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 kmem_cache_t * 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__ int addr_bit_set(void *token, int fn_bit)
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, SLAB_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,
172 .tb6_lock = RW_LOCK_UNLOCKED,
174 .leaf = &ip6_null_entry,
175 .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO,
179 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
180 #define FIB_TABLE_HASHSZ 256
182 #define FIB_TABLE_HASHSZ 1
184 static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ];
186 static void fib6_link_table(struct fib6_table *tb)
190 h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1);
193 * No protection necessary, this is the only list mutatation
194 * operation, tables never disappear once they exist.
196 hlist_add_head_rcu(&tb->tb6_hlist, &fib_table_hash[h]);
199 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
200 static struct fib6_table fib6_local_tbl = {
201 .tb6_id = RT6_TABLE_LOCAL,
202 .tb6_lock = RW_LOCK_UNLOCKED,
204 .leaf = &ip6_null_entry,
205 .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO,
209 static struct fib6_table *fib6_alloc_table(u32 id)
211 struct fib6_table *table;
213 table = kzalloc(sizeof(*table), GFP_ATOMIC);
216 table->tb6_lock = RW_LOCK_UNLOCKED;
217 table->tb6_root.leaf = &ip6_null_entry;
218 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
224 struct fib6_table *fib6_new_table(u32 id)
226 struct fib6_table *tb;
230 tb = fib6_get_table(id);
234 tb = fib6_alloc_table(id);
241 struct fib6_table *fib6_get_table(u32 id)
243 struct fib6_table *tb;
244 struct hlist_node *node;
249 h = id & (FIB_TABLE_HASHSZ - 1);
251 hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb6_hlist) {
252 if (tb->tb6_id == id) {
262 static void __init fib6_tables_init(void)
264 fib6_link_table(&fib6_main_tbl);
265 fib6_link_table(&fib6_local_tbl);
270 struct fib6_table *fib6_new_table(u32 id)
272 return fib6_get_table(id);
275 struct fib6_table *fib6_get_table(u32 id)
277 return &fib6_main_tbl;
280 struct dst_entry *fib6_rule_lookup(struct flowi *fl, int flags,
283 return (struct dst_entry *) lookup(&fib6_main_tbl, fl, flags);
286 static void __init fib6_tables_init(void)
288 fib6_link_table(&fib6_main_tbl);
293 static int fib6_dump_node(struct fib6_walker_t *w)
298 for (rt = w->leaf; rt; rt = rt->u.next) {
299 res = rt6_dump_route(rt, w->args);
301 /* Frame is full, suspend walking */
311 static void fib6_dump_end(struct netlink_callback *cb)
313 struct fib6_walker_t *w = (void*)cb->args[2];
319 cb->done = (void*)cb->args[3];
323 static int fib6_dump_done(struct netlink_callback *cb)
326 return cb->done ? cb->done(cb) : 0;
329 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
330 struct netlink_callback *cb)
332 struct fib6_walker_t *w;
335 w = (void *)cb->args[2];
336 w->root = &table->tb6_root;
338 if (cb->args[4] == 0) {
339 read_lock_bh(&table->tb6_lock);
341 read_unlock_bh(&table->tb6_lock);
345 read_lock_bh(&table->tb6_lock);
346 res = fib6_walk_continue(w);
347 read_unlock_bh(&table->tb6_lock);
350 fib6_walker_unlink(w);
353 fib6_walker_unlink(w);
360 int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
363 unsigned int e = 0, s_e;
364 struct rt6_rtnl_dump_arg arg;
365 struct fib6_walker_t *w;
366 struct fib6_table *tb;
367 struct hlist_node *node;
373 w = (void *)cb->args[2];
377 * 1. hook callback destructor.
379 cb->args[3] = (long)cb->done;
380 cb->done = fib6_dump_done;
383 * 2. allocate and initialize walker.
385 w = kzalloc(sizeof(*w), GFP_ATOMIC);
388 w->func = fib6_dump_node;
389 cb->args[2] = (long)w;
396 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
398 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb6_hlist) {
401 res = fib6_dump_table(tb, skb, cb);
412 res = res < 0 ? res : skb->len;
421 * return the appropriate node for a routing tree "add" operation
422 * by either creating and inserting or by returning an existing
426 static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr,
427 int addrlen, int plen,
430 struct fib6_node *fn, *in, *ln;
431 struct fib6_node *pn = NULL;
435 __u32 sernum = fib6_new_sernum();
437 RT6_TRACE("fib6_add_1\n");
439 /* insert node in tree */
444 key = (struct rt6key *)((u8 *)fn->leaf + offset);
449 if (plen < fn->fn_bit ||
450 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
457 if (plen == fn->fn_bit) {
458 /* clean up an intermediate node */
459 if ((fn->fn_flags & RTN_RTINFO) == 0) {
460 rt6_release(fn->leaf);
464 fn->fn_sernum = sernum;
470 * We have more bits to go
473 /* Try to walk down on tree. */
474 fn->fn_sernum = sernum;
475 dir = addr_bit_set(addr, fn->fn_bit);
477 fn = dir ? fn->right: fn->left;
481 * We walked to the bottom of tree.
