2 * Linux INET6 implementation
3 * Forwarding Information Database
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * $Id: ip6_fib.c,v 1.25 2001/10/31 21:55:55 davem Exp $
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
18 * Yuji SEKIYA @USAGI: Support default route on router node;
19 * remove ip6_null_entry from the top of
21 * Ville Nuorvala: Fixed routing subtrees.
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/net.h>
26 #include <linux/route.h>
27 #include <linux/netdevice.h>
28 #include <linux/in6.h>
29 #include <linux/init.h>
30 #include <linux/list.h>
33 #include <linux/proc_fs.h>
37 #include <net/ndisc.h>
38 #include <net/addrconf.h>
40 #include <net/ip6_fib.h>
41 #include <net/ip6_route.h>
46 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
48 #define RT6_TRACE(x...) do { ; } while (0)
51 struct rt6_statistics rt6_stats;
53 static struct kmem_cache * fib6_node_kmem __read_mostly;
57 #ifdef CONFIG_IPV6_SUBTREES
68 struct fib6_walker_t w;
69 int (*func)(struct rt6_info *, void *arg);
73 static DEFINE_RWLOCK(fib6_walker_lock);
75 #ifdef CONFIG_IPV6_SUBTREES
76 #define FWS_INIT FWS_S
78 #define FWS_INIT FWS_L
81 static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt);
82 static struct rt6_info * fib6_find_prefix(struct fib6_node *fn);
83 static struct fib6_node * fib6_repair_tree(struct fib6_node *fn);
84 static int fib6_walk(struct fib6_walker_t *w);
85 static int fib6_walk_continue(struct fib6_walker_t *w);
88 * A routing update causes an increase of the serial number on the
89 * affected subtree. This allows for cached routes to be asynchronously
90 * tested when modifications are made to the destination cache as a
91 * result of redirects, path MTU changes, etc.
94 static __u32 rt_sernum;
96 static DEFINE_TIMER(ip6_fib_timer, fib6_run_gc, 0, 0);
98 static struct fib6_walker_t fib6_walker_list = {
99 .prev = &fib6_walker_list,
100 .next = &fib6_walker_list,
103 #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next)
105 static inline void fib6_walker_link(struct fib6_walker_t *w)
107 write_lock_bh(&fib6_walker_lock);
108 w->next = fib6_walker_list.next;
109 w->prev = &fib6_walker_list;
112 write_unlock_bh(&fib6_walker_lock);
115 static inline void fib6_walker_unlink(struct fib6_walker_t *w)
117 write_lock_bh(&fib6_walker_lock);
118 w->next->prev = w->prev;
119 w->prev->next = w->next;
120 w->prev = w->next = w;
121 write_unlock_bh(&fib6_walker_lock);
123 static __inline__ u32 fib6_new_sernum(void)
132 * Auxiliary address test functions for the radix tree.
134 * These assume a 32bit processor (although it will work on
142 static __inline__ __be32 addr_bit_set(void *token, int fn_bit)
144 __be32 *addr = token;
146 return htonl(1 << ((~fn_bit)&0x1F)) & addr[fn_bit>>5];
149 static __inline__ struct fib6_node * node_alloc(void)
151 struct fib6_node *fn;
153 fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
158 static __inline__ void node_free(struct fib6_node * fn)
160 kmem_cache_free(fib6_node_kmem, fn);
163 static __inline__ void rt6_release(struct rt6_info *rt)
165 if (atomic_dec_and_test(&rt->rt6i_ref))
166 dst_free(&rt->u.dst);
169 static struct fib6_table *fib6_main_tbl;
171 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
172 #define FIB_TABLE_HASHSZ 256
174 #define FIB_TABLE_HASHSZ 1
176 static struct hlist_head *fib_table_hash;
178 static void fib6_link_table(struct fib6_table *tb)
183 * Initialize table lock at a single place to give lockdep a key,
184 * tables aren't visible prior to being linked to the list.
186 rwlock_init(&tb->tb6_lock);
188 h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1);
191 * No protection necessary, this is the only list mutatation
192 * operation, tables never disappear once they exist.
