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, bool bump_id);
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 link is ready: is it up and is a valid qdisc available */
290 static inline bool addrconf_link_ready(const struct net_device *dev)
292 return netif_oper_up(dev) && !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,
319 if (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_link_ready(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;
992 /* No need to add the TENTATIVE flag for addresses with NODAD */
993 if (!(flags & IFA_F_NODAD))
994 ifa->flags |= IFA_F_TENTATIVE;
995 ifa->valid_lft = valid_lft;
996 ifa->prefered_lft = prefered_lft;
997 ifa->cstamp = ifa->tstamp = jiffies;
998 ifa->tokenized = false;
1007 /* Add to big hash table */
1008 hash = inet6_addr_hash(addr);
1010 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1011 spin_unlock(&addrconf_hash_lock);
1013 write_lock(&idev->lock);
1014 /* Add to inet6_dev unicast addr list. */
1015 ipv6_link_dev_addr(idev, ifa);
1017 if (ifa->flags&IFA_F_TEMPORARY) {
1018 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1023 write_unlock(&idev->lock);
1025 rcu_read_unlock_bh();
1027 if (likely(err == 0))
1028 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1036 spin_unlock(&addrconf_hash_lock);
1040 enum cleanup_prefix_rt_t {
1041 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1042 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1043 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1047 * Check, whether the prefix for ifp would still need a prefix route
1048 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1051 * 1) we don't purge prefix if address was not permanent.
1052 * prefix is managed by its own lifetime.
1053 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1054 * 3) if there are no addresses, delete prefix.
1055 * 4) if there are still other permanent address(es),
1056 * corresponding prefix is still permanent.
1057 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1058 * don't purge the prefix, assume user space is managing it.
1059 * 6) otherwise, update prefix lifetime to the
1060 * longest valid lifetime among the corresponding
1061 * addresses on the device.
1062 * Note: subsequent RA will update lifetime.
1064 static enum cleanup_prefix_rt_t
1065 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1067 struct inet6_ifaddr *ifa;
1068 struct inet6_dev *idev = ifp->idev;
1069 unsigned long lifetime;
1070 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1074 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1077 if (ifa->prefix_len != ifp->prefix_len ||
1078 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1081 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1082 return CLEANUP_PREFIX_RT_NOP;
1084 action = CLEANUP_PREFIX_RT_EXPIRE;
1086 spin_lock(&ifa->lock);
1088 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1090 * Note: Because this address is
1091 * not permanent, lifetime <
1092 * LONG_MAX / HZ here.
1094 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1095 *expires = ifa->tstamp + lifetime * HZ;
1096 spin_unlock(&ifa->lock);
1103 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1105 struct rt6_info *rt;
1107 rt = addrconf_get_prefix_route(&ifp->addr,
1110 0, RTF_GATEWAY | RTF_DEFAULT);
1115 if (!(rt->rt6i_flags & RTF_EXPIRES))
1116 rt6_set_expires(rt, expires);
1123 /* This function wants to get referenced ifp and releases it before return */
1125 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1128 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1129 unsigned long expires;
1133 spin_lock_bh(&ifp->lock);
1135 ifp->state = INET6_IFADDR_STATE_DEAD;
1136 spin_unlock_bh(&ifp->lock);
1138 if (state == INET6_IFADDR_STATE_DEAD)
1141 spin_lock_bh(&addrconf_hash_lock);
1142 hlist_del_init_rcu(&ifp->addr_lst);
1143 spin_unlock_bh(&addrconf_hash_lock);
1145 write_lock_bh(&ifp->idev->lock);
1147 if (ifp->flags&IFA_F_TEMPORARY) {
1148 list_del(&ifp->tmp_list);
1150 in6_ifa_put(ifp->ifpub);
1156 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1157 action = check_cleanup_prefix_route(ifp, &expires);
1159 list_del_init(&ifp->if_list);
1162 write_unlock_bh(&ifp->idev->lock);
1164 addrconf_del_dad_work(ifp);
1166 ipv6_ifa_notify(RTM_DELADDR, ifp);
1168 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1170 if (action != CLEANUP_PREFIX_RT_NOP) {
1171 cleanup_prefix_route(ifp, expires,
1172 action == CLEANUP_PREFIX_RT_DEL);
1175 /* clean up prefsrc entries */
1176 rt6_remove_prefsrc(ifp);
1181 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1183 struct inet6_dev *idev = ifp->idev;
1184 struct in6_addr addr, *tmpaddr;
1185 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1186 unsigned long regen_advance;
1190 unsigned long now = jiffies;
1191 long max_desync_factor;
1192 s32 cnf_temp_preferred_lft;
1194 write_lock_bh(&idev->lock);
1196 spin_lock_bh(&ift->lock);
1197 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1198 spin_unlock_bh(&ift->lock);
1205 if (idev->cnf.use_tempaddr <= 0) {
1206 write_unlock_bh(&idev->lock);
1207 pr_info("%s: use_tempaddr is disabled\n", __func__);
1212 spin_lock_bh(&ifp->lock);
1213 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1214 idev->cnf.use_tempaddr = -1; /*XXX*/
1215 spin_unlock_bh(&ifp->lock);
1216 write_unlock_bh(&idev->lock);
1217 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1224 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1225 ipv6_try_regen_rndid(idev, tmpaddr);
1226 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1227 age = (now - ifp->tstamp) / HZ;
1229 regen_advance = idev->cnf.regen_max_retry *
1230 idev->cnf.dad_transmits *
1231 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1233 /* recalculate max_desync_factor each time and update
1234 * idev->desync_factor if it's larger
1236 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1237 max_desync_factor = min_t(__u32,
1238 idev->cnf.max_desync_factor,
1239 cnf_temp_preferred_lft - regen_advance);
1241 if (unlikely(idev->desync_factor > max_desync_factor)) {
1242 if (max_desync_factor > 0) {
1243 get_random_bytes(&idev->desync_factor,
1244 sizeof(idev->desync_factor));
1245 idev->desync_factor %= max_desync_factor;
1247 idev->desync_factor = 0;
1251 tmp_valid_lft = min_t(__u32,
1253 idev->cnf.temp_valid_lft + age);
1254 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1255 idev->desync_factor;
1256 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1257 tmp_plen = ifp->prefix_len;
1258 tmp_tstamp = ifp->tstamp;
1259 spin_unlock_bh(&ifp->lock);
1261 write_unlock_bh(&idev->lock);
1263 /* A temporary address is created only if this calculated Preferred
1264 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1265 * an implementation must not create a temporary address with a zero
1266 * Preferred Lifetime.
1267 * Use age calculation as in addrconf_verify to avoid unnecessary
1268 * temporary addresses being generated.
1270 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1271 if (tmp_prefered_lft <= regen_advance + age) {
1278 addr_flags = IFA_F_TEMPORARY;
1279 /* set in addrconf_prefix_rcv() */
1280 if (ifp->flags & IFA_F_OPTIMISTIC)
1281 addr_flags |= IFA_F_OPTIMISTIC;
1283 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1284 ipv6_addr_scope(&addr), addr_flags,
1285 tmp_valid_lft, tmp_prefered_lft);
1289 pr_info("%s: retry temporary address regeneration\n", __func__);
1291 write_lock_bh(&idev->lock);
1295 spin_lock_bh(&ift->lock);
1298 ift->tstamp = tmp_tstamp;
1299 spin_unlock_bh(&ift->lock);
1301 addrconf_dad_start(ift);
1309 * Choose an appropriate source address (RFC3484)
1312 IPV6_SADDR_RULE_INIT = 0,
1313 IPV6_SADDR_RULE_LOCAL,
1314 IPV6_SADDR_RULE_SCOPE,
1315 IPV6_SADDR_RULE_PREFERRED,
1316 #ifdef CONFIG_IPV6_MIP6
1317 IPV6_SADDR_RULE_HOA,
1319 IPV6_SADDR_RULE_OIF,
1320 IPV6_SADDR_RULE_LABEL,
1321 IPV6_SADDR_RULE_PRIVACY,
1322 IPV6_SADDR_RULE_ORCHID,
1323 IPV6_SADDR_RULE_PREFIX,
1324 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1325 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1330 struct ipv6_saddr_score {
1333 struct inet6_ifaddr *ifa;
1334 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1339 struct ipv6_saddr_dst {
1340 const struct in6_addr *addr;
1347 static inline int ipv6_saddr_preferred(int type)
1349 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1354 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1356 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1357 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1363 static int ipv6_get_saddr_eval(struct net *net,
1364 struct ipv6_saddr_score *score,
1365 struct ipv6_saddr_dst *dst,
1370 if (i <= score->rule) {
1372 case IPV6_SADDR_RULE_SCOPE:
1373 ret = score->scopedist;
1375 case IPV6_SADDR_RULE_PREFIX:
1376 ret = score->matchlen;
1379 ret = !!test_bit(i, score->scorebits);
1385 case IPV6_SADDR_RULE_INIT:
1386 /* Rule 0: remember if hiscore is not ready yet */
1389 case IPV6_SADDR_RULE_LOCAL:
1390 /* Rule 1: Prefer same address */
1391 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1393 case IPV6_SADDR_RULE_SCOPE:
1394 /* Rule 2: Prefer appropriate scope
1399 * ---+--+-+---> scope
1401 * | d is scope of the destination.
