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 described 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.csc.kth.se/~snilsson/software/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
26 * Code from fib_hash has been reused which includes the following header:
29 * INET An implementation of the TCP/IP protocol suite for the LINUX
30 * operating system. INET is implemented using the BSD Socket
31 * interface as the means of communication with the user level.
33 * IPv4 FIB: lookup engine and maintenance routines.
36 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
38 * This program is free software; you can redistribute it and/or
39 * modify it under the terms of the GNU General Public License
40 * as published by the Free Software Foundation; either version
41 * 2 of the License, or (at your option) any later version.
43 * Substantial contributions to this work comes from:
45 * David S. Miller, <davem@davemloft.net>
46 * Stephen Hemminger <shemminger@osdl.org>
47 * Paul E. McKenney <paulmck@us.ibm.com>
48 * Patrick McHardy <kaber@trash.net>
51 #define VERSION "0.409"
53 #include <asm/uaccess.h>
54 #include <linux/bitops.h>
55 #include <linux/types.h>
56 #include <linux/kernel.h>
58 #include <linux/string.h>
59 #include <linux/socket.h>
60 #include <linux/sockios.h>
61 #include <linux/errno.h>
63 #include <linux/inet.h>
64 #include <linux/inetdevice.h>
65 #include <linux/netdevice.h>
66 #include <linux/if_arp.h>
67 #include <linux/proc_fs.h>
68 #include <linux/rcupdate.h>
69 #include <linux/skbuff.h>
70 #include <linux/netlink.h>
71 #include <linux/init.h>
72 #include <linux/list.h>
73 #include <linux/slab.h>
74 #include <linux/export.h>
75 #include <linux/vmalloc.h>
76 #include <net/net_namespace.h>
78 #include <net/protocol.h>
79 #include <net/route.h>
82 #include <net/ip_fib.h>
83 #include <net/switchdev.h>
84 #include "fib_lookup.h"
86 #define MAX_STAT_DEPTH 32
88 #define KEYLENGTH (8*sizeof(t_key))
89 #define KEY_MAX ((t_key)~0)
91 typedef unsigned int t_key;
93 #define IS_TRIE(n) ((n)->pos >= KEYLENGTH)
94 #define IS_TNODE(n) ((n)->bits)
95 #define IS_LEAF(n) (!(n)->bits)
99 unsigned char pos; /* 2log(KEYLENGTH) bits needed */
100 unsigned char bits; /* 2log(KEYLENGTH) bits needed */
103 /* This list pointer if valid if (pos | bits) == 0 (LEAF) */
104 struct hlist_head leaf;
105 /* This array is valid if (pos | bits) > 0 (TNODE) */
106 struct key_vector __rcu *tnode[0];
112 t_key empty_children; /* KEYLENGTH bits needed */
113 t_key full_children; /* KEYLENGTH bits needed */
114 struct key_vector __rcu *parent;
115 struct key_vector kv[1];
116 #define tn_bits kv[0].bits
119 #define TNODE_SIZE(n) offsetof(struct tnode, kv[0].tnode[n])
120 #define LEAF_SIZE TNODE_SIZE(1)
122 #ifdef CONFIG_IP_FIB_TRIE_STATS
123 struct trie_use_stats {
125 unsigned int backtrack;
126 unsigned int semantic_match_passed;
127 unsigned int semantic_match_miss;
128 unsigned int null_node_hit;
129 unsigned int resize_node_skipped;
134 unsigned int totdepth;
135 unsigned int maxdepth;
138 unsigned int nullpointers;
139 unsigned int prefixes;
140 unsigned int nodesizes[MAX_STAT_DEPTH];
144 struct key_vector kv[1];
145 #ifdef CONFIG_IP_FIB_TRIE_STATS
146 struct trie_use_stats __percpu *stats;
150 static struct key_vector *resize(struct trie *t, struct key_vector *tn);
151 static size_t tnode_free_size;
154 * synchronize_rcu after call_rcu for that many pages; it should be especially
155 * useful before resizing the root node with PREEMPT_NONE configs; the value was
156 * obtained experimentally, aiming to avoid visible slowdown.
158 static const int sync_pages = 128;
160 static struct kmem_cache *fn_alias_kmem __read_mostly;
161 static struct kmem_cache *trie_leaf_kmem __read_mostly;
163 static inline struct tnode *tn_info(struct key_vector *kv)
165 return container_of(kv, struct tnode, kv[0]);
168 /* caller must hold RTNL */
169 #define node_parent(tn) rtnl_dereference(tn_info(tn)->parent)
170 #define get_child(tn, i) rtnl_dereference((tn)->tnode[i])
172 /* caller must hold RCU read lock or RTNL */
173 #define node_parent_rcu(tn) rcu_dereference_rtnl(tn_info(tn)->parent)
174 #define get_child_rcu(tn, i) rcu_dereference_rtnl((tn)->tnode[i])
176 /* wrapper for rcu_assign_pointer */
177 static inline void node_set_parent(struct key_vector *n, struct key_vector *tp)
180 rcu_assign_pointer(tn_info(n)->parent, tp);
183 #define NODE_INIT_PARENT(n, p) RCU_INIT_POINTER(tn_info(n)->parent, p)
185 /* This provides us with the number of children in this node, in the case of a
186 * leaf this will return 0 meaning none of the children are accessible.
188 static inline unsigned long child_length(const struct key_vector *tn)
190 return (1ul << tn->bits) & ~(1ul);
193 #define get_cindex(key, kv) (((key) ^ (kv)->key) >> (kv)->pos)
195 static inline unsigned long get_index(t_key key, struct key_vector *kv)
197 unsigned long index = key ^ kv->key;
199 if ((BITS_PER_LONG <= KEYLENGTH) && (KEYLENGTH == kv->pos))
202 return index >> kv->pos;
205 /* To understand this stuff, an understanding of keys and all their bits is
206 * necessary. Every node in the trie has a key associated with it, but not
207 * all of the bits in that key are significant.
209 * Consider a node 'n' and its parent 'tp'.
211 * If n is a leaf, every bit in its key is significant. Its presence is
212 * necessitated by path compression, since during a tree traversal (when
213 * searching for a leaf - unless we are doing an insertion) we will completely
214 * ignore all skipped bits we encounter. Thus we need to verify, at the end of
215 * a potentially successful search, that we have indeed been walking the
218 * Note that we can never "miss" the correct key in the tree if present by
219 * following the wrong path. Path compression ensures that segments of the key
220 * that are the same for all keys with a given prefix are skipped, but the
221 * skipped part *is* identical for each node in the subtrie below the skipped
222 * bit! trie_insert() in this implementation takes care of that.
224 * if n is an internal node - a 'tnode' here, the various parts of its key
225 * have many different meanings.
228 * _________________________________________________________________
229 * | i | i | i | i | i | i | i | N | N | N | S | S | S | S | S | C |
230 * -----------------------------------------------------------------
231 * 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16
233 * _________________________________________________________________
234 * | C | C | C | u | u | u | u | u | u | u | u | u | u | u | u | u |
235 * -----------------------------------------------------------------
236 * 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
243 * First, let's just ignore the bits that come before the parent tp, that is
244 * the bits from (tp->pos + tp->bits) to 31. They are *known* but at this
245 * point we do not use them for anything.
247 * The bits from (tp->pos) to (tp->pos + tp->bits - 1) - "N", above - are the
248 * index into the parent's child array. That is, they will be used to find
249 * 'n' among tp's children.
251 * The bits from (n->pos + n->bits) to (tn->pos - 1) - "S" - are skipped bits
254 * All the bits we have seen so far are significant to the node n. The rest
255 * of the bits are really not needed or indeed known in n->key.
257 * The bits from (n->pos) to (n->pos + n->bits - 1) - "C" - are the index into
258 * n's child array, and will of course be different for each child.