482 * Create new leaf node without children.
492 ln->fn_sernum = sernum;
504 * split since we don't have a common prefix anymore or
505 * we have a less significant route.
506 * we've to insert an intermediate node on the list
507 * this new node will point to the one we need to create
513 /* find 1st bit in difference between the 2 addrs.
515 See comment in __ipv6_addr_diff: bit may be an invalid value,
516 but if it is >= plen, the value is ignored in any case.
519 bit = __ipv6_addr_diff(addr, &key->addr, addrlen);
524 * (new leaf node)[ln] (old node)[fn]
530 if (in == NULL || ln == NULL) {
539 * new intermediate node.
541 * be off since that an address that chooses one of
542 * the branches would not match less specific routes
543 * in the other branch
550 atomic_inc(&in->leaf->rt6i_ref);
552 in->fn_sernum = sernum;
554 /* update parent pointer */
565 ln->fn_sernum = sernum;
567 if (addr_bit_set(addr, bit)) {
574 } else { /* plen <= bit */
577 * (new leaf node)[ln]
579 * (old node)[fn] NULL
591 ln->fn_sernum = sernum;
598 if (addr_bit_set(&key->addr, plen))
609 * Insert routing information in a node.
612 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
613 struct nl_info *info)
615 struct rt6_info *iter = NULL;
616 struct rt6_info **ins;
620 if (fn->fn_flags&RTN_TL_ROOT &&
621 fn->leaf == &ip6_null_entry &&
622 !(rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ){
628 for (iter = fn->leaf; iter; iter=iter->u.next) {
630 * Search for duplicates
633 if (iter->rt6i_metric == rt->rt6i_metric) {
635 * Same priority level
638 if (iter->rt6i_dev == rt->rt6i_dev &&
639 iter->rt6i_idev == rt->rt6i_idev &&
640 ipv6_addr_equal(&iter->rt6i_gateway,
641 &rt->rt6i_gateway)) {
642 if (!(iter->rt6i_flags&RTF_EXPIRES))
644 iter->rt6i_expires = rt->rt6i_expires;
645 if (!(rt->rt6i_flags&RTF_EXPIRES)) {
646 iter->rt6i_flags &= ~RTF_EXPIRES;
647 iter->rt6i_expires = 0;
653 if (iter->rt6i_metric > rt->rt6i_metric)
667 atomic_inc(&rt->rt6i_ref);
668 inet6_rt_notify(RTM_NEWROUTE, rt, info);
669 rt6_stats.fib_rt_entries++;
671 if ((fn->fn_flags & RTN_RTINFO) == 0) {
672 rt6_stats.fib_route_nodes++;
673 fn->fn_flags |= RTN_RTINFO;
679 static __inline__ void fib6_start_gc(struct rt6_info *rt)
681 if (ip6_fib_timer.expires == 0 &&
682 (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE)))
683 mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval);
686 void fib6_force_start_gc(void)
688 if (ip6_fib_timer.expires == 0)
689 mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval);
693 * Add routing information to the routing tree.
694 * <destination addr>/<source addr>
695 * with source addr info in sub-trees
698 int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info)
700 struct fib6_node *fn, *pn = NULL;
703 fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr),
704 rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst));
711 #ifdef CONFIG_IPV6_SUBTREES
712 if (rt->rt6i_src.plen) {
713 struct fib6_node *sn;
715 if (fn->subtree == NULL) {
716 struct fib6_node *sfn;
728 /* Create subtree root node */
733 sfn->leaf = &ip6_null_entry;
734 atomic_inc(&ip6_null_entry.rt6i_ref);
735 sfn->fn_flags = RTN_ROOT;
736 sfn->fn_sernum = fib6_new_sernum();
738 /* Now add the first leaf node to new subtree */
740 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
741 sizeof(struct in6_addr), rt->rt6i_src.plen,
742 offsetof(struct rt6_info, rt6i_src));
745 /* If it is failed, discard just allocated
746 root, and then (in st_failure) stale node
753 /* Now link new subtree to main tree */
757 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
758 sizeof(struct in6_addr), rt->rt6i_src.plen,
759 offsetof(struct rt6_info, rt6i_src));
765 if (fn->leaf == NULL) {
767 atomic_inc(&rt->rt6i_ref);
773 err = fib6_add_rt2node(fn, rt, info);
777 if (!(rt->rt6i_flags&RTF_CACHE))
778 fib6_prune_clones(pn, rt);
783 #ifdef CONFIG_IPV6_SUBTREES
785 * If fib6_add_1 has cleared the old leaf pointer in the
786 * super-tree leaf node we have to find a new one for it.