1403 * | \ <- smaller scope is better if
1404 * B-15 | \ if scope is enough for destination.
1405 * | ret = B - scope (-1 <= scope >= d <= 15).
1407 * |/ <- greater is better
1408 * -C / if scope is not enough for destination.
1409 * /| ret = scope - C (-1 <= d < scope <= 15).
1411 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1412 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1413 * Assume B = 0 and we get C > 29.
1415 ret = __ipv6_addr_src_scope(score->addr_type);
1416 if (ret >= dst->scope)
1419 ret -= 128; /* 30 is enough */
1420 score->scopedist = ret;
1422 case IPV6_SADDR_RULE_PREFERRED:
1424 /* Rule 3: Avoid deprecated and optimistic addresses */
1425 u8 avoid = IFA_F_DEPRECATED;
1427 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1428 avoid |= IFA_F_OPTIMISTIC;
1429 ret = ipv6_saddr_preferred(score->addr_type) ||
1430 !(score->ifa->flags & avoid);
1433 #ifdef CONFIG_IPV6_MIP6
1434 case IPV6_SADDR_RULE_HOA:
1436 /* Rule 4: Prefer home address */
1437 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1438 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1442 case IPV6_SADDR_RULE_OIF:
1443 /* Rule 5: Prefer outgoing interface */
1444 ret = (!dst->ifindex ||
1445 dst->ifindex == score->ifa->idev->dev->ifindex);
1447 case IPV6_SADDR_RULE_LABEL:
1448 /* Rule 6: Prefer matching label */
1449 ret = ipv6_addr_label(net,
1450 &score->ifa->addr, score->addr_type,
1451 score->ifa->idev->dev->ifindex) == dst->label;
1453 case IPV6_SADDR_RULE_PRIVACY:
1455 /* Rule 7: Prefer public address
1456 * Note: prefer temporary address if use_tempaddr >= 2
1458 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1459 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1460 score->ifa->idev->cnf.use_tempaddr >= 2;
1461 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1464 case IPV6_SADDR_RULE_ORCHID:
1465 /* Rule 8-: Prefer ORCHID vs ORCHID or
1466 * non-ORCHID vs non-ORCHID
1468 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1469 ipv6_addr_orchid(dst->addr));
1471 case IPV6_SADDR_RULE_PREFIX:
1472 /* Rule 8: Use longest matching prefix */
1473 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1474 if (ret > score->ifa->prefix_len)
1475 ret = score->ifa->prefix_len;
1476 score->matchlen = ret;
1478 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1479 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1480 /* Optimistic addresses still have lower precedence than other
1481 * preferred addresses.
1483 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1491 __set_bit(i, score->scorebits);
1497 static int __ipv6_dev_get_saddr(struct net *net,
1498 struct ipv6_saddr_dst *dst,
1499 struct inet6_dev *idev,
1500 struct ipv6_saddr_score *scores,
1503 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1505 read_lock_bh(&idev->lock);
1506 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1510 * - Tentative Address (RFC2462 section 5.4)
1511 * - A tentative address is not considered
1512 * "assigned to an interface" in the traditional
1513 * sense, unless it is also flagged as optimistic.
1514 * - Candidate Source Address (section 4)
1515 * - In any case, anycast addresses, multicast
1516 * addresses, and the unspecified address MUST
1517 * NOT be included in a candidate set.
1519 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1520 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1523 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1525 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1526 score->addr_type & IPV6_ADDR_MULTICAST)) {
1527 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1533 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1535 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1536 int minihiscore, miniscore;
1538 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1539 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1541 if (minihiscore > miniscore) {
1542 if (i == IPV6_SADDR_RULE_SCOPE &&
1543 score->scopedist > 0) {
1546 * each remaining entry
1547 * has too small (not enough)
1548 * scope, because ifa entries
1549 * are sorted by their scope
1555 } else if (minihiscore < miniscore) {
1557 in6_ifa_put(hiscore->ifa);
1559 in6_ifa_hold(score->ifa);
1561 swap(hiscore, score);
1562 hiscore_idx = 1 - hiscore_idx;
1564 /* restore our iterator */
1565 score->ifa = hiscore->ifa;
1572 read_unlock_bh(&idev->lock);
1576 static int ipv6_get_saddr_master(struct net *net,
1577 const struct net_device *dst_dev,
1578 const struct net_device *master,
1579 struct ipv6_saddr_dst *dst,
1580 struct ipv6_saddr_score *scores,
1583 struct inet6_dev *idev;
1585 idev = __in6_dev_get(dst_dev);
1587 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1588 scores, hiscore_idx);
1590 idev = __in6_dev_get(master);
1592 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1593 scores, hiscore_idx);
1598 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1599 const struct in6_addr *daddr, unsigned int prefs,
1600 struct in6_addr *saddr)
1602 struct ipv6_saddr_score scores[2], *hiscore;
1603 struct ipv6_saddr_dst dst;
1604 struct inet6_dev *idev;
1605 struct net_device *dev;
1607 bool use_oif_addr = false;
1608 int hiscore_idx = 0;
1610 dst_type = __ipv6_addr_type(daddr);
1612 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1613 dst.scope = __ipv6_addr_src_scope(dst_type);
1614 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1617 scores[hiscore_idx].rule = -1;
1618 scores[hiscore_idx].ifa = NULL;
1622 /* Candidate Source Address (section 4)
1623 * - multicast and link-local destination address,
1624 * the set of candidate source address MUST only
1625 * include addresses assigned to interfaces
1626 * belonging to the same link as the outgoing
1628 * (- For site-local destination addresses, the
1629 * set of candidate source addresses MUST only
1630 * include addresses assigned to interfaces
1631 * belonging to the same site as the outgoing
1633 * - "It is RECOMMENDED that the candidate source addresses
1634 * be the set of unicast addresses assigned to the
1635 * interface that will be used to send to the destination
1636 * (the 'outgoing' interface)." (RFC 6724)
1639 idev = __in6_dev_get(dst_dev);
1640 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1641 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1642 (idev && idev->cnf.use_oif_addrs_only)) {
1643 use_oif_addr = true;
1649 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1651 const struct net_device *master;
1654 /* if dst_dev exists and is enslaved to an L3 device, then
1655 * prefer addresses from dst_dev and then the master over
1656 * any other enslaved devices in the L3 domain.
1658 master = l3mdev_master_dev_rcu(dst_dev);
1660 master_idx = master->ifindex;
1662 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1664 scores, hiscore_idx);
1666 if (scores[hiscore_idx].ifa)
1670 for_each_netdev_rcu(net, dev) {
1671 /* only consider addresses on devices in the
1674 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1676 idev = __in6_dev_get(dev);
1679 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1686 hiscore = &scores[hiscore_idx];
1688 return -EADDRNOTAVAIL;
1690 *saddr = hiscore->ifa->addr;
1691 in6_ifa_put(hiscore->ifa);
1694 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1696 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1699 struct inet6_ifaddr *ifp;
1700 int err = -EADDRNOTAVAIL;
1702 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1703 if (ifp->scope > IFA_LINK)
1705 if (ifp->scope == IFA_LINK &&
1706 !(ifp->flags & banned_flags)) {
1715 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1718 struct inet6_dev *idev;
1719 int err = -EADDRNOTAVAIL;
1722 idev = __in6_dev_get(dev);
1724 read_lock_bh(&idev->lock);
1725 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1726 read_unlock_bh(&idev->lock);
1732 static int ipv6_count_addresses(struct inet6_dev *idev)
1735 struct inet6_ifaddr *ifp;
1737 read_lock_bh(&idev->lock);
1738 list_for_each_entry(ifp, &idev->addr_list, if_list)
1740 read_unlock_bh(&idev->lock);
1744 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1745 const struct net_device *dev, int strict)
1747 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1749 EXPORT_SYMBOL(ipv6_chk_addr);
1751 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1752 const struct net_device *dev, int strict,
1755 struct inet6_ifaddr *ifp;
1756 unsigned int hash = inet6_addr_hash(addr);
1760 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1761 if (!net_eq(dev_net(ifp->idev->dev), net))
1763 /* Decouple optimistic from tentative for evaluation here.
