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 = {
170 .tb6_id = RT6_TABLE_MAIN,
172 .leaf = &ip6_null_entry,
173 .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO,
177 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
178 #define FIB_TABLE_HASHSZ 256
180 #define FIB_TABLE_HASHSZ 1
182 static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ];
184 static void fib6_link_table(struct fib6_table *tb)
189 * Initialize table lock at a single place to give lockdep a key,
190 * tables aren't visible prior to being linked to the list.
192 rwlock_init(&tb->tb6_lock);
194 h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1);
197 * No protection necessary, this is the only list mutatation
198 * operation, tables never disappear once they exist.
200 hlist_add_head_rcu(&tb->tb6_hlist, &fib_table_hash[h]);
203 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
204 static struct fib6_table fib6_local_tbl = {
205 .tb6_id = RT6_TABLE_LOCAL,
207 .leaf = &ip6_null_entry,
208 .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO,
212 static struct fib6_table *fib6_alloc_table(u32 id)
214 struct fib6_table *table;
216 table = kzalloc(sizeof(*table), GFP_ATOMIC);
219 table->tb6_root.leaf = &ip6_null_entry;
220 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
226 struct fib6_table *fib6_new_table(u32 id)
228 struct fib6_table *tb;
232 tb = fib6_get_table(id);
236 tb = fib6_alloc_table(id);
243 struct fib6_table *fib6_get_table(u32 id)
245 struct fib6_table *tb;
246 struct hlist_node *node;
251 h = id & (FIB_TABLE_HASHSZ - 1);
253 hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb6_hlist) {
254 if (tb->tb6_id == id) {
264 static void __init fib6_tables_init(void)
266 fib6_link_table(&fib6_main_tbl);
267 fib6_link_table(&fib6_local_tbl);
272 struct fib6_table *fib6_new_table(u32 id)
274 return fib6_get_table(id);
277 struct fib6_table *fib6_get_table(u32 id)
279 return &fib6_main_tbl;
282 struct dst_entry *fib6_rule_lookup(struct flowi *fl, int flags,
285 return (struct dst_entry *) lookup(&fib6_main_tbl, fl, flags);
288 static void __init fib6_tables_init(void)
290 fib6_link_table(&fib6_main_tbl);
295 static int fib6_dump_node(struct fib6_walker_t *w)
300 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
301 res = rt6_dump_route(rt, w->args);
303 /* Frame is full, suspend walking */
313 static void fib6_dump_end(struct netlink_callback *cb)
315 struct fib6_walker_t *w = (void*)cb->args[2];
321 cb->done = (void*)cb->args[3];
325 static int fib6_dump_done(struct netlink_callback *cb)
328 return cb->done ? cb->done(cb) : 0;
331 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
332 struct netlink_callback *cb)
334 struct fib6_walker_t *w;
337 w = (void *)cb->args[2];
338 w->root = &table->tb6_root;
340 if (cb->args[4] == 0) {
341 read_lock_bh(&table->tb6_lock);
343 read_unlock_bh(&table->tb6_lock);
347 read_lock_bh(&table->tb6_lock);
348 res = fib6_walk_continue(w);
349 read_unlock_bh(&table->tb6_lock);
352 fib6_walker_unlink(w);
355 fib6_walker_unlink(w);
362 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
364 struct net *net = skb->sk->sk_net;
366 unsigned int e = 0, s_e;
367 struct rt6_rtnl_dump_arg arg;
368 struct fib6_walker_t *w;
369 struct fib6_table *tb;
370 struct hlist_node *node;
373 if (net != &init_net)
379 w = (void *)cb->args[2];
383 * 1. hook callback destructor.
385 cb->args[3] = (long)cb->done;
386 cb->done = fib6_dump_done;
389 * 2. allocate and initialize walker.