194 hlist_add_head_rcu(&tb->tb6_hlist, &fib_table_hash[h]);
197 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
199 static struct fib6_table *fib6_local_tbl;
201 static struct fib6_table *fib6_alloc_table(u32 id)
203 struct fib6_table *table;
205 table = kzalloc(sizeof(*table), GFP_ATOMIC);
208 table->tb6_root.leaf = &ip6_null_entry;
209 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
215 struct fib6_table *fib6_new_table(u32 id)
217 struct fib6_table *tb;
221 tb = fib6_get_table(id);
225 tb = fib6_alloc_table(id);
232 struct fib6_table *fib6_get_table(u32 id)
234 struct fib6_table *tb;
235 struct hlist_node *node;
240 h = id & (FIB_TABLE_HASHSZ - 1);
242 hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb6_hlist) {
243 if (tb->tb6_id == id) {
253 static void __init fib6_tables_init(void)
255 fib6_link_table(fib6_main_tbl);
256 fib6_link_table(fib6_local_tbl);
261 struct fib6_table *fib6_new_table(u32 id)
263 return fib6_get_table(id);
266 struct fib6_table *fib6_get_table(u32 id)
268 return fib6_main_tbl;
271 struct dst_entry *fib6_rule_lookup(struct flowi *fl, int flags,
274 return (struct dst_entry *) lookup(fib6_main_tbl, fl, flags);
277 static void __init fib6_tables_init(void)
279 fib6_link_table(fib6_main_tbl);
284 static int fib6_dump_node(struct fib6_walker_t *w)
289 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
290 res = rt6_dump_route(rt, w->args);
292 /* Frame is full, suspend walking */
302 static void fib6_dump_end(struct netlink_callback *cb)
304 struct fib6_walker_t *w = (void*)cb->args[2];
310 cb->done = (void*)cb->args[3];
314 static int fib6_dump_done(struct netlink_callback *cb)
317 return cb->done ? cb->done(cb) : 0;
320 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
321 struct netlink_callback *cb)
323 struct fib6_walker_t *w;
326 w = (void *)cb->args[2];
327 w->root = &table->tb6_root;
329 if (cb->args[4] == 0) {
330 read_lock_bh(&table->tb6_lock);
332 read_unlock_bh(&table->tb6_lock);
336 read_lock_bh(&table->tb6_lock);
337 res = fib6_walk_continue(w);
338 read_unlock_bh(&table->tb6_lock);
341 fib6_walker_unlink(w);
344 fib6_walker_unlink(w);
351 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
353 struct net *net = skb->sk->sk_net;
355 unsigned int e = 0, s_e;
356 struct rt6_rtnl_dump_arg arg;
357 struct fib6_walker_t *w;
358 struct fib6_table *tb;
359 struct hlist_node *node;
362 if (net != &init_net)
368 w = (void *)cb->args[2];
372 * 1. hook callback destructor.
374 cb->args[3] = (long)cb->done;
375 cb->done = fib6_dump_done;
378 * 2. allocate and initialize walker.
380 w = kzalloc(sizeof(*w), GFP_ATOMIC);
383 w->func = fib6_dump_node;
384 cb->args[2] = (long)w;
391 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
393 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb6_hlist) {
396 res = fib6_dump_table(tb, skb, cb);
407 res = res < 0 ? res : skb->len;
416 * return the appropriate node for a routing tree "add" operation
417 * by either creating and inserting or by returning an existing
421 static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr,
422 int addrlen, int plen,
425 struct fib6_node *fn, *in, *ln;
426 struct fib6_node *pn = NULL;
430 __u32 sernum = fib6_new_sernum();
432 RT6_TRACE("fib6_add_1\n");
434 /* insert node in tree */
439 key = (struct rt6key *)((u8 *)fn->leaf + offset);
444 if (plen < fn->fn_bit ||
445 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
452 if (plen == fn->fn_bit) {
453 /* clean up an intermediate node */
454 if ((fn->fn_flags & RTN_RTINFO) == 0) {
455 rt6_release(fn->leaf);
459 fn->fn_sernum = sernum;
465 * We have more bits to go
468 /* Try to walk down on tree. */
469 fn->fn_sernum = sernum;
470 dir = addr_bit_set(addr, fn->fn_bit);
472 fn = dir ? fn->right: fn->left;
476 * We walked to the bottom of tree.
