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>
31 #include <linux/proc_fs.h>
35 #include <net/ndisc.h>
36 #include <net/addrconf.h>
38 #include <net/ip6_fib.h>
39 #include <net/ip6_route.h>
44 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
46 #define RT6_TRACE(x...) do { ; } while (0)
49 static struct kmem_cache * fib6_node_kmem __read_mostly;
53 #ifdef CONFIG_IPV6_SUBTREES
64 struct fib6_walker_t w;
66 int (*func)(struct rt6_info *, void *arg);
70 static DEFINE_RWLOCK(fib6_walker_lock);
72 #ifdef CONFIG_IPV6_SUBTREES
73 #define FWS_INIT FWS_S
75 #define FWS_INIT FWS_L
78 static void fib6_prune_clones(struct net *net, struct fib6_node *fn,
80 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn);
81 static struct fib6_node *fib6_repair_tree(struct net *net, struct fib6_node *fn);
82 static int fib6_walk(struct fib6_walker_t *w);
83 static int fib6_walk_continue(struct fib6_walker_t *w);
86 * A routing update causes an increase of the serial number on the
87 * affected subtree. This allows for cached routes to be asynchronously
88 * tested when modifications are made to the destination cache as a
89 * result of redirects, path MTU changes, etc.
92 static __u32 rt_sernum;
94 static void fib6_gc_timer_cb(unsigned long arg);
96 static struct fib6_walker_t fib6_walker_list = {
97 .prev = &fib6_walker_list,
98 .next = &fib6_walker_list,
101 #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next)
103 static inline void fib6_walker_link(struct fib6_walker_t *w)
105 write_lock_bh(&fib6_walker_lock);
106 w->next = fib6_walker_list.next;
107 w->prev = &fib6_walker_list;
110 write_unlock_bh(&fib6_walker_lock);
113 static inline void fib6_walker_unlink(struct fib6_walker_t *w)
115 write_lock_bh(&fib6_walker_lock);
116 w->next->prev = w->prev;
117 w->prev->next = w->next;
118 w->prev = w->next = w;
119 write_unlock_bh(&fib6_walker_lock);
121 static __inline__ u32 fib6_new_sernum(void)
130 * Auxiliary address test functions for the radix tree.
132 * These assume a 32bit processor (although it will work on
140 static __inline__ __be32 addr_bit_set(void *token, int fn_bit)
142 __be32 *addr = token;
144 return htonl(1 << ((~fn_bit)&0x1F)) & addr[fn_bit>>5];
147 static __inline__ struct fib6_node * node_alloc(void)
149 struct fib6_node *fn;
151 fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
156 static __inline__ void node_free(struct fib6_node * fn)
158 kmem_cache_free(fib6_node_kmem, fn);
161 static __inline__ void rt6_release(struct rt6_info *rt)
163 if (atomic_dec_and_test(&rt->rt6i_ref))
164 dst_free(&rt->u.dst);
167 static void fib6_link_table(struct net *net, struct fib6_table *tb)
172 * Initialize table lock at a single place to give lockdep a key,
173 * tables aren't visible prior to being linked to the list.
175 rwlock_init(&tb->tb6_lock);
177 h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1);
180 * No protection necessary, this is the only list mutatation
181 * operation, tables never disappear once they exist.
