2 * Linux INET6 implementation
3 * Forwarding Information Database
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * Yuji SEKIYA @USAGI: Support default route on router node;
17 * remove ip6_null_entry from the top of
19 * Ville Nuorvala: Fixed routing subtrees.
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/net.h>
24 #include <linux/route.h>
25 #include <linux/netdevice.h>
26 #include <linux/in6.h>
27 #include <linux/init.h>
28 #include <linux/list.h>
29 #include <linux/slab.h>
32 #include <linux/proc_fs.h>
36 #include <net/ndisc.h>
37 #include <net/addrconf.h>
39 #include <net/ip6_fib.h>
40 #include <net/ip6_route.h>
45 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
47 #define RT6_TRACE(x...) do { ; } while (0)
50 static struct kmem_cache * fib6_node_kmem __read_mostly;
54 #ifdef CONFIG_IPV6_SUBTREES
65 struct fib6_walker_t w;
67 int (*func)(struct rt6_info *, void *arg);
71 static DEFINE_RWLOCK(fib6_walker_lock);
73 #ifdef CONFIG_IPV6_SUBTREES
74 #define FWS_INIT FWS_S
76 #define FWS_INIT FWS_L
79 static void fib6_prune_clones(struct net *net, struct fib6_node *fn,
81 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn);
82 static struct fib6_node *fib6_repair_tree(struct net *net, struct fib6_node *fn);
83 static int fib6_walk(struct fib6_walker_t *w);
84 static int fib6_walk_continue(struct fib6_walker_t *w);
87 * A routing update causes an increase of the serial number on the
88 * affected subtree. This allows for cached routes to be asynchronously
89 * tested when modifications are made to the destination cache as a
90 * result of redirects, path MTU changes, etc.
93 static __u32 rt_sernum;
95 static void fib6_gc_timer_cb(unsigned long arg);
97 static LIST_HEAD(fib6_walkers);
98 #define FOR_WALKERS(w) list_for_each_entry(w, &fib6_walkers, lh)
100 static inline void fib6_walker_link(struct fib6_walker_t *w)
102 write_lock_bh(&fib6_walker_lock);
103 list_add(&w->lh, &fib6_walkers);
104 write_unlock_bh(&fib6_walker_lock);
107 static inline void fib6_walker_unlink(struct fib6_walker_t *w)
109 write_lock_bh(&fib6_walker_lock);
111 write_unlock_bh(&fib6_walker_lock);
113 static __inline__ u32 fib6_new_sernum(void)
122 * Auxiliary address test functions for the radix tree.
124 * These assume a 32bit processor (although it will work on
131 #if defined(__LITTLE_ENDIAN)
132 # define BITOP_BE32_SWIZZLE (0x1F & ~7)
134 # define BITOP_BE32_SWIZZLE 0
137 static __inline__ __be32 addr_bit_set(void *token, int fn_bit)
139 __be32 *addr = token;
142 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
143 * is optimized version of
144 * htonl(1 << ((~fn_bit)&0x1F))
145 * See include/asm-generic/bitops/le.h.
147 return (1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) & addr[fn_bit >> 5];
150 static __inline__ struct fib6_node * node_alloc(void)
152 struct fib6_node *fn;
154 fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
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 void fib6_link_table(struct net *net, struct fib6_table *tb)
175 * Initialize table lock at a single place to give lockdep a key,
176 * tables aren't visible prior to being linked to the list.
178 rwlock_init(&tb->tb6_lock);
180 h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1);
183 * No protection necessary, this is the only list mutatation
184 * operation, tables never disappear once they exist.
