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 if ((fn = kmem_cache_alloc(fib6_node_kmem, GFP_ATOMIC)) != NULL)
154 memset(fn, 0, sizeof(struct fib6_node));
159 static __inline__ void node_free(struct fib6_node * fn)
161 kmem_cache_free(fib6_node_kmem, fn);
164 static __inline__ void rt6_release(struct rt6_info *rt)
166 if (atomic_dec_and_test(&rt->rt6i_ref))
167 dst_free(&rt->u.dst);
170 static struct fib6_table fib6_main_tbl = {
171 .tb6_id = RT6_TABLE_MAIN,
173 .leaf = &ip6_null_entry,
174 .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO,
178 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
179 #define FIB_TABLE_HASHSZ 256
181 #define FIB_TABLE_HASHSZ 1
183 static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ];
185 static void fib6_link_table(struct fib6_table *tb)
190 * Initialize table lock at a single place to give lockdep a key,
191 * tables aren't visible prior to being linked to the list.
193 rwlock_init(&tb->tb6_lock);
195 h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1);
198 * No protection necessary, this is the only list mutatation
199 * operation, tables never disappear once they exist.
201 hlist_add_head_rcu(&tb->tb6_hlist, &fib_table_hash[h]);
204 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
205 static struct fib6_table fib6_local_tbl = {
206 .tb6_id = RT6_TABLE_LOCAL,
208 .leaf = &ip6_null_entry,
209 .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO,
213 static struct fib6_table *fib6_alloc_table(u32 id)
215 struct fib6_table *table;
217 table = kzalloc(sizeof(*table), GFP_ATOMIC);
220 table->tb6_root.leaf = &ip6_null_entry;
221 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
227 struct fib6_table *fib6_new_table(u32 id)
229 struct fib6_table *tb;
233 tb = fib6_get_table(id);
237 tb = fib6_alloc_table(id);
244 struct fib6_table *fib6_get_table(u32 id)
246 struct fib6_table *tb;
247 struct hlist_node *node;
252 h = id & (FIB_TABLE_HASHSZ - 1);
254 hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb6_hlist) {
255 if (tb->tb6_id == id) {
265 static void __init fib6_tables_init(void)
267 fib6_link_table(&fib6_main_tbl);
268 fib6_link_table(&fib6_local_tbl);
273 struct fib6_table *fib6_new_table(u32 id)
275 return fib6_get_table(id);
278 struct fib6_table *fib6_get_table(u32 id)
280 return &fib6_main_tbl;
283 struct dst_entry *fib6_rule_lookup(struct flowi *fl, int flags,
286 return (struct dst_entry *) lookup(&fib6_main_tbl, fl, flags);
289 static void __init fib6_tables_init(void)
291 fib6_link_table(&fib6_main_tbl);
296 static int fib6_dump_node(struct fib6_walker_t *w)
301 for (rt = w->leaf; rt; rt = rt->u.next) {
302 res = rt6_dump_route(rt, w->args);
304 /* Frame is full, suspend walking */
314 static void fib6_dump_end(struct netlink_callback *cb)
316 struct fib6_walker_t *w = (void*)cb->args[2];
322 cb->done = (void*)cb->args[3];
326 static int fib6_dump_done(struct netlink_callback *cb)
329 return cb->done ? cb->done(cb) : 0;
332 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
333 struct netlink_callback *cb)
335 struct fib6_walker_t *w;
338 w = (void *)cb->args[2];
339 w->root = &table->tb6_root;
341 if (cb->args[4] == 0) {
342 read_lock_bh(&table->tb6_lock);
344 read_unlock_bh(&table->tb6_lock);
348 read_lock_bh(&table->tb6_lock);
349 res = fib6_walk_continue(w);
350 read_unlock_bh(&table->tb6_lock);
353 fib6_walker_unlink(w);
356 fib6_walker_unlink(w);
363 int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
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;
376 w = (void *)cb->args[2];
380 * 1. hook callback destructor.