183 hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]);
186 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
188 static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
190 struct fib6_table *table;
192 table = kzalloc(sizeof(*table), GFP_ATOMIC);
195 table->tb6_root.leaf = net->ipv6.ip6_null_entry;
196 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
202 struct fib6_table *fib6_new_table(struct net *net, u32 id)
204 struct fib6_table *tb;
208 tb = fib6_get_table(net, id);
212 tb = fib6_alloc_table(net, id);
214 fib6_link_table(net, tb);
219 struct fib6_table *fib6_get_table(struct net *net, u32 id)
221 struct fib6_table *tb;
222 struct hlist_head *head;
223 struct hlist_node *node;
228 h = id & (FIB6_TABLE_HASHSZ - 1);
230 head = &net->ipv6.fib_table_hash[h];
231 hlist_for_each_entry_rcu(tb, node, head, tb6_hlist) {
232 if (tb->tb6_id == id) {
242 static void __net_init fib6_tables_init(struct net *net)
244 fib6_link_table(net, net->ipv6.fib6_main_tbl);
245 fib6_link_table(net, net->ipv6.fib6_local_tbl);
249 struct fib6_table *fib6_new_table(struct net *net, u32 id)
251 return fib6_get_table(net, id);
254 struct fib6_table *fib6_get_table(struct net *net, u32 id)
256 return net->ipv6.fib6_main_tbl;
259 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi *fl,
260 int flags, pol_lookup_t lookup)
262 return (struct dst_entry *) lookup(net, net->ipv6.fib6_main_tbl, fl, flags);
265 static void __net_init fib6_tables_init(struct net *net)
267 fib6_link_table(net, net->ipv6.fib6_main_tbl);
272 static int fib6_dump_node(struct fib6_walker_t *w)
277 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
278 res = rt6_dump_route(rt, w->args);
280 /* Frame is full, suspend walking */
290 static void fib6_dump_end(struct netlink_callback *cb)
292 struct fib6_walker_t *w = (void*)cb->args[2];
297 fib6_walker_unlink(w);
302 cb->done = (void*)cb->args[3];
306 static int fib6_dump_done(struct netlink_callback *cb)
309 return cb->done ? cb->done(cb) : 0;
312 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
313 struct netlink_callback *cb)
315 struct fib6_walker_t *w;
318 w = (void *)cb->args[2];
319 w->root = &table->tb6_root;
321 if (cb->args[4] == 0) {
325 read_lock_bh(&table->tb6_lock);
327 read_unlock_bh(&table->tb6_lock);
330 cb->args[5] = w->root->fn_sernum;
333 if (cb->args[5] != w->root->fn_sernum) {
334 /* Begin at the root if the tree changed */
335 cb->args[5] = w->root->fn_sernum;
342 read_lock_bh(&table->tb6_lock);
343 res = fib6_walk_continue(w);
344 read_unlock_bh(&table->tb6_lock);
346 fib6_walker_unlink(w);
354 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
356 struct net *net = sock_net(skb->sk);
358 unsigned int e = 0, s_e;
359 struct rt6_rtnl_dump_arg arg;
360 struct fib6_walker_t *w;
361 struct fib6_table *tb;
362 struct hlist_node *node;
363 struct hlist_head *head;
369 w = (void *)cb->args[2];
373 * 1. hook callback destructor.
375 cb->args[3] = (long)cb->done;
376 cb->done = fib6_dump_done;
379 * 2. allocate and initialize walker.
381 w = kzalloc(sizeof(*w), GFP_ATOMIC);
384 w->func = fib6_dump_node;
385 cb->args[2] = (long)w;
393 for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) {
395 head = &net->ipv6.fib_table_hash[h];
396 hlist_for_each_entry(tb, node, head, tb6_hlist) {
399 res = fib6_dump_table(tb, skb, cb);
410 res = res < 0 ? res : skb->len;
419 * return the appropriate node for a routing tree "add" operation
420 * by either creating and inserting or by returning an existing
424 static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr,
425 int addrlen, int plen,
428 struct fib6_node *fn, *in, *ln;
429 struct fib6_node *pn = NULL;
433 __u32 sernum = fib6_new_sernum();
435 RT6_TRACE("fib6_add_1\n");
437 /* insert node in tree */
442 key = (struct rt6key *)((u8 *)fn->leaf + offset);
447 if (plen < fn->fn_bit ||
448 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
455 if (plen == fn->fn_bit) {
456 /* clean up an intermediate node */
457 if ((fn->fn_flags & RTN_RTINFO) == 0) {
458 rt6_release(fn->leaf);
462 fn->fn_sernum = sernum;
468 * We have more bits to go
471 /* Try to walk down on tree. */
472 fn->fn_sernum = sernum;
473 dir = addr_bit_set(addr, fn->fn_bit);
475 fn = dir ? fn->right: fn->left;
479 * We walked to the bottom of tree.