1764 * Ban optimistic addresses explicitly, when required.
1766 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1767 ? (ifp->flags&~IFA_F_TENTATIVE)
1769 if (ipv6_addr_equal(&ifp->addr, addr) &&
1770 !(ifp_flags&banned_flags) &&
1771 (!dev || ifp->idev->dev == dev ||
1772 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1773 rcu_read_unlock_bh();
1778 rcu_read_unlock_bh();
1781 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1783 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1784 struct net_device *dev)
1786 unsigned int hash = inet6_addr_hash(addr);
1787 struct inet6_ifaddr *ifp;
1789 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1790 if (!net_eq(dev_net(ifp->idev->dev), net))
1792 if (ipv6_addr_equal(&ifp->addr, addr)) {
1793 if (!dev || ifp->idev->dev == dev)
1800 /* Compares an address/prefix_len with addresses on device @dev.
1801 * If one is found it returns true.
1803 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1804 const unsigned int prefix_len, struct net_device *dev)
1806 struct inet6_dev *idev;
1807 struct inet6_ifaddr *ifa;
1811 idev = __in6_dev_get(dev);
1813 read_lock_bh(&idev->lock);
1814 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1815 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1819 read_unlock_bh(&idev->lock);
1825 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1827 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1829 struct inet6_dev *idev;
1830 struct inet6_ifaddr *ifa;
1835 idev = __in6_dev_get(dev);
1837 read_lock_bh(&idev->lock);
1838 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1839 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1844 read_unlock_bh(&idev->lock);
1849 EXPORT_SYMBOL(ipv6_chk_prefix);
1851 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1852 struct net_device *dev, int strict)
1854 struct inet6_ifaddr *ifp, *result = NULL;
1855 unsigned int hash = inet6_addr_hash(addr);
1858 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1859 if (!net_eq(dev_net(ifp->idev->dev), net))
1861 if (ipv6_addr_equal(&ifp->addr, addr)) {
1862 if (!dev || ifp->idev->dev == dev ||
1863 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1870 rcu_read_unlock_bh();
1875 /* Gets referenced address, destroys ifaddr */
1877 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1880 ifp->flags |= IFA_F_DADFAILED;
1882 if (ifp->flags&IFA_F_TEMPORARY) {
1883 struct inet6_ifaddr *ifpub;
1884 spin_lock_bh(&ifp->lock);
1887 in6_ifa_hold(ifpub);
1888 spin_unlock_bh(&ifp->lock);
1889 ipv6_create_tempaddr(ifpub, ifp);
1892 spin_unlock_bh(&ifp->lock);
1895 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1896 spin_lock_bh(&ifp->lock);
1897 addrconf_del_dad_work(ifp);
1898 ifp->flags |= IFA_F_TENTATIVE;
1899 spin_unlock_bh(&ifp->lock);
1901 ipv6_ifa_notify(0, ifp);
1908 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1912 spin_lock_bh(&ifp->lock);
1913 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1914 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1917 spin_unlock_bh(&ifp->lock);
1922 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1924 struct inet6_dev *idev = ifp->idev;
1925 struct net *net = dev_net(ifp->idev->dev);
1927 if (addrconf_dad_end(ifp)) {
1932 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1933 ifp->idev->dev->name, &ifp->addr);
1935 spin_lock_bh(&ifp->lock);
1937 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1938 int scope = ifp->scope;
1939 u32 flags = ifp->flags;
1940 struct in6_addr new_addr;
1941 struct inet6_ifaddr *ifp2;
1942 u32 valid_lft, preferred_lft;
1943 int pfxlen = ifp->prefix_len;
1944 int retries = ifp->stable_privacy_retry + 1;
1946 if (retries > net->ipv6.sysctl.idgen_retries) {
1947 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1948 ifp->idev->dev->name);
1952 new_addr = ifp->addr;
1953 if (ipv6_generate_stable_address(&new_addr, retries,
1957 valid_lft = ifp->valid_lft;
1958 preferred_lft = ifp->prefered_lft;
1960 spin_unlock_bh(&ifp->lock);
1962 if (idev->cnf.max_addresses &&
1963 ipv6_count_addresses(idev) >=
1964 idev->cnf.max_addresses)
1967 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1968 ifp->idev->dev->name);
1970 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1971 scope, flags, valid_lft,
1976 spin_lock_bh(&ifp2->lock);
1977 ifp2->stable_privacy_retry = retries;
1978 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1979 spin_unlock_bh(&ifp2->lock);
1981 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1984 spin_lock_bh(&ifp->lock);
1988 /* transition from _POSTDAD to _ERRDAD */
1989 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1990 spin_unlock_bh(&ifp->lock);
1992 addrconf_mod_dad_work(ifp, 0);
1996 /* Join to solicited addr multicast group.
1997 * caller must hold RTNL */
1998 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2000 struct in6_addr maddr;
2002 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2005 addrconf_addr_solict_mult(addr, &maddr);
2006 ipv6_dev_mc_inc(dev, &maddr);
2009 /* caller must hold RTNL */
2010 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2012 struct in6_addr maddr;
2014 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2017 addrconf_addr_solict_mult(addr, &maddr);
2018 __ipv6_dev_mc_dec(idev, &maddr);
2021 /* caller must hold RTNL */
2022 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2024 struct in6_addr addr;
2026 if (ifp->prefix_len >= 127) /* RFC 6164 */
2028 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2029 if (ipv6_addr_any(&addr))
2031 __ipv6_dev_ac_inc(ifp->idev, &addr);
2034 /* caller must hold RTNL */
2035 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2037 struct in6_addr addr;
2039 if (ifp->prefix_len >= 127) /* RFC 6164 */
2041 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2042 if (ipv6_addr_any(&addr))
2044 __ipv6_dev_ac_dec(ifp->idev, &addr);
2047 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2049 if (dev->addr_len != EUI64_ADDR_LEN)
2051 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2056 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2058 union fwnet_hwaddr *ha;
2060 if (dev->addr_len != FWNET_ALEN)
2063 ha = (union fwnet_hwaddr *)dev->dev_addr;
2065 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2070 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2072 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2073 if (dev->addr_len != ARCNET_ALEN)
2076 eui[7] = *(u8 *)dev->dev_addr;
2080 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2082 if (dev->addr_len != INFINIBAND_ALEN)
2084 memcpy(eui, dev->dev_addr + 12, 8);
2089 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2093 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2094 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2095 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2096 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2097 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2098 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2102 memcpy(eui + 4, &addr, 4);
2106 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2108 if (dev->priv_flags & IFF_ISATAP)
2109 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2113 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2115 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2118 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2120 memcpy(eui, dev->perm_addr, 3);
2121 memcpy(eui + 5, dev->perm_addr + 3, 3);
2128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2130 switch (dev->type) {
2133 return addrconf_ifid_eui48(eui, dev);
2135 return addrconf_ifid_arcnet(eui, dev);
2136 case ARPHRD_INFINIBAND:
2137 return addrconf_ifid_infiniband(eui, dev);
2139 return addrconf_ifid_sit(eui, dev);
2141 return addrconf_ifid_gre(eui, dev);
2142 case ARPHRD_6LOWPAN:
2143 return addrconf_ifid_eui64(eui, dev);
2144 case ARPHRD_IEEE1394:
2145 return addrconf_ifid_ieee1394(eui, dev);
2146 case ARPHRD_TUNNEL6:
2147 return addrconf_ifid_ip6tnl(eui, dev);
2152 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2155 struct inet6_ifaddr *ifp;
2157 read_lock_bh(&idev->lock);
2158 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2159 if (ifp->scope > IFA_LINK)
2161 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2162 memcpy(eui, ifp->addr.s6_addr+8, 8);
2167 read_unlock_bh(&idev->lock);
2171 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2172 static void ipv6_regen_rndid(struct inet6_dev *idev)
2175 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2176 idev->rndid[0] &= ~0x02;
2179 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2180 * check if generated address is not inappropriate
2182 * - Reserved subnet anycast (RFC 2526)
2183 * 11111101 11....11 1xxxxxxx
2184 * - ISATAP (RFC4214) 6.1
2185 * 00-00-5E-FE-xx-xx-xx-xx
2187 * - XXX: already assigned to an address on the device
2189 if (idev->rndid[0] == 0xfd &&
2190 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2191 (idev->rndid[7]&0x80))
2193 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2194 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2196 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2201 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2203 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2204 ipv6_regen_rndid(idev);
2212 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2213 unsigned long expires, u32 flags)
2215 struct fib6_config cfg = {
2216 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2217 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2218 .fc_ifindex = dev->ifindex,
2219 .fc_expires = expires,
2221 .fc_flags = RTF_UP | flags,
2222 .fc_nlinfo.nl_net = dev_net(dev),
2223 .fc_protocol = RTPROT_KERNEL,
2228 /* Prevent useless cloning on PtP SIT.