477 * Create new leaf node without children.
487 ln->fn_sernum = sernum;
499 * split since we don't have a common prefix anymore or
500 * we have a less significant route.
501 * we've to insert an intermediate node on the list
502 * this new node will point to the one we need to create
508 /* find 1st bit in difference between the 2 addrs.
510 See comment in __ipv6_addr_diff: bit may be an invalid value,
511 but if it is >= plen, the value is ignored in any case.
514 bit = __ipv6_addr_diff(addr, &key->addr, addrlen);
519 * (new leaf node)[ln] (old node)[fn]
525 if (in == NULL || ln == NULL) {
534 * new intermediate node.
536 * be off since that an address that chooses one of
537 * the branches would not match less specific routes
538 * in the other branch
545 atomic_inc(&in->leaf->rt6i_ref);
547 in->fn_sernum = sernum;
549 /* update parent pointer */
560 ln->fn_sernum = sernum;
562 if (addr_bit_set(addr, bit)) {
569 } else { /* plen <= bit */
572 * (new leaf node)[ln]
574 * (old node)[fn] NULL
586 ln->fn_sernum = sernum;
593 if (addr_bit_set(&key->addr, plen))
604 * Insert routing information in a node.
607 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
608 struct nl_info *info)
610 struct rt6_info *iter = NULL;
611 struct rt6_info **ins;
615 for (iter = fn->leaf; iter; iter=iter->u.dst.rt6_next) {
617 * Search for duplicates
620 if (iter->rt6i_metric == rt->rt6i_metric) {
622 * Same priority level
625 if (iter->rt6i_dev == rt->rt6i_dev &&
626 iter->rt6i_idev == rt->rt6i_idev &&
627 ipv6_addr_equal(&iter->rt6i_gateway,
628 &rt->rt6i_gateway)) {
629 if (!(iter->rt6i_flags&RTF_EXPIRES))
631 iter->rt6i_expires = rt->rt6i_expires;
632 if (!(rt->rt6i_flags&RTF_EXPIRES)) {
633 iter->rt6i_flags &= ~RTF_EXPIRES;
634 iter->rt6i_expires = 0;
640 if (iter->rt6i_metric > rt->rt6i_metric)
643 ins = &iter->u.dst.rt6_next;
646 /* Reset round-robin state, if necessary */
647 if (ins == &fn->leaf)
654 rt->u.dst.rt6_next = iter;
657 atomic_inc(&rt->rt6i_ref);
658 inet6_rt_notify(RTM_NEWROUTE, rt, info);
659 rt6_stats.fib_rt_entries++;
661 if ((fn->fn_flags & RTN_RTINFO) == 0) {
662 rt6_stats.fib_route_nodes++;
663 fn->fn_flags |= RTN_RTINFO;
669 static __inline__ void fib6_start_gc(struct rt6_info *rt)
671 if (ip6_fib_timer.expires == 0 &&
672 (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE)))
673 mod_timer(&ip6_fib_timer, jiffies +
674 init_net.ipv6.sysctl.ip6_rt_gc_interval);
677 void fib6_force_start_gc(void)
679 if (ip6_fib_timer.expires == 0)
680 mod_timer(&ip6_fib_timer, jiffies +
681 init_net.ipv6.sysctl.ip6_rt_gc_interval);
685 * Add routing information to the routing tree.
