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.
14 * Yuji SEKIYA @USAGI: Support default route on router node;
15 * remove ip6_null_entry from the top of
17 * Ville Nuorvala: Fixed routing subtrees.
20 #define pr_fmt(fmt) "IPv6: " fmt
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/net.h>
25 #include <linux/route.h>
26 #include <linux/netdevice.h>
27 #include <linux/in6.h>
28 #include <linux/init.h>
29 #include <linux/list.h>
30 #include <linux/slab.h>
33 #include <net/ndisc.h>
34 #include <net/addrconf.h>
35 #include <net/lwtunnel.h>
37 #include <net/ip6_fib.h>
38 #include <net/ip6_route.h>
43 #define RT6_TRACE(x...) pr_debug(x)
45 #define RT6_TRACE(x...) do { ; } while (0)
48 static struct kmem_cache *fib6_node_kmem __read_mostly;
53 int (*func)(struct rt6_info *, void *arg);
58 #ifdef CONFIG_IPV6_SUBTREES
59 #define FWS_INIT FWS_S
61 #define FWS_INIT FWS_L
64 static void fib6_prune_clones(struct net *net, struct fib6_node *fn);
65 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn);
66 static struct fib6_node *fib6_repair_tree(struct net *net, struct fib6_node *fn);
67 static int fib6_walk(struct net *net, struct fib6_walker *w);
68 static int fib6_walk_continue(struct fib6_walker *w);
71 * A routing update causes an increase of the serial number on the
72 * affected subtree. This allows for cached routes to be asynchronously
73 * tested when modifications are made to the destination cache as a
74 * result of redirects, path MTU changes, etc.
77 static void fib6_gc_timer_cb(unsigned long arg);
79 #define FOR_WALKERS(net, w) \
80 list_for_each_entry(w, &(net)->ipv6.fib6_walkers, lh)
82 static void fib6_walker_link(struct net *net, struct fib6_walker *w)
84 write_lock_bh(&net->ipv6.fib6_walker_lock);
85 list_add(&w->lh, &net->ipv6.fib6_walkers);
86 write_unlock_bh(&net->ipv6.fib6_walker_lock);
89 static void fib6_walker_unlink(struct net *net, struct fib6_walker *w)
91 write_lock_bh(&net->ipv6.fib6_walker_lock);
93 write_unlock_bh(&net->ipv6.fib6_walker_lock);
96 static int fib6_new_sernum(struct net *net)
101 old = atomic_read(&net->ipv6.fib6_sernum);
102 new = old < INT_MAX ? old + 1 : 1;
103 } while (atomic_cmpxchg(&net->ipv6.fib6_sernum,
109 FIB6_NO_SERNUM_CHANGE = 0,
113 * Auxiliary address test functions for the radix tree.
115 * These assume a 32bit processor (although it will work on
122 #if defined(__LITTLE_ENDIAN)
123 # define BITOP_BE32_SWIZZLE (0x1F & ~7)
125 # define BITOP_BE32_SWIZZLE 0
128 static __be32 addr_bit_set(const void *token, int fn_bit)
130 const __be32 *addr = token;
133 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
134 * is optimized version of
135 * htonl(1 << ((~fn_bit)&0x1F))
136 * See include/asm-generic/bitops/le.h.
138 return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) &
142 static struct fib6_node *node_alloc(void)
144 struct fib6_node *fn;
146 fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
151 static void node_free_immediate(struct fib6_node *fn)
153 kmem_cache_free(fib6_node_kmem, fn);
156 static void node_free_rcu(struct rcu_head *head)
158 struct fib6_node *fn = container_of(head, struct fib6_node, rcu);
160 kmem_cache_free(fib6_node_kmem, fn);
163 static void node_free(struct fib6_node *fn)
165 call_rcu(&fn->rcu, node_free_rcu);
168 static void rt6_rcu_free(struct rt6_info *rt)
170 call_rcu(&rt->dst.rcu_head, dst_rcu_free);
173 static void rt6_free_pcpu(struct rt6_info *non_pcpu_rt)
177 if (!non_pcpu_rt->rt6i_pcpu)
180 for_each_possible_cpu(cpu) {
181 struct rt6_info **ppcpu_rt;
182 struct rt6_info *pcpu_rt;
184 ppcpu_rt = per_cpu_ptr(non_pcpu_rt->rt6i_pcpu, cpu);
187 rt6_rcu_free(pcpu_rt);
192 free_percpu(non_pcpu_rt->rt6i_pcpu);
193 non_pcpu_rt->rt6i_pcpu = NULL;
196 static void rt6_release(struct rt6_info *rt)
198 if (atomic_dec_and_test(&rt->rt6i_ref)) {
204 static void fib6_free_table(struct fib6_table *table)
206 inetpeer_invalidate_tree(&table->tb6_peers);
210 static void fib6_link_table(struct net *net, struct fib6_table *tb)
215 * Initialize table lock at a single place to give lockdep a key,
216 * tables aren't visible prior to being linked to the list.
218 rwlock_init(&tb->tb6_lock);
220 h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1);
223 * No protection necessary, this is the only list mutatation
224 * operation, tables never disappear once they exist.
