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 void fib6_gc_timer_cb(unsigned long arg);
98 static DEFINE_TIMER(ip6_fib_timer, fib6_gc_timer_cb, 0,
99 (unsigned long)&init_net);
101 static struct fib6_walker_t fib6_walker_list = {
102 .prev = &fib6_walker_list,
103 .next = &fib6_walker_list,
106 #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next)
108 static inline void fib6_walker_link(struct fib6_walker_t *w)
110 write_lock_bh(&fib6_walker_lock);
111 w->next = fib6_walker_list.next;
112 w->prev = &fib6_walker_list;
115 write_unlock_bh(&fib6_walker_lock);
118 static inline void fib6_walker_unlink(struct fib6_walker_t *w)
120 write_lock_bh(&fib6_walker_lock);
121 w->next->prev = w->prev;
122 w->prev->next = w->next;
123 w->prev = w->next = w;
124 write_unlock_bh(&fib6_walker_lock);
126 static __inline__ u32 fib6_new_sernum(void)
135 * Auxiliary address test functions for the radix tree.
137 * These assume a 32bit processor (although it will work on
145 static __inline__ __be32 addr_bit_set(void *token, int fn_bit)
147 __be32 *addr = token;
149 return htonl(1 << ((~fn_bit)&0x1F)) & addr[fn_bit>>5];
152 static __inline__ struct fib6_node * node_alloc(void)
154 struct fib6_node *fn;
156 fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
161 static __inline__ void node_free(struct fib6_node * fn)
163 kmem_cache_free(fib6_node_kmem, fn);
166 static __inline__ void rt6_release(struct rt6_info *rt)
168 if (atomic_dec_and_test(&rt->rt6i_ref))
169 dst_free(&rt->u.dst);
172 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
173 #define FIB_TABLE_HASHSZ 256
175 #define FIB_TABLE_HASHSZ 1
178 static void fib6_link_table(struct net *net, 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, &net->ipv6.fib_table_hash[h]);
197 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
199 static struct fib6_table *fib6_alloc_table(u32 id)
201 struct fib6_table *table;
203 table = kzalloc(sizeof(*table), GFP_ATOMIC);
206 table->tb6_root.leaf = &ip6_null_entry;
207 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
213 struct fib6_table *fib6_new_table(struct net *net, u32 id)
215 struct fib6_table *tb;
219 tb = fib6_get_table(net, id);
223 tb = fib6_alloc_table(id);
225 fib6_link_table(net, tb);
230 struct fib6_table *fib6_get_table(struct net *net, u32 id)
232 struct fib6_table *tb;
233 struct hlist_head *head;
234 struct hlist_node *node;
239 h = id & (FIB_TABLE_HASHSZ - 1);
241 head = &net->ipv6.fib_table_hash[h];
242 hlist_for_each_entry_rcu(tb, node, head, tb6_hlist) {
243 if (tb->tb6_id == id) {
253 static void fib6_tables_init(struct net *net)
255 fib6_link_table(net, net->ipv6.fib6_main_tbl);
256 fib6_link_table(net, net->ipv6.fib6_local_tbl);
260 struct fib6_table *fib6_new_table(struct net *net, u32 id)
262 return fib6_get_table(net, id);
265 struct fib6_table *fib6_get_table(struct net *net, u32 id)
267 return net->ipv6.fib6_main_tbl;
270 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi *fl,
271 int flags, pol_lookup_t lookup)
273 return (struct dst_entry *) lookup(net->ipv6.fib6_main_tbl, fl, flags);
276 static void fib6_tables_init(struct net *net)
278 fib6_link_table(net, net->ipv6.fib6_main_tbl);
283 static int fib6_dump_node(struct fib6_walker_t *w)
288 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
289 res = rt6_dump_route(rt, w->args);
291 /* Frame is full, suspend walking */
301 static void fib6_dump_end(struct netlink_callback *cb)
303 struct fib6_walker_t *w = (void*)cb->args[2];
309 cb->done = (void*)cb->args[3];
313 static int fib6_dump_done(struct netlink_callback *cb)
316 return cb->done ? cb->done(cb) : 0;
319 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
320 struct netlink_callback *cb)
322 struct fib6_walker_t *w;
325 w = (void *)cb->args[2];
326 w->root = &table->tb6_root;
328 if (cb->args[4] == 0) {
329 read_lock_bh(&table->tb6_lock);
331 read_unlock_bh(&table->tb6_lock);
335 read_lock_bh(&table->tb6_lock);
336 res = fib6_walk_continue(w);
337 read_unlock_bh(&table->tb6_lock);
340 fib6_walker_unlink(w);
343 fib6_walker_unlink(w);
350 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
352 struct net *net = skb->sk->sk_net;
354 unsigned int e = 0, s_e;
355 struct rt6_rtnl_dump_arg arg;
356 struct fib6_walker_t *w;
357 struct fib6_table *tb;
358 struct hlist_node *node;
359 struct hlist_head *head;
365 w = (void *)cb->args[2];
369 * 1. hook callback destructor.