480 * Create new leaf node without children.
490 ln->fn_sernum = sernum;
502 * split since we don't have a common prefix anymore or
503 * we have a less significant route.
504 * we've to insert an intermediate node on the list
505 * this new node will point to the one we need to create
511 /* find 1st bit in difference between the 2 addrs.
513 See comment in __ipv6_addr_diff: bit may be an invalid value,
514 but if it is >= plen, the value is ignored in any case.
517 bit = __ipv6_addr_diff(addr, &key->addr, addrlen);
522 * (new leaf node)[ln] (old node)[fn]
528 if (in == NULL || ln == NULL) {
537 * new intermediate node.
539 * be off since that an address that chooses one of
540 * the branches would not match less specific routes
541 * in the other branch
548 atomic_inc(&in->leaf->rt6i_ref);
550 in->fn_sernum = sernum;
552 /* update parent pointer */
563 ln->fn_sernum = sernum;
565 if (addr_bit_set(addr, bit)) {
572 } else { /* plen <= bit */
575 * (new leaf node)[ln]
577 * (old node)[fn] NULL
589 ln->fn_sernum = sernum;
596 if (addr_bit_set(&key->addr, plen))
607 * Insert routing information in a node.
610 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
611 struct nl_info *info)
613 struct rt6_info *iter = NULL;
614 struct rt6_info **ins;
618 for (iter = fn->leaf; iter; iter=iter->u.dst.rt6_next) {
620 * Search for duplicates
623 if (iter->rt6i_metric == rt->rt6i_metric) {
625 * Same priority level
628 if (iter->rt6i_dev == rt->rt6i_dev &&
629 iter->rt6i_idev == rt->rt6i_idev &&
630 ipv6_addr_equal(&iter->rt6i_gateway,
631 &rt->rt6i_gateway)) {
632 if (!(iter->rt6i_flags&RTF_EXPIRES))
634 iter->rt6i_expires = rt->rt6i_expires;
635 if (!(rt->rt6i_flags&RTF_EXPIRES)) {
636 iter->rt6i_flags &= ~RTF_EXPIRES;
637 iter->rt6i_expires = 0;
643 if (iter->rt6i_metric > rt->rt6i_metric)
646 ins = &iter->u.dst.rt6_next;
649 /* Reset round-robin state, if necessary */
650 if (ins == &fn->leaf)
657 rt->u.dst.rt6_next = iter;
660 atomic_inc(&rt->rt6i_ref);
661 inet6_rt_notify(RTM_NEWROUTE, rt, info);
662 info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
664 if ((fn->fn_flags & RTN_RTINFO) == 0) {
665 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
666 fn->fn_flags |= RTN_RTINFO;
672 static __inline__ void fib6_start_gc(struct net *net, struct rt6_info *rt)
674 if (!timer_pending(&net->ipv6.ip6_fib_timer) &&
675 (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE)))
676 mod_timer(&net->ipv6.ip6_fib_timer,
677 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
680 void fib6_force_start_gc(struct net *net)
682 if (!timer_pending(&net->ipv6.ip6_fib_timer))
683 mod_timer(&net->ipv6.ip6_fib_timer,
684 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
688 * Add routing information to the routing tree.