2229 This thing is done here expecting that the whole
2230 class of non-broadcast devices need not cloning.
2232 #if IS_ENABLED(CONFIG_IPV6_SIT)
2233 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2234 cfg.fc_flags |= RTF_NONEXTHOP;
2237 ip6_route_add(&cfg);
2241 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2243 const struct net_device *dev,
2244 u32 flags, u32 noflags)
2246 struct fib6_node *fn;
2247 struct rt6_info *rt = NULL;
2248 struct fib6_table *table;
2249 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2251 table = fib6_get_table(dev_net(dev), tb_id);
2255 read_lock_bh(&table->tb6_lock);
2256 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2260 noflags |= RTF_CACHE;
2261 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2262 if (rt->dst.dev->ifindex != dev->ifindex)
2264 if ((rt->rt6i_flags & flags) != flags)
2266 if ((rt->rt6i_flags & noflags) != 0)
2272 read_unlock_bh(&table->tb6_lock);
2277 /* Create "default" multicast route to the interface */
2279 static void addrconf_add_mroute(struct net_device *dev)
2281 struct fib6_config cfg = {
2282 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2283 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2284 .fc_ifindex = dev->ifindex,
2287 .fc_nlinfo.nl_net = dev_net(dev),
2288 .fc_protocol = RTPROT_KERNEL,
2291 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2293 ip6_route_add(&cfg);
2296 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2298 struct inet6_dev *idev;
2302 idev = ipv6_find_idev(dev);
2304 return ERR_PTR(-ENOBUFS);
2306 if (idev->cnf.disable_ipv6)
2307 return ERR_PTR(-EACCES);
2309 /* Add default multicast route */
2310 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2311 addrconf_add_mroute(dev);
2316 static void manage_tempaddrs(struct inet6_dev *idev,
2317 struct inet6_ifaddr *ifp,
2318 __u32 valid_lft, __u32 prefered_lft,
2319 bool create, unsigned long now)
2322 struct inet6_ifaddr *ift;
2324 read_lock_bh(&idev->lock);
2325 /* update all temporary addresses in the list */
2326 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2327 int age, max_valid, max_prefered;
2329 if (ifp != ift->ifpub)
2332 /* RFC 4941 section 3.3:
2333 * If a received option will extend the lifetime of a public
2334 * address, the lifetimes of temporary addresses should
2335 * be extended, subject to the overall constraint that no
2336 * temporary addresses should ever remain "valid" or "preferred"
2337 * for a time longer than (TEMP_VALID_LIFETIME) or
2338 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2340 age = (now - ift->cstamp) / HZ;
2341 max_valid = idev->cnf.temp_valid_lft - age;
2345 max_prefered = idev->cnf.temp_prefered_lft -
2346 idev->desync_factor - age;
2347 if (max_prefered < 0)
2350 if (valid_lft > max_valid)
2351 valid_lft = max_valid;
2353 if (prefered_lft > max_prefered)
2354 prefered_lft = max_prefered;
2356 spin_lock(&ift->lock);
2358 ift->valid_lft = valid_lft;
2359 ift->prefered_lft = prefered_lft;
2361 if (prefered_lft > 0)
2362 ift->flags &= ~IFA_F_DEPRECATED;
2364 spin_unlock(&ift->lock);
2365 if (!(flags&IFA_F_TENTATIVE))
2366 ipv6_ifa_notify(0, ift);
2369 if ((create || list_empty(&idev->tempaddr_list)) &&
2370 idev->cnf.use_tempaddr > 0) {
2371 /* When a new public address is created as described
2372 * in [ADDRCONF], also create a new temporary address.
2373 * Also create a temporary address if it's enabled but
2374 * no temporary address currently exists.
2376 read_unlock_bh(&idev->lock);
2377 ipv6_create_tempaddr(ifp, NULL);
2379 read_unlock_bh(&idev->lock);
2383 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2385 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2386 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2389 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2390 const struct prefix_info *pinfo,
2391 struct inet6_dev *in6_dev,
2392 const struct in6_addr *addr, int addr_type,
2393 u32 addr_flags, bool sllao, bool tokenized,
2394 __u32 valid_lft, u32 prefered_lft)
2396 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2397 int create = 0, update_lft = 0;
2399 if (!ifp && valid_lft) {
2400 int max_addresses = in6_dev->cnf.max_addresses;
2402 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2403 if (in6_dev->cnf.optimistic_dad &&
2404 !net->ipv6.devconf_all->forwarding && sllao)
2405 addr_flags |= IFA_F_OPTIMISTIC;
2408 /* Do not allow to create too much of autoconfigured
2409 * addresses; this would be too easy way to crash kernel.
2411 if (!max_addresses ||
2412 ipv6_count_addresses(in6_dev) < max_addresses)
2413 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2415 addr_type&IPV6_ADDR_SCOPE_MASK,
2416 addr_flags, valid_lft,
2419 if (IS_ERR_OR_NULL(ifp))
2424 spin_lock_bh(&ifp->lock);
2425 ifp->flags |= IFA_F_MANAGETEMPADDR;
2426 ifp->cstamp = jiffies;
2427 ifp->tokenized = tokenized;
2428 spin_unlock_bh(&ifp->lock);
2429 addrconf_dad_start(ifp);
2437 /* update lifetime (RFC2462 5.5.3 e) */
2438 spin_lock_bh(&ifp->lock);
2440 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2441 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2444 if (!update_lft && !create && stored_lft) {
2445 const u32 minimum_lft = min_t(u32,
2446 stored_lft, MIN_VALID_LIFETIME);
2447 valid_lft = max(valid_lft, minimum_lft);
2449 /* RFC4862 Section 5.5.3e:
2450 * "Note that the preferred lifetime of the
2451 * corresponding address is always reset to
2452 * the Preferred Lifetime in the received
2453 * Prefix Information option, regardless of
2454 * whether the valid lifetime is also reset or
2457 * So we should always update prefered_lft here.
