2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
61 #include <linux/sysctl.h>
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
69 #include <net/net_namespace.h>
73 #include <net/af_ieee802154.h>
74 #include <net/firewire.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
96 /* Set to 3 to get tracing... */
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
105 #define INFINITY_LIFE_TIME 0xFFFFFFFF
107 #define IPV6_MAX_STRLEN \
108 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
110 static inline u32 cstamp_delta(unsigned long cstamp)
112 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
115 static inline s32 rfc3315_s14_backoff_init(s32 irt)
117 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
118 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
119 do_div(tmp, 1000000);
123 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
125 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
126 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
127 do_div(tmp, 1000000);
128 if ((s32)tmp > mrt) {
129 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
130 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
131 do_div(tmp, 1000000);
137 static int addrconf_sysctl_register(struct inet6_dev *idev);
138 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
140 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
145 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
150 static void __ipv6_regen_rndid(struct inet6_dev *idev);
151 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
152 static void ipv6_regen_rndid(unsigned long data);
154 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
155 static int ipv6_count_addresses(struct inet6_dev *idev);
156 static int ipv6_generate_stable_address(struct in6_addr *addr,
158 const struct inet6_dev *idev);
161 * Configured unicast address hash table
163 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
164 static DEFINE_SPINLOCK(addrconf_hash_lock);
166 static void addrconf_verify(void);
167 static void addrconf_verify_rtnl(void);
168 static void addrconf_verify_work(struct work_struct *);
170 static struct workqueue_struct *addrconf_wq;
171 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
173 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
174 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
176 static void addrconf_type_change(struct net_device *dev,
177 unsigned long event);
178 static int addrconf_ifdown(struct net_device *dev, int how);
180 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
182 const struct net_device *dev,
183 u32 flags, u32 noflags);
185 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
186 static void addrconf_dad_work(struct work_struct *w);
187 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
188 static void addrconf_dad_run(struct inet6_dev *idev);
189 static void addrconf_rs_timer(unsigned long data);
190 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
191 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
193 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
194 struct prefix_info *pinfo);
195 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
196 struct net_device *dev);
198 static struct ipv6_devconf ipv6_devconf __read_mostly = {
200 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
201 .mtu6 = IPV6_MIN_MTU,
203 .accept_redirects = 1,
205 .force_mld_version = 0,
206 .mldv1_unsolicited_report_interval = 10 * HZ,
207 .mldv2_unsolicited_report_interval = HZ,
209 .rtr_solicits = MAX_RTR_SOLICITATIONS,
210 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
211 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
212 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
214 .temp_valid_lft = TEMP_VALID_LIFETIME,
215 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
216 .regen_max_retry = REGEN_MAX_RETRY,
217 .max_desync_factor = MAX_DESYNC_FACTOR,
218 .max_addresses = IPV6_MAX_ADDRESSES,
219 .accept_ra_defrtr = 1,
220 .accept_ra_from_local = 0,
221 .accept_ra_min_hop_limit= 1,
222 .accept_ra_pinfo = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 .accept_ra_rtr_pref = 1,
225 .rtr_probe_interval = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 .accept_ra_rt_info_min_plen = 0,
228 .accept_ra_rt_info_max_plen = 0,
231 .accept_ra_rt_table = 0,
233 .accept_source_route = 0, /* we do not accept RH0 by default. */
236 .suppress_frag_ndisc = 1,
239 .initialized = false,
241 .use_oif_addrs_only = 0,
242 .ignore_routes_with_linkdown = 0,
245 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
247 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
248 .mtu6 = IPV6_MIN_MTU,
250 .accept_redirects = 1,
252 .force_mld_version = 0,
253 .mldv1_unsolicited_report_interval = 10 * HZ,
254 .mldv2_unsolicited_report_interval = HZ,
256 .rtr_solicits = MAX_RTR_SOLICITATIONS,
257 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
258 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
259 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
261 .temp_valid_lft = TEMP_VALID_LIFETIME,
262 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
263 .regen_max_retry = REGEN_MAX_RETRY,
264 .max_desync_factor = MAX_DESYNC_FACTOR,
265 .max_addresses = IPV6_MAX_ADDRESSES,
266 .accept_ra_defrtr = 1,
267 .accept_ra_from_local = 0,
268 .accept_ra_min_hop_limit= 1,
269 .accept_ra_pinfo = 1,
270 #ifdef CONFIG_IPV6_ROUTER_PREF
271 .accept_ra_rtr_pref = 1,
272 .rtr_probe_interval = 60 * HZ,
273 #ifdef CONFIG_IPV6_ROUTE_INFO
274 .accept_ra_rt_info_min_plen = 0,
275 .accept_ra_rt_info_max_plen = 0,
278 .accept_ra_rt_table = 0,
280 .accept_source_route = 0, /* we do not accept RH0 by default. */
283 .suppress_frag_ndisc = 1,
286 .initialized = false,
288 .use_oif_addrs_only = 0,
289 .ignore_routes_with_linkdown = 0,
292 /* Check if a valid qdisc is available */
293 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
295 return !qdisc_tx_is_noop(dev);
298 static void addrconf_del_rs_timer(struct inet6_dev *idev)
300 if (del_timer(&idev->rs_timer))
304 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
306 if (cancel_delayed_work(&ifp->dad_work))
310 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
313 if (!timer_pending(&idev->rs_timer))
315 mod_timer(&idev->rs_timer, jiffies + when);
318 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
322 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
326 static int snmp6_alloc_dev(struct inet6_dev *idev)
330 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
331 if (!idev->stats.ipv6)
334 for_each_possible_cpu(i) {
335 struct ipstats_mib *addrconf_stats;
336 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
337 u64_stats_init(&addrconf_stats->syncp);
341 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
343 if (!idev->stats.icmpv6dev)
345 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
347 if (!idev->stats.icmpv6msgdev)
353 kfree(idev->stats.icmpv6dev);
355 free_percpu(idev->stats.ipv6);
360 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
362 struct inet6_dev *ndev;
367 if (dev->mtu < IPV6_MIN_MTU)
368 return ERR_PTR(-EINVAL);
370 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
374 rwlock_init(&ndev->lock);
376 INIT_LIST_HEAD(&ndev->addr_list);
377 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
378 (unsigned long)ndev);
379 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
381 if (ndev->cnf.stable_secret.initialized)
382 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
384 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
386 ndev->cnf.mtu6 = dev->mtu;
387 ndev->cnf.sysctl = NULL;
388 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
389 if (!ndev->nd_parms) {
393 if (ndev->cnf.forwarding)
394 dev_disable_lro(dev);
395 /* We refer to the device */
398 if (snmp6_alloc_dev(ndev) < 0) {
400 "%s: cannot allocate memory for statistics; dev=%s.\n",
401 __func__, dev->name);
402 neigh_parms_release(&nd_tbl, ndev->nd_parms);
408 if (snmp6_register_dev(ndev) < 0) {
410 "%s: cannot create /proc/net/dev_snmp6/%s\n",
411 __func__, dev->name);
415 /* One reference from device. We must do this before
416 * we invoke __ipv6_regen_rndid().
420 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
421 ndev->cnf.accept_dad = -1;
423 #if IS_ENABLED(CONFIG_IPV6_SIT)
424 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
425 pr_info("%s: Disabled Multicast RS\n", dev->name);
426 ndev->cnf.rtr_solicits = 0;
430 INIT_LIST_HEAD(&ndev->tempaddr_list);
431 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
432 if ((dev->flags&IFF_LOOPBACK) ||
433 dev->type == ARPHRD_TUNNEL ||
434 dev->type == ARPHRD_TUNNEL6 ||
435 dev->type == ARPHRD_SIT ||
436 dev->type == ARPHRD_NONE) {
437 ndev->cnf.use_tempaddr = -1;
440 ipv6_regen_rndid((unsigned long) ndev);
443 ndev->token = in6addr_any;
445 if (netif_running(dev) && addrconf_qdisc_ok(dev))
446 ndev->if_flags |= IF_READY;
448 ipv6_mc_init_dev(ndev);
449 ndev->tstamp = jiffies;
450 err = addrconf_sysctl_register(ndev);
452 ipv6_mc_destroy_dev(ndev);
453 del_timer(&ndev->regen_timer);
454 snmp6_unregister_dev(ndev);
457 /* protected by rtnl_lock */
458 rcu_assign_pointer(dev->ip6_ptr, ndev);
460 /* Join interface-local all-node multicast group */
461 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
463 /* Join all-node multicast group */
464 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
466 /* Join all-router multicast group if forwarding is set */
467 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
468 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
473 neigh_parms_release(&nd_tbl, ndev->nd_parms);
475 in6_dev_finish_destroy(ndev);
479 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
481 struct inet6_dev *idev;
485 idev = __in6_dev_get(dev);
487 idev = ipv6_add_dev(dev);
492 if (dev->flags&IFF_UP)
497 static int inet6_netconf_msgsize_devconf(int type)
499 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
500 + nla_total_size(4); /* NETCONFA_IFINDEX */
502 /* type -1 is used for ALL */
503 if (type == -1 || type == NETCONFA_FORWARDING)
504 size += nla_total_size(4);
505 #ifdef CONFIG_IPV6_MROUTE
506 if (type == -1 || type == NETCONFA_MC_FORWARDING)
507 size += nla_total_size(4);
509 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
510 size += nla_total_size(4);
512 if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
513 size += nla_total_size(4);
518 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
519 struct ipv6_devconf *devconf, u32 portid,
520 u32 seq, int event, unsigned int flags,
523 struct nlmsghdr *nlh;
524 struct netconfmsg *ncm;
526 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
531 ncm = nlmsg_data(nlh);
532 ncm->ncm_family = AF_INET6;
534 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
535 goto nla_put_failure;
537 /* type -1 is used for ALL */
538 if ((type == -1 || type == NETCONFA_FORWARDING) &&
539 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
540 goto nla_put_failure;
541 #ifdef CONFIG_IPV6_MROUTE
542 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
543 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
544 devconf->mc_forwarding) < 0)
545 goto nla_put_failure;
547 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
548 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
549 goto nla_put_failure;
551 if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
552 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
553 devconf->ignore_routes_with_linkdown) < 0)
554 goto nla_put_failure;
560 nlmsg_cancel(skb, nlh);
564 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
565 struct ipv6_devconf *devconf)
570 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
574 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
575 RTM_NEWNETCONF, 0, type);
577 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
578 WARN_ON(err == -EMSGSIZE);
582 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
585 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
588 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
589 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
590 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
591 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
592 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
595 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
596 struct nlmsghdr *nlh)
598 struct net *net = sock_net(in_skb->sk);
599 struct nlattr *tb[NETCONFA_MAX+1];
600 struct netconfmsg *ncm;
602 struct ipv6_devconf *devconf;
603 struct inet6_dev *in6_dev;
604 struct net_device *dev;
608 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
609 devconf_ipv6_policy);
614 if (!tb[NETCONFA_IFINDEX])
617 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
619 case NETCONFA_IFINDEX_ALL:
620 devconf = net->ipv6.devconf_all;
622 case NETCONFA_IFINDEX_DEFAULT:
623 devconf = net->ipv6.devconf_dflt;
626 dev = __dev_get_by_index(net, ifindex);
629 in6_dev = __in6_dev_get(dev);
632 devconf = &in6_dev->cnf;
637 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
641 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
642 NETLINK_CB(in_skb).portid,
643 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
646 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
647 WARN_ON(err == -EMSGSIZE);
651 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
656 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
657 struct netlink_callback *cb)
659 struct net *net = sock_net(skb->sk);
662 struct net_device *dev;
663 struct inet6_dev *idev;
664 struct hlist_head *head;
667 s_idx = idx = cb->args[1];
669 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
671 head = &net->dev_index_head[h];
673 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
675 hlist_for_each_entry_rcu(dev, head, index_hlist) {
678 idev = __in6_dev_get(dev);
682 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
684 NETLINK_CB(cb->skb).portid,
692 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
698 if (h == NETDEV_HASHENTRIES) {
699 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
700 net->ipv6.devconf_all,
701 NETLINK_CB(cb->skb).portid,
703 RTM_NEWNETCONF, NLM_F_MULTI,
709 if (h == NETDEV_HASHENTRIES + 1) {
710 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
711 net->ipv6.devconf_dflt,
712 NETLINK_CB(cb->skb).portid,
714 RTM_NEWNETCONF, NLM_F_MULTI,
728 static void dev_forward_change(struct inet6_dev *idev)
730 struct net_device *dev;
731 struct inet6_ifaddr *ifa;
736 if (idev->cnf.forwarding)
737 dev_disable_lro(dev);
738 if (dev->flags & IFF_MULTICAST) {
739 if (idev->cnf.forwarding) {
740 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
741 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
742 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
744 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
745 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
746 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
750 list_for_each_entry(ifa, &idev->addr_list, if_list) {
751 if (ifa->flags&IFA_F_TENTATIVE)
753 if (idev->cnf.forwarding)
754 addrconf_join_anycast(ifa);
756 addrconf_leave_anycast(ifa);
758 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
759 dev->ifindex, &idev->cnf);
763 static void addrconf_forward_change(struct net *net, __s32 newf)
765 struct net_device *dev;
766 struct inet6_dev *idev;
768 for_each_netdev(net, dev) {
769 idev = __in6_dev_get(dev);
771 int changed = (!idev->cnf.forwarding) ^ (!newf);
772 idev->cnf.forwarding = newf;
774 dev_forward_change(idev);
779 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
785 return restart_syscall();
787 net = (struct net *)table->extra2;
791 if (p == &net->ipv6.devconf_dflt->forwarding) {
792 if ((!newf) ^ (!old))
793 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
794 NETCONFA_IFINDEX_DEFAULT,
795 net->ipv6.devconf_dflt);
800 if (p == &net->ipv6.devconf_all->forwarding) {
801 net->ipv6.devconf_dflt->forwarding = newf;
802 addrconf_forward_change(net, newf);
803 if ((!newf) ^ (!old))
804 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
805 NETCONFA_IFINDEX_ALL,
806 net->ipv6.devconf_all);
807 } else if ((!newf) ^ (!old))
808 dev_forward_change((struct inet6_dev *)table->extra1);
812 rt6_purge_dflt_routers(net);
816 static void addrconf_linkdown_change(struct net *net, __s32 newf)
818 struct net_device *dev;
819 struct inet6_dev *idev;
821 for_each_netdev(net, dev) {
822 idev = __in6_dev_get(dev);
824 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
826 idev->cnf.ignore_routes_with_linkdown = newf;
828 inet6_netconf_notify_devconf(dev_net(dev),
829 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
836 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
842 return restart_syscall();
844 net = (struct net *)table->extra2;
848 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
849 if ((!newf) ^ (!old))
850 inet6_netconf_notify_devconf(net,
851 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
852 NETCONFA_IFINDEX_DEFAULT,
853 net->ipv6.devconf_dflt);
858 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
859 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
860 addrconf_linkdown_change(net, newf);
861 if ((!newf) ^ (!old))
862 inet6_netconf_notify_devconf(net,
863 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
864 NETCONFA_IFINDEX_ALL,
865 net->ipv6.devconf_all);
874 /* Nobody refers to this ifaddr, destroy it */
875 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
877 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
879 #ifdef NET_REFCNT_DEBUG
880 pr_debug("%s\n", __func__);
883 in6_dev_put(ifp->idev);
885 if (cancel_delayed_work(&ifp->dad_work))
886 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
889 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
890 pr_warn("Freeing alive inet6 address %p\n", ifp);
899 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
902 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
905 * Each device address list is sorted in order of scope -
906 * global before linklocal.