226 hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]);
229 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
231 static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
233 struct fib6_table *table;
235 table = kzalloc(sizeof(*table), GFP_ATOMIC);
238 table->tb6_root.leaf = net->ipv6.ip6_null_entry;
239 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
240 inet_peer_base_init(&table->tb6_peers);
246 struct fib6_table *fib6_new_table(struct net *net, u32 id)
248 struct fib6_table *tb;
252 tb = fib6_get_table(net, id);
256 tb = fib6_alloc_table(net, id);
258 fib6_link_table(net, tb);
262 EXPORT_SYMBOL_GPL(fib6_new_table);
264 struct fib6_table *fib6_get_table(struct net *net, u32 id)
266 struct fib6_table *tb;
267 struct hlist_head *head;
272 h = id & (FIB6_TABLE_HASHSZ - 1);
274 head = &net->ipv6.fib_table_hash[h];
275 hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
276 if (tb->tb6_id == id) {
285 EXPORT_SYMBOL_GPL(fib6_get_table);
287 static void __net_init fib6_tables_init(struct net *net)
289 fib6_link_table(net, net->ipv6.fib6_main_tbl);
290 fib6_link_table(net, net->ipv6.fib6_local_tbl);
294 struct fib6_table *fib6_new_table(struct net *net, u32 id)
296 return fib6_get_table(net, id);
299 struct fib6_table *fib6_get_table(struct net *net, u32 id)
301 return net->ipv6.fib6_main_tbl;
304 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
305 int flags, pol_lookup_t lookup)
309 rt = lookup(net, net->ipv6.fib6_main_tbl, fl6, flags);
310 if (rt->dst.error == -EAGAIN) {
312 rt = net->ipv6.ip6_null_entry;
319 static void __net_init fib6_tables_init(struct net *net)
321 fib6_link_table(net, net->ipv6.fib6_main_tbl);
326 static int fib6_dump_node(struct fib6_walker *w)
331 for (rt = w->leaf; rt; rt = rt->dst.rt6_next) {
332 res = rt6_dump_route(rt, w->args);
334 /* Frame is full, suspend walking */
343 static void fib6_dump_end(struct netlink_callback *cb)
345 struct net *net = sock_net(cb->skb->sk);
346 struct fib6_walker *w = (void *)cb->args[2];
351 fib6_walker_unlink(net, w);
356 cb->done = (void *)cb->args[3];
360 static int fib6_dump_done(struct netlink_callback *cb)
363 return cb->done ? cb->done(cb) : 0;
366 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
367 struct netlink_callback *cb)
369 struct net *net = sock_net(skb->sk);
370 struct fib6_walker *w;
373 w = (void *)cb->args[2];
374 w->root = &table->tb6_root;
376 if (cb->args[4] == 0) {
380 read_lock_bh(&table->tb6_lock);
381 res = fib6_walk(net, w);
382 read_unlock_bh(&table->tb6_lock);
385 cb->args[5] = w->root->fn_sernum;
388 if (cb->args[5] != w->root->fn_sernum) {
389 /* Begin at the root if the tree changed */
390 cb->args[5] = w->root->fn_sernum;
397 read_lock_bh(&table->tb6_lock);
398 res = fib6_walk_continue(w);
399 read_unlock_bh(&table->tb6_lock);
401 fib6_walker_unlink(net, w);
409 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
411 struct net *net = sock_net(skb->sk);
413 unsigned int e = 0, s_e;
414 struct rt6_rtnl_dump_arg arg;
415 struct fib6_walker *w;
416 struct fib6_table *tb;
417 struct hlist_head *head;
423 w = (void *)cb->args[2];
427 * 1. hook callback destructor.
429 cb->args[3] = (long)cb->done;
430 cb->done = fib6_dump_done;
433 * 2. allocate and initialize walker.
435 w = kzalloc(sizeof(*w), GFP_ATOMIC);
438 w->func = fib6_dump_node;
439 cb->args[2] = (long)w;
448 for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) {
450 head = &net->ipv6.fib_table_hash[h];
451 hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
454 res = fib6_dump_table(tb, skb, cb);
466 res = res < 0 ? res : skb->len;
475 * return the appropriate node for a routing tree "add" operation
476 * by either creating and inserting or by returning an existing
480 static struct fib6_node *fib6_add_1(struct fib6_node *root,
481 struct in6_addr *addr, int plen,
482 int offset, int allow_create,
483 int replace_required, int sernum)
485 struct fib6_node *fn, *in, *ln;
486 struct fib6_node *pn = NULL;
491 RT6_TRACE("fib6_add_1\n");
493 /* insert node in tree */
498 key = (struct rt6key *)((u8 *)fn->leaf + offset);
503 if (plen < fn->fn_bit ||
504 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) {
506 if (replace_required) {
507 pr_warn("Can't replace route, no match found\n");
508 return ERR_PTR(-ENOENT);
510 pr_warn("NLM_F_CREATE should be set when creating new route\n");
519 if (plen == fn->fn_bit) {
520 /* clean up an intermediate node */
521 if (!(fn->fn_flags & RTN_RTINFO)) {
522 rt6_release(fn->leaf);
526 fn->fn_sernum = sernum;
532 * We have more bits to go
535 /* Try to walk down on tree. */
536 fn->fn_sernum = sernum;
537 dir = addr_bit_set(addr, fn->fn_bit);
539 fn = dir ? fn->right : fn->left;
543 /* We should not create new node because
544 * NLM_F_REPLACE was specified without NLM_F_CREATE
545 * I assume it is safe to require NLM_F_CREATE when
546 * REPLACE flag is used! Later we may want to remove the
547 * check for replace_required, because according
548 * to netlink specification, NLM_F_CREATE
549 * MUST be specified if new route is created.
550 * That would keep IPv6 consistent with IPv4
552 if (replace_required) {
553 pr_warn("Can't replace route, no match found\n");
554 return ERR_PTR(-ENOENT);
556 pr_warn("NLM_F_CREATE should be set when creating new route\n");
559 * We walked to the bottom of tree.
560 * Create new leaf node without children.
566 return ERR_PTR(-ENOMEM);
570 ln->fn_sernum = sernum;
582 * split since we don't have a common prefix anymore or
583 * we have a less significant route.