260 * The rest of the bits, from 0 to (n->pos + n->bits), are completely unknown
264 static const int halve_threshold = 25;
265 static const int inflate_threshold = 50;
266 static const int halve_threshold_root = 15;
267 static const int inflate_threshold_root = 30;
269 static void __alias_free_mem(struct rcu_head *head)
271 struct fib_alias *fa = container_of(head, struct fib_alias, rcu);
272 kmem_cache_free(fn_alias_kmem, fa);
275 static inline void alias_free_mem_rcu(struct fib_alias *fa)
277 call_rcu(&fa->rcu, __alias_free_mem);
280 #define TNODE_KMALLOC_MAX \
281 ilog2((PAGE_SIZE - TNODE_SIZE(0)) / sizeof(struct key_vector *))
282 #define TNODE_VMALLOC_MAX \
283 ilog2((SIZE_MAX - TNODE_SIZE(0)) / sizeof(struct key_vector *))
285 static void __node_free_rcu(struct rcu_head *head)
287 struct tnode *n = container_of(head, struct tnode, rcu);
290 kmem_cache_free(trie_leaf_kmem, n);
291 else if (n->tn_bits <= TNODE_KMALLOC_MAX)
297 #define node_free(n) call_rcu(&tn_info(n)->rcu, __node_free_rcu)
299 static struct tnode *tnode_alloc(int bits)
303 /* verify bits is within bounds */
304 if (bits > TNODE_VMALLOC_MAX)
307 /* determine size and verify it is non-zero and didn't overflow */
308 size = TNODE_SIZE(1ul << bits);
310 if (size <= PAGE_SIZE)
311 return kzalloc(size, GFP_KERNEL);
313 return vzalloc(size);
316 static inline void empty_child_inc(struct key_vector *n)
318 ++tn_info(n)->empty_children ? : ++tn_info(n)->full_children;
321 static inline void empty_child_dec(struct key_vector *n)
323 tn_info(n)->empty_children-- ? : tn_info(n)->full_children--;
326 static struct key_vector *leaf_new(t_key key, struct fib_alias *fa)
328 struct tnode *kv = kmem_cache_alloc(trie_leaf_kmem, GFP_KERNEL);
329 struct key_vector *l = kv->kv;
334 /* initialize key vector */
338 l->slen = fa->fa_slen;
340 /* link leaf to fib alias */
341 INIT_HLIST_HEAD(&l->leaf);
342 hlist_add_head(&fa->fa_list, &l->leaf);
347 static struct key_vector *tnode_new(t_key key, int pos, int bits)
349 struct tnode *tnode = tnode_alloc(bits);
350 unsigned int shift = pos + bits;
351 struct key_vector *tn = tnode->kv;
353 /* verify bits and pos their msb bits clear and values are valid */
354 BUG_ON(!bits || (shift > KEYLENGTH));
356 pr_debug("AT %p s=%zu %zu\n", tnode, TNODE_SIZE(0),
357 sizeof(struct key_vector *) << bits);
362 if (bits == KEYLENGTH)
363 tnode->full_children = 1;
365 tnode->empty_children = 1ul << bits;
367 tn->key = (shift < KEYLENGTH) ? (key >> shift) << shift : 0;
375 /* Check whether a tnode 'n' is "full", i.e. it is an internal node
376 * and no bits are skipped. See discussion in dyntree paper p. 6
378 static inline int tnode_full(struct key_vector *tn, struct key_vector *n)
380 return n && ((n->pos + n->bits) == tn->pos) && IS_TNODE(n);
383 /* Add a child at position i overwriting the old value.
384 * Update the value of full_children and empty_children.
386 static void put_child(struct key_vector *tn, unsigned long i,
387 struct key_vector *n)
389 struct key_vector *chi = get_child(tn, i);
392 BUG_ON(i >= child_length(tn));
394 /* update emptyChildren, overflow into fullChildren */
400 /* update fullChildren */
401 wasfull = tnode_full(tn, chi);
402 isfull = tnode_full(tn, n);
404 if (wasfull && !isfull)
405 tn_info(tn)->full_children--;
406 else if (!wasfull && isfull)
407 tn_info(tn)->full_children++;
409 if (n && (tn->slen < n->slen))
412 rcu_assign_pointer(tn->tnode[i], n);
415 static void update_children(struct key_vector *tn)
419 /* update all of the child parent pointers */
420 for (i = child_length(tn); i;) {
421 struct key_vector *inode = get_child(tn, --i);
426 /* Either update the children of a tnode that
427 * already belongs to us or update the child
428 * to point to ourselves.
430 if (node_parent(inode) == tn)
431 update_children(inode);
433 node_set_parent(inode, tn);
437 static inline void put_child_root(struct key_vector *tp, t_key key,
438 struct key_vector *n)
441 rcu_assign_pointer(tp->tnode[0], n);
443 put_child(tp, get_index(key, tp), n);
446 static inline void tnode_free_init(struct key_vector *tn)
448 tn_info(tn)->rcu.next = NULL;
451 static inline void tnode_free_append(struct key_vector *tn,
452 struct key_vector *n)
454 tn_info(n)->rcu.next = tn_info(tn)->rcu.next;
455 tn_info(tn)->rcu.next = &tn_info(n)->rcu;
458 static void tnode_free(struct key_vector *tn)
460 struct callback_head *head = &tn_info(tn)->rcu;
464 tnode_free_size += TNODE_SIZE(1ul << tn->bits);
467 tn = container_of(head, struct tnode, rcu)->kv;
470 if (tnode_free_size >= PAGE_SIZE * sync_pages) {
476 static struct key_vector *replace(struct trie *t,
477 struct key_vector *oldtnode,
478 struct key_vector *tn)
480 struct key_vector *tp = node_parent(oldtnode);
483 /* setup the parent pointer out of and back into this node */
484 NODE_INIT_PARENT(tn, tp);
485 put_child_root(tp, tn->key, tn);
487 /* update all of the child parent pointers */
490 /* all pointers should be clean so we are done */
491 tnode_free(oldtnode);
493 /* resize children now that oldtnode is freed */
494 for (i = child_length(tn); i;) {
495 struct key_vector *inode = get_child(tn, --i);
497 /* resize child node */
498 if (tnode_full(tn, inode))
499 tn = resize(t, inode);
505 static struct key_vector *inflate(struct trie *t,
506 struct key_vector *oldtnode)
508 struct key_vector *tn;
512 pr_debug("In inflate\n");
514 tn = tnode_new(oldtnode->key, oldtnode->pos - 1, oldtnode->bits + 1);
518 /* prepare oldtnode to be freed */
519 tnode_free_init(oldtnode);
521 /* Assemble all of the pointers in our cluster, in this case that
522 * represents all of the pointers out of our allocated nodes that
523 * point to existing tnodes and the links between our allocated
526 for (i = child_length(oldtnode), m = 1u << tn->pos; i;) {
527 struct key_vector *inode = get_child(oldtnode, --i);
528 struct key_vector *node0, *node1;
535 /* A leaf or an internal node with skipped bits */
536 if (!tnode_full(oldtnode, inode)) {
537 put_child(tn, get_index(inode->key, tn), inode);
541 /* drop the node in the old tnode free list */
542 tnode_free_append(oldtnode, inode);
544 /* An internal node with two children */
545 if (inode->bits == 1) {
546 put_child(tn, 2 * i + 1, get_child(inode, 1));
547 put_child(tn, 2 * i, get_child(inode, 0));
551 /* We will replace this node 'inode' with two new
552 * ones, 'node0' and 'node1', each with half of the
553 * original children. The two new nodes will have
554 * a position one bit further down the key and this
555 * means that the "significant" part of their keys
556 * (see the discussion near the top of this file)
557 * will differ by one bit, which will be "0" in
558 * node0's key and "1" in node1's key. Since we are
559 * moving the key position by one step, the bit that
560 * we are moving away from - the bit at position
561 * (tn->pos) - is the one that will differ between
562 * node0 and node1. So... we synthesize that bit in the
565 node1 = tnode_new(inode->key | m, inode->pos, inode->bits - 1);
568 node0 = tnode_new(inode->key, inode->pos, inode->bits - 1);
570 tnode_free_append(tn, node1);
573 tnode_free_append(tn, node0);
575 /* populate child pointers in new nodes */
576 for (k = child_length(inode), j = k / 2; j;) {
577 put_child(node1, --j, get_child(inode, --k));
578 put_child(node0, j, get_child(inode, j));
579 put_child(node1, --j, get_child(inode, --k));
580 put_child(node0, j, get_child(inode, j));
583 /* link new nodes to parent */
584 NODE_INIT_PARENT(node1, tn);
585 NODE_INIT_PARENT(node0, tn);
587 /* link parent to nodes */
588 put_child(tn, 2 * i + 1, node1);
589 put_child(tn, 2 * i, node0);
592 /* setup the parent pointers into and out of this node */
593 return replace(t, oldtnode, tn);
595 /* all pointers should be clean so we are done */
601 static struct key_vector *halve(struct trie *t,
602 struct key_vector *oldtnode)
604 struct key_vector *tn;
607 pr_debug("In halve\n");
609 tn = tnode_new(oldtnode->key, oldtnode->pos + 1, oldtnode->bits - 1);
613 /* prepare oldtnode to be freed */
614 tnode_free_init(oldtnode);
616 /* Assemble all of the pointers in our cluster, in this case that
617 * represents all of the pointers out of our allocated nodes that
618 * point to existing tnodes and the links between our allocated
621 for (i = child_length(oldtnode); i;) {
622 struct key_vector *node1 = get_child(oldtnode, --i);
623 struct key_vector *node0 = get_child(oldtnode, --i);
624 struct key_vector *inode;
626 /* At least one of the children is empty */
627 if (!node1 || !node0) {
628 put_child(tn, i / 2, node1 ? : node0);
632 /* Two nonempty children */
633 inode = tnode_new(node0->key, oldtnode->pos, 1);
636 tnode_free_append(tn, inode);
638 /* initialize pointers out of node */
639 put_child(inode, 1, node1);
640 put_child(inode, 0, node0);
641 NODE_INIT_PARENT(inode, tn);
643 /* link parent to node */
644 put_child(tn, i / 2, inode);
647 /* setup the parent pointers into and out of this node */
648 return replace(t, oldtnode, tn);
650 /* all pointers should be clean so we are done */
656 static struct key_vector *collapse(struct trie *t,
657 struct key_vector *oldtnode)
659 struct key_vector *n, *tp;
662 /* scan the tnode looking for that one child that might still exist */
663 for (n = NULL, i = child_length(oldtnode); !n && i;)
664 n = get_child(oldtnode, --i);
666 /* compress one level */
667 tp = node_parent(oldtnode);
668 put_child_root(tp, oldtnode->key, n);
669 node_set_parent(n, tp);
677 static unsigned char update_suffix(struct key_vector *tn)
679 unsigned char slen = tn->pos;
680 unsigned long stride, i;
682 /* search though the list of children looking for nodes that might
683 * have a suffix greater than the one we currently have. This is
684 * why we start with a stride of 2 since a stride of 1 would
685 * represent the nodes with suffix length equal to tn->pos
687 for (i = 0, stride = 0x2ul ; i < child_length(tn); i += stride) {
688 struct key_vector *n = get_child(tn, i);
690 if (!n || (n->slen <= slen))
693 /* update stride and slen based on new value */
694 stride <<= (n->slen - slen);
698 /* if slen covers all but the last bit we can stop here
699 * there will be nothing longer than that since only node
700 * 0 and 1 << (bits - 1) could have that as their suffix
703 if ((slen + 1) >= (tn->pos + tn->bits))
712 /* From "Implementing a dynamic compressed trie" by Stefan Nilsson of
713 * the Helsinki University of Technology and Matti Tikkanen of Nokia
714 * Telecommunications, page 6:
715 * "A node is doubled if the ratio of non-empty children to all
716 * children in the *doubled* node is at least 'high'."