391 w = kzalloc(sizeof(*w), GFP_ATOMIC);
394 w->func = fib6_dump_node;
395 cb->args[2] = (long)w;
402 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
404 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb6_hlist) {
407 res = fib6_dump_table(tb, skb, cb);
418 res = res < 0 ? res : skb->len;
427 * return the appropriate node for a routing tree "add" operation
428 * by either creating and inserting or by returning an existing
432 static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr,
433 int addrlen, int plen,
436 struct fib6_node *fn, *in, *ln;
437 struct fib6_node *pn = NULL;
441 __u32 sernum = fib6_new_sernum();
443 RT6_TRACE("fib6_add_1\n");
445 /* insert node in tree */
450 key = (struct rt6key *)((u8 *)fn->leaf + offset);
455 if (plen < fn->fn_bit ||
456 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
463 if (plen == fn->fn_bit) {
464 /* clean up an intermediate node */
465 if ((fn->fn_flags & RTN_RTINFO) == 0) {
466 rt6_release(fn->leaf);
470 fn->fn_sernum = sernum;
476 * We have more bits to go
479 /* Try to walk down on tree. */
480 fn->fn_sernum = sernum;
481 dir = addr_bit_set(addr, fn->fn_bit);
483 fn = dir ? fn->right: fn->left;
487 * We walked to the bottom of tree.
488 * Create new leaf node without children.
498 ln->fn_sernum = sernum;
510 * split since we don't have a common prefix anymore or
511 * we have a less significant route.
512 * we've to insert an intermediate node on the list
513 * this new node will point to the one we need to create
519 /* find 1st bit in difference between the 2 addrs.
521 See comment in __ipv6_addr_diff: bit may be an invalid value,
522 but if it is >= plen, the value is ignored in any case.
525 bit = __ipv6_addr_diff(addr, &key->addr, addrlen);
530 * (new leaf node)[ln] (old node)[fn]
536 if (in == NULL || ln == NULL) {
545 * new intermediate node.
547 * be off since that an address that chooses one of
548 * the branches would not match less specific routes
549 * in the other branch
556 atomic_inc(&in->leaf->rt6i_ref);
558 in->fn_sernum = sernum;
560 /* update parent pointer */
571 ln->fn_sernum = sernum;
573 if (addr_bit_set(addr, bit)) {
580 } else { /* plen <= bit */
583 * (new leaf node)[ln]
585 * (old node)[fn] NULL
597 ln->fn_sernum = sernum;
604 if (addr_bit_set(&key->addr, plen))
615 * Insert routing information in a node.
618 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
619 struct nl_info *info)
621 struct rt6_info *iter = NULL;
622 struct rt6_info **ins;
626 for (iter = fn->leaf; iter; iter=iter->u.dst.rt6_next) {
628 * Search for duplicates
631 if (iter->rt6i_metric == rt->rt6i_metric) {
633 * Same priority level
636 if (iter->rt6i_dev == rt->rt6i_dev &&
637 iter->rt6i_idev == rt->rt6i_idev &&
638 ipv6_addr_equal(&iter->rt6i_gateway,
639 &rt->rt6i_gateway)) {
640 if (!(iter->rt6i_flags&RTF_EXPIRES))
642 iter->rt6i_expires = rt->rt6i_expires;
643 if (!(rt->rt6i_flags&RTF_EXPIRES)) {
644 iter->rt6i_flags &= ~RTF_EXPIRES;
645 iter->rt6i_expires = 0;
651 if (iter->rt6i_metric > rt->rt6i_metric)
654 ins = &iter->u.dst.rt6_next;
657 /* Reset round-robin state, if necessary */
658 if (ins == &fn->leaf)
665 rt->u.dst.rt6_next = iter;
668 atomic_inc(&rt->rt6i_ref);
669 inet6_rt_notify(RTM_NEWROUTE, rt, info);
670 rt6_stats.fib_rt_entries++;
672 if ((fn->fn_flags & RTN_RTINFO) == 0) {
673 rt6_stats.fib_route_nodes++;
674 fn->fn_flags |= RTN_RTINFO;
680 static __inline__ void fib6_start_gc(struct rt6_info *rt)
682 if (ip6_fib_timer.expires == 0 &&
683 (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE)))
684 mod_timer(&ip6_fib_timer, jiffies +
685 init_net.ipv6.sysctl.ip6_rt_gc_interval);
688 void fib6_force_start_gc(void)
690 if (ip6_fib_timer.expires == 0)
691 mod_timer(&ip6_fib_timer, jiffies +
692 init_net.ipv6.sysctl.ip6_rt_gc_interval);
696 * Add routing information to the routing tree.