686 * <destination addr>/<source addr>
687 * with source addr info in sub-trees
690 int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info)
692 struct fib6_node *fn, *pn = NULL;
695 fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr),
696 rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst));
703 #ifdef CONFIG_IPV6_SUBTREES
704 if (rt->rt6i_src.plen) {
705 struct fib6_node *sn;
707 if (fn->subtree == NULL) {
708 struct fib6_node *sfn;
720 /* Create subtree root node */
725 sfn->leaf = &ip6_null_entry;
726 atomic_inc(&ip6_null_entry.rt6i_ref);
727 sfn->fn_flags = RTN_ROOT;
728 sfn->fn_sernum = fib6_new_sernum();
730 /* Now add the first leaf node to new subtree */
732 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
733 sizeof(struct in6_addr), rt->rt6i_src.plen,
734 offsetof(struct rt6_info, rt6i_src));
737 /* If it is failed, discard just allocated
738 root, and then (in st_failure) stale node
745 /* Now link new subtree to main tree */
749 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
750 sizeof(struct in6_addr), rt->rt6i_src.plen,
751 offsetof(struct rt6_info, rt6i_src));
757 if (fn->leaf == NULL) {
759 atomic_inc(&rt->rt6i_ref);
765 err = fib6_add_rt2node(fn, rt, info);
769 if (!(rt->rt6i_flags&RTF_CACHE))
770 fib6_prune_clones(pn, rt);
775 #ifdef CONFIG_IPV6_SUBTREES
777 * If fib6_add_1 has cleared the old leaf pointer in the
778 * super-tree leaf node we have to find a new one for it.
780 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
781 pn->leaf = fib6_find_prefix(pn);
784 BUG_TRAP(pn->leaf != NULL);
785 pn->leaf = &ip6_null_entry;
788 atomic_inc(&pn->leaf->rt6i_ref);
791 dst_free(&rt->u.dst);
795 #ifdef CONFIG_IPV6_SUBTREES
796 /* Subtree creation failed, probably main tree node
797 is orphan. If it is, shoot it.
800 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
801 fib6_repair_tree(fn);
802 dst_free(&rt->u.dst);
808 * Routing tree lookup
813 int offset; /* key offset on rt6_info */
814 struct in6_addr *addr; /* search key */
817 static struct fib6_node * fib6_lookup_1(struct fib6_node *root,
818 struct lookup_args *args)
820 struct fib6_node *fn;
823 if (unlikely(args->offset == 0))
833 struct fib6_node *next;
835 dir = addr_bit_set(args->addr, fn->fn_bit);
837 next = dir ? fn->right : fn->left;
848 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
851 key = (struct rt6key *) ((u8 *) fn->leaf +
854 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
855 #ifdef CONFIG_IPV6_SUBTREES
857 fn = fib6_lookup_1(fn->subtree, args + 1);
859 if (!fn || fn->fn_flags & RTN_RTINFO)
864 if (fn->fn_flags & RTN_ROOT)
873 struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr,
874 struct in6_addr *saddr)
876 struct fib6_node *fn;
877 struct lookup_args args[] = {
879 .offset = offsetof(struct rt6_info, rt6i_dst),
882 #ifdef CONFIG_IPV6_SUBTREES
884 .offset = offsetof(struct rt6_info, rt6i_src),
889 .offset = 0, /* sentinel */
893 fn = fib6_lookup_1(root, daddr ? args : args + 1);
895 if (fn == NULL || fn->fn_flags & RTN_TL_ROOT)
902 * Get node with specified destination prefix (and source prefix,
903 * if subtrees are used)
907 static struct fib6_node * fib6_locate_1(struct fib6_node *root,
908 struct in6_addr *addr,
909 int plen, int offset)
911 struct fib6_node *fn;
913 for (fn = root; fn ; ) {
914 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
919 if (plen < fn->fn_bit ||
920 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
923 if (plen == fn->fn_bit)
927 * We have more bits to go
929 if (addr_bit_set(addr, fn->fn_bit))
937 struct fib6_node * fib6_locate(struct fib6_node *root,
938 struct in6_addr *daddr, int dst_len,
939 struct in6_addr *saddr, int src_len)
941 struct fib6_node *fn;
943 fn = fib6_locate_1(root, daddr, dst_len,
944 offsetof(struct rt6_info, rt6i_dst));
946 #ifdef CONFIG_IPV6_SUBTREES
948 BUG_TRAP(saddr!=NULL);
949 if (fn && fn->subtree)
950 fn = fib6_locate_1(fn->subtree, saddr, src_len,
951 offsetof(struct rt6_info, rt6i_src));
955 if (fn && fn->fn_flags&RTN_RTINFO)
967 static struct rt6_info * fib6_find_prefix(struct fib6_node *fn)
969 if (fn->fn_flags&RTN_ROOT)
970 return &ip6_null_entry;
974 return fn->left->leaf;
977 return fn->right->leaf;
979 fn = FIB6_SUBTREE(fn);
985 * Called to trim the tree of intermediate nodes when possible. "fn"
986 * is the node we want to try and remove.