382 cb->args[3] = (long)cb->done;
383 cb->done = fib6_dump_done;
386 * 2. allocate and initialize walker.
388 w = kzalloc(sizeof(*w), GFP_ATOMIC);
391 w->func = fib6_dump_node;
392 cb->args[2] = (long)w;
399 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
401 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb6_hlist) {
404 res = fib6_dump_table(tb, skb, cb);
415 res = res < 0 ? res : skb->len;
424 * return the appropriate node for a routing tree "add" operation
425 * by either creating and inserting or by returning an existing
429 static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr,
430 int addrlen, int plen,
433 struct fib6_node *fn, *in, *ln;
434 struct fib6_node *pn = NULL;
438 __u32 sernum = fib6_new_sernum();
440 RT6_TRACE("fib6_add_1\n");
442 /* insert node in tree */
447 key = (struct rt6key *)((u8 *)fn->leaf + offset);
452 if (plen < fn->fn_bit ||
453 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
460 if (plen == fn->fn_bit) {
461 /* clean up an intermediate node */
462 if ((fn->fn_flags & RTN_RTINFO) == 0) {
463 rt6_release(fn->leaf);
467 fn->fn_sernum = sernum;
473 * We have more bits to go
476 /* Try to walk down on tree. */
477 fn->fn_sernum = sernum;
478 dir = addr_bit_set(addr, fn->fn_bit);
480 fn = dir ? fn->right: fn->left;
484 * We walked to the bottom of tree.
485 * Create new leaf node without children.
495 ln->fn_sernum = sernum;
507 * split since we don't have a common prefix anymore or
508 * we have a less significant route.
509 * we've to insert an intermediate node on the list
510 * this new node will point to the one we need to create
516 /* find 1st bit in difference between the 2 addrs.
518 See comment in __ipv6_addr_diff: bit may be an invalid value,
519 but if it is >= plen, the value is ignored in any case.
522 bit = __ipv6_addr_diff(addr, &key->addr, addrlen);
527 * (new leaf node)[ln] (old node)[fn]
533 if (in == NULL || ln == NULL) {
542 * new intermediate node.
544 * be off since that an address that chooses one of
545 * the branches would not match less specific routes
546 * in the other branch
553 atomic_inc(&in->leaf->rt6i_ref);
555 in->fn_sernum = sernum;
557 /* update parent pointer */
568 ln->fn_sernum = sernum;
570 if (addr_bit_set(addr, bit)) {
577 } else { /* plen <= bit */
580 * (new leaf node)[ln]
582 * (old node)[fn] NULL
594 ln->fn_sernum = sernum;
601 if (addr_bit_set(&key->addr, plen))
612 * Insert routing information in a node.
615 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
616 struct nl_info *info)
618 struct rt6_info *iter = NULL;
619 struct rt6_info **ins;
623 if (fn->fn_flags&RTN_TL_ROOT &&
624 fn->leaf == &ip6_null_entry &&
625 !(rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ){
631 for (iter = fn->leaf; iter; iter=iter->u.next) {
633 * Search for duplicates
636 if (iter->rt6i_metric == rt->rt6i_metric) {
638 * Same priority level
641 if (iter->rt6i_dev == rt->rt6i_dev &&
642 iter->rt6i_idev == rt->rt6i_idev &&
643 ipv6_addr_equal(&iter->rt6i_gateway,
644 &rt->rt6i_gateway)) {
645 if (!(iter->rt6i_flags&RTF_EXPIRES))
647 iter->rt6i_expires = rt->rt6i_expires;
648 if (!(rt->rt6i_flags&RTF_EXPIRES)) {
649 iter->rt6i_flags &= ~RTF_EXPIRES;
650 iter->rt6i_expires = 0;
656 if (iter->rt6i_metric > rt->rt6i_metric)
662 /* Reset round-robin state, if necessary */
663 if (ins == &fn->leaf)
674 atomic_inc(&rt->rt6i_ref);
675 inet6_rt_notify(RTM_NEWROUTE, rt, info);
676 rt6_stats.fib_rt_entries++;
678 if ((fn->fn_flags & RTN_RTINFO) == 0) {
679 rt6_stats.fib_route_nodes++;
680 fn->fn_flags |= RTN_RTINFO;
686 static __inline__ void fib6_start_gc(struct rt6_info *rt)
688 if (ip6_fib_timer.expires == 0 &&
689 (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE)))
690 mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval);
693 void fib6_force_start_gc(void)
695 if (ip6_fib_timer.expires == 0)
696 mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval);
700 * Add routing information to the routing tree.