689 * <destination addr>/<source addr>
690 * with source addr info in sub-trees
693 int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info)
695 struct fib6_node *fn, *pn = NULL;
698 fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr),
699 rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst));
706 #ifdef CONFIG_IPV6_SUBTREES
707 if (rt->rt6i_src.plen) {
708 struct fib6_node *sn;
710 if (fn->subtree == NULL) {
711 struct fib6_node *sfn;
723 /* Create subtree root node */
728 sfn->leaf = info->nl_net->ipv6.ip6_null_entry;
729 atomic_inc(&info->nl_net->ipv6.ip6_null_entry->rt6i_ref);
730 sfn->fn_flags = RTN_ROOT;
731 sfn->fn_sernum = fib6_new_sernum();
733 /* Now add the first leaf node to new subtree */
735 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
736 sizeof(struct in6_addr), rt->rt6i_src.plen,
737 offsetof(struct rt6_info, rt6i_src));
740 /* If it is failed, discard just allocated
741 root, and then (in st_failure) stale node
748 /* Now link new subtree to main tree */
752 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
753 sizeof(struct in6_addr), rt->rt6i_src.plen,
754 offsetof(struct rt6_info, rt6i_src));
760 if (fn->leaf == NULL) {
762 atomic_inc(&rt->rt6i_ref);
768 err = fib6_add_rt2node(fn, rt, info);
771 fib6_start_gc(info->nl_net, rt);
772 if (!(rt->rt6i_flags&RTF_CACHE))
773 fib6_prune_clones(info->nl_net, pn, rt);
778 #ifdef CONFIG_IPV6_SUBTREES
780 * If fib6_add_1 has cleared the old leaf pointer in the
781 * super-tree leaf node we have to find a new one for it.
783 if (pn != fn && pn->leaf == rt) {
785 atomic_dec(&rt->rt6i_ref);
787 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
788 pn->leaf = fib6_find_prefix(info->nl_net, pn);
791 WARN_ON(pn->leaf == NULL);
792 pn->leaf = info->nl_net->ipv6.ip6_null_entry;
795 atomic_inc(&pn->leaf->rt6i_ref);
798 dst_free(&rt->u.dst);
802 #ifdef CONFIG_IPV6_SUBTREES
803 /* Subtree creation failed, probably main tree node
804 is orphan. If it is, shoot it.
807 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
808 fib6_repair_tree(info->nl_net, fn);
809 dst_free(&rt->u.dst);
815 * Routing tree lookup
820 int offset; /* key offset on rt6_info */
821 struct in6_addr *addr; /* search key */
824 static struct fib6_node * fib6_lookup_1(struct fib6_node *root,
825 struct lookup_args *args)
827 struct fib6_node *fn;
830 if (unlikely(args->offset == 0))
840 struct fib6_node *next;
842 dir = addr_bit_set(args->addr, fn->fn_bit);
844 next = dir ? fn->right : fn->left;
855 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
858 key = (struct rt6key *) ((u8 *) fn->leaf +
861 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
862 #ifdef CONFIG_IPV6_SUBTREES
864 fn = fib6_lookup_1(fn->subtree, args + 1);
866 if (!fn || fn->fn_flags & RTN_RTINFO)
871 if (fn->fn_flags & RTN_ROOT)
880 struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr,
881 struct in6_addr *saddr)
883 struct fib6_node *fn;
884 struct lookup_args args[] = {
886 .offset = offsetof(struct rt6_info, rt6i_dst),
889 #ifdef CONFIG_IPV6_SUBTREES
891 .offset = offsetof(struct rt6_info, rt6i_src),
896 .offset = 0, /* sentinel */
900 fn = fib6_lookup_1(root, daddr ? args : args + 1);
902 if (fn == NULL || fn->fn_flags & RTN_TL_ROOT)
909 * Get node with specified destination prefix (and source prefix,
910 * if subtrees are used)
914 static struct fib6_node * fib6_locate_1(struct fib6_node *root,
915 struct in6_addr *addr,
916 int plen, int offset)
918 struct fib6_node *fn;
920 for (fn = root; fn ; ) {
921 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
926 if (plen < fn->fn_bit ||
927 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
930 if (plen == fn->fn_bit)
934 * We have more bits to go
936 if (addr_bit_set(addr, fn->fn_bit))
944 struct fib6_node * fib6_locate(struct fib6_node *root,
945 struct in6_addr *daddr, int dst_len,
946 struct in6_addr *saddr, int src_len)
948 struct fib6_node *fn;
950 fn = fib6_locate_1(root, daddr, dst_len,
951 offsetof(struct rt6_info, rt6i_dst));
953 #ifdef CONFIG_IPV6_SUBTREES
955 WARN_ON(saddr == NULL);
956 if (fn && fn->subtree)
957 fn = fib6_locate_1(fn->subtree, saddr, src_len,
958 offsetof(struct rt6_info, rt6i_src));
962 if (fn && fn->fn_flags&RTN_RTINFO)
974 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn)
976 if (fn->fn_flags&RTN_ROOT)
977 return net->ipv6.ip6_null_entry;
981 return fn->left->leaf;
984 return fn->right->leaf;
986 fn = FIB6_SUBTREE(fn);
992 * Called to trim the tree of intermediate nodes when possible. "fn"
993 * is the node we want to try and remove.