186 hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]);
189 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
191 static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
193 struct fib6_table *table;
195 table = kzalloc(sizeof(*table), GFP_ATOMIC);
198 table->tb6_root.leaf = net->ipv6.ip6_null_entry;
199 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
205 struct fib6_table *fib6_new_table(struct net *net, u32 id)
207 struct fib6_table *tb;
211 tb = fib6_get_table(net, id);
215 tb = fib6_alloc_table(net, id);
217 fib6_link_table(net, tb);
222 struct fib6_table *fib6_get_table(struct net *net, u32 id)
224 struct fib6_table *tb;
225 struct hlist_head *head;
226 struct hlist_node *node;
231 h = id & (FIB6_TABLE_HASHSZ - 1);
233 head = &net->ipv6.fib_table_hash[h];
234 hlist_for_each_entry_rcu(tb, node, head, tb6_hlist) {
235 if (tb->tb6_id == id) {
245 static void __net_init fib6_tables_init(struct net *net)
247 fib6_link_table(net, net->ipv6.fib6_main_tbl);
248 fib6_link_table(net, net->ipv6.fib6_local_tbl);
252 struct fib6_table *fib6_new_table(struct net *net, u32 id)
254 return fib6_get_table(net, id);
257 struct fib6_table *fib6_get_table(struct net *net, u32 id)
259 return net->ipv6.fib6_main_tbl;
262 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi *fl,
263 int flags, pol_lookup_t lookup)
265 return (struct dst_entry *) lookup(net, net->ipv6.fib6_main_tbl, fl, flags);
268 static void __net_init fib6_tables_init(struct net *net)
270 fib6_link_table(net, net->ipv6.fib6_main_tbl);
275 static int fib6_dump_node(struct fib6_walker_t *w)
280 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
281 res = rt6_dump_route(rt, w->args);
283 /* Frame is full, suspend walking */
293 static void fib6_dump_end(struct netlink_callback *cb)
295 struct fib6_walker_t *w = (void*)cb->args[2];
300 fib6_walker_unlink(w);
305 cb->done = (void*)cb->args[3];
309 static int fib6_dump_done(struct netlink_callback *cb)
312 return cb->done ? cb->done(cb) : 0;
315 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
316 struct netlink_callback *cb)
318 struct fib6_walker_t *w;
321 w = (void *)cb->args[2];
322 w->root = &table->tb6_root;
324 if (cb->args[4] == 0) {
328 read_lock_bh(&table->tb6_lock);
330 read_unlock_bh(&table->tb6_lock);
333 cb->args[5] = w->root->fn_sernum;
336 if (cb->args[5] != w->root->fn_sernum) {
337 /* Begin at the root if the tree changed */
338 cb->args[5] = w->root->fn_sernum;
345 read_lock_bh(&table->tb6_lock);
346 res = fib6_walk_continue(w);
347 read_unlock_bh(&table->tb6_lock);
349 fib6_walker_unlink(w);
357 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
359 struct net *net = sock_net(skb->sk);
361 unsigned int e = 0, s_e;
362 struct rt6_rtnl_dump_arg arg;
363 struct fib6_walker_t *w;
364 struct fib6_table *tb;
365 struct hlist_node *node;
366 struct hlist_head *head;
372 w = (void *)cb->args[2];
376 * 1. hook callback destructor.
378 cb->args[3] = (long)cb->done;
379 cb->done = fib6_dump_done;
382 * 2. allocate and initialize walker.
384 w = kzalloc(sizeof(*w), GFP_ATOMIC);
387 w->func = fib6_dump_node;
388 cb->args[2] = (long)w;
396 for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) {
398 head = &net->ipv6.fib_table_hash[h];
399 hlist_for_each_entry(tb, node, head, tb6_hlist) {
402 res = fib6_dump_table(tb, skb, cb);
413 res = res < 0 ? res : skb->len;
422 * return the appropriate node for a routing tree "add" operation
423 * by either creating and inserting or by returning an existing
427 static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr,
428 int addrlen, int plen,
431 struct fib6_node *fn, *in, *ln;
432 struct fib6_node *pn = NULL;
436 __u32 sernum = fib6_new_sernum();
438 RT6_TRACE("fib6_add_1\n");
440 /* insert node in tree */
445 key = (struct rt6key *)((u8 *)fn->leaf + offset);
450 if (plen < fn->fn_bit ||
451 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
458 if (plen == fn->fn_bit) {
459 /* clean up an intermediate node */
460 if ((fn->fn_flags & RTN_RTINFO) == 0) {
461 rt6_release(fn->leaf);
465 fn->fn_sernum = sernum;
471 * We have more bits to go
474 /* Try to walk down on tree. */
475 fn->fn_sernum = sernum;
476 dir = addr_bit_set(addr, fn->fn_bit);
478 fn = dir ? fn->right: fn->left;
482 * We walked to the bottom of tree.