989 static struct fib6_node * fib6_repair_tree(struct fib6_node *fn)
993 struct fib6_node *child, *pn;
994 struct fib6_walker_t *w;
998 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1001 BUG_TRAP(!(fn->fn_flags&RTN_RTINFO));
1002 BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT));
1003 BUG_TRAP(fn->leaf==NULL);
1007 if (fn->right) child = fn->right, children |= 1;
1008 if (fn->left) child = fn->left, children |= 2;
1010 if (children == 3 || FIB6_SUBTREE(fn)
1011 #ifdef CONFIG_IPV6_SUBTREES
1012 /* Subtree root (i.e. fn) may have one child */
1013 || (children && fn->fn_flags&RTN_ROOT)
1016 fn->leaf = fib6_find_prefix(fn);
1018 if (fn->leaf==NULL) {
1020 fn->leaf = &ip6_null_entry;
1023 atomic_inc(&fn->leaf->rt6i_ref);
1028 #ifdef CONFIG_IPV6_SUBTREES
1029 if (FIB6_SUBTREE(pn) == fn) {
1030 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1031 FIB6_SUBTREE(pn) = NULL;
1034 BUG_TRAP(!(fn->fn_flags&RTN_ROOT));
1036 if (pn->right == fn) pn->right = child;
1037 else if (pn->left == fn) pn->left = child;
1044 #ifdef CONFIG_IPV6_SUBTREES
1048 read_lock(&fib6_walker_lock);
1050 if (child == NULL) {
1051 if (w->root == fn) {
1052 w->root = w->node = NULL;
1053 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1054 } else if (w->node == fn) {
1055 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1060 if (w->root == fn) {
1062 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1064 if (w->node == fn) {
1067 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1068 w->state = w->state>=FWS_R ? FWS_U : FWS_INIT;
1070 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1071 w->state = w->state>=FWS_C ? FWS_U : FWS_INIT;
1076 read_unlock(&fib6_walker_lock);
1079 if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn))
1082 rt6_release(pn->leaf);
1088 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1089 struct nl_info *info)
1091 struct fib6_walker_t *w;
1092 struct rt6_info *rt = *rtp;
1094 RT6_TRACE("fib6_del_route\n");
1097 *rtp = rt->u.dst.rt6_next;
1098 rt->rt6i_node = NULL;
1099 rt6_stats.fib_rt_entries--;
1100 rt6_stats.fib_discarded_routes++;
1102 /* Reset round-robin state, if necessary */
1103 if (fn->rr_ptr == rt)
1106 /* Adjust walkers */
1107 read_lock(&fib6_walker_lock);
1109 if (w->state == FWS_C && w->leaf == rt) {
1110 RT6_TRACE("walker %p adjusted by delroute\n", w);
1111 w->leaf = rt->u.dst.rt6_next;
1112 if (w->leaf == NULL)
1116 read_unlock(&fib6_walker_lock);
1118 rt->u.dst.rt6_next = NULL;
1120 /* If it was last route, expunge its radix tree node */
1121 if (fn->leaf == NULL) {
1122 fn->fn_flags &= ~RTN_RTINFO;
1123 rt6_stats.fib_route_nodes--;
1124 fn = fib6_repair_tree(fn);
1127 if (atomic_read(&rt->rt6i_ref) != 1) {
1128 /* This route is used as dummy address holder in some split
1129 * nodes. It is not leaked, but it still holds other resources,
1130 * which must be released in time. So, scan ascendant nodes
1131 * and replace dummy references to this route with references
1132 * to still alive ones.