2463 ifp->valid_lft = valid_lft;
2464 ifp->prefered_lft = prefered_lft;
2467 ifp->flags &= ~IFA_F_DEPRECATED;
2468 spin_unlock_bh(&ifp->lock);
2470 if (!(flags&IFA_F_TENTATIVE))
2471 ipv6_ifa_notify(0, ifp);
2473 spin_unlock_bh(&ifp->lock);
2475 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2484 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2486 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2488 struct prefix_info *pinfo;
2493 struct inet6_dev *in6_dev;
2494 struct net *net = dev_net(dev);
2496 pinfo = (struct prefix_info *) opt;
2498 if (len < sizeof(struct prefix_info)) {
2499 ADBG("addrconf: prefix option too short\n");
2504 * Validation checks ([ADDRCONF], page 19)
2507 addr_type = ipv6_addr_type(&pinfo->prefix);
2509 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2512 valid_lft = ntohl(pinfo->valid);
2513 prefered_lft = ntohl(pinfo->prefered);
2515 if (prefered_lft > valid_lft) {
2516 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2520 in6_dev = in6_dev_get(dev);
2523 net_dbg_ratelimited("addrconf: device %s not configured\n",
2529 * Two things going on here:
2530 * 1) Add routes for on-link prefixes
2531 * 2) Configure prefixes with the auto flag set
2534 if (pinfo->onlink) {
2535 struct rt6_info *rt;
2536 unsigned long rt_expires;
2538 /* Avoid arithmetic overflow. Really, we could
2539 * save rt_expires in seconds, likely valid_lft,
2540 * but it would require division in fib gc, that it
2544 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2546 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2548 if (addrconf_finite_timeout(rt_expires))
2551 rt = addrconf_get_prefix_route(&pinfo->prefix,
2554 RTF_ADDRCONF | RTF_PREFIX_RT,
2555 RTF_GATEWAY | RTF_DEFAULT);
2558 /* Autoconf prefix route */
2559 if (valid_lft == 0) {
2562 } else if (addrconf_finite_timeout(rt_expires)) {
2564 rt6_set_expires(rt, jiffies + rt_expires);
2566 rt6_clean_expires(rt);
2568 } else if (valid_lft) {
2569 clock_t expires = 0;
2570 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2571 if (addrconf_finite_timeout(rt_expires)) {
2573 flags |= RTF_EXPIRES;
2574 expires = jiffies_to_clock_t(rt_expires);
2576 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2577 dev, expires, flags);
2582 /* Try to figure out our local address for this prefix */
2584 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2585 struct in6_addr addr;
2586 bool tokenized = false, dev_addr_generated = false;
2588 if (pinfo->prefix_len == 64) {
2589 memcpy(&addr, &pinfo->prefix, 8);
2591 if (!ipv6_addr_any(&in6_dev->token)) {
2592 read_lock_bh(&in6_dev->lock);
2593 memcpy(addr.s6_addr + 8,
2594 in6_dev->token.s6_addr + 8, 8);
2595 read_unlock_bh(&in6_dev->lock);
2597 } else if (is_addr_mode_generate_stable(in6_dev) &&
2598 !ipv6_generate_stable_address(&addr, 0,
2600 addr_flags |= IFA_F_STABLE_PRIVACY;
2602 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2603 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2606 dev_addr_generated = true;
2610 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2615 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2618 tokenized, valid_lft,
2623 /* Ignore error case here because previous prefix add addr was
2624 * successful which will be notified.
2626 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2627 addr_type, addr_flags, sllao,
2628 tokenized, valid_lft,
2630 dev_addr_generated);
2632 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2634 in6_dev_put(in6_dev);
2638 * Set destination address.
2639 * Special case for SIT interfaces where we create a new "virtual"
2642 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2644 struct in6_ifreq ireq;
2645 struct net_device *dev;
2651 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2654 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2660 #if IS_ENABLED(CONFIG_IPV6_SIT)
2661 if (dev->type == ARPHRD_SIT) {
2662 const struct net_device_ops *ops = dev->netdev_ops;
2664 struct ip_tunnel_parm p;
2666 err = -EADDRNOTAVAIL;
2667 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2670 memset(&p, 0, sizeof(p));
2671 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2675 p.iph.protocol = IPPROTO_IPV6;
2677 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2679 if (ops->ndo_do_ioctl) {
2680 mm_segment_t oldfs = get_fs();
2683 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2690 dev = __dev_get_by_name(net, p.name);
2693 err = dev_open(dev);
2703 static int ipv6_mc_config(struct sock *sk, bool join,
2704 const struct in6_addr *addr, int ifindex)
2712 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2714 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2721 * Manual configuration of address on an interface
2723 static int inet6_addr_add(struct net *net, int ifindex,
2724 const struct in6_addr *pfx,
2725 const struct in6_addr *peer_pfx,
2726 unsigned int plen, __u32 ifa_flags,
2727 __u32 prefered_lft, __u32 valid_lft)
2729 struct inet6_ifaddr *ifp;
2730 struct inet6_dev *idev;
2731 struct net_device *dev;
2732 unsigned long timeout;
2742 /* check the lifetime */
2743 if (!valid_lft || prefered_lft > valid_lft)
2746 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2749 dev = __dev_get_by_index(net, ifindex);
2753 idev = addrconf_add_dev(dev);
2755 return PTR_ERR(idev);
2757 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2758 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2759 true, pfx, ifindex);
2765 scope = ipv6_addr_scope(pfx);
2767 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2768 if (addrconf_finite_timeout(timeout)) {
2769 expires = jiffies_to_clock_t(timeout * HZ);
2770 valid_lft = timeout;
2771 flags = RTF_EXPIRES;
2775 ifa_flags |= IFA_F_PERMANENT;
2778 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2779 if (addrconf_finite_timeout(timeout)) {
2781 ifa_flags |= IFA_F_DEPRECATED;
2782 prefered_lft = timeout;
2785 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2786 valid_lft, prefered_lft);
2789 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2790 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2795 * Note that section 3.1 of RFC 4429 indicates
2796 * that the Optimistic flag should not be set for
2797 * manually configured addresses
2799 addrconf_dad_start(ifp);
2800 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2801 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2804 addrconf_verify_rtnl();
2806 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2807 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2808 false, pfx, ifindex);
2811 return PTR_ERR(ifp);
2814 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2815 const struct in6_addr *pfx, unsigned int plen)
2817 struct inet6_ifaddr *ifp;
2818 struct inet6_dev *idev;
2819 struct net_device *dev;
2824 dev = __dev_get_by_index(net, ifindex);
2828 idev = __in6_dev_get(dev);
2832 read_lock_bh(&idev->lock);
2833 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2834 if (ifp->prefix_len == plen &&
2835 ipv6_addr_equal(pfx, &ifp->addr)) {
2837 read_unlock_bh(&idev->lock);
2839 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2840 (ifa_flags & IFA_F_MANAGETEMPADDR))
2841 manage_tempaddrs(idev, ifp, 0, 0, false,
2844 addrconf_verify_rtnl();
2845 if (ipv6_addr_is_multicast(pfx)) {
2846 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2847 false, pfx, dev->ifindex);
2852 read_unlock_bh(&idev->lock);
2853 return -EADDRNOTAVAIL;
2857 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2859 struct in6_ifreq ireq;
2862 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2865 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2869 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2870 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2871 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2876 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2878 struct in6_ifreq ireq;
2881 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2884 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2888 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2889 ireq.ifr6_prefixlen);
2894 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2895 int plen, int scope)
2897 struct inet6_ifaddr *ifp;
2899 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2900 scope, IFA_F_PERMANENT,
2901 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2903 spin_lock_bh(&ifp->lock);
2904 ifp->flags &= ~IFA_F_TENTATIVE;
2905 spin_unlock_bh(&ifp->lock);
2906 rt_genid_bump_ipv6(dev_net(idev->dev));
2907 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2912 #if IS_ENABLED(CONFIG_IPV6_SIT)
2913 static void sit_add_v4_addrs(struct inet6_dev *idev)
2915 struct in6_addr addr;
2916 struct net_device *dev;
2917 struct net *net = dev_net(idev->dev);
2923 memset(&addr, 0, sizeof(struct in6_addr));
2924 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2926 if (idev->dev->flags&IFF_POINTOPOINT) {
2927 addr.s6_addr32[0] = htonl(0xfe800000);
2931 scope = IPV6_ADDR_COMPATv4;
2933 pflags |= RTF_NONEXTHOP;
2936 if (addr.s6_addr32[3]) {
2937 add_addr(idev, &addr, plen, scope);
2938 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2942 for_each_netdev(net, dev) {
2943 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2944 if (in_dev && (dev->flags & IFF_UP)) {
2945 struct in_ifaddr *ifa;
2949 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2951 addr.s6_addr32[3] = ifa->ifa_local;
2953 if (ifa->ifa_scope == RT_SCOPE_LINK)
2955 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2956 if (idev->dev->flags&IFF_POINTOPOINT)
2961 add_addr(idev, &addr, plen, flag);
2962 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2970 static void init_loopback(struct net_device *dev)
2972 struct inet6_dev *idev;
2973 struct net_device *sp_dev;
2974 struct inet6_ifaddr *sp_ifa;
2975 struct rt6_info *sp_rt;
2981 idev = ipv6_find_idev(dev);
2983 pr_debug("%s: add_dev failed\n", __func__);
2987 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2989 /* Add routes to other interface's IPv6 addresses */
2990 for_each_netdev(dev_net(dev), sp_dev) {
2991 if (!strcmp(sp_dev->name, dev->name))
2994 idev = __in6_dev_get(sp_dev);
2998 read_lock_bh(&idev->lock);
2999 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
3001 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
3005 /* This dst has been added to garbage list when
3006 * lo device down, release this obsolete dst and
3007 * reallocate a new router for ifa.