483 * Create new leaf node without children.
493 ln->fn_sernum = sernum;
505 * split since we don't have a common prefix anymore or
506 * we have a less significant route.
507 * we've to insert an intermediate node on the list
508 * this new node will point to the one we need to create
514 /* find 1st bit in difference between the 2 addrs.
516 See comment in __ipv6_addr_diff: bit may be an invalid value,
517 but if it is >= plen, the value is ignored in any case.
520 bit = __ipv6_addr_diff(addr, &key->addr, addrlen);
525 * (new leaf node)[ln] (old node)[fn]
531 if (in == NULL || ln == NULL) {
540 * new intermediate node.
542 * be off since that an address that chooses one of
543 * the branches would not match less specific routes
544 * in the other branch
551 atomic_inc(&in->leaf->rt6i_ref);
553 in->fn_sernum = sernum;
555 /* update parent pointer */
566 ln->fn_sernum = sernum;
568 if (addr_bit_set(addr, bit)) {
575 } else { /* plen <= bit */
578 * (new leaf node)[ln]
580 * (old node)[fn] NULL
592 ln->fn_sernum = sernum;
599 if (addr_bit_set(&key->addr, plen))
610 * Insert routing information in a node.
613 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
614 struct nl_info *info)
616 struct rt6_info *iter = NULL;
617 struct rt6_info **ins;
621 for (iter = fn->leaf; iter; iter=iter->u.dst.rt6_next) {
623 * Search for duplicates
626 if (iter->rt6i_metric == rt->rt6i_metric) {
628 * Same priority level
631 if (iter->rt6i_dev == rt->rt6i_dev &&
632 iter->rt6i_idev == rt->rt6i_idev &&
633 ipv6_addr_equal(&iter->rt6i_gateway,
634 &rt->rt6i_gateway)) {
635 if (!(iter->rt6i_flags&RTF_EXPIRES))
637 iter->rt6i_expires = rt->rt6i_expires;
638 if (!(rt->rt6i_flags&RTF_EXPIRES)) {
639 iter->rt6i_flags &= ~RTF_EXPIRES;
640 iter->rt6i_expires = 0;
646 if (iter->rt6i_metric > rt->rt6i_metric)
649 ins = &iter->u.dst.rt6_next;
652 /* Reset round-robin state, if necessary */
653 if (ins == &fn->leaf)
660 rt->u.dst.rt6_next = iter;
663 atomic_inc(&rt->rt6i_ref);
664 inet6_rt_notify(RTM_NEWROUTE, rt, info);
665 info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
667 if ((fn->fn_flags & RTN_RTINFO) == 0) {
668 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
669 fn->fn_flags |= RTN_RTINFO;
675 static __inline__ void fib6_start_gc(struct net *net, struct rt6_info *rt)
677 if (!timer_pending(&net->ipv6.ip6_fib_timer) &&
678 (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE)))
679 mod_timer(&net->ipv6.ip6_fib_timer,
680 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
683 void fib6_force_start_gc(struct net *net)
685 if (!timer_pending(&net->ipv6.ip6_fib_timer))
686 mod_timer(&net->ipv6.ip6_fib_timer,
687 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
691 * Add routing information to the routing tree.
692 * <destination addr>/<source addr>
693 * with source addr info in sub-trees
696 int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info)
698 struct fib6_node *fn, *pn = NULL;
701 fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr),
702 rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst));
709 #ifdef CONFIG_IPV6_SUBTREES
710 if (rt->rt6i_src.plen) {
711 struct fib6_node *sn;
713 if (fn->subtree == NULL) {
714 struct fib6_node *sfn;
726 /* Create subtree root node */
731 sfn->leaf = info->nl_net->ipv6.ip6_null_entry;
732 atomic_inc(&info->nl_net->ipv6.ip6_null_entry->rt6i_ref);
733 sfn->fn_flags = RTN_ROOT;
734 sfn->fn_sernum = fib6_new_sernum();
736 /* Now add the first leaf node to new subtree */
738 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
739 sizeof(struct in6_addr), rt->rt6i_src.plen,
740 offsetof(struct rt6_info, rt6i_src));
743 /* If it is failed, discard just allocated
744 root, and then (in st_failure) stale node
751 /* Now link new subtree to main tree */
755 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
756 sizeof(struct in6_addr), rt->rt6i_src.plen,
757 offsetof(struct rt6_info, rt6i_src));
763 if (fn->leaf == NULL) {
765 atomic_inc(&rt->rt6i_ref);
771 err = fib6_add_rt2node(fn, rt, info);
774 fib6_start_gc(info->nl_net, rt);
775 if (!(rt->rt6i_flags&RTF_CACHE))
776 fib6_prune_clones(info->nl_net, pn, rt);
781 #ifdef CONFIG_IPV6_SUBTREES
783 * If fib6_add_1 has cleared the old leaf pointer in the
784 * super-tree leaf node we have to find a new one for it.