996 static struct fib6_node *fib6_repair_tree(struct net *net,
997 struct fib6_node *fn)
1001 struct fib6_node *child, *pn;
1002 struct fib6_walker_t *w;
1006 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1009 WARN_ON(fn->fn_flags & RTN_RTINFO);
1010 WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1011 WARN_ON(fn->leaf != NULL);
1015 if (fn->right) child = fn->right, children |= 1;
1016 if (fn->left) child = fn->left, children |= 2;
1018 if (children == 3 || FIB6_SUBTREE(fn)
1019 #ifdef CONFIG_IPV6_SUBTREES
1020 /* Subtree root (i.e. fn) may have one child */
1021 || (children && fn->fn_flags&RTN_ROOT)
1024 fn->leaf = fib6_find_prefix(net, fn);
1026 if (fn->leaf==NULL) {
1028 fn->leaf = net->ipv6.ip6_null_entry;
1031 atomic_inc(&fn->leaf->rt6i_ref);
1036 #ifdef CONFIG_IPV6_SUBTREES
1037 if (FIB6_SUBTREE(pn) == fn) {
1038 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1039 FIB6_SUBTREE(pn) = NULL;
1042 WARN_ON(fn->fn_flags & RTN_ROOT);
1044 if (pn->right == fn) pn->right = child;
1045 else if (pn->left == fn) pn->left = child;
1053 #ifdef CONFIG_IPV6_SUBTREES
1057 read_lock(&fib6_walker_lock);
1059 if (child == NULL) {
1060 if (w->root == fn) {
1061 w->root = w->node = NULL;
1062 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1063 } else if (w->node == fn) {
1064 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1069 if (w->root == fn) {
1071 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1073 if (w->node == fn) {
1076 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1077 w->state = w->state>=FWS_R ? FWS_U : FWS_INIT;
1079 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1080 w->state = w->state>=FWS_C ? FWS_U : FWS_INIT;
1085 read_unlock(&fib6_walker_lock);
1088 if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn))
1091 rt6_release(pn->leaf);
1097 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1098 struct nl_info *info)
1100 struct fib6_walker_t *w;
1101 struct rt6_info *rt = *rtp;
1102 struct net *net = info->nl_net;
1104 RT6_TRACE("fib6_del_route\n");
1107 *rtp = rt->u.dst.rt6_next;
1108 rt->rt6i_node = NULL;
1109 net->ipv6.rt6_stats->fib_rt_entries--;
1110 net->ipv6.rt6_stats->fib_discarded_routes++;
1112 /* Reset round-robin state, if necessary */
1113 if (fn->rr_ptr == rt)
1116 /* Adjust walkers */
1117 read_lock(&fib6_walker_lock);
1119 if (w->state == FWS_C && w->leaf == rt) {
1120 RT6_TRACE("walker %p adjusted by delroute\n", w);
1121 w->leaf = rt->u.dst.rt6_next;
1122 if (w->leaf == NULL)
1126 read_unlock(&fib6_walker_lock);
1128 rt->u.dst.rt6_next = NULL;
1130 /* If it was last route, expunge its radix tree node */
1131 if (fn->leaf == NULL) {
1132 fn->fn_flags &= ~RTN_RTINFO;
1133 net->ipv6.rt6_stats->fib_route_nodes--;
1134 fn = fib6_repair_tree(net, fn);
1137 if (atomic_read(&rt->rt6i_ref) != 1) {
1138 /* This route is used as dummy address holder in some split
1139 * nodes. It is not leaked, but it still holds other resources,
1140 * which must be released in time. So, scan ascendant nodes
1141 * and replace dummy references to this route with references
1142 * to still alive ones.