371 cb->args[3] = (long)cb->done;
372 cb->done = fib6_dump_done;
375 * 2. allocate and initialize walker.
377 w = kzalloc(sizeof(*w), GFP_ATOMIC);
380 w->func = fib6_dump_node;
381 cb->args[2] = (long)w;
388 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
390 head = &net->ipv6.fib_table_hash[h];
391 hlist_for_each_entry(tb, node, head, tb6_hlist) {
394 res = fib6_dump_table(tb, skb, cb);
405 res = res < 0 ? res : skb->len;
414 * return the appropriate node for a routing tree "add" operation
415 * by either creating and inserting or by returning an existing
419 static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr,
420 int addrlen, int plen,
423 struct fib6_node *fn, *in, *ln;
424 struct fib6_node *pn = NULL;
428 __u32 sernum = fib6_new_sernum();
430 RT6_TRACE("fib6_add_1\n");
432 /* insert node in tree */
437 key = (struct rt6key *)((u8 *)fn->leaf + offset);
442 if (plen < fn->fn_bit ||
443 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
450 if (plen == fn->fn_bit) {
451 /* clean up an intermediate node */
452 if ((fn->fn_flags & RTN_RTINFO) == 0) {
453 rt6_release(fn->leaf);
457 fn->fn_sernum = sernum;
463 * We have more bits to go
466 /* Try to walk down on tree. */
467 fn->fn_sernum = sernum;
468 dir = addr_bit_set(addr, fn->fn_bit);
470 fn = dir ? fn->right: fn->left;
474 * We walked to the bottom of tree.
475 * Create new leaf node without children.
485 ln->fn_sernum = sernum;
497 * split since we don't have a common prefix anymore or
498 * we have a less significant route.
499 * we've to insert an intermediate node on the list
500 * this new node will point to the one we need to create
506 /* find 1st bit in difference between the 2 addrs.
508 See comment in __ipv6_addr_diff: bit may be an invalid value,
509 but if it is >= plen, the value is ignored in any case.
512 bit = __ipv6_addr_diff(addr, &key->addr, addrlen);
517 * (new leaf node)[ln] (old node)[fn]
523 if (in == NULL || ln == NULL) {
532 * new intermediate node.
534 * be off since that an address that chooses one of
535 * the branches would not match less specific routes
536 * in the other branch
543 atomic_inc(&in->leaf->rt6i_ref);
545 in->fn_sernum = sernum;
547 /* update parent pointer */
558 ln->fn_sernum = sernum;
560 if (addr_bit_set(addr, bit)) {
567 } else { /* plen <= bit */
570 * (new leaf node)[ln]
572 * (old node)[fn] NULL
584 ln->fn_sernum = sernum;
591 if (addr_bit_set(&key->addr, plen))
602 * Insert routing information in a node.