697 * <destination addr>/<source addr>
698 * with source addr info in sub-trees
701 int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info)
703 struct fib6_node *fn, *pn = NULL;
706 fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr),
707 rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst));
714 #ifdef CONFIG_IPV6_SUBTREES
715 if (rt->rt6i_src.plen) {
716 struct fib6_node *sn;
718 if (fn->subtree == NULL) {
719 struct fib6_node *sfn;
731 /* Create subtree root node */
736 sfn->leaf = &ip6_null_entry;
737 atomic_inc(&ip6_null_entry.rt6i_ref);
738 sfn->fn_flags = RTN_ROOT;
739 sfn->fn_sernum = fib6_new_sernum();
741 /* Now add the first leaf node to new subtree */
743 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
744 sizeof(struct in6_addr), rt->rt6i_src.plen,
745 offsetof(struct rt6_info, rt6i_src));
748 /* If it is failed, discard just allocated
749 root, and then (in st_failure) stale node
756 /* Now link new subtree to main tree */
760 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
761 sizeof(struct in6_addr), rt->rt6i_src.plen,
762 offsetof(struct rt6_info, rt6i_src));
768 if (fn->leaf == NULL) {
770 atomic_inc(&rt->rt6i_ref);
776 err = fib6_add_rt2node(fn, rt, info);
780 if (!(rt->rt6i_flags&RTF_CACHE))
781 fib6_prune_clones(pn, rt);
786 #ifdef CONFIG_IPV6_SUBTREES
788 * If fib6_add_1 has cleared the old leaf pointer in the
789 * super-tree leaf node we have to find a new one for it.
791 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
792 pn->leaf = fib6_find_prefix(pn);
795 BUG_TRAP(pn->leaf != NULL);
796 pn->leaf = &ip6_null_entry;
799 atomic_inc(&pn->leaf->rt6i_ref);
802 dst_free(&rt->u.dst);
806 #ifdef CONFIG_IPV6_SUBTREES
807 /* Subtree creation failed, probably main tree node
808 is orphan. If it is, shoot it.
811 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
812 fib6_repair_tree(fn);
813 dst_free(&rt->u.dst);
819 * Routing tree lookup
824 int offset; /* key offset on rt6_info */
825 struct in6_addr *addr; /* search key */
828 static struct fib6_node * fib6_lookup_1(struct fib6_node *root,
829 struct lookup_args *args)
831 struct fib6_node *fn;
834 if (unlikely(args->offset == 0))
844 struct fib6_node *next;
846 dir = addr_bit_set(args->addr, fn->fn_bit);
848 next = dir ? fn->right : fn->left;
859 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
862 key = (struct rt6key *) ((u8 *) fn->leaf +
865 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
866 #ifdef CONFIG_IPV6_SUBTREES
868 fn = fib6_lookup_1(fn->subtree, args + 1);
870 if (!fn || fn->fn_flags & RTN_RTINFO)
875 if (fn->fn_flags & RTN_ROOT)
884 struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr,
885 struct in6_addr *saddr)
887 struct fib6_node *fn;
888 struct lookup_args args[] = {
890 .offset = offsetof(struct rt6_info, rt6i_dst),
893 #ifdef CONFIG_IPV6_SUBTREES
895 .offset = offsetof(struct rt6_info, rt6i_src),
900 .offset = 0, /* sentinel */
904 fn = fib6_lookup_1(root, daddr ? args : args + 1);
906 if (fn == NULL || fn->fn_flags & RTN_TL_ROOT)
913 * Get node with specified destination prefix (and source prefix,
914 * if subtrees are used)
918 static struct fib6_node * fib6_locate_1(struct fib6_node *root,
919 struct in6_addr *addr,
920 int plen, int offset)
922 struct fib6_node *fn;
924 for (fn = root; fn ; ) {
925 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
930 if (plen < fn->fn_bit ||
931 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
934 if (plen == fn->fn_bit)
938 * We have more bits to go
940 if (addr_bit_set(addr, fn->fn_bit))
948 struct fib6_node * fib6_locate(struct fib6_node *root,
949 struct in6_addr *daddr, int dst_len,
950 struct in6_addr *saddr, int src_len)