1145 if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) {
1146 fn->leaf = fib6_find_prefix(net, fn);
1147 atomic_inc(&fn->leaf->rt6i_ref);
1152 /* No more references are possible at this point. */
1153 BUG_ON(atomic_read(&rt->rt6i_ref) != 1);
1156 inet6_rt_notify(RTM_DELROUTE, rt, info);
1160 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1162 struct net *net = info->nl_net;
1163 struct fib6_node *fn = rt->rt6i_node;
1164 struct rt6_info **rtp;
1167 if (rt->u.dst.obsolete>0) {
1168 WARN_ON(fn != NULL);
1172 if (fn == NULL || rt == net->ipv6.ip6_null_entry)
1175 WARN_ON(!(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(info->nl_net, 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) {
1269 if (w->count < w->skip) {
1287 #ifdef CONFIG_IPV6_SUBTREES
1288 if (FIB6_SUBTREE(pn) == fn) {
1289 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1294 if (pn->left == fn) {
1298 if (pn->right == fn) {
1300 w->leaf = w->node->leaf;
1310 static int fib6_walk(struct fib6_walker_t *w)
1314 w->state = FWS_INIT;
1317 fib6_walker_link(w);
1318 res = fib6_walk_continue(w);
1320 fib6_walker_unlink(w);
1324 static int fib6_clean_node(struct fib6_walker_t *w)
1327 struct rt6_info *rt;
1328 struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w);
1329 struct nl_info info = {
1333 for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) {
1334 res = c->func(rt, c->arg);
1337 res = fib6_del(rt, &info);
1340 printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res);
1353 * Convenient frontend to tree walker.
1355 * func is called on each route.
1356 * It may return -1 -> delete this route.
1357 * 0 -> continue walking
1359 * prune==1 -> only immediate children of node (certainly,
1360 * ignoring pure split nodes) will be scanned.
1363 static void fib6_clean_tree(struct net *net, struct fib6_node *root,
1364 int (*func)(struct rt6_info *, void *arg),
1365 int prune, void *arg)
1367 struct fib6_cleaner_t c;
1370 c.w.func = fib6_clean_node;
1381 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg),
1382 int prune, void *arg)
1384 struct fib6_table *table;
1385 struct hlist_node *node;
1386 struct hlist_head *head;
1390 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
1391 head = &net->ipv6.fib_table_hash[h];
1392 hlist_for_each_entry_rcu(table, node, head, tb6_hlist) {
1393 write_lock_bh(&table->tb6_lock);
1394 fib6_clean_tree(net, &table->tb6_root,
1396 write_unlock_bh(&table->tb6_lock);
1402 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1404 if (rt->rt6i_flags & RTF_CACHE) {
1405 RT6_TRACE("pruning clone %p\n", rt);
1412 static void fib6_prune_clones(struct net *net, struct fib6_node *fn,
1413 struct rt6_info *rt)
1415 fib6_clean_tree(net, fn, fib6_prune_clone, 1, rt);
1419 * Garbage collection
1422 static struct fib6_gc_args
1428 static int fib6_age(struct rt6_info *rt, void *arg)
1430 unsigned long now = jiffies;
1433 * check addrconf expiration here.
1434 * Routes are expired even if they are in use.
1436 * Also age clones. Note, that clones are aged out
1437 * only if they are not in use now.