788 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
789 pn->leaf = fib6_find_prefix(pn);
792 BUG_TRAP(pn->leaf != NULL);
793 pn->leaf = &ip6_null_entry;
796 atomic_inc(&pn->leaf->rt6i_ref);
799 dst_free(&rt->u.dst);
803 #ifdef CONFIG_IPV6_SUBTREES
804 /* Subtree creation failed, probably main tree node
805 is orphan. If it is, shoot it.
808 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
809 fib6_repair_tree(fn);
810 dst_free(&rt->u.dst);
816 * Routing tree lookup
821 int offset; /* key offset on rt6_info */
822 struct in6_addr *addr; /* search key */
825 static struct fib6_node * fib6_lookup_1(struct fib6_node *root,
826 struct lookup_args *args)
828 struct fib6_node *fn;
831 if (unlikely(args->offset == 0))
841 struct fib6_node *next;
843 dir = addr_bit_set(args->addr, fn->fn_bit);
845 next = dir ? fn->right : fn->left;
856 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
859 key = (struct rt6key *) ((u8 *) fn->leaf +
862 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
863 #ifdef CONFIG_IPV6_SUBTREES
865 fn = fib6_lookup_1(fn->subtree, args + 1);
867 if (!fn || fn->fn_flags & RTN_RTINFO)
872 if (fn->fn_flags & RTN_ROOT)
881 struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr,
882 struct in6_addr *saddr)
884 struct fib6_node *fn;
885 struct lookup_args args[] = {
887 .offset = offsetof(struct rt6_info, rt6i_dst),
890 #ifdef CONFIG_IPV6_SUBTREES
892 .offset = offsetof(struct rt6_info, rt6i_src),
897 .offset = 0, /* sentinel */
901 fn = fib6_lookup_1(root, daddr ? args : args + 1);
903 if (fn == NULL || fn->fn_flags & RTN_TL_ROOT)
910 * Get node with specified destination prefix (and source prefix,
911 * if subtrees are used)
915 static struct fib6_node * fib6_locate_1(struct fib6_node *root,
916 struct in6_addr *addr,
917 int plen, int offset)
919 struct fib6_node *fn;
921 for (fn = root; fn ; ) {
922 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
927 if (plen < fn->fn_bit ||
928 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
931 if (plen == fn->fn_bit)
935 * We have more bits to go
937 if (addr_bit_set(addr, fn->fn_bit))
945 struct fib6_node * fib6_locate(struct fib6_node *root,
946 struct in6_addr *daddr, int dst_len,
947 struct in6_addr *saddr, int src_len)
949 struct fib6_node *fn;
951 fn = fib6_locate_1(root, daddr, dst_len,
952 offsetof(struct rt6_info, rt6i_dst));
954 #ifdef CONFIG_IPV6_SUBTREES
956 BUG_TRAP(saddr!=NULL);
957 if (fn && fn->subtree)
958 fn = fib6_locate_1(fn->subtree, saddr, src_len,
959 offsetof(struct rt6_info, rt6i_src));
963 if (fn && fn->fn_flags&RTN_RTINFO)
975 static struct rt6_info * fib6_find_prefix(struct fib6_node *fn)
977 if (fn->fn_flags&RTN_ROOT)
978 return &ip6_null_entry;
982 return fn->left->leaf;
985 return fn->right->leaf;
987 fn = FIB6_SUBTREE(fn);
993 * Called to trim the tree of intermediate nodes when possible. "fn"
994 * is the node we want to try and remove.