952 struct fib6_node *fn;
954 fn = fib6_locate_1(root, daddr, dst_len,
955 offsetof(struct rt6_info, rt6i_dst));
957 #ifdef CONFIG_IPV6_SUBTREES
959 BUG_TRAP(saddr!=NULL);
960 if (fn && fn->subtree)
961 fn = fib6_locate_1(fn->subtree, saddr, src_len,
962 offsetof(struct rt6_info, rt6i_src));
966 if (fn && fn->fn_flags&RTN_RTINFO)
978 static struct rt6_info * fib6_find_prefix(struct fib6_node *fn)
980 if (fn->fn_flags&RTN_ROOT)
981 return &ip6_null_entry;
985 return fn->left->leaf;
988 return fn->right->leaf;
990 fn = FIB6_SUBTREE(fn);
996 * Called to trim the tree of intermediate nodes when possible. "fn"
997 * is the node we want to try and remove.
1000 static struct fib6_node * fib6_repair_tree(struct fib6_node *fn)
1004 struct fib6_node *child, *pn;
1005 struct fib6_walker_t *w;
1009 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1012 BUG_TRAP(!(fn->fn_flags&RTN_RTINFO));
1013 BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT));
1014 BUG_TRAP(fn->leaf==NULL);
1018 if (fn->right) child = fn->right, children |= 1;
1019 if (fn->left) child = fn->left, children |= 2;
1021 if (children == 3 || FIB6_SUBTREE(fn)
1022 #ifdef CONFIG_IPV6_SUBTREES
1023 /* Subtree root (i.e. fn) may have one child */
1024 || (children && fn->fn_flags&RTN_ROOT)
1027 fn->leaf = fib6_find_prefix(fn);
1029 if (fn->leaf==NULL) {
1031 fn->leaf = &ip6_null_entry;
1034 atomic_inc(&fn->leaf->rt6i_ref);
1039 #ifdef CONFIG_IPV6_SUBTREES
1040 if (FIB6_SUBTREE(pn) == fn) {
1041 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1042 FIB6_SUBTREE(pn) = NULL;
1045 BUG_TRAP(!(fn->fn_flags&RTN_ROOT));
1047 if (pn->right == fn) pn->right = child;
1048 else if (pn->left == fn) pn->left = child;
1055 #ifdef CONFIG_IPV6_SUBTREES
1059 read_lock(&fib6_walker_lock);
1061 if (child == NULL) {
1062 if (w->root == fn) {
1063 w->root = w->node = NULL;
1064 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1065 } else if (w->node == fn) {
1066 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1071 if (w->root == fn) {
1073 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1075 if (w->node == fn) {
1078 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1079 w->state = w->state>=FWS_R ? FWS_U : FWS_INIT;
1081 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1082 w->state = w->state>=FWS_C ? FWS_U : FWS_INIT;
1087 read_unlock(&fib6_walker_lock);
1090 if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn))
1093 rt6_release(pn->leaf);
1099 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1100 struct nl_info *info)
1102 struct fib6_walker_t *w;
1103 struct rt6_info *rt = *rtp;
1105 RT6_TRACE("fib6_del_route\n");
1108 *rtp = rt->u.dst.rt6_next;
1109 rt->rt6i_node = NULL;
1110 rt6_stats.fib_rt_entries--;
1111 rt6_stats.fib_discarded_routes++;
1113 /* Reset round-robin state, if necessary */
1114 if (fn->rr_ptr == rt)
1117 /* Adjust walkers */
1118 read_lock(&fib6_walker_lock);
1120 if (w->state == FWS_C && w->leaf == rt) {
1121 RT6_TRACE("walker %p adjusted by delroute\n", w);
1122 w->leaf = rt->u.dst.rt6_next;
1123 if (w->leaf == NULL)
1127 read_unlock(&fib6_walker_lock);
1129 rt->u.dst.rt6_next = NULL;
1131 /* If it was last route, expunge its radix tree node */
1132 if (fn->leaf == NULL) {
1133 fn->fn_flags &= ~RTN_RTINFO;
1134 rt6_stats.fib_route_nodes--;
1135 fn = fib6_repair_tree(fn);
1138 if (atomic_read(&rt->rt6i_ref) != 1) {
1139 /* This route is used as dummy address holder in some split
1140 * nodes. It is not leaked, but it still holds other resources,
1141 * which must be released in time. So, scan ascendant nodes
1142 * and replace dummy references to this route with references
1143 * to still alive ones.