718 * 'high' in this instance is the variable 'inflate_threshold'. It
719 * is expressed as a percentage, so we multiply it with
720 * child_length() and instead of multiplying by 2 (since the
721 * child array will be doubled by inflate()) and multiplying
722 * the left-hand side by 100 (to handle the percentage thing) we
723 * multiply the left-hand side by 50.
725 * The left-hand side may look a bit weird: child_length(tn)
726 * - tn->empty_children is of course the number of non-null children
727 * in the current node. tn->full_children is the number of "full"
728 * children, that is non-null tnodes with a skip value of 0.
729 * All of those will be doubled in the resulting inflated tnode, so
730 * we just count them one extra time here.
732 * A clearer way to write this would be:
734 * to_be_doubled = tn->full_children;
735 * not_to_be_doubled = child_length(tn) - tn->empty_children -
738 * new_child_length = child_length(tn) * 2;
740 * new_fill_factor = 100 * (not_to_be_doubled + 2*to_be_doubled) /
742 * if (new_fill_factor >= inflate_threshold)
744 * ...and so on, tho it would mess up the while () loop.
747 * 100 * (not_to_be_doubled + 2*to_be_doubled) / new_child_length >=
751 * 100 * (not_to_be_doubled + 2*to_be_doubled) >=
752 * inflate_threshold * new_child_length
754 * expand not_to_be_doubled and to_be_doubled, and shorten:
755 * 100 * (child_length(tn) - tn->empty_children +
756 * tn->full_children) >= inflate_threshold * new_child_length
758 * expand new_child_length:
759 * 100 * (child_length(tn) - tn->empty_children +
760 * tn->full_children) >=
761 * inflate_threshold * child_length(tn) * 2
764 * 50 * (tn->full_children + child_length(tn) -
765 * tn->empty_children) >= inflate_threshold *
769 static inline bool should_inflate(struct key_vector *tp, struct key_vector *tn)
771 unsigned long used = child_length(tn);
772 unsigned long threshold = used;
774 /* Keep root node larger */
775 threshold *= IS_TRIE(tp) ? inflate_threshold_root : inflate_threshold;
776 used -= tn_info(tn)->empty_children;
777 used += tn_info(tn)->full_children;
779 /* if bits == KEYLENGTH then pos = 0, and will fail below */
781 return (used > 1) && tn->pos && ((50 * used) >= threshold);
784 static inline bool should_halve(struct key_vector *tp, struct key_vector *tn)
786 unsigned long used = child_length(tn);
787 unsigned long threshold = used;
789 /* Keep root node larger */
790 threshold *= IS_TRIE(tp) ? halve_threshold_root : halve_threshold;
791 used -= tn_info(tn)->empty_children;
793 /* if bits == KEYLENGTH then used = 100% on wrap, and will fail below */
795 return (used > 1) && (tn->bits > 1) && ((100 * used) < threshold);
798 static inline bool should_collapse(struct key_vector *tn)
800 unsigned long used = child_length(tn);
802 used -= tn_info(tn)->empty_children;
804 /* account for bits == KEYLENGTH case */
805 if ((tn->bits == KEYLENGTH) && tn_info(tn)->full_children)
808 /* One child or none, time to drop us from the trie */
813 static struct key_vector *resize(struct trie *t, struct key_vector *tn)
815 #ifdef CONFIG_IP_FIB_TRIE_STATS
816 struct trie_use_stats __percpu *stats = t->stats;
818 struct key_vector *tp = node_parent(tn);
819 unsigned long cindex = get_index(tn->key, tp);
820 int max_work = MAX_WORK;
822 pr_debug("In tnode_resize %p inflate_threshold=%d threshold=%d\n",
823 tn, inflate_threshold, halve_threshold);
825 /* track the tnode via the pointer from the parent instead of
826 * doing it ourselves. This way we can let RCU fully do its
827 * thing without us interfering
829 BUG_ON(tn != get_child(tp, cindex));
831 /* Double as long as the resulting node has a number of
832 * nonempty nodes that are above the threshold.
834 while (should_inflate(tp, tn) && max_work) {
837 #ifdef CONFIG_IP_FIB_TRIE_STATS
838 this_cpu_inc(stats->resize_node_skipped);
844 tn = get_child(tp, cindex);
847 /* update parent in case inflate failed */
848 tp = node_parent(tn);
850 /* Return if at least one inflate is run */
851 if (max_work != MAX_WORK)
854 /* Halve as long as the number of empty children in this
855 * node is above threshold.
857 while (should_halve(tp, tn) && max_work) {
860 #ifdef CONFIG_IP_FIB_TRIE_STATS
861 this_cpu_inc(stats->resize_node_skipped);
867 tn = get_child(tp, cindex);
870 /* Only one child remains */
871 if (should_collapse(tn))
872 return collapse(t, tn);
874 /* update parent in case halve failed */
875 tp = node_parent(tn);
877 /* Return if at least one deflate was run */
878 if (max_work != MAX_WORK)
881 /* push the suffix length to the parent node */
882 if (tn->slen > tn->pos) {
883 unsigned char slen = update_suffix(tn);
892 static void leaf_pull_suffix(struct key_vector *tp, struct key_vector *l)
894 while ((tp->slen > tp->pos) && (tp->slen > l->slen)) {
895 if (update_suffix(tp) > l->slen)
897 tp = node_parent(tp);
901 static void leaf_push_suffix(struct key_vector *tn, struct key_vector *l)
903 /* if this is a new leaf then tn will be NULL and we can sort
904 * out parent suffix lengths as a part of trie_rebalance
906 while (tn->slen < l->slen) {
908 tn = node_parent(tn);
912 /* rcu_read_lock needs to be hold by caller from readside */
913 static struct key_vector *fib_find_node(struct trie *t,
914 struct key_vector **tp, u32 key)
916 struct key_vector *pn, *n = t->kv;
917 unsigned long index = 0;
921 n = get_child_rcu(n, index);
926 index = get_cindex(key, n);
928 /* This bit of code is a bit tricky but it combines multiple
929 * checks into a single check. The prefix consists of the
930 * prefix plus zeros for the bits in the cindex. The index
931 * is the difference between the key and this value. From
932 * this we can actually derive several pieces of data.
933 * if (index >= (1ul << bits))
934 * we have a mismatch in skip bits and failed
936 * we know the value is cindex
938 * This check is safe even if bits == KEYLENGTH due to the
939 * fact that we can only allocate a node with 32 bits if a
940 * long is greater than 32 bits.