584 * we've to insert an intermediate node on the list
585 * this new node will point to the one we need to create
591 /* find 1st bit in difference between the 2 addrs.
593 See comment in __ipv6_addr_diff: bit may be an invalid value,
594 but if it is >= plen, the value is ignored in any case.
597 bit = __ipv6_addr_diff(addr, &key->addr, sizeof(*addr));
602 * (new leaf node)[ln] (old node)[fn]
610 node_free_immediate(in);
612 node_free_immediate(ln);
613 return ERR_PTR(-ENOMEM);
617 * new intermediate node.
619 * be off since that an address that chooses one of
620 * the branches would not match less specific routes
621 * in the other branch
628 atomic_inc(&in->leaf->rt6i_ref);
630 in->fn_sernum = sernum;
632 /* update parent pointer */
643 ln->fn_sernum = sernum;
645 if (addr_bit_set(addr, bit)) {
652 } else { /* plen <= bit */
655 * (new leaf node)[ln]
657 * (old node)[fn] NULL
663 return ERR_PTR(-ENOMEM);
669 ln->fn_sernum = sernum;
676 if (addr_bit_set(&key->addr, plen))
686 static bool rt6_qualify_for_ecmp(struct rt6_info *rt)
688 return (rt->rt6i_flags & (RTF_GATEWAY|RTF_ADDRCONF|RTF_DYNAMIC)) ==
692 static void fib6_copy_metrics(u32 *mp, const struct mx6_config *mxc)
696 for (i = 0; i < RTAX_MAX; i++) {
697 if (test_bit(i, mxc->mx_valid))
702 static int fib6_commit_metrics(struct dst_entry *dst, struct mx6_config *mxc)
707 if (dst->flags & DST_HOST) {
708 u32 *mp = dst_metrics_write_ptr(dst);
713 fib6_copy_metrics(mp, mxc);
715 dst_init_metrics(dst, mxc->mx, false);
717 /* We've stolen mx now. */
724 static void fib6_purge_rt(struct rt6_info *rt, struct fib6_node *fn,
727 if (atomic_read(&rt->rt6i_ref) != 1) {
728 /* This route is used as dummy address holder in some split
729 * nodes. It is not leaked, but it still holds other resources,
730 * which must be released in time. So, scan ascendant nodes
731 * and replace dummy references to this route with references
732 * to still alive ones.
735 if (!(fn->fn_flags & RTN_RTINFO) && fn->leaf == rt) {
736 fn->leaf = fib6_find_prefix(net, fn);
737 atomic_inc(&fn->leaf->rt6i_ref);
742 /* No more references are possible at this point. */
743 BUG_ON(atomic_read(&rt->rt6i_ref) != 1);
748 * Insert routing information in a node.
751 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
752 struct nl_info *info, struct mx6_config *mxc)
754 struct rt6_info *iter = NULL;
755 struct rt6_info **ins;
756 struct rt6_info **fallback_ins = NULL;
757 int replace = (info->nlh &&
758 (info->nlh->nlmsg_flags & NLM_F_REPLACE));
759 int add = (!info->nlh ||
760 (info->nlh->nlmsg_flags & NLM_F_CREATE));
762 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
763 u16 nlflags = NLM_F_EXCL;
768 for (iter = fn->leaf; iter; iter = iter->dst.rt6_next) {
770 * Search for duplicates
773 if (iter->rt6i_metric == rt->rt6i_metric) {
775 * Same priority level
778 (info->nlh->nlmsg_flags & NLM_F_EXCL))
781 nlflags &= ~NLM_F_EXCL;
783 if (rt_can_ecmp == rt6_qualify_for_ecmp(iter)) {
787 fallback_ins = fallback_ins ?: ins;
791 if (rt6_duplicate_nexthop(iter, rt)) {
792 if (rt->rt6i_nsiblings)
793 rt->rt6i_nsiblings = 0;
794 if (!(iter->rt6i_flags & RTF_EXPIRES))
796 if (!(rt->rt6i_flags & RTF_EXPIRES))
797 rt6_clean_expires(iter);
799 rt6_set_expires(iter, rt->dst.expires);
800 iter->rt6i_pmtu = rt->rt6i_pmtu;
803 /* If we have the same destination and the same metric,
804 * but not the same gateway, then the route we try to
805 * add is sibling to this route, increment our counter
806 * of siblings, and later we will add our route to the
808 * Only static routes (which don't have flag
809 * RTF_EXPIRES) are used for ECMPv6.
811 * To avoid long list, we only had siblings if the
812 * route have a gateway.