908 list_for_each(p, &idev->addr_list) {
909 struct inet6_ifaddr *ifa
910 = list_entry(p, struct inet6_ifaddr, if_list);
911 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
915 list_add_tail(&ifp->if_list, p);
918 static u32 inet6_addr_hash(const struct in6_addr *addr)
920 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
923 /* On success it returns ifp with increased reference count */
925 static struct inet6_ifaddr *
926 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
927 const struct in6_addr *peer_addr, int pfxlen,
928 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
930 struct inet6_ifaddr *ifa = NULL;
934 int addr_type = ipv6_addr_type(addr);
936 if (addr_type == IPV6_ADDR_ANY ||
937 addr_type & IPV6_ADDR_MULTICAST ||
938 (!(idev->dev->flags & IFF_LOOPBACK) &&
939 addr_type & IPV6_ADDR_LOOPBACK))
940 return ERR_PTR(-EADDRNOTAVAIL);
944 err = -ENODEV; /*XXX*/
948 if (idev->cnf.disable_ipv6) {
953 spin_lock(&addrconf_hash_lock);
955 /* Ignore adding duplicate addresses on an interface */
956 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
957 ADBG("ipv6_add_addr: already assigned\n");
962 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
965 ADBG("ipv6_add_addr: malloc failed\n");
970 rt = addrconf_dst_alloc(idev, addr, false);
976 neigh_parms_data_state_setall(idev->nd_parms);
980 ifa->peer_addr = *peer_addr;
982 spin_lock_init(&ifa->lock);
983 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
984 INIT_HLIST_NODE(&ifa->addr_lst);
986 ifa->prefix_len = pfxlen;
988 /* No need to add the TENTATIVE flag for addresses with NODAD */
989 if (!(flags & IFA_F_NODAD))
990 ifa->flags |= IFA_F_TENTATIVE;
991 ifa->valid_lft = valid_lft;
992 ifa->prefered_lft = prefered_lft;
993 ifa->cstamp = ifa->tstamp = jiffies;
994 ifa->tokenized = false;
1003 /* Add to big hash table */
1004 hash = inet6_addr_hash(addr);
1006 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1007 spin_unlock(&addrconf_hash_lock);
1009 write_lock(&idev->lock);
1010 /* Add to inet6_dev unicast addr list. */
1011 ipv6_link_dev_addr(idev, ifa);
1013 if (ifa->flags&IFA_F_TEMPORARY) {
1014 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1019 write_unlock(&idev->lock);
1021 rcu_read_unlock_bh();
1023 if (likely(err == 0))
1024 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1032 spin_unlock(&addrconf_hash_lock);
1036 enum cleanup_prefix_rt_t {
1037 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1038 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1039 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1043 * Check, whether the prefix for ifp would still need a prefix route
1044 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1047 * 1) we don't purge prefix if address was not permanent.
1048 * prefix is managed by its own lifetime.
1049 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1050 * 3) if there are no addresses, delete prefix.
1051 * 4) if there are still other permanent address(es),
1052 * corresponding prefix is still permanent.
1053 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1054 * don't purge the prefix, assume user space is managing it.
1055 * 6) otherwise, update prefix lifetime to the
1056 * longest valid lifetime among the corresponding
1057 * addresses on the device.
1058 * Note: subsequent RA will update lifetime.
1060 static enum cleanup_prefix_rt_t
1061 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1063 struct inet6_ifaddr *ifa;
1064 struct inet6_dev *idev = ifp->idev;
1065 unsigned long lifetime;
1066 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1070 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1073 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1076 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1077 return CLEANUP_PREFIX_RT_NOP;
1079 action = CLEANUP_PREFIX_RT_EXPIRE;
1081 spin_lock(&ifa->lock);
1083 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1085 * Note: Because this address is
1086 * not permanent, lifetime <
1087 * LONG_MAX / HZ here.
1089 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1090 *expires = ifa->tstamp + lifetime * HZ;
1091 spin_unlock(&ifa->lock);
1098 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1100 struct rt6_info *rt;
1102 rt = addrconf_get_prefix_route(&ifp->addr,
1105 0, RTF_GATEWAY | RTF_DEFAULT);
1110 if (!(rt->rt6i_flags & RTF_EXPIRES))
1111 rt6_set_expires(rt, expires);
1118 /* This function wants to get referenced ifp and releases it before return */
1120 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1123 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1124 unsigned long expires;
1128 spin_lock_bh(&ifp->lock);
1130 ifp->state = INET6_IFADDR_STATE_DEAD;
1131 spin_unlock_bh(&ifp->lock);
1133 if (state == INET6_IFADDR_STATE_DEAD)
1136 spin_lock_bh(&addrconf_hash_lock);
1137 hlist_del_init_rcu(&ifp->addr_lst);
1138 spin_unlock_bh(&addrconf_hash_lock);
1140 write_lock_bh(&ifp->idev->lock);
1142 if (ifp->flags&IFA_F_TEMPORARY) {
1143 list_del(&ifp->tmp_list);
1145 in6_ifa_put(ifp->ifpub);
1151 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1152 action = check_cleanup_prefix_route(ifp, &expires);
1154 list_del_init(&ifp->if_list);
1157 write_unlock_bh(&ifp->idev->lock);
1159 addrconf_del_dad_work(ifp);
1161 ipv6_ifa_notify(RTM_DELADDR, ifp);
1163 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1165 if (action != CLEANUP_PREFIX_RT_NOP) {
1166 cleanup_prefix_route(ifp, expires,
1167 action == CLEANUP_PREFIX_RT_DEL);
1170 /* clean up prefsrc entries */
1171 rt6_remove_prefsrc(ifp);
1176 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1178 struct inet6_dev *idev = ifp->idev;
1179 struct in6_addr addr, *tmpaddr;
1180 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1181 unsigned long regen_advance;
1185 unsigned long now = jiffies;
1187 write_lock_bh(&idev->lock);
1189 spin_lock_bh(&ift->lock);
1190 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1191 spin_unlock_bh(&ift->lock);
1198 if (idev->cnf.use_tempaddr <= 0) {
1199 write_unlock_bh(&idev->lock);
1200 pr_info("%s: use_tempaddr is disabled\n", __func__);
1205 spin_lock_bh(&ifp->lock);
1206 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1207 idev->cnf.use_tempaddr = -1; /*XXX*/
1208 spin_unlock_bh(&ifp->lock);
1209 write_unlock_bh(&idev->lock);
1210 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1217 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1218 __ipv6_try_regen_rndid(idev, tmpaddr);
1219 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1220 age = (now - ifp->tstamp) / HZ;
1221 tmp_valid_lft = min_t(__u32,
1223 idev->cnf.temp_valid_lft + age);
1224 tmp_prefered_lft = min_t(__u32,
1226 idev->cnf.temp_prefered_lft + age -
1227 idev->cnf.max_desync_factor);
1228 tmp_plen = ifp->prefix_len;
1229 tmp_tstamp = ifp->tstamp;
1230 spin_unlock_bh(&ifp->lock);
1232 regen_advance = idev->cnf.regen_max_retry *
1233 idev->cnf.dad_transmits *
1234 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1235 write_unlock_bh(&idev->lock);
1237 /* A temporary address is created only if this calculated Preferred
1238 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1239 * an implementation must not create a temporary address with a zero
1240 * Preferred Lifetime.
1241 * Use age calculation as in addrconf_verify to avoid unnecessary
1242 * temporary addresses being generated.
1244 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1245 if (tmp_prefered_lft <= regen_advance + age) {
1252 addr_flags = IFA_F_TEMPORARY;
1253 /* set in addrconf_prefix_rcv() */
1254 if (ifp->flags & IFA_F_OPTIMISTIC)
1255 addr_flags |= IFA_F_OPTIMISTIC;
1257 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1258 ipv6_addr_scope(&addr), addr_flags,
1259 tmp_valid_lft, tmp_prefered_lft);
1263 pr_info("%s: retry temporary address regeneration\n", __func__);
1265 write_lock_bh(&idev->lock);
1269 spin_lock_bh(&ift->lock);
1272 ift->tstamp = tmp_tstamp;
1273 spin_unlock_bh(&ift->lock);
1275 addrconf_dad_start(ift);
1283 * Choose an appropriate source address (RFC3484)
1286 IPV6_SADDR_RULE_INIT = 0,
1287 IPV6_SADDR_RULE_LOCAL,
1288 IPV6_SADDR_RULE_SCOPE,
1289 IPV6_SADDR_RULE_PREFERRED,
1290 #ifdef CONFIG_IPV6_MIP6
1291 IPV6_SADDR_RULE_HOA,
1293 IPV6_SADDR_RULE_OIF,
1294 IPV6_SADDR_RULE_LABEL,
1295 IPV6_SADDR_RULE_PRIVACY,
1296 IPV6_SADDR_RULE_ORCHID,
1297 IPV6_SADDR_RULE_PREFIX,
1298 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1299 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1304 struct ipv6_saddr_score {
1307 struct inet6_ifaddr *ifa;
1308 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1313 struct ipv6_saddr_dst {
1314 const struct in6_addr *addr;
1321 static inline int ipv6_saddr_preferred(int type)
1323 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1328 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1330 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1331 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1337 static int ipv6_get_saddr_eval(struct net *net,
1338 struct ipv6_saddr_score *score,
1339 struct ipv6_saddr_dst *dst,
1344 if (i <= score->rule) {
1346 case IPV6_SADDR_RULE_SCOPE:
1347 ret = score->scopedist;
1349 case IPV6_SADDR_RULE_PREFIX:
1350 ret = score->matchlen;
1353 ret = !!test_bit(i, score->scorebits);
1359 case IPV6_SADDR_RULE_INIT:
1360 /* Rule 0: remember if hiscore is not ready yet */
1363 case IPV6_SADDR_RULE_LOCAL:
1364 /* Rule 1: Prefer same address */
1365 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1367 case IPV6_SADDR_RULE_SCOPE:
1368 /* Rule 2: Prefer appropriate scope
1373 * ---+--+-+---> scope
1375 * | d is scope of the destination.
1377 * | \ <- smaller scope is better if
1378 * B-15 | \ if scope is enough for destination.
1379 * | ret = B - scope (-1 <= scope >= d <= 15).
1381 * |/ <- greater is better
1382 * -C / if scope is not enough for destination.
1383 * /| ret = scope - C (-1 <= d < scope <= 15).
1385 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1386 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1387 * Assume B = 0 and we get C > 29.
1389 ret = __ipv6_addr_src_scope(score->addr_type);
1390 if (ret >= dst->scope)
1393 ret -= 128; /* 30 is enough */
1394 score->scopedist = ret;
1396 case IPV6_SADDR_RULE_PREFERRED:
1398 /* Rule 3: Avoid deprecated and optimistic addresses */
1399 u8 avoid = IFA_F_DEPRECATED;
1401 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1402 avoid |= IFA_F_OPTIMISTIC;
1403 ret = ipv6_saddr_preferred(score->addr_type) ||
1404 !(score->ifa->flags & avoid);
1407 #ifdef CONFIG_IPV6_MIP6
1408 case IPV6_SADDR_RULE_HOA:
1410 /* Rule 4: Prefer home address */
1411 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1412 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1416 case IPV6_SADDR_RULE_OIF:
1417 /* Rule 5: Prefer outgoing interface */
1418 ret = (!dst->ifindex ||
1419 dst->ifindex == score->ifa->idev->dev->ifindex);
1421 case IPV6_SADDR_RULE_LABEL:
1422 /* Rule 6: Prefer matching label */
1423 ret = ipv6_addr_label(net,
1424 &score->ifa->addr, score->addr_type,
1425 score->ifa->idev->dev->ifindex) == dst->label;
1427 case IPV6_SADDR_RULE_PRIVACY:
1429 /* Rule 7: Prefer public address
1430 * Note: prefer temporary address if use_tempaddr >= 2
1432 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1433 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1434 score->ifa->idev->cnf.use_tempaddr >= 2;
1435 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1438 case IPV6_SADDR_RULE_ORCHID:
1439 /* Rule 8-: Prefer ORCHID vs ORCHID or
1440 * non-ORCHID vs non-ORCHID
1442 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1443 ipv6_addr_orchid(dst->addr));
1445 case IPV6_SADDR_RULE_PREFIX:
1446 /* Rule 8: Use longest matching prefix */
1447 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1448 if (ret > score->ifa->prefix_len)
1449 ret = score->ifa->prefix_len;
1450 score->matchlen = ret;
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1454 /* Optimistic addresses still have lower precedence than other
1455 * preferred addresses.
1457 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1465 __set_bit(i, score->scorebits);
1471 static int __ipv6_dev_get_saddr(struct net *net,
1472 struct ipv6_saddr_dst *dst,
1473 struct inet6_dev *idev,
1474 struct ipv6_saddr_score *scores,
1477 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1479 read_lock_bh(&idev->lock);
1480 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1484 * - Tentative Address (RFC2462 section 5.4)
1485 * - A tentative address is not considered
1486 * "assigned to an interface" in the traditional
1487 * sense, unless it is also flagged as optimistic.
1488 * - Candidate Source Address (section 4)
1489 * - In any case, anycast addresses, multicast
1490 * addresses, and the unspecified address MUST
1491 * NOT be included in a candidate set.