997 static struct fib6_node * fib6_repair_tree(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 BUG_TRAP(!(fn->fn_flags&RTN_RTINFO));
1010 BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT));
1011 BUG_TRAP(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(fn);
1026 if (fn->leaf==NULL) {
1028 fn->leaf = &ip6_null_entry;
1031 atomic_inc(&fn->leaf->rt6i_ref);
1036 #ifdef CONFIG_IPV6_SUBTREES
1037 if (FIB6_SUBTREE(pn) == fn) {
1038 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1039 FIB6_SUBTREE(pn) = NULL;
1042 BUG_TRAP(!(fn->fn_flags&RTN_ROOT));
1044 if (pn->right == fn) pn->right = child;
1045 else if (pn->left == fn) pn->left = child;
1052 #ifdef CONFIG_IPV6_SUBTREES
1056 read_lock(&fib6_walker_lock);
1058 if (child == NULL) {
1059 if (w->root == fn) {
1060 w->root = w->node = NULL;
1061 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1062 } else if (w->node == fn) {
1063 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1068 if (w->root == fn) {
1070 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1072 if (w->node == fn) {
1075 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1076 w->state = w->state>=FWS_R ? FWS_U : FWS_INIT;
1078 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1079 w->state = w->state>=FWS_C ? FWS_U : FWS_INIT;
1084 read_unlock(&fib6_walker_lock);
1087 if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn))
1090 rt6_release(pn->leaf);
1096 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1097 struct nl_info *info)
1099 struct fib6_walker_t *w;
1100 struct rt6_info *rt = *rtp;
1102 RT6_TRACE("fib6_del_route\n");
1106 rt->rt6i_node = NULL;
1107 rt6_stats.fib_rt_entries--;
1108 rt6_stats.fib_discarded_routes++;
1110 /* Adjust walkers */
1111 read_lock(&fib6_walker_lock);
1113 if (w->state == FWS_C && w->leaf == rt) {
1114 RT6_TRACE("walker %p adjusted by delroute\n", w);
1115 w->leaf = rt->u.next;
1116 if (w->leaf == NULL)
1120 read_unlock(&fib6_walker_lock);
1124 if (fn->leaf == NULL && fn->fn_flags&RTN_TL_ROOT)
1125 fn->leaf = &ip6_null_entry;
1127 /* If it was last route, expunge its radix tree node */
1128 if (fn->leaf == NULL) {
1129 fn->fn_flags &= ~RTN_RTINFO;
1130 rt6_stats.fib_route_nodes--;
1131 fn = fib6_repair_tree(fn);
1134 if (atomic_read(&rt->rt6i_ref) != 1) {
1135 /* This route is used as dummy address holder in some split
1136 * nodes. It is not leaked, but it still holds other resources,
1137 * which must be released in time. So, scan ascendant nodes
1138 * and replace dummy references to this route with references
1139 * to still alive ones.
1142 if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) {
1143 fn->leaf = fib6_find_prefix(fn);
1144 atomic_inc(&fn->leaf->rt6i_ref);
1149 /* No more references are possible at this point. */
1150 if (atomic_read(&rt->rt6i_ref) != 1) BUG();
1153 inet6_rt_notify(RTM_DELROUTE, rt, info);
1157 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1159 struct fib6_node *fn = rt->rt6i_node;
1160 struct rt6_info **rtp;
1163 if (rt->u.dst.obsolete>0) {
1168 if (fn == NULL || rt == &ip6_null_entry)
1171 BUG_TRAP(fn->fn_flags&RTN_RTINFO);
1173 if (!(rt->rt6i_flags&RTF_CACHE)) {
1174 struct fib6_node *pn = fn;
1175 #ifdef CONFIG_IPV6_SUBTREES
1176 /* clones of this route might be in another subtree */
1177 if (rt->rt6i_src.plen) {
1178 while (!(pn->fn_flags&RTN_ROOT))
1183 fib6_prune_clones(pn, rt);
1187 * Walk the leaf entries looking for ourself
1190 for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.next) {
1192 fib6_del_route(fn, rtp, info);
1200 * Tree traversal function.
1202 * Certainly, it is not interrupt safe.
1203 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1204 * It means, that we can modify tree during walking
1205 * and use this function for garbage collection, clone pruning,
1206 * cleaning tree when a device goes down etc. etc.
1208 * It guarantees that every node will be traversed,
1209 * and that it will be traversed only once.
1211 * Callback function w->func may return:
1212 * 0 -> continue walking.
1213 * positive value -> walking is suspended (used by tree dumps,
1214 * and probably by gc, if it will be split to several slices)
1215 * negative value -> terminate walking.
1217 * The function itself returns:
1218 * 0 -> walk is complete.
1219 * >0 -> walk is incomplete (i.e. suspended)
1220 * <0 -> walk is terminated by an error.