605 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
606 struct nl_info *info)
608 struct rt6_info *iter = NULL;
609 struct rt6_info **ins;
613 for (iter = fn->leaf; iter; iter=iter->u.dst.rt6_next) {
615 * Search for duplicates
618 if (iter->rt6i_metric == rt->rt6i_metric) {
620 * Same priority level
623 if (iter->rt6i_dev == rt->rt6i_dev &&
624 iter->rt6i_idev == rt->rt6i_idev &&
625 ipv6_addr_equal(&iter->rt6i_gateway,
626 &rt->rt6i_gateway)) {
627 if (!(iter->rt6i_flags&RTF_EXPIRES))
629 iter->rt6i_expires = rt->rt6i_expires;
630 if (!(rt->rt6i_flags&RTF_EXPIRES)) {
631 iter->rt6i_flags &= ~RTF_EXPIRES;
632 iter->rt6i_expires = 0;
638 if (iter->rt6i_metric > rt->rt6i_metric)
641 ins = &iter->u.dst.rt6_next;
644 /* Reset round-robin state, if necessary */
645 if (ins == &fn->leaf)
652 rt->u.dst.rt6_next = iter;
655 atomic_inc(&rt->rt6i_ref);
656 inet6_rt_notify(RTM_NEWROUTE, rt, info);
657 rt6_stats.fib_rt_entries++;
659 if ((fn->fn_flags & RTN_RTINFO) == 0) {
660 rt6_stats.fib_route_nodes++;
661 fn->fn_flags |= RTN_RTINFO;
667 static __inline__ void fib6_start_gc(struct rt6_info *rt)
669 if (ip6_fib_timer.expires == 0 &&
670 (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE)))
671 mod_timer(&ip6_fib_timer, jiffies +
672 init_net.ipv6.sysctl.ip6_rt_gc_interval);
675 void fib6_force_start_gc(void)
677 if (ip6_fib_timer.expires == 0)
678 mod_timer(&ip6_fib_timer, jiffies +
679 init_net.ipv6.sysctl.ip6_rt_gc_interval);
683 * Add routing information to the routing tree.
684 * <destination addr>/<source addr>
685 * with source addr info in sub-trees
688 int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info)
690 struct fib6_node *fn, *pn = NULL;
693 fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr),
694 rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst));
701 #ifdef CONFIG_IPV6_SUBTREES
702 if (rt->rt6i_src.plen) {
703 struct fib6_node *sn;
705 if (fn->subtree == NULL) {
706 struct fib6_node *sfn;
718 /* Create subtree root node */
723 sfn->leaf = &ip6_null_entry;
724 atomic_inc(&ip6_null_entry.rt6i_ref);
725 sfn->fn_flags = RTN_ROOT;
726 sfn->fn_sernum = fib6_new_sernum();
728 /* Now add the first leaf node to new subtree */
730 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
731 sizeof(struct in6_addr), rt->rt6i_src.plen,
732 offsetof(struct rt6_info, rt6i_src));
735 /* If it is failed, discard just allocated
736 root, and then (in st_failure) stale node
743 /* Now link new subtree to main tree */
747 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
748 sizeof(struct in6_addr), rt->rt6i_src.plen,
749 offsetof(struct rt6_info, rt6i_src));
755 if (fn->leaf == NULL) {
757 atomic_inc(&rt->rt6i_ref);
763 err = fib6_add_rt2node(fn, rt, info);
767 if (!(rt->rt6i_flags&RTF_CACHE))
768 fib6_prune_clones(pn, rt);
773 #ifdef CONFIG_IPV6_SUBTREES
775 * If fib6_add_1 has cleared the old leaf pointer in the
776 * super-tree leaf node we have to find a new one for it.