1135 if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) {
1136 fn->leaf = fib6_find_prefix(fn);
1137 atomic_inc(&fn->leaf->rt6i_ref);
1142 /* No more references are possible at this point. */
1143 BUG_ON(atomic_read(&rt->rt6i_ref) != 1);
1146 inet6_rt_notify(RTM_DELROUTE, rt, info);
1150 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1152 struct fib6_node *fn = rt->rt6i_node;
1153 struct rt6_info **rtp;
1156 if (rt->u.dst.obsolete>0) {
1161 if (fn == NULL || rt == &ip6_null_entry)
1164 BUG_TRAP(fn->fn_flags&RTN_RTINFO);
1166 if (!(rt->rt6i_flags&RTF_CACHE)) {
1167 struct fib6_node *pn = fn;
1168 #ifdef CONFIG_IPV6_SUBTREES
1169 /* clones of this route might be in another subtree */
1170 if (rt->rt6i_src.plen) {
1171 while (!(pn->fn_flags&RTN_ROOT))
1176 fib6_prune_clones(pn, rt);
1180 * Walk the leaf entries looking for ourself
1183 for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.dst.rt6_next) {
1185 fib6_del_route(fn, rtp, info);
1193 * Tree traversal function.
1195 * Certainly, it is not interrupt safe.
1196 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1197 * It means, that we can modify tree during walking
1198 * and use this function for garbage collection, clone pruning,
1199 * cleaning tree when a device goes down etc. etc.
1201 * It guarantees that every node will be traversed,
1202 * and that it will be traversed only once.
1204 * Callback function w->func may return:
1205 * 0 -> continue walking.
1206 * positive value -> walking is suspended (used by tree dumps,
1207 * and probably by gc, if it will be split to several slices)
1208 * negative value -> terminate walking.
1210 * The function itself returns:
1211 * 0 -> walk is complete.
1212 * >0 -> walk is incomplete (i.e. suspended)
1213 * <0 -> walk is terminated by an error.
1216 static int fib6_walk_continue(struct fib6_walker_t *w)
1218 struct fib6_node *fn, *pn;
1225 if (w->prune && fn != w->root &&
1226 fn->fn_flags&RTN_RTINFO && w->state < FWS_C) {
1231 #ifdef CONFIG_IPV6_SUBTREES
1233 if (FIB6_SUBTREE(fn)) {
1234 w->node = FIB6_SUBTREE(fn);
1242 w->state = FWS_INIT;
1248 w->node = fn->right;
1249 w->state = FWS_INIT;
1255 if (w->leaf && fn->fn_flags&RTN_RTINFO) {
1256 int err = w->func(w);
1267 #ifdef CONFIG_IPV6_SUBTREES
1268 if (FIB6_SUBTREE(pn) == fn) {
1269 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1274 if (pn->left == fn) {
1278 if (pn->right == fn) {
1280 w->leaf = w->node->leaf;
1290 static int fib6_walk(struct fib6_walker_t *w)
1294 w->state = FWS_INIT;
1297 fib6_walker_link(w);
1298 res = fib6_walk_continue(w);
1300 fib6_walker_unlink(w);
1304 static int fib6_clean_node(struct fib6_walker_t *w)
1306 struct nl_info info = {
1307 .nl_net = &init_net,
1310 struct rt6_info *rt;
1311 struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w);
1313 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
1314 res = c->func(rt, c->arg);
1317 res = fib6_del(rt, &info);
1320 printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res);
1333 * Convenient frontend to tree walker.
1335 * func is called on each route.
1336 * It may return -1 -> delete this route.
1337 * 0 -> continue walking
1339 * prune==1 -> only immediate children of node (certainly,
1340 * ignoring pure split nodes) will be scanned.