1493 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1494 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1497 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1499 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1500 score->addr_type & IPV6_ADDR_MULTICAST)) {
1501 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1507 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1509 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1510 int minihiscore, miniscore;
1512 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1513 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1515 if (minihiscore > miniscore) {
1516 if (i == IPV6_SADDR_RULE_SCOPE &&
1517 score->scopedist > 0) {
1520 * each remaining entry
1521 * has too small (not enough)
1522 * scope, because ifa entries
1523 * are sorted by their scope
1529 } else if (minihiscore < miniscore) {
1531 in6_ifa_put(hiscore->ifa);
1533 in6_ifa_hold(score->ifa);
1535 swap(hiscore, score);
1536 hiscore_idx = 1 - hiscore_idx;
1538 /* restore our iterator */
1539 score->ifa = hiscore->ifa;
1546 read_unlock_bh(&idev->lock);
1550 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1551 const struct in6_addr *daddr, unsigned int prefs,
1552 struct in6_addr *saddr)
1554 struct ipv6_saddr_score scores[2], *hiscore;
1555 struct ipv6_saddr_dst dst;
1556 struct inet6_dev *idev;
1557 struct net_device *dev;
1559 bool use_oif_addr = false;
1560 int hiscore_idx = 0;
1562 dst_type = __ipv6_addr_type(daddr);
1564 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1565 dst.scope = __ipv6_addr_src_scope(dst_type);
1566 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1569 scores[hiscore_idx].rule = -1;
1570 scores[hiscore_idx].ifa = NULL;
1574 /* Candidate Source Address (section 4)
1575 * - multicast and link-local destination address,
1576 * the set of candidate source address MUST only
1577 * include addresses assigned to interfaces
1578 * belonging to the same link as the outgoing
1580 * (- For site-local destination addresses, the
1581 * set of candidate source addresses MUST only
1582 * include addresses assigned to interfaces
1583 * belonging to the same site as the outgoing
1585 * - "It is RECOMMENDED that the candidate source addresses
1586 * be the set of unicast addresses assigned to the
1587 * interface that will be used to send to the destination
1588 * (the 'outgoing' interface)." (RFC 6724)
1591 idev = __in6_dev_get(dst_dev);
1592 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1593 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1594 (idev && idev->cnf.use_oif_addrs_only)) {
1595 use_oif_addr = true;
1601 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1603 for_each_netdev_rcu(net, dev) {
1604 idev = __in6_dev_get(dev);
1607 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1612 hiscore = &scores[hiscore_idx];
1614 return -EADDRNOTAVAIL;
1616 *saddr = hiscore->ifa->addr;
1617 in6_ifa_put(hiscore->ifa);
1620 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1622 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1625 struct inet6_ifaddr *ifp;
1626 int err = -EADDRNOTAVAIL;
1628 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1629 if (ifp->scope > IFA_LINK)
1631 if (ifp->scope == IFA_LINK &&
1632 !(ifp->flags & banned_flags)) {
1641 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1644 struct inet6_dev *idev;
1645 int err = -EADDRNOTAVAIL;
1648 idev = __in6_dev_get(dev);
1650 read_lock_bh(&idev->lock);
1651 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1652 read_unlock_bh(&idev->lock);
1658 static int ipv6_count_addresses(struct inet6_dev *idev)
1661 struct inet6_ifaddr *ifp;
1663 read_lock_bh(&idev->lock);
1664 list_for_each_entry(ifp, &idev->addr_list, if_list)
1666 read_unlock_bh(&idev->lock);
1670 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1671 const struct net_device *dev, int strict)
1673 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1675 EXPORT_SYMBOL(ipv6_chk_addr);
1677 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1678 const struct net_device *dev, int strict,
1681 struct inet6_ifaddr *ifp;
1682 unsigned int hash = inet6_addr_hash(addr);
1686 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1687 if (!net_eq(dev_net(ifp->idev->dev), net))
1689 /* Decouple optimistic from tentative for evaluation here.
1690 * Ban optimistic addresses explicitly, when required.
1692 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1693 ? (ifp->flags&~IFA_F_TENTATIVE)
1695 if (ipv6_addr_equal(&ifp->addr, addr) &&
1696 !(ifp_flags&banned_flags) &&
1697 (!dev || ifp->idev->dev == dev ||
1698 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1699 rcu_read_unlock_bh();
1704 rcu_read_unlock_bh();
1707 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1709 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1710 struct net_device *dev)
1712 unsigned int hash = inet6_addr_hash(addr);
1713 struct inet6_ifaddr *ifp;
1715 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1716 if (!net_eq(dev_net(ifp->idev->dev), net))
1718 if (ipv6_addr_equal(&ifp->addr, addr)) {
1719 if (!dev || ifp->idev->dev == dev)
1726 /* Compares an address/prefix_len with addresses on device @dev.
1727 * If one is found it returns true.
1729 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1730 const unsigned int prefix_len, struct net_device *dev)
1732 struct inet6_dev *idev;
1733 struct inet6_ifaddr *ifa;
1737 idev = __in6_dev_get(dev);
1739 read_lock_bh(&idev->lock);
1740 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1741 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1745 read_unlock_bh(&idev->lock);
1751 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1753 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1755 struct inet6_dev *idev;
1756 struct inet6_ifaddr *ifa;
1761 idev = __in6_dev_get(dev);
1763 read_lock_bh(&idev->lock);
1764 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1765 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1770 read_unlock_bh(&idev->lock);
1775 EXPORT_SYMBOL(ipv6_chk_prefix);
1777 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1778 struct net_device *dev, int strict)
1780 struct inet6_ifaddr *ifp, *result = NULL;
1781 unsigned int hash = inet6_addr_hash(addr);
1784 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1785 if (!net_eq(dev_net(ifp->idev->dev), net))
1787 if (ipv6_addr_equal(&ifp->addr, addr)) {
1788 if (!dev || ifp->idev->dev == dev ||
1789 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1796 rcu_read_unlock_bh();
1801 /* Gets referenced address, destroys ifaddr */
1803 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1805 if (ifp->flags&IFA_F_TEMPORARY) {
1806 struct inet6_ifaddr *ifpub;
1807 spin_lock_bh(&ifp->lock);
1810 in6_ifa_hold(ifpub);
1811 spin_unlock_bh(&ifp->lock);
1812 ipv6_create_tempaddr(ifpub, ifp);
1815 spin_unlock_bh(&ifp->lock);
1818 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1819 spin_lock_bh(&ifp->lock);
1820 addrconf_del_dad_work(ifp);
1821 ifp->flags |= IFA_F_TENTATIVE;
1823 ifp->flags |= IFA_F_DADFAILED;
1824 spin_unlock_bh(&ifp->lock);
1826 ipv6_ifa_notify(0, ifp);
1833 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1837 spin_lock_bh(&ifp->lock);
1838 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1839 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1842 spin_unlock_bh(&ifp->lock);
1847 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1849 struct in6_addr addr;
1850 struct inet6_dev *idev = ifp->idev;
1851 struct net *net = dev_net(ifp->idev->dev);
1853 if (addrconf_dad_end(ifp)) {
1858 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1859 ifp->idev->dev->name, &ifp->addr);
1861 spin_lock_bh(&ifp->lock);
1863 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1864 int scope = ifp->scope;
1865 u32 flags = ifp->flags;
1866 struct in6_addr new_addr;
1867 struct inet6_ifaddr *ifp2;
1868 u32 valid_lft, preferred_lft;
1869 int pfxlen = ifp->prefix_len;
1870 int retries = ifp->stable_privacy_retry + 1;
1872 if (retries > net->ipv6.sysctl.idgen_retries) {
1873 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1874 ifp->idev->dev->name);
1878 new_addr = ifp->addr;
1879 if (ipv6_generate_stable_address(&new_addr, retries,
1883 valid_lft = ifp->valid_lft;
1884 preferred_lft = ifp->prefered_lft;
1886 spin_unlock_bh(&ifp->lock);
1888 if (idev->cnf.max_addresses &&
1889 ipv6_count_addresses(idev) >=
1890 idev->cnf.max_addresses)
1893 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1894 ifp->idev->dev->name);
1896 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1897 scope, flags, valid_lft,
1902 spin_lock_bh(&ifp2->lock);
1903 ifp2->stable_privacy_retry = retries;
1904 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1905 spin_unlock_bh(&ifp2->lock);
1907 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1910 spin_lock_bh(&ifp->lock);
1911 } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1912 addr.s6_addr32[0] = htonl(0xfe800000);
1913 addr.s6_addr32[1] = 0;
1915 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1916 ipv6_addr_equal(&ifp->addr, &addr)) {
1917 /* DAD failed for link-local based on MAC address */
1918 idev->cnf.disable_ipv6 = 1;
1920 pr_info("%s: IPv6 being disabled!\n",
1921 ifp->idev->dev->name);
1926 /* transition from _POSTDAD to _ERRDAD */
1927 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1928 spin_unlock_bh(&ifp->lock);
1930 addrconf_mod_dad_work(ifp, 0);
1934 /* Join to solicited addr multicast group.
1935 * caller must hold RTNL */
1936 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1938 struct in6_addr maddr;
1940 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1943 addrconf_addr_solict_mult(addr, &maddr);
1944 ipv6_dev_mc_inc(dev, &maddr);
1947 /* caller must hold RTNL */
1948 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1950 struct in6_addr maddr;
1952 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1955 addrconf_addr_solict_mult(addr, &maddr);
1956 __ipv6_dev_mc_dec(idev, &maddr);
1959 /* caller must hold RTNL */
1960 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1962 struct in6_addr addr;
1964 if (ifp->prefix_len >= 127) /* RFC 6164 */
1966 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1967 if (ipv6_addr_any(&addr))
1969 __ipv6_dev_ac_inc(ifp->idev, &addr);
1972 /* caller must hold RTNL */
1973 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1975 struct in6_addr addr;
1977 if (ifp->prefix_len >= 127) /* RFC 6164 */
1979 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1980 if (ipv6_addr_any(&addr))
1982 __ipv6_dev_ac_dec(ifp->idev, &addr);
1985 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1987 if (dev->addr_len != IEEE802154_ADDR_LEN)
1989 memcpy(eui, dev->dev_addr, 8);
1994 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1996 union fwnet_hwaddr *ha;
1998 if (dev->addr_len != FWNET_ALEN)
2001 ha = (union fwnet_hwaddr *)dev->dev_addr;
2003 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2008 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2010 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2011 if (dev->addr_len != ARCNET_ALEN)
2014 eui[7] = *(u8 *)dev->dev_addr;
2018 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2020 if (dev->addr_len != INFINIBAND_ALEN)
2022 memcpy(eui, dev->dev_addr + 12, 8);
2027 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2031 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2032 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2033 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2034 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2035 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2036 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2040 memcpy(eui + 4, &addr, 4);
2044 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2046 if (dev->priv_flags & IFF_ISATAP)
2047 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2051 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2053 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2056 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2058 memcpy(eui, dev->perm_addr, 3);
2059 memcpy(eui + 5, dev->perm_addr + 3, 3);
2066 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2068 switch (dev->type) {
2071 return addrconf_ifid_eui48(eui, dev);
2073 return addrconf_ifid_arcnet(eui, dev);
2074 case ARPHRD_INFINIBAND:
2075 return addrconf_ifid_infiniband(eui, dev);
2077 return addrconf_ifid_sit(eui, dev);
2079 return addrconf_ifid_gre(eui, dev);
2080 case ARPHRD_6LOWPAN:
2081 case ARPHRD_IEEE802154:
2082 return addrconf_ifid_eui64(eui, dev);
2083 case ARPHRD_IEEE1394:
2084 return addrconf_ifid_ieee1394(eui, dev);
2085 case ARPHRD_TUNNEL6:
2086 return addrconf_ifid_ip6tnl(eui, dev);
2091 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2094 struct inet6_ifaddr *ifp;
2096 read_lock_bh(&idev->lock);
2097 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2098 if (ifp->scope > IFA_LINK)
2100 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2101 memcpy(eui, ifp->addr.s6_addr+8, 8);
2106 read_unlock_bh(&idev->lock);
2110 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2111 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2114 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2115 idev->rndid[0] &= ~0x02;
2118 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2119 * check if generated address is not inappropriate
2121 * - Reserved subnet anycast (RFC 2526)
2122 * 11111101 11....11 1xxxxxxx
2123 * - ISATAP (RFC4214) 6.1
2124 * 00-00-5E-FE-xx-xx-xx-xx
2126 * - XXX: already assigned to an address on the device
2128 if (idev->rndid[0] == 0xfd &&
2129 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2130 (idev->rndid[7]&0x80))
2132 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2133 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2135 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2140 static void ipv6_regen_rndid(unsigned long data)
2142 struct inet6_dev *idev = (struct inet6_dev *) data;
2143 unsigned long expires;
2146 write_lock_bh(&idev->lock);
2151 __ipv6_regen_rndid(idev);
2154 idev->cnf.temp_prefered_lft * HZ -
2155 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2156 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2157 idev->cnf.max_desync_factor * HZ;
2158 if (time_before(expires, jiffies)) {
2159 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2160 __func__, idev->dev->name);
2164 if (!mod_timer(&idev->regen_timer, expires))
2168 write_unlock_bh(&idev->lock);
2169 rcu_read_unlock_bh();
2173 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2175 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2176 __ipv6_regen_rndid(idev);
2179 u32 addrconf_rt_table(const struct net_device *dev, u32 default_table) {
2180 /* Determines into what table to put autoconf PIO/RIO/default routes
2181 * learned on this device.
2183 * - If 0, use the same table for every device. This puts routes into
2184 * one of RT_TABLE_{PREFIX,INFO,DFLT} depending on the type of route
2185 * (but note that these three are currently all equal to
2187 * - If > 0, use the specified table.
2188 * - If < 0, put routes into table dev->ifindex + (-rt_table).
2190 struct inet6_dev *idev = in6_dev_get(dev);
2192 int sysctl = idev->cnf.accept_ra_rt_table;
2194 table = default_table;
2195 } else if (sysctl > 0) {
2196 table = (u32) sysctl;
2198 table = (unsigned) dev->ifindex + (-sysctl);
2209 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2210 unsigned long expires, u32 flags)
2212 struct fib6_config cfg = {
2213 .fc_table = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_PREFIX),
2214 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2215 .fc_ifindex = dev->ifindex,
2216 .fc_expires = expires,
2218 .fc_flags = RTF_UP | flags,
2219 .fc_nlinfo.nl_net = dev_net(dev),
2220 .fc_protocol = RTPROT_KERNEL,
2225 /* Prevent useless cloning on PtP SIT.
2226 This thing is done here expecting that the whole
2227 class of non-broadcast devices need not cloning.
2229 #if IS_ENABLED(CONFIG_IPV6_SIT)
2230 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2231 cfg.fc_flags |= RTF_NONEXTHOP;
2234 ip6_route_add(&cfg);
2238 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2240 const struct net_device *dev,
2241 u32 flags, u32 noflags)
2243 struct fib6_node *fn;
2244 struct rt6_info *rt = NULL;
2245 struct fib6_table *table;
2246 u32 tb_id = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_PREFIX);
2248 table = fib6_get_table(dev_net(dev), tb_id);
2252 read_lock_bh(&table->tb6_lock);
2253 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2257 noflags |= RTF_CACHE;
2258 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2259 if (rt->dst.dev->ifindex != dev->ifindex)
2261 if ((rt->rt6i_flags & flags) != flags)
2263 if ((rt->rt6i_flags & noflags) != 0)
2269 read_unlock_bh(&table->tb6_lock);
2274 /* Create "default" multicast route to the interface */
2276 static void addrconf_add_mroute(struct net_device *dev)
2278 struct fib6_config cfg = {
2279 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2280 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2281 .fc_ifindex = dev->ifindex,
2284 .fc_nlinfo.nl_net = dev_net(dev),
2287 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2289 ip6_route_add(&cfg);
2292 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2294 struct inet6_dev *idev;
2298 idev = ipv6_find_idev(dev);
2300 return ERR_PTR(-ENOBUFS);
2302 if (idev->cnf.disable_ipv6)
2303 return ERR_PTR(-EACCES);
2305 /* Add default multicast route */
2306 if (!(dev->flags & IFF_LOOPBACK))
2307 addrconf_add_mroute(dev);
2312 static void manage_tempaddrs(struct inet6_dev *idev,
2313 struct inet6_ifaddr *ifp,
2314 __u32 valid_lft, __u32 prefered_lft,
2315 bool create, unsigned long now)
2318 struct inet6_ifaddr *ift;
2320 read_lock_bh(&idev->lock);
2321 /* update all temporary addresses in the list */
2322 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2323 int age, max_valid, max_prefered;
2325 if (ifp != ift->ifpub)
2328 /* RFC 4941 section 3.3:
2329 * If a received option will extend the lifetime of a public
2330 * address, the lifetimes of temporary addresses should
2331 * be extended, subject to the overall constraint that no
2332 * temporary addresses should ever remain "valid" or "preferred"
2333 * for a time longer than (TEMP_VALID_LIFETIME) or
2334 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2336 age = (now - ift->cstamp) / HZ;
2337 max_valid = idev->cnf.temp_valid_lft - age;
2341 max_prefered = idev->cnf.temp_prefered_lft -
2342 idev->cnf.max_desync_factor - age;
2343 if (max_prefered < 0)
2346 if (valid_lft > max_valid)
2347 valid_lft = max_valid;
2349 if (prefered_lft > max_prefered)
2350 prefered_lft = max_prefered;
2352 spin_lock(&ift->lock);
2354 ift->valid_lft = valid_lft;
2355 ift->prefered_lft = prefered_lft;
2357 if (prefered_lft > 0)
2358 ift->flags &= ~IFA_F_DEPRECATED;
2360 spin_unlock(&ift->lock);
2361 if (!(flags&IFA_F_TENTATIVE))
2362 ipv6_ifa_notify(0, ift);
2365 if ((create || list_empty(&idev->tempaddr_list)) &&
2366 idev->cnf.use_tempaddr > 0) {
2367 /* When a new public address is created as described
2368 * in [ADDRCONF], also create a new temporary address.