778 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
779 pn->leaf = fib6_find_prefix(pn);
782 BUG_TRAP(pn->leaf != NULL);
783 pn->leaf = &ip6_null_entry;
786 atomic_inc(&pn->leaf->rt6i_ref);
789 dst_free(&rt->u.dst);
793 #ifdef CONFIG_IPV6_SUBTREES
794 /* Subtree creation failed, probably main tree node
795 is orphan. If it is, shoot it.
798 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
799 fib6_repair_tree(fn);
800 dst_free(&rt->u.dst);
806 * Routing tree lookup
811 int offset; /* key offset on rt6_info */
812 struct in6_addr *addr; /* search key */
815 static struct fib6_node * fib6_lookup_1(struct fib6_node *root,
816 struct lookup_args *args)
818 struct fib6_node *fn;
821 if (unlikely(args->offset == 0))
831 struct fib6_node *next;
833 dir = addr_bit_set(args->addr, fn->fn_bit);
835 next = dir ? fn->right : fn->left;
846 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
849 key = (struct rt6key *) ((u8 *) fn->leaf +
852 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
853 #ifdef CONFIG_IPV6_SUBTREES
855 fn = fib6_lookup_1(fn->subtree, args + 1);
857 if (!fn || fn->fn_flags & RTN_RTINFO)
862 if (fn->fn_flags & RTN_ROOT)
871 struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr,
872 struct in6_addr *saddr)
874 struct fib6_node *fn;
875 struct lookup_args args[] = {
877 .offset = offsetof(struct rt6_info, rt6i_dst),
880 #ifdef CONFIG_IPV6_SUBTREES
882 .offset = offsetof(struct rt6_info, rt6i_src),
887 .offset = 0, /* sentinel */
891 fn = fib6_lookup_1(root, daddr ? args : args + 1);
893 if (fn == NULL || fn->fn_flags & RTN_TL_ROOT)
900 * Get node with specified destination prefix (and source prefix,
901 * if subtrees are used)
905 static struct fib6_node * fib6_locate_1(struct fib6_node *root,
906 struct in6_addr *addr,
907 int plen, int offset)
909 struct fib6_node *fn;
911 for (fn = root; fn ; ) {
912 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
917 if (plen < fn->fn_bit ||
918 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
921 if (plen == fn->fn_bit)
925 * We have more bits to go
927 if (addr_bit_set(addr, fn->fn_bit))
935 struct fib6_node * fib6_locate(struct fib6_node *root,
936 struct in6_addr *daddr, int dst_len,
937 struct in6_addr *saddr, int src_len)
939 struct fib6_node *fn;
941 fn = fib6_locate_1(root, daddr, dst_len,
942 offsetof(struct rt6_info, rt6i_dst));
944 #ifdef CONFIG_IPV6_SUBTREES
946 BUG_TRAP(saddr!=NULL);
947 if (fn && fn->subtree)
948 fn = fib6_locate_1(fn->subtree, saddr, src_len,
949 offsetof(struct rt6_info, rt6i_src));
953 if (fn && fn->fn_flags&RTN_RTINFO)
965 static struct rt6_info * fib6_find_prefix(struct fib6_node *fn)
967 if (fn->fn_flags&RTN_ROOT)
968 return &ip6_null_entry;
972 return fn->left->leaf;
975 return fn->right->leaf;
977 fn = FIB6_SUBTREE(fn);
983 * Called to trim the tree of intermediate nodes when possible. "fn"
984 * is the node we want to try and remove.