1343 static void fib6_clean_tree(struct fib6_node *root,
1344 int (*func)(struct rt6_info *, void *arg),
1345 int prune, void *arg)
1347 struct fib6_cleaner_t c;
1350 c.w.func = fib6_clean_node;
1358 void fib6_clean_all(int (*func)(struct rt6_info *, void *arg),
1359 int prune, void *arg)
1361 struct fib6_table *table;
1362 struct hlist_node *node;
1366 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
1367 hlist_for_each_entry_rcu(table, node, &fib_table_hash[h],
1369 write_lock_bh(&table->tb6_lock);
1370 fib6_clean_tree(&table->tb6_root, func, prune, arg);
1371 write_unlock_bh(&table->tb6_lock);
1377 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1379 if (rt->rt6i_flags & RTF_CACHE) {
1380 RT6_TRACE("pruning clone %p\n", rt);
1387 static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt)
1389 fib6_clean_tree(fn, fib6_prune_clone, 1, rt);
1393 * Garbage collection
1396 static struct fib6_gc_args
1402 static int fib6_age(struct rt6_info *rt, void *arg)
1404 unsigned long now = jiffies;
1407 * check addrconf expiration here.
1408 * Routes are expired even if they are in use.
1410 * Also age clones. Note, that clones are aged out
1411 * only if they are not in use now.
1414 if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) {
1415 if (time_after(now, rt->rt6i_expires)) {
1416 RT6_TRACE("expiring %p\n", rt);
1420 } else if (rt->rt6i_flags & RTF_CACHE) {
1421 if (atomic_read(&rt->u.dst.__refcnt) == 0 &&
1422 time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) {
1423 RT6_TRACE("aging clone %p\n", rt);
1425 } else if ((rt->rt6i_flags & RTF_GATEWAY) &&
1426 (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) {
1427 RT6_TRACE("purging route %p via non-router but gateway\n",
1437 static DEFINE_SPINLOCK(fib6_gc_lock);
1439 void fib6_run_gc(unsigned long dummy)
1441 if (dummy != ~0UL) {
1442 spin_lock_bh(&fib6_gc_lock);
1443 gc_args.timeout = dummy ? (int)dummy :
1444 init_net.ipv6.sysctl.ip6_rt_gc_interval;
1447 if (!spin_trylock(&fib6_gc_lock)) {
1448 mod_timer(&ip6_fib_timer, jiffies + HZ);
1452 gc_args.timeout = init_net.ipv6.sysctl.ip6_rt_gc_interval;
1456 icmp6_dst_gc(&gc_args.more);
1457 fib6_clean_all(fib6_age, 0, NULL);
1460 mod_timer(&ip6_fib_timer, jiffies +
1461 init_net.ipv6.sysctl.ip6_rt_gc_interval);
1463 del_timer(&ip6_fib_timer);
1464 ip6_fib_timer.expires = 0;
1466 spin_unlock_bh(&fib6_gc_lock);
1469 int __init fib6_init(void)
1472 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1473 sizeof(struct fib6_node),
1474 0, SLAB_HWCACHE_ALIGN,
1476 if (!fib6_node_kmem)
1479 fib_table_hash = kzalloc(sizeof(*fib_table_hash)*FIB_TABLE_HASHSZ,
1481 if (!fib_table_hash)
1482 goto out_kmem_cache_create;
1484 fib6_main_tbl = kzalloc(sizeof(*fib6_main_tbl), GFP_KERNEL);
1486 goto out_fib_table_hash;
1488 fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
1489 fib6_main_tbl->tb6_root.leaf = &ip6_null_entry;
1490 fib6_main_tbl->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1492 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1493 fib6_local_tbl = kzalloc(sizeof(*fib6_local_tbl), GFP_KERNEL);
1494 if (!fib6_local_tbl)
1495 goto out_fib6_main_tbl;
1497 fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
1498 fib6_local_tbl->tb6_root.leaf = &ip6_null_entry;
1499 fib6_local_tbl->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1504 ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib);
1506 goto out_fib6_local_tbl;
1511 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1512 kfree(fib6_local_tbl);
1515 kfree(fib6_main_tbl);
1517 kfree(fib_table_hash);
1518 out_kmem_cache_create:
1519 kmem_cache_destroy(fib6_node_kmem);
1523 void fib6_gc_cleanup(void)
1525 del_timer(&ip6_fib_timer);
1526 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1527 kfree(fib6_local_tbl);
1529 kfree(fib6_main_tbl);
1530 kfree(fib_table_hash);
1531 kmem_cache_destroy(fib6_node_kmem);