1223 static int fib6_walk_continue(struct fib6_walker_t *w)
1225 struct fib6_node *fn, *pn;
1232 if (w->prune && fn != w->root &&
1233 fn->fn_flags&RTN_RTINFO && w->state < FWS_C) {
1238 #ifdef CONFIG_IPV6_SUBTREES
1240 if (FIB6_SUBTREE(fn)) {
1241 w->node = FIB6_SUBTREE(fn);
1249 w->state = FWS_INIT;
1255 w->node = fn->right;
1256 w->state = FWS_INIT;
1262 if (w->leaf && fn->fn_flags&RTN_RTINFO) {
1263 int err = w->func(w);
1274 #ifdef CONFIG_IPV6_SUBTREES
1275 if (FIB6_SUBTREE(pn) == fn) {
1276 BUG_TRAP(fn->fn_flags&RTN_ROOT);
1281 if (pn->left == fn) {
1285 if (pn->right == fn) {
1287 w->leaf = w->node->leaf;
1297 static int fib6_walk(struct fib6_walker_t *w)
1301 w->state = FWS_INIT;
1304 fib6_walker_link(w);
1305 res = fib6_walk_continue(w);
1307 fib6_walker_unlink(w);
1311 static int fib6_clean_node(struct fib6_walker_t *w)
1314 struct rt6_info *rt;
1315 struct fib6_cleaner_t *c = (struct fib6_cleaner_t*)w;
1317 for (rt = w->leaf; rt; rt = rt->u.next) {
1318 res = c->func(rt, c->arg);
1321 res = fib6_del(rt, NULL);
1324 printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res);
1337 * Convenient frontend to tree walker.
1339 * func is called on each route.
1340 * It may return -1 -> delete this route.
1341 * 0 -> continue walking
1343 * prune==1 -> only immediate children of node (certainly,
1344 * ignoring pure split nodes) will be scanned.
1347 static void fib6_clean_tree(struct fib6_node *root,
1348 int (*func)(struct rt6_info *, void *arg),
1349 int prune, void *arg)
1351 struct fib6_cleaner_t c;
1354 c.w.func = fib6_clean_node;
1362 void fib6_clean_all(int (*func)(struct rt6_info *, void *arg),
1363 int prune, void *arg)
1365 struct fib6_table *table;
1366 struct hlist_node *node;
1370 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
1371 hlist_for_each_entry_rcu(table, node, &fib_table_hash[h],
1373 write_lock_bh(&table->tb6_lock);
1374 fib6_clean_tree(&table->tb6_root, func, prune, arg);
1375 write_unlock_bh(&table->tb6_lock);
1381 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1383 if (rt->rt6i_flags & RTF_CACHE) {
1384 RT6_TRACE("pruning clone %p\n", rt);
1391 static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt)
1393 fib6_clean_tree(fn, fib6_prune_clone, 1, rt);
1397 * Garbage collection
1400 static struct fib6_gc_args
1406 static int fib6_age(struct rt6_info *rt, void *arg)
1408 unsigned long now = jiffies;
1411 * check addrconf expiration here.
1412 * Routes are expired even if they are in use.
1414 * Also age clones. Note, that clones are aged out
1415 * only if they are not in use now.
1418 if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) {
1419 if (time_after(now, rt->rt6i_expires)) {
1420 RT6_TRACE("expiring %p\n", rt);
1424 } else if (rt->rt6i_flags & RTF_CACHE) {
1425 if (atomic_read(&rt->u.dst.__refcnt) == 0 &&
1426 time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) {
1427 RT6_TRACE("aging clone %p\n", rt);
1429 } else if ((rt->rt6i_flags & RTF_GATEWAY) &&
1430 (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) {
1431 RT6_TRACE("purging route %p via non-router but gateway\n",
1441 static DEFINE_SPINLOCK(fib6_gc_lock);
1443 void fib6_run_gc(unsigned long dummy)
1445 if (dummy != ~0UL) {
1446 spin_lock_bh(&fib6_gc_lock);
1447 gc_args.timeout = dummy ? (int)dummy : ip6_rt_gc_interval;
1450 if (!spin_trylock(&fib6_gc_lock)) {
1451 mod_timer(&ip6_fib_timer, jiffies + HZ);
1455 gc_args.timeout = ip6_rt_gc_interval;
1459 ndisc_dst_gc(&gc_args.more);
1460 fib6_clean_all(fib6_age, 0, NULL);
1463 mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval);
1465 del_timer(&ip6_fib_timer);
1466 ip6_fib_timer.expires = 0;
1468 spin_unlock_bh(&fib6_gc_lock);
1471 void __init fib6_init(void)
1473 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1474 sizeof(struct fib6_node),
1475 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1481 void fib6_gc_cleanup(void)
1483 del_timer(&ip6_fib_timer);
1484 kmem_cache_destroy(fib6_node_kmem);