815 rt6_qualify_for_ecmp(iter))
816 rt->rt6i_nsiblings++;
819 if (iter->rt6i_metric > rt->rt6i_metric)
823 ins = &iter->dst.rt6_next;
826 if (fallback_ins && !found) {
827 /* No matching route with same ecmp-able-ness found, replace
828 * first matching route
835 /* Reset round-robin state, if necessary */
836 if (ins == &fn->leaf)
839 /* Link this route to others same route. */
840 if (rt->rt6i_nsiblings) {
841 unsigned int rt6i_nsiblings;
842 struct rt6_info *sibling, *temp_sibling;
844 /* Find the first route that have the same metric */
847 if (sibling->rt6i_metric == rt->rt6i_metric &&
848 rt6_qualify_for_ecmp(sibling)) {
849 list_add_tail(&rt->rt6i_siblings,
850 &sibling->rt6i_siblings);
853 sibling = sibling->dst.rt6_next;
855 /* For each sibling in the list, increment the counter of
856 * siblings. BUG() if counters does not match, list of siblings
860 list_for_each_entry_safe(sibling, temp_sibling,
861 &rt->rt6i_siblings, rt6i_siblings) {
862 sibling->rt6i_nsiblings++;
863 BUG_ON(sibling->rt6i_nsiblings != rt->rt6i_nsiblings);
866 BUG_ON(rt6i_nsiblings != rt->rt6i_nsiblings);
874 pr_warn("NLM_F_CREATE should be set when creating new route\n");
877 nlflags |= NLM_F_CREATE;
878 err = fib6_commit_metrics(&rt->dst, mxc);
882 rt->dst.rt6_next = iter;
884 rcu_assign_pointer(rt->rt6i_node, fn);
885 atomic_inc(&rt->rt6i_ref);
886 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
887 info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
889 if (!(fn->fn_flags & RTN_RTINFO)) {
890 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
891 fn->fn_flags |= RTN_RTINFO;
900 pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
904 err = fib6_commit_metrics(&rt->dst, mxc);
909 rcu_assign_pointer(rt->rt6i_node, fn);
910 rt->dst.rt6_next = iter->dst.rt6_next;
911 atomic_inc(&rt->rt6i_ref);
912 inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE);
913 if (!(fn->fn_flags & RTN_RTINFO)) {
914 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
915 fn->fn_flags |= RTN_RTINFO;
917 nsiblings = iter->rt6i_nsiblings;
918 fib6_purge_rt(iter, fn, info->nl_net);
919 if (fn->rr_ptr == iter)
924 /* Replacing an ECMP route, remove all siblings */
925 ins = &rt->dst.rt6_next;
928 if (iter->rt6i_metric > rt->rt6i_metric)
930 if (rt6_qualify_for_ecmp(iter)) {
931 *ins = iter->dst.rt6_next;
932 fib6_purge_rt(iter, fn, info->nl_net);
933 if (fn->rr_ptr == iter)
938 ins = &iter->dst.rt6_next;
942 WARN_ON(nsiblings != 0);
949 static void fib6_start_gc(struct net *net, struct rt6_info *rt)
951 if (!timer_pending(&net->ipv6.ip6_fib_timer) &&
952 (rt->rt6i_flags & (RTF_EXPIRES | RTF_CACHE)))
953 mod_timer(&net->ipv6.ip6_fib_timer,
954 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
957 void fib6_force_start_gc(struct net *net)
959 if (!timer_pending(&net->ipv6.ip6_fib_timer))
960 mod_timer(&net->ipv6.ip6_fib_timer,
961 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
965 * Add routing information to the routing tree.
966 * <destination addr>/<source addr>
967 * with source addr info in sub-trees
970 int fib6_add(struct fib6_node *root, struct rt6_info *rt,
971 struct nl_info *info, struct mx6_config *mxc)
973 struct fib6_node *fn, *pn = NULL;
975 int allow_create = 1;
976 int replace_required = 0;
977 int sernum = fib6_new_sernum(info->nl_net);
979 if (WARN_ON_ONCE((rt->dst.flags & DST_NOCACHE) &&
980 !atomic_read(&rt->dst.__refcnt)))
984 if (!(info->nlh->nlmsg_flags & NLM_F_CREATE))
986 if (info->nlh->nlmsg_flags & NLM_F_REPLACE)
987 replace_required = 1;
989 if (!allow_create && !replace_required)
990 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
992 fn = fib6_add_1(root, &rt->rt6i_dst.addr, rt->rt6i_dst.plen,
993 offsetof(struct rt6_info, rt6i_dst), allow_create,
994 replace_required, sernum);
1003 #ifdef CONFIG_IPV6_SUBTREES
1004 if (rt->rt6i_src.plen) {
1005 struct fib6_node *sn;
1008 struct fib6_node *sfn;
1020 /* Create subtree root node */
1025 sfn->leaf = info->nl_net->ipv6.ip6_null_entry;
1026 atomic_inc(&info->nl_net->ipv6.ip6_null_entry->rt6i_ref);
1027 sfn->fn_flags = RTN_ROOT;
1028 sfn->fn_sernum = sernum;
1030 /* Now add the first leaf node to new subtree */
1032 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
1034 offsetof(struct rt6_info, rt6i_src),
1035 allow_create, replace_required, sernum);
1038 /* If it is failed, discard just allocated
1039 root, and then (in failure) stale node
1042 node_free_immediate(sfn);
1047 /* Now link new subtree to main tree */
1051 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
1053 offsetof(struct rt6_info, rt6i_src),
1054 allow_create, replace_required, sernum);
1064 atomic_inc(&rt->rt6i_ref);
1070 err = fib6_add_rt2node(fn, rt, info, mxc);
1072 fib6_start_gc(info->nl_net, rt);
1073 if (!(rt->rt6i_flags & RTF_CACHE))
1074 fib6_prune_clones(info->nl_net, pn);
1075 rt->dst.flags &= ~DST_NOCACHE;
1080 #ifdef CONFIG_IPV6_SUBTREES
1082 * If fib6_add_1 has cleared the old leaf pointer in the
1083 * super-tree leaf node we have to find a new one for it.
1085 if (pn != fn && pn->leaf == rt) {
1087 atomic_dec(&rt->rt6i_ref);
1089 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
1090 pn->leaf = fib6_find_prefix(info->nl_net, pn);
1093 WARN_ON(pn->leaf == NULL);
1094 pn->leaf = info->nl_net->ipv6.ip6_null_entry;
1097 atomic_inc(&pn->leaf->rt6i_ref);
1105 /* fn->leaf could be NULL if fn is an intermediate node and we
1106 * failed to add the new route to it in both subtree creation
1107 * failure and fib6_add_rt2node() failure case.
