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;
1188 s32 cnf_temp_preferred_lft;
1190 write_lock_bh(&idev->lock);
1192 spin_lock_bh(&ift->lock);
1193 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1194 spin_unlock_bh(&ift->lock);
1201 if (idev->cnf.use_tempaddr <= 0) {
1202 write_unlock_bh(&idev->lock);
1203 pr_info("%s: use_tempaddr is disabled\n", __func__);
1208 spin_lock_bh(&ifp->lock);
1209 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1210 idev->cnf.use_tempaddr = -1; /*XXX*/
1211 spin_unlock_bh(&ifp->lock);
1212 write_unlock_bh(&idev->lock);
1213 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1220 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1221 ipv6_try_regen_rndid(idev, tmpaddr);
1222 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1223 age = (now - ifp->tstamp) / HZ;
1225 regen_advance = idev->cnf.regen_max_retry *
1226 idev->cnf.dad_transmits *
1227 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1229 /* recalculate max_desync_factor each time and update
1230 * idev->desync_factor if it's larger
1232 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1233 max_desync_factor = min_t(__u32,
1234 idev->cnf.max_desync_factor,
1235 cnf_temp_preferred_lft - regen_advance);
1237 if (unlikely(idev->desync_factor > max_desync_factor)) {
1238 if (max_desync_factor > 0) {
1239 get_random_bytes(&idev->desync_factor,
1240 sizeof(idev->desync_factor));
1241 idev->desync_factor %= max_desync_factor;
1243 idev->desync_factor = 0;
1247 tmp_valid_lft = min_t(__u32,
1249 idev->cnf.temp_valid_lft + age);
1250 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1251 idev->desync_factor;
1252 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1253 tmp_plen = ifp->prefix_len;
1254 tmp_tstamp = ifp->tstamp;
1255 spin_unlock_bh(&ifp->lock);
1257 write_unlock_bh(&idev->lock);
1259 /* A temporary address is created only if this calculated Preferred
1260 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1261 * an implementation must not create a temporary address with a zero
1262 * Preferred Lifetime.
1263 * Use age calculation as in addrconf_verify to avoid unnecessary
1264 * temporary addresses being generated.
1266 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1267 if (tmp_prefered_lft <= regen_advance + age) {
1274 addr_flags = IFA_F_TEMPORARY;
1275 /* set in addrconf_prefix_rcv() */
1276 if (ifp->flags & IFA_F_OPTIMISTIC)
1277 addr_flags |= IFA_F_OPTIMISTIC;
1279 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1280 ipv6_addr_scope(&addr), addr_flags,
1281 tmp_valid_lft, tmp_prefered_lft);
1285 pr_info("%s: retry temporary address regeneration\n", __func__);
1287 write_lock_bh(&idev->lock);
1291 spin_lock_bh(&ift->lock);
1294 ift->tstamp = tmp_tstamp;
1295 spin_unlock_bh(&ift->lock);
1297 addrconf_dad_start(ift);
1305 * Choose an appropriate source address (RFC3484)
1308 IPV6_SADDR_RULE_INIT = 0,
1309 IPV6_SADDR_RULE_LOCAL,
1310 IPV6_SADDR_RULE_SCOPE,
1311 IPV6_SADDR_RULE_PREFERRED,
1312 #ifdef CONFIG_IPV6_MIP6
1313 IPV6_SADDR_RULE_HOA,
1315 IPV6_SADDR_RULE_OIF,
1316 IPV6_SADDR_RULE_LABEL,
1317 IPV6_SADDR_RULE_PRIVACY,
1318 IPV6_SADDR_RULE_ORCHID,
1319 IPV6_SADDR_RULE_PREFIX,
1320 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1321 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1326 struct ipv6_saddr_score {
1329 struct inet6_ifaddr *ifa;
1330 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1335 struct ipv6_saddr_dst {
1336 const struct in6_addr *addr;
1343 static inline int ipv6_saddr_preferred(int type)
1345 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1350 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1352 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1353 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1359 static int ipv6_get_saddr_eval(struct net *net,
1360 struct ipv6_saddr_score *score,
1361 struct ipv6_saddr_dst *dst,
1366 if (i <= score->rule) {
1368 case IPV6_SADDR_RULE_SCOPE:
1369 ret = score->scopedist;
1371 case IPV6_SADDR_RULE_PREFIX:
1372 ret = score->matchlen;
1375 ret = !!test_bit(i, score->scorebits);
1381 case IPV6_SADDR_RULE_INIT:
1382 /* Rule 0: remember if hiscore is not ready yet */
1385 case IPV6_SADDR_RULE_LOCAL:
1386 /* Rule 1: Prefer same address */
1387 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1389 case IPV6_SADDR_RULE_SCOPE:
1390 /* Rule 2: Prefer appropriate scope
1395 * ---+--+-+---> scope
1397 * | d is scope of the destination.
1399 * | \ <- smaller scope is better if
1400 * B-15 | \ if scope is enough for destination.
1401 * | ret = B - scope (-1 <= scope >= d <= 15).
1403 * |/ <- greater is better
1404 * -C / if scope is not enough for destination.
1405 * /| ret = scope - C (-1 <= d < scope <= 15).
1407 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1408 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1409 * Assume B = 0 and we get C > 29.
1411 ret = __ipv6_addr_src_scope(score->addr_type);
1412 if (ret >= dst->scope)
1415 ret -= 128; /* 30 is enough */
1416 score->scopedist = ret;
1418 case IPV6_SADDR_RULE_PREFERRED:
1420 /* Rule 3: Avoid deprecated and optimistic addresses */
1421 u8 avoid = IFA_F_DEPRECATED;
1423 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1424 avoid |= IFA_F_OPTIMISTIC;
1425 ret = ipv6_saddr_preferred(score->addr_type) ||
1426 !(score->ifa->flags & avoid);
1429 #ifdef CONFIG_IPV6_MIP6
1430 case IPV6_SADDR_RULE_HOA:
1432 /* Rule 4: Prefer home address */
1433 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1434 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1438 case IPV6_SADDR_RULE_OIF:
1439 /* Rule 5: Prefer outgoing interface */
1440 ret = (!dst->ifindex ||
1441 dst->ifindex == score->ifa->idev->dev->ifindex);
1443 case IPV6_SADDR_RULE_LABEL:
1444 /* Rule 6: Prefer matching label */
1445 ret = ipv6_addr_label(net,
1446 &score->ifa->addr, score->addr_type,
1447 score->ifa->idev->dev->ifindex) == dst->label;
1449 case IPV6_SADDR_RULE_PRIVACY:
1451 /* Rule 7: Prefer public address
1452 * Note: prefer temporary address if use_tempaddr >= 2
1454 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1455 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1456 score->ifa->idev->cnf.use_tempaddr >= 2;
1457 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1460 case IPV6_SADDR_RULE_ORCHID:
1461 /* Rule 8-: Prefer ORCHID vs ORCHID or
1462 * non-ORCHID vs non-ORCHID
1464 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1465 ipv6_addr_orchid(dst->addr));
1467 case IPV6_SADDR_RULE_PREFIX:
1468 /* Rule 8: Use longest matching prefix */
1469 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1470 if (ret > score->ifa->prefix_len)
1471 ret = score->ifa->prefix_len;
1472 score->matchlen = ret;
1474 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1475 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1476 /* Optimistic addresses still have lower precedence than other
1477 * preferred addresses.
1479 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1487 __set_bit(i, score->scorebits);
1493 static int __ipv6_dev_get_saddr(struct net *net,
1494 struct ipv6_saddr_dst *dst,
1495 struct inet6_dev *idev,
1496 struct ipv6_saddr_score *scores,
1499 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1501 read_lock_bh(&idev->lock);
1502 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1506 * - Tentative Address (RFC2462 section 5.4)
1507 * - A tentative address is not considered
1508 * "assigned to an interface" in the traditional
1509 * sense, unless it is also flagged as optimistic.
1510 * - Candidate Source Address (section 4)
1511 * - In any case, anycast addresses, multicast
1512 * addresses, and the unspecified address MUST
1513 * NOT be included in a candidate set.
1515 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1516 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1519 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1521 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1522 score->addr_type & IPV6_ADDR_MULTICAST)) {
1523 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1529 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1531 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1532 int minihiscore, miniscore;
1534 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1535 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1537 if (minihiscore > miniscore) {
1538 if (i == IPV6_SADDR_RULE_SCOPE &&
1539 score->scopedist > 0) {
1542 * each remaining entry
1543 * has too small (not enough)
1544 * scope, because ifa entries
1545 * are sorted by their scope
1551 } else if (minihiscore < miniscore) {
1553 in6_ifa_put(hiscore->ifa);
1555 in6_ifa_hold(score->ifa);
1557 swap(hiscore, score);
1558 hiscore_idx = 1 - hiscore_idx;
1560 /* restore our iterator */
1561 score->ifa = hiscore->ifa;
1568 read_unlock_bh(&idev->lock);
1572 static int ipv6_get_saddr_master(struct net *net,
1573 const struct net_device *dst_dev,
1574 const struct net_device *master,
1575 struct ipv6_saddr_dst *dst,
1576 struct ipv6_saddr_score *scores,
1579 struct inet6_dev *idev;
1581 idev = __in6_dev_get(dst_dev);
1583 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1584 scores, hiscore_idx);
1586 idev = __in6_dev_get(master);
1588 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1589 scores, hiscore_idx);
1594 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1595 const struct in6_addr *daddr, unsigned int prefs,
1596 struct in6_addr *saddr)
1598 struct ipv6_saddr_score scores[2], *hiscore;
1599 struct ipv6_saddr_dst dst;
1600 struct inet6_dev *idev;
1601 struct net_device *dev;
1603 bool use_oif_addr = false;
1604 int hiscore_idx = 0;
1606 dst_type = __ipv6_addr_type(daddr);
1608 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1609 dst.scope = __ipv6_addr_src_scope(dst_type);
1610 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1613 scores[hiscore_idx].rule = -1;
1614 scores[hiscore_idx].ifa = NULL;
1618 /* Candidate Source Address (section 4)
1619 * - multicast and link-local destination address,
1620 * the set of candidate source address MUST only
1621 * include addresses assigned to interfaces
1622 * belonging to the same link as the outgoing
1624 * (- For site-local destination addresses, the
1625 * set of candidate source addresses MUST only
1626 * include addresses assigned to interfaces
1627 * belonging to the same site as the outgoing
1629 * - "It is RECOMMENDED that the candidate source addresses
1630 * be the set of unicast addresses assigned to the
1631 * interface that will be used to send to the destination
1632 * (the 'outgoing' interface)." (RFC 6724)
1635 idev = __in6_dev_get(dst_dev);
1636 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1637 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1638 (idev && idev->cnf.use_oif_addrs_only)) {
1639 use_oif_addr = true;
1645 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1647 const struct net_device *master;
1650 /* if dst_dev exists and is enslaved to an L3 device, then
1651 * prefer addresses from dst_dev and then the master over
1652 * any other enslaved devices in the L3 domain.
1654 master = l3mdev_master_dev_rcu(dst_dev);
1656 master_idx = master->ifindex;
1658 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1660 scores, hiscore_idx);
1662 if (scores[hiscore_idx].ifa)
1666 for_each_netdev_rcu(net, dev) {
1667 /* only consider addresses on devices in the
1670 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1672 idev = __in6_dev_get(dev);
1675 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1682 hiscore = &scores[hiscore_idx];
1684 return -EADDRNOTAVAIL;
1686 *saddr = hiscore->ifa->addr;
1687 in6_ifa_put(hiscore->ifa);
1690 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1692 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1695 struct inet6_ifaddr *ifp;
1696 int err = -EADDRNOTAVAIL;
1698 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1699 if (ifp->scope > IFA_LINK)
1701 if (ifp->scope == IFA_LINK &&
1702 !(ifp->flags & banned_flags)) {
1711 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1714 struct inet6_dev *idev;
1715 int err = -EADDRNOTAVAIL;
1718 idev = __in6_dev_get(dev);
1720 read_lock_bh(&idev->lock);
1721 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1722 read_unlock_bh(&idev->lock);
1728 static int ipv6_count_addresses(struct inet6_dev *idev)
1731 struct inet6_ifaddr *ifp;
1733 read_lock_bh(&idev->lock);
1734 list_for_each_entry(ifp, &idev->addr_list, if_list)
1736 read_unlock_bh(&idev->lock);
1740 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1741 const struct net_device *dev, int strict)
1743 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1745 EXPORT_SYMBOL(ipv6_chk_addr);
1747 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1748 const struct net_device *dev, int strict,
1751 struct inet6_ifaddr *ifp;
1752 unsigned int hash = inet6_addr_hash(addr);
1756 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1757 if (!net_eq(dev_net(ifp->idev->dev), net))
1759 /* Decouple optimistic from tentative for evaluation here.
1760 * Ban optimistic addresses explicitly, when required.