1146 if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) {
1147 fn->leaf = fib6_find_prefix(fn);
1148 atomic_inc(&fn->leaf->rt6i_ref);
1153 /* No more references are possible at this point. */
1154 if (atomic_read(&rt->rt6i_ref) != 1) BUG();
1157 inet6_rt_notify(RTM_DELROUTE, rt, info);
1161 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1163 struct fib6_node *fn = rt->rt6i_node;
1164 struct rt6_info **rtp;
1167 if (rt->u.dst.obsolete>0) {
1172 if (fn == NULL || rt == &ip6_null_entry)
1175 BUG_TRAP(fn->fn_flags&RTN_RTINFO);
1177 if (!(rt->rt6i_flags&RTF_CACHE)) {
1178 struct fib6_node *pn = fn;
1179 #ifdef CONFIG_IPV6_SUBTREES
1180 /* clones of this route might be in another subtree */
1181 if (rt->rt6i_src.plen) {
1182 while (!(pn->fn_flags&RTN_ROOT))
1187 fib6_prune_clones(pn, rt);
1191 * Walk the leaf entries looking for ourself
1194 for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.dst.rt6_next) {
1196 fib6_del_route(fn, rtp, info);
1204 * Tree traversal function.
1206 * Certainly, it is not interrupt safe.
1207 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1208 * It means, that we can modify tree during walking
1209 * and use this function for garbage collection, clone pruning,
1210 * cleaning tree when a device goes down etc. etc.
1212 * It guarantees that every node will be traversed,
1213 * and that it will be traversed only once.
1215 * Callback function w->func may return:
1216 * 0 -> continue walking.
1217 * positive value -> walking is suspended (used by tree dumps,
1218 * and probably by gc, if it will be split to several slices)
1219 * negative value -> terminate walking.
1221 * The function itself returns:
1222 * 0 -> walk is complete.
1223 * >0 -> walk is incomplete (i.e. suspended)
1224 * <0 -> walk is terminated by an error.
1227 static int fib6_walk_continue(struct fib6_walker_t *w)
1229 struct fib6_node *fn, *pn;
1236 if (w->prune && fn != w->root &&
1237 fn->fn_flags&RTN_RTINFO && w->state < FWS_C) {
1242 #ifdef CONFIG_IPV6_SUBTREES
1244 if (FIB6_SUBTREE(fn)) {
1245 w->node = FIB6_SUBTREE(fn);
1253 w->state = FWS_INIT;
1259 w->node = fn->right;
1260 w->state = FWS_INIT;
1266 if (w->leaf && fn->fn_flags&RTN_RTINFO) {
1267 int err = w->func(w);
1278 #ifdef CONFIG_IPV6_SUBTREES
1279 if (FIB6_SUBTREE(pn) == fn) {
1280 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1285 if (pn->left == fn) {
1289 if (pn->right == fn) {
1291 w->leaf = w->node->leaf;
1301 static int fib6_walk(struct fib6_walker_t *w)
1305 w->state = FWS_INIT;
1308 fib6_walker_link(w);
1309 res = fib6_walk_continue(w);
1311 fib6_walker_unlink(w);
1315 static int fib6_clean_node(struct fib6_walker_t *w)
1317 struct nl_info info = {
1318 .nl_net = &init_net,
1321 struct rt6_info *rt;
1322 struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w);
1324 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
1325 res = c->func(rt, c->arg);
1328 res = fib6_del(rt, &info);
1331 printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res);
1344 * Convenient frontend to tree walker.