3009 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3010 ip6_rt_put(sp_ifa->rt);
3017 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3019 /* Failure cases are ignored */
3020 if (!IS_ERR(sp_rt)) {
3025 read_unlock_bh(&idev->lock);
3029 void addrconf_add_linklocal(struct inet6_dev *idev,
3030 const struct in6_addr *addr, u32 flags)
3032 struct inet6_ifaddr *ifp;
3033 u32 addr_flags = flags | IFA_F_PERMANENT;
3035 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3036 if (idev->cnf.optimistic_dad &&
3037 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3038 addr_flags |= IFA_F_OPTIMISTIC;
3041 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3042 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3044 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3045 addrconf_dad_start(ifp);
3049 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3051 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3053 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3056 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3057 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3060 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3061 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3067 static int ipv6_generate_stable_address(struct in6_addr *address,
3069 const struct inet6_dev *idev)
3071 static DEFINE_SPINLOCK(lock);
3072 static __u32 digest[SHA_DIGEST_WORDS];
3073 static __u32 workspace[SHA_WORKSPACE_WORDS];
3076 char __data[SHA_MESSAGE_BYTES];
3078 struct in6_addr secret;
3080 unsigned char hwaddr[MAX_ADDR_LEN];
3085 struct in6_addr secret;
3086 struct in6_addr temp;
3087 struct net *net = dev_net(idev->dev);
3089 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3091 if (idev->cnf.stable_secret.initialized)
3092 secret = idev->cnf.stable_secret.secret;
3093 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3094 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3099 spin_lock_bh(&lock);
3102 memset(&data, 0, sizeof(data));
3103 memset(workspace, 0, sizeof(workspace));
3104 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3105 data.prefix[0] = address->s6_addr32[0];
3106 data.prefix[1] = address->s6_addr32[1];
3107 data.secret = secret;
3108 data.dad_count = dad_count;
3110 sha_transform(digest, data.__data, workspace);
3113 temp.s6_addr32[2] = (__force __be32)digest[0];
3114 temp.s6_addr32[3] = (__force __be32)digest[1];
3116 spin_unlock_bh(&lock);
3118 if (ipv6_reserved_interfaceid(temp)) {
3120 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3129 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3131 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3135 s = &idev->cnf.stable_secret;
3136 get_random_bytes(&s->secret, sizeof(s->secret));
3137 s->initialized = true;
3140 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3142 struct in6_addr addr;
3144 /* no link local addresses on L3 master devices */
3145 if (netif_is_l3_master(idev->dev))
3148 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3150 switch (idev->addr_gen_mode) {
3151 case IN6_ADDR_GEN_MODE_RANDOM:
3152 ipv6_gen_mode_random_init(idev);
3154 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3155 if (!ipv6_generate_stable_address(&addr, 0, idev))
3156 addrconf_add_linklocal(idev, &addr,
3157 IFA_F_STABLE_PRIVACY);
3158 else if (prefix_route)
3159 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3161 case IN6_ADDR_GEN_MODE_EUI64:
3162 /* addrconf_add_linklocal also adds a prefix_route and we
3163 * only need to care about prefix routes if ipv6_generate_eui64
3164 * couldn't generate one.
3166 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3167 addrconf_add_linklocal(idev, &addr, 0);
3168 else if (prefix_route)
3169 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3171 case IN6_ADDR_GEN_MODE_NONE:
3173 /* will not add any link local address */
3178 static void addrconf_dev_config(struct net_device *dev)
3180 struct inet6_dev *idev;
3184 if ((dev->type != ARPHRD_ETHER) &&
3185 (dev->type != ARPHRD_FDDI) &&
3186 (dev->type != ARPHRD_ARCNET) &&
3187 (dev->type != ARPHRD_INFINIBAND) &&
3188 (dev->type != ARPHRD_IEEE1394) &&
3189 (dev->type != ARPHRD_TUNNEL6) &&
3190 (dev->type != ARPHRD_6LOWPAN) &&
3191 (dev->type != ARPHRD_NONE)) {
3192 /* Alas, we support only Ethernet autoconfiguration. */
3193 idev = __in6_dev_get(dev);
3194 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3195 dev->flags & IFF_MULTICAST)
3200 idev = addrconf_add_dev(dev);
3204 /* this device type has no EUI support */
3205 if (dev->type == ARPHRD_NONE &&
3206 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3207 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3209 addrconf_addr_gen(idev, false);
3212 #if IS_ENABLED(CONFIG_IPV6_SIT)
3213 static void addrconf_sit_config(struct net_device *dev)
3215 struct inet6_dev *idev;
3220 * Configure the tunnel with one of our IPv4
3221 * addresses... we should configure all of
3222 * our v4 addrs in the tunnel
3225 idev = ipv6_find_idev(dev);
3227 pr_debug("%s: add_dev failed\n", __func__);
3231 if (dev->priv_flags & IFF_ISATAP) {
3232 addrconf_addr_gen(idev, false);
3236 sit_add_v4_addrs(idev);
3238 if (dev->flags&IFF_POINTOPOINT)
3239 addrconf_add_mroute(dev);
3243 #if IS_ENABLED(CONFIG_NET_IPGRE)
3244 static void addrconf_gre_config(struct net_device *dev)
3246 struct inet6_dev *idev;
3250 idev = ipv6_find_idev(dev);
3252 pr_debug("%s: add_dev failed\n", __func__);
3256 addrconf_addr_gen(idev, true);
3257 if (dev->flags & IFF_POINTOPOINT)
3258 addrconf_add_mroute(dev);
3262 static int fixup_permanent_addr(struct inet6_dev *idev,
3263 struct inet6_ifaddr *ifp)
3265 /* rt6i_ref == 0 means the host route was removed from the
3266 * FIB, for example, if 'lo' device is taken down. In that
3267 * case regenerate the host route.
3269 if (!ifp->rt || !atomic_read(&ifp->rt->rt6i_ref)) {
3270 struct rt6_info *rt, *prev;
3272 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3273 if (unlikely(IS_ERR(rt)))
3276 /* ifp->rt can be accessed outside of rtnl */
3277 spin_lock(&ifp->lock);
3280 spin_unlock(&ifp->lock);
3285 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3286 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3290 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3291 addrconf_dad_start(ifp);
3296 static void addrconf_permanent_addr(struct net_device *dev)
3298 struct inet6_ifaddr *ifp, *tmp;
3299 struct inet6_dev *idev;
3301 idev = __in6_dev_get(dev);
3305 write_lock_bh(&idev->lock);
3307 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3308 if ((ifp->flags & IFA_F_PERMANENT) &&
3309 fixup_permanent_addr(idev, ifp) < 0) {
3310 write_unlock_bh(&idev->lock);
3313 write_lock_bh(&idev->lock);
3315 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3316 idev->dev->name, &ifp->addr);
3320 write_unlock_bh(&idev->lock);
3323 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3326 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3327 struct netdev_notifier_changeupper_info *info;
3328 struct inet6_dev *idev = __in6_dev_get(dev);
3329 struct net *net = dev_net(dev);
3330 int run_pending = 0;
3334 case NETDEV_REGISTER:
3335 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3336 idev = ipv6_add_dev(dev);
3338 return notifier_from_errno(PTR_ERR(idev));
3342 case NETDEV_CHANGEMTU:
3343 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3344 if (dev->mtu < IPV6_MIN_MTU) {
3345 addrconf_ifdown(dev, dev != net->loopback_dev);
3350 rt6_mtu_change(dev, dev->mtu);
3351 idev->cnf.mtu6 = dev->mtu;
3355 /* allocate new idev */
3356 idev = ipv6_add_dev(dev);
3360 /* device is still not ready */
3361 if (!(idev->if_flags & IF_READY))
3370 if (dev->flags & IFF_SLAVE)
3373 if (idev && idev->cnf.disable_ipv6)
3376 if (event == NETDEV_UP) {
3377 /* restore routes for permanent addresses */
3378 addrconf_permanent_addr(dev);
3380 if (!addrconf_link_ready(dev)) {
3381 /* device is not ready yet. */
3382 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3387 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3388 idev = ipv6_add_dev(dev);
3390 if (!IS_ERR_OR_NULL(idev)) {
3391 idev->if_flags |= IF_READY;
3394 } else if (event == NETDEV_CHANGE) {
3395 if (!addrconf_link_ready(dev)) {
3396 /* device is still not ready. */
3401 if (idev->if_flags & IF_READY) {
3402 /* device is already configured -
3403 * but resend MLD reports, we might
3404 * have roamed and need to update
3405 * multicast snooping switches
3410 idev->if_flags |= IF_READY;
3413 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3419 switch (dev->type) {
3420 #if IS_ENABLED(CONFIG_IPV6_SIT)
3422 addrconf_sit_config(dev);
3425 #if IS_ENABLED(CONFIG_NET_IPGRE)
3427 addrconf_gre_config(dev);
3430 case ARPHRD_LOOPBACK:
3435 addrconf_dev_config(dev);
3439 if (!IS_ERR_OR_NULL(idev)) {
3441 addrconf_dad_run(idev);
3444 * If the MTU changed during the interface down,
3445 * when the interface up, the changed MTU must be
3446 * reflected in the idev as well as routers.