701 * <destination addr>/<source addr>
702 * with source addr info in sub-trees
705 int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info)
707 struct fib6_node *fn, *pn = NULL;
710 fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr),
711 rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst));
718 #ifdef CONFIG_IPV6_SUBTREES
719 if (rt->rt6i_src.plen) {
720 struct fib6_node *sn;
722 if (fn->subtree == NULL) {
723 struct fib6_node *sfn;
735 /* Create subtree root node */
740 sfn->leaf = &ip6_null_entry;
741 atomic_inc(&ip6_null_entry.rt6i_ref);
742 sfn->fn_flags = RTN_ROOT;
743 sfn->fn_sernum = fib6_new_sernum();
745 /* Now add the first leaf node to new subtree */
747 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
748 sizeof(struct in6_addr), rt->rt6i_src.plen,
749 offsetof(struct rt6_info, rt6i_src));
752 /* If it is failed, discard just allocated
753 root, and then (in st_failure) stale node
760 /* Now link new subtree to main tree */
764 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
765 sizeof(struct in6_addr), rt->rt6i_src.plen,
766 offsetof(struct rt6_info, rt6i_src));
772 if (fn->leaf == NULL) {
774 atomic_inc(&rt->rt6i_ref);
780 err = fib6_add_rt2node(fn, rt, info);
784 if (!(rt->rt6i_flags&RTF_CACHE))
785 fib6_prune_clones(pn, rt);
790 #ifdef CONFIG_IPV6_SUBTREES
792 * If fib6_add_1 has cleared the old leaf pointer in the
793 * super-tree leaf node we have to find a new one for it.
795 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
796 pn->leaf = fib6_find_prefix(pn);
799 BUG_TRAP(pn->leaf != NULL);
800 pn->leaf = &ip6_null_entry;
803 atomic_inc(&pn->leaf->rt6i_ref);
806 dst_free(&rt->u.dst);
810 #ifdef CONFIG_IPV6_SUBTREES
811 /* Subtree creation failed, probably main tree node
812 is orphan. If it is, shoot it.
815 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
816 fib6_repair_tree(fn);
817 dst_free(&rt->u.dst);
823 * Routing tree lookup
828 int offset; /* key offset on rt6_info */
829 struct in6_addr *addr; /* search key */
832 static struct fib6_node * fib6_lookup_1(struct fib6_node *root,
833 struct lookup_args *args)
835 struct fib6_node *fn;
838 if (unlikely(args->offset == 0))
848 struct fib6_node *next;
850 dir = addr_bit_set(args->addr, fn->fn_bit);
852 next = dir ? fn->right : fn->left;
863 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
866 key = (struct rt6key *) ((u8 *) fn->leaf +
869 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
870 #ifdef CONFIG_IPV6_SUBTREES
872 fn = fib6_lookup_1(fn->subtree, args + 1);
874 if (!fn || fn->fn_flags & RTN_RTINFO)
879 if (fn->fn_flags & RTN_ROOT)
888 struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr,
889 struct in6_addr *saddr)
891 struct fib6_node *fn;
892 struct lookup_args args[] = {
894 .offset = offsetof(struct rt6_info, rt6i_dst),
897 #ifdef CONFIG_IPV6_SUBTREES
899 .offset = offsetof(struct rt6_info, rt6i_src),
904 .offset = 0, /* sentinel */
908 fn = fib6_lookup_1(root, daddr ? args : args + 1);
910 if (fn == NULL || fn->fn_flags & RTN_TL_ROOT)
917 * Get node with specified destination prefix (and source prefix,
918 * if subtrees are used)
922 static struct fib6_node * fib6_locate_1(struct fib6_node *root,
923 struct in6_addr *addr,
924 int plen, int offset)
926 struct fib6_node *fn;
928 for (fn = root; fn ; ) {
929 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
934 if (plen < fn->fn_bit ||
935 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