1108 * In both cases, fib6_repair_tree() should be called to fix
1111 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
1112 fib6_repair_tree(info->nl_net, fn);
1113 if (!(rt->dst.flags & DST_NOCACHE))
1119 * Routing tree lookup
1123 struct lookup_args {
1124 int offset; /* key offset on rt6_info */
1125 const struct in6_addr *addr; /* search key */
1128 static struct fib6_node *fib6_lookup_1(struct fib6_node *root,
1129 struct lookup_args *args)
1131 struct fib6_node *fn;
1134 if (unlikely(args->offset == 0))
1144 struct fib6_node *next;
1146 dir = addr_bit_set(args->addr, fn->fn_bit);
1148 next = dir ? fn->right : fn->left;
1158 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
1161 key = (struct rt6key *) ((u8 *) fn->leaf +
1164 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
1165 #ifdef CONFIG_IPV6_SUBTREES
1167 struct fib6_node *sfn;
1168 sfn = fib6_lookup_1(fn->subtree,
1175 if (fn->fn_flags & RTN_RTINFO)
1179 #ifdef CONFIG_IPV6_SUBTREES
1182 if (fn->fn_flags & RTN_ROOT)
1191 struct fib6_node *fib6_lookup(struct fib6_node *root, const struct in6_addr *daddr,
1192 const struct in6_addr *saddr)
1194 struct fib6_node *fn;
1195 struct lookup_args args[] = {
1197 .offset = offsetof(struct rt6_info, rt6i_dst),
1200 #ifdef CONFIG_IPV6_SUBTREES
1202 .offset = offsetof(struct rt6_info, rt6i_src),
1207 .offset = 0, /* sentinel */
1211 fn = fib6_lookup_1(root, daddr ? args : args + 1);
1212 if (!fn || fn->fn_flags & RTN_TL_ROOT)
1219 * Get node with specified destination prefix (and source prefix,
1220 * if subtrees are used)
1224 static struct fib6_node *fib6_locate_1(struct fib6_node *root,
1225 const struct in6_addr *addr,
1226 int plen, int offset)
1228 struct fib6_node *fn;
1230 for (fn = root; fn ; ) {
1231 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
1236 if (plen < fn->fn_bit ||
1237 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
1240 if (plen == fn->fn_bit)
1244 * We have more bits to go
1246 if (addr_bit_set(addr, fn->fn_bit))
1254 struct fib6_node *fib6_locate(struct fib6_node *root,
1255 const struct in6_addr *daddr, int dst_len,
1256 const struct in6_addr *saddr, int src_len)
1258 struct fib6_node *fn;
1260 fn = fib6_locate_1(root, daddr, dst_len,
1261 offsetof(struct rt6_info, rt6i_dst));
1263 #ifdef CONFIG_IPV6_SUBTREES
1265 WARN_ON(saddr == NULL);
1266 if (fn && fn->subtree)
1267 fn = fib6_locate_1(fn->subtree, saddr, src_len,
1268 offsetof(struct rt6_info, rt6i_src));
1272 if (fn && fn->fn_flags & RTN_RTINFO)
1284 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn)
1286 if (fn->fn_flags & RTN_ROOT)
1287 return net->ipv6.ip6_null_entry;
1291 return fn->left->leaf;
1293 return fn->right->leaf;
1295 fn = FIB6_SUBTREE(fn);
1301 * Called to trim the tree of intermediate nodes when possible. "fn"
1302 * is the node we want to try and remove.
1305 static struct fib6_node *fib6_repair_tree(struct net *net,
1306 struct fib6_node *fn)
1310 struct fib6_node *child, *pn;
1311 struct fib6_walker *w;
1315 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1318 WARN_ON(fn->fn_flags & RTN_RTINFO);
1319 WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1325 child = fn->right, children |= 1;
1327 child = fn->left, children |= 2;
1329 if (children == 3 || FIB6_SUBTREE(fn)
1330 #ifdef CONFIG_IPV6_SUBTREES
1331 /* Subtree root (i.e. fn) may have one child */
1332 || (children && fn->fn_flags & RTN_ROOT)
1335 fn->leaf = fib6_find_prefix(net, fn);
1339 fn->leaf = net->ipv6.ip6_null_entry;
1342 atomic_inc(&fn->leaf->rt6i_ref);
1347 #ifdef CONFIG_IPV6_SUBTREES
1348 if (FIB6_SUBTREE(pn) == fn) {
1349 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1350 FIB6_SUBTREE(pn) = NULL;
1353 WARN_ON(fn->fn_flags & RTN_ROOT);
1355 if (pn->right == fn)
1357 else if (pn->left == fn)
1366 #ifdef CONFIG_IPV6_SUBTREES
1370 read_lock(&net->ipv6.fib6_walker_lock);
1371 FOR_WALKERS(net, w) {
1373 if (w->root == fn) {
1374 w->root = w->node = NULL;
1375 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1376 } else if (w->node == fn) {
1377 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1382 if (w->root == fn) {
1384 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1386 if (w->node == fn) {
1389 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1390 w->state = w->state >= FWS_R ? FWS_U : FWS_INIT;
1392 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1393 w->state = w->state >= FWS_C ? FWS_U : FWS_INIT;
1398 read_unlock(&net->ipv6.fib6_walker_lock);
1401 if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn))
1404 rt6_release(pn->leaf);