1346 * func is called on each route.
1347 * It may return -1 -> delete this route.
1348 * 0 -> continue walking
1350 * prune==1 -> only immediate children of node (certainly,
1351 * ignoring pure split nodes) will be scanned.
1354 static void fib6_clean_tree(struct fib6_node *root,
1355 int (*func)(struct rt6_info *, void *arg),
1356 int prune, void *arg)
1358 struct fib6_cleaner_t c;
1361 c.w.func = fib6_clean_node;
1369 void fib6_clean_all(int (*func)(struct rt6_info *, void *arg),
1370 int prune, void *arg)
1372 struct fib6_table *table;
1373 struct hlist_node *node;
1377 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
1378 hlist_for_each_entry_rcu(table, node, &fib_table_hash[h],
1380 write_lock_bh(&table->tb6_lock);
1381 fib6_clean_tree(&table->tb6_root, func, prune, arg);
1382 write_unlock_bh(&table->tb6_lock);
1388 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1390 if (rt->rt6i_flags & RTF_CACHE) {
1391 RT6_TRACE("pruning clone %p\n", rt);
1398 static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt)
1400 fib6_clean_tree(fn, fib6_prune_clone, 1, rt);
1404 * Garbage collection
1407 static struct fib6_gc_args
1413 static int fib6_age(struct rt6_info *rt, void *arg)
1415 unsigned long now = jiffies;
1418 * check addrconf expiration here.
1419 * Routes are expired even if they are in use.
1421 * Also age clones. Note, that clones are aged out
1422 * only if they are not in use now.
1425 if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) {
1426 if (time_after(now, rt->rt6i_expires)) {
1427 RT6_TRACE("expiring %p\n", rt);
1431 } else if (rt->rt6i_flags & RTF_CACHE) {
1432 if (atomic_read(&rt->u.dst.__refcnt) == 0 &&
1433 time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) {
1434 RT6_TRACE("aging clone %p\n", rt);
1436 } else if ((rt->rt6i_flags & RTF_GATEWAY) &&
1437 (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) {
1438 RT6_TRACE("purging route %p via non-router but gateway\n",
1448 static DEFINE_SPINLOCK(fib6_gc_lock);
1450 void fib6_run_gc(unsigned long dummy)
1452 if (dummy != ~0UL) {
1453 spin_lock_bh(&fib6_gc_lock);
1454 gc_args.timeout = dummy ? (int)dummy :
1455 init_net.ipv6.sysctl.ip6_rt_gc_interval;
1458 if (!spin_trylock(&fib6_gc_lock)) {
1459 mod_timer(&ip6_fib_timer, jiffies + HZ);
1463 gc_args.timeout = init_net.ipv6.sysctl.ip6_rt_gc_interval;
1467 ndisc_dst_gc(&gc_args.more);
1468 fib6_clean_all(fib6_age, 0, NULL);
1471 mod_timer(&ip6_fib_timer, jiffies +
1472 init_net.ipv6.sysctl.ip6_rt_gc_interval);
1474 del_timer(&ip6_fib_timer);
1475 ip6_fib_timer.expires = 0;
1477 spin_unlock_bh(&fib6_gc_lock);
1480 int __init fib6_init(void)
1483 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1484 sizeof(struct fib6_node),
1485 0, SLAB_HWCACHE_ALIGN,
1487 if (!fib6_node_kmem)
1492 ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib);
1494 goto out_kmem_cache_create;
1498 out_kmem_cache_create:
1499 kmem_cache_destroy(fib6_node_kmem);
1503 void fib6_gc_cleanup(void)
1505 del_timer(&ip6_fib_timer);
1506 kmem_cache_destroy(fib6_node_kmem);