786 if (pn != fn && pn->leaf == rt) {
788 atomic_dec(&rt->rt6i_ref);
790 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
791 pn->leaf = fib6_find_prefix(info->nl_net, pn);
794 WARN_ON(pn->leaf == NULL);
795 pn->leaf = info->nl_net->ipv6.ip6_null_entry;
798 atomic_inc(&pn->leaf->rt6i_ref);
801 dst_free(&rt->u.dst);
805 #ifdef CONFIG_IPV6_SUBTREES
806 /* Subtree creation failed, probably main tree node
807 is orphan. If it is, shoot it.
810 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
811 fib6_repair_tree(info->nl_net, fn);
812 dst_free(&rt->u.dst);
818 * Routing tree lookup
823 int offset; /* key offset on rt6_info */
824 struct in6_addr *addr; /* search key */
827 static struct fib6_node * fib6_lookup_1(struct fib6_node *root,
828 struct lookup_args *args)
830 struct fib6_node *fn;
833 if (unlikely(args->offset == 0))
843 struct fib6_node *next;
845 dir = addr_bit_set(args->addr, fn->fn_bit);
847 next = dir ? fn->right : fn->left;
858 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
861 key = (struct rt6key *) ((u8 *) fn->leaf +
864 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
865 #ifdef CONFIG_IPV6_SUBTREES
867 fn = fib6_lookup_1(fn->subtree, args + 1);
869 if (!fn || fn->fn_flags & RTN_RTINFO)
874 if (fn->fn_flags & RTN_ROOT)
883 struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr,
884 struct in6_addr *saddr)
886 struct fib6_node *fn;
887 struct lookup_args args[] = {
889 .offset = offsetof(struct rt6_info, rt6i_dst),
892 #ifdef CONFIG_IPV6_SUBTREES
894 .offset = offsetof(struct rt6_info, rt6i_src),
899 .offset = 0, /* sentinel */
903 fn = fib6_lookup_1(root, daddr ? args : args + 1);
905 if (fn == NULL || fn->fn_flags & RTN_TL_ROOT)
912 * Get node with specified destination prefix (and source prefix,
913 * if subtrees are used)
917 static struct fib6_node * fib6_locate_1(struct fib6_node *root,
918 struct in6_addr *addr,
919 int plen, int offset)
921 struct fib6_node *fn;
923 for (fn = root; fn ; ) {
924 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
929 if (plen < fn->fn_bit ||
930 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
933 if (plen == fn->fn_bit)
937 * We have more bits to go
939 if (addr_bit_set(addr, fn->fn_bit))
947 struct fib6_node * fib6_locate(struct fib6_node *root,
948 struct in6_addr *daddr, int dst_len,
949 struct in6_addr *saddr, int src_len)
951 struct fib6_node *fn;
953 fn = fib6_locate_1(root, daddr, dst_len,
954 offsetof(struct rt6_info, rt6i_dst));
956 #ifdef CONFIG_IPV6_SUBTREES
958 WARN_ON(saddr == NULL);
959 if (fn && fn->subtree)
960 fn = fib6_locate_1(fn->subtree, saddr, src_len,
961 offsetof(struct rt6_info, rt6i_src));
965 if (fn && fn->fn_flags&RTN_RTINFO)
977 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn)
979 if (fn->fn_flags&RTN_ROOT)
980 return net->ipv6.ip6_null_entry;
984 return fn->left->leaf;
987 return fn->right->leaf;
989 fn = FIB6_SUBTREE(fn);
995 * Called to trim the tree of intermediate nodes when possible. "fn"
996 * is the node we want to try and remove.