1762 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1763 ? (ifp->flags&~IFA_F_TENTATIVE)
1765 if (ipv6_addr_equal(&ifp->addr, addr) &&
1766 !(ifp_flags&banned_flags) &&
1767 (!dev || ifp->idev->dev == dev ||
1768 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1769 rcu_read_unlock_bh();
1774 rcu_read_unlock_bh();
1777 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1779 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1780 struct net_device *dev)
1782 unsigned int hash = inet6_addr_hash(addr);
1783 struct inet6_ifaddr *ifp;
1785 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1786 if (!net_eq(dev_net(ifp->idev->dev), net))
1788 if (ipv6_addr_equal(&ifp->addr, addr)) {
1789 if (!dev || ifp->idev->dev == dev)
1796 /* Compares an address/prefix_len with addresses on device @dev.
1797 * If one is found it returns true.
1799 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1800 const unsigned int prefix_len, struct net_device *dev)
1802 struct inet6_dev *idev;
1803 struct inet6_ifaddr *ifa;
1807 idev = __in6_dev_get(dev);
1809 read_lock_bh(&idev->lock);
1810 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1811 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1815 read_unlock_bh(&idev->lock);
1821 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1823 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1825 struct inet6_dev *idev;
1826 struct inet6_ifaddr *ifa;
1831 idev = __in6_dev_get(dev);
1833 read_lock_bh(&idev->lock);
1834 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1835 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1840 read_unlock_bh(&idev->lock);
1845 EXPORT_SYMBOL(ipv6_chk_prefix);
1847 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1848 struct net_device *dev, int strict)
1850 struct inet6_ifaddr *ifp, *result = NULL;
1851 unsigned int hash = inet6_addr_hash(addr);
1854 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1855 if (!net_eq(dev_net(ifp->idev->dev), net))
1857 if (ipv6_addr_equal(&ifp->addr, addr)) {
1858 if (!dev || ifp->idev->dev == dev ||
1859 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1866 rcu_read_unlock_bh();
1871 /* Gets referenced address, destroys ifaddr */
1873 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1876 ifp->flags |= IFA_F_DADFAILED;
1878 if (ifp->flags&IFA_F_PERMANENT) {
1879 spin_lock_bh(&ifp->lock);
1880 addrconf_del_dad_work(ifp);
1881 ifp->flags |= IFA_F_TENTATIVE;
1882 spin_unlock_bh(&ifp->lock);
1884 ipv6_ifa_notify(0, ifp);
1886 } else if (ifp->flags&IFA_F_TEMPORARY) {
1887 struct inet6_ifaddr *ifpub;
1888 spin_lock_bh(&ifp->lock);
1891 in6_ifa_hold(ifpub);
1892 spin_unlock_bh(&ifp->lock);
1893 ipv6_create_tempaddr(ifpub, ifp);
1896 spin_unlock_bh(&ifp->lock);
1904 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1908 spin_lock_bh(&ifp->lock);
1909 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1910 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1913 spin_unlock_bh(&ifp->lock);
1918 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1920 struct inet6_dev *idev = ifp->idev;
1921 struct net *net = dev_net(ifp->idev->dev);
1923 if (addrconf_dad_end(ifp)) {
1928 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1929 ifp->idev->dev->name, &ifp->addr);
1931 spin_lock_bh(&ifp->lock);
1933 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1934 int scope = ifp->scope;
1935 u32 flags = ifp->flags;
1936 struct in6_addr new_addr;
1937 struct inet6_ifaddr *ifp2;
1938 u32 valid_lft, preferred_lft;
1939 int pfxlen = ifp->prefix_len;
1940 int retries = ifp->stable_privacy_retry + 1;
1942 if (retries > net->ipv6.sysctl.idgen_retries) {
1943 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1944 ifp->idev->dev->name);
1948 new_addr = ifp->addr;
1949 if (ipv6_generate_stable_address(&new_addr, retries,
1953 valid_lft = ifp->valid_lft;
1954 preferred_lft = ifp->prefered_lft;
1956 spin_unlock_bh(&ifp->lock);
1958 if (idev->cnf.max_addresses &&
1959 ipv6_count_addresses(idev) >=
1960 idev->cnf.max_addresses)
1963 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1964 ifp->idev->dev->name);
1966 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1967 scope, flags, valid_lft,
1972 spin_lock_bh(&ifp2->lock);
1973 ifp2->stable_privacy_retry = retries;
1974 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1975 spin_unlock_bh(&ifp2->lock);
1977 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1980 spin_lock_bh(&ifp->lock);
1984 /* transition from _POSTDAD to _ERRDAD */
1985 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1986 spin_unlock_bh(&ifp->lock);
1988 addrconf_mod_dad_work(ifp, 0);
1992 /* Join to solicited addr multicast group.
1993 * caller must hold RTNL */
1994 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1996 struct in6_addr maddr;
1998 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2001 addrconf_addr_solict_mult(addr, &maddr);
2002 ipv6_dev_mc_inc(dev, &maddr);
2005 /* caller must hold RTNL */
2006 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2008 struct in6_addr maddr;
2010 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2013 addrconf_addr_solict_mult(addr, &maddr);
2014 __ipv6_dev_mc_dec(idev, &maddr);
2017 /* caller must hold RTNL */
2018 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2020 struct in6_addr addr;
2022 if (ifp->prefix_len >= 127) /* RFC 6164 */
2024 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2025 if (ipv6_addr_any(&addr))
2027 __ipv6_dev_ac_inc(ifp->idev, &addr);
2030 /* caller must hold RTNL */
2031 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2033 struct in6_addr addr;
2035 if (ifp->prefix_len >= 127) /* RFC 6164 */
2037 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2038 if (ipv6_addr_any(&addr))
2040 __ipv6_dev_ac_dec(ifp->idev, &addr);
2043 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2045 if (dev->addr_len != EUI64_ADDR_LEN)
2047 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2052 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2054 union fwnet_hwaddr *ha;
2056 if (dev->addr_len != FWNET_ALEN)
2059 ha = (union fwnet_hwaddr *)dev->dev_addr;
2061 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2066 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2068 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2069 if (dev->addr_len != ARCNET_ALEN)
2072 eui[7] = *(u8 *)dev->dev_addr;
2076 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2078 if (dev->addr_len != INFINIBAND_ALEN)
2080 memcpy(eui, dev->dev_addr + 12, 8);
2085 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2089 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2090 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2091 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2092 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2093 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2094 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2098 memcpy(eui + 4, &addr, 4);
2102 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2104 if (dev->priv_flags & IFF_ISATAP)
2105 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2109 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2111 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2114 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2116 memcpy(eui, dev->perm_addr, 3);
2117 memcpy(eui + 5, dev->perm_addr + 3, 3);
2124 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2126 switch (dev->type) {
2129 return addrconf_ifid_eui48(eui, dev);
2131 return addrconf_ifid_arcnet(eui, dev);
2132 case ARPHRD_INFINIBAND:
2133 return addrconf_ifid_infiniband(eui, dev);
2135 return addrconf_ifid_sit(eui, dev);
2137 return addrconf_ifid_gre(eui, dev);
2138 case ARPHRD_6LOWPAN:
2139 return addrconf_ifid_eui64(eui, dev);
2140 case ARPHRD_IEEE1394:
2141 return addrconf_ifid_ieee1394(eui, dev);
2142 case ARPHRD_TUNNEL6:
2143 return addrconf_ifid_ip6tnl(eui, dev);
2148 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2151 struct inet6_ifaddr *ifp;
2153 read_lock_bh(&idev->lock);
2154 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2155 if (ifp->scope > IFA_LINK)
2157 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2158 memcpy(eui, ifp->addr.s6_addr+8, 8);
2163 read_unlock_bh(&idev->lock);
2167 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2168 static void ipv6_regen_rndid(struct inet6_dev *idev)
2171 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2172 idev->rndid[0] &= ~0x02;
2175 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2176 * check if generated address is not inappropriate
2178 * - Reserved subnet anycast (RFC 2526)
2179 * 11111101 11....11 1xxxxxxx
2180 * - ISATAP (RFC4214) 6.1
2181 * 00-00-5E-FE-xx-xx-xx-xx
2183 * - XXX: already assigned to an address on the device
2185 if (idev->rndid[0] == 0xfd &&
2186 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2187 (idev->rndid[7]&0x80))
2189 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2190 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2192 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2197 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2199 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2200 ipv6_regen_rndid(idev);
2208 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2209 unsigned long expires, u32 flags)
2211 struct fib6_config cfg = {
2212 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2213 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2214 .fc_ifindex = dev->ifindex,
2215 .fc_expires = expires,
2217 .fc_flags = RTF_UP | flags,
2218 .fc_nlinfo.nl_net = dev_net(dev),
2219 .fc_protocol = RTPROT_KERNEL,
2224 /* Prevent useless cloning on PtP SIT.
2225 This thing is done here expecting that the whole
2226 class of non-broadcast devices need not cloning.
2228 #if IS_ENABLED(CONFIG_IPV6_SIT)
2229 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2230 cfg.fc_flags |= RTF_NONEXTHOP;
2233 ip6_route_add(&cfg);
2237 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2239 const struct net_device *dev,
2240 u32 flags, u32 noflags)
2242 struct fib6_node *fn;
2243 struct rt6_info *rt = NULL;
2244 struct fib6_table *table;
2245 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2247 table = fib6_get_table(dev_net(dev), tb_id);
2251 read_lock_bh(&table->tb6_lock);
2252 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2256 noflags |= RTF_CACHE;
2257 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2258 if (rt->dst.dev->ifindex != dev->ifindex)
2260 if ((rt->rt6i_flags & flags) != flags)
2262 if ((rt->rt6i_flags & noflags) != 0)
2268 read_unlock_bh(&table->tb6_lock);
2273 /* Create "default" multicast route to the interface */
2275 static void addrconf_add_mroute(struct net_device *dev)
2277 struct fib6_config cfg = {
2278 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2279 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2280 .fc_ifindex = dev->ifindex,
2283 .fc_nlinfo.nl_net = dev_net(dev),
2286 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2288 ip6_route_add(&cfg);
2291 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2293 struct inet6_dev *idev;
2297 idev = ipv6_find_idev(dev);
2299 return ERR_PTR(-ENOBUFS);
2301 if (idev->cnf.disable_ipv6)
2302 return ERR_PTR(-EACCES);
2304 /* Add default multicast route */
2305 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2306 addrconf_add_mroute(dev);
2311 static void manage_tempaddrs(struct inet6_dev *idev,
2312 struct inet6_ifaddr *ifp,
2313 __u32 valid_lft, __u32 prefered_lft,
2314 bool create, unsigned long now)
2317 struct inet6_ifaddr *ift;
2319 read_lock_bh(&idev->lock);
2320 /* update all temporary addresses in the list */
2321 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2322 int age, max_valid, max_prefered;
2324 if (ifp != ift->ifpub)
2327 /* RFC 4941 section 3.3:
2328 * If a received option will extend the lifetime of a public
2329 * address, the lifetimes of temporary addresses should
2330 * be extended, subject to the overall constraint that no
2331 * temporary addresses should ever remain "valid" or "preferred"
2332 * for a time longer than (TEMP_VALID_LIFETIME) or
2333 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2335 age = (now - ift->cstamp) / HZ;
2336 max_valid = idev->cnf.temp_valid_lft - age;
2340 max_prefered = idev->cnf.temp_prefered_lft -
2341 idev->desync_factor - age;
2342 if (max_prefered < 0)
2345 if (valid_lft > max_valid)
2346 valid_lft = max_valid;
2348 if (prefered_lft > max_prefered)
2349 prefered_lft = max_prefered;
2351 spin_lock(&ift->lock);
2353 ift->valid_lft = valid_lft;
2354 ift->prefered_lft = prefered_lft;
2356 if (prefered_lft > 0)
2357 ift->flags &= ~IFA_F_DEPRECATED;
2359 spin_unlock(&ift->lock);
2360 if (!(flags&IFA_F_TENTATIVE))
2361 ipv6_ifa_notify(0, ift);
2364 if ((create || list_empty(&idev->tempaddr_list)) &&
2365 idev->cnf.use_tempaddr > 0) {
2366 /* When a new public address is created as described
2367 * in [ADDRCONF], also create a new temporary address.
2368 * Also create a temporary address if it's enabled but
2369 * no temporary address currently exists.
2371 read_unlock_bh(&idev->lock);
2372 ipv6_create_tempaddr(ifp, NULL);
2374 read_unlock_bh(&idev->lock);
2378 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2380 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2381 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2384 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2385 const struct prefix_info *pinfo,
2386 struct inet6_dev *in6_dev,
2387 const struct in6_addr *addr, int addr_type,
2388 u32 addr_flags, bool sllao, bool tokenized,
2389 __u32 valid_lft, u32 prefered_lft)
2391 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2392 int create = 0, update_lft = 0;
2394 if (!ifp && valid_lft) {
2395 int max_addresses = in6_dev->cnf.max_addresses;
2397 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2398 if (in6_dev->cnf.optimistic_dad &&
2399 !net->ipv6.devconf_all->forwarding && sllao)
2400 addr_flags |= IFA_F_OPTIMISTIC;
2403 /* Do not allow to create too much of autoconfigured
2404 * addresses; this would be too easy way to crash kernel.
