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/6lowpan.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);
153 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
154 static int ipv6_count_addresses(struct inet6_dev *idev);
155 static int ipv6_generate_stable_address(struct in6_addr *addr,
157 const struct inet6_dev *idev);
160 * Configured unicast address hash table
162 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
163 static DEFINE_SPINLOCK(addrconf_hash_lock);
165 static void addrconf_verify(void);
166 static void addrconf_verify_rtnl(void);
167 static void addrconf_verify_work(struct work_struct *);
169 static struct workqueue_struct *addrconf_wq;
170 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
172 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
173 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
175 static void addrconf_type_change(struct net_device *dev,
176 unsigned long event);
177 static int addrconf_ifdown(struct net_device *dev, int how);
179 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
181 const struct net_device *dev,
182 u32 flags, u32 noflags);
184 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
185 static void addrconf_dad_work(struct work_struct *w);
186 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
187 static void addrconf_dad_run(struct inet6_dev *idev);
188 static void addrconf_rs_timer(unsigned long data);
189 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
190 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
192 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
193 struct prefix_info *pinfo);
194 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
195 struct net_device *dev);
197 static struct ipv6_devconf ipv6_devconf __read_mostly = {
199 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
200 .mtu6 = IPV6_MIN_MTU,
202 .accept_redirects = 1,
204 .force_mld_version = 0,
205 .mldv1_unsolicited_report_interval = 10 * HZ,
206 .mldv2_unsolicited_report_interval = HZ,
208 .rtr_solicits = MAX_RTR_SOLICITATIONS,
209 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
210 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
211 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
213 .temp_valid_lft = TEMP_VALID_LIFETIME,
214 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
215 .regen_max_retry = REGEN_MAX_RETRY,
216 .max_desync_factor = MAX_DESYNC_FACTOR,
217 .max_addresses = IPV6_MAX_ADDRESSES,
218 .accept_ra_defrtr = 1,
219 .accept_ra_from_local = 0,
220 .accept_ra_min_hop_limit= 1,
221 .accept_ra_pinfo = 1,
222 #ifdef CONFIG_IPV6_ROUTER_PREF
223 .accept_ra_rtr_pref = 1,
224 .rtr_probe_interval = 60 * HZ,
225 #ifdef CONFIG_IPV6_ROUTE_INFO
226 .accept_ra_rt_info_max_plen = 0,
230 .accept_source_route = 0, /* we do not accept RH0 by default. */
233 .suppress_frag_ndisc = 1,
236 .initialized = false,
238 .use_oif_addrs_only = 0,
239 .ignore_routes_with_linkdown = 0,
240 .keep_addr_on_down = 0,
243 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
245 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
246 .mtu6 = IPV6_MIN_MTU,
248 .accept_redirects = 1,
250 .force_mld_version = 0,
251 .mldv1_unsolicited_report_interval = 10 * HZ,
252 .mldv2_unsolicited_report_interval = HZ,
254 .rtr_solicits = MAX_RTR_SOLICITATIONS,
255 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
256 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
257 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
259 .temp_valid_lft = TEMP_VALID_LIFETIME,
260 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
261 .regen_max_retry = REGEN_MAX_RETRY,
262 .max_desync_factor = MAX_DESYNC_FACTOR,
263 .max_addresses = IPV6_MAX_ADDRESSES,
264 .accept_ra_defrtr = 1,
265 .accept_ra_from_local = 0,
266 .accept_ra_min_hop_limit= 1,
267 .accept_ra_pinfo = 1,
268 #ifdef CONFIG_IPV6_ROUTER_PREF
269 .accept_ra_rtr_pref = 1,
270 .rtr_probe_interval = 60 * HZ,
271 #ifdef CONFIG_IPV6_ROUTE_INFO
272 .accept_ra_rt_info_max_plen = 0,
276 .accept_source_route = 0, /* we do not accept RH0 by default. */
279 .suppress_frag_ndisc = 1,
282 .initialized = false,
284 .use_oif_addrs_only = 0,
285 .ignore_routes_with_linkdown = 0,
286 .keep_addr_on_down = 0,
289 /* Check if a valid qdisc is available */
290 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
292 return !qdisc_tx_is_noop(dev);
295 static void addrconf_del_rs_timer(struct inet6_dev *idev)
297 if (del_timer(&idev->rs_timer))
301 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
303 if (cancel_delayed_work(&ifp->dad_work))
307 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
310 if (!timer_pending(&idev->rs_timer))
312 mod_timer(&idev->rs_timer, jiffies + when);
315 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
318 if (!delayed_work_pending(&ifp->dad_work))
320 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
323 static int snmp6_alloc_dev(struct inet6_dev *idev)
327 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
328 if (!idev->stats.ipv6)
331 for_each_possible_cpu(i) {
332 struct ipstats_mib *addrconf_stats;
333 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
334 u64_stats_init(&addrconf_stats->syncp);
338 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
340 if (!idev->stats.icmpv6dev)
342 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
344 if (!idev->stats.icmpv6msgdev)
350 kfree(idev->stats.icmpv6dev);
352 free_percpu(idev->stats.ipv6);
357 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
359 struct inet6_dev *ndev;
364 if (dev->mtu < IPV6_MIN_MTU)
365 return ERR_PTR(-EINVAL);
367 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
371 rwlock_init(&ndev->lock);
373 INIT_LIST_HEAD(&ndev->addr_list);
374 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
375 (unsigned long)ndev);
376 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
378 if (ndev->cnf.stable_secret.initialized)
379 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
381 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
383 ndev->cnf.mtu6 = dev->mtu;
384 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
385 if (!ndev->nd_parms) {
389 if (ndev->cnf.forwarding)
390 dev_disable_lro(dev);
391 /* We refer to the device */
394 if (snmp6_alloc_dev(ndev) < 0) {
396 "%s: cannot allocate memory for statistics; dev=%s.\n",
397 __func__, dev->name);
398 neigh_parms_release(&nd_tbl, ndev->nd_parms);
404 if (snmp6_register_dev(ndev) < 0) {
406 "%s: cannot create /proc/net/dev_snmp6/%s\n",
407 __func__, dev->name);
411 /* One reference from device. */
414 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
415 ndev->cnf.accept_dad = -1;
417 #if IS_ENABLED(CONFIG_IPV6_SIT)
418 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
419 pr_info("%s: Disabled Multicast RS\n", dev->name);
420 ndev->cnf.rtr_solicits = 0;
424 INIT_LIST_HEAD(&ndev->tempaddr_list);
425 ndev->desync_factor = U32_MAX;
426 if ((dev->flags&IFF_LOOPBACK) ||
427 dev->type == ARPHRD_TUNNEL ||
428 dev->type == ARPHRD_TUNNEL6 ||
429 dev->type == ARPHRD_SIT ||
430 dev->type == ARPHRD_NONE) {
431 ndev->cnf.use_tempaddr = -1;
433 ipv6_regen_rndid(ndev);
435 ndev->token = in6addr_any;
437 if (netif_running(dev) && addrconf_qdisc_ok(dev))
438 ndev->if_flags |= IF_READY;
440 ipv6_mc_init_dev(ndev);
441 ndev->tstamp = jiffies;
442 err = addrconf_sysctl_register(ndev);
444 ipv6_mc_destroy_dev(ndev);
445 snmp6_unregister_dev(ndev);
448 /* protected by rtnl_lock */
449 rcu_assign_pointer(dev->ip6_ptr, ndev);
451 /* Join interface-local all-node multicast group */
452 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
454 /* Join all-node multicast group */
455 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
457 /* Join all-router multicast group if forwarding is set */
458 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
459 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
464 neigh_parms_release(&nd_tbl, ndev->nd_parms);
466 in6_dev_finish_destroy(ndev);
470 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
472 struct inet6_dev *idev;
476 idev = __in6_dev_get(dev);
478 idev = ipv6_add_dev(dev);
483 if (dev->flags&IFF_UP)
488 static int inet6_netconf_msgsize_devconf(int type)
490 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
491 + nla_total_size(4); /* NETCONFA_IFINDEX */
494 if (type == NETCONFA_ALL)
497 if (all || type == NETCONFA_FORWARDING)
498 size += nla_total_size(4);
499 #ifdef CONFIG_IPV6_MROUTE
500 if (all || type == NETCONFA_MC_FORWARDING)
501 size += nla_total_size(4);
503 if (all || type == NETCONFA_PROXY_NEIGH)
504 size += nla_total_size(4);
506 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
507 size += nla_total_size(4);
512 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
513 struct ipv6_devconf *devconf, u32 portid,
514 u32 seq, int event, unsigned int flags,
517 struct nlmsghdr *nlh;
518 struct netconfmsg *ncm;
521 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
526 if (type == NETCONFA_ALL)
529 ncm = nlmsg_data(nlh);
530 ncm->ncm_family = AF_INET6;
532 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
533 goto nla_put_failure;
535 if ((all || type == NETCONFA_FORWARDING) &&
536 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
537 goto nla_put_failure;
538 #ifdef CONFIG_IPV6_MROUTE
539 if ((all || type == NETCONFA_MC_FORWARDING) &&
540 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
541 devconf->mc_forwarding) < 0)
542 goto nla_put_failure;
544 if ((all || type == NETCONFA_PROXY_NEIGH) &&
545 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
546 goto nla_put_failure;
548 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
549 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
550 devconf->ignore_routes_with_linkdown) < 0)
551 goto nla_put_failure;
557 nlmsg_cancel(skb, nlh);
561 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
562 struct ipv6_devconf *devconf)
567 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
571 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
572 RTM_NEWNETCONF, 0, type);
574 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
575 WARN_ON(err == -EMSGSIZE);
579 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
582 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
585 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
586 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
587 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
588 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
589 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
592 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
593 struct nlmsghdr *nlh)
595 struct net *net = sock_net(in_skb->sk);
596 struct nlattr *tb[NETCONFA_MAX+1];
597 struct netconfmsg *ncm;
599 struct ipv6_devconf *devconf;
600 struct inet6_dev *in6_dev;
601 struct net_device *dev;
605 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
606 devconf_ipv6_policy);
611 if (!tb[NETCONFA_IFINDEX])
614 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
616 case NETCONFA_IFINDEX_ALL:
617 devconf = net->ipv6.devconf_all;
619 case NETCONFA_IFINDEX_DEFAULT:
620 devconf = net->ipv6.devconf_dflt;
623 dev = __dev_get_by_index(net, ifindex);
626 in6_dev = __in6_dev_get(dev);
629 devconf = &in6_dev->cnf;
634 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
638 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
639 NETLINK_CB(in_skb).portid,
640 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
643 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
644 WARN_ON(err == -EMSGSIZE);
648 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
653 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
654 struct netlink_callback *cb)
656 struct net *net = sock_net(skb->sk);
659 struct net_device *dev;
660 struct inet6_dev *idev;
661 struct hlist_head *head;
664 s_idx = idx = cb->args[1];
666 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
668 head = &net->dev_index_head[h];
670 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
672 hlist_for_each_entry_rcu(dev, head, index_hlist) {
675 idev = __in6_dev_get(dev);
679 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
681 NETLINK_CB(cb->skb).portid,
689 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
695 if (h == NETDEV_HASHENTRIES) {
696 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
697 net->ipv6.devconf_all,
698 NETLINK_CB(cb->skb).portid,
700 RTM_NEWNETCONF, NLM_F_MULTI,
706 if (h == NETDEV_HASHENTRIES + 1) {
707 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
708 net->ipv6.devconf_dflt,
709 NETLINK_CB(cb->skb).portid,
711 RTM_NEWNETCONF, NLM_F_MULTI,
725 static void dev_forward_change(struct inet6_dev *idev)
727 struct net_device *dev;
728 struct inet6_ifaddr *ifa;
733 if (idev->cnf.forwarding)
734 dev_disable_lro(dev);
735 if (dev->flags & IFF_MULTICAST) {
736 if (idev->cnf.forwarding) {
737 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
738 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
739 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
741 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
742 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
743 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
747 list_for_each_entry(ifa, &idev->addr_list, if_list) {
748 if (ifa->flags&IFA_F_TENTATIVE)
750 if (idev->cnf.forwarding)
751 addrconf_join_anycast(ifa);
753 addrconf_leave_anycast(ifa);
755 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
756 dev->ifindex, &idev->cnf);
760 static void addrconf_forward_change(struct net *net, __s32 newf)
762 struct net_device *dev;
763 struct inet6_dev *idev;
765 for_each_netdev(net, dev) {
766 idev = __in6_dev_get(dev);
768 int changed = (!idev->cnf.forwarding) ^ (!newf);
769 idev->cnf.forwarding = newf;
771 dev_forward_change(idev);
776 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
782 return restart_syscall();
784 net = (struct net *)table->extra2;
788 if (p == &net->ipv6.devconf_dflt->forwarding) {
789 if ((!newf) ^ (!old))
790 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
791 NETCONFA_IFINDEX_DEFAULT,
792 net->ipv6.devconf_dflt);
797 if (p == &net->ipv6.devconf_all->forwarding) {
798 int old_dflt = net->ipv6.devconf_dflt->forwarding;
800 net->ipv6.devconf_dflt->forwarding = newf;
801 if ((!newf) ^ (!old_dflt))
802 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
803 NETCONFA_IFINDEX_DEFAULT,
804 net->ipv6.devconf_dflt);
806 addrconf_forward_change(net, newf);
807 if ((!newf) ^ (!old))
808 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
809 NETCONFA_IFINDEX_ALL,
810 net->ipv6.devconf_all);
811 } else if ((!newf) ^ (!old))
812 dev_forward_change((struct inet6_dev *)table->extra1);
816 rt6_purge_dflt_routers(net);
820 static void addrconf_linkdown_change(struct net *net, __s32 newf)
822 struct net_device *dev;
823 struct inet6_dev *idev;
825 for_each_netdev(net, dev) {
826 idev = __in6_dev_get(dev);
828 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
830 idev->cnf.ignore_routes_with_linkdown = newf;
832 inet6_netconf_notify_devconf(dev_net(dev),
833 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
840 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
846 return restart_syscall();
848 net = (struct net *)table->extra2;
852 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
853 if ((!newf) ^ (!old))
854 inet6_netconf_notify_devconf(net,
855 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
856 NETCONFA_IFINDEX_DEFAULT,
857 net->ipv6.devconf_dflt);
862 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
863 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
864 addrconf_linkdown_change(net, newf);
865 if ((!newf) ^ (!old))
866 inet6_netconf_notify_devconf(net,
867 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
868 NETCONFA_IFINDEX_ALL,
869 net->ipv6.devconf_all);
878 /* Nobody refers to this ifaddr, destroy it */
879 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
881 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
883 #ifdef NET_REFCNT_DEBUG
884 pr_debug("%s\n", __func__);
887 in6_dev_put(ifp->idev);
889 if (cancel_delayed_work(&ifp->dad_work))
890 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
893 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
894 pr_warn("Freeing alive inet6 address %p\n", ifp);
903 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
906 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
909 * Each device address list is sorted in order of scope -
910 * global before linklocal.