999 static struct fib6_node *fib6_repair_tree(struct net *net,
1000 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 WARN_ON(fn->fn_flags & RTN_RTINFO);
1013 WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1014 WARN_ON(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(net, fn);
1029 if (fn->leaf==NULL) {
1031 fn->leaf = net->ipv6.ip6_null_entry;
1034 atomic_inc(&fn->leaf->rt6i_ref);
1039 #ifdef CONFIG_IPV6_SUBTREES
1040 if (FIB6_SUBTREE(pn) == fn) {
1041 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1042 FIB6_SUBTREE(pn) = NULL;
1045 WARN_ON(fn->fn_flags & RTN_ROOT);
1047 if (pn->right == fn) pn->right = child;
1048 else if (pn->left == fn) pn->left = child;
1056 #ifdef CONFIG_IPV6_SUBTREES
1060 read_lock(&fib6_walker_lock);
1062 if (child == NULL) {
1063 if (w->root == fn) {
1064 w->root = w->node = NULL;
1065 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1066 } else if (w->node == fn) {
1067 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1072 if (w->root == fn) {
1074 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1076 if (w->node == fn) {
1079 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1080 w->state = w->state>=FWS_R ? FWS_U : FWS_INIT;
1082 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1083 w->state = w->state>=FWS_C ? FWS_U : FWS_INIT;
1088 read_unlock(&fib6_walker_lock);
1091 if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn))
1094 rt6_release(pn->leaf);
1100 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1101 struct nl_info *info)
1103 struct fib6_walker_t *w;
1104 struct rt6_info *rt = *rtp;
1105 struct net *net = info->nl_net;
1107 RT6_TRACE("fib6_del_route\n");
1110 *rtp = rt->u.dst.rt6_next;
1111 rt->rt6i_node = NULL;
1112 net->ipv6.rt6_stats->fib_rt_entries--;
1113 net->ipv6.rt6_stats->fib_discarded_routes++;
1115 /* Reset round-robin state, if necessary */
1116 if (fn->rr_ptr == rt)
1119 /* Adjust walkers */
1120 read_lock(&fib6_walker_lock);
1122 if (w->state == FWS_C && w->leaf == rt) {
1123 RT6_TRACE("walker %p adjusted by delroute\n", w);
1124 w->leaf = rt->u.dst.rt6_next;
1125 if (w->leaf == NULL)
1129 read_unlock(&fib6_walker_lock);
1131 rt->u.dst.rt6_next = NULL;
1133 /* If it was last route, expunge its radix tree node */
1134 if (fn->leaf == NULL) {
1135 fn->fn_flags &= ~RTN_RTINFO;
1136 net->ipv6.rt6_stats->fib_route_nodes--;
1137 fn = fib6_repair_tree(net, fn);
1140 if (atomic_read(&rt->rt6i_ref) != 1) {
1141 /* This route is used as dummy address holder in some split
1142 * nodes. It is not leaked, but it still holds other resources,
1143 * which must be released in time. So, scan ascendant nodes
1144 * and replace dummy references to this route with references
1145 * to still alive ones.
1148 if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) {
1149 fn->leaf = fib6_find_prefix(net, fn);
1150 atomic_inc(&fn->leaf->rt6i_ref);
1155 /* No more references are possible at this point. */
1156 BUG_ON(atomic_read(&rt->rt6i_ref) != 1);
1159 inet6_rt_notify(RTM_DELROUTE, rt, info);
1163 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1165 struct net *net = info->nl_net;
1166 struct fib6_node *fn = rt->rt6i_node;
1167 struct rt6_info **rtp;
1170 if (rt->u.dst.obsolete>0) {
1171 WARN_ON(fn != NULL);
1175 if (fn == NULL || rt == net->ipv6.ip6_null_entry)
1178 WARN_ON(!(fn->fn_flags & RTN_RTINFO));
1180 if (!(rt->rt6i_flags&RTF_CACHE)) {
1181 struct fib6_node *pn = fn;
1182 #ifdef CONFIG_IPV6_SUBTREES
1183 /* clones of this route might be in another subtree */
1184 if (rt->rt6i_src.plen) {
1185 while (!(pn->fn_flags&RTN_ROOT))
1190 fib6_prune_clones(info->nl_net, pn, rt);
1194 * Walk the leaf entries looking for ourself
1197 for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.dst.rt6_next) {
1199 fib6_del_route(fn, rtp, info);
1207 * Tree traversal function.