2406 if (!max_addresses ||
2407 ipv6_count_addresses(in6_dev) < max_addresses)
2408 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2410 addr_type&IPV6_ADDR_SCOPE_MASK,
2411 addr_flags, valid_lft,
2414 if (IS_ERR_OR_NULL(ifp))
2419 spin_lock_bh(&ifp->lock);
2420 ifp->flags |= IFA_F_MANAGETEMPADDR;
2421 ifp->cstamp = jiffies;
2422 ifp->tokenized = tokenized;
2423 spin_unlock_bh(&ifp->lock);
2424 addrconf_dad_start(ifp);
2432 /* update lifetime (RFC2462 5.5.3 e) */
2433 spin_lock_bh(&ifp->lock);
2435 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2436 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2439 if (!update_lft && !create && stored_lft) {
2440 const u32 minimum_lft = min_t(u32,
2441 stored_lft, MIN_VALID_LIFETIME);
2442 valid_lft = max(valid_lft, minimum_lft);
2444 /* RFC4862 Section 5.5.3e:
2445 * "Note that the preferred lifetime of the
2446 * corresponding address is always reset to
2447 * the Preferred Lifetime in the received
2448 * Prefix Information option, regardless of
2449 * whether the valid lifetime is also reset or
2452 * So we should always update prefered_lft here.
2458 ifp->valid_lft = valid_lft;
2459 ifp->prefered_lft = prefered_lft;
2462 ifp->flags &= ~IFA_F_DEPRECATED;
2463 spin_unlock_bh(&ifp->lock);
2465 if (!(flags&IFA_F_TENTATIVE))
2466 ipv6_ifa_notify(0, ifp);
2468 spin_unlock_bh(&ifp->lock);
2470 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2479 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2481 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2483 struct prefix_info *pinfo;
2488 struct inet6_dev *in6_dev;
2489 struct net *net = dev_net(dev);
2491 pinfo = (struct prefix_info *) opt;
2493 if (len < sizeof(struct prefix_info)) {
2494 ADBG("addrconf: prefix option too short\n");
2499 * Validation checks ([ADDRCONF], page 19)
2502 addr_type = ipv6_addr_type(&pinfo->prefix);
2504 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2507 valid_lft = ntohl(pinfo->valid);
2508 prefered_lft = ntohl(pinfo->prefered);
2510 if (prefered_lft > valid_lft) {
2511 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2515 in6_dev = in6_dev_get(dev);
2518 net_dbg_ratelimited("addrconf: device %s not configured\n",
2524 * Two things going on here:
2525 * 1) Add routes for on-link prefixes
2526 * 2) Configure prefixes with the auto flag set
2529 if (pinfo->onlink) {
2530 struct rt6_info *rt;
2531 unsigned long rt_expires;
2533 /* Avoid arithmetic overflow. Really, we could
2534 * save rt_expires in seconds, likely valid_lft,
2535 * but it would require division in fib gc, that it
2539 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2541 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2543 if (addrconf_finite_timeout(rt_expires))
2546 rt = addrconf_get_prefix_route(&pinfo->prefix,
2549 RTF_ADDRCONF | RTF_PREFIX_RT,
2550 RTF_GATEWAY | RTF_DEFAULT);
2553 /* Autoconf prefix route */
2554 if (valid_lft == 0) {
2557 } else if (addrconf_finite_timeout(rt_expires)) {
2559 rt6_set_expires(rt, jiffies + rt_expires);
2561 rt6_clean_expires(rt);
2563 } else if (valid_lft) {
2564 clock_t expires = 0;
2565 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2566 if (addrconf_finite_timeout(rt_expires)) {
2568 flags |= RTF_EXPIRES;
2569 expires = jiffies_to_clock_t(rt_expires);
2571 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2572 dev, expires, flags);
2577 /* Try to figure out our local address for this prefix */
2579 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2580 struct in6_addr addr;
2581 bool tokenized = false, dev_addr_generated = false;
2583 if (pinfo->prefix_len == 64) {
2584 memcpy(&addr, &pinfo->prefix, 8);
2586 if (!ipv6_addr_any(&in6_dev->token)) {
2587 read_lock_bh(&in6_dev->lock);
2588 memcpy(addr.s6_addr + 8,
2589 in6_dev->token.s6_addr + 8, 8);
2590 read_unlock_bh(&in6_dev->lock);
2592 } else if (is_addr_mode_generate_stable(in6_dev) &&
2593 !ipv6_generate_stable_address(&addr, 0,
2595 addr_flags |= IFA_F_STABLE_PRIVACY;
2597 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2598 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2601 dev_addr_generated = true;
2605 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2610 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2613 tokenized, valid_lft,
2618 /* Ignore error case here because previous prefix add addr was
2619 * successful which will be notified.
2621 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2622 addr_type, addr_flags, sllao,
2623 tokenized, valid_lft,
2625 dev_addr_generated);
2627 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2629 in6_dev_put(in6_dev);
2633 * Set destination address.
2634 * Special case for SIT interfaces where we create a new "virtual"
2637 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2639 struct in6_ifreq ireq;
2640 struct net_device *dev;
2646 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2649 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2655 #if IS_ENABLED(CONFIG_IPV6_SIT)
2656 if (dev->type == ARPHRD_SIT) {
2657 const struct net_device_ops *ops = dev->netdev_ops;
2659 struct ip_tunnel_parm p;
2661 err = -EADDRNOTAVAIL;
2662 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2665 memset(&p, 0, sizeof(p));
2666 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2670 p.iph.protocol = IPPROTO_IPV6;
2672 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2674 if (ops->ndo_do_ioctl) {
2675 mm_segment_t oldfs = get_fs();
2678 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2685 dev = __dev_get_by_name(net, p.name);
2688 err = dev_open(dev);
2698 static int ipv6_mc_config(struct sock *sk, bool join,
2699 const struct in6_addr *addr, int ifindex)
2707 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2709 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2716 * Manual configuration of address on an interface
2718 static int inet6_addr_add(struct net *net, int ifindex,
2719 const struct in6_addr *pfx,
2720 const struct in6_addr *peer_pfx,
2721 unsigned int plen, __u32 ifa_flags,
2722 __u32 prefered_lft, __u32 valid_lft)
2724 struct inet6_ifaddr *ifp;
2725 struct inet6_dev *idev;
2726 struct net_device *dev;
2727 unsigned long timeout;
2737 /* check the lifetime */
2738 if (!valid_lft || prefered_lft > valid_lft)
2741 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2744 dev = __dev_get_by_index(net, ifindex);
2748 idev = addrconf_add_dev(dev);
2750 return PTR_ERR(idev);
2752 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2753 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2754 true, pfx, ifindex);
2760 scope = ipv6_addr_scope(pfx);
2762 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2763 if (addrconf_finite_timeout(timeout)) {
2764 expires = jiffies_to_clock_t(timeout * HZ);
2765 valid_lft = timeout;
2766 flags = RTF_EXPIRES;
2770 ifa_flags |= IFA_F_PERMANENT;
2773 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2774 if (addrconf_finite_timeout(timeout)) {
2776 ifa_flags |= IFA_F_DEPRECATED;
2777 prefered_lft = timeout;
2780 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2781 valid_lft, prefered_lft);
2784 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2785 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2790 * Note that section 3.1 of RFC 4429 indicates
2791 * that the Optimistic flag should not be set for
2792 * manually configured addresses
2794 addrconf_dad_start(ifp);
2795 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2796 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2799 addrconf_verify_rtnl();
2801 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2802 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2803 false, pfx, ifindex);
2806 return PTR_ERR(ifp);
2809 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2810 const struct in6_addr *pfx, unsigned int plen)
2812 struct inet6_ifaddr *ifp;
2813 struct inet6_dev *idev;
2814 struct net_device *dev;
2819 dev = __dev_get_by_index(net, ifindex);
2823 idev = __in6_dev_get(dev);
2827 read_lock_bh(&idev->lock);
2828 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2829 if (ifp->prefix_len == plen &&
2830 ipv6_addr_equal(pfx, &ifp->addr)) {
2832 read_unlock_bh(&idev->lock);
2834 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2835 (ifa_flags & IFA_F_MANAGETEMPADDR))
2836 manage_tempaddrs(idev, ifp, 0, 0, false,
2839 addrconf_verify_rtnl();
2840 if (ipv6_addr_is_multicast(pfx)) {
2841 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2842 false, pfx, dev->ifindex);
2847 read_unlock_bh(&idev->lock);
2848 return -EADDRNOTAVAIL;
2852 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2854 struct in6_ifreq ireq;
2857 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2860 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2864 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2865 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2866 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2871 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2873 struct in6_ifreq ireq;
2876 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2879 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2883 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2884 ireq.ifr6_prefixlen);
2889 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2890 int plen, int scope)
2892 struct inet6_ifaddr *ifp;
2894 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2895 scope, IFA_F_PERMANENT,
2896 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2898 spin_lock_bh(&ifp->lock);
2899 ifp->flags &= ~IFA_F_TENTATIVE;
2900 spin_unlock_bh(&ifp->lock);
2901 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2906 #if IS_ENABLED(CONFIG_IPV6_SIT)
2907 static void sit_add_v4_addrs(struct inet6_dev *idev)
2909 struct in6_addr addr;
2910 struct net_device *dev;
2911 struct net *net = dev_net(idev->dev);
2917 memset(&addr, 0, sizeof(struct in6_addr));
2918 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2920 if (idev->dev->flags&IFF_POINTOPOINT) {
2921 addr.s6_addr32[0] = htonl(0xfe800000);
2925 scope = IPV6_ADDR_COMPATv4;
2927 pflags |= RTF_NONEXTHOP;
2930 if (addr.s6_addr32[3]) {
2931 add_addr(idev, &addr, plen, scope);
2932 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2936 for_each_netdev(net, dev) {
2937 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2938 if (in_dev && (dev->flags & IFF_UP)) {
2939 struct in_ifaddr *ifa;
2943 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2945 addr.s6_addr32[3] = ifa->ifa_local;
2947 if (ifa->ifa_scope == RT_SCOPE_LINK)
2949 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2950 if (idev->dev->flags&IFF_POINTOPOINT)
2955 add_addr(idev, &addr, plen, flag);
2956 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2964 static void init_loopback(struct net_device *dev)
2966 struct inet6_dev *idev;
2967 struct net_device *sp_dev;
2968 struct inet6_ifaddr *sp_ifa;
2969 struct rt6_info *sp_rt;
2975 idev = ipv6_find_idev(dev);
2977 pr_debug("%s: add_dev failed\n", __func__);
2981 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2983 /* Add routes to other interface's IPv6 addresses */
2984 for_each_netdev(dev_net(dev), sp_dev) {
2985 if (!strcmp(sp_dev->name, dev->name))
2988 idev = __in6_dev_get(sp_dev);
2992 read_lock_bh(&idev->lock);
2993 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2995 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2999 /* This dst has been added to garbage list when
3000 * lo device down, release this obsolete dst and
3001 * reallocate a new router for ifa.
3003 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3004 ip6_rt_put(sp_ifa->rt);
3011 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3013 /* Failure cases are ignored */
3014 if (!IS_ERR(sp_rt)) {
3019 read_unlock_bh(&idev->lock);
3023 void addrconf_add_linklocal(struct inet6_dev *idev,
3024 const struct in6_addr *addr, u32 flags)
3026 struct inet6_ifaddr *ifp;
3027 u32 addr_flags = flags | IFA_F_PERMANENT;
3029 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3030 if (idev->cnf.optimistic_dad &&
3031 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3032 addr_flags |= IFA_F_OPTIMISTIC;
3035 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3036 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3038 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3039 addrconf_dad_start(ifp);
3043 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3045 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3047 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3050 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3051 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3054 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3055 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3061 static int ipv6_generate_stable_address(struct in6_addr *address,
3063 const struct inet6_dev *idev)
3065 static DEFINE_SPINLOCK(lock);
3066 static __u32 digest[SHA_DIGEST_WORDS];
3067 static __u32 workspace[SHA_WORKSPACE_WORDS];
3070 char __data[SHA_MESSAGE_BYTES];
3072 struct in6_addr secret;
3074 unsigned char hwaddr[MAX_ADDR_LEN];
3079 struct in6_addr secret;
3080 struct in6_addr temp;
3081 struct net *net = dev_net(idev->dev);
3083 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3085 if (idev->cnf.stable_secret.initialized)
3086 secret = idev->cnf.stable_secret.secret;
3087 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3088 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3093 spin_lock_bh(&lock);
3096 memset(&data, 0, sizeof(data));
3097 memset(workspace, 0, sizeof(workspace));
3098 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3099 data.prefix[0] = address->s6_addr32[0];
3100 data.prefix[1] = address->s6_addr32[1];
3101 data.secret = secret;
3102 data.dad_count = dad_count;
3104 sha_transform(digest, data.__data, workspace);
3107 temp.s6_addr32[2] = (__force __be32)digest[0];
3108 temp.s6_addr32[3] = (__force __be32)digest[1];
3110 spin_unlock_bh(&lock);
3112 if (ipv6_reserved_interfaceid(temp)) {
3114 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3123 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3125 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3129 s = &idev->cnf.stable_secret;
3130 get_random_bytes(&s->secret, sizeof(s->secret));
3131 s->initialized = true;
3134 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3136 struct in6_addr addr;
3138 /* no link local addresses on L3 master devices */
3139 if (netif_is_l3_master(idev->dev))
3142 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3144 switch (idev->addr_gen_mode) {
3145 case IN6_ADDR_GEN_MODE_RANDOM:
3146 ipv6_gen_mode_random_init(idev);
3148 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3149 if (!ipv6_generate_stable_address(&addr, 0, idev))
3150 addrconf_add_linklocal(idev, &addr,
3151 IFA_F_STABLE_PRIVACY);
3152 else if (prefix_route)
3153 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3155 case IN6_ADDR_GEN_MODE_EUI64:
3156 /* addrconf_add_linklocal also adds a prefix_route and we
3157 * only need to care about prefix routes if ipv6_generate_eui64
3158 * couldn't generate one.