912 list_for_each(p, &idev->addr_list) {
913 struct inet6_ifaddr *ifa
914 = list_entry(p, struct inet6_ifaddr, if_list);
915 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
919 list_add_tail(&ifp->if_list, p);
922 static u32 inet6_addr_hash(const struct in6_addr *addr)
924 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
927 /* On success it returns ifp with increased reference count */
929 static struct inet6_ifaddr *
930 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
931 const struct in6_addr *peer_addr, int pfxlen,
932 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
934 struct inet6_ifaddr *ifa = NULL;
938 int addr_type = ipv6_addr_type(addr);
940 if (addr_type == IPV6_ADDR_ANY ||
941 addr_type & IPV6_ADDR_MULTICAST ||
942 (!(idev->dev->flags & IFF_LOOPBACK) &&
943 addr_type & IPV6_ADDR_LOOPBACK))
944 return ERR_PTR(-EADDRNOTAVAIL);
948 err = -ENODEV; /*XXX*/
952 if (idev->cnf.disable_ipv6) {
957 spin_lock(&addrconf_hash_lock);
959 /* Ignore adding duplicate addresses on an interface */
960 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
961 ADBG("ipv6_add_addr: already assigned\n");
966 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
969 ADBG("ipv6_add_addr: malloc failed\n");
974 rt = addrconf_dst_alloc(idev, addr, false);
980 neigh_parms_data_state_setall(idev->nd_parms);
984 ifa->peer_addr = *peer_addr;
986 spin_lock_init(&ifa->lock);
987 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
988 INIT_HLIST_NODE(&ifa->addr_lst);
990 ifa->prefix_len = pfxlen;
991 ifa->flags = flags | IFA_F_TENTATIVE;
992 ifa->valid_lft = valid_lft;
993 ifa->prefered_lft = prefered_lft;
994 ifa->cstamp = ifa->tstamp = jiffies;
995 ifa->tokenized = false;
1004 /* Add to big hash table */
1005 hash = inet6_addr_hash(addr);
1007 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1008 spin_unlock(&addrconf_hash_lock);
1010 write_lock(&idev->lock);
1011 /* Add to inet6_dev unicast addr list. */
1012 ipv6_link_dev_addr(idev, ifa);
1014 if (ifa->flags&IFA_F_TEMPORARY) {
1015 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1020 write_unlock(&idev->lock);
1022 rcu_read_unlock_bh();
1024 if (likely(err == 0))
1025 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1033 spin_unlock(&addrconf_hash_lock);
1037 enum cleanup_prefix_rt_t {
1038 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1039 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1040 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1044 * Check, whether the prefix for ifp would still need a prefix route
1045 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1048 * 1) we don't purge prefix if address was not permanent.
1049 * prefix is managed by its own lifetime.
1050 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1051 * 3) if there are no addresses, delete prefix.
1052 * 4) if there are still other permanent address(es),
1053 * corresponding prefix is still permanent.
1054 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1055 * don't purge the prefix, assume user space is managing it.
1056 * 6) otherwise, update prefix lifetime to the
1057 * longest valid lifetime among the corresponding
1058 * addresses on the device.
1059 * Note: subsequent RA will update lifetime.
1061 static enum cleanup_prefix_rt_t
1062 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1064 struct inet6_ifaddr *ifa;
1065 struct inet6_dev *idev = ifp->idev;
1066 unsigned long lifetime;
1067 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1071 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1074 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1077 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1078 return CLEANUP_PREFIX_RT_NOP;
1080 action = CLEANUP_PREFIX_RT_EXPIRE;
1082 spin_lock(&ifa->lock);
1084 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1086 * Note: Because this address is
1087 * not permanent, lifetime <
1088 * LONG_MAX / HZ here.
1090 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1091 *expires = ifa->tstamp + lifetime * HZ;
1092 spin_unlock(&ifa->lock);
1099 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1101 struct rt6_info *rt;
1103 rt = addrconf_get_prefix_route(&ifp->addr,
1106 0, RTF_GATEWAY | RTF_DEFAULT);
1111 if (!(rt->rt6i_flags & RTF_EXPIRES))
1112 rt6_set_expires(rt, expires);
1119 /* This function wants to get referenced ifp and releases it before return */
1121 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1124 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1125 unsigned long expires;
1129 spin_lock_bh(&ifp->lock);
1131 ifp->state = INET6_IFADDR_STATE_DEAD;
1132 spin_unlock_bh(&ifp->lock);
1134 if (state == INET6_IFADDR_STATE_DEAD)
1137 spin_lock_bh(&addrconf_hash_lock);
1138 hlist_del_init_rcu(&ifp->addr_lst);
1139 spin_unlock_bh(&addrconf_hash_lock);
1141 write_lock_bh(&ifp->idev->lock);
1143 if (ifp->flags&IFA_F_TEMPORARY) {
1144 list_del(&ifp->tmp_list);
1146 in6_ifa_put(ifp->ifpub);
1152 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1153 action = check_cleanup_prefix_route(ifp, &expires);
1155 list_del_init(&ifp->if_list);
1158 write_unlock_bh(&ifp->idev->lock);
1160 addrconf_del_dad_work(ifp);
1162 ipv6_ifa_notify(RTM_DELADDR, ifp);
1164 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1166 if (action != CLEANUP_PREFIX_RT_NOP) {
1167 cleanup_prefix_route(ifp, expires,
1168 action == CLEANUP_PREFIX_RT_DEL);
1171 /* clean up prefsrc entries */
1172 rt6_remove_prefsrc(ifp);
1177 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1179 struct inet6_dev *idev = ifp->idev;
1180 struct in6_addr addr, *tmpaddr;
1181 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1182 unsigned long regen_advance;
1186 unsigned long now = jiffies;
1187 long max_desync_factor;
1189 write_lock_bh(&idev->lock);
1191 spin_lock_bh(&ift->lock);
1192 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1193 spin_unlock_bh(&ift->lock);
1200 if (idev->cnf.use_tempaddr <= 0) {
1201 write_unlock_bh(&idev->lock);
1202 pr_info("%s: use_tempaddr is disabled\n", __func__);
1207 spin_lock_bh(&ifp->lock);
1208 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1209 idev->cnf.use_tempaddr = -1; /*XXX*/
1210 spin_unlock_bh(&ifp->lock);
1211 write_unlock_bh(&idev->lock);
1212 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1219 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1220 ipv6_try_regen_rndid(idev, tmpaddr);
1221 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1222 age = (now - ifp->tstamp) / HZ;
1224 regen_advance = idev->cnf.regen_max_retry *
1225 idev->cnf.dad_transmits *
1226 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1228 /* recalculate max_desync_factor each time and update
1229 * idev->desync_factor if it's larger
1231 max_desync_factor = min_t(__u32,
1232 idev->cnf.max_desync_factor,
1233 idev->cnf.temp_prefered_lft - regen_advance);
1235 if (unlikely(idev->desync_factor > max_desync_factor)) {
1236 if (max_desync_factor > 0) {
1237 get_random_bytes(&idev->desync_factor,
1238 sizeof(idev->desync_factor));
1239 idev->desync_factor %= max_desync_factor;
1241 idev->desync_factor = 0;
1245 tmp_valid_lft = min_t(__u32,
1247 idev->cnf.temp_valid_lft + age);
1248 tmp_prefered_lft = idev->cnf.temp_prefered_lft + age -
1249 idev->desync_factor;
1250 /* guard against underflow in case of concurrent updates to cnf */
1251 if (unlikely(tmp_prefered_lft < 0))
1252 tmp_prefered_lft = 0;
1253 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1254 tmp_plen = ifp->prefix_len;
1255 tmp_tstamp = ifp->tstamp;
1256 spin_unlock_bh(&ifp->lock);
1258 write_unlock_bh(&idev->lock);
1260 /* A temporary address is created only if this calculated Preferred
1261 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1262 * an implementation must not create a temporary address with a zero
1263 * Preferred Lifetime.
1264 * Use age calculation as in addrconf_verify to avoid unnecessary
1265 * temporary addresses being generated.
1267 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1268 if (tmp_prefered_lft <= regen_advance + age) {
1275 addr_flags = IFA_F_TEMPORARY;
1276 /* set in addrconf_prefix_rcv() */
1277 if (ifp->flags & IFA_F_OPTIMISTIC)
1278 addr_flags |= IFA_F_OPTIMISTIC;
1280 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1281 ipv6_addr_scope(&addr), addr_flags,
1282 tmp_valid_lft, tmp_prefered_lft);
1286 pr_info("%s: retry temporary address regeneration\n", __func__);
1288 write_lock_bh(&idev->lock);
1292 spin_lock_bh(&ift->lock);
1295 ift->tstamp = tmp_tstamp;
1296 spin_unlock_bh(&ift->lock);
1298 addrconf_dad_start(ift);
1306 * Choose an appropriate source address (RFC3484)
1309 IPV6_SADDR_RULE_INIT = 0,
1310 IPV6_SADDR_RULE_LOCAL,
1311 IPV6_SADDR_RULE_SCOPE,
1312 IPV6_SADDR_RULE_PREFERRED,
1313 #ifdef CONFIG_IPV6_MIP6
1314 IPV6_SADDR_RULE_HOA,
1316 IPV6_SADDR_RULE_OIF,
1317 IPV6_SADDR_RULE_LABEL,
1318 IPV6_SADDR_RULE_PRIVACY,
1319 IPV6_SADDR_RULE_ORCHID,
1320 IPV6_SADDR_RULE_PREFIX,
1321 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1322 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1327 struct ipv6_saddr_score {
1330 struct inet6_ifaddr *ifa;
1331 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1336 struct ipv6_saddr_dst {
1337 const struct in6_addr *addr;
1344 static inline int ipv6_saddr_preferred(int type)
1346 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1351 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1353 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1354 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1360 static int ipv6_get_saddr_eval(struct net *net,
1361 struct ipv6_saddr_score *score,
1362 struct ipv6_saddr_dst *dst,
1367 if (i <= score->rule) {
1369 case IPV6_SADDR_RULE_SCOPE:
1370 ret = score->scopedist;
1372 case IPV6_SADDR_RULE_PREFIX:
1373 ret = score->matchlen;
1376 ret = !!test_bit(i, score->scorebits);
1382 case IPV6_SADDR_RULE_INIT:
1383 /* Rule 0: remember if hiscore is not ready yet */
1386 case IPV6_SADDR_RULE_LOCAL:
1387 /* Rule 1: Prefer same address */
1388 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1390 case IPV6_SADDR_RULE_SCOPE:
1391 /* Rule 2: Prefer appropriate scope
1396 * ---+--+-+---> scope
1398 * | d is scope of the destination.
1400 * | \ <- smaller scope is better if
1401 * B-15 | \ if scope is enough for destination.
1402 * | ret = B - scope (-1 <= scope >= d <= 15).
1404 * |/ <- greater is better
1405 * -C / if scope is not enough for destination.
1406 * /| ret = scope - C (-1 <= d < scope <= 15).
1408 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1409 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1410 * Assume B = 0 and we get C > 29.
1412 ret = __ipv6_addr_src_scope(score->addr_type);
1413 if (ret >= dst->scope)
1416 ret -= 128; /* 30 is enough */
1417 score->scopedist = ret;
1419 case IPV6_SADDR_RULE_PREFERRED:
1421 /* Rule 3: Avoid deprecated and optimistic addresses */
1422 u8 avoid = IFA_F_DEPRECATED;
1424 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1425 avoid |= IFA_F_OPTIMISTIC;
1426 ret = ipv6_saddr_preferred(score->addr_type) ||
1427 !(score->ifa->flags & avoid);
1430 #ifdef CONFIG_IPV6_MIP6
1431 case IPV6_SADDR_RULE_HOA:
1433 /* Rule 4: Prefer home address */
1434 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1435 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1439 case IPV6_SADDR_RULE_OIF:
1440 /* Rule 5: Prefer outgoing interface */
1441 ret = (!dst->ifindex ||
1442 dst->ifindex == score->ifa->idev->dev->ifindex);
1444 case IPV6_SADDR_RULE_LABEL:
1445 /* Rule 6: Prefer matching label */
1446 ret = ipv6_addr_label(net,
1447 &score->ifa->addr, score->addr_type,
1448 score->ifa->idev->dev->ifindex) == dst->label;
1450 case IPV6_SADDR_RULE_PRIVACY:
1452 /* Rule 7: Prefer public address
1453 * Note: prefer temporary address if use_tempaddr >= 2
1455 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1456 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1457 score->ifa->idev->cnf.use_tempaddr >= 2;
1458 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1461 case IPV6_SADDR_RULE_ORCHID:
1462 /* Rule 8-: Prefer ORCHID vs ORCHID or
1463 * non-ORCHID vs non-ORCHID
1465 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1466 ipv6_addr_orchid(dst->addr));
1468 case IPV6_SADDR_RULE_PREFIX:
1469 /* Rule 8: Use longest matching prefix */
1470 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1471 if (ret > score->ifa->prefix_len)
1472 ret = score->ifa->prefix_len;
1473 score->matchlen = ret;
1475 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1476 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1477 /* Optimistic addresses still have lower precedence than other
1478 * preferred addresses.
1480 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1488 __set_bit(i, score->scorebits);
1494 static int __ipv6_dev_get_saddr(struct net *net,
1495 struct ipv6_saddr_dst *dst,
1496 struct inet6_dev *idev,
1497 struct ipv6_saddr_score *scores,
1500 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1502 read_lock_bh(&idev->lock);
1503 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1507 * - Tentative Address (RFC2462 section 5.4)
1508 * - A tentative address is not considered
1509 * "assigned to an interface" in the traditional
1510 * sense, unless it is also flagged as optimistic.
1511 * - Candidate Source Address (section 4)
1512 * - In any case, anycast addresses, multicast
1513 * addresses, and the unspecified address MUST
1514 * NOT be included in a candidate set.
1516 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1517 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1520 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1522 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1523 score->addr_type & IPV6_ADDR_MULTICAST)) {
1524 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1530 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1532 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1533 int minihiscore, miniscore;
1535 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1536 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1538 if (minihiscore > miniscore) {
1539 if (i == IPV6_SADDR_RULE_SCOPE &&
1540 score->scopedist > 0) {
1543 * each remaining entry
1544 * has too small (not enough)
1545 * scope, because ifa entries
1546 * are sorted by their scope
1552 } else if (minihiscore < miniscore) {
1554 in6_ifa_put(hiscore->ifa);
1556 in6_ifa_hold(score->ifa);
1558 swap(hiscore, score);
1559 hiscore_idx = 1 - hiscore_idx;
1561 /* restore our iterator */
1562 score->ifa = hiscore->ifa;
1569 read_unlock_bh(&idev->lock);
1573 static int ipv6_get_saddr_master(struct net *net,
1574 const struct net_device *dst_dev,
1575 const struct net_device *master,
1576 struct ipv6_saddr_dst *dst,
1577 struct ipv6_saddr_score *scores,
1580 struct inet6_dev *idev;
1582 idev = __in6_dev_get(dst_dev);
1584 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1585 scores, hiscore_idx);
1587 idev = __in6_dev_get(master);
1589 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1590 scores, hiscore_idx);
1595 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1596 const struct in6_addr *daddr, unsigned int prefs,
1597 struct in6_addr *saddr)
1599 struct ipv6_saddr_score scores[2], *hiscore;
1600 struct ipv6_saddr_dst dst;
1601 struct inet6_dev *idev;
1602 struct net_device *dev;
1604 bool use_oif_addr = false;
1605 int hiscore_idx = 0;
1607 dst_type = __ipv6_addr_type(daddr);
1609 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1610 dst.scope = __ipv6_addr_src_scope(dst_type);
1611 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1614 scores[hiscore_idx].rule = -1;
1615 scores[hiscore_idx].ifa = NULL;
1619 /* Candidate Source Address (section 4)
1620 * - multicast and link-local destination address,
1621 * the set of candidate source address MUST only
1622 * include addresses assigned to interfaces
1623 * belonging to the same link as the outgoing
1625 * (- For site-local destination addresses, the
1626 * set of candidate source addresses MUST only
1627 * include addresses assigned to interfaces
1628 * belonging to the same site as the outgoing
1630 * - "It is RECOMMENDED that the candidate source addresses
1631 * be the set of unicast addresses assigned to the
1632 * interface that will be used to send to the destination
1633 * (the 'outgoing' interface)." (RFC 6724)
1636 idev = __in6_dev_get(dst_dev);
1637 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1638 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1639 (idev && idev->cnf.use_oif_addrs_only)) {
1640 use_oif_addr = true;
1646 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1648 const struct net_device *master;
1651 /* if dst_dev exists and is enslaved to an L3 device, then
1652 * prefer addresses from dst_dev and then the master over
1653 * any other enslaved devices in the L3 domain.