2369 * Also create a temporary address if it's enabled but
2370 * no temporary address currently exists.
2372 read_unlock_bh(&idev->lock);
2373 ipv6_create_tempaddr(ifp, NULL);
2375 read_unlock_bh(&idev->lock);
2379 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2381 struct prefix_info *pinfo;
2386 struct inet6_dev *in6_dev;
2387 struct net *net = dev_net(dev);
2389 pinfo = (struct prefix_info *) opt;
2391 if (len < sizeof(struct prefix_info)) {
2392 ADBG("addrconf: prefix option too short\n");
2397 * Validation checks ([ADDRCONF], page 19)
2400 addr_type = ipv6_addr_type(&pinfo->prefix);
2402 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2405 valid_lft = ntohl(pinfo->valid);
2406 prefered_lft = ntohl(pinfo->prefered);
2408 if (prefered_lft > valid_lft) {
2409 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2413 in6_dev = in6_dev_get(dev);
2416 net_dbg_ratelimited("addrconf: device %s not configured\n",
2422 * Two things going on here:
2423 * 1) Add routes for on-link prefixes
2424 * 2) Configure prefixes with the auto flag set
2427 if (pinfo->onlink) {
2428 struct rt6_info *rt;
2429 unsigned long rt_expires;
2431 /* Avoid arithmetic overflow. Really, we could
2432 * save rt_expires in seconds, likely valid_lft,
2433 * but it would require division in fib gc, that it
2437 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2439 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2441 if (addrconf_finite_timeout(rt_expires))
2444 rt = addrconf_get_prefix_route(&pinfo->prefix,
2447 RTF_ADDRCONF | RTF_PREFIX_RT,
2448 RTF_GATEWAY | RTF_DEFAULT);
2451 /* Autoconf prefix route */
2452 if (valid_lft == 0) {
2455 } else if (addrconf_finite_timeout(rt_expires)) {
2457 rt6_set_expires(rt, jiffies + rt_expires);
2459 rt6_clean_expires(rt);
2461 } else if (valid_lft) {
2462 clock_t expires = 0;
2463 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2464 if (addrconf_finite_timeout(rt_expires)) {
2466 flags |= RTF_EXPIRES;
2467 expires = jiffies_to_clock_t(rt_expires);
2469 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2470 dev, expires, flags);
2475 /* Try to figure out our local address for this prefix */
2477 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2478 struct inet6_ifaddr *ifp;
2479 struct in6_addr addr;
2480 int create = 0, update_lft = 0;
2481 bool tokenized = false;
2483 if (pinfo->prefix_len == 64) {
2484 memcpy(&addr, &pinfo->prefix, 8);
2486 if (!ipv6_addr_any(&in6_dev->token)) {
2487 read_lock_bh(&in6_dev->lock);
2488 memcpy(addr.s6_addr + 8,
2489 in6_dev->token.s6_addr + 8, 8);
2490 read_unlock_bh(&in6_dev->lock);
2492 } else if (in6_dev->addr_gen_mode ==
2493 IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2494 !ipv6_generate_stable_address(&addr, 0,
2496 addr_flags |= IFA_F_STABLE_PRIVACY;
2498 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2499 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2500 in6_dev_put(in6_dev);
2505 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2507 in6_dev_put(in6_dev);
2512 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2514 if (!ifp && valid_lft) {
2515 int max_addresses = in6_dev->cnf.max_addresses;
2517 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2518 if (in6_dev->cnf.optimistic_dad &&
2519 !net->ipv6.devconf_all->forwarding && sllao)
2520 addr_flags |= IFA_F_OPTIMISTIC;
2523 /* Do not allow to create too much of autoconfigured
2524 * addresses; this would be too easy way to crash kernel.
2526 if (!max_addresses ||
2527 ipv6_count_addresses(in6_dev) < max_addresses)
2528 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2530 addr_type&IPV6_ADDR_SCOPE_MASK,
2531 addr_flags, valid_lft,
2534 if (IS_ERR_OR_NULL(ifp)) {
2535 in6_dev_put(in6_dev);
2541 spin_lock_bh(&ifp->lock);
2542 ifp->flags |= IFA_F_MANAGETEMPADDR;
2543 ifp->cstamp = jiffies;
2544 ifp->tokenized = tokenized;
2545 spin_unlock_bh(&ifp->lock);
2546 addrconf_dad_start(ifp);
2554 /* update lifetime (RFC2462 5.5.3 e) */
2555 spin_lock_bh(&ifp->lock);
2557 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2558 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2561 if (!update_lft && !create && stored_lft) {
2562 const u32 minimum_lft = min_t(u32,
2563 stored_lft, MIN_VALID_LIFETIME);
2564 valid_lft = max(valid_lft, minimum_lft);
2566 /* RFC4862 Section 5.5.3e:
2567 * "Note that the preferred lifetime of the
2568 * corresponding address is always reset to
2569 * the Preferred Lifetime in the received
2570 * Prefix Information option, regardless of
2571 * whether the valid lifetime is also reset or
2574 * So we should always update prefered_lft here.
2580 ifp->valid_lft = valid_lft;
2581 ifp->prefered_lft = prefered_lft;
2584 ifp->flags &= ~IFA_F_DEPRECATED;
2585 spin_unlock_bh(&ifp->lock);
2587 if (!(flags&IFA_F_TENTATIVE))
2588 ipv6_ifa_notify(0, ifp);
2590 spin_unlock_bh(&ifp->lock);
2592 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2599 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2600 in6_dev_put(in6_dev);
2604 * Set destination address.
2605 * Special case for SIT interfaces where we create a new "virtual"
2608 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2610 struct in6_ifreq ireq;
2611 struct net_device *dev;
2617 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2620 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2626 #if IS_ENABLED(CONFIG_IPV6_SIT)
2627 if (dev->type == ARPHRD_SIT) {
2628 const struct net_device_ops *ops = dev->netdev_ops;
2630 struct ip_tunnel_parm p;
2632 err = -EADDRNOTAVAIL;
2633 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2636 memset(&p, 0, sizeof(p));
2637 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2641 p.iph.protocol = IPPROTO_IPV6;
2643 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2645 if (ops->ndo_do_ioctl) {
2646 mm_segment_t oldfs = get_fs();
2649 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2656 dev = __dev_get_by_name(net, p.name);
2659 err = dev_open(dev);
2669 static int ipv6_mc_config(struct sock *sk, bool join,
2670 const struct in6_addr *addr, int ifindex)
2678 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2680 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2687 * Manual configuration of address on an interface
2689 static int inet6_addr_add(struct net *net, int ifindex,
2690 const struct in6_addr *pfx,
2691 const struct in6_addr *peer_pfx,
2692 unsigned int plen, __u32 ifa_flags,
2693 __u32 prefered_lft, __u32 valid_lft)
2695 struct inet6_ifaddr *ifp;
2696 struct inet6_dev *idev;
2697 struct net_device *dev;
2698 unsigned long timeout;
2708 /* check the lifetime */
2709 if (!valid_lft || prefered_lft > valid_lft)
2712 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2715 dev = __dev_get_by_index(net, ifindex);
2719 idev = addrconf_add_dev(dev);
2721 return PTR_ERR(idev);
2723 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2724 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2725 true, pfx, ifindex);
2731 scope = ipv6_addr_scope(pfx);
2733 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2734 if (addrconf_finite_timeout(timeout)) {
2735 expires = jiffies_to_clock_t(timeout * HZ);
2736 valid_lft = timeout;
2737 flags = RTF_EXPIRES;
2741 ifa_flags |= IFA_F_PERMANENT;
2744 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2745 if (addrconf_finite_timeout(timeout)) {
2747 ifa_flags |= IFA_F_DEPRECATED;
2748 prefered_lft = timeout;
2751 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2752 valid_lft, prefered_lft);
2755 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2756 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2761 * Note that section 3.1 of RFC 4429 indicates
2762 * that the Optimistic flag should not be set for
2763 * manually configured addresses
2765 addrconf_dad_start(ifp);
2766 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2767 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2770 addrconf_verify_rtnl();
2772 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2773 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2774 false, pfx, ifindex);
2777 return PTR_ERR(ifp);
2780 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2781 const struct in6_addr *pfx, unsigned int plen)
2783 struct inet6_ifaddr *ifp;
2784 struct inet6_dev *idev;
2785 struct net_device *dev;
2790 dev = __dev_get_by_index(net, ifindex);
2794 idev = __in6_dev_get(dev);
2798 read_lock_bh(&idev->lock);
2799 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2800 if (ifp->prefix_len == plen &&
2801 ipv6_addr_equal(pfx, &ifp->addr)) {
2803 read_unlock_bh(&idev->lock);
2805 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2806 (ifa_flags & IFA_F_MANAGETEMPADDR))
2807 manage_tempaddrs(idev, ifp, 0, 0, false,
2810 addrconf_verify_rtnl();
2811 if (ipv6_addr_is_multicast(pfx)) {
2812 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2813 false, pfx, dev->ifindex);
2818 read_unlock_bh(&idev->lock);
2819 return -EADDRNOTAVAIL;
2823 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2825 struct in6_ifreq ireq;
2828 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2831 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2835 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2836 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2837 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2842 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2844 struct in6_ifreq ireq;
2847 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2850 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2854 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2855 ireq.ifr6_prefixlen);
2860 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2861 int plen, int scope)
2863 struct inet6_ifaddr *ifp;
2865 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2866 scope, IFA_F_PERMANENT,
2867 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2869 spin_lock_bh(&ifp->lock);
2870 ifp->flags &= ~IFA_F_TENTATIVE;
2871 spin_unlock_bh(&ifp->lock);
2872 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2877 #if IS_ENABLED(CONFIG_IPV6_SIT)
2878 static void sit_add_v4_addrs(struct inet6_dev *idev)
2880 struct in6_addr addr;
2881 struct net_device *dev;
2882 struct net *net = dev_net(idev->dev);
2888 memset(&addr, 0, sizeof(struct in6_addr));
2889 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2891 if (idev->dev->flags&IFF_POINTOPOINT) {
2892 addr.s6_addr32[0] = htonl(0xfe800000);
2896 scope = IPV6_ADDR_COMPATv4;
2898 pflags |= RTF_NONEXTHOP;
2901 if (addr.s6_addr32[3]) {
2902 add_addr(idev, &addr, plen, scope);
2903 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2907 for_each_netdev(net, dev) {
2908 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2909 if (in_dev && (dev->flags & IFF_UP)) {
2910 struct in_ifaddr *ifa;
2914 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2916 addr.s6_addr32[3] = ifa->ifa_local;
2918 if (ifa->ifa_scope == RT_SCOPE_LINK)
2920 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2921 if (idev->dev->flags&IFF_POINTOPOINT)
2926 add_addr(idev, &addr, plen, flag);
2927 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2935 static void init_loopback(struct net_device *dev)
2937 struct inet6_dev *idev;
2938 struct net_device *sp_dev;
2939 struct inet6_ifaddr *sp_ifa;
2940 struct rt6_info *sp_rt;
2946 idev = ipv6_find_idev(dev);
2948 pr_debug("%s: add_dev failed\n", __func__);
2952 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2954 /* Add routes to other interface's IPv6 addresses */
2955 for_each_netdev(dev_net(dev), sp_dev) {
2956 if (!strcmp(sp_dev->name, dev->name))
2959 idev = __in6_dev_get(sp_dev);
2963 read_lock_bh(&idev->lock);
2964 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2966 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2970 /* This dst has been added to garbage list when
2971 * lo device down, release this obsolete dst and
2972 * reallocate a new router for ifa.
2974 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
2975 ip6_rt_put(sp_ifa->rt);
2982 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2984 /* Failure cases are ignored */
2985 if (!IS_ERR(sp_rt)) {
2990 read_unlock_bh(&idev->lock);
2994 static void addrconf_add_linklocal(struct inet6_dev *idev,
2995 const struct in6_addr *addr, u32 flags)
2997 struct inet6_ifaddr *ifp;
2998 u32 addr_flags = flags | IFA_F_PERMANENT;
3000 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3001 if (idev->cnf.optimistic_dad &&
3002 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3003 addr_flags |= IFA_F_OPTIMISTIC;
3006 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3007 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3009 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3010 addrconf_dad_start(ifp);
3015 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3017 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3020 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3021 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3024 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3025 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3031 static int ipv6_generate_stable_address(struct in6_addr *address,
3033 const struct inet6_dev *idev)
3035 static DEFINE_SPINLOCK(lock);
3036 static __u32 digest[SHA_DIGEST_WORDS];
3037 static __u32 workspace[SHA_WORKSPACE_WORDS];
3040 char __data[SHA_MESSAGE_BYTES];
3042 struct in6_addr secret;
3044 unsigned char hwaddr[MAX_ADDR_LEN];
3049 struct in6_addr secret;
3050 struct in6_addr temp;
3051 struct net *net = dev_net(idev->dev);
3053 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3055 if (idev->cnf.stable_secret.initialized)
3056 secret = idev->cnf.stable_secret.secret;
3057 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3058 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3063 spin_lock_bh(&lock);
3066 memset(&data, 0, sizeof(data));
3067 memset(workspace, 0, sizeof(workspace));
3068 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3069 data.prefix[0] = address->s6_addr32[0];
3070 data.prefix[1] = address->s6_addr32[1];
3071 data.secret = secret;
3072 data.dad_count = dad_count;
3074 sha_transform(digest, data.__data, workspace);
3077 temp.s6_addr32[2] = (__force __be32)digest[0];
3078 temp.s6_addr32[3] = (__force __be32)digest[1];
3080 spin_unlock_bh(&lock);
3082 if (ipv6_reserved_interfaceid(temp)) {
3084 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3093 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3095 struct in6_addr addr;
3097 /* no link local addresses on L3 master devices */
3098 if (netif_is_l3_master(idev->dev))
3101 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3103 if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3104 if (!ipv6_generate_stable_address(&addr, 0, idev))
3105 addrconf_add_linklocal(idev, &addr,
3106 IFA_F_STABLE_PRIVACY);
3107 else if (prefix_route)
3108 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3109 } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3110 /* addrconf_add_linklocal also adds a prefix_route and we
3111 * only need to care about prefix routes if ipv6_generate_eui64
3112 * couldn't generate one.