942 if (index >= (1ul << n->bits)) {
947 /* keep searching until we find a perfect match leaf or NULL */
948 } while (IS_TNODE(n));
955 /* Return the first fib alias matching TOS with
956 * priority less than or equal to PRIO.
958 static struct fib_alias *fib_find_alias(struct hlist_head *fah, u8 slen,
959 u8 tos, u32 prio, u32 tb_id)
961 struct fib_alias *fa;
966 hlist_for_each_entry(fa, fah, fa_list) {
967 if (fa->fa_slen < slen)
969 if (fa->fa_slen != slen)
971 if (fa->tb_id > tb_id)
973 if (fa->tb_id != tb_id)
975 if (fa->fa_tos > tos)
977 if (fa->fa_info->fib_priority >= prio || fa->fa_tos < tos)
984 static void trie_rebalance(struct trie *t, struct key_vector *tn)
990 static int fib_insert_node(struct trie *t, struct key_vector *tp,
991 struct fib_alias *new, t_key key)
993 struct key_vector *n, *l;
995 l = leaf_new(key, new);
999 /* retrieve child from parent node */
1000 n = get_child(tp, get_index(key, tp));
1002 /* Case 2: n is a LEAF or a TNODE and the key doesn't match.
1004 * Add a new tnode here
1005 * first tnode need some special handling
1006 * leaves us in position for handling as case 3
1009 struct key_vector *tn;
1011 tn = tnode_new(key, __fls(key ^ n->key), 1);
1015 /* initialize routes out of node */
1016 NODE_INIT_PARENT(tn, tp);
1017 put_child(tn, get_index(key, tn) ^ 1, n);
1019 /* start adding routes into the node */
1020 put_child_root(tp, key, tn);
1021 node_set_parent(n, tn);
1023 /* parent now has a NULL spot where the leaf can go */
1027 /* Case 3: n is NULL, and will just insert a new leaf */
1028 NODE_INIT_PARENT(l, tp);
1029 put_child_root(tp, key, l);
1030 trie_rebalance(t, tp);
1039 static int fib_insert_alias(struct trie *t, struct key_vector *tp,
1040 struct key_vector *l, struct fib_alias *new,
1041 struct fib_alias *fa, t_key key)
1044 return fib_insert_node(t, tp, new, key);
1047 hlist_add_before_rcu(&new->fa_list, &fa->fa_list);
1049 struct fib_alias *last;
1051 hlist_for_each_entry(last, &l->leaf, fa_list) {
1052 if (new->fa_slen < last->fa_slen)
1054 if ((new->fa_slen == last->fa_slen) &&
1055 (new->tb_id > last->tb_id))
1061 hlist_add_behind_rcu(&new->fa_list, &fa->fa_list);
1063 hlist_add_head_rcu(&new->fa_list, &l->leaf);
1066 /* if we added to the tail node then we need to update slen */
1067 if (l->slen < new->fa_slen) {
1068 l->slen = new->fa_slen;
1069 leaf_push_suffix(tp, l);
1075 /* Caller must hold RTNL. */
1076 int fib_table_insert(struct fib_table *tb, struct fib_config *cfg)
1078 struct trie *t = (struct trie *)tb->tb_data;
1079 struct fib_alias *fa, *new_fa;
1080 struct key_vector *l, *tp;
1081 struct fib_info *fi;
1082 u8 plen = cfg->fc_dst_len;
1083 u8 slen = KEYLENGTH - plen;
1084 u8 tos = cfg->fc_tos;
1088 if (plen > KEYLENGTH)
1091 key = ntohl(cfg->fc_dst);
1093 pr_debug("Insert table=%u %08x/%d\n", tb->tb_id, key, plen);
1095 if ((plen < KEYLENGTH) && (key << plen))
1098 fi = fib_create_info(cfg);
1104 l = fib_find_node(t, &tp, key);
1105 fa = l ? fib_find_alias(&l->leaf, slen, tos, fi->fib_priority,
1108 /* Now fa, if non-NULL, points to the first fib alias
1109 * with the same keys [prefix,tos,priority], if such key already
1110 * exists or to the node before which we will insert new one.
1112 * If fa is NULL, we will need to allocate a new one and
1113 * insert to the tail of the section matching the suffix length
1117 if (fa && fa->fa_tos == tos &&
1118 fa->fa_info->fib_priority == fi->fib_priority) {
1119 struct fib_alias *fa_first, *fa_match;
1122 if (cfg->fc_nlflags & NLM_F_EXCL)
1126 * 1. Find exact match for type, scope, fib_info to avoid
1128 * 2. Find next 'fa' (or head), NLM_F_APPEND inserts before it
1132 hlist_for_each_entry_from(fa, fa_list) {
1133 if ((fa->fa_slen != slen) ||
1134 (fa->tb_id != tb->tb_id) ||
1135 (fa->fa_tos != tos))
1137 if (fa->fa_info->fib_priority != fi->fib_priority)
1139 if (fa->fa_type == cfg->fc_type &&
1140 fa->fa_info == fi) {
1146 if (cfg->fc_nlflags & NLM_F_REPLACE) {
1147 struct fib_info *fi_drop;
1157 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
1161 fi_drop = fa->fa_info;
1162 new_fa->fa_tos = fa->fa_tos;
1163 new_fa->fa_info = fi;
1164 new_fa->fa_type = cfg->fc_type;
1165 state = fa->fa_state;
1166 new_fa->fa_state = state & ~FA_S_ACCESSED;
1167 new_fa->fa_slen = fa->fa_slen;
1168 new_fa->tb_id = tb->tb_id;
1170 err = switchdev_fib_ipv4_add(key, plen, fi,
1176 switchdev_fib_ipv4_abort(fi);
1177 kmem_cache_free(fn_alias_kmem, new_fa);
1181 hlist_replace_rcu(&fa->fa_list, &new_fa->fa_list);
1183 alias_free_mem_rcu(fa);
1185 fib_release_info(fi_drop);
1186 if (state & FA_S_ACCESSED)
1187 rt_cache_flush(cfg->fc_nlinfo.nl_net);
1188 rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen,
1189 tb->tb_id, &cfg->fc_nlinfo, NLM_F_REPLACE);
1193 /* Error if we find a perfect match which
1194 * uses the same scope, type, and nexthop
1200 if (!(cfg->fc_nlflags & NLM_F_APPEND))
1204 if (!(cfg->fc_nlflags & NLM_F_CREATE))
1208 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
1212 new_fa->fa_info = fi;
1213 new_fa->fa_tos = tos;
1214 new_fa->fa_type = cfg->fc_type;
1215 new_fa->fa_state = 0;
1216 new_fa->fa_slen = slen;
1217 new_fa->tb_id = tb->tb_id;
1219 /* (Optionally) offload fib entry to switch hardware. */
1220 err = switchdev_fib_ipv4_add(key, plen, fi, tos, cfg->fc_type,
1221 cfg->fc_nlflags, tb->tb_id);
1223 switchdev_fib_ipv4_abort(fi);
1224 goto out_free_new_fa;
1227 /* Insert new entry to the list. */
1228 err = fib_insert_alias(t, tp, l, new_fa, fa, key);
1230 goto out_sw_fib_del;
1233 tb->tb_num_default++;
1235 rt_cache_flush(cfg->fc_nlinfo.nl_net);
1236 rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen, new_fa->tb_id,
1237 &cfg->fc_nlinfo, 0);
1242 switchdev_fib_ipv4_del(key, plen, fi, tos, cfg->fc_type, tb->tb_id);
1244 kmem_cache_free(fn_alias_kmem, new_fa);
1246 fib_release_info(fi);
1251 static inline t_key prefix_mismatch(t_key key, struct key_vector *n)
1253 t_key prefix = n->key;
1255 return (key ^ prefix) & (prefix | -prefix);
1258 /* should be called with rcu_read_lock */
1259 int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp,
1260 struct fib_result *res, int fib_flags)
1262 struct trie *t = (struct trie *) tb->tb_data;
1263 #ifdef CONFIG_IP_FIB_TRIE_STATS
1264 struct trie_use_stats __percpu *stats = t->stats;
1266 const t_key key = ntohl(flp->daddr);
1267 struct key_vector *n, *pn;
1268 struct fib_alias *fa;
1269 unsigned long index;
1275 n = get_child_rcu(pn, cindex);
1279 #ifdef CONFIG_IP_FIB_TRIE_STATS
1280 this_cpu_inc(stats->gets);
1283 /* Step 1: Travel to the longest prefix match in the trie */
1285 index = get_cindex(key, n);
1287 /* This bit of code is a bit tricky but it combines multiple
1288 * checks into a single check. The prefix consists of the
1289 * prefix plus zeros for the "bits" in the prefix. The index
1290 * is the difference between the key and this value. From
1291 * this we can actually derive several pieces of data.