1655 master = l3mdev_master_dev_rcu(dst_dev);
1657 master_idx = master->ifindex;
1659 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1661 scores, hiscore_idx);
1663 if (scores[hiscore_idx].ifa)
1667 for_each_netdev_rcu(net, dev) {
1668 /* only consider addresses on devices in the
1671 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1673 idev = __in6_dev_get(dev);
1676 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1683 hiscore = &scores[hiscore_idx];
1685 return -EADDRNOTAVAIL;
1687 *saddr = hiscore->ifa->addr;
1688 in6_ifa_put(hiscore->ifa);
1691 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1693 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1696 struct inet6_ifaddr *ifp;
1697 int err = -EADDRNOTAVAIL;
1699 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1700 if (ifp->scope > IFA_LINK)
1702 if (ifp->scope == IFA_LINK &&
1703 !(ifp->flags & banned_flags)) {
1712 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1715 struct inet6_dev *idev;
1716 int err = -EADDRNOTAVAIL;
1719 idev = __in6_dev_get(dev);
1721 read_lock_bh(&idev->lock);
1722 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1723 read_unlock_bh(&idev->lock);
1729 static int ipv6_count_addresses(struct inet6_dev *idev)
1732 struct inet6_ifaddr *ifp;
1734 read_lock_bh(&idev->lock);
1735 list_for_each_entry(ifp, &idev->addr_list, if_list)
1737 read_unlock_bh(&idev->lock);
1741 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1742 const struct net_device *dev, int strict)
1744 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1746 EXPORT_SYMBOL(ipv6_chk_addr);
1748 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1749 const struct net_device *dev, int strict,
1752 struct inet6_ifaddr *ifp;
1753 unsigned int hash = inet6_addr_hash(addr);
1757 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1758 if (!net_eq(dev_net(ifp->idev->dev), net))
1760 /* Decouple optimistic from tentative for evaluation here.
1761 * Ban optimistic addresses explicitly, when required.
1763 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1764 ? (ifp->flags&~IFA_F_TENTATIVE)
1766 if (ipv6_addr_equal(&ifp->addr, addr) &&
1767 !(ifp_flags&banned_flags) &&
1768 (!dev || ifp->idev->dev == dev ||
1769 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1770 rcu_read_unlock_bh();
1775 rcu_read_unlock_bh();
1778 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1780 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1781 struct net_device *dev)
1783 unsigned int hash = inet6_addr_hash(addr);
1784 struct inet6_ifaddr *ifp;
1786 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1787 if (!net_eq(dev_net(ifp->idev->dev), net))
1789 if (ipv6_addr_equal(&ifp->addr, addr)) {
1790 if (!dev || ifp->idev->dev == dev)
1797 /* Compares an address/prefix_len with addresses on device @dev.
1798 * If one is found it returns true.
1800 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1801 const unsigned int prefix_len, struct net_device *dev)
1803 struct inet6_dev *idev;
1804 struct inet6_ifaddr *ifa;
1808 idev = __in6_dev_get(dev);
1810 read_lock_bh(&idev->lock);
1811 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1812 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1816 read_unlock_bh(&idev->lock);
1822 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1824 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1826 struct inet6_dev *idev;
1827 struct inet6_ifaddr *ifa;
1832 idev = __in6_dev_get(dev);
1834 read_lock_bh(&idev->lock);
1835 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1836 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1841 read_unlock_bh(&idev->lock);
1846 EXPORT_SYMBOL(ipv6_chk_prefix);
1848 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1849 struct net_device *dev, int strict)
1851 struct inet6_ifaddr *ifp, *result = NULL;
1852 unsigned int hash = inet6_addr_hash(addr);
1855 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1856 if (!net_eq(dev_net(ifp->idev->dev), net))
1858 if (ipv6_addr_equal(&ifp->addr, addr)) {
1859 if (!dev || ifp->idev->dev == dev ||
1860 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1867 rcu_read_unlock_bh();
1872 /* Gets referenced address, destroys ifaddr */
1874 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1877 ifp->flags |= IFA_F_DADFAILED;
1879 if (ifp->flags&IFA_F_PERMANENT) {
1880 spin_lock_bh(&ifp->lock);
1881 addrconf_del_dad_work(ifp);
1882 ifp->flags |= IFA_F_TENTATIVE;
1883 spin_unlock_bh(&ifp->lock);
1885 ipv6_ifa_notify(0, ifp);
1887 } else if (ifp->flags&IFA_F_TEMPORARY) {
1888 struct inet6_ifaddr *ifpub;
1889 spin_lock_bh(&ifp->lock);
1892 in6_ifa_hold(ifpub);
1893 spin_unlock_bh(&ifp->lock);
1894 ipv6_create_tempaddr(ifpub, ifp);
1897 spin_unlock_bh(&ifp->lock);
1905 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1909 spin_lock_bh(&ifp->lock);
1910 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1911 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1914 spin_unlock_bh(&ifp->lock);
1919 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1921 struct inet6_dev *idev = ifp->idev;
1922 struct net *net = dev_net(ifp->idev->dev);
1924 if (addrconf_dad_end(ifp)) {
1929 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1930 ifp->idev->dev->name, &ifp->addr);
1932 spin_lock_bh(&ifp->lock);
1934 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1935 int scope = ifp->scope;
1936 u32 flags = ifp->flags;
1937 struct in6_addr new_addr;
1938 struct inet6_ifaddr *ifp2;
1939 u32 valid_lft, preferred_lft;
1940 int pfxlen = ifp->prefix_len;
1941 int retries = ifp->stable_privacy_retry + 1;
1943 if (retries > net->ipv6.sysctl.idgen_retries) {
1944 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1945 ifp->idev->dev->name);
1949 new_addr = ifp->addr;
1950 if (ipv6_generate_stable_address(&new_addr, retries,
1954 valid_lft = ifp->valid_lft;
1955 preferred_lft = ifp->prefered_lft;
1957 spin_unlock_bh(&ifp->lock);
1959 if (idev->cnf.max_addresses &&
1960 ipv6_count_addresses(idev) >=
1961 idev->cnf.max_addresses)
1964 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1965 ifp->idev->dev->name);
1967 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1968 scope, flags, valid_lft,
1973 spin_lock_bh(&ifp2->lock);
1974 ifp2->stable_privacy_retry = retries;
1975 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1976 spin_unlock_bh(&ifp2->lock);
1978 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1981 spin_lock_bh(&ifp->lock);
1985 /* transition from _POSTDAD to _ERRDAD */
1986 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1987 spin_unlock_bh(&ifp->lock);
1989 addrconf_mod_dad_work(ifp, 0);
1993 /* Join to solicited addr multicast group.
1994 * caller must hold RTNL */
1995 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1997 struct in6_addr maddr;
1999 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2002 addrconf_addr_solict_mult(addr, &maddr);
2003 ipv6_dev_mc_inc(dev, &maddr);
2006 /* caller must hold RTNL */
2007 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2009 struct in6_addr maddr;
2011 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2014 addrconf_addr_solict_mult(addr, &maddr);
2015 __ipv6_dev_mc_dec(idev, &maddr);
2018 /* caller must hold RTNL */
2019 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2021 struct in6_addr addr;
2023 if (ifp->prefix_len >= 127) /* RFC 6164 */
2025 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2026 if (ipv6_addr_any(&addr))
2028 __ipv6_dev_ac_inc(ifp->idev, &addr);
2031 /* caller must hold RTNL */
2032 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2034 struct in6_addr addr;
2036 if (ifp->prefix_len >= 127) /* RFC 6164 */
2038 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2039 if (ipv6_addr_any(&addr))
2041 __ipv6_dev_ac_dec(ifp->idev, &addr);
2044 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2046 if (dev->addr_len != EUI64_ADDR_LEN)
2048 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2053 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2055 union fwnet_hwaddr *ha;
2057 if (dev->addr_len != FWNET_ALEN)
2060 ha = (union fwnet_hwaddr *)dev->dev_addr;
2062 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2067 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2069 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2070 if (dev->addr_len != ARCNET_ALEN)
2073 eui[7] = *(u8 *)dev->dev_addr;
2077 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2079 if (dev->addr_len != INFINIBAND_ALEN)
2081 memcpy(eui, dev->dev_addr + 12, 8);
2086 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2090 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2091 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2092 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2093 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2094 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2095 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2099 memcpy(eui + 4, &addr, 4);
2103 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2105 if (dev->priv_flags & IFF_ISATAP)
2106 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2110 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2112 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2115 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2117 memcpy(eui, dev->perm_addr, 3);
2118 memcpy(eui + 5, dev->perm_addr + 3, 3);
2125 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2127 switch (dev->type) {
2130 return addrconf_ifid_eui48(eui, dev);
2132 return addrconf_ifid_arcnet(eui, dev);
2133 case ARPHRD_INFINIBAND:
2134 return addrconf_ifid_infiniband(eui, dev);
2136 return addrconf_ifid_sit(eui, dev);
2138 return addrconf_ifid_gre(eui, dev);
2139 case ARPHRD_6LOWPAN:
2140 return addrconf_ifid_eui64(eui, dev);
2141 case ARPHRD_IEEE1394:
2142 return addrconf_ifid_ieee1394(eui, dev);
2143 case ARPHRD_TUNNEL6:
2144 return addrconf_ifid_ip6tnl(eui, dev);
2149 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2152 struct inet6_ifaddr *ifp;
2154 read_lock_bh(&idev->lock);
2155 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2156 if (ifp->scope > IFA_LINK)
2158 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2159 memcpy(eui, ifp->addr.s6_addr+8, 8);
2164 read_unlock_bh(&idev->lock);
2168 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2169 static void ipv6_regen_rndid(struct inet6_dev *idev)
2172 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2173 idev->rndid[0] &= ~0x02;
2176 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2177 * check if generated address is not inappropriate
2179 * - Reserved subnet anycast (RFC 2526)
2180 * 11111101 11....11 1xxxxxxx
2181 * - ISATAP (RFC4214) 6.1
2182 * 00-00-5E-FE-xx-xx-xx-xx
2184 * - XXX: already assigned to an address on the device
2186 if (idev->rndid[0] == 0xfd &&
2187 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2188 (idev->rndid[7]&0x80))
2190 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2191 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2193 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2198 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2200 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2201 ipv6_regen_rndid(idev);
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) ? : 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) ? : 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) && !netif_is_l3_master(dev))
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->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 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2381 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2382 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2385 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2386 const struct prefix_info *pinfo,
2387 struct inet6_dev *in6_dev,
2388 const struct in6_addr *addr, int addr_type,
2389 u32 addr_flags, bool sllao, bool tokenized,
2390 __u32 valid_lft, u32 prefered_lft)
2392 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2393 int create = 0, update_lft = 0;
2395 if (!ifp && valid_lft) {
2396 int max_addresses = in6_dev->cnf.max_addresses;
2398 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2399 if (in6_dev->cnf.optimistic_dad &&
2400 !net->ipv6.devconf_all->forwarding && sllao)
2401 addr_flags |= IFA_F_OPTIMISTIC;
2404 /* Do not allow to create too much of autoconfigured
2405 * addresses; this would be too easy way to crash kernel.
2407 if (!max_addresses ||
2408 ipv6_count_addresses(in6_dev) < max_addresses)
2409 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2411 addr_type&IPV6_ADDR_SCOPE_MASK,
2412 addr_flags, valid_lft,
2415 if (IS_ERR_OR_NULL(ifp))
2420 spin_lock_bh(&ifp->lock);
2421 ifp->flags |= IFA_F_MANAGETEMPADDR;
2422 ifp->cstamp = jiffies;
2423 ifp->tokenized = tokenized;
2424 spin_unlock_bh(&ifp->lock);
2425 addrconf_dad_start(ifp);
2433 /* update lifetime (RFC2462 5.5.3 e) */
2434 spin_lock_bh(&ifp->lock);
2436 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2437 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2440 if (!update_lft && !create && stored_lft) {
2441 const u32 minimum_lft = min_t(u32,
2442 stored_lft, MIN_VALID_LIFETIME);
2443 valid_lft = max(valid_lft, minimum_lft);
2445 /* RFC4862 Section 5.5.3e:
2446 * "Note that the preferred lifetime of the
2447 * corresponding address is always reset to
2448 * the Preferred Lifetime in the received
2449 * Prefix Information option, regardless of
2450 * whether the valid lifetime is also reset or
2453 * So we should always update prefered_lft here.