3114 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3115 addrconf_add_linklocal(idev, &addr, 0);
3116 else if (prefix_route)
3117 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3121 static void addrconf_dev_config(struct net_device *dev)
3123 struct inet6_dev *idev;
3127 if ((dev->type != ARPHRD_ETHER) &&
3128 (dev->type != ARPHRD_FDDI) &&
3129 (dev->type != ARPHRD_ARCNET) &&
3130 (dev->type != ARPHRD_INFINIBAND) &&
3131 (dev->type != ARPHRD_IEEE802154) &&
3132 (dev->type != ARPHRD_IEEE1394) &&
3133 (dev->type != ARPHRD_TUNNEL6) &&
3134 (dev->type != ARPHRD_6LOWPAN)) {
3135 /* Alas, we support only Ethernet autoconfiguration. */
3139 idev = addrconf_add_dev(dev);
3143 addrconf_addr_gen(idev, false);
3146 #if IS_ENABLED(CONFIG_IPV6_SIT)
3147 static void addrconf_sit_config(struct net_device *dev)
3149 struct inet6_dev *idev;
3154 * Configure the tunnel with one of our IPv4
3155 * addresses... we should configure all of
3156 * our v4 addrs in the tunnel
3159 idev = ipv6_find_idev(dev);
3161 pr_debug("%s: add_dev failed\n", __func__);
3165 if (dev->priv_flags & IFF_ISATAP) {
3166 addrconf_addr_gen(idev, false);
3170 sit_add_v4_addrs(idev);
3172 if (dev->flags&IFF_POINTOPOINT)
3173 addrconf_add_mroute(dev);
3177 #if IS_ENABLED(CONFIG_NET_IPGRE)
3178 static void addrconf_gre_config(struct net_device *dev)
3180 struct inet6_dev *idev;
3184 idev = ipv6_find_idev(dev);
3186 pr_debug("%s: add_dev failed\n", __func__);
3190 addrconf_addr_gen(idev, true);
3191 if (dev->flags & IFF_POINTOPOINT)
3192 addrconf_add_mroute(dev);
3196 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3199 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3200 struct inet6_dev *idev = __in6_dev_get(dev);
3201 struct net *net = dev_net(dev);
3202 int run_pending = 0;
3206 case NETDEV_REGISTER:
3207 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3208 idev = ipv6_add_dev(dev);
3210 return notifier_from_errno(PTR_ERR(idev));
3214 case NETDEV_CHANGEMTU:
3215 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3216 if (dev->mtu < IPV6_MIN_MTU) {
3217 addrconf_ifdown(dev, dev != net->loopback_dev);
3222 rt6_mtu_change(dev, dev->mtu);
3223 idev->cnf.mtu6 = dev->mtu;
3227 /* allocate new idev */
3228 idev = ipv6_add_dev(dev);
3232 /* device is still not ready */
3233 if (!(idev->if_flags & IF_READY))
3242 if (dev->flags & IFF_SLAVE)
3245 if (idev && idev->cnf.disable_ipv6)
3248 if (event == NETDEV_UP) {
3249 if (!addrconf_qdisc_ok(dev)) {
3250 /* device is not ready yet. */
3251 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3256 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3257 idev = ipv6_add_dev(dev);
3259 if (!IS_ERR_OR_NULL(idev)) {
3260 idev->if_flags |= IF_READY;
3263 } else if (event == NETDEV_CHANGE) {
3264 if (!addrconf_qdisc_ok(dev)) {
3265 /* device is still not ready. */
3270 if (idev->if_flags & IF_READY) {
3271 /* device is already configured -
3272 * but resend MLD reports, we might
3273 * have roamed and need to update
3274 * multicast snooping switches
3279 idev->if_flags |= IF_READY;
3282 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3288 switch (dev->type) {
3289 #if IS_ENABLED(CONFIG_IPV6_SIT)
3291 addrconf_sit_config(dev);
3294 #if IS_ENABLED(CONFIG_NET_IPGRE)
3296 addrconf_gre_config(dev);
3299 case ARPHRD_LOOPBACK:
3304 addrconf_dev_config(dev);
3308 if (!IS_ERR_OR_NULL(idev)) {
3310 addrconf_dad_run(idev);
3313 * If the MTU changed during the interface down,
3314 * when the interface up, the changed MTU must be
3315 * reflected in the idev as well as routers.
3317 if (idev->cnf.mtu6 != dev->mtu &&
3318 dev->mtu >= IPV6_MIN_MTU) {
3319 rt6_mtu_change(dev, dev->mtu);
3320 idev->cnf.mtu6 = dev->mtu;
3322 idev->tstamp = jiffies;
3323 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3326 * If the changed mtu during down is lower than
3327 * IPV6_MIN_MTU stop IPv6 on this interface.
3329 if (dev->mtu < IPV6_MIN_MTU)
3330 addrconf_ifdown(dev, dev != net->loopback_dev);
3335 case NETDEV_UNREGISTER:
3337 * Remove all addresses from this interface.
3339 addrconf_ifdown(dev, event != NETDEV_DOWN);
3342 case NETDEV_CHANGENAME:
3344 snmp6_unregister_dev(idev);
3345 addrconf_sysctl_unregister(idev);
3346 err = addrconf_sysctl_register(idev);
3348 return notifier_from_errno(err);
3349 err = snmp6_register_dev(idev);
3351 addrconf_sysctl_unregister(idev);
3352 return notifier_from_errno(err);
3357 case NETDEV_PRE_TYPE_CHANGE:
3358 case NETDEV_POST_TYPE_CHANGE:
3359 addrconf_type_change(dev, event);
3367 * addrconf module should be notified of a device going up
3369 static struct notifier_block ipv6_dev_notf = {
3370 .notifier_call = addrconf_notify,
3371 .priority = ADDRCONF_NOTIFY_PRIORITY,
3374 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3376 struct inet6_dev *idev;
3379 idev = __in6_dev_get(dev);
3381 if (event == NETDEV_POST_TYPE_CHANGE)
3382 ipv6_mc_remap(idev);
3383 else if (event == NETDEV_PRE_TYPE_CHANGE)
3384 ipv6_mc_unmap(idev);
3387 static int addrconf_ifdown(struct net_device *dev, int how)
3389 struct net *net = dev_net(dev);
3390 struct inet6_dev *idev;
3391 struct inet6_ifaddr *ifa;
3396 rt6_ifdown(net, dev);
3397 neigh_ifdown(&nd_tbl, dev);
3399 idev = __in6_dev_get(dev);
3404 * Step 1: remove reference to ipv6 device from parent device.
3410 /* protected by rtnl_lock */
3411 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3413 /* Step 1.5: remove snmp6 entry */
3414 snmp6_unregister_dev(idev);
3418 /* Step 2: clear hash table */
3419 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3420 struct hlist_head *h = &inet6_addr_lst[i];
3422 spin_lock_bh(&addrconf_hash_lock);
3424 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3425 if (ifa->idev == idev) {
3426 hlist_del_init_rcu(&ifa->addr_lst);
3427 addrconf_del_dad_work(ifa);
3431 spin_unlock_bh(&addrconf_hash_lock);
3434 write_lock_bh(&idev->lock);
3436 addrconf_del_rs_timer(idev);
3438 /* Step 2: clear flags for stateless addrconf */
3440 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3442 if (how && del_timer(&idev->regen_timer))
3445 /* Step 3: clear tempaddr list */
3446 while (!list_empty(&idev->tempaddr_list)) {
3447 ifa = list_first_entry(&idev->tempaddr_list,
3448 struct inet6_ifaddr, tmp_list);
3449 list_del(&ifa->tmp_list);
3450 write_unlock_bh(&idev->lock);
3451 spin_lock_bh(&ifa->lock);
3454 in6_ifa_put(ifa->ifpub);
3457 spin_unlock_bh(&ifa->lock);
3459 write_lock_bh(&idev->lock);
3462 while (!list_empty(&idev->addr_list)) {
3463 ifa = list_first_entry(&idev->addr_list,
3464 struct inet6_ifaddr, if_list);
3465 addrconf_del_dad_work(ifa);
3467 list_del(&ifa->if_list);
3469 write_unlock_bh(&idev->lock);
3471 spin_lock_bh(&ifa->lock);
3473 ifa->state = INET6_IFADDR_STATE_DEAD;
3474 spin_unlock_bh(&ifa->lock);
3476 if (state != INET6_IFADDR_STATE_DEAD) {
3477 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3478 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3482 write_lock_bh(&idev->lock);
3485 write_unlock_bh(&idev->lock);
3487 /* Step 5: Discard anycast and multicast list */
3489 ipv6_ac_destroy_dev(idev);
3490 ipv6_mc_destroy_dev(idev);
3495 idev->tstamp = jiffies;
3497 /* Last: Shot the device (if unregistered) */
3499 addrconf_sysctl_unregister(idev);
3500 neigh_parms_release(&nd_tbl, idev->nd_parms);
3501 neigh_ifdown(&nd_tbl, dev);
3507 static void addrconf_rs_timer(unsigned long data)
3509 struct inet6_dev *idev = (struct inet6_dev *)data;
3510 struct net_device *dev = idev->dev;
3511 struct in6_addr lladdr;
3513 write_lock(&idev->lock);
3514 if (idev->dead || !(idev->if_flags & IF_READY))
3517 if (!ipv6_accept_ra(idev))
3520 /* Announcement received after solicitation was sent */
3521 if (idev->if_flags & IF_RA_RCVD)
3524 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3525 write_unlock(&idev->lock);
3526 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3527 ndisc_send_rs(dev, &lladdr,
3528 &in6addr_linklocal_allrouters);
3532 write_lock(&idev->lock);
3533 idev->rs_interval = rfc3315_s14_backoff_update(
3534 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3535 /* The wait after the last probe can be shorter */
3536 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3537 idev->cnf.rtr_solicits) ?
3538 idev->cnf.rtr_solicit_delay :
3542 * Note: we do not support deprecated "all on-link"
3543 * assumption any longer.
3545 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3549 write_unlock(&idev->lock);
3555 * Duplicate Address Detection
3557 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3559 unsigned long rand_num;
3560 struct inet6_dev *idev = ifp->idev;
3562 if (ifp->flags & IFA_F_OPTIMISTIC)
3565 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3567 ifp->dad_probes = idev->cnf.dad_transmits;
3568 addrconf_mod_dad_work(ifp, rand_num);
3571 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3573 struct inet6_dev *idev = ifp->idev;
3574 struct net_device *dev = idev->dev;
3575 bool notify = false;
3577 addrconf_join_solict(dev, &ifp->addr);
3579 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3581 read_lock_bh(&idev->lock);
3582 spin_lock(&ifp->lock);
3583 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3586 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3587 idev->cnf.accept_dad < 1 ||
3588 !(ifp->flags&IFA_F_TENTATIVE) ||
3589 ifp->flags & IFA_F_NODAD) {
3590 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3591 spin_unlock(&ifp->lock);
3592 read_unlock_bh(&idev->lock);
3594 addrconf_dad_completed(ifp);
3598 if (!(idev->if_flags & IF_READY)) {
3599 spin_unlock(&ifp->lock);
3600 read_unlock_bh(&idev->lock);
3602 * If the device is not ready:
3603 * - keep it tentative if it is a permanent address.
3604 * - otherwise, kill it.
3607 addrconf_dad_stop(ifp, 0);
3612 * Optimistic nodes can start receiving
3615 if (ifp->flags & IFA_F_OPTIMISTIC) {
3616 ip6_ins_rt(ifp->rt);
3617 if (ipv6_use_optimistic_addr(idev)) {
3618 /* Because optimistic nodes can use this address,
3619 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3625 addrconf_dad_kick(ifp);
3627 spin_unlock(&ifp->lock);
3628 read_unlock_bh(&idev->lock);
3630 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3633 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3635 bool begin_dad = false;
3637 spin_lock_bh(&ifp->lock);
3638 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3639 ifp->state = INET6_IFADDR_STATE_PREDAD;
3642 spin_unlock_bh(&ifp->lock);
3645 addrconf_mod_dad_work(ifp, 0);
3648 static void addrconf_dad_work(struct work_struct *w)
3650 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3651 struct inet6_ifaddr,
3653 struct inet6_dev *idev = ifp->idev;
3654 struct in6_addr mcaddr;
3660 } action = DAD_PROCESS;
3664 spin_lock_bh(&ifp->lock);
3665 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3667 ifp->state = INET6_IFADDR_STATE_DAD;
3668 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3670 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3672 spin_unlock_bh(&ifp->lock);
3674 if (action == DAD_BEGIN) {
3675 addrconf_dad_begin(ifp);
3677 } else if (action == DAD_ABORT) {
3679 addrconf_dad_stop(ifp, 1);
3683 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3686 write_lock_bh(&idev->lock);
3687 if (idev->dead || !(idev->if_flags & IF_READY)) {
3688 write_unlock_bh(&idev->lock);
3692 spin_lock(&ifp->lock);
3693 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3694 spin_unlock(&ifp->lock);
3695 write_unlock_bh(&idev->lock);
3699 if (ifp->dad_probes == 0) {
3701 * DAD was successful
3704 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3705 spin_unlock(&ifp->lock);
3706 write_unlock_bh(&idev->lock);
3708 addrconf_dad_completed(ifp);
3714 addrconf_mod_dad_work(ifp,
3715 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3716 spin_unlock(&ifp->lock);
3717 write_unlock_bh(&idev->lock);
3719 /* send a neighbour solicitation for our addr */
3720 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3721 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3727 /* ifp->idev must be at least read locked */
3728 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3730 struct inet6_ifaddr *ifpiter;
3731 struct inet6_dev *idev = ifp->idev;
3733 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3734 if (ifpiter->scope > IFA_LINK)
3736 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3737 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3738 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3745 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3747 struct net_device *dev = ifp->idev->dev;
3748 struct in6_addr lladdr;
3749 bool send_rs, send_mld;
3751 addrconf_del_dad_work(ifp);
3754 * Configure the address for reception. Now it is valid.
3757 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3759 /* If added prefix is link local and we are prepared to process
3760 router advertisements, start sending router solicitations.
3763 read_lock_bh(&ifp->idev->lock);
3764 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3765 send_rs = send_mld &&
3766 ipv6_accept_ra(ifp->idev) &&
3767 ifp->idev->cnf.rtr_solicits != 0 &&
3768 (dev->flags&IFF_LOOPBACK) == 0;
3769 read_unlock_bh(&ifp->idev->lock);
3771 /* While dad is in progress mld report's source address is in6_addrany.
3772 * Resend with proper ll now.