1292 * if (index >= (1ul << bits))
1293 * we have a mismatch in skip bits and failed
1295 * we know the value is cindex
1297 * This check is safe even if bits == KEYLENGTH due to the
1298 * fact that we can only allocate a node with 32 bits if a
1299 * long is greater than 32 bits.
1301 if (index >= (1ul << n->bits))
1304 /* we have found a leaf. Prefixes have already been compared */
1308 /* only record pn and cindex if we are going to be chopping
1309 * bits later. Otherwise we are just wasting cycles.
1311 if (n->slen > n->pos) {
1316 n = get_child_rcu(n, index);
1321 /* Step 2: Sort out leaves and begin backtracing for longest prefix */
1323 /* record the pointer where our next node pointer is stored */
1324 struct key_vector __rcu **cptr = n->tnode;
1326 /* This test verifies that none of the bits that differ
1327 * between the key and the prefix exist in the region of
1328 * the lsb and higher in the prefix.
1330 if (unlikely(prefix_mismatch(key, n)) || (n->slen == n->pos))
1333 /* exit out and process leaf */
1334 if (unlikely(IS_LEAF(n)))
1337 /* Don't bother recording parent info. Since we are in
1338 * prefix match mode we will have to come back to wherever
1339 * we started this traversal anyway
1342 while ((n = rcu_dereference(*cptr)) == NULL) {
1344 #ifdef CONFIG_IP_FIB_TRIE_STATS
1346 this_cpu_inc(stats->null_node_hit);
1348 /* If we are at cindex 0 there are no more bits for
1349 * us to strip at this level so we must ascend back
1350 * up one level to see if there are any more bits to
1351 * be stripped there.
1354 t_key pkey = pn->key;
1356 /* If we don't have a parent then there is
1357 * nothing for us to do as we do not have any
1358 * further nodes to parse.
1362 #ifdef CONFIG_IP_FIB_TRIE_STATS
1363 this_cpu_inc(stats->backtrack);
1365 /* Get Child's index */
1366 pn = node_parent_rcu(pn);
1367 cindex = get_index(pkey, pn);
1370 /* strip the least significant bit from the cindex */
1371 cindex &= cindex - 1;
1373 /* grab pointer for next child node */
1374 cptr = &pn->tnode[cindex];
1379 /* this line carries forward the xor from earlier in the function */
1380 index = key ^ n->key;
1382 /* Step 3: Process the leaf, if that fails fall back to backtracing */
1383 hlist_for_each_entry_rcu(fa, &n->leaf, fa_list) {
1384 struct fib_info *fi = fa->fa_info;
1387 if ((index >= (1ul << fa->fa_slen)) &&
1388 ((BITS_PER_LONG > KEYLENGTH) || (fa->fa_slen != KEYLENGTH)))
1390 if (fa->fa_tos && fa->fa_tos != flp->flowi4_tos)
1394 if (fa->fa_info->fib_scope < flp->flowi4_scope)
1396 fib_alias_accessed(fa);
1397 err = fib_props[fa->fa_type].error;
1398 if (unlikely(err < 0)) {
1399 #ifdef CONFIG_IP_FIB_TRIE_STATS
1400 this_cpu_inc(stats->semantic_match_passed);
1404 if (fi->fib_flags & RTNH_F_DEAD)
1406 for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) {
1407 const struct fib_nh *nh = &fi->fib_nh[nhsel];
1409 if (nh->nh_flags & RTNH_F_DEAD)
1411 if (flp->flowi4_oif && flp->flowi4_oif != nh->nh_oif)
1414 if (!(fib_flags & FIB_LOOKUP_NOREF))
1415 atomic_inc(&fi->fib_clntref);
1417 res->prefixlen = KEYLENGTH - fa->fa_slen;
1418 res->nh_sel = nhsel;
1419 res->type = fa->fa_type;
1420 res->scope = fi->fib_scope;
1423 res->fa_head = &n->leaf;
1424 #ifdef CONFIG_IP_FIB_TRIE_STATS
1425 this_cpu_inc(stats->semantic_match_passed);
1430 #ifdef CONFIG_IP_FIB_TRIE_STATS
1431 this_cpu_inc(stats->semantic_match_miss);
1435 EXPORT_SYMBOL_GPL(fib_table_lookup);
1437 static void fib_remove_alias(struct trie *t, struct key_vector *tp,
1438 struct key_vector *l, struct fib_alias *old)
1440 /* record the location of the previous list_info entry */
1441 struct hlist_node **pprev = old->fa_list.pprev;
1442 struct fib_alias *fa = hlist_entry(pprev, typeof(*fa), fa_list.next);
1444 /* remove the fib_alias from the list */
1445 hlist_del_rcu(&old->fa_list);
1447 /* if we emptied the list this leaf will be freed and we can sort
1448 * out parent suffix lengths as a part of trie_rebalance
1450 if (hlist_empty(&l->leaf)) {
1451 put_child_root(tp, l->key, NULL);
1453 trie_rebalance(t, tp);
1457 /* only access fa if it is pointing at the last valid hlist_node */
1461 /* update the trie with the latest suffix length */
1462 l->slen = fa->fa_slen;
1463 leaf_pull_suffix(tp, l);
1466 /* Caller must hold RTNL. */
1467 int fib_table_delete(struct fib_table *tb, struct fib_config *cfg)
1469 struct trie *t = (struct trie *) tb->tb_data;
1470 struct fib_alias *fa, *fa_to_delete;
1471 struct key_vector *l, *tp;
1472 u8 plen = cfg->fc_dst_len;
1473 u8 slen = KEYLENGTH - plen;
1474 u8 tos = cfg->fc_tos;
1477 if (plen > KEYLENGTH)
1480 key = ntohl(cfg->fc_dst);
1482 if ((plen < KEYLENGTH) && (key << plen))
1485 l = fib_find_node(t, &tp, key);
1489 fa = fib_find_alias(&l->leaf, slen, tos, 0, tb->tb_id);
1493 pr_debug("Deleting %08x/%d tos=%d t=%p\n", key, plen, tos, t);
1495 fa_to_delete = NULL;
1496 hlist_for_each_entry_from(fa, fa_list) {
1497 struct fib_info *fi = fa->fa_info;
1499 if ((fa->fa_slen != slen) ||
1500 (fa->tb_id != tb->tb_id) ||
1501 (fa->fa_tos != tos))
1504 if ((!cfg->fc_type || fa->fa_type == cfg->fc_type) &&
1505 (cfg->fc_scope == RT_SCOPE_NOWHERE ||
1506 fa->fa_info->fib_scope == cfg->fc_scope) &&
1507 (!cfg->fc_prefsrc ||
1508 fi->fib_prefsrc == cfg->fc_prefsrc) &&
1509 (!cfg->fc_protocol ||
1510 fi->fib_protocol == cfg->fc_protocol) &&
1511 fib_nh_match(cfg, fi) == 0) {
1520 switchdev_fib_ipv4_del(key, plen, fa_to_delete->fa_info, tos,
1521 cfg->fc_type, tb->tb_id);
1523 rtmsg_fib(RTM_DELROUTE, htonl(key), fa_to_delete, plen, tb->tb_id,
1524 &cfg->fc_nlinfo, 0);
1527 tb->tb_num_default--;
1529 fib_remove_alias(t, tp, l, fa_to_delete);
1531 if (fa_to_delete->fa_state & FA_S_ACCESSED)
1532 rt_cache_flush(cfg->fc_nlinfo.nl_net);
1534 fib_release_info(fa_to_delete->fa_info);
1535 alias_free_mem_rcu(fa_to_delete);
1539 /* Scan for the next leaf starting at the provided key value */
1540 static struct key_vector *leaf_walk_rcu(struct key_vector **tn, t_key key)
1542 struct key_vector *pn, *n = *tn;
1543 unsigned long cindex;
1545 /* this loop is meant to try and find the key in the trie */
1547 /* record parent and next child index */
1549 cindex = key ? get_index(key, pn) : 0;
1551 if (cindex >> pn->bits)