938 if (plen == fn->fn_bit)
942 * We have more bits to go
944 if (addr_bit_set(addr, fn->fn_bit))
952 struct fib6_node * fib6_locate(struct fib6_node *root,
953 struct in6_addr *daddr, int dst_len,
954 struct in6_addr *saddr, int src_len)
956 struct fib6_node *fn;
958 fn = fib6_locate_1(root, daddr, dst_len,
959 offsetof(struct rt6_info, rt6i_dst));
961 #ifdef CONFIG_IPV6_SUBTREES
963 BUG_TRAP(saddr!=NULL);
964 if (fn && fn->subtree)
965 fn = fib6_locate_1(fn->subtree, saddr, src_len,
966 offsetof(struct rt6_info, rt6i_src));
970 if (fn && fn->fn_flags&RTN_RTINFO)
982 static struct rt6_info * fib6_find_prefix(struct fib6_node *fn)
984 if (fn->fn_flags&RTN_ROOT)
985 return &ip6_null_entry;
989 return fn->left->leaf;
992 return fn->right->leaf;
994 fn = FIB6_SUBTREE(fn);
1000 * Called to trim the tree of intermediate nodes when possible. "fn"
1001 * is the node we want to try and remove.
1004 static struct fib6_node * fib6_repair_tree(struct fib6_node *fn)
1008 struct fib6_node *child, *pn;
1009 struct fib6_walker_t *w;
1013 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1016 BUG_TRAP(!(fn->fn_flags&RTN_RTINFO));
1017 BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT));
1018 BUG_TRAP(fn->leaf==NULL);
1022 if (fn->right) child = fn->right, children |= 1;
1023 if (fn->left) child = fn->left, children |= 2;
1025 if (children == 3 || FIB6_SUBTREE(fn)
1026 #ifdef CONFIG_IPV6_SUBTREES
1027 /* Subtree root (i.e. fn) may have one child */
1028 || (children && fn->fn_flags&RTN_ROOT)
1031 fn->leaf = fib6_find_prefix(fn);
1033 if (fn->leaf==NULL) {
1035 fn->leaf = &ip6_null_entry;
1038 atomic_inc(&fn->leaf->rt6i_ref);
1043 #ifdef CONFIG_IPV6_SUBTREES
1044 if (FIB6_SUBTREE(pn) == fn) {
1045 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1046 FIB6_SUBTREE(pn) = NULL;
1049 BUG_TRAP(!(fn->fn_flags&RTN_ROOT));
1051 if (pn->right == fn) pn->right = child;
1052 else if (pn->left == fn) pn->left = child;
1059 #ifdef CONFIG_IPV6_SUBTREES
1063 read_lock(&fib6_walker_lock);
1065 if (child == NULL) {
1066 if (w->root == fn) {
1067 w->root = w->node = NULL;
1068 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1069 } else if (w->node == fn) {
1070 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1075 if (w->root == fn) {
1077 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1079 if (w->node == fn) {
1082 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1083 w->state = w->state>=FWS_R ? FWS_U : FWS_INIT;
1085 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1086 w->state = w->state>=FWS_C ? FWS_U : FWS_INIT;
1091 read_unlock(&fib6_walker_lock);
1094 if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn))
1097 rt6_release(pn->leaf);
1103 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1104 struct nl_info *info)
1106 struct fib6_walker_t *w;
1107 struct rt6_info *rt = *rtp;
1109 RT6_TRACE("fib6_del_route\n");
1113 rt->rt6i_node = NULL;
1114 rt6_stats.fib_rt_entries--;
1115 rt6_stats.fib_discarded_routes++;
1117 /* Reset round-robin state, if necessary */
1118 if (fn->rr_ptr == rt)
1121 /* Adjust walkers */
1122 read_lock(&fib6_walker_lock);
1124 if (w->state == FWS_C && w->leaf == rt) {
1125 RT6_TRACE("walker %p adjusted by delroute\n", w);
1126 w->leaf = rt->u.next;
1127 if (w->leaf == NULL)
1131 read_unlock(&fib6_walker_lock);
1135 if (fn->leaf == NULL && fn->fn_flags&RTN_TL_ROOT)
1136 fn->leaf = &ip6_null_entry;
1138 /* If it was last route, expunge its radix tree node */