3775 ipv6_mc_dad_complete(ifp->idev);
3779 * If a host as already performed a random delay
3780 * [...] as part of DAD [...] there is no need
3781 * to delay again before sending the first RS
3783 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3785 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3787 write_lock_bh(&ifp->idev->lock);
3788 spin_lock(&ifp->lock);
3789 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
3790 ifp->idev->cnf.rtr_solicit_interval);
3791 ifp->idev->rs_probes = 1;
3792 ifp->idev->if_flags |= IF_RS_SENT;
3793 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
3794 spin_unlock(&ifp->lock);
3795 write_unlock_bh(&ifp->idev->lock);
3799 static void addrconf_dad_run(struct inet6_dev *idev)
3801 struct inet6_ifaddr *ifp;
3803 read_lock_bh(&idev->lock);
3804 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3805 spin_lock(&ifp->lock);
3806 if (ifp->flags & IFA_F_TENTATIVE &&
3807 ifp->state == INET6_IFADDR_STATE_DAD)
3808 addrconf_dad_kick(ifp);
3809 spin_unlock(&ifp->lock);
3811 read_unlock_bh(&idev->lock);
3814 #ifdef CONFIG_PROC_FS
3815 struct if6_iter_state {
3816 struct seq_net_private p;
3821 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3823 struct inet6_ifaddr *ifa = NULL;
3824 struct if6_iter_state *state = seq->private;
3825 struct net *net = seq_file_net(seq);
3828 /* initial bucket if pos is 0 */
3834 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3835 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3837 if (!net_eq(dev_net(ifa->idev->dev), net))
3839 /* sync with offset */
3840 if (p < state->offset) {
3847 /* prepare for next bucket */
3854 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3855 struct inet6_ifaddr *ifa)
3857 struct if6_iter_state *state = seq->private;
3858 struct net *net = seq_file_net(seq);
3860 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3861 if (!net_eq(dev_net(ifa->idev->dev), net))
3868 while (++state->bucket < IN6_ADDR_HSIZE) {
3869 hlist_for_each_entry_rcu_bh(ifa,
3870 &inet6_addr_lst[state->bucket], addr_lst) {
3871 if (!net_eq(dev_net(ifa->idev->dev), net))
3880 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3884 return if6_get_first(seq, *pos);
3887 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3889 struct inet6_ifaddr *ifa;
3891 ifa = if6_get_next(seq, v);
3896 static void if6_seq_stop(struct seq_file *seq, void *v)
3899 rcu_read_unlock_bh();
3902 static int if6_seq_show(struct seq_file *seq, void *v)
3904 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3905 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3907 ifp->idev->dev->ifindex,
3911 ifp->idev->dev->name);
3915 static const struct seq_operations if6_seq_ops = {
3916 .start = if6_seq_start,
3917 .next = if6_seq_next,
3918 .show = if6_seq_show,
3919 .stop = if6_seq_stop,
3922 static int if6_seq_open(struct inode *inode, struct file *file)
3924 return seq_open_net(inode, file, &if6_seq_ops,
3925 sizeof(struct if6_iter_state));
3928 static const struct file_operations if6_fops = {
3929 .owner = THIS_MODULE,
3930 .open = if6_seq_open,
3932 .llseek = seq_lseek,
3933 .release = seq_release_net,
3936 static int __net_init if6_proc_net_init(struct net *net)
3938 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3943 static void __net_exit if6_proc_net_exit(struct net *net)
3945 remove_proc_entry("if_inet6", net->proc_net);
3948 static struct pernet_operations if6_proc_net_ops = {
3949 .init = if6_proc_net_init,
3950 .exit = if6_proc_net_exit,
3953 int __init if6_proc_init(void)
3955 return register_pernet_subsys(&if6_proc_net_ops);
3958 void if6_proc_exit(void)
3960 unregister_pernet_subsys(&if6_proc_net_ops);
3962 #endif /* CONFIG_PROC_FS */
3964 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3965 /* Check if address is a home address configured on any interface. */
3966 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3969 struct inet6_ifaddr *ifp = NULL;
3970 unsigned int hash = inet6_addr_hash(addr);
3973 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3974 if (!net_eq(dev_net(ifp->idev->dev), net))
3976 if (ipv6_addr_equal(&ifp->addr, addr) &&
3977 (ifp->flags & IFA_F_HOMEADDRESS)) {
3982 rcu_read_unlock_bh();
3988 * Periodic address status verification
3991 static void addrconf_verify_rtnl(void)
3993 unsigned long now, next, next_sec, next_sched;
3994 struct inet6_ifaddr *ifp;
4001 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4003 cancel_delayed_work(&addr_chk_work);
4005 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4007 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4010 /* When setting preferred_lft to a value not zero or
4011 * infinity, while valid_lft is infinity
4012 * IFA_F_PERMANENT has a non-infinity life time.
4014 if ((ifp->flags & IFA_F_PERMANENT) &&
4015 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4018 spin_lock(&ifp->lock);
4019 /* We try to batch several events at once. */
4020 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4022 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4023 age >= ifp->valid_lft) {
4024 spin_unlock(&ifp->lock);
4028 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4029 spin_unlock(&ifp->lock);
4031 } else if (age >= ifp->prefered_lft) {
4032 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4035 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4037 ifp->flags |= IFA_F_DEPRECATED;
4040 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4041 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4042 next = ifp->tstamp + ifp->valid_lft * HZ;
4044 spin_unlock(&ifp->lock);
4049 ipv6_ifa_notify(0, ifp);
4053 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4054 !(ifp->flags&IFA_F_TENTATIVE)) {
4055 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4056 ifp->idev->cnf.dad_transmits *
4057 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4059 if (age >= ifp->prefered_lft - regen_advance) {
4060 struct inet6_ifaddr *ifpub = ifp->ifpub;
4061 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4062 next = ifp->tstamp + ifp->prefered_lft * HZ;
4063 if (!ifp->regen_count && ifpub) {
4066 in6_ifa_hold(ifpub);
4067 spin_unlock(&ifp->lock);
4069 spin_lock(&ifpub->lock);
4070 ifpub->regen_count = 0;
4071 spin_unlock(&ifpub->lock);
4072 ipv6_create_tempaddr(ifpub, ifp);
4077 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4078 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4079 spin_unlock(&ifp->lock);
4081 /* ifp->prefered_lft <= ifp->valid_lft */
4082 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4083 next = ifp->tstamp + ifp->prefered_lft * HZ;
4084 spin_unlock(&ifp->lock);
4089 next_sec = round_jiffies_up(next);
4092 /* If rounded timeout is accurate enough, accept it. */
4093 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4094 next_sched = next_sec;
4096 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4097 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4098 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4100 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4101 now, next, next_sec, next_sched);
4102 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4103 rcu_read_unlock_bh();
4106 static void addrconf_verify_work(struct work_struct *w)
4109 addrconf_verify_rtnl();
4113 static void addrconf_verify(void)
4115 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4118 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4119 struct in6_addr **peer_pfx)
4121 struct in6_addr *pfx = NULL;
4126 pfx = nla_data(addr);
4129 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4131 pfx = nla_data(local);
4137 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4138 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4139 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4140 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4141 [IFA_FLAGS] = { .len = sizeof(u32) },
4145 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4147 struct net *net = sock_net(skb->sk);
4148 struct ifaddrmsg *ifm;
4149 struct nlattr *tb[IFA_MAX+1];
4150 struct in6_addr *pfx, *peer_pfx;
4154 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4158 ifm = nlmsg_data(nlh);
4159 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4163 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4165 /* We ignore other flags so far. */
4166 ifa_flags &= IFA_F_MANAGETEMPADDR;
4168 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4169 ifm->ifa_prefixlen);
4172 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4173 u32 prefered_lft, u32 valid_lft)
4177 unsigned long timeout;
4178 bool was_managetempaddr;
4179 bool had_prefixroute;
4183 if (!valid_lft || (prefered_lft > valid_lft))
4186 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4187 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4190 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4191 if (addrconf_finite_timeout(timeout)) {
4192 expires = jiffies_to_clock_t(timeout * HZ);
4193 valid_lft = timeout;
4194 flags = RTF_EXPIRES;
4198 ifa_flags |= IFA_F_PERMANENT;
4201 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4202 if (addrconf_finite_timeout(timeout)) {
4204 ifa_flags |= IFA_F_DEPRECATED;
4205 prefered_lft = timeout;
4208 spin_lock_bh(&ifp->lock);
4209 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4210 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4211 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4212 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4213 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4214 IFA_F_NOPREFIXROUTE);
4215 ifp->flags |= ifa_flags;
4216 ifp->tstamp = jiffies;
4217 ifp->valid_lft = valid_lft;
4218 ifp->prefered_lft = prefered_lft;
4220 spin_unlock_bh(&ifp->lock);
4221 if (!(ifp->flags&IFA_F_TENTATIVE))
4222 ipv6_ifa_notify(0, ifp);
4224 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4225 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4227 } else if (had_prefixroute) {
4228 enum cleanup_prefix_rt_t action;
4229 unsigned long rt_expires;
4231 write_lock_bh(&ifp->idev->lock);
4232 action = check_cleanup_prefix_route(ifp, &rt_expires);
4233 write_unlock_bh(&ifp->idev->lock);
4235 if (action != CLEANUP_PREFIX_RT_NOP) {
4236 cleanup_prefix_route(ifp, rt_expires,
4237 action == CLEANUP_PREFIX_RT_DEL);
4241 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4242 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4243 valid_lft = prefered_lft = 0;
4244 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4245 !was_managetempaddr, jiffies);
4248 addrconf_verify_rtnl();
4254 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4256 struct net *net = sock_net(skb->sk);
4257 struct ifaddrmsg *ifm;
4258 struct nlattr *tb[IFA_MAX+1];
4259 struct in6_addr *pfx, *peer_pfx;
4260 struct inet6_ifaddr *ifa;
4261 struct net_device *dev;
4262 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4266 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4270 ifm = nlmsg_data(nlh);
4271 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4275 if (tb[IFA_CACHEINFO]) {
4276 struct ifa_cacheinfo *ci;
4278 ci = nla_data(tb[IFA_CACHEINFO]);
4279 valid_lft = ci->ifa_valid;
4280 preferred_lft = ci->ifa_prefered;
4282 preferred_lft = INFINITY_LIFE_TIME;
4283 valid_lft = INFINITY_LIFE_TIME;
4286 dev = __dev_get_by_index(net, ifm->ifa_index);
4290 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4292 /* We ignore other flags so far. */
4293 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4294 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4296 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4299 * It would be best to check for !NLM_F_CREATE here but
4300 * userspace already relies on not having to provide this.
4302 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4303 ifm->ifa_prefixlen, ifa_flags,
4304 preferred_lft, valid_lft);
4307 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4308 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4311 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4318 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4319 u8 scope, int ifindex)
4321 struct ifaddrmsg *ifm;
4323 ifm = nlmsg_data(nlh);
4324 ifm->ifa_family = AF_INET6;
4325 ifm->ifa_prefixlen = prefixlen;
4326 ifm->ifa_flags = flags;
4327 ifm->ifa_scope = scope;
4328 ifm->ifa_index = ifindex;
4331 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4332 unsigned long tstamp, u32 preferred, u32 valid)
4334 struct ifa_cacheinfo ci;
4336 ci.cstamp = cstamp_delta(cstamp);
4337 ci.tstamp = cstamp_delta(tstamp);
4338 ci.ifa_prefered = preferred;
4339 ci.ifa_valid = valid;
4341 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4344 static inline int rt_scope(int ifa_scope)
4346 if (ifa_scope & IFA_HOST)
4347 return RT_SCOPE_HOST;
4348 else if (ifa_scope & IFA_LINK)
4349 return RT_SCOPE_LINK;
4350 else if (ifa_scope & IFA_SITE)
4351 return RT_SCOPE_SITE;
4353 return RT_SCOPE_UNIVERSE;
4356 static inline int inet6_ifaddr_msgsize(void)
4358 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4359 + nla_total_size(16) /* IFA_LOCAL */
4360 + nla_total_size(16) /* IFA_ADDRESS */
4361 + nla_total_size(sizeof(struct ifa_cacheinfo))
4362 + nla_total_size(4) /* IFA_FLAGS */;
4365 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4366 u32 portid, u32 seq, int event, unsigned int flags)
4368 struct nlmsghdr *nlh;
4369 u32 preferred, valid;
4371 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4375 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4376 ifa->idev->dev->ifindex);
4378 if (!((ifa->flags&IFA_F_PERMANENT) &&
4379 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4380 preferred = ifa->prefered_lft;
4381 valid = ifa->valid_lft;
4382 if (preferred != INFINITY_LIFE_TIME) {
4383 long tval = (jiffies - ifa->tstamp)/HZ;
4384 if (preferred > tval)
4388 if (valid != INFINITY_LIFE_TIME) {
4396 preferred = INFINITY_LIFE_TIME;
4397 valid = INFINITY_LIFE_TIME;
4400 if (!ipv6_addr_any(&ifa->peer_addr)) {
4401 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4402 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4405 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4408 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4411 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4414 nlmsg_end(skb, nlh);
4418 nlmsg_cancel(skb, nlh);
4422 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4423 u32 portid, u32 seq, int event, u16 flags)
4425 struct nlmsghdr *nlh;
4426 u8 scope = RT_SCOPE_UNIVERSE;
4427 int ifindex = ifmca->idev->dev->ifindex;
4429 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4430 scope = RT_SCOPE_SITE;
4432 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4436 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4437 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4438 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4439 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4440 nlmsg_cancel(skb, nlh);
4444 nlmsg_end(skb, nlh);
4448 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4449 u32 portid, u32 seq, int event, unsigned int flags)
4451 struct nlmsghdr *nlh;
4452 u8 scope = RT_SCOPE_UNIVERSE;
4453 int ifindex = ifaca->aca_idev->dev->ifindex;
4455 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4456 scope = RT_SCOPE_SITE;
4458 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4462 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4463 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4464 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4465 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4466 nlmsg_cancel(skb, nlh);
4470 nlmsg_end(skb, nlh);
4480 /* called with rcu_read_lock() */
4481 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4482 struct netlink_callback *cb, enum addr_type_t type,
4483 int s_ip_idx, int *p_ip_idx)
4485 struct ifmcaddr6 *ifmca;
4486 struct ifacaddr6 *ifaca;
4488 int ip_idx = *p_ip_idx;
4490 read_lock_bh(&idev->lock);
4492 case UNICAST_ADDR: {
4493 struct inet6_ifaddr *ifa;
4495 /* unicast address incl. temp addr */
4496 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4497 if (++ip_idx < s_ip_idx)
4499 err = inet6_fill_ifaddr(skb, ifa,
4500 NETLINK_CB(cb->skb).portid,
4506 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4510 case MULTICAST_ADDR:
4511 /* multicast address */
4512 for (ifmca = idev->mc_list; ifmca;
4513 ifmca = ifmca->next, ip_idx++) {
4514 if (ip_idx < s_ip_idx)
4516 err = inet6_fill_ifmcaddr(skb, ifmca,
4517 NETLINK_CB(cb->skb).portid,
4526 /* anycast address */
4527 for (ifaca = idev->ac_list; ifaca;
4528 ifaca = ifaca->aca_next, ip_idx++) {
4529 if (ip_idx < s_ip_idx)
4531 err = inet6_fill_ifacaddr(skb, ifaca,
4532 NETLINK_CB(cb->skb).portid,
4543 read_unlock_bh(&idev->lock);
4548 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4549 enum addr_type_t type)
4551 struct net *net = sock_net(skb->sk);
4554 int s_idx, s_ip_idx;
4555 struct net_device *dev;
4556 struct inet6_dev *idev;
4557 struct hlist_head *head;
4560 s_idx = idx = cb->args[1];
4561 s_ip_idx = ip_idx = cb->args[2];
4564 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4565 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4567 head = &net->dev_index_head[h];
4568 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4571 if (h > s_h || idx > s_idx)
4574 idev = __in6_dev_get(dev);
4578 if (in6_dump_addrs(idev, skb, cb, type,
4579 s_ip_idx, &ip_idx) < 0)
4589 cb->args[2] = ip_idx;
4594 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4596 enum addr_type_t type = UNICAST_ADDR;
4598 return inet6_dump_addr(skb, cb, type);
4601 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4603 enum addr_type_t type = MULTICAST_ADDR;
4605 return inet6_dump_addr(skb, cb, type);
4609 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4611 enum addr_type_t type = ANYCAST_ADDR;
4613 return inet6_dump_addr(skb, cb, type);