2459 ifp->valid_lft = valid_lft;
2460 ifp->prefered_lft = prefered_lft;
2463 ifp->flags &= ~IFA_F_DEPRECATED;
2464 spin_unlock_bh(&ifp->lock);
2466 if (!(flags&IFA_F_TENTATIVE))
2467 ipv6_ifa_notify(0, ifp);
2469 spin_unlock_bh(&ifp->lock);
2471 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2480 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2482 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2484 struct prefix_info *pinfo;
2489 struct inet6_dev *in6_dev;
2490 struct net *net = dev_net(dev);
2492 pinfo = (struct prefix_info *) opt;
2494 if (len < sizeof(struct prefix_info)) {
2495 ADBG("addrconf: prefix option too short\n");
2500 * Validation checks ([ADDRCONF], page 19)
2503 addr_type = ipv6_addr_type(&pinfo->prefix);
2505 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2508 valid_lft = ntohl(pinfo->valid);
2509 prefered_lft = ntohl(pinfo->prefered);
2511 if (prefered_lft > valid_lft) {
2512 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2516 in6_dev = in6_dev_get(dev);
2519 net_dbg_ratelimited("addrconf: device %s not configured\n",
2525 * Two things going on here:
2526 * 1) Add routes for on-link prefixes
2527 * 2) Configure prefixes with the auto flag set
2530 if (pinfo->onlink) {
2531 struct rt6_info *rt;
2532 unsigned long rt_expires;
2534 /* Avoid arithmetic overflow. Really, we could
2535 * save rt_expires in seconds, likely valid_lft,
2536 * but it would require division in fib gc, that it
2540 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2542 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2544 if (addrconf_finite_timeout(rt_expires))
2547 rt = addrconf_get_prefix_route(&pinfo->prefix,
2550 RTF_ADDRCONF | RTF_PREFIX_RT,
2551 RTF_GATEWAY | RTF_DEFAULT);
2554 /* Autoconf prefix route */
2555 if (valid_lft == 0) {
2558 } else if (addrconf_finite_timeout(rt_expires)) {
2560 rt6_set_expires(rt, jiffies + rt_expires);
2562 rt6_clean_expires(rt);
2564 } else if (valid_lft) {
2565 clock_t expires = 0;
2566 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2567 if (addrconf_finite_timeout(rt_expires)) {
2569 flags |= RTF_EXPIRES;
2570 expires = jiffies_to_clock_t(rt_expires);
2572 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2573 dev, expires, flags);
2578 /* Try to figure out our local address for this prefix */
2580 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2581 struct in6_addr addr;
2582 bool tokenized = false, dev_addr_generated = false;
2584 if (pinfo->prefix_len == 64) {
2585 memcpy(&addr, &pinfo->prefix, 8);
2587 if (!ipv6_addr_any(&in6_dev->token)) {
2588 read_lock_bh(&in6_dev->lock);
2589 memcpy(addr.s6_addr + 8,
2590 in6_dev->token.s6_addr + 8, 8);
2591 read_unlock_bh(&in6_dev->lock);
2593 } else if (is_addr_mode_generate_stable(in6_dev) &&
2594 !ipv6_generate_stable_address(&addr, 0,
2596 addr_flags |= IFA_F_STABLE_PRIVACY;
2598 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2599 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2602 dev_addr_generated = true;
2606 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2611 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2614 tokenized, valid_lft,
2619 /* Ignore error case here because previous prefix add addr was
2620 * successful which will be notified.
2622 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2623 addr_type, addr_flags, sllao,
2624 tokenized, valid_lft,
2626 dev_addr_generated);
2628 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2630 in6_dev_put(in6_dev);
2634 * Set destination address.
2635 * Special case for SIT interfaces where we create a new "virtual"
2638 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2640 struct in6_ifreq ireq;
2641 struct net_device *dev;
2647 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2650 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2656 #if IS_ENABLED(CONFIG_IPV6_SIT)
2657 if (dev->type == ARPHRD_SIT) {
2658 const struct net_device_ops *ops = dev->netdev_ops;
2660 struct ip_tunnel_parm p;
2662 err = -EADDRNOTAVAIL;
2663 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2666 memset(&p, 0, sizeof(p));
2667 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2671 p.iph.protocol = IPPROTO_IPV6;
2673 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2675 if (ops->ndo_do_ioctl) {
2676 mm_segment_t oldfs = get_fs();
2679 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2686 dev = __dev_get_by_name(net, p.name);
2689 err = dev_open(dev);
2699 static int ipv6_mc_config(struct sock *sk, bool join,
2700 const struct in6_addr *addr, int ifindex)
2708 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2710 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2717 * Manual configuration of address on an interface
2719 static int inet6_addr_add(struct net *net, int ifindex,
2720 const struct in6_addr *pfx,
2721 const struct in6_addr *peer_pfx,
2722 unsigned int plen, __u32 ifa_flags,
2723 __u32 prefered_lft, __u32 valid_lft)
2725 struct inet6_ifaddr *ifp;
2726 struct inet6_dev *idev;
2727 struct net_device *dev;
2728 unsigned long timeout;
2738 /* check the lifetime */
2739 if (!valid_lft || prefered_lft > valid_lft)
2742 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2745 dev = __dev_get_by_index(net, ifindex);
2749 idev = addrconf_add_dev(dev);
2751 return PTR_ERR(idev);
2753 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2754 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2755 true, pfx, ifindex);
2761 scope = ipv6_addr_scope(pfx);
2763 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2764 if (addrconf_finite_timeout(timeout)) {
2765 expires = jiffies_to_clock_t(timeout * HZ);
2766 valid_lft = timeout;
2767 flags = RTF_EXPIRES;
2771 ifa_flags |= IFA_F_PERMANENT;
2774 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2775 if (addrconf_finite_timeout(timeout)) {
2777 ifa_flags |= IFA_F_DEPRECATED;
2778 prefered_lft = timeout;
2781 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2782 valid_lft, prefered_lft);
2785 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2786 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2791 * Note that section 3.1 of RFC 4429 indicates
2792 * that the Optimistic flag should not be set for
2793 * manually configured addresses
2795 addrconf_dad_start(ifp);
2796 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2797 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2800 addrconf_verify_rtnl();
2802 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2803 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2804 false, pfx, ifindex);
2807 return PTR_ERR(ifp);
2810 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2811 const struct in6_addr *pfx, unsigned int plen)
2813 struct inet6_ifaddr *ifp;
2814 struct inet6_dev *idev;
2815 struct net_device *dev;
2820 dev = __dev_get_by_index(net, ifindex);
2824 idev = __in6_dev_get(dev);
2828 read_lock_bh(&idev->lock);
2829 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2830 if (ifp->prefix_len == plen &&
2831 ipv6_addr_equal(pfx, &ifp->addr)) {
2833 read_unlock_bh(&idev->lock);
2835 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2836 (ifa_flags & IFA_F_MANAGETEMPADDR))
2837 manage_tempaddrs(idev, ifp, 0, 0, false,
2840 addrconf_verify_rtnl();
2841 if (ipv6_addr_is_multicast(pfx)) {
2842 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2843 false, pfx, dev->ifindex);
2848 read_unlock_bh(&idev->lock);
2849 return -EADDRNOTAVAIL;
2853 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2855 struct in6_ifreq ireq;
2858 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2861 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2865 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2866 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2867 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2872 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2874 struct in6_ifreq ireq;
2877 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2880 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2884 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2885 ireq.ifr6_prefixlen);
2890 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2891 int plen, int scope)
2893 struct inet6_ifaddr *ifp;
2895 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2896 scope, IFA_F_PERMANENT,
2897 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2899 spin_lock_bh(&ifp->lock);
2900 ifp->flags &= ~IFA_F_TENTATIVE;
2901 spin_unlock_bh(&ifp->lock);
2902 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2907 #if IS_ENABLED(CONFIG_IPV6_SIT)
2908 static void sit_add_v4_addrs(struct inet6_dev *idev)
2910 struct in6_addr addr;
2911 struct net_device *dev;
2912 struct net *net = dev_net(idev->dev);
2918 memset(&addr, 0, sizeof(struct in6_addr));
2919 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2921 if (idev->dev->flags&IFF_POINTOPOINT) {
2922 addr.s6_addr32[0] = htonl(0xfe800000);
2926 scope = IPV6_ADDR_COMPATv4;
2928 pflags |= RTF_NONEXTHOP;
2931 if (addr.s6_addr32[3]) {
2932 add_addr(idev, &addr, plen, scope);
2933 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2937 for_each_netdev(net, dev) {
2938 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2939 if (in_dev && (dev->flags & IFF_UP)) {
2940 struct in_ifaddr *ifa;
2944 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2946 addr.s6_addr32[3] = ifa->ifa_local;
2948 if (ifa->ifa_scope == RT_SCOPE_LINK)
2950 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2951 if (idev->dev->flags&IFF_POINTOPOINT)
2956 add_addr(idev, &addr, plen, flag);
2957 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2965 static void init_loopback(struct net_device *dev)
2967 struct inet6_dev *idev;
2968 struct net_device *sp_dev;
2969 struct inet6_ifaddr *sp_ifa;
2970 struct rt6_info *sp_rt;
2976 idev = ipv6_find_idev(dev);
2978 pr_debug("%s: add_dev failed\n", __func__);
2982 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2984 /* Add routes to other interface's IPv6 addresses */
2985 for_each_netdev(dev_net(dev), sp_dev) {
2986 if (!strcmp(sp_dev->name, dev->name))
2989 idev = __in6_dev_get(sp_dev);
2993 read_lock_bh(&idev->lock);
2994 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2996 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
3000 /* This dst has been added to garbage list when
3001 * lo device down, release this obsolete dst and
3002 * reallocate a new router for ifa.
3004 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3005 ip6_rt_put(sp_ifa->rt);
3012 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3014 /* Failure cases are ignored */
3015 if (!IS_ERR(sp_rt)) {
3020 read_unlock_bh(&idev->lock);
3024 void addrconf_add_linklocal(struct inet6_dev *idev,
3025 const struct in6_addr *addr, u32 flags)
3027 struct inet6_ifaddr *ifp;
3028 u32 addr_flags = flags | IFA_F_PERMANENT;
3030 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3031 if (idev->cnf.optimistic_dad &&
3032 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3033 addr_flags |= IFA_F_OPTIMISTIC;
3036 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3037 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3039 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3040 addrconf_dad_start(ifp);
3044 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3046 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3048 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3051 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3052 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3055 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3056 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3062 static int ipv6_generate_stable_address(struct in6_addr *address,
3064 const struct inet6_dev *idev)
3066 static DEFINE_SPINLOCK(lock);
3067 static __u32 digest[SHA_DIGEST_WORDS];
3068 static __u32 workspace[SHA_WORKSPACE_WORDS];
3071 char __data[SHA_MESSAGE_BYTES];
3073 struct in6_addr secret;
3075 unsigned char hwaddr[MAX_ADDR_LEN];
3080 struct in6_addr secret;
3081 struct in6_addr temp;
3082 struct net *net = dev_net(idev->dev);
3084 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3086 if (idev->cnf.stable_secret.initialized)
3087 secret = idev->cnf.stable_secret.secret;
3088 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3089 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3094 spin_lock_bh(&lock);
3097 memset(&data, 0, sizeof(data));
3098 memset(workspace, 0, sizeof(workspace));
3099 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3100 data.prefix[0] = address->s6_addr32[0];
3101 data.prefix[1] = address->s6_addr32[1];
3102 data.secret = secret;
3103 data.dad_count = dad_count;
3105 sha_transform(digest, data.__data, workspace);
3108 temp.s6_addr32[2] = (__force __be32)digest[0];
3109 temp.s6_addr32[3] = (__force __be32)digest[1];
3111 spin_unlock_bh(&lock);
3113 if (ipv6_reserved_interfaceid(temp)) {
3115 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3124 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3126 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3130 s = &idev->cnf.stable_secret;
3131 get_random_bytes(&s->secret, sizeof(s->secret));
3132 s->initialized = true;
3135 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3137 struct in6_addr addr;
3139 /* no link local addresses on L3 master devices */
3140 if (netif_is_l3_master(idev->dev))
3143 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3145 switch (idev->addr_gen_mode) {
3146 case IN6_ADDR_GEN_MODE_RANDOM:
3147 ipv6_gen_mode_random_init(idev);
3149 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3150 if (!ipv6_generate_stable_address(&addr, 0, idev))
3151 addrconf_add_linklocal(idev, &addr,
3152 IFA_F_STABLE_PRIVACY);
3153 else if (prefix_route)
3154 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3156 case IN6_ADDR_GEN_MODE_EUI64:
3157 /* addrconf_add_linklocal also adds a prefix_route and we
3158 * only need to care about prefix routes if ipv6_generate_eui64
3159 * couldn't generate one.
3161 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3162 addrconf_add_linklocal(idev, &addr, 0);
3163 else if (prefix_route)
3164 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3166 case IN6_ADDR_GEN_MODE_NONE:
3168 /* will not add any link local address */
3173 static void addrconf_dev_config(struct net_device *dev)
3175 struct inet6_dev *idev;
3179 if ((dev->type != ARPHRD_ETHER) &&
3180 (dev->type != ARPHRD_FDDI) &&
3181 (dev->type != ARPHRD_ARCNET) &&
3182 (dev->type != ARPHRD_INFINIBAND) &&
3183 (dev->type != ARPHRD_IEEE1394) &&
3184 (dev->type != ARPHRD_TUNNEL6) &&
3185 (dev->type != ARPHRD_6LOWPAN) &&
3186 (dev->type != ARPHRD_NONE)) {
3187 /* Alas, we support only Ethernet autoconfiguration. */
3191 idev = addrconf_add_dev(dev);
3195 /* this device type has no EUI support */
3196 if (dev->type == ARPHRD_NONE &&
3197 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3198 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3200 addrconf_addr_gen(idev, false);
3203 #if IS_ENABLED(CONFIG_IPV6_SIT)
3204 static void addrconf_sit_config(struct net_device *dev)
3206 struct inet6_dev *idev;
3211 * Configure the tunnel with one of our IPv4
3212 * addresses... we should configure all of
3213 * our v4 addrs in the tunnel
3216 idev = ipv6_find_idev(dev);
3218 pr_debug("%s: add_dev failed\n", __func__);
3222 if (dev->priv_flags & IFF_ISATAP) {
3223 addrconf_addr_gen(idev, false);
3227 sit_add_v4_addrs(idev);
3229 if (dev->flags&IFF_POINTOPOINT)
3230 addrconf_add_mroute(dev);
3234 #if IS_ENABLED(CONFIG_NET_IPGRE)
3235 static void addrconf_gre_config(struct net_device *dev)
3237 struct inet6_dev *idev;
3241 idev = ipv6_find_idev(dev);
3243 pr_debug("%s: add_dev failed\n", __func__);
3247 addrconf_addr_gen(idev, true);
3248 if (dev->flags & IFF_POINTOPOINT)
3249 addrconf_add_mroute(dev);
3253 static int fixup_permanent_addr(struct inet6_dev *idev,
3254 struct inet6_ifaddr *ifp)
3257 struct rt6_info *rt;
3259 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3260 if (unlikely(IS_ERR(rt)))
3266 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3267 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3271 addrconf_dad_start(ifp);
3276 static void addrconf_permanent_addr(struct net_device *dev)
3278 struct inet6_ifaddr *ifp, *tmp;
3279 struct inet6_dev *idev;
3281 idev = __in6_dev_get(dev);
3285 write_lock_bh(&idev->lock);
3287 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3288 if ((ifp->flags & IFA_F_PERMANENT) &&
3289 fixup_permanent_addr(idev, ifp) < 0) {
3290 write_unlock_bh(&idev->lock);
3292 write_lock_bh(&idev->lock);
3294 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3295 idev->dev->name, &ifp->addr);
3299 write_unlock_bh(&idev->lock);
3302 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3305 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3306 struct netdev_notifier_changeupper_info *info;
3307 struct inet6_dev *idev = __in6_dev_get(dev);
3308 int run_pending = 0;
3312 case NETDEV_REGISTER:
3313 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3314 idev = ipv6_add_dev(dev);
3316 return notifier_from_errno(PTR_ERR(idev));
3320 case NETDEV_CHANGEMTU:
3321 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3322 if (dev->mtu < IPV6_MIN_MTU) {
3323 addrconf_ifdown(dev, 1);
3328 rt6_mtu_change(dev, dev->mtu);
3329 idev->cnf.mtu6 = dev->mtu;
3333 /* allocate new idev */
3334 idev = ipv6_add_dev(dev);
3338 /* device is still not ready */
3339 if (!(idev->if_flags & IF_READY))
3348 if (dev->flags & IFF_SLAVE)
3351 if (idev && idev->cnf.disable_ipv6)
3354 if (event == NETDEV_UP) {
3355 /* restore routes for permanent addresses */
3356 addrconf_permanent_addr(dev);
3358 if (!addrconf_qdisc_ok(dev)) {
3359 /* device is not ready yet. */
3360 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3365 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3366 idev = ipv6_add_dev(dev);
3368 if (!IS_ERR_OR_NULL(idev)) {
3369 idev->if_flags |= IF_READY;
3372 } else if (event == NETDEV_CHANGE) {
3373 if (!addrconf_qdisc_ok(dev)) {
3374 /* device is still not ready. */
3379 if (idev->if_flags & IF_READY)
3380 /* device is already configured. */
3382 idev->if_flags |= IF_READY;
3385 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3391 switch (dev->type) {
3392 #if IS_ENABLED(CONFIG_IPV6_SIT)
3394 addrconf_sit_config(dev);
3397 #if IS_ENABLED(CONFIG_NET_IPGRE)
3399 addrconf_gre_config(dev);
3402 case ARPHRD_LOOPBACK:
3407 addrconf_dev_config(dev);
3411 if (!IS_ERR_OR_NULL(idev)) {
3413 addrconf_dad_run(idev);
3416 * If the MTU changed during the interface down,
3417 * when the interface up, the changed MTU must be
3418 * reflected in the idev as well as routers.