1410 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1411 struct nl_info *info)
1413 struct fib6_walker *w;
1414 struct rt6_info *rt = *rtp;
1415 struct net *net = info->nl_net;
1417 RT6_TRACE("fib6_del_route\n");
1420 *rtp = rt->dst.rt6_next;
1421 rt->rt6i_node = NULL;
1422 net->ipv6.rt6_stats->fib_rt_entries--;
1423 net->ipv6.rt6_stats->fib_discarded_routes++;
1425 /* Reset round-robin state, if necessary */
1426 if (fn->rr_ptr == rt)
1429 /* Remove this entry from other siblings */
1430 if (rt->rt6i_nsiblings) {
1431 struct rt6_info *sibling, *next_sibling;
1433 list_for_each_entry_safe(sibling, next_sibling,
1434 &rt->rt6i_siblings, rt6i_siblings)
1435 sibling->rt6i_nsiblings--;
1436 rt->rt6i_nsiblings = 0;
1437 list_del_init(&rt->rt6i_siblings);
1440 /* Adjust walkers */
1441 read_lock(&net->ipv6.fib6_walker_lock);
1442 FOR_WALKERS(net, w) {
1443 if (w->state == FWS_C && w->leaf == rt) {
1444 RT6_TRACE("walker %p adjusted by delroute\n", w);
1445 w->leaf = rt->dst.rt6_next;
1450 read_unlock(&net->ipv6.fib6_walker_lock);
1452 rt->dst.rt6_next = NULL;
1454 /* If it was last route, expunge its radix tree node */
1456 fn->fn_flags &= ~RTN_RTINFO;
1457 net->ipv6.rt6_stats->fib_route_nodes--;
1458 fn = fib6_repair_tree(net, fn);
1461 fib6_purge_rt(rt, fn, net);
1463 inet6_rt_notify(RTM_DELROUTE, rt, info, 0);
1467 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1469 struct fib6_node *fn = rcu_dereference_protected(rt->rt6i_node,
1470 lockdep_is_held(&rt->rt6i_table->tb6_lock));
1471 struct net *net = info->nl_net;
1472 struct rt6_info **rtp;
1475 if (rt->dst.obsolete > 0) {
1480 if (!fn || rt == net->ipv6.ip6_null_entry)
1483 WARN_ON(!(fn->fn_flags & RTN_RTINFO));
1485 if (!(rt->rt6i_flags & RTF_CACHE)) {
1486 struct fib6_node *pn = fn;
1487 #ifdef CONFIG_IPV6_SUBTREES
1488 /* clones of this route might be in another subtree */
1489 if (rt->rt6i_src.plen) {
1490 while (!(pn->fn_flags & RTN_ROOT))
1495 fib6_prune_clones(info->nl_net, pn);
1499 * Walk the leaf entries looking for ourself
1502 for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->dst.rt6_next) {
1504 fib6_del_route(fn, rtp, info);
1512 * Tree traversal function.
1514 * Certainly, it is not interrupt safe.
1515 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1516 * It means, that we can modify tree during walking
1517 * and use this function for garbage collection, clone pruning,
1518 * cleaning tree when a device goes down etc. etc.
1520 * It guarantees that every node will be traversed,
1521 * and that it will be traversed only once.
1523 * Callback function w->func may return:
1524 * 0 -> continue walking.
1525 * positive value -> walking is suspended (used by tree dumps,
1526 * and probably by gc, if it will be split to several slices)
1527 * negative value -> terminate walking.
1529 * The function itself returns:
1530 * 0 -> walk is complete.
1531 * >0 -> walk is incomplete (i.e. suspended)
1532 * <0 -> walk is terminated by an error.
1535 static int fib6_walk_continue(struct fib6_walker *w)
1537 struct fib6_node *fn, *pn;
1544 if (w->prune && fn != w->root &&
1545 fn->fn_flags & RTN_RTINFO && w->state < FWS_C) {
1550 #ifdef CONFIG_IPV6_SUBTREES
1552 if (FIB6_SUBTREE(fn)) {
1553 w->node = FIB6_SUBTREE(fn);
1561 w->state = FWS_INIT;
1567 w->node = fn->right;
1568 w->state = FWS_INIT;
1574 if (w->leaf && fn->fn_flags & RTN_RTINFO) {
1596 #ifdef CONFIG_IPV6_SUBTREES
1597 if (FIB6_SUBTREE(pn) == fn) {
1598 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1603 if (pn->left == fn) {
1607 if (pn->right == fn) {
1609 w->leaf = w->node->leaf;
1619 static int fib6_walk(struct net *net, struct fib6_walker *w)
1623 w->state = FWS_INIT;
1626 fib6_walker_link(net, w);
1627 res = fib6_walk_continue(w);
1629 fib6_walker_unlink(net, w);
1633 static int fib6_clean_node(struct fib6_walker *w)
1636 struct rt6_info *rt;
1637 struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w);
1638 struct nl_info info = {
1642 if (c->sernum != FIB6_NO_SERNUM_CHANGE &&
1643 w->node->fn_sernum != c->sernum)
1644 w->node->fn_sernum = c->sernum;
1647 WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE);
1652 for (rt = w->leaf; rt; rt = rt->dst.rt6_next) {
1653 res = c->func(rt, c->arg);
1656 res = fib6_del(rt, &info);
1659 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
1661 rcu_access_pointer(rt->rt6i_node),
1675 * Convenient frontend to tree walker.
1677 * func is called on each route.
1678 * It may return -1 -> delete this route.
1679 * 0 -> continue walking
1681 * prune==1 -> only immediate children of node (certainly,
1682 * ignoring pure split nodes) will be scanned.
