
2504 if (!proc_net_fops_create("route", S_IRUGO, &fib_route_fops))
2505 goto out3;
2507 return 0;
2509 out3:
2510 proc_net_remove("fib_triestat");
2511 out2:
2512 proc_net_remove("fib_trie");
2513 out1:
2514 return -ENOMEM;
2517 void __init fib_proc_exit(void)
2519 proc_net_remove("fib_trie");
2520 proc_net_remove("fib_triestat");
2521 proc_net_remove("route");
2524 #endif /* CONFIG_PROC_FS */