3420 if (idev->cnf.mtu6 != dev->mtu &&
3421 dev->mtu >= IPV6_MIN_MTU) {
3422 rt6_mtu_change(dev, dev->mtu);
3423 idev->cnf.mtu6 = dev->mtu;
3425 idev->tstamp = jiffies;
3426 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3429 * If the changed mtu during down is lower than
3430 * IPV6_MIN_MTU stop IPv6 on this interface.
3432 if (dev->mtu < IPV6_MIN_MTU)
3433 addrconf_ifdown(dev, 1);
3438 case NETDEV_UNREGISTER:
3440 * Remove all addresses from this interface.
3442 addrconf_ifdown(dev, event != NETDEV_DOWN);
3445 case NETDEV_CHANGENAME:
3447 snmp6_unregister_dev(idev);
3448 addrconf_sysctl_unregister(idev);
3449 err = addrconf_sysctl_register(idev);
3451 return notifier_from_errno(err);
3452 err = snmp6_register_dev(idev);
3454 addrconf_sysctl_unregister(idev);
3455 return notifier_from_errno(err);
3460 case NETDEV_PRE_TYPE_CHANGE:
3461 case NETDEV_POST_TYPE_CHANGE:
3463 addrconf_type_change(dev, event);
3466 case NETDEV_CHANGEUPPER:
3469 /* flush all routes if dev is linked to or unlinked from
3470 * an L3 master device (e.g., VRF)
3472 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3473 addrconf_ifdown(dev, 0);
3480 * addrconf module should be notified of a device going up
3482 static struct notifier_block ipv6_dev_notf = {
3483 .notifier_call = addrconf_notify,
3486 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3488 struct inet6_dev *idev;
3491 idev = __in6_dev_get(dev);
3493 if (event == NETDEV_POST_TYPE_CHANGE)
3494 ipv6_mc_remap(idev);
3495 else if (event == NETDEV_PRE_TYPE_CHANGE)
3496 ipv6_mc_unmap(idev);
3499 static bool addr_is_local(const struct in6_addr *addr)
3501 return ipv6_addr_type(addr) &
3502 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3505 static int addrconf_ifdown(struct net_device *dev, int how)
3507 struct net *net = dev_net(dev);
3508 struct inet6_dev *idev;
3509 struct inet6_ifaddr *ifa, *tmp;
3510 struct list_head del_list;
3517 rt6_ifdown(net, dev);
3518 neigh_ifdown(&nd_tbl, dev);
3520 idev = __in6_dev_get(dev);
3525 * Step 1: remove reference to ipv6 device from parent device.
3531 /* protected by rtnl_lock */
3532 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3534 /* Step 1.5: remove snmp6 entry */
3535 snmp6_unregister_dev(idev);
3539 /* aggregate the system setting and interface setting */
3540 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3542 _keep_addr = idev->cnf.keep_addr_on_down;
3544 /* combine the user config with event to determine if permanent
3545 * addresses are to be removed from address hash table
3547 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3549 /* Step 2: clear hash table */
3550 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3551 struct hlist_head *h = &inet6_addr_lst[i];
3553 spin_lock_bh(&addrconf_hash_lock);
3555 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3556 if (ifa->idev == idev) {
3557 addrconf_del_dad_work(ifa);
3558 /* combined flag + permanent flag decide if
3559 * address is retained on a down event
3562 !(ifa->flags & IFA_F_PERMANENT) ||
3563 addr_is_local(&ifa->addr)) {
3564 hlist_del_init_rcu(&ifa->addr_lst);
3569 spin_unlock_bh(&addrconf_hash_lock);
3572 write_lock_bh(&idev->lock);
3574 addrconf_del_rs_timer(idev);
3576 /* Step 2: clear flags for stateless addrconf */
3578 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3580 /* Step 3: clear tempaddr list */
3581 while (!list_empty(&idev->tempaddr_list)) {
3582 ifa = list_first_entry(&idev->tempaddr_list,
3583 struct inet6_ifaddr, tmp_list);
3584 list_del(&ifa->tmp_list);
3585 write_unlock_bh(&idev->lock);
3586 spin_lock_bh(&ifa->lock);
3589 in6_ifa_put(ifa->ifpub);
3592 spin_unlock_bh(&ifa->lock);
3594 write_lock_bh(&idev->lock);
3597 /* re-combine the user config with event to determine if permanent
3598 * addresses are to be removed from the interface list
3600 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3602 INIT_LIST_HEAD(&del_list);
3603 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3604 struct rt6_info *rt = NULL;
3606 addrconf_del_dad_work(ifa);
3608 write_unlock_bh(&idev->lock);
3609 spin_lock_bh(&ifa->lock);
3611 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3612 !addr_is_local(&ifa->addr)) {
3613 /* set state to skip the notifier below */
3614 state = INET6_IFADDR_STATE_DEAD;
3616 if (!(ifa->flags & IFA_F_NODAD))
3617 ifa->flags |= IFA_F_TENTATIVE;
3623 ifa->state = INET6_IFADDR_STATE_DEAD;
3625 list_move(&ifa->if_list, &del_list);
3628 spin_unlock_bh(&ifa->lock);
3633 if (state != INET6_IFADDR_STATE_DEAD) {
3634 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3635 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3637 if (idev->cnf.forwarding)
3638 addrconf_leave_anycast(ifa);
3639 addrconf_leave_solict(ifa->idev, &ifa->addr);
3642 write_lock_bh(&idev->lock);
3645 write_unlock_bh(&idev->lock);
3647 /* now clean up addresses to be removed */
3648 while (!list_empty(&del_list)) {
3649 ifa = list_first_entry(&del_list,
3650 struct inet6_ifaddr, if_list);
3651 list_del(&ifa->if_list);
3656 /* Step 5: Discard anycast and multicast list */
3658 ipv6_ac_destroy_dev(idev);
3659 ipv6_mc_destroy_dev(idev);
3664 idev->tstamp = jiffies;
3666 /* Last: Shot the device (if unregistered) */
3668 addrconf_sysctl_unregister(idev);
3669 neigh_parms_release(&nd_tbl, idev->nd_parms);
3670 neigh_ifdown(&nd_tbl, dev);
3676 static void addrconf_rs_timer(unsigned long data)
3678 struct inet6_dev *idev = (struct inet6_dev *)data;
3679 struct net_device *dev = idev->dev;
3680 struct in6_addr lladdr;
3682 write_lock(&idev->lock);
3683 if (idev->dead || !(idev->if_flags & IF_READY))
3686 if (!ipv6_accept_ra(idev))
3689 /* Announcement received after solicitation was sent */
3690 if (idev->if_flags & IF_RA_RCVD)
3693 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3694 write_unlock(&idev->lock);
3695 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3696 ndisc_send_rs(dev, &lladdr,
3697 &in6addr_linklocal_allrouters);
3701 write_lock(&idev->lock);
3702 idev->rs_interval = rfc3315_s14_backoff_update(
3703 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3704 /* The wait after the last probe can be shorter */
3705 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3706 idev->cnf.rtr_solicits) ?
3707 idev->cnf.rtr_solicit_delay :
3711 * Note: we do not support deprecated "all on-link"
3712 * assumption any longer.
3714 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3718 write_unlock(&idev->lock);
3724 * Duplicate Address Detection
3726 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3728 unsigned long rand_num;
3729 struct inet6_dev *idev = ifp->idev;
3731 if (ifp->flags & IFA_F_OPTIMISTIC)
3734 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3736 ifp->dad_probes = idev->cnf.dad_transmits;
3737 addrconf_mod_dad_work(ifp, rand_num);
3740 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3742 struct inet6_dev *idev = ifp->idev;
3743 struct net_device *dev = idev->dev;
3744 bool notify = false;
3746 addrconf_join_solict(dev, &ifp->addr);
3748 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3750 read_lock_bh(&idev->lock);
3751 spin_lock(&ifp->lock);
3752 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3755 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3756 idev->cnf.accept_dad < 1 ||
3757 !(ifp->flags&IFA_F_TENTATIVE) ||
3758 ifp->flags & IFA_F_NODAD) {
3759 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3760 spin_unlock(&ifp->lock);
3761 read_unlock_bh(&idev->lock);
3763 addrconf_dad_completed(ifp);
3767 if (!(idev->if_flags & IF_READY)) {
3768 spin_unlock(&ifp->lock);
3769 read_unlock_bh(&idev->lock);
3771 * If the device is not ready:
3772 * - keep it tentative if it is a permanent address.
3773 * - otherwise, kill it.
3776 addrconf_dad_stop(ifp, 0);
3781 * Optimistic nodes can start receiving
3784 if (ifp->flags & IFA_F_OPTIMISTIC) {
3785 ip6_ins_rt(ifp->rt);
3786 if (ipv6_use_optimistic_addr(idev)) {
3787 /* Because optimistic nodes can use this address,
3788 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3794 addrconf_dad_kick(ifp);
3796 spin_unlock(&ifp->lock);
3797 read_unlock_bh(&idev->lock);
3799 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3802 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3804 bool begin_dad = false;
3806 spin_lock_bh(&ifp->lock);
3807 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3808 ifp->state = INET6_IFADDR_STATE_PREDAD;
3811 spin_unlock_bh(&ifp->lock);
3814 addrconf_mod_dad_work(ifp, 0);
3817 static void addrconf_dad_work(struct work_struct *w)
3819 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3820 struct inet6_ifaddr,
3822 struct inet6_dev *idev = ifp->idev;
3823 struct in6_addr mcaddr;
3824 bool disable_ipv6 = false;
3830 } action = DAD_PROCESS;
3834 spin_lock_bh(&ifp->lock);
3835 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3837 ifp->state = INET6_IFADDR_STATE_DAD;
3838 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3840 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3842 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3843 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3844 struct in6_addr addr;
3846 addr.s6_addr32[0] = htonl(0xfe800000);
3847 addr.s6_addr32[1] = 0;
3849 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3850 ipv6_addr_equal(&ifp->addr, &addr)) {
3851 /* DAD failed for link-local based on MAC */
3852 idev->cnf.disable_ipv6 = 1;
3854 pr_info("%s: IPv6 being disabled!\n",
3855 ifp->idev->dev->name);
3856 disable_ipv6 = true;
3860 spin_unlock_bh(&ifp->lock);
3862 if (action == DAD_BEGIN) {
3863 addrconf_dad_begin(ifp);
3865 } else if (action == DAD_ABORT) {
3867 addrconf_dad_stop(ifp, 1);
3869 addrconf_ifdown(idev->dev, 0);
3873 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3876 write_lock_bh(&idev->lock);
3877 if (idev->dead || !(idev->if_flags & IF_READY)) {
3878 write_unlock_bh(&idev->lock);
3882 spin_lock(&ifp->lock);
3883 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3884 spin_unlock(&ifp->lock);
3885 write_unlock_bh(&idev->lock);
3889 if (ifp->dad_probes == 0) {
3891 * DAD was successful
3894 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3895 spin_unlock(&ifp->lock);
3896 write_unlock_bh(&idev->lock);
3898 addrconf_dad_completed(ifp);
3904 addrconf_mod_dad_work(ifp,
3905 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3906 spin_unlock(&ifp->lock);
3907 write_unlock_bh(&idev->lock);
3909 /* send a neighbour solicitation for our addr */
3910 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3911 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3917 /* ifp->idev must be at least read locked */
3918 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3920 struct inet6_ifaddr *ifpiter;
3921 struct inet6_dev *idev = ifp->idev;
3923 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3924 if (ifpiter->scope > IFA_LINK)
3926 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3927 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3928 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3935 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3937 struct net_device *dev = ifp->idev->dev;
3938 struct in6_addr lladdr;
3939 bool send_rs, send_mld;
3941 addrconf_del_dad_work(ifp);
3944 * Configure the address for reception. Now it is valid.
3947 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3949 /* If added prefix is link local and we are prepared to process
3950 router advertisements, start sending router solicitations.
3953 read_lock_bh(&ifp->idev->lock);
3954 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3955 send_rs = send_mld &&
3956 ipv6_accept_ra(ifp->idev) &&
3957 ifp->idev->cnf.rtr_solicits != 0 &&
3958 (dev->flags&IFF_LOOPBACK) == 0;
3959 read_unlock_bh(&ifp->idev->lock);
3961 /* While dad is in progress mld report's source address is in6_addrany.
3962 * Resend with proper ll now.