1685 static void fib6_clean_tree(struct net *net, struct fib6_node *root,
1686 int (*func)(struct rt6_info *, void *arg),
1687 bool prune, int sernum, void *arg)
1689 struct fib6_cleaner c;
1692 c.w.func = fib6_clean_node;
1701 fib6_walk(net, &c.w);
1704 static void __fib6_clean_all(struct net *net,
1705 int (*func)(struct rt6_info *, void *),
1706 int sernum, void *arg)
1708 struct fib6_table *table;
1709 struct hlist_head *head;
1713 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
1714 head = &net->ipv6.fib_table_hash[h];
1715 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
1716 write_lock_bh(&table->tb6_lock);
1717 fib6_clean_tree(net, &table->tb6_root,
1718 func, false, sernum, arg);
1719 write_unlock_bh(&table->tb6_lock);
1725 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *),
1728 __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg);
1731 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1733 if (rt->rt6i_flags & RTF_CACHE) {
1734 RT6_TRACE("pruning clone %p\n", rt);
1741 static void fib6_prune_clones(struct net *net, struct fib6_node *fn)
1743 fib6_clean_tree(net, fn, fib6_prune_clone, true,
1744 FIB6_NO_SERNUM_CHANGE, NULL);
1747 static void fib6_flush_trees(struct net *net)
1749 int new_sernum = fib6_new_sernum(net);
1751 __fib6_clean_all(net, NULL, new_sernum, NULL);
1755 * Garbage collection
1764 static int fib6_age(struct rt6_info *rt, void *arg)
1766 struct fib6_gc_args *gc_args = arg;
1767 unsigned long now = jiffies;
1770 * check addrconf expiration here.
1771 * Routes are expired even if they are in use.
1773 * Also age clones. Note, that clones are aged out
1774 * only if they are not in use now.
1777 if (rt->rt6i_flags & RTF_EXPIRES && rt->dst.expires) {
1778 if (time_after(now, rt->dst.expires)) {
1779 RT6_TRACE("expiring %p\n", rt);
1783 } else if (rt->rt6i_flags & RTF_CACHE) {
1784 if (atomic_read(&rt->dst.__refcnt) == 0 &&
1785 time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
1786 RT6_TRACE("aging clone %p\n", rt);
1788 } else if (rt->rt6i_flags & RTF_GATEWAY) {
1789 struct neighbour *neigh;
1790 __u8 neigh_flags = 0;
1792 neigh = dst_neigh_lookup(&rt->dst, &rt->rt6i_gateway);
1794 neigh_flags = neigh->flags;
1795 neigh_release(neigh);
1797 if (!(neigh_flags & NTF_ROUTER)) {
1798 RT6_TRACE("purging route %p via non-router but gateway\n",
1809 void fib6_run_gc(unsigned long expires, struct net *net, bool force)
1811 struct fib6_gc_args gc_args;
1815 spin_lock_bh(&net->ipv6.fib6_gc_lock);
1816 } else if (!spin_trylock_bh(&net->ipv6.fib6_gc_lock)) {
1817 mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
1820 gc_args.timeout = expires ? (int)expires :
1821 net->ipv6.sysctl.ip6_rt_gc_interval;
1823 gc_args.more = icmp6_dst_gc();
1825 fib6_clean_all(net, fib6_age, &gc_args);
1827 net->ipv6.ip6_rt_last_gc = now;
1830 mod_timer(&net->ipv6.ip6_fib_timer,
1832 + net->ipv6.sysctl.ip6_rt_gc_interval));
1834 del_timer(&net->ipv6.ip6_fib_timer);
1835 spin_unlock_bh(&net->ipv6.fib6_gc_lock);
1838 static void fib6_gc_timer_cb(unsigned long arg)
1840 fib6_run_gc(0, (struct net *)arg, true);
1843 static int __net_init fib6_net_init(struct net *net)
1845 size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ;
1847 spin_lock_init(&net->ipv6.fib6_gc_lock);
1848 rwlock_init(&net->ipv6.fib6_walker_lock);
1849 INIT_LIST_HEAD(&net->ipv6.fib6_walkers);
1850 setup_timer(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, (unsigned long)net);
1852 net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL);
1853 if (!net->ipv6.rt6_stats)
1856 /* Avoid false sharing : Use at least a full cache line */
1857 size = max_t(size_t, size, L1_CACHE_BYTES);
1859 net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL);
1860 if (!net->ipv6.fib_table_hash)
1863 net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
1865 if (!net->ipv6.fib6_main_tbl)
1866 goto out_fib_table_hash;
1868 net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
1869 net->ipv6.fib6_main_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1870 net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
1871 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1872 inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers);
1874 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1875 net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
1877 if (!net->ipv6.fib6_local_tbl)
1878 goto out_fib6_main_tbl;
1879 net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
1880 net->ipv6.fib6_local_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1881 net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
1882 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1883 inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers);
1885 fib6_tables_init(net);
1889 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1891 kfree(net->ipv6.fib6_main_tbl);
1894 kfree(net->ipv6.fib_table_hash);
1896 kfree(net->ipv6.rt6_stats);
1901 static void fib6_net_exit(struct net *net)
1905 rt6_ifdown(net, NULL);
1906 del_timer_sync(&net->ipv6.ip6_fib_timer);