4616 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4618 struct net *net = sock_net(in_skb->sk);
4619 struct ifaddrmsg *ifm;
4620 struct nlattr *tb[IFA_MAX+1];
4621 struct in6_addr *addr = NULL, *peer;
4622 struct net_device *dev = NULL;
4623 struct inet6_ifaddr *ifa;
4624 struct sk_buff *skb;
4627 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4631 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4637 ifm = nlmsg_data(nlh);
4639 dev = __dev_get_by_index(net, ifm->ifa_index);
4641 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4643 err = -EADDRNOTAVAIL;
4647 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4653 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4654 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4656 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4657 WARN_ON(err == -EMSGSIZE);
4661 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4668 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4670 struct sk_buff *skb;
4671 struct net *net = dev_net(ifa->idev->dev);
4674 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4678 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4680 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4681 WARN_ON(err == -EMSGSIZE);
4685 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4689 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4692 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4693 __s32 *array, int bytes)
4695 BUG_ON(bytes < (DEVCONF_MAX * 4));
4697 memset(array, 0, bytes);
4698 array[DEVCONF_FORWARDING] = cnf->forwarding;
4699 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4700 array[DEVCONF_MTU6] = cnf->mtu6;
4701 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4702 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4703 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4704 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4705 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4706 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4707 jiffies_to_msecs(cnf->rtr_solicit_interval);
4708 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4709 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4710 array[DEVCONF_RTR_SOLICIT_DELAY] =
4711 jiffies_to_msecs(cnf->rtr_solicit_delay);
4712 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4713 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4714 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4715 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4716 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4717 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4718 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4719 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4720 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4721 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4722 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4723 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4724 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4725 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4726 #ifdef CONFIG_IPV6_ROUTER_PREF
4727 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4728 array[DEVCONF_RTR_PROBE_INTERVAL] =
4729 jiffies_to_msecs(cnf->rtr_probe_interval);
4730 #ifdef CONFIG_IPV6_ROUTE_INFO
4731 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
4732 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4735 array[DEVCONF_ACCEPT_RA_RT_TABLE] = cnf->accept_ra_rt_table;
4736 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4737 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4738 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4739 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4740 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4742 #ifdef CONFIG_IPV6_MROUTE
4743 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4745 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4746 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4747 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4748 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4749 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4750 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4751 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4752 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4753 /* we omit DEVCONF_STABLE_SECRET for now */
4754 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4757 static inline size_t inet6_ifla6_size(void)
4759 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4760 + nla_total_size(sizeof(struct ifla_cacheinfo))
4761 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4762 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4763 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4764 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4767 static inline size_t inet6_if_nlmsg_size(void)
4769 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4770 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4771 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4772 + nla_total_size(4) /* IFLA_MTU */
4773 + nla_total_size(4) /* IFLA_LINK */
4774 + nla_total_size(1) /* IFLA_OPERSTATE */
4775 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4778 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4779 int items, int bytes)
4782 int pad = bytes - sizeof(u64) * items;
4785 /* Use put_unaligned() because stats may not be aligned for u64. */
4786 put_unaligned(items, &stats[0]);
4787 for (i = 1; i < items; i++)
4788 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4790 memset(&stats[items], 0, pad);
4793 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4794 int bytes, size_t syncpoff)
4797 u64 buff[IPSTATS_MIB_MAX];
4798 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4802 memset(buff, 0, sizeof(buff));
4803 buff[0] = IPSTATS_MIB_MAX;
4805 for_each_possible_cpu(c) {
4806 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4807 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4810 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4811 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4814 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4818 case IFLA_INET6_STATS:
4819 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4820 offsetof(struct ipstats_mib, syncp));
4822 case IFLA_INET6_ICMP6STATS:
4823 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4828 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4829 u32 ext_filter_mask)
4832 struct ifla_cacheinfo ci;
4834 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4835 goto nla_put_failure;
4836 ci.max_reasm_len = IPV6_MAXPLEN;
4837 ci.tstamp = cstamp_delta(idev->tstamp);
4838 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4839 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4840 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4841 goto nla_put_failure;
4842 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4844 goto nla_put_failure;
4845 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4847 /* XXX - MC not implemented */
4849 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4852 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4854 goto nla_put_failure;
4855 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4857 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4859 goto nla_put_failure;
4860 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4862 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4864 goto nla_put_failure;
4866 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4867 goto nla_put_failure;
4869 read_lock_bh(&idev->lock);
4870 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4871 read_unlock_bh(&idev->lock);
4879 static size_t inet6_get_link_af_size(const struct net_device *dev,
4880 u32 ext_filter_mask)
4882 if (!__in6_dev_get(dev))
4885 return inet6_ifla6_size();
4888 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4889 u32 ext_filter_mask)
4891 struct inet6_dev *idev = __in6_dev_get(dev);
4896 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4902 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4904 struct inet6_ifaddr *ifp;
4905 struct net_device *dev = idev->dev;
4906 bool update_rs = false;
4907 struct in6_addr ll_addr;
4913 if (ipv6_addr_any(token))
4915 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4917 if (!ipv6_accept_ra(idev))
4919 if (idev->cnf.rtr_solicits == 0)
4922 write_lock_bh(&idev->lock);
4924 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4925 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4927 write_unlock_bh(&idev->lock);
4929 if (!idev->dead && (idev->if_flags & IF_READY) &&
4930 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4931 IFA_F_OPTIMISTIC)) {
4933 /* If we're not ready, then normal ifup will take care
4934 * of this. Otherwise, we need to request our rs here.
4936 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4940 write_lock_bh(&idev->lock);
4943 idev->if_flags |= IF_RS_SENT;
4944 idev->rs_interval = rfc3315_s14_backoff_init(
4945 idev->cnf.rtr_solicit_interval);
4946 idev->rs_probes = 1;
4947 addrconf_mod_rs_timer(idev, idev->rs_interval);
4950 /* Well, that's kinda nasty ... */
4951 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4952 spin_lock(&ifp->lock);
4953 if (ifp->tokenized) {
4955 ifp->prefered_lft = 0;
4957 spin_unlock(&ifp->lock);
4960 write_unlock_bh(&idev->lock);
4961 inet6_ifinfo_notify(RTM_NEWLINK, idev);
4962 addrconf_verify_rtnl();
4966 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4967 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
4968 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
4971 static int inet6_validate_link_af(const struct net_device *dev,
4972 const struct nlattr *nla)
4974 struct nlattr *tb[IFLA_INET6_MAX + 1];
4976 if (dev && !__in6_dev_get(dev))
4977 return -EAFNOSUPPORT;
4979 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4982 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4985 struct inet6_dev *idev = __in6_dev_get(dev);
4986 struct nlattr *tb[IFLA_INET6_MAX + 1];
4989 return -EAFNOSUPPORT;
4991 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4994 if (tb[IFLA_INET6_TOKEN]) {
4995 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5000 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5001 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5003 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5004 mode != IN6_ADDR_GEN_MODE_NONE &&
5005 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
5008 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5009 !idev->cnf.stable_secret.initialized &&
5010 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5013 idev->addr_gen_mode = mode;
5020 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5021 u32 portid, u32 seq, int event, unsigned int flags)
5023 struct net_device *dev = idev->dev;
5024 struct ifinfomsg *hdr;
5025 struct nlmsghdr *nlh;
5028 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5032 hdr = nlmsg_data(nlh);
5033 hdr->ifi_family = AF_INET6;
5035 hdr->ifi_type = dev->type;
5036 hdr->ifi_index = dev->ifindex;
5037 hdr->ifi_flags = dev_get_flags(dev);
5038 hdr->ifi_change = 0;
5040 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5042 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5043 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5044 (dev->ifindex != dev_get_iflink(dev) &&
5045 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5046 nla_put_u8(skb, IFLA_OPERSTATE,
5047 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5048 goto nla_put_failure;
5049 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5051 goto nla_put_failure;
5053 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5054 goto nla_put_failure;
5056 nla_nest_end(skb, protoinfo);
5057 nlmsg_end(skb, nlh);
5061 nlmsg_cancel(skb, nlh);
5065 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5067 struct net *net = sock_net(skb->sk);
5070 struct net_device *dev;
5071 struct inet6_dev *idev;
5072 struct hlist_head *head;
5075 s_idx = cb->args[1];
5078 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5080 head = &net->dev_index_head[h];
5081 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5084 idev = __in6_dev_get(dev);
5087 if (inet6_fill_ifinfo(skb, idev,
5088 NETLINK_CB(cb->skb).portid,
5090 RTM_NEWLINK, NLM_F_MULTI) < 0)
5104 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5106 struct sk_buff *skb;
5107 struct net *net = dev_net(idev->dev);
5110 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5114 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5116 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5117 WARN_ON(err == -EMSGSIZE);
5121 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5125 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5128 static inline size_t inet6_prefix_nlmsg_size(void)
5130 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5131 + nla_total_size(sizeof(struct in6_addr))
5132 + nla_total_size(sizeof(struct prefix_cacheinfo));
5135 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5136 struct prefix_info *pinfo, u32 portid, u32 seq,
5137 int event, unsigned int flags)
5139 struct prefixmsg *pmsg;
5140 struct nlmsghdr *nlh;
5141 struct prefix_cacheinfo ci;
5143 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5147 pmsg = nlmsg_data(nlh);
5148 pmsg->prefix_family = AF_INET6;
5149 pmsg->prefix_pad1 = 0;
5150 pmsg->prefix_pad2 = 0;
5151 pmsg->prefix_ifindex = idev->dev->ifindex;
5152 pmsg->prefix_len = pinfo->prefix_len;
5153 pmsg->prefix_type = pinfo->type;
5154 pmsg->prefix_pad3 = 0;
5155 pmsg->prefix_flags = 0;
5157 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5158 if (pinfo->autoconf)
5159 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5161 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5162 goto nla_put_failure;
5163 ci.preferred_time = ntohl(pinfo->prefered);
5164 ci.valid_time = ntohl(pinfo->valid);
5165 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5166 goto nla_put_failure;
5167 nlmsg_end(skb, nlh);
5171 nlmsg_cancel(skb, nlh);
5175 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5176 struct prefix_info *pinfo)
5178 struct sk_buff *skb;
5179 struct net *net = dev_net(idev->dev);
5182 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5186 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5188 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5189 WARN_ON(err == -EMSGSIZE);
5193 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5197 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5200 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5202 struct net *net = dev_net(ifp->idev->dev);
5207 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5212 * If the address was optimistic
5213 * we inserted the route at the start of
5214 * our DAD process, so we don't need
5217 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5218 ip6_ins_rt(ifp->rt);
5219 if (ifp->idev->cnf.forwarding)
5220 addrconf_join_anycast(ifp);
5221 if (!ipv6_addr_any(&ifp->peer_addr))
5222 addrconf_prefix_route(&ifp->peer_addr, 128,
5223 ifp->idev->dev, 0, 0);
5226 if (ifp->idev->cnf.forwarding)
5227 addrconf_leave_anycast(ifp);
5228 addrconf_leave_solict(ifp->idev, &ifp->addr);
5229 if (!ipv6_addr_any(&ifp->peer_addr)) {
5230 struct rt6_info *rt;
5232 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5233 ifp->idev->dev, 0, 0);
5237 dst_hold(&ifp->rt->dst);
5239 ip6_del_rt(ifp->rt);
5241 rt_genid_bump_ipv6(net);
5244 atomic_inc(&net->ipv6.dev_addr_genid);
5247 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5250 if (likely(ifp->idev->dead == 0))
5251 __ipv6_ifa_notify(event, ifp);
5252 rcu_read_unlock_bh();
5255 #ifdef CONFIG_SYSCTL
5258 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5259 void __user *buffer, size_t *lenp, loff_t *ppos)
5261 int *valp = ctl->data;
5264 struct ctl_table lctl;
5268 * ctl->data points to idev->cnf.forwarding, we should
5269 * not modify it until we get the rtnl lock.
5274 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5277 ret = addrconf_fixup_forwarding(ctl, valp, val);
5284 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5285 void __user *buffer, size_t *lenp, loff_t *ppos)
5287 struct inet6_dev *idev = ctl->extra1;
5288 int min_mtu = IPV6_MIN_MTU;
5289 struct ctl_table lctl;
5292 lctl.extra1 = &min_mtu;
5293 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5295 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5298 static void dev_disable_change(struct inet6_dev *idev)
5300 struct netdev_notifier_info info;
5302 if (!idev || !idev->dev)
5305 netdev_notifier_info_init(&info, idev->dev);
5306 if (idev->cnf.disable_ipv6)
5307 addrconf_notify(NULL, NETDEV_DOWN, &info);
5309 addrconf_notify(NULL, NETDEV_UP, &info);
5312 static void addrconf_disable_change(struct net *net, __s32 newf)
5314 struct net_device *dev;
5315 struct inet6_dev *idev;
5317 for_each_netdev(net, dev) {
5318 idev = __in6_dev_get(dev);
5320 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5321 idev->cnf.disable_ipv6 = newf;
5323 dev_disable_change(idev);
5328 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5333 if (!rtnl_trylock())
5334 return restart_syscall();
5336 net = (struct net *)table->extra2;
5340 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5345 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5346 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5347 addrconf_disable_change(net, newf);
5348 } else if ((!newf) ^ (!old))
5349 dev_disable_change((struct inet6_dev *)table->extra1);
5356 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5357 void __user *buffer, size_t *lenp, loff_t *ppos)
5359 int *valp = ctl->data;
5362 struct ctl_table lctl;
5366 * ctl->data points to idev->cnf.disable_ipv6, we should
5367 * not modify it until we get the rtnl lock.