987 static struct fib6_node * fib6_repair_tree(struct fib6_node *fn)
991 struct fib6_node *child, *pn;
992 struct fib6_walker_t *w;
996 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
999 BUG_TRAP(!(fn->fn_flags&RTN_RTINFO));
1000 BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT));
1001 BUG_TRAP(fn->leaf==NULL);
1005 if (fn->right) child = fn->right, children |= 1;
1006 if (fn->left) child = fn->left, children |= 2;
1008 if (children == 3 || FIB6_SUBTREE(fn)
1009 #ifdef CONFIG_IPV6_SUBTREES
1010 /* Subtree root (i.e. fn) may have one child */
1011 || (children && fn->fn_flags&RTN_ROOT)
1014 fn->leaf = fib6_find_prefix(fn);
1016 if (fn->leaf==NULL) {
1018 fn->leaf = &ip6_null_entry;
1021 atomic_inc(&fn->leaf->rt6i_ref);
1026 #ifdef CONFIG_IPV6_SUBTREES
1027 if (FIB6_SUBTREE(pn) == fn) {
1028 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1029 FIB6_SUBTREE(pn) = NULL;
1032 BUG_TRAP(!(fn->fn_flags&RTN_ROOT));
1034 if (pn->right == fn) pn->right = child;
1035 else if (pn->left == fn) pn->left = child;
1042 #ifdef CONFIG_IPV6_SUBTREES
1046 read_lock(&fib6_walker_lock);
1048 if (child == NULL) {
1049 if (w->root == fn) {
1050 w->root = w->node = NULL;
1051 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1052 } else if (w->node == fn) {
1053 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1058 if (w->root == fn) {
1060 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1062 if (w->node == fn) {
1065 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1066 w->state = w->state>=FWS_R ? FWS_U : FWS_INIT;
1068 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1069 w->state = w->state>=FWS_C ? FWS_U : FWS_INIT;
1074 read_unlock(&fib6_walker_lock);
1077 if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn))
1080 rt6_release(pn->leaf);
1086 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1087 struct nl_info *info)
1089 struct fib6_walker_t *w;
1090 struct rt6_info *rt = *rtp;
1092 RT6_TRACE("fib6_del_route\n");
1095 *rtp = rt->u.dst.rt6_next;
1096 rt->rt6i_node = NULL;
1097 rt6_stats.fib_rt_entries--;
1098 rt6_stats.fib_discarded_routes++;
1100 /* Reset round-robin state, if necessary */
1101 if (fn->rr_ptr == rt)
1104 /* Adjust walkers */
1105 read_lock(&fib6_walker_lock);
1107 if (w->state == FWS_C && w->leaf == rt) {
1108 RT6_TRACE("walker %p adjusted by delroute\n", w);
1109 w->leaf = rt->u.dst.rt6_next;
1110 if (w->leaf == NULL)
1114 read_unlock(&fib6_walker_lock);
1116 rt->u.dst.rt6_next = NULL;
1118 /* If it was last route, expunge its radix tree node */
1119 if (fn->leaf == NULL) {
1120 fn->fn_flags &= ~RTN_RTINFO;
1121 rt6_stats.fib_route_nodes--;
1122 fn = fib6_repair_tree(fn);
1125 if (atomic_read(&rt->rt6i_ref) != 1) {
1126 /* This route is used as dummy address holder in some split
1127 * nodes. It is not leaked, but it still holds other resources,
1128 * which must be released in time. So, scan ascendant nodes
1129 * and replace dummy references to this route with references
1130 * to still alive ones.
1133 if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) {
1134 fn->leaf = fib6_find_prefix(fn);
1135 atomic_inc(&fn->leaf->rt6i_ref);
1140 /* No more references are possible at this point. */
1141 BUG_ON(atomic_read(&rt->rt6i_ref) != 1);
1144 inet6_rt_notify(RTM_DELROUTE, rt, info);
1148 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1150 struct fib6_node *fn = rt->rt6i_node;
1151 struct rt6_info **rtp;
1154 if (rt->u.dst.obsolete>0) {
1159 if (fn == NULL || rt == &ip6_null_entry)
1162 BUG_TRAP(fn->fn_flags&RTN_RTINFO);
1164 if (!(rt->rt6i_flags&RTF_CACHE)) {
1165 struct fib6_node *pn = fn;
1166 #ifdef CONFIG_IPV6_SUBTREES
1167 /* clones of this route might be in another subtree */
1168 if (rt->rt6i_src.plen) {
1169 while (!(pn->fn_flags&RTN_ROOT))
1174 fib6_prune_clones(pn, rt);
1178 * Walk the leaf entries looking for ourself
1181 for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.dst.rt6_next) {
1183 fib6_del_route(fn, rtp, info);
1191 * Tree traversal function.