3965 ipv6_mc_dad_complete(ifp->idev);
3969 * If a host as already performed a random delay
3970 * [...] as part of DAD [...] there is no need
3971 * to delay again before sending the first RS
3973 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3975 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3977 write_lock_bh(&ifp->idev->lock);
3978 spin_lock(&ifp->lock);
3979 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
3980 ifp->idev->cnf.rtr_solicit_interval);
3981 ifp->idev->rs_probes = 1;
3982 ifp->idev->if_flags |= IF_RS_SENT;
3983 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
3984 spin_unlock(&ifp->lock);
3985 write_unlock_bh(&ifp->idev->lock);
3989 static void addrconf_dad_run(struct inet6_dev *idev)
3991 struct inet6_ifaddr *ifp;
3993 read_lock_bh(&idev->lock);
3994 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3995 spin_lock(&ifp->lock);
3996 if (ifp->flags & IFA_F_TENTATIVE &&
3997 ifp->state == INET6_IFADDR_STATE_DAD)
3998 addrconf_dad_kick(ifp);
3999 spin_unlock(&ifp->lock);
4001 read_unlock_bh(&idev->lock);
4004 #ifdef CONFIG_PROC_FS
4005 struct if6_iter_state {
4006 struct seq_net_private p;
4011 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4013 struct inet6_ifaddr *ifa = NULL;
4014 struct if6_iter_state *state = seq->private;
4015 struct net *net = seq_file_net(seq);
4018 /* initial bucket if pos is 0 */
4024 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4025 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4027 if (!net_eq(dev_net(ifa->idev->dev), net))
4029 /* sync with offset */
4030 if (p < state->offset) {
4038 /* prepare for next bucket */
4045 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4046 struct inet6_ifaddr *ifa)
4048 struct if6_iter_state *state = seq->private;
4049 struct net *net = seq_file_net(seq);
4051 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4052 if (!net_eq(dev_net(ifa->idev->dev), net))
4058 while (++state->bucket < IN6_ADDR_HSIZE) {
4060 hlist_for_each_entry_rcu_bh(ifa,
4061 &inet6_addr_lst[state->bucket], addr_lst) {
4062 if (!net_eq(dev_net(ifa->idev->dev), net))
4072 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4076 return if6_get_first(seq, *pos);
4079 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4081 struct inet6_ifaddr *ifa;
4083 ifa = if6_get_next(seq, v);
4088 static void if6_seq_stop(struct seq_file *seq, void *v)
4091 rcu_read_unlock_bh();
4094 static int if6_seq_show(struct seq_file *seq, void *v)
4096 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4097 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4099 ifp->idev->dev->ifindex,
4103 ifp->idev->dev->name);
4107 static const struct seq_operations if6_seq_ops = {
4108 .start = if6_seq_start,
4109 .next = if6_seq_next,
4110 .show = if6_seq_show,
4111 .stop = if6_seq_stop,
4114 static int if6_seq_open(struct inode *inode, struct file *file)
4116 return seq_open_net(inode, file, &if6_seq_ops,
4117 sizeof(struct if6_iter_state));
4120 static const struct file_operations if6_fops = {
4121 .owner = THIS_MODULE,
4122 .open = if6_seq_open,
4124 .llseek = seq_lseek,
4125 .release = seq_release_net,
4128 static int __net_init if6_proc_net_init(struct net *net)
4130 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4135 static void __net_exit if6_proc_net_exit(struct net *net)
4137 remove_proc_entry("if_inet6", net->proc_net);
4140 static struct pernet_operations if6_proc_net_ops = {
4141 .init = if6_proc_net_init,
4142 .exit = if6_proc_net_exit,
4145 int __init if6_proc_init(void)
4147 return register_pernet_subsys(&if6_proc_net_ops);
4150 void if6_proc_exit(void)
4152 unregister_pernet_subsys(&if6_proc_net_ops);
4154 #endif /* CONFIG_PROC_FS */
4156 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4157 /* Check if address is a home address configured on any interface. */
4158 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4161 struct inet6_ifaddr *ifp = NULL;
4162 unsigned int hash = inet6_addr_hash(addr);
4165 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4166 if (!net_eq(dev_net(ifp->idev->dev), net))
4168 if (ipv6_addr_equal(&ifp->addr, addr) &&
4169 (ifp->flags & IFA_F_HOMEADDRESS)) {
4174 rcu_read_unlock_bh();
4180 * Periodic address status verification
4183 static void addrconf_verify_rtnl(void)
4185 unsigned long now, next, next_sec, next_sched;
4186 struct inet6_ifaddr *ifp;
4193 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4195 cancel_delayed_work(&addr_chk_work);
4197 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4199 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4202 /* When setting preferred_lft to a value not zero or
4203 * infinity, while valid_lft is infinity
4204 * IFA_F_PERMANENT has a non-infinity life time.
4206 if ((ifp->flags & IFA_F_PERMANENT) &&
4207 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4210 spin_lock(&ifp->lock);
4211 /* We try to batch several events at once. */
4212 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4214 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4215 age >= ifp->valid_lft) {
4216 spin_unlock(&ifp->lock);
4220 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4221 spin_unlock(&ifp->lock);
4223 } else if (age >= ifp->prefered_lft) {
4224 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4227 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4229 ifp->flags |= IFA_F_DEPRECATED;
4232 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4233 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4234 next = ifp->tstamp + ifp->valid_lft * HZ;
4236 spin_unlock(&ifp->lock);
4241 ipv6_ifa_notify(0, ifp);
4245 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4246 !(ifp->flags&IFA_F_TENTATIVE)) {
4247 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4248 ifp->idev->cnf.dad_transmits *
4249 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4251 if (age >= ifp->prefered_lft - regen_advance) {
4252 struct inet6_ifaddr *ifpub = ifp->ifpub;
4253 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4254 next = ifp->tstamp + ifp->prefered_lft * HZ;
4255 if (!ifp->regen_count && ifpub) {
4258 in6_ifa_hold(ifpub);
4259 spin_unlock(&ifp->lock);
4261 spin_lock(&ifpub->lock);
4262 ifpub->regen_count = 0;
4263 spin_unlock(&ifpub->lock);
4264 ipv6_create_tempaddr(ifpub, ifp);
4269 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4270 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4271 spin_unlock(&ifp->lock);
4273 /* ifp->prefered_lft <= ifp->valid_lft */
4274 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4275 next = ifp->tstamp + ifp->prefered_lft * HZ;
4276 spin_unlock(&ifp->lock);
4281 next_sec = round_jiffies_up(next);
4284 /* If rounded timeout is accurate enough, accept it. */
4285 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4286 next_sched = next_sec;
4288 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4289 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4290 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4292 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4293 now, next, next_sec, next_sched);
4294 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4295 rcu_read_unlock_bh();
4298 static void addrconf_verify_work(struct work_struct *w)
4301 addrconf_verify_rtnl();
4305 static void addrconf_verify(void)
4307 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4310 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4311 struct in6_addr **peer_pfx)
4313 struct in6_addr *pfx = NULL;
4318 pfx = nla_data(addr);
4321 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4323 pfx = nla_data(local);
4329 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4330 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4331 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4332 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4333 [IFA_FLAGS] = { .len = sizeof(u32) },
4337 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4339 struct net *net = sock_net(skb->sk);
4340 struct ifaddrmsg *ifm;
4341 struct nlattr *tb[IFA_MAX+1];
4342 struct in6_addr *pfx, *peer_pfx;
4346 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4350 ifm = nlmsg_data(nlh);
4351 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4355 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4357 /* We ignore other flags so far. */
4358 ifa_flags &= IFA_F_MANAGETEMPADDR;
4360 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4361 ifm->ifa_prefixlen);
4364 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4365 u32 prefered_lft, u32 valid_lft)
4369 unsigned long timeout;
4370 bool was_managetempaddr;
4371 bool had_prefixroute;
4375 if (!valid_lft || (prefered_lft > valid_lft))
4378 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4379 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4382 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4383 if (addrconf_finite_timeout(timeout)) {
4384 expires = jiffies_to_clock_t(timeout * HZ);
4385 valid_lft = timeout;
4386 flags = RTF_EXPIRES;
4390 ifa_flags |= IFA_F_PERMANENT;
4393 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4394 if (addrconf_finite_timeout(timeout)) {
4396 ifa_flags |= IFA_F_DEPRECATED;
4397 prefered_lft = timeout;
4400 spin_lock_bh(&ifp->lock);
4401 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4402 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4403 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4404 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4405 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4406 IFA_F_NOPREFIXROUTE);
4407 ifp->flags |= ifa_flags;
4408 ifp->tstamp = jiffies;
4409 ifp->valid_lft = valid_lft;
4410 ifp->prefered_lft = prefered_lft;
4412 spin_unlock_bh(&ifp->lock);
4413 if (!(ifp->flags&IFA_F_TENTATIVE))
4414 ipv6_ifa_notify(0, ifp);
4416 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4417 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4419 } else if (had_prefixroute) {
4420 enum cleanup_prefix_rt_t action;
4421 unsigned long rt_expires;
4423 write_lock_bh(&ifp->idev->lock);
4424 action = check_cleanup_prefix_route(ifp, &rt_expires);
4425 write_unlock_bh(&ifp->idev->lock);
4427 if (action != CLEANUP_PREFIX_RT_NOP) {
4428 cleanup_prefix_route(ifp, rt_expires,
4429 action == CLEANUP_PREFIX_RT_DEL);
4433 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4434 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4435 valid_lft = prefered_lft = 0;
4436 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4437 !was_managetempaddr, jiffies);
4440 addrconf_verify_rtnl();
4446 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4448 struct net *net = sock_net(skb->sk);
4449 struct ifaddrmsg *ifm;
4450 struct nlattr *tb[IFA_MAX+1];
4451 struct in6_addr *pfx, *peer_pfx;
4452 struct inet6_ifaddr *ifa;
4453 struct net_device *dev;
4454 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4458 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4462 ifm = nlmsg_data(nlh);
4463 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4467 if (tb[IFA_CACHEINFO]) {
4468 struct ifa_cacheinfo *ci;
4470 ci = nla_data(tb[IFA_CACHEINFO]);
4471 valid_lft = ci->ifa_valid;
4472 preferred_lft = ci->ifa_prefered;
4474 preferred_lft = INFINITY_LIFE_TIME;
4475 valid_lft = INFINITY_LIFE_TIME;
4478 dev = __dev_get_by_index(net, ifm->ifa_index);
4482 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4484 /* We ignore other flags so far. */
4485 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4486 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4488 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4491 * It would be best to check for !NLM_F_CREATE here but
4492 * userspace already relies on not having to provide this.
4494 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4495 ifm->ifa_prefixlen, ifa_flags,
4496 preferred_lft, valid_lft);
4499 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4500 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4503 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4510 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4511 u8 scope, int ifindex)
4513 struct ifaddrmsg *ifm;
4515 ifm = nlmsg_data(nlh);
4516 ifm->ifa_family = AF_INET6;
4517 ifm->ifa_prefixlen = prefixlen;
4518 ifm->ifa_flags = flags;
4519 ifm->ifa_scope = scope;
4520 ifm->ifa_index = ifindex;
4523 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4524 unsigned long tstamp, u32 preferred, u32 valid)
4526 struct ifa_cacheinfo ci;
4528 ci.cstamp = cstamp_delta(cstamp);
4529 ci.tstamp = cstamp_delta(tstamp);
4530 ci.ifa_prefered = preferred;
4531 ci.ifa_valid = valid;
4533 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4536 static inline int rt_scope(int ifa_scope)
4538 if (ifa_scope & IFA_HOST)
4539 return RT_SCOPE_HOST;
4540 else if (ifa_scope & IFA_LINK)
4541 return RT_SCOPE_LINK;
4542 else if (ifa_scope & IFA_SITE)
4543 return RT_SCOPE_SITE;
4545 return RT_SCOPE_UNIVERSE;
4548 static inline int inet6_ifaddr_msgsize(void)
4550 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4551 + nla_total_size(16) /* IFA_LOCAL */
4552 + nla_total_size(16) /* IFA_ADDRESS */
4553 + nla_total_size(sizeof(struct ifa_cacheinfo))
4554 + nla_total_size(4) /* IFA_FLAGS */;
4557 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4558 u32 portid, u32 seq, int event, unsigned int flags)
4560 struct nlmsghdr *nlh;
4561 u32 preferred, valid;
4563 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4567 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4568 ifa->idev->dev->ifindex);
4570 if (!((ifa->flags&IFA_F_PERMANENT) &&
4571 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4572 preferred = ifa->prefered_lft;
4573 valid = ifa->valid_lft;
4574 if (preferred != INFINITY_LIFE_TIME) {
4575 long tval = (jiffies - ifa->tstamp)/HZ;
4576 if (preferred > tval)
4580 if (valid != INFINITY_LIFE_TIME) {
4588 preferred = INFINITY_LIFE_TIME;
4589 valid = INFINITY_LIFE_TIME;
4592 if (!ipv6_addr_any(&ifa->peer_addr)) {
4593 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4594 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4597 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4600 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4603 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4606 nlmsg_end(skb, nlh);
4610 nlmsg_cancel(skb, nlh);
4614 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4615 u32 portid, u32 seq, int event, u16 flags)
4617 struct nlmsghdr *nlh;
4618 u8 scope = RT_SCOPE_UNIVERSE;
4619 int ifindex = ifmca->idev->dev->ifindex;
4621 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4622 scope = RT_SCOPE_SITE;
4624 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4628 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4629 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4630 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4631 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4632 nlmsg_cancel(skb, nlh);
4636 nlmsg_end(skb, nlh);
4640 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4641 u32 portid, u32 seq, int event, unsigned int flags)
4643 struct nlmsghdr *nlh;
4644 u8 scope = RT_SCOPE_UNIVERSE;
4645 int ifindex = ifaca->aca_idev->dev->ifindex;
4647 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4648 scope = RT_SCOPE_SITE;
4650 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4654 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4655 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4656 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4657 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4658 nlmsg_cancel(skb, nlh);
4662 nlmsg_end(skb, nlh);
4672 /* called with rcu_read_lock() */
4673 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4674 struct netlink_callback *cb, enum addr_type_t type,
4675 int s_ip_idx, int *p_ip_idx)
4677 struct ifmcaddr6 *ifmca;
4678 struct ifacaddr6 *ifaca;
4680 int ip_idx = *p_ip_idx;
4682 read_lock_bh(&idev->lock);
4684 case UNICAST_ADDR: {
4685 struct inet6_ifaddr *ifa;
4687 /* unicast address incl. temp addr */
4688 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4689 if (++ip_idx < s_ip_idx)
4691 err = inet6_fill_ifaddr(skb, ifa,
4692 NETLINK_CB(cb->skb).portid,
4698 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4702 case MULTICAST_ADDR:
4703 /* multicast address */
4704 for (ifmca = idev->mc_list; ifmca;
4705 ifmca = ifmca->next, ip_idx++) {
4706 if (ip_idx < s_ip_idx)
4708 err = inet6_fill_ifmcaddr(skb, ifmca,
4709 NETLINK_CB(cb->skb).portid,
4718 /* anycast address */
4719 for (ifaca = idev->ac_list; ifaca;
4720 ifaca = ifaca->aca_next, ip_idx++) {
4721 if (ip_idx < s_ip_idx)
4723 err = inet6_fill_ifacaddr(skb, ifaca,
4724 NETLINK_CB(cb->skb).portid,
4735 read_unlock_bh(&idev->lock);
4740 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4741 enum addr_type_t type)
4743 struct net *net = sock_net(skb->sk);
4746 int s_idx, s_ip_idx;
4747 struct net_device *dev;
4748 struct inet6_dev *idev;
4749 struct hlist_head *head;
4752 s_idx = idx = cb->args[1];
4753 s_ip_idx = ip_idx = cb->args[2];
4756 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4757 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4759 head = &net->dev_index_head[h];
4760 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4763 if (h > s_h || idx > s_idx)
4766 idev = __in6_dev_get(dev);
4770 if (in6_dump_addrs(idev, skb, cb, type,
4771 s_ip_idx, &ip_idx) < 0)
4781 cb->args[2] = ip_idx;
4786 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4788 enum addr_type_t type = UNICAST_ADDR;
4790 return inet6_dump_addr(skb, cb, type);
4793 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4795 enum addr_type_t type = MULTICAST_ADDR;
4797 return inet6_dump_addr(skb, cb, type);
4801 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4803 enum addr_type_t type = ANYCAST_ADDR;
4805 return inet6_dump_addr(skb, cb, type);
4808 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4810 struct net *net = sock_net(in_skb->sk);
4811 struct ifaddrmsg *ifm;
4812 struct nlattr *tb[IFA_MAX+1];
4813 struct in6_addr *addr = NULL, *peer;
4814 struct net_device *dev = NULL;
4815 struct inet6_ifaddr *ifa;
4816 struct sk_buff *skb;
4819 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4823 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4829 ifm = nlmsg_data(nlh);
4831 dev = __dev_get_by_index(net, ifm->ifa_index);
4833 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4835 err = -EADDRNOTAVAIL;
4839 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4845 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4846 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4848 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4849 WARN_ON(err == -EMSGSIZE);
4853 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4860 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4862 struct sk_buff *skb;
4863 struct net *net = dev_net(ifa->idev->dev);
4866 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4870 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4872 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4873 WARN_ON(err == -EMSGSIZE);