1209 * Certainly, it is not interrupt safe.
1210 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1211 * It means, that we can modify tree during walking
1212 * and use this function for garbage collection, clone pruning,
1213 * cleaning tree when a device goes down etc. etc.
1215 * It guarantees that every node will be traversed,
1216 * and that it will be traversed only once.
1218 * Callback function w->func may return:
1219 * 0 -> continue walking.
1220 * positive value -> walking is suspended (used by tree dumps,
1221 * and probably by gc, if it will be split to several slices)
1222 * negative value -> terminate walking.
1224 * The function itself returns:
1225 * 0 -> walk is complete.
1226 * >0 -> walk is incomplete (i.e. suspended)
1227 * <0 -> walk is terminated by an error.
1230 static int fib6_walk_continue(struct fib6_walker_t *w)
1232 struct fib6_node *fn, *pn;
1239 if (w->prune && fn != w->root &&
1240 fn->fn_flags&RTN_RTINFO && w->state < FWS_C) {
1245 #ifdef CONFIG_IPV6_SUBTREES
1247 if (FIB6_SUBTREE(fn)) {
1248 w->node = FIB6_SUBTREE(fn);
1256 w->state = FWS_INIT;
1262 w->node = fn->right;
1263 w->state = FWS_INIT;
1269 if (w->leaf && fn->fn_flags&RTN_RTINFO) {
1272 if (w->count < w->skip) {
1290 #ifdef CONFIG_IPV6_SUBTREES
1291 if (FIB6_SUBTREE(pn) == fn) {
1292 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1297 if (pn->left == fn) {
1301 if (pn->right == fn) {
1303 w->leaf = w->node->leaf;
1313 static int fib6_walk(struct fib6_walker_t *w)
1317 w->state = FWS_INIT;
1320 fib6_walker_link(w);
1321 res = fib6_walk_continue(w);
1323 fib6_walker_unlink(w);
1327 static int fib6_clean_node(struct fib6_walker_t *w)
1330 struct rt6_info *rt;
1331 struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w);
1332 struct nl_info info = {
1336 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
1337 res = c->func(rt, c->arg);
1340 res = fib6_del(rt, &info);
1343 printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res);
1356 * Convenient frontend to tree walker.
1358 * func is called on each route.
1359 * It may return -1 -> delete this route.
1360 * 0 -> continue walking
1362 * prune==1 -> only immediate children of node (certainly,
1363 * ignoring pure split nodes) will be scanned.
1366 static void fib6_clean_tree(struct net *net, struct fib6_node *root,
1367 int (*func)(struct rt6_info *, void *arg),
1368 int prune, void *arg)
1370 struct fib6_cleaner_t c;
1373 c.w.func = fib6_clean_node;
1384 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg),
1385 int prune, void *arg)
1387 struct fib6_table *table;
1388 struct hlist_node *node;
1389 struct hlist_head *head;
1393 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
1394 head = &net->ipv6.fib_table_hash[h];
1395 hlist_for_each_entry_rcu(table, node, head, tb6_hlist) {
1396 write_lock_bh(&table->tb6_lock);
1397 fib6_clean_tree(net, &table->tb6_root,
1399 write_unlock_bh(&table->tb6_lock);
1405 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1407 if (rt->rt6i_flags & RTF_CACHE) {
1408 RT6_TRACE("pruning clone %p\n", rt);
1415 static void fib6_prune_clones(struct net *net, struct fib6_node *fn,
1416 struct rt6_info *rt)
1418 fib6_clean_tree(net, fn, fib6_prune_clone, 1, rt);
1422 * Garbage collection
1425 static struct fib6_gc_args
1431 static int fib6_age(struct rt6_info *rt, void *arg)
1433 unsigned long now = jiffies;
1436 * check addrconf expiration here.