3448 if (idev->cnf.mtu6 != dev->mtu &&
3449 dev->mtu >= IPV6_MIN_MTU) {
3450 rt6_mtu_change(dev, dev->mtu);
3451 idev->cnf.mtu6 = dev->mtu;
3453 idev->tstamp = jiffies;
3454 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3457 * If the changed mtu during down is lower than
3458 * IPV6_MIN_MTU stop IPv6 on this interface.
3460 if (dev->mtu < IPV6_MIN_MTU)
3461 addrconf_ifdown(dev, dev != net->loopback_dev);
3466 case NETDEV_UNREGISTER:
3468 * Remove all addresses from this interface.
3470 addrconf_ifdown(dev, event != NETDEV_DOWN);
3473 case NETDEV_CHANGENAME:
3475 snmp6_unregister_dev(idev);
3476 addrconf_sysctl_unregister(idev);
3477 err = addrconf_sysctl_register(idev);
3479 return notifier_from_errno(err);
3480 err = snmp6_register_dev(idev);
3482 addrconf_sysctl_unregister(idev);
3483 return notifier_from_errno(err);
3488 case NETDEV_PRE_TYPE_CHANGE:
3489 case NETDEV_POST_TYPE_CHANGE:
3491 addrconf_type_change(dev, event);
3494 case NETDEV_CHANGEUPPER:
3497 /* flush all routes if dev is linked to or unlinked from
3498 * an L3 master device (e.g., VRF)
3500 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3501 addrconf_ifdown(dev, 0);
3508 * addrconf module should be notified of a device going up
3510 static struct notifier_block ipv6_dev_notf = {
3511 .notifier_call = addrconf_notify,
3512 .priority = ADDRCONF_NOTIFY_PRIORITY,
3515 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3517 struct inet6_dev *idev;
3520 idev = __in6_dev_get(dev);
3522 if (event == NETDEV_POST_TYPE_CHANGE)
3523 ipv6_mc_remap(idev);
3524 else if (event == NETDEV_PRE_TYPE_CHANGE)
3525 ipv6_mc_unmap(idev);
3528 static bool addr_is_local(const struct in6_addr *addr)
3530 return ipv6_addr_type(addr) &
3531 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3534 static int addrconf_ifdown(struct net_device *dev, int how)
3536 struct net *net = dev_net(dev);
3537 struct inet6_dev *idev;
3538 struct inet6_ifaddr *ifa, *tmp;
3539 struct list_head del_list;
3546 rt6_ifdown(net, dev);
3547 neigh_ifdown(&nd_tbl, dev);
3549 idev = __in6_dev_get(dev);
3554 * Step 1: remove reference to ipv6 device from parent device.
3560 /* protected by rtnl_lock */
3561 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3563 /* Step 1.5: remove snmp6 entry */
3564 snmp6_unregister_dev(idev);
3568 /* aggregate the system setting and interface setting */
3569 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3571 _keep_addr = idev->cnf.keep_addr_on_down;
3573 /* combine the user config with event to determine if permanent
3574 * addresses are to be removed from address hash table
3576 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3578 /* Step 2: clear hash table */
3579 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3580 struct hlist_head *h = &inet6_addr_lst[i];
3582 spin_lock_bh(&addrconf_hash_lock);
3584 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3585 if (ifa->idev == idev) {
3586 addrconf_del_dad_work(ifa);
3587 /* combined flag + permanent flag decide if
3588 * address is retained on a down event
3591 !(ifa->flags & IFA_F_PERMANENT) ||
3592 addr_is_local(&ifa->addr)) {
3593 hlist_del_init_rcu(&ifa->addr_lst);
3598 spin_unlock_bh(&addrconf_hash_lock);
3601 write_lock_bh(&idev->lock);
3603 addrconf_del_rs_timer(idev);
3605 /* Step 2: clear flags for stateless addrconf */
3607 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3609 /* Step 3: clear tempaddr list */
3610 while (!list_empty(&idev->tempaddr_list)) {
3611 ifa = list_first_entry(&idev->tempaddr_list,
3612 struct inet6_ifaddr, tmp_list);
3613 list_del(&ifa->tmp_list);
3614 write_unlock_bh(&idev->lock);
3615 spin_lock_bh(&ifa->lock);
3618 in6_ifa_put(ifa->ifpub);
3621 spin_unlock_bh(&ifa->lock);
3623 write_lock_bh(&idev->lock);
3626 /* re-combine the user config with event to determine if permanent
3627 * addresses are to be removed from the interface list
3629 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3631 INIT_LIST_HEAD(&del_list);
3632 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3633 struct rt6_info *rt = NULL;
3636 addrconf_del_dad_work(ifa);
3638 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3639 !addr_is_local(&ifa->addr);
3641 list_move(&ifa->if_list, &del_list);
3643 write_unlock_bh(&idev->lock);
3644 spin_lock_bh(&ifa->lock);
3647 /* set state to skip the notifier below */
3648 state = INET6_IFADDR_STATE_DEAD;
3649 ifa->state = INET6_IFADDR_STATE_PREDAD;
3650 if (!(ifa->flags & IFA_F_NODAD))
3651 ifa->flags |= IFA_F_TENTATIVE;
3657 ifa->state = INET6_IFADDR_STATE_DEAD;
3660 spin_unlock_bh(&ifa->lock);
3665 if (state != INET6_IFADDR_STATE_DEAD) {
3666 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3667 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3669 if (idev->cnf.forwarding)
3670 addrconf_leave_anycast(ifa);
3671 addrconf_leave_solict(ifa->idev, &ifa->addr);
3674 write_lock_bh(&idev->lock);
3677 write_unlock_bh(&idev->lock);
3679 /* now clean up addresses to be removed */
3680 while (!list_empty(&del_list)) {
3681 ifa = list_first_entry(&del_list,
3682 struct inet6_ifaddr, if_list);
3683 list_del(&ifa->if_list);
3688 /* Step 5: Discard anycast and multicast list */
3690 ipv6_ac_destroy_dev(idev);
3691 ipv6_mc_destroy_dev(idev);
3696 idev->tstamp = jiffies;
3698 /* Last: Shot the device (if unregistered) */
3700 addrconf_sysctl_unregister(idev);
3701 neigh_parms_release(&nd_tbl, idev->nd_parms);
3702 neigh_ifdown(&nd_tbl, dev);
3708 static void addrconf_rs_timer(unsigned long data)
3710 struct inet6_dev *idev = (struct inet6_dev *)data;
3711 struct net_device *dev = idev->dev;
3712 struct in6_addr lladdr;
3714 write_lock(&idev->lock);
3715 if (idev->dead || !(idev->if_flags & IF_READY))
3718 if (!ipv6_accept_ra(idev))
3721 /* Announcement received after solicitation was sent */
3722 if (idev->if_flags & IF_RA_RCVD)
3725 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3726 write_unlock(&idev->lock);
3727 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3728 ndisc_send_rs(dev, &lladdr,
3729 &in6addr_linklocal_allrouters);
3733 write_lock(&idev->lock);
3734 idev->rs_interval = rfc3315_s14_backoff_update(
3735 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3736 /* The wait after the last probe can be shorter */
3737 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3738 idev->cnf.rtr_solicits) ?