4877 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4881 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4884 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4885 __s32 *array, int bytes)
4887 BUG_ON(bytes < (DEVCONF_MAX * 4));
4889 memset(array, 0, bytes);
4890 array[DEVCONF_FORWARDING] = cnf->forwarding;
4891 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4892 array[DEVCONF_MTU6] = cnf->mtu6;
4893 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4894 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4895 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4896 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4897 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4898 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4899 jiffies_to_msecs(cnf->rtr_solicit_interval);
4900 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4901 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4902 array[DEVCONF_RTR_SOLICIT_DELAY] =
4903 jiffies_to_msecs(cnf->rtr_solicit_delay);
4904 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4905 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4906 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4907 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4908 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4909 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4910 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4911 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4912 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4913 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4914 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4915 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4916 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4917 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4918 #ifdef CONFIG_IPV6_ROUTER_PREF
4919 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4920 array[DEVCONF_RTR_PROBE_INTERVAL] =
4921 jiffies_to_msecs(cnf->rtr_probe_interval);
4922 #ifdef CONFIG_IPV6_ROUTE_INFO
4923 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4926 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4927 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4928 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4929 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4930 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4932 #ifdef CONFIG_IPV6_MROUTE
4933 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4935 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4936 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4937 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4938 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4939 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4940 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4941 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4942 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4943 /* we omit DEVCONF_STABLE_SECRET for now */
4944 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4945 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4946 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4947 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4950 static inline size_t inet6_ifla6_size(void)
4952 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4953 + nla_total_size(sizeof(struct ifla_cacheinfo))
4954 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4955 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4956 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4957 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4960 static inline size_t inet6_if_nlmsg_size(void)
4962 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4963 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4964 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4965 + nla_total_size(4) /* IFLA_MTU */
4966 + nla_total_size(4) /* IFLA_LINK */
4967 + nla_total_size(1) /* IFLA_OPERSTATE */
4968 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4971 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4975 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
4978 /* Use put_unaligned() because stats may not be aligned for u64. */
4979 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
4980 for (i = 1; i < ICMP6_MIB_MAX; i++)
4981 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4983 memset(&stats[ICMP6_MIB_MAX], 0, pad);
4986 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4987 int bytes, size_t syncpoff)
4990 u64 buff[IPSTATS_MIB_MAX];
4991 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4995 memset(buff, 0, sizeof(buff));
4996 buff[0] = IPSTATS_MIB_MAX;
4998 for_each_possible_cpu(c) {
4999 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5000 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5003 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5004 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5007 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5011 case IFLA_INET6_STATS:
5012 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5013 offsetof(struct ipstats_mib, syncp));
5015 case IFLA_INET6_ICMP6STATS:
5016 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5021 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5022 u32 ext_filter_mask)
5025 struct ifla_cacheinfo ci;
5027 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5028 goto nla_put_failure;
5029 ci.max_reasm_len = IPV6_MAXPLEN;
5030 ci.tstamp = cstamp_delta(idev->tstamp);
5031 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5032 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5033 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5034 goto nla_put_failure;
5035 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5037 goto nla_put_failure;
5038 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5040 /* XXX - MC not implemented */
5042 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5045 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5047 goto nla_put_failure;
5048 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5050 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5052 goto nla_put_failure;
5053 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5055 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5057 goto nla_put_failure;
5059 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5060 goto nla_put_failure;
5062 read_lock_bh(&idev->lock);
5063 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5064 read_unlock_bh(&idev->lock);
5072 static size_t inet6_get_link_af_size(const struct net_device *dev,
5073 u32 ext_filter_mask)
5075 if (!__in6_dev_get(dev))
5078 return inet6_ifla6_size();
5081 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5082 u32 ext_filter_mask)
5084 struct inet6_dev *idev = __in6_dev_get(dev);
5089 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5095 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5097 struct inet6_ifaddr *ifp;
5098 struct net_device *dev = idev->dev;
5099 bool clear_token, update_rs = false;
5100 struct in6_addr ll_addr;
5106 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5108 if (!ipv6_accept_ra(idev))
5110 if (idev->cnf.rtr_solicits == 0)
5113 write_lock_bh(&idev->lock);
5115 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5116 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5118 write_unlock_bh(&idev->lock);
5120 clear_token = ipv6_addr_any(token);
5124 if (!idev->dead && (idev->if_flags & IF_READY) &&
5125 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5126 IFA_F_OPTIMISTIC)) {
5127 /* If we're not ready, then normal ifup will take care
5128 * of this. Otherwise, we need to request our rs here.
5130 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5135 write_lock_bh(&idev->lock);
5138 idev->if_flags |= IF_RS_SENT;
5139 idev->rs_interval = rfc3315_s14_backoff_init(
5140 idev->cnf.rtr_solicit_interval);
5141 idev->rs_probes = 1;
5142 addrconf_mod_rs_timer(idev, idev->rs_interval);
5145 /* Well, that's kinda nasty ... */
5146 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5147 spin_lock(&ifp->lock);
5148 if (ifp->tokenized) {
5150 ifp->prefered_lft = 0;
5152 spin_unlock(&ifp->lock);
5155 write_unlock_bh(&idev->lock);
5156 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5157 addrconf_verify_rtnl();
5161 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5162 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5163 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5166 static int inet6_validate_link_af(const struct net_device *dev,
5167 const struct nlattr *nla)
5169 struct nlattr *tb[IFLA_INET6_MAX + 1];
5171 if (dev && !__in6_dev_get(dev))
5172 return -EAFNOSUPPORT;
5174 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5177 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5180 struct inet6_dev *idev = __in6_dev_get(dev);
5181 struct nlattr *tb[IFLA_INET6_MAX + 1];
5184 return -EAFNOSUPPORT;
5186 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5189 if (tb[IFLA_INET6_TOKEN]) {
5190 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5195 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5196 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5198 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5199 mode != IN6_ADDR_GEN_MODE_NONE &&
5200 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5201 mode != IN6_ADDR_GEN_MODE_RANDOM)
5204 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5205 !idev->cnf.stable_secret.initialized &&
5206 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5209 idev->addr_gen_mode = mode;
5216 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5217 u32 portid, u32 seq, int event, unsigned int flags)
5219 struct net_device *dev = idev->dev;
5220 struct ifinfomsg *hdr;
5221 struct nlmsghdr *nlh;
5224 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5228 hdr = nlmsg_data(nlh);
5229 hdr->ifi_family = AF_INET6;
5231 hdr->ifi_type = dev->type;
5232 hdr->ifi_index = dev->ifindex;
5233 hdr->ifi_flags = dev_get_flags(dev);
5234 hdr->ifi_change = 0;
5236 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5238 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5239 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5240 (dev->ifindex != dev_get_iflink(dev) &&
5241 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5242 nla_put_u8(skb, IFLA_OPERSTATE,
5243 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5244 goto nla_put_failure;
5245 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5247 goto nla_put_failure;
5249 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5250 goto nla_put_failure;
5252 nla_nest_end(skb, protoinfo);
5253 nlmsg_end(skb, nlh);
5257 nlmsg_cancel(skb, nlh);
5261 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5263 struct net *net = sock_net(skb->sk);
5266 struct net_device *dev;
5267 struct inet6_dev *idev;
5268 struct hlist_head *head;
5271 s_idx = cb->args[1];
5274 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5276 head = &net->dev_index_head[h];
5277 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5280 idev = __in6_dev_get(dev);
5283 if (inet6_fill_ifinfo(skb, idev,
5284 NETLINK_CB(cb->skb).portid,
5286 RTM_NEWLINK, NLM_F_MULTI) < 0)
5300 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5302 struct sk_buff *skb;
5303 struct net *net = dev_net(idev->dev);
5306 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5310 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5312 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5313 WARN_ON(err == -EMSGSIZE);
5317 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5321 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5324 static inline size_t inet6_prefix_nlmsg_size(void)
5326 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5327 + nla_total_size(sizeof(struct in6_addr))
5328 + nla_total_size(sizeof(struct prefix_cacheinfo));
5331 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5332 struct prefix_info *pinfo, u32 portid, u32 seq,
5333 int event, unsigned int flags)
5335 struct prefixmsg *pmsg;
5336 struct nlmsghdr *nlh;
5337 struct prefix_cacheinfo ci;
5339 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5343 pmsg = nlmsg_data(nlh);
5344 pmsg->prefix_family = AF_INET6;
5345 pmsg->prefix_pad1 = 0;
5346 pmsg->prefix_pad2 = 0;
5347 pmsg->prefix_ifindex = idev->dev->ifindex;
5348 pmsg->prefix_len = pinfo->prefix_len;
5349 pmsg->prefix_type = pinfo->type;
5350 pmsg->prefix_pad3 = 0;
5351 pmsg->prefix_flags = 0;
5353 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5354 if (pinfo->autoconf)
5355 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5357 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5358 goto nla_put_failure;
5359 ci.preferred_time = ntohl(pinfo->prefered);
5360 ci.valid_time = ntohl(pinfo->valid);
5361 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5362 goto nla_put_failure;
5363 nlmsg_end(skb, nlh);
5367 nlmsg_cancel(skb, nlh);
5371 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5372 struct prefix_info *pinfo)
5374 struct sk_buff *skb;
5375 struct net *net = dev_net(idev->dev);
5378 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5382 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5384 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5385 WARN_ON(err == -EMSGSIZE);
5389 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5393 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5396 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5398 struct net *net = dev_net(ifp->idev->dev);
5403 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5408 * If the address was optimistic
5409 * we inserted the route at the start of
5410 * our DAD process, so we don't need
5413 if (!(ifp->rt->rt6i_node))
5414 ip6_ins_rt(ifp->rt);
5415 if (ifp->idev->cnf.forwarding)
5416 addrconf_join_anycast(ifp);
5417 if (!ipv6_addr_any(&ifp->peer_addr))
5418 addrconf_prefix_route(&ifp->peer_addr, 128,
5419 ifp->idev->dev, 0, 0);
5422 if (ifp->idev->cnf.forwarding)
5423 addrconf_leave_anycast(ifp);
5424 addrconf_leave_solict(ifp->idev, &ifp->addr);
5425 if (!ipv6_addr_any(&ifp->peer_addr)) {
5426 struct rt6_info *rt;
5428 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5429 ifp->idev->dev, 0, 0);
5434 dst_hold(&ifp->rt->dst);
5435 ip6_del_rt(ifp->rt);
5437 rt_genid_bump_ipv6(net);
5440 atomic_inc(&net->ipv6.dev_addr_genid);
5443 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5446 if (likely(ifp->idev->dead == 0))
5447 __ipv6_ifa_notify(event, ifp);
5448 rcu_read_unlock_bh();
5451 #ifdef CONFIG_SYSCTL
5454 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5455 void __user *buffer, size_t *lenp, loff_t *ppos)
5457 int *valp = ctl->data;
5460 struct ctl_table lctl;
5464 * ctl->data points to idev->cnf.forwarding, we should
5465 * not modify it until we get the rtnl lock.
5470 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5473 ret = addrconf_fixup_forwarding(ctl, valp, val);
5480 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5481 void __user *buffer, size_t *lenp, loff_t *ppos)
5483 struct inet6_dev *idev = ctl->extra1;
5484 int min_mtu = IPV6_MIN_MTU;
5485 struct ctl_table lctl;
5488 lctl.extra1 = &min_mtu;
5489 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5491 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5494 static void dev_disable_change(struct inet6_dev *idev)
5496 struct netdev_notifier_info info;
5498 if (!idev || !idev->dev)
5501 netdev_notifier_info_init(&info, idev->dev);
5502 if (idev->cnf.disable_ipv6)
5503 addrconf_notify(NULL, NETDEV_DOWN, &info);
5505 addrconf_notify(NULL, NETDEV_UP, &info);
5508 static void addrconf_disable_change(struct net *net, __s32 newf)
5510 struct net_device *dev;
5511 struct inet6_dev *idev;
5514 for_each_netdev_rcu(net, dev) {
5515 idev = __in6_dev_get(dev);
5517 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5518 idev->cnf.disable_ipv6 = newf;
5520 dev_disable_change(idev);
5526 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5531 if (!rtnl_trylock())
5532 return restart_syscall();
5534 net = (struct net *)table->extra2;
5538 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5543 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5544 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5545 addrconf_disable_change(net, newf);
5546 } else if ((!newf) ^ (!old))
5547 dev_disable_change((struct inet6_dev *)table->extra1);
5554 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5555 void __user *buffer, size_t *lenp, loff_t *ppos)
5557 int *valp = ctl->data;
5560 struct ctl_table lctl;
5564 * ctl->data points to idev->cnf.disable_ipv6, we should
5565 * not modify it until we get the rtnl lock.
5570 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5573 ret = addrconf_disable_ipv6(ctl, valp, val);
5580 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5581 void __user *buffer, size_t *lenp, loff_t *ppos)
5583 int *valp = ctl->data;
5588 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5591 if (write && old != new) {
5592 struct net *net = ctl->extra2;
5594 if (!rtnl_trylock())
5595 return restart_syscall();
5597 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5598 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5599 NETCONFA_IFINDEX_DEFAULT,
5600 net->ipv6.devconf_dflt);
5601 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5602 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5603 NETCONFA_IFINDEX_ALL,
5604 net->ipv6.devconf_all);
5606 struct inet6_dev *idev = ctl->extra1;
5608 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5618 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5619 void __user *buffer, size_t *lenp,
5623 struct in6_addr addr;
5624 char str[IPV6_MAX_STRLEN];
5625 struct ctl_table lctl = *ctl;
5626 struct net *net = ctl->extra2;
5627 struct ipv6_stable_secret *secret = ctl->data;
5629 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5632 lctl.maxlen = IPV6_MAX_STRLEN;
5635 if (!rtnl_trylock())
5636 return restart_syscall();
5638 if (!write && !secret->initialized) {
5643 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5644 if (err >= sizeof(str)) {
5649 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5653 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5658 secret->initialized = true;
5659 secret->secret = addr;
5661 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5662 struct net_device *dev;
5664 for_each_netdev(net, dev) {
5665 struct inet6_dev *idev = __in6_dev_get(dev);
5668 idev->addr_gen_mode =
5669 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5673 struct inet6_dev *idev = ctl->extra1;
5675 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5685 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5687 void __user *buffer,
5691 int *valp = ctl->data;
5694 struct ctl_table lctl;
5697 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5698 * we should not modify it until we get the rtnl lock.