3160 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3161 addrconf_add_linklocal(idev, &addr, 0);
3162 else if (prefix_route)
3163 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3165 case IN6_ADDR_GEN_MODE_NONE:
3167 /* will not add any link local address */
3172 static void addrconf_dev_config(struct net_device *dev)
3174 struct inet6_dev *idev;
3178 if ((dev->type != ARPHRD_ETHER) &&
3179 (dev->type != ARPHRD_FDDI) &&
3180 (dev->type != ARPHRD_ARCNET) &&
3181 (dev->type != ARPHRD_INFINIBAND) &&
3182 (dev->type != ARPHRD_IEEE1394) &&
3183 (dev->type != ARPHRD_TUNNEL6) &&
3184 (dev->type != ARPHRD_6LOWPAN) &&
3185 (dev->type != ARPHRD_NONE)) {
3186 /* Alas, we support only Ethernet autoconfiguration. */
3190 idev = addrconf_add_dev(dev);
3194 /* this device type has no EUI support */
3195 if (dev->type == ARPHRD_NONE &&
3196 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3197 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3199 addrconf_addr_gen(idev, false);
3202 #if IS_ENABLED(CONFIG_IPV6_SIT)
3203 static void addrconf_sit_config(struct net_device *dev)
3205 struct inet6_dev *idev;
3210 * Configure the tunnel with one of our IPv4
3211 * addresses... we should configure all of
3212 * our v4 addrs in the tunnel
3215 idev = ipv6_find_idev(dev);
3217 pr_debug("%s: add_dev failed\n", __func__);
3221 if (dev->priv_flags & IFF_ISATAP) {
3222 addrconf_addr_gen(idev, false);
3226 sit_add_v4_addrs(idev);
3228 if (dev->flags&IFF_POINTOPOINT)
3229 addrconf_add_mroute(dev);
3233 #if IS_ENABLED(CONFIG_NET_IPGRE)
3234 static void addrconf_gre_config(struct net_device *dev)
3236 struct inet6_dev *idev;
3240 idev = ipv6_find_idev(dev);
3242 pr_debug("%s: add_dev failed\n", __func__);
3246 addrconf_addr_gen(idev, true);
3247 if (dev->flags & IFF_POINTOPOINT)
3248 addrconf_add_mroute(dev);
3252 static int fixup_permanent_addr(struct inet6_dev *idev,
3253 struct inet6_ifaddr *ifp)
3256 struct rt6_info *rt;
3258 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3259 if (unlikely(IS_ERR(rt)))
3265 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3266 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3270 addrconf_dad_start(ifp);
3275 static void addrconf_permanent_addr(struct net_device *dev)
3277 struct inet6_ifaddr *ifp, *tmp;
3278 struct inet6_dev *idev;
3280 idev = __in6_dev_get(dev);
3284 write_lock_bh(&idev->lock);
3286 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3287 if ((ifp->flags & IFA_F_PERMANENT) &&
3288 fixup_permanent_addr(idev, ifp) < 0) {
3289 write_unlock_bh(&idev->lock);
3291 write_lock_bh(&idev->lock);
3293 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3294 idev->dev->name, &ifp->addr);
3298 write_unlock_bh(&idev->lock);
3301 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3304 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3305 struct netdev_notifier_changeupper_info *info;
3306 struct inet6_dev *idev = __in6_dev_get(dev);
3307 int run_pending = 0;
3311 case NETDEV_REGISTER:
3312 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3313 idev = ipv6_add_dev(dev);
3315 return notifier_from_errno(PTR_ERR(idev));
3319 case NETDEV_CHANGEMTU:
3320 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3321 if (dev->mtu < IPV6_MIN_MTU) {
3322 addrconf_ifdown(dev, 1);
3327 rt6_mtu_change(dev, dev->mtu);
3328 idev->cnf.mtu6 = dev->mtu;
3332 /* allocate new idev */
3333 idev = ipv6_add_dev(dev);
3337 /* device is still not ready */
3338 if (!(idev->if_flags & IF_READY))
3347 if (dev->flags & IFF_SLAVE)
3350 if (idev && idev->cnf.disable_ipv6)
3353 if (event == NETDEV_UP) {
3354 /* restore routes for permanent addresses */
3355 addrconf_permanent_addr(dev);
3357 if (!addrconf_qdisc_ok(dev)) {
3358 /* device is not ready yet. */
3359 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3364 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3365 idev = ipv6_add_dev(dev);
3367 if (!IS_ERR_OR_NULL(idev)) {
3368 idev->if_flags |= IF_READY;
3371 } else if (event == NETDEV_CHANGE) {
3372 if (!addrconf_qdisc_ok(dev)) {
3373 /* device is still not ready. */
3378 if (idev->if_flags & IF_READY)
3379 /* device is already configured. */
3381 idev->if_flags |= IF_READY;
3384 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3390 switch (dev->type) {
3391 #if IS_ENABLED(CONFIG_IPV6_SIT)
3393 addrconf_sit_config(dev);
3396 #if IS_ENABLED(CONFIG_NET_IPGRE)
3398 addrconf_gre_config(dev);
3401 case ARPHRD_LOOPBACK:
3406 addrconf_dev_config(dev);
3410 if (!IS_ERR_OR_NULL(idev)) {
3412 addrconf_dad_run(idev);
3415 * If the MTU changed during the interface down,
3416 * when the interface up, the changed MTU must be
3417 * reflected in the idev as well as routers.
3419 if (idev->cnf.mtu6 != dev->mtu &&
3420 dev->mtu >= IPV6_MIN_MTU) {
3421 rt6_mtu_change(dev, dev->mtu);
3422 idev->cnf.mtu6 = dev->mtu;
3424 idev->tstamp = jiffies;
3425 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3428 * If the changed mtu during down is lower than
3429 * IPV6_MIN_MTU stop IPv6 on this interface.
3431 if (dev->mtu < IPV6_MIN_MTU)
3432 addrconf_ifdown(dev, 1);
3437 case NETDEV_UNREGISTER:
3439 * Remove all addresses from this interface.
3441 addrconf_ifdown(dev, event != NETDEV_DOWN);
3444 case NETDEV_CHANGENAME:
3446 snmp6_unregister_dev(idev);
3447 addrconf_sysctl_unregister(idev);
3448 err = addrconf_sysctl_register(idev);
3450 return notifier_from_errno(err);
3451 err = snmp6_register_dev(idev);
3453 addrconf_sysctl_unregister(idev);
3454 return notifier_from_errno(err);
3459 case NETDEV_PRE_TYPE_CHANGE:
3460 case NETDEV_POST_TYPE_CHANGE:
3462 addrconf_type_change(dev, event);
3465 case NETDEV_CHANGEUPPER:
3468 /* flush all routes if dev is linked to or unlinked from
3469 * an L3 master device (e.g., VRF)
3471 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3472 addrconf_ifdown(dev, 0);
3479 * addrconf module should be notified of a device going up
3481 static struct notifier_block ipv6_dev_notf = {
3482 .notifier_call = addrconf_notify,
3485 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3487 struct inet6_dev *idev;
3490 idev = __in6_dev_get(dev);
3492 if (event == NETDEV_POST_TYPE_CHANGE)
3493 ipv6_mc_remap(idev);
3494 else if (event == NETDEV_PRE_TYPE_CHANGE)
3495 ipv6_mc_unmap(idev);
3498 static bool addr_is_local(const struct in6_addr *addr)
3500 return ipv6_addr_type(addr) &
3501 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3504 static int addrconf_ifdown(struct net_device *dev, int how)
3506 struct net *net = dev_net(dev);
3507 struct inet6_dev *idev;
3508 struct inet6_ifaddr *ifa, *tmp;
3509 struct list_head del_list;
3516 rt6_ifdown(net, dev);
3517 neigh_ifdown(&nd_tbl, dev);
3519 idev = __in6_dev_get(dev);
3524 * Step 1: remove reference to ipv6 device from parent device.
3530 /* protected by rtnl_lock */
3531 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3533 /* Step 1.5: remove snmp6 entry */
3534 snmp6_unregister_dev(idev);
3538 /* aggregate the system setting and interface setting */
3539 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3541 _keep_addr = idev->cnf.keep_addr_on_down;
3543 /* combine the user config with event to determine if permanent
3544 * addresses are to be removed from address hash table
3546 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3548 /* Step 2: clear hash table */
3549 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3550 struct hlist_head *h = &inet6_addr_lst[i];
3552 spin_lock_bh(&addrconf_hash_lock);
3554 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3555 if (ifa->idev == idev) {
3556 addrconf_del_dad_work(ifa);
3557 /* combined flag + permanent flag decide if
3558 * address is retained on a down event
3561 !(ifa->flags & IFA_F_PERMANENT) ||
3562 addr_is_local(&ifa->addr)) {
3563 hlist_del_init_rcu(&ifa->addr_lst);
3568 spin_unlock_bh(&addrconf_hash_lock);
3571 write_lock_bh(&idev->lock);
3573 addrconf_del_rs_timer(idev);
3575 /* Step 2: clear flags for stateless addrconf */
3577 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3579 /* Step 3: clear tempaddr list */
3580 while (!list_empty(&idev->tempaddr_list)) {
3581 ifa = list_first_entry(&idev->tempaddr_list,
3582 struct inet6_ifaddr, tmp_list);
3583 list_del(&ifa->tmp_list);
3584 write_unlock_bh(&idev->lock);
3585 spin_lock_bh(&ifa->lock);
3588 in6_ifa_put(ifa->ifpub);
3591 spin_unlock_bh(&ifa->lock);
3593 write_lock_bh(&idev->lock);
3596 /* re-combine the user config with event to determine if permanent
3597 * addresses are to be removed from the interface list
3599 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3601 INIT_LIST_HEAD(&del_list);
3602 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3603 struct rt6_info *rt = NULL;
3605 addrconf_del_dad_work(ifa);
3607 write_unlock_bh(&idev->lock);
3608 spin_lock_bh(&ifa->lock);
3610 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3611 !addr_is_local(&ifa->addr)) {
3612 /* set state to skip the notifier below */
3613 state = INET6_IFADDR_STATE_DEAD;
3615 if (!(ifa->flags & IFA_F_NODAD))
3616 ifa->flags |= IFA_F_TENTATIVE;
3622 ifa->state = INET6_IFADDR_STATE_DEAD;
3624 list_move(&ifa->if_list, &del_list);
3627 spin_unlock_bh(&ifa->lock);
3632 if (state != INET6_IFADDR_STATE_DEAD) {
3633 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3634 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3636 if (idev->cnf.forwarding)
3637 addrconf_leave_anycast(ifa);
3638 addrconf_leave_solict(ifa->idev, &ifa->addr);
3641 write_lock_bh(&idev->lock);
3644 write_unlock_bh(&idev->lock);
3646 /* now clean up addresses to be removed */
3647 while (!list_empty(&del_list)) {
3648 ifa = list_first_entry(&del_list,
3649 struct inet6_ifaddr, if_list);
3650 list_del(&ifa->if_list);
3655 /* Step 5: Discard anycast and multicast list */
3657 ipv6_ac_destroy_dev(idev);
3658 ipv6_mc_destroy_dev(idev);
3663 idev->tstamp = jiffies;
3665 /* Last: Shot the device (if unregistered) */
3667 addrconf_sysctl_unregister(idev);
3668 neigh_parms_release(&nd_tbl, idev->nd_parms);
3669 neigh_ifdown(&nd_tbl, dev);
3675 static void addrconf_rs_timer(unsigned long data)
3677 struct inet6_dev *idev = (struct inet6_dev *)data;
3678 struct net_device *dev = idev->dev;
3679 struct in6_addr lladdr;
3681 write_lock(&idev->lock);
3682 if (idev->dead || !(idev->if_flags & IF_READY))
3685 if (!ipv6_accept_ra(idev))
3688 /* Announcement received after solicitation was sent */
3689 if (idev->if_flags & IF_RA_RCVD)
3692 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3693 write_unlock(&idev->lock);
3694 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3695 ndisc_send_rs(dev, &lladdr,
3696 &in6addr_linklocal_allrouters);
3700 write_lock(&idev->lock);
3701 idev->rs_interval = rfc3315_s14_backoff_update(
3702 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3703 /* The wait after the last probe can be shorter */
3704 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3705 idev->cnf.rtr_solicits) ?