1193 * Certainly, it is not interrupt safe.
1194 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1195 * It means, that we can modify tree during walking
1196 * and use this function for garbage collection, clone pruning,
1197 * cleaning tree when a device goes down etc. etc.
1199 * It guarantees that every node will be traversed,
1200 * and that it will be traversed only once.
1202 * Callback function w->func may return:
1203 * 0 -> continue walking.
1204 * positive value -> walking is suspended (used by tree dumps,
1205 * and probably by gc, if it will be split to several slices)
1206 * negative value -> terminate walking.
1208 * The function itself returns:
1209 * 0 -> walk is complete.
1210 * >0 -> walk is incomplete (i.e. suspended)
1211 * <0 -> walk is terminated by an error.
1214 static int fib6_walk_continue(struct fib6_walker_t *w)
1216 struct fib6_node *fn, *pn;
1223 if (w->prune && fn != w->root &&
1224 fn->fn_flags&RTN_RTINFO && w->state < FWS_C) {
1229 #ifdef CONFIG_IPV6_SUBTREES
1231 if (FIB6_SUBTREE(fn)) {
1232 w->node = FIB6_SUBTREE(fn);
1240 w->state = FWS_INIT;
1246 w->node = fn->right;
1247 w->state = FWS_INIT;
1253 if (w->leaf && fn->fn_flags&RTN_RTINFO) {
1254 int err = w->func(w);
1265 #ifdef CONFIG_IPV6_SUBTREES
1266 if (FIB6_SUBTREE(pn) == fn) {
1267 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1272 if (pn->left == fn) {
1276 if (pn->right == fn) {
1278 w->leaf = w->node->leaf;
1288 static int fib6_walk(struct fib6_walker_t *w)
1292 w->state = FWS_INIT;
1295 fib6_walker_link(w);
1296 res = fib6_walk_continue(w);
1298 fib6_walker_unlink(w);
1302 static int fib6_clean_node(struct fib6_walker_t *w)
1304 struct nl_info info = {
1305 .nl_net = &init_net,
1308 struct rt6_info *rt;
1309 struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w);
1311 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
1312 res = c->func(rt, c->arg);
1315 res = fib6_del(rt, &info);
1318 printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res);
1331 * Convenient frontend to tree walker.
1333 * func is called on each route.
1334 * It may return -1 -> delete this route.
1335 * 0 -> continue walking
1337 * prune==1 -> only immediate children of node (certainly,
1338 * ignoring pure split nodes) will be scanned.
1341 static void fib6_clean_tree(struct fib6_node *root,
1342 int (*func)(struct rt6_info *, void *arg),
1343 int prune, void *arg)
1345 struct fib6_cleaner_t c;
1348 c.w.func = fib6_clean_node;
1356 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg),
1357 int prune, void *arg)
1359 struct fib6_table *table;
1360 struct hlist_node *node;
1361 struct hlist_head *head;
1365 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
1366 head = &net->ipv6.fib_table_hash[h];
1367 hlist_for_each_entry_rcu(table, node, head, tb6_hlist) {
1368 write_lock_bh(&table->tb6_lock);
1369 fib6_clean_tree(&table->tb6_root, func, prune, arg);
1370 write_unlock_bh(&table->tb6_lock);
1376 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1378 if (rt->rt6i_flags & RTF_CACHE) {
1379 RT6_TRACE("pruning clone %p\n", rt);
1386 static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt)
1388 fib6_clean_tree(fn, fib6_prune_clone, 1, rt);
1392 * Garbage collection
1395 static struct fib6_gc_args
1401 static int fib6_age(struct rt6_info *rt, void *arg)
1403 unsigned long now = jiffies;
1406 * check addrconf expiration here.