1554 /* descend into the next child */
1555 n = get_child_rcu(pn, cindex++);
1559 /* guarantee forward progress on the keys */
1560 if (IS_LEAF(n) && (n->key >= key))
1562 } while (IS_TNODE(n));
1564 /* this loop will search for the next leaf with a greater key */
1565 while (!IS_TRIE(pn)) {
1566 /* if we exhausted the parent node we will need to climb */
1567 if (cindex >= (1ul << pn->bits)) {
1568 t_key pkey = pn->key;
1570 pn = node_parent_rcu(pn);
1571 cindex = get_index(pkey, pn) + 1;
1575 /* grab the next available node */
1576 n = get_child_rcu(pn, cindex++);
1580 /* no need to compare keys since we bumped the index */
1584 /* Rescan start scanning in new node */
1590 return NULL; /* Root of trie */
1592 /* if we are at the limit for keys just return NULL for the tnode */
1597 static void fib_trie_free(struct fib_table *tb)
1599 struct trie *t = (struct trie *)tb->tb_data;
1600 struct key_vector *pn = t->kv;
1601 unsigned long cindex = 1;
1602 struct hlist_node *tmp;
1603 struct fib_alias *fa;
1605 /* walk trie in reverse order and free everything */
1607 struct key_vector *n;
1610 t_key pkey = pn->key;
1616 pn = node_parent(pn);
1618 /* drop emptied tnode */
1619 put_child_root(pn, n->key, NULL);
1622 cindex = get_index(pkey, pn);
1627 /* grab the next available node */
1628 n = get_child(pn, cindex);
1633 /* record pn and cindex for leaf walking */
1635 cindex = 1ul << n->bits;
1640 hlist_for_each_entry_safe(fa, tmp, &n->leaf, fa_list) {
1641 hlist_del_rcu(&fa->fa_list);
1642 alias_free_mem_rcu(fa);
1645 put_child_root(pn, n->key, NULL);
1649 #ifdef CONFIG_IP_FIB_TRIE_STATS
1650 free_percpu(t->stats);
1655 struct fib_table *fib_trie_unmerge(struct fib_table *oldtb)
1657 struct trie *ot = (struct trie *)oldtb->tb_data;
1658 struct key_vector *l, *tp = ot->kv;
1659 struct fib_table *local_tb;
1660 struct fib_alias *fa;
1664 if (oldtb->tb_data == oldtb->__data)
1667 local_tb = fib_trie_table(RT_TABLE_LOCAL, NULL);
1671 lt = (struct trie *)local_tb->tb_data;
1673 while ((l = leaf_walk_rcu(&tp, key)) != NULL) {
1674 struct key_vector *local_l = NULL, *local_tp;
1676 hlist_for_each_entry_rcu(fa, &l->leaf, fa_list) {
1677 struct fib_alias *new_fa;
1679 if (local_tb->tb_id != fa->tb_id)
1682 /* clone fa for new local table */
1683 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
1687 memcpy(new_fa, fa, sizeof(*fa));
1689 /* insert clone into table */
1691 local_l = fib_find_node(lt, &local_tp, l->key);
1693 if (fib_insert_alias(lt, local_tp, local_l, new_fa,
1698 /* stop loop if key wrapped back to 0 */
1706 fib_trie_free(local_tb);
1711 /* Caller must hold RTNL */
1712 void fib_table_flush_external(struct fib_table *tb)
1714 struct trie *t = (struct trie *)tb->tb_data;
1715 struct key_vector *pn = t->kv;
1716 unsigned long cindex = 1;
1717 struct hlist_node *tmp;
1718 struct fib_alias *fa;
1720 /* walk trie in reverse order */
1722 unsigned char slen = 0;
1723 struct key_vector *n;
1726 t_key pkey = pn->key;
1728 /* cannot resize the trie vector */
1732 /* resize completed node */
1734 cindex = get_index(pkey, pn);
1739 /* grab the next available node */
1740 n = get_child(pn, cindex);
1745 /* record pn and cindex for leaf walking */
1747 cindex = 1ul << n->bits;
1752 hlist_for_each_entry_safe(fa, tmp, &n->leaf, fa_list) {
1753 struct fib_info *fi = fa->fa_info;
1755 /* if alias was cloned to local then we just
1756 * need to remove the local copy from main
1758 if (tb->tb_id != fa->tb_id) {
1759 hlist_del_rcu(&fa->fa_list);
1760 alias_free_mem_rcu(fa);
1764 /* record local slen */
1767 if (!fi || !(fi->fib_flags & RTNH_F_OFFLOAD))
1770 switchdev_fib_ipv4_del(n->key, KEYLENGTH - fa->fa_slen,
1771 fi, fa->fa_tos, fa->fa_type,
1775 /* update leaf slen */
1778 if (hlist_empty(&n->leaf)) {
1779 put_child_root(pn, n->key, NULL);
1782 leaf_pull_suffix(pn, n);
1787 /* Caller must hold RTNL. */
1788 int fib_table_flush(struct fib_table *tb)
1790 struct trie *t = (struct trie *)tb->tb_data;
1791 struct key_vector *pn = t->kv;
1792 unsigned long cindex = 1;
1793 struct hlist_node *tmp;
1794 struct fib_alias *fa;
1797 /* walk trie in reverse order */
1799 unsigned char slen = 0;
1800 struct key_vector *n;
1803 t_key pkey = pn->key;
1805 /* cannot resize the trie vector */
1809 /* resize completed node */
1811 cindex = get_index(pkey, pn);
1816 /* grab the next available node */
1817 n = get_child(pn, cindex);
1822 /* record pn and cindex for leaf walking */
1824 cindex = 1ul << n->bits;
1829 hlist_for_each_entry_safe(fa, tmp, &n->leaf, fa_list) {
1830 struct fib_info *fi = fa->fa_info;
1832 if (!fi || !(fi->fib_flags & RTNH_F_DEAD)) {
1837 switchdev_fib_ipv4_del(n->key, KEYLENGTH - fa->fa_slen,
1838 fi, fa->fa_tos, fa->fa_type,
1840 hlist_del_rcu(&fa->fa_list);
1841 fib_release_info(fa->fa_info);
1842 alias_free_mem_rcu(fa);
1846 /* update leaf slen */
1849 if (hlist_empty(&n->leaf)) {
1850 put_child_root(pn, n->key, NULL);
1853 leaf_pull_suffix(pn, n);
1857 pr_debug("trie_flush found=%d\n", found);
1861 static void __trie_free_rcu(struct rcu_head *head)
1863 struct fib_table *tb = container_of(head, struct fib_table, rcu);
1864 #ifdef CONFIG_IP_FIB_TRIE_STATS
1865 struct trie *t = (struct trie *)tb->tb_data;
1867 if (tb->tb_data == tb->__data)
1868 free_percpu(t->stats);
1869 #endif /* CONFIG_IP_FIB_TRIE_STATS */
1873 void fib_free_table(struct fib_table *tb)
1875 call_rcu(&tb->rcu, __trie_free_rcu);
1878 static int fn_trie_dump_leaf(struct key_vector *l, struct fib_table *tb,
1879 struct sk_buff *skb, struct netlink_callback *cb)
1881 __be32 xkey = htonl(l->key);
1882 struct fib_alias *fa;
1888 /* rcu_read_lock is hold by caller */
1889 hlist_for_each_entry_rcu(fa, &l->leaf, fa_list) {
1895 if (tb->tb_id != fa->tb_id) {
1900 if (fib_dump_info(skb, NETLINK_CB(cb->skb).portid,
1906 KEYLENGTH - fa->fa_slen,
1908 fa->fa_info, NLM_F_MULTI) < 0) {
1919 /* rcu_read_lock needs to be hold by caller from readside */
1920 int fib_table_dump(struct fib_table *tb, struct sk_buff *skb,
1921 struct netlink_callback *cb)
1923 struct trie *t = (struct trie *)tb->tb_data;
1924 struct key_vector *l, *tp = t->kv;
1925 /* Dump starting at last key.
1926 * Note: 0.0.0.0/0 (ie default) is first key.