3706 idev->cnf.rtr_solicit_delay :
3710 * Note: we do not support deprecated "all on-link"
3711 * assumption any longer.
3713 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3717 write_unlock(&idev->lock);
3723 * Duplicate Address Detection
3725 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3727 unsigned long rand_num;
3728 struct inet6_dev *idev = ifp->idev;
3730 if (ifp->flags & IFA_F_OPTIMISTIC)
3733 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3735 ifp->dad_probes = idev->cnf.dad_transmits;
3736 addrconf_mod_dad_work(ifp, rand_num);
3739 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3741 struct inet6_dev *idev = ifp->idev;
3742 struct net_device *dev = idev->dev;
3743 bool notify = false;
3745 addrconf_join_solict(dev, &ifp->addr);
3747 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3749 read_lock_bh(&idev->lock);
3750 spin_lock(&ifp->lock);
3751 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3754 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3755 idev->cnf.accept_dad < 1 ||
3756 !(ifp->flags&IFA_F_TENTATIVE) ||
3757 ifp->flags & IFA_F_NODAD) {
3758 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3759 spin_unlock(&ifp->lock);
3760 read_unlock_bh(&idev->lock);
3762 addrconf_dad_completed(ifp);
3766 if (!(idev->if_flags & IF_READY)) {
3767 spin_unlock(&ifp->lock);
3768 read_unlock_bh(&idev->lock);
3770 * If the device is not ready:
3771 * - keep it tentative if it is a permanent address.
3772 * - otherwise, kill it.
3775 addrconf_dad_stop(ifp, 0);
3780 * Optimistic nodes can start receiving
3783 if (ifp->flags & IFA_F_OPTIMISTIC) {
3784 ip6_ins_rt(ifp->rt);
3785 if (ipv6_use_optimistic_addr(idev)) {
3786 /* Because optimistic nodes can use this address,
3787 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3793 addrconf_dad_kick(ifp);
3795 spin_unlock(&ifp->lock);
3796 read_unlock_bh(&idev->lock);
3798 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3801 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3803 bool begin_dad = false;
3805 spin_lock_bh(&ifp->lock);
3806 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3807 ifp->state = INET6_IFADDR_STATE_PREDAD;
3810 spin_unlock_bh(&ifp->lock);
3813 addrconf_mod_dad_work(ifp, 0);
3816 static void addrconf_dad_work(struct work_struct *w)
3818 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3819 struct inet6_ifaddr,
3821 struct inet6_dev *idev = ifp->idev;
3822 struct in6_addr mcaddr;
3823 bool disable_ipv6 = false;
3829 } action = DAD_PROCESS;
3833 spin_lock_bh(&ifp->lock);
3834 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3836 ifp->state = INET6_IFADDR_STATE_DAD;
3837 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3839 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3841 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3842 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3843 struct in6_addr addr;
3845 addr.s6_addr32[0] = htonl(0xfe800000);
3846 addr.s6_addr32[1] = 0;
3848 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3849 ipv6_addr_equal(&ifp->addr, &addr)) {
3850 /* DAD failed for link-local based on MAC */
3851 idev->cnf.disable_ipv6 = 1;
3853 pr_info("%s: IPv6 being disabled!\n",
3854 ifp->idev->dev->name);
3855 disable_ipv6 = true;
3859 spin_unlock_bh(&ifp->lock);
3861 if (action == DAD_BEGIN) {
3862 addrconf_dad_begin(ifp);
3864 } else if (action == DAD_ABORT) {
3866 addrconf_dad_stop(ifp, 1);
3868 addrconf_ifdown(idev->dev, 0);
3872 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3875 write_lock_bh(&idev->lock);
3876 if (idev->dead || !(idev->if_flags & IF_READY)) {
3877 write_unlock_bh(&idev->lock);
3881 spin_lock(&ifp->lock);
3882 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3883 spin_unlock(&ifp->lock);
3884 write_unlock_bh(&idev->lock);
3888 if (ifp->dad_probes == 0) {
3890 * DAD was successful
3893 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3894 spin_unlock(&ifp->lock);
3895 write_unlock_bh(&idev->lock);
3897 addrconf_dad_completed(ifp);
3903 addrconf_mod_dad_work(ifp,
3904 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3905 spin_unlock(&ifp->lock);
3906 write_unlock_bh(&idev->lock);
3908 /* send a neighbour solicitation for our addr */
3909 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3910 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3916 /* ifp->idev must be at least read locked */
3917 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3919 struct inet6_ifaddr *ifpiter;
3920 struct inet6_dev *idev = ifp->idev;
3922 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3923 if (ifpiter->scope > IFA_LINK)
3925 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3926 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3927 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3934 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3936 struct net_device *dev = ifp->idev->dev;
3937 struct in6_addr lladdr;
3938 bool send_rs, send_mld;
3940 addrconf_del_dad_work(ifp);
3943 * Configure the address for reception. Now it is valid.
3946 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3948 /* If added prefix is link local and we are prepared to process
3949 router advertisements, start sending router solicitations.
3952 read_lock_bh(&ifp->idev->lock);
3953 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3954 send_rs = send_mld &&
3955 ipv6_accept_ra(ifp->idev) &&
3956 ifp->idev->cnf.rtr_solicits != 0 &&
3957 (dev->flags&IFF_LOOPBACK) == 0;
3958 read_unlock_bh(&ifp->idev->lock);
3960 /* While dad is in progress mld report's source address is in6_addrany.
3961 * Resend with proper ll now.
3964 ipv6_mc_dad_complete(ifp->idev);
3968 * If a host as already performed a random delay
3969 * [...] as part of DAD [...] there is no need
3970 * to delay again before sending the first RS
3972 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3974 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3976 write_lock_bh(&ifp->idev->lock);
3977 spin_lock(&ifp->lock);
3978 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
3979 ifp->idev->cnf.rtr_solicit_interval);
3980 ifp->idev->rs_probes = 1;
3981 ifp->idev->if_flags |= IF_RS_SENT;
3982 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
3983 spin_unlock(&ifp->lock);
3984 write_unlock_bh(&ifp->idev->lock);
3988 static void addrconf_dad_run(struct inet6_dev *idev)
3990 struct inet6_ifaddr *ifp;
3992 read_lock_bh(&idev->lock);
3993 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3994 spin_lock(&ifp->lock);
3995 if (ifp->flags & IFA_F_TENTATIVE &&
3996 ifp->state == INET6_IFADDR_STATE_DAD)
3997 addrconf_dad_kick(ifp);
3998 spin_unlock(&ifp->lock);
4000 read_unlock_bh(&idev->lock);
4003 #ifdef CONFIG_PROC_FS
4004 struct if6_iter_state {
4005 struct seq_net_private p;
4010 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4012 struct inet6_ifaddr *ifa = NULL;
4013 struct if6_iter_state *state = seq->private;
4014 struct net *net = seq_file_net(seq);
4017 /* initial bucket if pos is 0 */
4023 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4024 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4026 if (!net_eq(dev_net(ifa->idev->dev), net))
4028 /* sync with offset */
4029 if (p < state->offset) {
4037 /* prepare for next bucket */
4044 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4045 struct inet6_ifaddr *ifa)
4047 struct if6_iter_state *state = seq->private;
4048 struct net *net = seq_file_net(seq);
4050 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4051 if (!net_eq(dev_net(ifa->idev->dev), net))
4057 while (++state->bucket < IN6_ADDR_HSIZE) {
4059 hlist_for_each_entry_rcu_bh(ifa,
4060 &inet6_addr_lst[state->bucket], addr_lst) {
4061 if (!net_eq(dev_net(ifa->idev->dev), net))
4071 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4075 return if6_get_first(seq, *pos);
4078 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4080 struct inet6_ifaddr *ifa;
4082 ifa = if6_get_next(seq, v);
4087 static void if6_seq_stop(struct seq_file *seq, void *v)
4090 rcu_read_unlock_bh();
4093 static int if6_seq_show(struct seq_file *seq, void *v)
4095 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4096 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4098 ifp->idev->dev->ifindex,
4102 ifp->idev->dev->name);
4106 static const struct seq_operations if6_seq_ops = {
4107 .start = if6_seq_start,
4108 .next = if6_seq_next,
4109 .show = if6_seq_show,
4110 .stop = if6_seq_stop,
4113 static int if6_seq_open(struct inode *inode, struct file *file)
4115 return seq_open_net(inode, file, &if6_seq_ops,
4116 sizeof(struct if6_iter_state));
4119 static const struct file_operations if6_fops = {
4120 .owner = THIS_MODULE,
4121 .open = if6_seq_open,
4123 .llseek = seq_lseek,
4124 .release = seq_release_net,
4127 static int __net_init if6_proc_net_init(struct net *net)
4129 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4134 static void __net_exit if6_proc_net_exit(struct net *net)
4136 remove_proc_entry("if_inet6", net->proc_net);
4139 static struct pernet_operations if6_proc_net_ops = {
4140 .init = if6_proc_net_init,
4141 .exit = if6_proc_net_exit,
4144 int __init if6_proc_init(void)
4146 return register_pernet_subsys(&if6_proc_net_ops);
4149 void if6_proc_exit(void)
4151 unregister_pernet_subsys(&if6_proc_net_ops);
4153 #endif /* CONFIG_PROC_FS */
4155 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4156 /* Check if address is a home address configured on any interface. */
4157 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4160 struct inet6_ifaddr *ifp = NULL;
4161 unsigned int hash = inet6_addr_hash(addr);
4164 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4165 if (!net_eq(dev_net(ifp->idev->dev), net))
4167 if (ipv6_addr_equal(&ifp->addr, addr) &&
4168 (ifp->flags & IFA_F_HOMEADDRESS)) {
4173 rcu_read_unlock_bh();
4179 * Periodic address status verification
4182 static void addrconf_verify_rtnl(void)
4184 unsigned long now, next, next_sec, next_sched;
4185 struct inet6_ifaddr *ifp;
4192 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4194 cancel_delayed_work(&addr_chk_work);
4196 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4198 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4201 /* When setting preferred_lft to a value not zero or
4202 * infinity, while valid_lft is infinity
4203 * IFA_F_PERMANENT has a non-infinity life time.