5372 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5375 ret = addrconf_disable_ipv6(ctl, valp, val);
5382 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5383 void __user *buffer, size_t *lenp, loff_t *ppos)
5385 int *valp = ctl->data;
5390 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5393 if (write && old != new) {
5394 struct net *net = ctl->extra2;
5396 if (!rtnl_trylock())
5397 return restart_syscall();
5399 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5400 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5401 NETCONFA_IFINDEX_DEFAULT,
5402 net->ipv6.devconf_dflt);
5403 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5404 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5405 NETCONFA_IFINDEX_ALL,
5406 net->ipv6.devconf_all);
5408 struct inet6_dev *idev = ctl->extra1;
5410 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5420 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5421 void __user *buffer, size_t *lenp,
5425 struct in6_addr addr;
5426 char str[IPV6_MAX_STRLEN];
5427 struct ctl_table lctl = *ctl;
5428 struct net *net = ctl->extra2;
5429 struct ipv6_stable_secret *secret = ctl->data;
5431 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5434 lctl.maxlen = IPV6_MAX_STRLEN;
5437 if (!rtnl_trylock())
5438 return restart_syscall();
5440 if (!write && !secret->initialized) {
5445 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5446 if (err >= sizeof(str)) {
5451 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5455 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5460 secret->initialized = true;
5461 secret->secret = addr;
5463 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5464 struct net_device *dev;
5466 for_each_netdev(net, dev) {
5467 struct inet6_dev *idev = __in6_dev_get(dev);
5470 idev->addr_gen_mode =
5471 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5475 struct inet6_dev *idev = ctl->extra1;
5477 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5487 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5489 void __user *buffer,
5493 int *valp = ctl->data;
5496 struct ctl_table lctl;
5499 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5500 * we should not modify it until we get the rtnl lock.
5505 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5508 ret = addrconf_fixup_linkdown(ctl, valp, val);
5514 static struct addrconf_sysctl_table
5516 struct ctl_table_header *sysctl_header;
5517 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5518 } addrconf_sysctl __read_mostly = {
5519 .sysctl_header = NULL,
5522 .procname = "forwarding",
5523 .data = &ipv6_devconf.forwarding,
5524 .maxlen = sizeof(int),
5526 .proc_handler = addrconf_sysctl_forward,
5529 .procname = "hop_limit",
5530 .data = &ipv6_devconf.hop_limit,
5531 .maxlen = sizeof(int),
5533 .proc_handler = proc_dointvec,
5537 .data = &ipv6_devconf.mtu6,
5538 .maxlen = sizeof(int),
5540 .proc_handler = addrconf_sysctl_mtu,
5543 .procname = "accept_ra",
5544 .data = &ipv6_devconf.accept_ra,
5545 .maxlen = sizeof(int),
5547 .proc_handler = proc_dointvec,
5550 .procname = "accept_redirects",
5551 .data = &ipv6_devconf.accept_redirects,
5552 .maxlen = sizeof(int),
5554 .proc_handler = proc_dointvec,
5557 .procname = "autoconf",
5558 .data = &ipv6_devconf.autoconf,
5559 .maxlen = sizeof(int),
5561 .proc_handler = proc_dointvec,
5564 .procname = "dad_transmits",
5565 .data = &ipv6_devconf.dad_transmits,
5566 .maxlen = sizeof(int),
5568 .proc_handler = proc_dointvec,
5571 .procname = "router_solicitations",
5572 .data = &ipv6_devconf.rtr_solicits,
5573 .maxlen = sizeof(int),
5575 .proc_handler = proc_dointvec,
5578 .procname = "router_solicitation_interval",
5579 .data = &ipv6_devconf.rtr_solicit_interval,
5580 .maxlen = sizeof(int),
5582 .proc_handler = proc_dointvec_jiffies,
5585 .procname = "router_solicitation_max_interval",
5586 .data = &ipv6_devconf.rtr_solicit_max_interval,
5587 .maxlen = sizeof(int),
5589 .proc_handler = proc_dointvec_jiffies,
5592 .procname = "router_solicitation_delay",
5593 .data = &ipv6_devconf.rtr_solicit_delay,
5594 .maxlen = sizeof(int),
5596 .proc_handler = proc_dointvec_jiffies,
5599 .procname = "force_mld_version",
5600 .data = &ipv6_devconf.force_mld_version,
5601 .maxlen = sizeof(int),
5603 .proc_handler = proc_dointvec,
5606 .procname = "mldv1_unsolicited_report_interval",
5608 &ipv6_devconf.mldv1_unsolicited_report_interval,
5609 .maxlen = sizeof(int),
5611 .proc_handler = proc_dointvec_ms_jiffies,
5614 .procname = "mldv2_unsolicited_report_interval",
5616 &ipv6_devconf.mldv2_unsolicited_report_interval,
5617 .maxlen = sizeof(int),
5619 .proc_handler = proc_dointvec_ms_jiffies,
5622 .procname = "use_tempaddr",
5623 .data = &ipv6_devconf.use_tempaddr,
5624 .maxlen = sizeof(int),
5626 .proc_handler = proc_dointvec,
5629 .procname = "temp_valid_lft",
5630 .data = &ipv6_devconf.temp_valid_lft,
5631 .maxlen = sizeof(int),
5633 .proc_handler = proc_dointvec,
5636 .procname = "temp_prefered_lft",
5637 .data = &ipv6_devconf.temp_prefered_lft,
5638 .maxlen = sizeof(int),
5640 .proc_handler = proc_dointvec,
5643 .procname = "regen_max_retry",
5644 .data = &ipv6_devconf.regen_max_retry,
5645 .maxlen = sizeof(int),
5647 .proc_handler = proc_dointvec,
5650 .procname = "max_desync_factor",
5651 .data = &ipv6_devconf.max_desync_factor,
5652 .maxlen = sizeof(int),
5654 .proc_handler = proc_dointvec,
5657 .procname = "max_addresses",
5658 .data = &ipv6_devconf.max_addresses,
5659 .maxlen = sizeof(int),
5661 .proc_handler = proc_dointvec,
5664 .procname = "accept_ra_defrtr",
5665 .data = &ipv6_devconf.accept_ra_defrtr,
5666 .maxlen = sizeof(int),
5668 .proc_handler = proc_dointvec,
5671 .procname = "accept_ra_min_hop_limit",
5672 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5673 .maxlen = sizeof(int),
5675 .proc_handler = proc_dointvec,
5678 .procname = "accept_ra_pinfo",
5679 .data = &ipv6_devconf.accept_ra_pinfo,
5680 .maxlen = sizeof(int),
5682 .proc_handler = proc_dointvec,
5684 #ifdef CONFIG_IPV6_ROUTER_PREF
5686 .procname = "accept_ra_rtr_pref",
5687 .data = &ipv6_devconf.accept_ra_rtr_pref,
5688 .maxlen = sizeof(int),
5690 .proc_handler = proc_dointvec,
5693 .procname = "router_probe_interval",
5694 .data = &ipv6_devconf.rtr_probe_interval,
5695 .maxlen = sizeof(int),
5697 .proc_handler = proc_dointvec_jiffies,
5699 #ifdef CONFIG_IPV6_ROUTE_INFO
5701 .procname = "accept_ra_rt_info_min_plen",
5702 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
5703 .maxlen = sizeof(int),
5705 .proc_handler = proc_dointvec,
5708 .procname = "accept_ra_rt_info_max_plen",
5709 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5710 .maxlen = sizeof(int),
5712 .proc_handler = proc_dointvec,
5717 .procname = "accept_ra_rt_table",
5718 .data = &ipv6_devconf.accept_ra_rt_table,
5719 .maxlen = sizeof(int),
5721 .proc_handler = proc_dointvec,
5724 .procname = "proxy_ndp",
5725 .data = &ipv6_devconf.proxy_ndp,
5726 .maxlen = sizeof(int),
5728 .proc_handler = addrconf_sysctl_proxy_ndp,
5731 .procname = "accept_source_route",
5732 .data = &ipv6_devconf.accept_source_route,
5733 .maxlen = sizeof(int),
5735 .proc_handler = proc_dointvec,
5737 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5739 .procname = "optimistic_dad",
5740 .data = &ipv6_devconf.optimistic_dad,
5741 .maxlen = sizeof(int),
5743 .proc_handler = proc_dointvec,
5747 .procname = "use_optimistic",
5748 .data = &ipv6_devconf.use_optimistic,
5749 .maxlen = sizeof(int),
5751 .proc_handler = proc_dointvec,
5755 #ifdef CONFIG_IPV6_MROUTE
5757 .procname = "mc_forwarding",
5758 .data = &ipv6_devconf.mc_forwarding,
5759 .maxlen = sizeof(int),
5761 .proc_handler = proc_dointvec,
5765 .procname = "disable_ipv6",
5766 .data = &ipv6_devconf.disable_ipv6,
5767 .maxlen = sizeof(int),
5769 .proc_handler = addrconf_sysctl_disable,
5772 .procname = "accept_dad",
5773 .data = &ipv6_devconf.accept_dad,
5774 .maxlen = sizeof(int),
5776 .proc_handler = proc_dointvec,
5779 .procname = "force_tllao",
5780 .data = &ipv6_devconf.force_tllao,
5781 .maxlen = sizeof(int),
5783 .proc_handler = proc_dointvec
5786 .procname = "ndisc_notify",
5787 .data = &ipv6_devconf.ndisc_notify,
5788 .maxlen = sizeof(int),
5790 .proc_handler = proc_dointvec
5793 .procname = "suppress_frag_ndisc",
5794 .data = &ipv6_devconf.suppress_frag_ndisc,
5795 .maxlen = sizeof(int),
5797 .proc_handler = proc_dointvec
5800 .procname = "accept_ra_from_local",
5801 .data = &ipv6_devconf.accept_ra_from_local,
5802 .maxlen = sizeof(int),
5804 .proc_handler = proc_dointvec,
5807 .procname = "accept_ra_mtu",
5808 .data = &ipv6_devconf.accept_ra_mtu,
5809 .maxlen = sizeof(int),
5811 .proc_handler = proc_dointvec,
5814 .procname = "stable_secret",
5815 .data = &ipv6_devconf.stable_secret,
5816 .maxlen = IPV6_MAX_STRLEN,
5818 .proc_handler = addrconf_sysctl_stable_secret,
5821 .procname = "use_oif_addrs_only",
5822 .data = &ipv6_devconf.use_oif_addrs_only,
5823 .maxlen = sizeof(int),
5825 .proc_handler = proc_dointvec,
5828 .procname = "ignore_routes_with_linkdown",
5829 .data = &ipv6_devconf.ignore_routes_with_linkdown,
5830 .maxlen = sizeof(int),
5832 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
5840 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5841 struct inet6_dev *idev, struct ipv6_devconf *p)
5844 struct addrconf_sysctl_table *t;
5845 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5847 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5851 for (i = 0; t->addrconf_vars[i].data; i++) {
5852 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5853 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5854 t->addrconf_vars[i].extra2 = net;
5857 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5859 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5860 if (!t->sysctl_header)
5872 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5874 struct addrconf_sysctl_table *t;
5881 unregister_net_sysctl_table(t->sysctl_header);
5885 static int addrconf_sysctl_register(struct inet6_dev *idev)
5889 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5892 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5893 &ndisc_ifinfo_sysctl_change);
5896 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5899 neigh_sysctl_unregister(idev->nd_parms);
5904 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5906 __addrconf_sysctl_unregister(&idev->cnf);
5907 neigh_sysctl_unregister(idev->nd_parms);
5913 static int __net_init addrconf_init_net(struct net *net)
5916 struct ipv6_devconf *all, *dflt;
5918 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5922 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5924 goto err_alloc_dflt;
5926 /* these will be inherited by all namespaces */
5927 dflt->autoconf = ipv6_defaults.autoconf;
5928 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5930 dflt->stable_secret.initialized = false;
5931 all->stable_secret.initialized = false;
5933 net->ipv6.devconf_all = all;
5934 net->ipv6.devconf_dflt = dflt;
5936 #ifdef CONFIG_SYSCTL
5937 err = __addrconf_sysctl_register(net, "all", NULL, all);
5941 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5947 #ifdef CONFIG_SYSCTL
5949 __addrconf_sysctl_unregister(all);
5959 static void __net_exit addrconf_exit_net(struct net *net)
5961 #ifdef CONFIG_SYSCTL
5962 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5963 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5965 kfree(net->ipv6.devconf_dflt);
5966 kfree(net->ipv6.devconf_all);
5969 static struct pernet_operations addrconf_ops = {
5970 .init = addrconf_init_net,
5971 .exit = addrconf_exit_net,
5974 static struct rtnl_af_ops inet6_ops __read_mostly = {
5976 .fill_link_af = inet6_fill_link_af,
5977 .get_link_af_size = inet6_get_link_af_size,
5978 .validate_link_af = inet6_validate_link_af,
5979 .set_link_af = inet6_set_link_af,
5983 * Init / cleanup code
5986 int __init addrconf_init(void)
5988 struct inet6_dev *idev;
5991 err = ipv6_addr_label_init();
5993 pr_crit("%s: cannot initialize default policy table: %d\n",
5998 err = register_pernet_subsys(&addrconf_ops);
6002 addrconf_wq = create_workqueue("ipv6_addrconf");
6008 /* The addrconf netdev notifier requires that loopback_dev
6009 * has it's ipv6 private information allocated and setup
6010 * before it can bring up and give link-local addresses
6011 * to other devices which are up.
6013 * Unfortunately, loopback_dev is not necessarily the first
6014 * entry in the global dev_base list of net devices. In fact,
6015 * it is likely to be the very last entry on that list.
6016 * So this causes the notifier registry below to try and
6017 * give link-local addresses to all devices besides loopback_dev
6018 * first, then loopback_dev, which cases all the non-loopback_dev
6019 * devices to fail to get a link-local address.
6021 * So, as a temporary fix, allocate the ipv6 structure for
6022 * loopback_dev first by hand.
6023 * Longer term, all of the dependencies ipv6 has upon the loopback
6024 * device and it being up should be removed.
6027 idev = ipv6_add_dev(init_net.loopback_dev);
6030 err = PTR_ERR(idev);
6034 ip6_route_init_special_entries();
6036 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6037 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6039 register_netdevice_notifier(&ipv6_dev_notf);
6043 rtnl_af_register(&inet6_ops);
6045 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6050 /* Only the first call to __rtnl_register can fail */
6051 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6052 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6053 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6054 inet6_dump_ifaddr, NULL);
6055 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6056 inet6_dump_ifmcaddr, NULL);
6057 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6058 inet6_dump_ifacaddr, NULL);
6059 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6060 inet6_netconf_dump_devconf, NULL);
6062 ipv6_addr_label_rtnl_register();
6066 rtnl_af_unregister(&inet6_ops);
6067 unregister_netdevice_notifier(&ipv6_dev_notf);
6069 destroy_workqueue(addrconf_wq);
6071 unregister_pernet_subsys(&addrconf_ops);
6073 ipv6_addr_label_cleanup();
6078 void addrconf_cleanup(void)
6080 struct net_device *dev;
6083 unregister_netdevice_notifier(&ipv6_dev_notf);
6084 unregister_pernet_subsys(&addrconf_ops);
6085 ipv6_addr_label_cleanup();
6089 __rtnl_af_unregister(&inet6_ops);
6091 /* clean dev list */
6092 for_each_netdev(&init_net, dev) {
6093 if (__in6_dev_get(dev) == NULL)
6095 addrconf_ifdown(dev, 1);
6097 addrconf_ifdown(init_net.loopback_dev, 2);
6102 spin_lock_bh(&addrconf_hash_lock);
6103 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6104 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6105 spin_unlock_bh(&addrconf_hash_lock);
6106 cancel_delayed_work(&addr_chk_work);
6109 destroy_workqueue(addrconf_wq);