1928 int count = cb->args[2];
1929 t_key key = cb->args[3];
1931 while ((l = leaf_walk_rcu(&tp, key)) != NULL) {
1932 if (fn_trie_dump_leaf(l, tb, skb, cb) < 0) {
1934 cb->args[2] = count;
1941 memset(&cb->args[4], 0,
1942 sizeof(cb->args) - 4*sizeof(cb->args[0]));
1944 /* stop loop if key wrapped back to 0 */
1950 cb->args[2] = count;
1955 void __init fib_trie_init(void)
1957 fn_alias_kmem = kmem_cache_create("ip_fib_alias",
1958 sizeof(struct fib_alias),
1959 0, SLAB_PANIC, NULL);
1961 trie_leaf_kmem = kmem_cache_create("ip_fib_trie",
1963 0, SLAB_PANIC, NULL);
1966 struct fib_table *fib_trie_table(u32 id, struct fib_table *alias)
1968 struct fib_table *tb;
1970 size_t sz = sizeof(*tb);
1973 sz += sizeof(struct trie);
1975 tb = kzalloc(sz, GFP_KERNEL);
1980 tb->tb_default = -1;
1981 tb->tb_num_default = 0;
1982 tb->tb_data = (alias ? alias->__data : tb->__data);
1987 t = (struct trie *) tb->tb_data;
1988 t->kv[0].pos = KEYLENGTH;
1989 t->kv[0].slen = KEYLENGTH;
1990 #ifdef CONFIG_IP_FIB_TRIE_STATS
1991 t->stats = alloc_percpu(struct trie_use_stats);
2001 #ifdef CONFIG_PROC_FS
2002 /* Depth first Trie walk iterator */
2003 struct fib_trie_iter {
2004 struct seq_net_private p;
2005 struct fib_table *tb;
2006 struct key_vector *tnode;
2011 static struct key_vector *fib_trie_get_next(struct fib_trie_iter *iter)
2013 unsigned long cindex = iter->index;
2014 struct key_vector *pn = iter->tnode;
2017 pr_debug("get_next iter={node=%p index=%d depth=%d}\n",
2018 iter->tnode, iter->index, iter->depth);
2020 while (!IS_TRIE(pn)) {
2021 while (cindex < child_length(pn)) {
2022 struct key_vector *n = get_child_rcu(pn, cindex++);
2029 iter->index = cindex;
2031 /* push down one level */
2040 /* Current node exhausted, pop back up */
2042 pn = node_parent_rcu(pn);
2043 cindex = get_index(pkey, pn) + 1;
2047 /* record root node so further searches know we are done */
2054 static struct key_vector *fib_trie_get_first(struct fib_trie_iter *iter,
2057 struct key_vector *n, *pn = t->kv;
2062 n = rcu_dereference(pn->tnode[0]);
2079 static void trie_collect_stats(struct trie *t, struct trie_stat *s)
2081 struct key_vector *n;
2082 struct fib_trie_iter iter;
2084 memset(s, 0, sizeof(*s));
2087 for (n = fib_trie_get_first(&iter, t); n; n = fib_trie_get_next(&iter)) {
2089 struct fib_alias *fa;
2092 s->totdepth += iter.depth;
2093 if (iter.depth > s->maxdepth)
2094 s->maxdepth = iter.depth;
2096 hlist_for_each_entry_rcu(fa, &n->leaf, fa_list)
2100 if (n->bits < MAX_STAT_DEPTH)
2101 s->nodesizes[n->bits]++;
2102 s->nullpointers += tn_info(n)->empty_children;
2109 * This outputs /proc/net/fib_triestats
2111 static void trie_show_stats(struct seq_file *seq, struct trie_stat *stat)
2113 unsigned int i, max, pointers, bytes, avdepth;
2116 avdepth = stat->totdepth*100 / stat->leaves;
2120 seq_printf(seq, "\tAver depth: %u.%02d\n",
2121 avdepth / 100, avdepth % 100);
2122 seq_printf(seq, "\tMax depth: %u\n", stat->maxdepth);
2124 seq_printf(seq, "\tLeaves: %u\n", stat->leaves);
2125 bytes = LEAF_SIZE * stat->leaves;
2127 seq_printf(seq, "\tPrefixes: %u\n", stat->prefixes);
2128 bytes += sizeof(struct fib_alias) * stat->prefixes;
2130 seq_printf(seq, "\tInternal nodes: %u\n\t", stat->tnodes);
2131 bytes += TNODE_SIZE(0) * stat->tnodes;
2133 max = MAX_STAT_DEPTH;
2134 while (max > 0 && stat->nodesizes[max-1] == 0)
2138 for (i = 1; i < max; i++)
2139 if (stat->nodesizes[i] != 0) {
2140 seq_printf(seq, " %u: %u", i, stat->nodesizes[i]);
2141 pointers += (1<<i) * stat->nodesizes[i];
2143 seq_putc(seq, '\n');
2144 seq_printf(seq, "\tPointers: %u\n", pointers);
2146 bytes += sizeof(struct key_vector *) * pointers;
2147 seq_printf(seq, "Null ptrs: %u\n", stat->nullpointers);
2148 seq_printf(seq, "Total size: %u kB\n", (bytes + 1023) / 1024);
2151 #ifdef CONFIG_IP_FIB_TRIE_STATS
2152 static void trie_show_usage(struct seq_file *seq,
2153 const struct trie_use_stats __percpu *stats)
2155 struct trie_use_stats s = { 0 };
2158 /* loop through all of the CPUs and gather up the stats */
2159 for_each_possible_cpu(cpu) {
2160 const struct trie_use_stats *pcpu = per_cpu_ptr(stats, cpu);
2162 s.gets += pcpu->gets;
2163 s.backtrack += pcpu->backtrack;
2164 s.semantic_match_passed += pcpu->semantic_match_passed;
2165 s.semantic_match_miss += pcpu->semantic_match_miss;
2166 s.null_node_hit += pcpu->null_node_hit;
2167 s.resize_node_skipped += pcpu->resize_node_skipped;
2170 seq_printf(seq, "\nCounters:\n---------\n");
2171 seq_printf(seq, "gets = %u\n", s.gets);
2172 seq_printf(seq, "backtracks = %u\n", s.backtrack);
2173 seq_printf(seq, "semantic match passed = %u\n",
2174 s.semantic_match_passed);
2175 seq_printf(seq, "semantic match miss = %u\n", s.semantic_match_miss);
2176 seq_printf(seq, "null node hit= %u\n", s.null_node_hit);
2177 seq_printf(seq, "skipped node resize = %u\n\n", s.resize_node_skipped);
2179 #endif /* CONFIG_IP_FIB_TRIE_STATS */
2181 static void fib_table_print(struct seq_file *seq, struct fib_table *tb)
2183 if (tb->tb_id == RT_TABLE_LOCAL)
2184 seq_puts(seq, "Local:\n");
2185 else if (tb->tb_id == RT_TABLE_MAIN)
2186 seq_puts(seq, "Main:\n");
2188 seq_printf(seq, "Id %d:\n", tb->tb_id);
2192 static int fib_triestat_seq_show(struct seq_file *seq, void *v)
2194 struct net *net = (struct net *)seq->private;
2198 "Basic info: size of leaf:"
2199 " %Zd bytes, size of tnode: %Zd bytes.\n",
2200 LEAF_SIZE, TNODE_SIZE(0));
2202 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
2203 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
2204 struct fib_table *tb;
2206 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
2207 struct trie *t = (struct trie *) tb->tb_data;
2208 struct trie_stat stat;
2213 fib_table_print(seq, tb);
2215 trie_collect_stats(t, &stat);
2216 trie_show_stats(seq, &stat);
2217 #ifdef CONFIG_IP_FIB_TRIE_STATS
2218 trie_show_usage(seq, t->stats);
2226 static int fib_triestat_seq_open(struct inode *inode, struct file *file)
2228 return single_open_net(inode, file, fib_triestat_seq_show);
2231 static const struct file_operations fib_triestat_fops = {
2232 .owner = THIS_MODULE,
2233 .open = fib_triestat_seq_open,
2235 .llseek = seq_lseek,
2236 .release = single_release_net,
2239 static struct key_vector *fib_trie_get_idx(struct seq_file *seq, loff_t pos)
2241 struct fib_trie_iter *iter = seq->private;
2242 struct net *net = seq_file_net(seq);
2246 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
2247 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
2248 struct fib_table *tb;
2250 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
2251 struct key_vector *n;
2253 for (n = fib_trie_get_first(iter,
2254 (struct trie *) tb->tb_data);
2255 n; n = fib_trie_get_next(iter))
2266 static void *fib_trie_seq_start(struct seq_file *seq, loff_t *pos)
2270 return fib_trie_get_idx(seq, *pos);
2273 static void *fib_trie_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2275 struct fib_trie_iter *iter = seq->private;
2276 struct net *net = seq_file_net(seq);
2277 struct fib_table *tb = iter->tb;
2278 struct hlist_node *tb_node;
2280 struct key_vector *n;
2283 /* next node in same table */
2284 n = fib_trie_get_next(iter);