4205 if ((ifp->flags & IFA_F_PERMANENT) &&
4206 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4209 spin_lock(&ifp->lock);
4210 /* We try to batch several events at once. */
4211 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4213 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4214 age >= ifp->valid_lft) {
4215 spin_unlock(&ifp->lock);
4219 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4220 spin_unlock(&ifp->lock);
4222 } else if (age >= ifp->prefered_lft) {
4223 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4226 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4228 ifp->flags |= IFA_F_DEPRECATED;
4231 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4232 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4233 next = ifp->tstamp + ifp->valid_lft * HZ;
4235 spin_unlock(&ifp->lock);
4240 ipv6_ifa_notify(0, ifp);
4244 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4245 !(ifp->flags&IFA_F_TENTATIVE)) {
4246 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4247 ifp->idev->cnf.dad_transmits *
4248 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4250 if (age >= ifp->prefered_lft - regen_advance) {
4251 struct inet6_ifaddr *ifpub = ifp->ifpub;
4252 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4253 next = ifp->tstamp + ifp->prefered_lft * HZ;
4254 if (!ifp->regen_count && ifpub) {
4257 in6_ifa_hold(ifpub);
4258 spin_unlock(&ifp->lock);
4260 spin_lock(&ifpub->lock);
4261 ifpub->regen_count = 0;
4262 spin_unlock(&ifpub->lock);
4263 ipv6_create_tempaddr(ifpub, ifp);
4268 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4269 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4270 spin_unlock(&ifp->lock);
4272 /* ifp->prefered_lft <= ifp->valid_lft */
4273 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4274 next = ifp->tstamp + ifp->prefered_lft * HZ;
4275 spin_unlock(&ifp->lock);
4280 next_sec = round_jiffies_up(next);
4283 /* If rounded timeout is accurate enough, accept it. */
4284 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4285 next_sched = next_sec;
4287 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4288 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4289 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4291 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4292 now, next, next_sec, next_sched);
4293 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4294 rcu_read_unlock_bh();
4297 static void addrconf_verify_work(struct work_struct *w)
4300 addrconf_verify_rtnl();
4304 static void addrconf_verify(void)
4306 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4309 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4310 struct in6_addr **peer_pfx)
4312 struct in6_addr *pfx = NULL;
4317 pfx = nla_data(addr);
4320 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4322 pfx = nla_data(local);
4328 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4329 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4330 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4331 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4332 [IFA_FLAGS] = { .len = sizeof(u32) },
4336 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4338 struct net *net = sock_net(skb->sk);
4339 struct ifaddrmsg *ifm;
4340 struct nlattr *tb[IFA_MAX+1];
4341 struct in6_addr *pfx, *peer_pfx;
4345 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4349 ifm = nlmsg_data(nlh);
4350 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4354 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4356 /* We ignore other flags so far. */
4357 ifa_flags &= IFA_F_MANAGETEMPADDR;
4359 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4360 ifm->ifa_prefixlen);
4363 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4364 u32 prefered_lft, u32 valid_lft)
4368 unsigned long timeout;
4369 bool was_managetempaddr;
4370 bool had_prefixroute;
4374 if (!valid_lft || (prefered_lft > valid_lft))
4377 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4378 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4381 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4382 if (addrconf_finite_timeout(timeout)) {
4383 expires = jiffies_to_clock_t(timeout * HZ);
4384 valid_lft = timeout;
4385 flags = RTF_EXPIRES;
4389 ifa_flags |= IFA_F_PERMANENT;
4392 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4393 if (addrconf_finite_timeout(timeout)) {
4395 ifa_flags |= IFA_F_DEPRECATED;
4396 prefered_lft = timeout;
4399 spin_lock_bh(&ifp->lock);
4400 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4401 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4402 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4403 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4404 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4405 IFA_F_NOPREFIXROUTE);
4406 ifp->flags |= ifa_flags;
4407 ifp->tstamp = jiffies;
4408 ifp->valid_lft = valid_lft;
4409 ifp->prefered_lft = prefered_lft;
4411 spin_unlock_bh(&ifp->lock);
4412 if (!(ifp->flags&IFA_F_TENTATIVE))
4413 ipv6_ifa_notify(0, ifp);
4415 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4416 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4418 } else if (had_prefixroute) {
4419 enum cleanup_prefix_rt_t action;
4420 unsigned long rt_expires;
4422 write_lock_bh(&ifp->idev->lock);
4423 action = check_cleanup_prefix_route(ifp, &rt_expires);
4424 write_unlock_bh(&ifp->idev->lock);
4426 if (action != CLEANUP_PREFIX_RT_NOP) {
4427 cleanup_prefix_route(ifp, rt_expires,
4428 action == CLEANUP_PREFIX_RT_DEL);
4432 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4433 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4434 valid_lft = prefered_lft = 0;
4435 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4436 !was_managetempaddr, jiffies);
4439 addrconf_verify_rtnl();
4445 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4447 struct net *net = sock_net(skb->sk);
4448 struct ifaddrmsg *ifm;
4449 struct nlattr *tb[IFA_MAX+1];
4450 struct in6_addr *pfx, *peer_pfx;
4451 struct inet6_ifaddr *ifa;
4452 struct net_device *dev;
4453 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4457 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4461 ifm = nlmsg_data(nlh);
4462 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4466 if (tb[IFA_CACHEINFO]) {
4467 struct ifa_cacheinfo *ci;
4469 ci = nla_data(tb[IFA_CACHEINFO]);
4470 valid_lft = ci->ifa_valid;
4471 preferred_lft = ci->ifa_prefered;
4473 preferred_lft = INFINITY_LIFE_TIME;
4474 valid_lft = INFINITY_LIFE_TIME;
4477 dev = __dev_get_by_index(net, ifm->ifa_index);
4481 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4483 /* We ignore other flags so far. */
4484 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4485 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4487 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4490 * It would be best to check for !NLM_F_CREATE here but
4491 * userspace already relies on not having to provide this.
4493 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4494 ifm->ifa_prefixlen, ifa_flags,
4495 preferred_lft, valid_lft);
4498 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4499 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4502 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4509 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4510 u8 scope, int ifindex)
4512 struct ifaddrmsg *ifm;
4514 ifm = nlmsg_data(nlh);
4515 ifm->ifa_family = AF_INET6;
4516 ifm->ifa_prefixlen = prefixlen;
4517 ifm->ifa_flags = flags;
4518 ifm->ifa_scope = scope;
4519 ifm->ifa_index = ifindex;
4522 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4523 unsigned long tstamp, u32 preferred, u32 valid)
4525 struct ifa_cacheinfo ci;
4527 ci.cstamp = cstamp_delta(cstamp);
4528 ci.tstamp = cstamp_delta(tstamp);
4529 ci.ifa_prefered = preferred;
4530 ci.ifa_valid = valid;
4532 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4535 static inline int rt_scope(int ifa_scope)
4537 if (ifa_scope & IFA_HOST)
4538 return RT_SCOPE_HOST;
4539 else if (ifa_scope & IFA_LINK)
4540 return RT_SCOPE_LINK;
4541 else if (ifa_scope & IFA_SITE)
4542 return RT_SCOPE_SITE;
4544 return RT_SCOPE_UNIVERSE;
4547 static inline int inet6_ifaddr_msgsize(void)
4549 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4550 + nla_total_size(16) /* IFA_LOCAL */
4551 + nla_total_size(16) /* IFA_ADDRESS */
4552 + nla_total_size(sizeof(struct ifa_cacheinfo))
4553 + nla_total_size(4) /* IFA_FLAGS */;
4556 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4557 u32 portid, u32 seq, int event, unsigned int flags)
4559 struct nlmsghdr *nlh;
4560 u32 preferred, valid;
4562 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4566 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4567 ifa->idev->dev->ifindex);
4569 if (!((ifa->flags&IFA_F_PERMANENT) &&
4570 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4571 preferred = ifa->prefered_lft;
4572 valid = ifa->valid_lft;
4573 if (preferred != INFINITY_LIFE_TIME) {
4574 long tval = (jiffies - ifa->tstamp)/HZ;
4575 if (preferred > tval)
4579 if (valid != INFINITY_LIFE_TIME) {
4587 preferred = INFINITY_LIFE_TIME;
4588 valid = INFINITY_LIFE_TIME;
4591 if (!ipv6_addr_any(&ifa->peer_addr)) {
4592 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4593 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4596 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4599 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4602 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4605 nlmsg_end(skb, nlh);
4609 nlmsg_cancel(skb, nlh);
4613 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4614 u32 portid, u32 seq, int event, u16 flags)
4616 struct nlmsghdr *nlh;
4617 u8 scope = RT_SCOPE_UNIVERSE;
4618 int ifindex = ifmca->idev->dev->ifindex;
4620 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4621 scope = RT_SCOPE_SITE;
4623 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4627 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4628 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4629 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4630 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4631 nlmsg_cancel(skb, nlh);
4635 nlmsg_end(skb, nlh);
4639 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4640 u32 portid, u32 seq, int event, unsigned int flags)
4642 struct nlmsghdr *nlh;
4643 u8 scope = RT_SCOPE_UNIVERSE;
4644 int ifindex = ifaca->aca_idev->dev->ifindex;
4646 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4647 scope = RT_SCOPE_SITE;
4649 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4653 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4654 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4655 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4656 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4657 nlmsg_cancel(skb, nlh);
4661 nlmsg_end(skb, nlh);
4671 /* called with rcu_read_lock() */
4672 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4673 struct netlink_callback *cb, enum addr_type_t type,
4674 int s_ip_idx, int *p_ip_idx)
4676 struct ifmcaddr6 *ifmca;
4677 struct ifacaddr6 *ifaca;
4679 int ip_idx = *p_ip_idx;
4681 read_lock_bh(&idev->lock);
4683 case UNICAST_ADDR: {
4684 struct inet6_ifaddr *ifa;
4686 /* unicast address incl. temp addr */
4687 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4688 if (++ip_idx < s_ip_idx)
4690 err = inet6_fill_ifaddr(skb, ifa,
4691 NETLINK_CB(cb->skb).portid,
4697 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4701 case MULTICAST_ADDR:
4702 /* multicast address */
4703 for (ifmca = idev->mc_list; ifmca;
4704 ifmca = ifmca->next, ip_idx++) {
4705 if (ip_idx < s_ip_idx)
4707 err = inet6_fill_ifmcaddr(skb, ifmca,
4708 NETLINK_CB(cb->skb).portid,
4717 /* anycast address */
4718 for (ifaca = idev->ac_list; ifaca;
4719 ifaca = ifaca->aca_next, ip_idx++) {
4720 if (ip_idx < s_ip_idx)
4722 err = inet6_fill_ifacaddr(skb, ifaca,
4723 NETLINK_CB(cb->skb).portid,
4734 read_unlock_bh(&idev->lock);
4739 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4740 enum addr_type_t type)
4742 struct net *net = sock_net(skb->sk);
4745 int s_idx, s_ip_idx;
4746 struct net_device *dev;
4747 struct inet6_dev *idev;
4748 struct hlist_head *head;
4751 s_idx = idx = cb->args[1];
4752 s_ip_idx = ip_idx = cb->args[2];
4755 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4756 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4758 head = &net->dev_index_head[h];
4759 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4762 if (h > s_h || idx > s_idx)
4765 idev = __in6_dev_get(dev);
4769 if (in6_dump_addrs(idev, skb, cb, type,
4770 s_ip_idx, &ip_idx) < 0)
4780 cb->args[2] = ip_idx;
4785 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4787 enum addr_type_t type = UNICAST_ADDR;
4789 return inet6_dump_addr(skb, cb, type);
4792 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4794 enum addr_type_t type = MULTICAST_ADDR;
4796 return inet6_dump_addr(skb, cb, type);
4800 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4802 enum addr_type_t type = ANYCAST_ADDR;
4804 return inet6_dump_addr(skb, cb, type);
4807 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4809 struct net *net = sock_net(in_skb->sk);
4810 struct ifaddrmsg *ifm;
4811 struct nlattr *tb[IFA_MAX+1];
4812 struct in6_addr *addr = NULL, *peer;
4813 struct net_device *dev = NULL;
4814 struct inet6_ifaddr *ifa;
4815 struct sk_buff *skb;
4818 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4822 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4828 ifm = nlmsg_data(nlh);
4830 dev = __dev_get_by_index(net, ifm->ifa_index);
4832 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4834 err = -EADDRNOTAVAIL;
4838 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4844 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4845 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4847 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4848 WARN_ON(err == -EMSGSIZE);
4852 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4859 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4861 struct sk_buff *skb;
4862 struct net *net = dev_net(ifa->idev->dev);
4865 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4869 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4871 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4872 WARN_ON(err == -EMSGSIZE);
4876 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4880 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4883 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4884 __s32 *array, int bytes)
4886 BUG_ON(bytes < (DEVCONF_MAX * 4));
4888 memset(array, 0, bytes);
4889 array[DEVCONF_FORWARDING] = cnf->forwarding;
4890 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4891 array[DEVCONF_MTU6] = cnf->mtu6;
4892 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4893 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4894 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4895 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4896 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4897 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4898 jiffies_to_msecs(cnf->rtr_solicit_interval);
4899 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4900 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4901 array[DEVCONF_RTR_SOLICIT_DELAY] =
4902 jiffies_to_msecs(cnf->rtr_solicit_delay);
4903 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4904 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4905 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4906 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4907 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4908 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4909 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4910 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4911 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4912 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4913 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4914 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4915 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4916 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4917 #ifdef CONFIG_IPV6_ROUTER_PREF
4918 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4919 array[DEVCONF_RTR_PROBE_INTERVAL] =
4920 jiffies_to_msecs(cnf->rtr_probe_interval);
4921 #ifdef CONFIG_IPV6_ROUTE_INFO
4922 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4925 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4926 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4927 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4928 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4929 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4931 #ifdef CONFIG_IPV6_MROUTE
4932 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4934 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4935 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4936 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4937 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4938 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4939 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4940 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4941 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4942 /* we omit DEVCONF_STABLE_SECRET for now */
4943 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4944 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4945 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4946 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4949 static inline size_t inet6_ifla6_size(void)
4951 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4952 + nla_total_size(sizeof(struct ifla_cacheinfo))
4953 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4954 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4955 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4956 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4959 static inline size_t inet6_if_nlmsg_size(void)
4961 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4962 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4963 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4964 + nla_total_size(4) /* IFLA_MTU */
4965 + nla_total_size(4) /* IFLA_LINK */
4966 + nla_total_size(1) /* IFLA_OPERSTATE */
4967 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4970 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4974 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
4977 /* Use put_unaligned() because stats may not be aligned for u64. */
4978 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
4979 for (i = 1; i < ICMP6_MIB_MAX; i++)
4980 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4982 memset(&stats[ICMP6_MIB_MAX], 0, pad);
4985 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4986 int bytes, size_t syncpoff)
4989 u64 buff[IPSTATS_MIB_MAX];
4990 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4994 memset(buff, 0, sizeof(buff));
4995 buff[0] = IPSTATS_MIB_MAX;
4997 for_each_possible_cpu(c) {
4998 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4999 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5002 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5003 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5006 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5010 case IFLA_INET6_STATS:
5011 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5012 offsetof(struct ipstats_mib, syncp));
5014 case IFLA_INET6_ICMP6STATS:
5015 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5020 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5021 u32 ext_filter_mask)
5024 struct ifla_cacheinfo ci;
5026 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5027 goto nla_put_failure;
5028 ci.max_reasm_len = IPV6_MAXPLEN;
5029 ci.tstamp = cstamp_delta(idev->tstamp);
5030 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5031 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5032 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5033 goto nla_put_failure;
5034 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5036 goto nla_put_failure;
5037 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5039 /* XXX - MC not implemented */
5041 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5044 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5046 goto nla_put_failure;
5047 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5049 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5051 goto nla_put_failure;
5052 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5054 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5056 goto nla_put_failure;
5058 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5059 goto nla_put_failure;
5061 read_lock_bh(&idev->lock);
5062 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5063 read_unlock_bh(&idev->lock);
5071 static size_t inet6_get_link_af_size(const struct net_device *dev,
5072 u32 ext_filter_mask)
5074 if (!__in6_dev_get(dev))
5077 return inet6_ifla6_size();
5080 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5081 u32 ext_filter_mask)
5083 struct inet6_dev *idev = __in6_dev_get(dev);
5088 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5094 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5096 struct inet6_ifaddr *ifp;
5097 struct net_device *dev = idev->dev;
5098 bool clear_token, update_rs = false;
5099 struct in6_addr ll_addr;
5105 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5107 if (!ipv6_accept_ra(idev))
5109 if (idev->cnf.rtr_solicits == 0)
5112 write_lock_bh(&idev->lock);
5114 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5115 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5117 write_unlock_bh(&idev->lock);
5119 clear_token = ipv6_addr_any(token);
5123 if (!idev->dead && (idev->if_flags & IF_READY) &&
5124 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5125 IFA_F_OPTIMISTIC)) {
5126 /* If we're not ready, then normal ifup will take care
5127 * of this. Otherwise, we need to request our rs here.