1139 if (fn->leaf == NULL) {
1140 fn->fn_flags &= ~RTN_RTINFO;
1141 rt6_stats.fib_route_nodes--;
1142 fn = fib6_repair_tree(fn);
1145 if (atomic_read(&rt->rt6i_ref) != 1) {
1146 /* This route is used as dummy address holder in some split
1147 * nodes. It is not leaked, but it still holds other resources,
1148 * which must be released in time. So, scan ascendant nodes
1149 * and replace dummy references to this route with references
1150 * to still alive ones.
1153 if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) {
1154 fn->leaf = fib6_find_prefix(fn);
1155 atomic_inc(&fn->leaf->rt6i_ref);
1160 /* No more references are possible at this point. */
1161 if (atomic_read(&rt->rt6i_ref) != 1) BUG();
1164 inet6_rt_notify(RTM_DELROUTE, rt, info);
1168 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1170 struct fib6_node *fn = rt->rt6i_node;
1171 struct rt6_info **rtp;
1174 if (rt->u.dst.obsolete>0) {
1179 if (fn == NULL || rt == &ip6_null_entry)
1182 BUG_TRAP(fn->fn_flags&RTN_RTINFO);
1184 if (!(rt->rt6i_flags&RTF_CACHE)) {
1185 struct fib6_node *pn = fn;
1186 #ifdef CONFIG_IPV6_SUBTREES
1187 /* clones of this route might be in another subtree */
1188 if (rt->rt6i_src.plen) {
1189 while (!(pn->fn_flags&RTN_ROOT))
1194 fib6_prune_clones(pn, rt);
1198 * Walk the leaf entries looking for ourself
1201 for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.next) {
1203 fib6_del_route(fn, rtp, info);
1211 * Tree traversal function.
1213 * Certainly, it is not interrupt safe.
1214 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1215 * It means, that we can modify tree during walking
1216 * and use this function for garbage collection, clone pruning,
1217 * cleaning tree when a device goes down etc. etc.
1219 * It guarantees that every node will be traversed,
1220 * and that it will be traversed only once.
1222 * Callback function w->func may return:
1223 * 0 -> continue walking.
1224 * positive value -> walking is suspended (used by tree dumps,
1225 * and probably by gc, if it will be split to several slices)
1226 * negative value -> terminate walking.
1228 * The function itself returns:
1229 * 0 -> walk is complete.
1230 * >0 -> walk is incomplete (i.e. suspended)
1231 * <0 -> walk is terminated by an error.
1234 static int fib6_walk_continue(struct fib6_walker_t *w)
1236 struct fib6_node *fn, *pn;
1243 if (w->prune && fn != w->root &&
1244 fn->fn_flags&RTN_RTINFO && w->state < FWS_C) {
1249 #ifdef CONFIG_IPV6_SUBTREES
1251 if (FIB6_SUBTREE(fn)) {
1252 w->node = FIB6_SUBTREE(fn);
1260 w->state = FWS_INIT;
1266 w->node = fn->right;
1267 w->state = FWS_INIT;
1273 if (w->leaf && fn->fn_flags&RTN_RTINFO) {
1274 int err = w->func(w);
1285 #ifdef CONFIG_IPV6_SUBTREES
1286 if (FIB6_SUBTREE(pn) == fn) {
1287 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1292 if (pn->left == fn) {
1296 if (pn->right == fn) {
1298 w->leaf = w->node->leaf;
1308 static int fib6_walk(struct fib6_walker_t *w)
1312 w->state = FWS_INIT;
1315 fib6_walker_link(w);
1316 res = fib6_walk_continue(w);
1318 fib6_walker_unlink(w);
1322 static int fib6_clean_node(struct fib6_walker_t *w)
1325 struct rt6_info *rt;
1326 struct fib6_cleaner_t *c = (struct fib6_cleaner_t*)w;
1328 for (rt = w->leaf; rt; rt = rt->u.next) {
1329 res = c->func(rt, c->arg);
1332 res = fib6_del(rt, NULL);
1335 printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res);
1348 * Convenient frontend to tree walker.