2288 /* walk rest of this hash chain */
2289 h = tb->tb_id & (FIB_TABLE_HASHSZ - 1);
2290 while ((tb_node = rcu_dereference(hlist_next_rcu(&tb->tb_hlist)))) {
2291 tb = hlist_entry(tb_node, struct fib_table, tb_hlist);
2292 n = fib_trie_get_first(iter, (struct trie *) tb->tb_data);
2297 /* new hash chain */
2298 while (++h < FIB_TABLE_HASHSZ) {
2299 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
2300 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
2301 n = fib_trie_get_first(iter, (struct trie *) tb->tb_data);
2313 static void fib_trie_seq_stop(struct seq_file *seq, void *v)
2319 static void seq_indent(struct seq_file *seq, int n)
2325 static inline const char *rtn_scope(char *buf, size_t len, enum rt_scope_t s)
2328 case RT_SCOPE_UNIVERSE: return "universe";
2329 case RT_SCOPE_SITE: return "site";
2330 case RT_SCOPE_LINK: return "link";
2331 case RT_SCOPE_HOST: return "host";
2332 case RT_SCOPE_NOWHERE: return "nowhere";
2334 snprintf(buf, len, "scope=%d", s);
2339 static const char *const rtn_type_names[__RTN_MAX] = {
2340 [RTN_UNSPEC] = "UNSPEC",
2341 [RTN_UNICAST] = "UNICAST",
2342 [RTN_LOCAL] = "LOCAL",
2343 [RTN_BROADCAST] = "BROADCAST",
2344 [RTN_ANYCAST] = "ANYCAST",
2345 [RTN_MULTICAST] = "MULTICAST",
2346 [RTN_BLACKHOLE] = "BLACKHOLE",
2347 [RTN_UNREACHABLE] = "UNREACHABLE",
2348 [RTN_PROHIBIT] = "PROHIBIT",
2349 [RTN_THROW] = "THROW",
2351 [RTN_XRESOLVE] = "XRESOLVE",
2354 static inline const char *rtn_type(char *buf, size_t len, unsigned int t)
2356 if (t < __RTN_MAX && rtn_type_names[t])
2357 return rtn_type_names[t];
2358 snprintf(buf, len, "type %u", t);
2362 /* Pretty print the trie */
2363 static int fib_trie_seq_show(struct seq_file *seq, void *v)
2365 const struct fib_trie_iter *iter = seq->private;
2366 struct key_vector *n = v;
2368 if (IS_TRIE(node_parent_rcu(n)))
2369 fib_table_print(seq, iter->tb);
2372 __be32 prf = htonl(n->key);
2374 seq_indent(seq, iter->depth-1);
2375 seq_printf(seq, " +-- %pI4/%zu %u %u %u\n",
2376 &prf, KEYLENGTH - n->pos - n->bits, n->bits,
2377 tn_info(n)->full_children,
2378 tn_info(n)->empty_children);
2380 __be32 val = htonl(n->key);
2381 struct fib_alias *fa;
2383 seq_indent(seq, iter->depth);
2384 seq_printf(seq, " |-- %pI4\n", &val);
2386 hlist_for_each_entry_rcu(fa, &n->leaf, fa_list) {
2387 char buf1[32], buf2[32];
2389 seq_indent(seq, iter->depth + 1);
2390 seq_printf(seq, " /%zu %s %s",
2391 KEYLENGTH - fa->fa_slen,
2392 rtn_scope(buf1, sizeof(buf1),
2393 fa->fa_info->fib_scope),
2394 rtn_type(buf2, sizeof(buf2),
2397 seq_printf(seq, " tos=%d", fa->fa_tos);
2398 seq_putc(seq, '\n');
2405 static const struct seq_operations fib_trie_seq_ops = {
2406 .start = fib_trie_seq_start,
2407 .next = fib_trie_seq_next,
2408 .stop = fib_trie_seq_stop,
2409 .show = fib_trie_seq_show,
2412 static int fib_trie_seq_open(struct inode *inode, struct file *file)
2414 return seq_open_net(inode, file, &fib_trie_seq_ops,
2415 sizeof(struct fib_trie_iter));
2418 static const struct file_operations fib_trie_fops = {
2419 .owner = THIS_MODULE,
2420 .open = fib_trie_seq_open,
2422 .llseek = seq_lseek,
2423 .release = seq_release_net,
2426 struct fib_route_iter {
2427 struct seq_net_private p;
2428 struct fib_table *main_tb;
2429 struct key_vector *tnode;
2434 static struct key_vector *fib_route_get_idx(struct fib_route_iter *iter,
2437 struct fib_table *tb = iter->main_tb;
2438 struct key_vector *l, **tp = &iter->tnode;
2442 /* use cache location of next-to-find key */
2443 if (iter->pos > 0 && pos >= iter->pos) {
2447 t = (struct trie *)tb->tb_data;
2448 iter->tnode = t->kv;
2453 while ((l = leaf_walk_rcu(tp, key)) != NULL) {
2462 /* handle unlikely case of a key wrap */
2468 iter->key = key; /* remember it */
2470 iter->pos = 0; /* forget it */
2475 static void *fib_route_seq_start(struct seq_file *seq, loff_t *pos)
2478 struct fib_route_iter *iter = seq->private;
2479 struct fib_table *tb;
2484 tb = fib_get_table(seq_file_net(seq), RT_TABLE_MAIN);
2491 return fib_route_get_idx(iter, *pos);
2493 t = (struct trie *)tb->tb_data;
2494 iter->tnode = t->kv;
2498 return SEQ_START_TOKEN;
2501 static void *fib_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2503 struct fib_route_iter *iter = seq->private;
2504 struct key_vector *l = NULL;
2505 t_key key = iter->key;
2509 /* only allow key of 0 for start of sequence */
2510 if ((v == SEQ_START_TOKEN) || key)
2511 l = leaf_walk_rcu(&iter->tnode, key);
2514 iter->key = l->key + 1;
2523 static void fib_route_seq_stop(struct seq_file *seq, void *v)
2529 static unsigned int fib_flag_trans(int type, __be32 mask, const struct fib_info *fi)
2531 unsigned int flags = 0;
2533 if (type == RTN_UNREACHABLE || type == RTN_PROHIBIT)
2535 if (fi && fi->fib_nh->nh_gw)
2536 flags |= RTF_GATEWAY;
2537 if (mask == htonl(0xFFFFFFFF))
2544 * This outputs /proc/net/route.
2545 * The format of the file is not supposed to be changed
2546 * and needs to be same as fib_hash output to avoid breaking
2549 static int fib_route_seq_show(struct seq_file *seq, void *v)
2551 struct fib_route_iter *iter = seq->private;
2552 struct fib_table *tb = iter->main_tb;
2553 struct fib_alias *fa;
2554 struct key_vector *l = v;
2557 if (v == SEQ_START_TOKEN) {
2558 seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
2559 "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
2564 prefix = htonl(l->key);
2566 hlist_for_each_entry_rcu(fa, &l->leaf, fa_list) {
2567 const struct fib_info *fi = fa->fa_info;
2568 __be32 mask = inet_make_mask(KEYLENGTH - fa->fa_slen);
2569 unsigned int flags = fib_flag_trans(fa->fa_type, mask, fi);
2571 if ((fa->fa_type == RTN_BROADCAST) ||
2572 (fa->fa_type == RTN_MULTICAST))
2575 if (fa->tb_id != tb->tb_id)
2578 seq_setwidth(seq, 127);
2582 "%s\t%08X\t%08X\t%04X\t%d\t%u\t"
2583 "%d\t%08X\t%d\t%u\t%u",
2584 fi->fib_dev ? fi->fib_dev->name : "*",
2586 fi->fib_nh->nh_gw, flags, 0, 0,
2590 fi->fib_advmss + 40 : 0),
2595 "*\t%08X\t%08X\t%04X\t%d\t%u\t"
2596 "%d\t%08X\t%d\t%u\t%u",
2597 prefix, 0, flags, 0, 0, 0,
2606 static const struct seq_operations fib_route_seq_ops = {
2607 .start = fib_route_seq_start,
2608 .next = fib_route_seq_next,
2609 .stop = fib_route_seq_stop,
2610 .show = fib_route_seq_show,
2613 static int fib_route_seq_open(struct inode *inode, struct file *file)
2615 return seq_open_net(inode, file, &fib_route_seq_ops,
2616 sizeof(struct fib_route_iter));
2619 static const struct file_operations fib_route_fops = {
2620 .owner = THIS_MODULE,
2621 .open = fib_route_seq_open,
2623 .llseek = seq_lseek,
2624 .release = seq_release_net,
2627 int __net_init fib_proc_init(struct net *net)
2629 if (!proc_create("fib_trie", S_IRUGO, net->proc_net, &fib_trie_fops))
2632 if (!proc_create("fib_triestat", S_IRUGO, net->proc_net,
2633 &fib_triestat_fops))
2636 if (!proc_create("route", S_IRUGO, net->proc_net, &fib_route_fops))
2642 remove_proc_entry("fib_triestat", net->proc_net);
2644 remove_proc_entry("fib_trie", net->proc_net);
2649 void __net_exit fib_proc_exit(struct net *net)
2651 remove_proc_entry("fib_trie", net->proc_net);
2652 remove_proc_entry("fib_triestat", net->proc_net);
2653 remove_proc_entry("route", net->proc_net);
2656 #endif /* CONFIG_PROC_FS */