5129 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5134 write_lock_bh(&idev->lock);
5137 idev->if_flags |= IF_RS_SENT;
5138 idev->rs_interval = rfc3315_s14_backoff_init(
5139 idev->cnf.rtr_solicit_interval);
5140 idev->rs_probes = 1;
5141 addrconf_mod_rs_timer(idev, idev->rs_interval);
5144 /* Well, that's kinda nasty ... */
5145 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5146 spin_lock(&ifp->lock);
5147 if (ifp->tokenized) {
5149 ifp->prefered_lft = 0;
5151 spin_unlock(&ifp->lock);
5154 write_unlock_bh(&idev->lock);
5155 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5156 addrconf_verify_rtnl();
5160 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5161 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5162 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5165 static int inet6_validate_link_af(const struct net_device *dev,
5166 const struct nlattr *nla)
5168 struct nlattr *tb[IFLA_INET6_MAX + 1];
5170 if (dev && !__in6_dev_get(dev))
5171 return -EAFNOSUPPORT;
5173 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5176 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5179 struct inet6_dev *idev = __in6_dev_get(dev);
5180 struct nlattr *tb[IFLA_INET6_MAX + 1];
5183 return -EAFNOSUPPORT;
5185 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5188 if (tb[IFLA_INET6_TOKEN]) {
5189 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5194 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5195 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5197 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5198 mode != IN6_ADDR_GEN_MODE_NONE &&
5199 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5200 mode != IN6_ADDR_GEN_MODE_RANDOM)
5203 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5204 !idev->cnf.stable_secret.initialized &&
5205 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5208 idev->addr_gen_mode = mode;
5215 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5216 u32 portid, u32 seq, int event, unsigned int flags)
5218 struct net_device *dev = idev->dev;
5219 struct ifinfomsg *hdr;
5220 struct nlmsghdr *nlh;
5223 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5227 hdr = nlmsg_data(nlh);
5228 hdr->ifi_family = AF_INET6;
5230 hdr->ifi_type = dev->type;
5231 hdr->ifi_index = dev->ifindex;
5232 hdr->ifi_flags = dev_get_flags(dev);
5233 hdr->ifi_change = 0;
5235 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5237 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5238 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5239 (dev->ifindex != dev_get_iflink(dev) &&
5240 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5241 nla_put_u8(skb, IFLA_OPERSTATE,
5242 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5243 goto nla_put_failure;
5244 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5246 goto nla_put_failure;
5248 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5249 goto nla_put_failure;
5251 nla_nest_end(skb, protoinfo);
5252 nlmsg_end(skb, nlh);
5256 nlmsg_cancel(skb, nlh);
5260 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5262 struct net *net = sock_net(skb->sk);
5265 struct net_device *dev;
5266 struct inet6_dev *idev;
5267 struct hlist_head *head;
5270 s_idx = cb->args[1];
5273 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5275 head = &net->dev_index_head[h];
5276 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5279 idev = __in6_dev_get(dev);
5282 if (inet6_fill_ifinfo(skb, idev,
5283 NETLINK_CB(cb->skb).portid,
5285 RTM_NEWLINK, NLM_F_MULTI) < 0)
5299 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5301 struct sk_buff *skb;
5302 struct net *net = dev_net(idev->dev);
5305 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5309 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5311 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5312 WARN_ON(err == -EMSGSIZE);
5316 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5320 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5323 static inline size_t inet6_prefix_nlmsg_size(void)
5325 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5326 + nla_total_size(sizeof(struct in6_addr))
5327 + nla_total_size(sizeof(struct prefix_cacheinfo));
5330 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5331 struct prefix_info *pinfo, u32 portid, u32 seq,
5332 int event, unsigned int flags)
5334 struct prefixmsg *pmsg;
5335 struct nlmsghdr *nlh;
5336 struct prefix_cacheinfo ci;
5338 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5342 pmsg = nlmsg_data(nlh);
5343 pmsg->prefix_family = AF_INET6;
5344 pmsg->prefix_pad1 = 0;
5345 pmsg->prefix_pad2 = 0;
5346 pmsg->prefix_ifindex = idev->dev->ifindex;
5347 pmsg->prefix_len = pinfo->prefix_len;
5348 pmsg->prefix_type = pinfo->type;
5349 pmsg->prefix_pad3 = 0;
5350 pmsg->prefix_flags = 0;
5352 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5353 if (pinfo->autoconf)
5354 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5356 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5357 goto nla_put_failure;
5358 ci.preferred_time = ntohl(pinfo->prefered);
5359 ci.valid_time = ntohl(pinfo->valid);
5360 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5361 goto nla_put_failure;
5362 nlmsg_end(skb, nlh);
5366 nlmsg_cancel(skb, nlh);
5370 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5371 struct prefix_info *pinfo)
5373 struct sk_buff *skb;
5374 struct net *net = dev_net(idev->dev);
5377 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5381 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5383 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5384 WARN_ON(err == -EMSGSIZE);
5388 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5392 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5395 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5397 struct net *net = dev_net(ifp->idev->dev);
5402 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5407 * If the address was optimistic
5408 * we inserted the route at the start of
5409 * our DAD process, so we don't need
5412 if (!(ifp->rt->rt6i_node))
5413 ip6_ins_rt(ifp->rt);
5414 if (ifp->idev->cnf.forwarding)
5415 addrconf_join_anycast(ifp);
5416 if (!ipv6_addr_any(&ifp->peer_addr))
5417 addrconf_prefix_route(&ifp->peer_addr, 128,
5418 ifp->idev->dev, 0, 0);
5421 if (ifp->idev->cnf.forwarding)
5422 addrconf_leave_anycast(ifp);
5423 addrconf_leave_solict(ifp->idev, &ifp->addr);
5424 if (!ipv6_addr_any(&ifp->peer_addr)) {
5425 struct rt6_info *rt;
5427 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5428 ifp->idev->dev, 0, 0);
5433 dst_hold(&ifp->rt->dst);
5434 ip6_del_rt(ifp->rt);
5436 rt_genid_bump_ipv6(net);
5439 atomic_inc(&net->ipv6.dev_addr_genid);
5442 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5445 if (likely(ifp->idev->dead == 0))
5446 __ipv6_ifa_notify(event, ifp);
5447 rcu_read_unlock_bh();
5450 #ifdef CONFIG_SYSCTL
5453 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5454 void __user *buffer, size_t *lenp, loff_t *ppos)
5456 int *valp = ctl->data;
5459 struct ctl_table lctl;
5463 * ctl->data points to idev->cnf.forwarding, we should
5464 * not modify it until we get the rtnl lock.
5469 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5472 ret = addrconf_fixup_forwarding(ctl, valp, val);
5479 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5480 void __user *buffer, size_t *lenp, loff_t *ppos)
5482 struct inet6_dev *idev = ctl->extra1;
5483 int min_mtu = IPV6_MIN_MTU;
5484 struct ctl_table lctl;
5487 lctl.extra1 = &min_mtu;
5488 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5490 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5493 static void dev_disable_change(struct inet6_dev *idev)
5495 struct netdev_notifier_info info;
5497 if (!idev || !idev->dev)
5500 netdev_notifier_info_init(&info, idev->dev);
5501 if (idev->cnf.disable_ipv6)
5502 addrconf_notify(NULL, NETDEV_DOWN, &info);
5504 addrconf_notify(NULL, NETDEV_UP, &info);
5507 static void addrconf_disable_change(struct net *net, __s32 newf)
5509 struct net_device *dev;
5510 struct inet6_dev *idev;
5513 for_each_netdev_rcu(net, dev) {
5514 idev = __in6_dev_get(dev);
5516 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5517 idev->cnf.disable_ipv6 = newf;
5519 dev_disable_change(idev);
5525 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5530 if (!rtnl_trylock())
5531 return restart_syscall();
5533 net = (struct net *)table->extra2;
5537 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5542 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5543 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5544 addrconf_disable_change(net, newf);
5545 } else if ((!newf) ^ (!old))
5546 dev_disable_change((struct inet6_dev *)table->extra1);
5553 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5554 void __user *buffer, size_t *lenp, loff_t *ppos)
5556 int *valp = ctl->data;
5559 struct ctl_table lctl;
5563 * ctl->data points to idev->cnf.disable_ipv6, we should
5564 * not modify it until we get the rtnl lock.
5569 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5572 ret = addrconf_disable_ipv6(ctl, valp, val);
5579 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5580 void __user *buffer, size_t *lenp, loff_t *ppos)
5582 int *valp = ctl->data;
5587 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5590 if (write && old != new) {
5591 struct net *net = ctl->extra2;
5593 if (!rtnl_trylock())
5594 return restart_syscall();
5596 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5597 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5598 NETCONFA_IFINDEX_DEFAULT,
5599 net->ipv6.devconf_dflt);
5600 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5601 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5602 NETCONFA_IFINDEX_ALL,
5603 net->ipv6.devconf_all);
5605 struct inet6_dev *idev = ctl->extra1;
5607 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5617 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5618 void __user *buffer, size_t *lenp,
5622 struct in6_addr addr;
5623 char str[IPV6_MAX_STRLEN];
5624 struct ctl_table lctl = *ctl;
5625 struct net *net = ctl->extra2;
5626 struct ipv6_stable_secret *secret = ctl->data;
5628 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5631 lctl.maxlen = IPV6_MAX_STRLEN;
5634 if (!rtnl_trylock())
5635 return restart_syscall();
5637 if (!write && !secret->initialized) {
5642 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5643 if (err >= sizeof(str)) {
5648 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5652 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5657 secret->initialized = true;
5658 secret->secret = addr;
5660 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5661 struct net_device *dev;
5663 for_each_netdev(net, dev) {
5664 struct inet6_dev *idev = __in6_dev_get(dev);
5667 idev->addr_gen_mode =
5668 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5672 struct inet6_dev *idev = ctl->extra1;
5674 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5684 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5686 void __user *buffer,
5690 int *valp = ctl->data;
5693 struct ctl_table lctl;
5696 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5697 * we should not modify it until we get the rtnl lock.