1440 if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) {
1441 if (time_after(now, rt->rt6i_expires)) {
1442 RT6_TRACE("expiring %p\n", rt);
1446 } else if (rt->rt6i_flags & RTF_CACHE) {
1447 if (atomic_read(&rt->u.dst.__refcnt) == 0 &&
1448 time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) {
1449 RT6_TRACE("aging clone %p\n", rt);
1451 } else if ((rt->rt6i_flags & RTF_GATEWAY) &&
1452 (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) {
1453 RT6_TRACE("purging route %p via non-router but gateway\n",
1463 static DEFINE_SPINLOCK(fib6_gc_lock);
1465 void fib6_run_gc(unsigned long expires, struct net *net)
1467 if (expires != ~0UL) {
1468 spin_lock_bh(&fib6_gc_lock);
1469 gc_args.timeout = expires ? (int)expires :
1470 net->ipv6.sysctl.ip6_rt_gc_interval;
1472 if (!spin_trylock_bh(&fib6_gc_lock)) {
1473 mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
1476 gc_args.timeout = net->ipv6.sysctl.ip6_rt_gc_interval;
1479 gc_args.more = icmp6_dst_gc();
1481 fib6_clean_all(net, fib6_age, 0, NULL);
1484 mod_timer(&net->ipv6.ip6_fib_timer,
1485 round_jiffies(jiffies
1486 + net->ipv6.sysctl.ip6_rt_gc_interval));
1488 del_timer(&net->ipv6.ip6_fib_timer);
1489 spin_unlock_bh(&fib6_gc_lock);
1492 static void fib6_gc_timer_cb(unsigned long arg)
1494 fib6_run_gc(0, (struct net *)arg);
1497 static int __net_init fib6_net_init(struct net *net)
1499 setup_timer(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, (unsigned long)net);
1501 net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL);
1502 if (!net->ipv6.rt6_stats)
1505 net->ipv6.fib_table_hash = kcalloc(FIB6_TABLE_HASHSZ,
1506 sizeof(*net->ipv6.fib_table_hash),
1508 if (!net->ipv6.fib_table_hash)
1511 net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
1513 if (!net->ipv6.fib6_main_tbl)
1514 goto out_fib_table_hash;
1516 net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
1517 net->ipv6.fib6_main_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1518 net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
1519 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1521 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1522 net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
1524 if (!net->ipv6.fib6_local_tbl)
1525 goto out_fib6_main_tbl;
1526 net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
1527 net->ipv6.fib6_local_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1528 net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
1529 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1531 fib6_tables_init(net);
1535 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1537 kfree(net->ipv6.fib6_main_tbl);
1540 kfree(net->ipv6.fib_table_hash);
1542 kfree(net->ipv6.rt6_stats);
1547 static void fib6_net_exit(struct net *net)
1549 rt6_ifdown(net, NULL);
1550 del_timer_sync(&net->ipv6.ip6_fib_timer);
1552 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1553 kfree(net->ipv6.fib6_local_tbl);
1555 kfree(net->ipv6.fib6_main_tbl);
1556 kfree(net->ipv6.fib_table_hash);
1557 kfree(net->ipv6.rt6_stats);
1560 static struct pernet_operations fib6_net_ops = {
1561 .init = fib6_net_init,
1562 .exit = fib6_net_exit,
1565 int __init fib6_init(void)
1569 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1570 sizeof(struct fib6_node),
1571 0, SLAB_HWCACHE_ALIGN,
1573 if (!fib6_node_kmem)
1576 ret = register_pernet_subsys(&fib6_net_ops);
1578 goto out_kmem_cache_create;
1580 ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib);
1582 goto out_unregister_subsys;
1586 out_unregister_subsys:
1587 unregister_pernet_subsys(&fib6_net_ops);
1588 out_kmem_cache_create:
1589 kmem_cache_destroy(fib6_node_kmem);
1593 void fib6_gc_cleanup(void)
1595 unregister_pernet_subsys(&fib6_net_ops);
1596 kmem_cache_destroy(fib6_node_kmem);