1407 * Routes are expired even if they are in use.
1409 * Also age clones. Note, that clones are aged out
1410 * only if they are not in use now.
1413 if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) {
1414 if (time_after(now, rt->rt6i_expires)) {
1415 RT6_TRACE("expiring %p\n", rt);
1419 } else if (rt->rt6i_flags & RTF_CACHE) {
1420 if (atomic_read(&rt->u.dst.__refcnt) == 0 &&
1421 time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) {
1422 RT6_TRACE("aging clone %p\n", rt);
1424 } else if ((rt->rt6i_flags & RTF_GATEWAY) &&
1425 (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) {
1426 RT6_TRACE("purging route %p via non-router but gateway\n",
1436 static DEFINE_SPINLOCK(fib6_gc_lock);
1438 void fib6_run_gc(unsigned long expires, struct net *net)
1440 if (expires != ~0UL) {
1441 spin_lock_bh(&fib6_gc_lock);
1442 gc_args.timeout = expires ? (int)expires :
1443 net->ipv6.sysctl.ip6_rt_gc_interval;
1446 if (!spin_trylock(&fib6_gc_lock)) {
1447 mod_timer(&ip6_fib_timer, jiffies + HZ);
1451 gc_args.timeout = net->ipv6.sysctl.ip6_rt_gc_interval;
1455 icmp6_dst_gc(&gc_args.more);
1457 fib6_clean_all(net, fib6_age, 0, NULL);
1460 mod_timer(&ip6_fib_timer, jiffies +
1461 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 static void fib6_gc_timer_cb(unsigned long arg)
1471 fib6_run_gc(0, (struct net *)arg);
1474 static int fib6_net_init(struct net *net)
1479 net->ipv6.fib_table_hash =
1480 kzalloc(sizeof(*net->ipv6.fib_table_hash)*FIB_TABLE_HASHSZ,
1482 if (!net->ipv6.fib_table_hash)
1485 net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
1487 if (!net->ipv6.fib6_main_tbl)
1488 goto out_fib_table_hash;
1490 net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
1491 net->ipv6.fib6_main_tbl->tb6_root.leaf = &ip6_null_entry;
1492 net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
1493 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1495 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1496 net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
1498 if (!net->ipv6.fib6_local_tbl)
1499 goto out_fib6_main_tbl;
1500 net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
1501 net->ipv6.fib6_local_tbl->tb6_root.leaf = &ip6_null_entry;
1502 net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
1503 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1505 fib6_tables_init(net);
1511 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1513 kfree(net->ipv6.fib6_main_tbl);
1516 kfree(net->ipv6.fib_table_hash);
1520 static void fib6_net_exit(struct net *net)
1522 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1523 kfree(net->ipv6.fib6_local_tbl);
1525 kfree(net->ipv6.fib6_main_tbl);
1526 kfree(net->ipv6.fib_table_hash);
1529 static struct pernet_operations fib6_net_ops = {
1530 .init = fib6_net_init,
1531 .exit = fib6_net_exit,
1534 int __init fib6_init(void)
1537 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1538 sizeof(struct fib6_node),
1539 0, SLAB_HWCACHE_ALIGN,
1541 if (!fib6_node_kmem)
1544 ret = register_pernet_subsys(&fib6_net_ops);
1546 goto out_kmem_cache_create;
1548 ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib);
1550 goto out_unregister_subsys;
1554 out_unregister_subsys:
1555 unregister_pernet_subsys(&fib6_net_ops);
1556 out_kmem_cache_create:
1557 kmem_cache_destroy(fib6_node_kmem);
1561 void fib6_gc_cleanup(void)
1563 del_timer(&ip6_fib_timer);
1564 unregister_pernet_subsys(&fib6_net_ops);
1565 kmem_cache_destroy(fib6_node_kmem);