5702 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5705 ret = addrconf_fixup_linkdown(ctl, valp, val);
5711 static int minus_one = -1;
5712 static const int one = 1;
5713 static const int two_five_five = 255;
5715 static const struct ctl_table addrconf_sysctl[] = {
5717 .procname = "forwarding",
5718 .data = &ipv6_devconf.forwarding,
5719 .maxlen = sizeof(int),
5721 .proc_handler = addrconf_sysctl_forward,
5724 .procname = "hop_limit",
5725 .data = &ipv6_devconf.hop_limit,
5726 .maxlen = sizeof(int),
5728 .proc_handler = proc_dointvec_minmax,
5729 .extra1 = (void *)&one,
5730 .extra2 = (void *)&two_five_five,
5734 .data = &ipv6_devconf.mtu6,
5735 .maxlen = sizeof(int),
5737 .proc_handler = addrconf_sysctl_mtu,
5740 .procname = "accept_ra",
5741 .data = &ipv6_devconf.accept_ra,
5742 .maxlen = sizeof(int),
5744 .proc_handler = proc_dointvec,
5747 .procname = "accept_redirects",
5748 .data = &ipv6_devconf.accept_redirects,
5749 .maxlen = sizeof(int),
5751 .proc_handler = proc_dointvec,
5754 .procname = "autoconf",
5755 .data = &ipv6_devconf.autoconf,
5756 .maxlen = sizeof(int),
5758 .proc_handler = proc_dointvec,
5761 .procname = "dad_transmits",
5762 .data = &ipv6_devconf.dad_transmits,
5763 .maxlen = sizeof(int),
5765 .proc_handler = proc_dointvec,
5768 .procname = "router_solicitations",
5769 .data = &ipv6_devconf.rtr_solicits,
5770 .maxlen = sizeof(int),
5772 .proc_handler = proc_dointvec_minmax,
5773 .extra1 = &minus_one,
5776 .procname = "router_solicitation_interval",
5777 .data = &ipv6_devconf.rtr_solicit_interval,
5778 .maxlen = sizeof(int),
5780 .proc_handler = proc_dointvec_jiffies,
5783 .procname = "router_solicitation_max_interval",
5784 .data = &ipv6_devconf.rtr_solicit_max_interval,
5785 .maxlen = sizeof(int),
5787 .proc_handler = proc_dointvec_jiffies,
5790 .procname = "router_solicitation_delay",
5791 .data = &ipv6_devconf.rtr_solicit_delay,
5792 .maxlen = sizeof(int),
5794 .proc_handler = proc_dointvec_jiffies,
5797 .procname = "force_mld_version",
5798 .data = &ipv6_devconf.force_mld_version,
5799 .maxlen = sizeof(int),
5801 .proc_handler = proc_dointvec,
5804 .procname = "mldv1_unsolicited_report_interval",
5806 &ipv6_devconf.mldv1_unsolicited_report_interval,
5807 .maxlen = sizeof(int),
5809 .proc_handler = proc_dointvec_ms_jiffies,
5812 .procname = "mldv2_unsolicited_report_interval",
5814 &ipv6_devconf.mldv2_unsolicited_report_interval,
5815 .maxlen = sizeof(int),
5817 .proc_handler = proc_dointvec_ms_jiffies,
5820 .procname = "use_tempaddr",
5821 .data = &ipv6_devconf.use_tempaddr,
5822 .maxlen = sizeof(int),
5824 .proc_handler = proc_dointvec,
5827 .procname = "temp_valid_lft",
5828 .data = &ipv6_devconf.temp_valid_lft,
5829 .maxlen = sizeof(int),
5831 .proc_handler = proc_dointvec,
5834 .procname = "temp_prefered_lft",
5835 .data = &ipv6_devconf.temp_prefered_lft,
5836 .maxlen = sizeof(int),
5838 .proc_handler = proc_dointvec,
5841 .procname = "regen_max_retry",
5842 .data = &ipv6_devconf.regen_max_retry,
5843 .maxlen = sizeof(int),
5845 .proc_handler = proc_dointvec,
5848 .procname = "max_desync_factor",
5849 .data = &ipv6_devconf.max_desync_factor,
5850 .maxlen = sizeof(int),
5852 .proc_handler = proc_dointvec,
5855 .procname = "max_addresses",
5856 .data = &ipv6_devconf.max_addresses,
5857 .maxlen = sizeof(int),
5859 .proc_handler = proc_dointvec,
5862 .procname = "accept_ra_defrtr",
5863 .data = &ipv6_devconf.accept_ra_defrtr,
5864 .maxlen = sizeof(int),
5866 .proc_handler = proc_dointvec,
5869 .procname = "accept_ra_min_hop_limit",
5870 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5871 .maxlen = sizeof(int),
5873 .proc_handler = proc_dointvec,
5876 .procname = "accept_ra_pinfo",
5877 .data = &ipv6_devconf.accept_ra_pinfo,
5878 .maxlen = sizeof(int),
5880 .proc_handler = proc_dointvec,
5882 #ifdef CONFIG_IPV6_ROUTER_PREF
5884 .procname = "accept_ra_rtr_pref",
5885 .data = &ipv6_devconf.accept_ra_rtr_pref,
5886 .maxlen = sizeof(int),
5888 .proc_handler = proc_dointvec,
5891 .procname = "router_probe_interval",
5892 .data = &ipv6_devconf.rtr_probe_interval,
5893 .maxlen = sizeof(int),
5895 .proc_handler = proc_dointvec_jiffies,
5897 #ifdef CONFIG_IPV6_ROUTE_INFO
5899 .procname = "accept_ra_rt_info_max_plen",
5900 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5901 .maxlen = sizeof(int),
5903 .proc_handler = proc_dointvec,
5908 .procname = "proxy_ndp",
5909 .data = &ipv6_devconf.proxy_ndp,
5910 .maxlen = sizeof(int),
5912 .proc_handler = addrconf_sysctl_proxy_ndp,
5915 .procname = "accept_source_route",
5916 .data = &ipv6_devconf.accept_source_route,
5917 .maxlen = sizeof(int),
5919 .proc_handler = proc_dointvec,
5921 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5923 .procname = "optimistic_dad",
5924 .data = &ipv6_devconf.optimistic_dad,
5925 .maxlen = sizeof(int),
5927 .proc_handler = proc_dointvec,
5930 .procname = "use_optimistic",
5931 .data = &ipv6_devconf.use_optimistic,
5932 .maxlen = sizeof(int),
5934 .proc_handler = proc_dointvec,
5937 #ifdef CONFIG_IPV6_MROUTE
5939 .procname = "mc_forwarding",
5940 .data = &ipv6_devconf.mc_forwarding,
5941 .maxlen = sizeof(int),
5943 .proc_handler = proc_dointvec,
5947 .procname = "disable_ipv6",
5948 .data = &ipv6_devconf.disable_ipv6,
5949 .maxlen = sizeof(int),
5951 .proc_handler = addrconf_sysctl_disable,
5954 .procname = "accept_dad",
5955 .data = &ipv6_devconf.accept_dad,
5956 .maxlen = sizeof(int),
5958 .proc_handler = proc_dointvec,
5961 .procname = "force_tllao",
5962 .data = &ipv6_devconf.force_tllao,
5963 .maxlen = sizeof(int),
5965 .proc_handler = proc_dointvec
5968 .procname = "ndisc_notify",
5969 .data = &ipv6_devconf.ndisc_notify,
5970 .maxlen = sizeof(int),
5972 .proc_handler = proc_dointvec
5975 .procname = "suppress_frag_ndisc",
5976 .data = &ipv6_devconf.suppress_frag_ndisc,
5977 .maxlen = sizeof(int),
5979 .proc_handler = proc_dointvec
5982 .procname = "accept_ra_from_local",
5983 .data = &ipv6_devconf.accept_ra_from_local,
5984 .maxlen = sizeof(int),
5986 .proc_handler = proc_dointvec,
5989 .procname = "accept_ra_mtu",
5990 .data = &ipv6_devconf.accept_ra_mtu,
5991 .maxlen = sizeof(int),
5993 .proc_handler = proc_dointvec,
5996 .procname = "stable_secret",
5997 .data = &ipv6_devconf.stable_secret,
5998 .maxlen = IPV6_MAX_STRLEN,
6000 .proc_handler = addrconf_sysctl_stable_secret,
6003 .procname = "use_oif_addrs_only",
6004 .data = &ipv6_devconf.use_oif_addrs_only,
6005 .maxlen = sizeof(int),
6007 .proc_handler = proc_dointvec,
6010 .procname = "ignore_routes_with_linkdown",
6011 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6012 .maxlen = sizeof(int),
6014 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6017 .procname = "drop_unicast_in_l2_multicast",
6018 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6019 .maxlen = sizeof(int),
6021 .proc_handler = proc_dointvec,
6024 .procname = "drop_unsolicited_na",
6025 .data = &ipv6_devconf.drop_unsolicited_na,
6026 .maxlen = sizeof(int),
6028 .proc_handler = proc_dointvec,
6031 .procname = "keep_addr_on_down",
6032 .data = &ipv6_devconf.keep_addr_on_down,
6033 .maxlen = sizeof(int),
6035 .proc_handler = proc_dointvec,
6043 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6044 struct inet6_dev *idev, struct ipv6_devconf *p)
6047 struct ctl_table *table;
6048 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6050 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6054 for (i = 0; table[i].data; i++) {
6055 table[i].data += (char *)p - (char *)&ipv6_devconf;
6056 /* If one of these is already set, then it is not safe to
6057 * overwrite either of them: this makes proc_dointvec_minmax
6060 if (!table[i].extra1 && !table[i].extra2) {
6061 table[i].extra1 = idev; /* embedded; no ref */
6062 table[i].extra2 = net;
6066 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6068 p->sysctl_header = register_net_sysctl(net, path, table);
6069 if (!p->sysctl_header)
6072 if (!strcmp(dev_name, "all"))
6073 ifindex = NETCONFA_IFINDEX_ALL;
6074 else if (!strcmp(dev_name, "default"))
6075 ifindex = NETCONFA_IFINDEX_DEFAULT;
6077 ifindex = idev->dev->ifindex;
6078 inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6087 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6089 struct ctl_table *table;
6091 if (!p->sysctl_header)
6094 table = p->sysctl_header->ctl_table_arg;
6095 unregister_net_sysctl_table(p->sysctl_header);
6096 p->sysctl_header = NULL;
6100 static int addrconf_sysctl_register(struct inet6_dev *idev)
6104 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6107 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6108 &ndisc_ifinfo_sysctl_change);
6111 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6114 neigh_sysctl_unregister(idev->nd_parms);
6119 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6121 __addrconf_sysctl_unregister(&idev->cnf);
6122 neigh_sysctl_unregister(idev->nd_parms);
6128 static int __net_init addrconf_init_net(struct net *net)
6131 struct ipv6_devconf *all, *dflt;
6133 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6137 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6139 goto err_alloc_dflt;
6141 /* these will be inherited by all namespaces */
6142 dflt->autoconf = ipv6_defaults.autoconf;
6143 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6145 dflt->stable_secret.initialized = false;
6146 all->stable_secret.initialized = false;
6148 net->ipv6.devconf_all = all;
6149 net->ipv6.devconf_dflt = dflt;
6151 #ifdef CONFIG_SYSCTL
6152 err = __addrconf_sysctl_register(net, "all", NULL, all);
6156 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6162 #ifdef CONFIG_SYSCTL
6164 __addrconf_sysctl_unregister(all);
6174 static void __net_exit addrconf_exit_net(struct net *net)
6176 #ifdef CONFIG_SYSCTL
6177 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6178 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6180 kfree(net->ipv6.devconf_dflt);
6181 kfree(net->ipv6.devconf_all);
6184 static struct pernet_operations addrconf_ops = {
6185 .init = addrconf_init_net,
6186 .exit = addrconf_exit_net,
6189 static struct rtnl_af_ops inet6_ops __read_mostly = {
6191 .fill_link_af = inet6_fill_link_af,
6192 .get_link_af_size = inet6_get_link_af_size,
6193 .validate_link_af = inet6_validate_link_af,
6194 .set_link_af = inet6_set_link_af,
6198 * Init / cleanup code
6201 int __init addrconf_init(void)
6203 struct inet6_dev *idev;
6206 err = ipv6_addr_label_init();
6208 pr_crit("%s: cannot initialize default policy table: %d\n",
6213 err = register_pernet_subsys(&addrconf_ops);
6217 addrconf_wq = create_workqueue("ipv6_addrconf");
6223 /* The addrconf netdev notifier requires that loopback_dev
6224 * has it's ipv6 private information allocated and setup
6225 * before it can bring up and give link-local addresses
6226 * to other devices which are up.
6228 * Unfortunately, loopback_dev is not necessarily the first
6229 * entry in the global dev_base list of net devices. In fact,
6230 * it is likely to be the very last entry on that list.
6231 * So this causes the notifier registry below to try and
6232 * give link-local addresses to all devices besides loopback_dev
6233 * first, then loopback_dev, which cases all the non-loopback_dev
6234 * devices to fail to get a link-local address.
6236 * So, as a temporary fix, allocate the ipv6 structure for
6237 * loopback_dev first by hand.
6238 * Longer term, all of the dependencies ipv6 has upon the loopback
6239 * device and it being up should be removed.
6242 idev = ipv6_add_dev(init_net.loopback_dev);
6245 err = PTR_ERR(idev);
6249 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6250 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6252 register_netdevice_notifier(&ipv6_dev_notf);
6256 rtnl_af_register(&inet6_ops);
6258 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6263 /* Only the first call to __rtnl_register can fail */
6264 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6265 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6266 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6267 inet6_dump_ifaddr, NULL);
6268 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6269 inet6_dump_ifmcaddr, NULL);
6270 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6271 inet6_dump_ifacaddr, NULL);
6272 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6273 inet6_netconf_dump_devconf, NULL);
6275 ipv6_addr_label_rtnl_register();
6279 rtnl_af_unregister(&inet6_ops);
6280 unregister_netdevice_notifier(&ipv6_dev_notf);
6282 destroy_workqueue(addrconf_wq);
6284 unregister_pernet_subsys(&addrconf_ops);
6286 ipv6_addr_label_cleanup();
6291 void addrconf_cleanup(void)
6293 struct net_device *dev;
6296 unregister_netdevice_notifier(&ipv6_dev_notf);
6297 unregister_pernet_subsys(&addrconf_ops);
6298 ipv6_addr_label_cleanup();
6302 __rtnl_af_unregister(&inet6_ops);
6304 /* clean dev list */
6305 for_each_netdev(&init_net, dev) {
6306 if (__in6_dev_get(dev) == NULL)
6308 addrconf_ifdown(dev, 1);
6310 addrconf_ifdown(init_net.loopback_dev, 2);
6315 spin_lock_bh(&addrconf_hash_lock);
6316 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6317 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6318 spin_unlock_bh(&addrconf_hash_lock);
6319 cancel_delayed_work(&addr_chk_work);
6322 destroy_workqueue(addrconf_wq);