1437 * Routes are expired even if they are in use.
1439 * Also age clones. Note, that clones are aged out
1440 * only if they are not in use now.
1443 if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) {
1444 if (time_after(now, rt->rt6i_expires)) {
1445 RT6_TRACE("expiring %p\n", rt);
1449 } else if (rt->rt6i_flags & RTF_CACHE) {
1450 if (atomic_read(&rt->u.dst.__refcnt) == 0 &&
1451 time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) {
1452 RT6_TRACE("aging clone %p\n", rt);
1454 } else if ((rt->rt6i_flags & RTF_GATEWAY) &&
1455 (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) {
1456 RT6_TRACE("purging route %p via non-router but gateway\n",
1466 static DEFINE_SPINLOCK(fib6_gc_lock);
1468 void fib6_run_gc(unsigned long expires, struct net *net)
1470 if (expires != ~0UL) {
1471 spin_lock_bh(&fib6_gc_lock);
1472 gc_args.timeout = expires ? (int)expires :
1473 net->ipv6.sysctl.ip6_rt_gc_interval;
1475 if (!spin_trylock_bh(&fib6_gc_lock)) {
1476 mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
1479 gc_args.timeout = net->ipv6.sysctl.ip6_rt_gc_interval;
1482 gc_args.more = icmp6_dst_gc();
1484 fib6_clean_all(net, fib6_age, 0, NULL);
1487 mod_timer(&net->ipv6.ip6_fib_timer,
1488 round_jiffies(jiffies
1489 + net->ipv6.sysctl.ip6_rt_gc_interval));
1491 del_timer(&net->ipv6.ip6_fib_timer);
1492 spin_unlock_bh(&fib6_gc_lock);
1495 static void fib6_gc_timer_cb(unsigned long arg)
1497 fib6_run_gc(0, (struct net *)arg);
1500 static int __net_init fib6_net_init(struct net *net)
1502 setup_timer(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, (unsigned long)net);
1504 net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL);
1505 if (!net->ipv6.rt6_stats)
1508 net->ipv6.fib_table_hash = kcalloc(FIB6_TABLE_HASHSZ,
1509 sizeof(*net->ipv6.fib_table_hash),
1511 if (!net->ipv6.fib_table_hash)
1514 net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
1516 if (!net->ipv6.fib6_main_tbl)
1517 goto out_fib_table_hash;
1519 net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
1520 net->ipv6.fib6_main_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1521 net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
1522 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1524 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1525 net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
1527 if (!net->ipv6.fib6_local_tbl)
1528 goto out_fib6_main_tbl;
1529 net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
1530 net->ipv6.fib6_local_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1531 net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
1532 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1534 fib6_tables_init(net);
1538 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1540 kfree(net->ipv6.fib6_main_tbl);
1543 kfree(net->ipv6.fib_table_hash);
1545 kfree(net->ipv6.rt6_stats);
1550 static void fib6_net_exit(struct net *net)
1552 rt6_ifdown(net, NULL);
1553 del_timer_sync(&net->ipv6.ip6_fib_timer);
1555 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1556 kfree(net->ipv6.fib6_local_tbl);
1558 kfree(net->ipv6.fib6_main_tbl);
1559 kfree(net->ipv6.fib_table_hash);
1560 kfree(net->ipv6.rt6_stats);
1563 static struct pernet_operations fib6_net_ops = {
1564 .init = fib6_net_init,
1565 .exit = fib6_net_exit,
1568 int __init fib6_init(void)
1572 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1573 sizeof(struct fib6_node),
1574 0, SLAB_HWCACHE_ALIGN,
1576 if (!fib6_node_kmem)
1579 ret = register_pernet_subsys(&fib6_net_ops);
1581 goto out_kmem_cache_create;
1583 ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib);
1585 goto out_unregister_subsys;
1589 out_unregister_subsys:
1590 unregister_pernet_subsys(&fib6_net_ops);
1591 out_kmem_cache_create:
1592 kmem_cache_destroy(fib6_node_kmem);
1596 void fib6_gc_cleanup(void)
1598 unregister_pernet_subsys(&fib6_net_ops);
1599 kmem_cache_destroy(fib6_node_kmem);