5703 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5706 ret = addrconf_fixup_linkdown(ctl, valp, val);
5712 static int minus_one = -1;
5713 static const int one = 1;
5714 static const int two_five_five = 255;
5716 static const struct ctl_table addrconf_sysctl[] = {
5718 .procname = "forwarding",
5719 .data = &ipv6_devconf.forwarding,
5720 .maxlen = sizeof(int),
5722 .proc_handler = addrconf_sysctl_forward,
5725 .procname = "hop_limit",
5726 .data = &ipv6_devconf.hop_limit,
5727 .maxlen = sizeof(int),
5729 .proc_handler = proc_dointvec_minmax,
5730 .extra1 = (void *)&one,
5731 .extra2 = (void *)&two_five_five,
5735 .data = &ipv6_devconf.mtu6,
5736 .maxlen = sizeof(int),
5738 .proc_handler = addrconf_sysctl_mtu,
5741 .procname = "accept_ra",
5742 .data = &ipv6_devconf.accept_ra,
5743 .maxlen = sizeof(int),
5745 .proc_handler = proc_dointvec,
5748 .procname = "accept_redirects",
5749 .data = &ipv6_devconf.accept_redirects,
5750 .maxlen = sizeof(int),
5752 .proc_handler = proc_dointvec,
5755 .procname = "autoconf",
5756 .data = &ipv6_devconf.autoconf,
5757 .maxlen = sizeof(int),
5759 .proc_handler = proc_dointvec,
5762 .procname = "dad_transmits",
5763 .data = &ipv6_devconf.dad_transmits,
5764 .maxlen = sizeof(int),
5766 .proc_handler = proc_dointvec,
5769 .procname = "router_solicitations",
5770 .data = &ipv6_devconf.rtr_solicits,
5771 .maxlen = sizeof(int),
5773 .proc_handler = proc_dointvec_minmax,
5774 .extra1 = &minus_one,
5777 .procname = "router_solicitation_interval",
5778 .data = &ipv6_devconf.rtr_solicit_interval,
5779 .maxlen = sizeof(int),
5781 .proc_handler = proc_dointvec_jiffies,
5784 .procname = "router_solicitation_max_interval",
5785 .data = &ipv6_devconf.rtr_solicit_max_interval,
5786 .maxlen = sizeof(int),
5788 .proc_handler = proc_dointvec_jiffies,
5791 .procname = "router_solicitation_delay",
5792 .data = &ipv6_devconf.rtr_solicit_delay,
5793 .maxlen = sizeof(int),
5795 .proc_handler = proc_dointvec_jiffies,
5798 .procname = "force_mld_version",
5799 .data = &ipv6_devconf.force_mld_version,
5800 .maxlen = sizeof(int),
5802 .proc_handler = proc_dointvec,
5805 .procname = "mldv1_unsolicited_report_interval",
5807 &ipv6_devconf.mldv1_unsolicited_report_interval,
5808 .maxlen = sizeof(int),
5810 .proc_handler = proc_dointvec_ms_jiffies,
5813 .procname = "mldv2_unsolicited_report_interval",
5815 &ipv6_devconf.mldv2_unsolicited_report_interval,
5816 .maxlen = sizeof(int),
5818 .proc_handler = proc_dointvec_ms_jiffies,
5821 .procname = "use_tempaddr",
5822 .data = &ipv6_devconf.use_tempaddr,
5823 .maxlen = sizeof(int),
5825 .proc_handler = proc_dointvec,
5828 .procname = "temp_valid_lft",
5829 .data = &ipv6_devconf.temp_valid_lft,
5830 .maxlen = sizeof(int),
5832 .proc_handler = proc_dointvec,
5835 .procname = "temp_prefered_lft",
5836 .data = &ipv6_devconf.temp_prefered_lft,
5837 .maxlen = sizeof(int),
5839 .proc_handler = proc_dointvec,
5842 .procname = "regen_max_retry",
5843 .data = &ipv6_devconf.regen_max_retry,
5844 .maxlen = sizeof(int),
5846 .proc_handler = proc_dointvec,
5849 .procname = "max_desync_factor",
5850 .data = &ipv6_devconf.max_desync_factor,
5851 .maxlen = sizeof(int),
5853 .proc_handler = proc_dointvec,
5856 .procname = "max_addresses",
5857 .data = &ipv6_devconf.max_addresses,
5858 .maxlen = sizeof(int),
5860 .proc_handler = proc_dointvec,
5863 .procname = "accept_ra_defrtr",
5864 .data = &ipv6_devconf.accept_ra_defrtr,
5865 .maxlen = sizeof(int),
5867 .proc_handler = proc_dointvec,
5870 .procname = "accept_ra_min_hop_limit",
5871 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5872 .maxlen = sizeof(int),
5874 .proc_handler = proc_dointvec,
5877 .procname = "accept_ra_pinfo",
5878 .data = &ipv6_devconf.accept_ra_pinfo,
5879 .maxlen = sizeof(int),
5881 .proc_handler = proc_dointvec,
5883 #ifdef CONFIG_IPV6_ROUTER_PREF
5885 .procname = "accept_ra_rtr_pref",
5886 .data = &ipv6_devconf.accept_ra_rtr_pref,
5887 .maxlen = sizeof(int),
5889 .proc_handler = proc_dointvec,
5892 .procname = "router_probe_interval",
5893 .data = &ipv6_devconf.rtr_probe_interval,
5894 .maxlen = sizeof(int),
5896 .proc_handler = proc_dointvec_jiffies,
5898 #ifdef CONFIG_IPV6_ROUTE_INFO
5900 .procname = "accept_ra_rt_info_max_plen",
5901 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5902 .maxlen = sizeof(int),
5904 .proc_handler = proc_dointvec,
5909 .procname = "proxy_ndp",
5910 .data = &ipv6_devconf.proxy_ndp,
5911 .maxlen = sizeof(int),
5913 .proc_handler = addrconf_sysctl_proxy_ndp,
5916 .procname = "accept_source_route",
5917 .data = &ipv6_devconf.accept_source_route,
5918 .maxlen = sizeof(int),
5920 .proc_handler = proc_dointvec,
5922 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5924 .procname = "optimistic_dad",
5925 .data = &ipv6_devconf.optimistic_dad,
5926 .maxlen = sizeof(int),
5928 .proc_handler = proc_dointvec,
5931 .procname = "use_optimistic",
5932 .data = &ipv6_devconf.use_optimistic,
5933 .maxlen = sizeof(int),
5935 .proc_handler = proc_dointvec,
5938 #ifdef CONFIG_IPV6_MROUTE
5940 .procname = "mc_forwarding",
5941 .data = &ipv6_devconf.mc_forwarding,
5942 .maxlen = sizeof(int),
5944 .proc_handler = proc_dointvec,
5948 .procname = "disable_ipv6",
5949 .data = &ipv6_devconf.disable_ipv6,
5950 .maxlen = sizeof(int),
5952 .proc_handler = addrconf_sysctl_disable,
5955 .procname = "accept_dad",
5956 .data = &ipv6_devconf.accept_dad,
5957 .maxlen = sizeof(int),
5959 .proc_handler = proc_dointvec,
5962 .procname = "force_tllao",
5963 .data = &ipv6_devconf.force_tllao,
5964 .maxlen = sizeof(int),
5966 .proc_handler = proc_dointvec
5969 .procname = "ndisc_notify",
5970 .data = &ipv6_devconf.ndisc_notify,
5971 .maxlen = sizeof(int),
5973 .proc_handler = proc_dointvec
5976 .procname = "suppress_frag_ndisc",
5977 .data = &ipv6_devconf.suppress_frag_ndisc,
5978 .maxlen = sizeof(int),
5980 .proc_handler = proc_dointvec
5983 .procname = "accept_ra_from_local",
5984 .data = &ipv6_devconf.accept_ra_from_local,
5985 .maxlen = sizeof(int),
5987 .proc_handler = proc_dointvec,
5990 .procname = "accept_ra_mtu",
5991 .data = &ipv6_devconf.accept_ra_mtu,
5992 .maxlen = sizeof(int),
5994 .proc_handler = proc_dointvec,
5997 .procname = "stable_secret",
5998 .data = &ipv6_devconf.stable_secret,
5999 .maxlen = IPV6_MAX_STRLEN,
6001 .proc_handler = addrconf_sysctl_stable_secret,
6004 .procname = "use_oif_addrs_only",
6005 .data = &ipv6_devconf.use_oif_addrs_only,
6006 .maxlen = sizeof(int),
6008 .proc_handler = proc_dointvec,
6011 .procname = "ignore_routes_with_linkdown",
6012 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6013 .maxlen = sizeof(int),
6015 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6018 .procname = "drop_unicast_in_l2_multicast",
6019 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6020 .maxlen = sizeof(int),
6022 .proc_handler = proc_dointvec,
6025 .procname = "drop_unsolicited_na",
6026 .data = &ipv6_devconf.drop_unsolicited_na,
6027 .maxlen = sizeof(int),
6029 .proc_handler = proc_dointvec,
6032 .procname = "keep_addr_on_down",
6033 .data = &ipv6_devconf.keep_addr_on_down,
6034 .maxlen = sizeof(int),
6036 .proc_handler = proc_dointvec,
6044 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6045 struct inet6_dev *idev, struct ipv6_devconf *p)
6048 struct ctl_table *table;
6049 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6051 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6055 for (i = 0; table[i].data; i++) {
6056 table[i].data += (char *)p - (char *)&ipv6_devconf;
6057 /* If one of these is already set, then it is not safe to
6058 * overwrite either of them: this makes proc_dointvec_minmax
6061 if (!table[i].extra1 && !table[i].extra2) {
6062 table[i].extra1 = idev; /* embedded; no ref */
6063 table[i].extra2 = net;
6067 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6069 p->sysctl_header = register_net_sysctl(net, path, table);
6070 if (!p->sysctl_header)
6073 if (!strcmp(dev_name, "all"))
6074 ifindex = NETCONFA_IFINDEX_ALL;
6075 else if (!strcmp(dev_name, "default"))
6076 ifindex = NETCONFA_IFINDEX_DEFAULT;
6078 ifindex = idev->dev->ifindex;
6079 inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6088 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6090 struct ctl_table *table;
6092 if (!p->sysctl_header)
6095 table = p->sysctl_header->ctl_table_arg;
6096 unregister_net_sysctl_table(p->sysctl_header);
6097 p->sysctl_header = NULL;
6101 static int addrconf_sysctl_register(struct inet6_dev *idev)
6105 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6108 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6109 &ndisc_ifinfo_sysctl_change);
6112 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6115 neigh_sysctl_unregister(idev->nd_parms);
6120 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6122 __addrconf_sysctl_unregister(&idev->cnf);
6123 neigh_sysctl_unregister(idev->nd_parms);
6129 static int __net_init addrconf_init_net(struct net *net)
6132 struct ipv6_devconf *all, *dflt;
6134 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6138 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6140 goto err_alloc_dflt;
6142 /* these will be inherited by all namespaces */
6143 dflt->autoconf = ipv6_defaults.autoconf;
6144 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6146 dflt->stable_secret.initialized = false;
6147 all->stable_secret.initialized = false;
6149 net->ipv6.devconf_all = all;
6150 net->ipv6.devconf_dflt = dflt;
6152 #ifdef CONFIG_SYSCTL
6153 err = __addrconf_sysctl_register(net, "all", NULL, all);
6157 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6163 #ifdef CONFIG_SYSCTL
6165 __addrconf_sysctl_unregister(all);
6175 static void __net_exit addrconf_exit_net(struct net *net)
6177 #ifdef CONFIG_SYSCTL
6178 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6179 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6181 kfree(net->ipv6.devconf_dflt);
6182 kfree(net->ipv6.devconf_all);
6185 static struct pernet_operations addrconf_ops = {
6186 .init = addrconf_init_net,
6187 .exit = addrconf_exit_net,
6190 static struct rtnl_af_ops inet6_ops __read_mostly = {
6192 .fill_link_af = inet6_fill_link_af,
6193 .get_link_af_size = inet6_get_link_af_size,
6194 .validate_link_af = inet6_validate_link_af,
6195 .set_link_af = inet6_set_link_af,
6199 * Init / cleanup code
6202 int __init addrconf_init(void)
6204 struct inet6_dev *idev;
6207 err = ipv6_addr_label_init();
6209 pr_crit("%s: cannot initialize default policy table: %d\n",
6214 err = register_pernet_subsys(&addrconf_ops);
6218 addrconf_wq = create_workqueue("ipv6_addrconf");
6224 /* The addrconf netdev notifier requires that loopback_dev
6225 * has it's ipv6 private information allocated and setup
6226 * before it can bring up and give link-local addresses
6227 * to other devices which are up.
6229 * Unfortunately, loopback_dev is not necessarily the first
6230 * entry in the global dev_base list of net devices. In fact,
6231 * it is likely to be the very last entry on that list.
6232 * So this causes the notifier registry below to try and
6233 * give link-local addresses to all devices besides loopback_dev
6234 * first, then loopback_dev, which cases all the non-loopback_dev
6235 * devices to fail to get a link-local address.
6237 * So, as a temporary fix, allocate the ipv6 structure for
6238 * loopback_dev first by hand.
6239 * Longer term, all of the dependencies ipv6 has upon the loopback
6240 * device and it being up should be removed.
6243 idev = ipv6_add_dev(init_net.loopback_dev);
6246 err = PTR_ERR(idev);
6250 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6251 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6253 register_netdevice_notifier(&ipv6_dev_notf);
6257 rtnl_af_register(&inet6_ops);
6259 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6264 /* Only the first call to __rtnl_register can fail */
6265 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6266 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6267 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6268 inet6_dump_ifaddr, NULL);
6269 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6270 inet6_dump_ifmcaddr, NULL);
6271 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6272 inet6_dump_ifacaddr, NULL);
6273 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6274 inet6_netconf_dump_devconf, NULL);
6276 ipv6_addr_label_rtnl_register();
6280 rtnl_af_unregister(&inet6_ops);
6281 unregister_netdevice_notifier(&ipv6_dev_notf);
6283 destroy_workqueue(addrconf_wq);
6285 unregister_pernet_subsys(&addrconf_ops);
6287 ipv6_addr_label_cleanup();
6292 void addrconf_cleanup(void)
6294 struct net_device *dev;
6297 unregister_netdevice_notifier(&ipv6_dev_notf);
6298 unregister_pernet_subsys(&addrconf_ops);
6299 ipv6_addr_label_cleanup();
6303 __rtnl_af_unregister(&inet6_ops);
6305 /* clean dev list */
6306 for_each_netdev(&init_net, dev) {
6307 if (__in6_dev_get(dev) == NULL)
6309 addrconf_ifdown(dev, 1);
6311 addrconf_ifdown(init_net.loopback_dev, 2);
6316 spin_lock_bh(&addrconf_hash_lock);
6317 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6318 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6319 spin_unlock_bh(&addrconf_hash_lock);
6320 cancel_delayed_work(&addr_chk_work);
6323 destroy_workqueue(addrconf_wq);