1350 * func is called on each route.
1351 * It may return -1 -> delete this route.
1352 * 0 -> continue walking
1354 * prune==1 -> only immediate children of node (certainly,
1355 * ignoring pure split nodes) will be scanned.
1358 static void fib6_clean_tree(struct fib6_node *root,
1359 int (*func)(struct rt6_info *, void *arg),
1360 int prune, void *arg)
1362 struct fib6_cleaner_t c;
1365 c.w.func = fib6_clean_node;
1373 void fib6_clean_all(int (*func)(struct rt6_info *, void *arg),
1374 int prune, void *arg)
1376 struct fib6_table *table;
1377 struct hlist_node *node;
1381 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
1382 hlist_for_each_entry_rcu(table, node, &fib_table_hash[h],
1384 write_lock_bh(&table->tb6_lock);
1385 fib6_clean_tree(&table->tb6_root, func, prune, arg);
1386 write_unlock_bh(&table->tb6_lock);
1392 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1394 if (rt->rt6i_flags & RTF_CACHE) {
1395 RT6_TRACE("pruning clone %p\n", rt);
1402 static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt)
1404 fib6_clean_tree(fn, fib6_prune_clone, 1, rt);
1408 * Garbage collection
1411 static struct fib6_gc_args
1417 static int fib6_age(struct rt6_info *rt, void *arg)
1419 unsigned long now = jiffies;
1422 * check addrconf expiration here.
1423 * Routes are expired even if they are in use.
1425 * Also age clones. Note, that clones are aged out
1426 * only if they are not in use now.
1429 if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) {
1430 if (time_after(now, rt->rt6i_expires)) {
1431 RT6_TRACE("expiring %p\n", rt);
1435 } else if (rt->rt6i_flags & RTF_CACHE) {
1436 if (atomic_read(&rt->u.dst.__refcnt) == 0 &&
1437 time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) {
1438 RT6_TRACE("aging clone %p\n", rt);
1440 } else if ((rt->rt6i_flags & RTF_GATEWAY) &&
1441 (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) {
1442 RT6_TRACE("purging route %p via non-router but gateway\n",
1452 static DEFINE_SPINLOCK(fib6_gc_lock);
1454 void fib6_run_gc(unsigned long dummy)
1456 if (dummy != ~0UL) {
1457 spin_lock_bh(&fib6_gc_lock);
1458 gc_args.timeout = dummy ? (int)dummy : ip6_rt_gc_interval;
1461 if (!spin_trylock(&fib6_gc_lock)) {
1462 mod_timer(&ip6_fib_timer, jiffies + HZ);
1466 gc_args.timeout = ip6_rt_gc_interval;
1470 ndisc_dst_gc(&gc_args.more);
1471 fib6_clean_all(fib6_age, 0, NULL);
1474 mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval);
1476 del_timer(&ip6_fib_timer);
1477 ip6_fib_timer.expires = 0;
1479 spin_unlock_bh(&fib6_gc_lock);
1482 void __init fib6_init(void)
1484 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1485 sizeof(struct fib6_node),
1486 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1492 void fib6_gc_cleanup(void)
1494 del_timer(&ip6_fib_timer);
1495 kmem_cache_destroy(fib6_node_kmem);