1908 for (i = 0; i < FIB6_TABLE_HASHSZ; i++) {
1909 struct hlist_head *head = &net->ipv6.fib_table_hash[i];
1910 struct hlist_node *tmp;
1911 struct fib6_table *tb;
1913 hlist_for_each_entry_safe(tb, tmp, head, tb6_hlist) {
1914 hlist_del(&tb->tb6_hlist);
1915 fib6_free_table(tb);
1919 kfree(net->ipv6.fib_table_hash);
1920 kfree(net->ipv6.rt6_stats);
1923 static struct pernet_operations fib6_net_ops = {
1924 .init = fib6_net_init,
1925 .exit = fib6_net_exit,
1928 int __init fib6_init(void)
1932 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1933 sizeof(struct fib6_node),
1934 0, SLAB_HWCACHE_ALIGN,
1936 if (!fib6_node_kmem)
1939 ret = register_pernet_subsys(&fib6_net_ops);
1941 goto out_kmem_cache_create;
1943 ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib,
1946 goto out_unregister_subsys;
1948 __fib6_flush_trees = fib6_flush_trees;
1952 out_unregister_subsys:
1953 unregister_pernet_subsys(&fib6_net_ops);
1954 out_kmem_cache_create:
1955 kmem_cache_destroy(fib6_node_kmem);
1959 void fib6_gc_cleanup(void)
1961 unregister_pernet_subsys(&fib6_net_ops);
1962 kmem_cache_destroy(fib6_node_kmem);
1965 #ifdef CONFIG_PROC_FS
1967 struct ipv6_route_iter {
1968 struct seq_net_private p;
1969 struct fib6_walker w;
1971 struct fib6_table *tbl;
1975 static int ipv6_route_seq_show(struct seq_file *seq, void *v)
1977 struct rt6_info *rt = v;
1978 struct ipv6_route_iter *iter = seq->private;
1980 seq_printf(seq, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
1982 #ifdef CONFIG_IPV6_SUBTREES
1983 seq_printf(seq, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
1985 seq_puts(seq, "00000000000000000000000000000000 00 ");
1987 if (rt->rt6i_flags & RTF_GATEWAY)
1988 seq_printf(seq, "%pi6", &rt->rt6i_gateway);
1990 seq_puts(seq, "00000000000000000000000000000000");
1992 seq_printf(seq, " %08x %08x %08x %08x %8s\n",
1993 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
1994 rt->dst.__use, rt->rt6i_flags,
1995 rt->dst.dev ? rt->dst.dev->name : "");
1996 iter->w.leaf = NULL;
2000 static int ipv6_route_yield(struct fib6_walker *w)
2002 struct ipv6_route_iter *iter = w->args;
2008 iter->w.leaf = iter->w.leaf->dst.rt6_next;
2010 if (!iter->skip && iter->w.leaf)
2012 } while (iter->w.leaf);
2017 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter,
2020 memset(&iter->w, 0, sizeof(iter->w));
2021 iter->w.func = ipv6_route_yield;
2022 iter->w.root = &iter->tbl->tb6_root;
2023 iter->w.state = FWS_INIT;
2024 iter->w.node = iter->w.root;
2025 iter->w.args = iter;
2026 iter->sernum = iter->w.root->fn_sernum;
2027 INIT_LIST_HEAD(&iter->w.lh);
2028 fib6_walker_link(net, &iter->w);
2031 static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl,
2035 struct hlist_node *node;
2038 h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1;
2039 node = rcu_dereference_bh(hlist_next_rcu(&tbl->tb6_hlist));
2045 while (!node && h < FIB6_TABLE_HASHSZ) {
2046 node = rcu_dereference_bh(
2047 hlist_first_rcu(&net->ipv6.fib_table_hash[h++]));
2049 return hlist_entry_safe(node, struct fib6_table, tb6_hlist);
2052 static void ipv6_route_check_sernum(struct ipv6_route_iter *iter)
2054 if (iter->sernum != iter->w.root->fn_sernum) {
2055 iter->sernum = iter->w.root->fn_sernum;
2056 iter->w.state = FWS_INIT;
2057 iter->w.node = iter->w.root;
2058 WARN_ON(iter->w.skip);
2059 iter->w.skip = iter->w.count;
2063 static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2067 struct net *net = seq_file_net(seq);
2068 struct ipv6_route_iter *iter = seq->private;
2073 n = ((struct rt6_info *)v)->dst.rt6_next;
2080 ipv6_route_check_sernum(iter);
2081 read_lock(&iter->tbl->tb6_lock);
2082 r = fib6_walk_continue(&iter->w);
2083 read_unlock(&iter->tbl->tb6_lock);
2087 return iter->w.leaf;
2089 fib6_walker_unlink(net, &iter->w);
2092 fib6_walker_unlink(net, &iter->w);
2094 iter->tbl = ipv6_route_seq_next_table(iter->tbl, net);
2098 ipv6_route_seq_setup_walk(iter, net);
2102 static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos)
2105 struct net *net = seq_file_net(seq);
2106 struct ipv6_route_iter *iter = seq->private;
2109 iter->tbl = ipv6_route_seq_next_table(NULL, net);
2113 ipv6_route_seq_setup_walk(iter, net);
2114 return ipv6_route_seq_next(seq, NULL, pos);
2120 static bool ipv6_route_iter_active(struct ipv6_route_iter *iter)
2122 struct fib6_walker *w = &iter->w;
2123 return w->node && !(w->state == FWS_U && w->node == w->root);
2126 static void ipv6_route_seq_stop(struct seq_file *seq, void *v)
2129 struct net *net = seq_file_net(seq);
2130 struct ipv6_route_iter *iter = seq->private;
2132 if (ipv6_route_iter_active(iter))
2133 fib6_walker_unlink(net, &iter->w);
2135 rcu_read_unlock_bh();
2138 static const struct seq_operations ipv6_route_seq_ops = {
2139 .start = ipv6_route_seq_start,
2140 .next = ipv6_route_seq_next,
2141 .stop = ipv6_route_seq_stop,
2142 .show = ipv6_route_seq_show
2145 int ipv6_route_open(struct inode *inode, struct file *file)
2147 return seq_open_net(inode, file, &ipv6_route_seq_ops,
2148 sizeof(struct ipv6_route_iter));
2151 #endif /* CONFIG_PROC_FS */