3739 idev->cnf.rtr_solicit_delay :
3743 * Note: we do not support deprecated "all on-link"
3744 * assumption any longer.
3746 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3750 write_unlock(&idev->lock);
3756 * Duplicate Address Detection
3758 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3760 unsigned long rand_num;
3761 struct inet6_dev *idev = ifp->idev;
3763 if (ifp->flags & IFA_F_OPTIMISTIC)
3766 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3768 ifp->dad_probes = idev->cnf.dad_transmits;
3769 addrconf_mod_dad_work(ifp, rand_num);
3772 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3774 struct inet6_dev *idev = ifp->idev;
3775 struct net_device *dev = idev->dev;
3776 bool bump_id, notify = false;
3778 addrconf_join_solict(dev, &ifp->addr);
3780 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3782 read_lock_bh(&idev->lock);
3783 spin_lock(&ifp->lock);
3784 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3787 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3788 idev->cnf.accept_dad < 1 ||
3789 !(ifp->flags&IFA_F_TENTATIVE) ||
3790 ifp->flags & IFA_F_NODAD) {
3791 bump_id = ifp->flags & IFA_F_TENTATIVE;
3792 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3793 spin_unlock(&ifp->lock);
3794 read_unlock_bh(&idev->lock);
3796 addrconf_dad_completed(ifp, bump_id);
3800 if (!(idev->if_flags & IF_READY)) {
3801 spin_unlock(&ifp->lock);
3802 read_unlock_bh(&idev->lock);
3804 * If the device is not ready:
3805 * - keep it tentative if it is a permanent address.
3806 * - otherwise, kill it.
3809 addrconf_dad_stop(ifp, 0);
3814 * Optimistic nodes can start receiving
3817 if (ifp->flags & IFA_F_OPTIMISTIC) {
3818 ip6_ins_rt(ifp->rt);
3819 if (ipv6_use_optimistic_addr(idev)) {
3820 /* Because optimistic nodes can use this address,
3821 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3827 addrconf_dad_kick(ifp);
3829 spin_unlock(&ifp->lock);
3830 read_unlock_bh(&idev->lock);
3832 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3835 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3837 bool begin_dad = false;
3839 spin_lock_bh(&ifp->lock);
3840 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3841 ifp->state = INET6_IFADDR_STATE_PREDAD;
3844 spin_unlock_bh(&ifp->lock);
3847 addrconf_mod_dad_work(ifp, 0);
3850 static void addrconf_dad_work(struct work_struct *w)
3852 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3853 struct inet6_ifaddr,
3855 struct inet6_dev *idev = ifp->idev;
3856 bool bump_id, disable_ipv6 = false;
3857 struct in6_addr mcaddr;
3863 } action = DAD_PROCESS;
3867 spin_lock_bh(&ifp->lock);
3868 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3870 ifp->state = INET6_IFADDR_STATE_DAD;
3871 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3873 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3875 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3876 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3877 struct in6_addr addr;
3879 addr.s6_addr32[0] = htonl(0xfe800000);
3880 addr.s6_addr32[1] = 0;
3882 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3883 ipv6_addr_equal(&ifp->addr, &addr)) {
3884 /* DAD failed for link-local based on MAC */
3885 idev->cnf.disable_ipv6 = 1;
3887 pr_info("%s: IPv6 being disabled!\n",
3888 ifp->idev->dev->name);
3889 disable_ipv6 = true;
3893 spin_unlock_bh(&ifp->lock);
3895 if (action == DAD_BEGIN) {
3896 addrconf_dad_begin(ifp);
3898 } else if (action == DAD_ABORT) {
3900 addrconf_dad_stop(ifp, 1);
3902 addrconf_ifdown(idev->dev, 0);
3906 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3909 write_lock_bh(&idev->lock);
3910 if (idev->dead || !(idev->if_flags & IF_READY)) {
3911 write_unlock_bh(&idev->lock);
3915 spin_lock(&ifp->lock);
3916 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3917 spin_unlock(&ifp->lock);
3918 write_unlock_bh(&idev->lock);
3922 if (ifp->dad_probes == 0) {
3924 * DAD was successful
3927 bump_id = ifp->flags & IFA_F_TENTATIVE;
3928 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3929 spin_unlock(&ifp->lock);
3930 write_unlock_bh(&idev->lock);
3932 addrconf_dad_completed(ifp, bump_id);
3938 addrconf_mod_dad_work(ifp,
3939 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3940 spin_unlock(&ifp->lock);
3941 write_unlock_bh(&idev->lock);
3943 /* send a neighbour solicitation for our addr */
3944 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3945 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3951 /* ifp->idev must be at least read locked */
3952 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3954 struct inet6_ifaddr *ifpiter;
3955 struct inet6_dev *idev = ifp->idev;
3957 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3958 if (ifpiter->scope > IFA_LINK)
3960 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3961 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3962 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3969 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
3971 struct net_device *dev = ifp->idev->dev;
3972 struct in6_addr lladdr;
3973 bool send_rs, send_mld;
3975 addrconf_del_dad_work(ifp);
3978 * Configure the address for reception. Now it is valid.
3981 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3983 /* If added prefix is link local and we are prepared to process
3984 router advertisements, start sending router solicitations.
3987 read_lock_bh(&ifp->idev->lock);
3988 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3989 send_rs = send_mld &&
3990 ipv6_accept_ra(ifp->idev) &&
3991 ifp->idev->cnf.rtr_solicits != 0 &&
3992 (dev->flags&IFF_LOOPBACK) == 0;
3993 read_unlock_bh(&ifp->idev->lock);
3995 /* While dad is in progress mld report's source address is in6_addrany.
3996 * Resend with proper ll now.
3999 ipv6_mc_dad_complete(ifp->idev);
4003 * If a host as already performed a random delay
4004 * [...] as part of DAD [...] there is no need
4005 * to delay again before sending the first RS
4007 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4009 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4011 write_lock_bh(&ifp->idev->lock);
4012 spin_lock(&ifp->lock);
4013 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4014 ifp->idev->cnf.rtr_solicit_interval);
4015 ifp->idev->rs_probes = 1;
4016 ifp->idev->if_flags |= IF_RS_SENT;
4017 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4018 spin_unlock(&ifp->lock);
4019 write_unlock_bh(&ifp->idev->lock);
4023 rt_genid_bump_ipv6(dev_net(dev));
4025 /* Make sure that a new temporary address will be created
4026 * before this temporary address becomes deprecated.
4028 if (ifp->flags & IFA_F_TEMPORARY)
4029 addrconf_verify_rtnl();
4032 static void addrconf_dad_run(struct inet6_dev *idev)
4034 struct inet6_ifaddr *ifp;
4036 read_lock_bh(&idev->lock);
4037 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4038 spin_lock(&ifp->lock);
4039 if (ifp->flags & IFA_F_TENTATIVE &&
4040 ifp->state == INET6_IFADDR_STATE_DAD)
4041 addrconf_dad_kick(ifp);
4042 spin_unlock(&ifp->lock);
4044 read_unlock_bh(&idev->lock);
4047 #ifdef CONFIG_PROC_FS
4048 struct if6_iter_state {
4049 struct seq_net_private p;
4054 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4056 struct inet6_ifaddr *ifa = NULL;
4057 struct if6_iter_state *state = seq->private;
4058 struct net *net = seq_file_net(seq);
4061 /* initial bucket if pos is 0 */
4067 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4068 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4070 if (!net_eq(dev_net(ifa->idev->dev), net))
4072 /* sync with offset */
4073 if (p < state->offset) {
4080 /* prepare for next bucket */
4087 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4088 struct inet6_ifaddr *ifa)
4090 struct if6_iter_state *state = seq->private;
4091 struct net *net = seq_file_net(seq);
4093 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4094 if (!net_eq(dev_net(ifa->idev->dev), net))
4101 while (++state->bucket < IN6_ADDR_HSIZE) {
4102 hlist_for_each_entry_rcu_bh(ifa,
4103 &inet6_addr_lst[state->bucket], addr_lst) {
4104 if (!net_eq(dev_net(ifa->idev->dev), net))
4113 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4117 return if6_get_first(seq, *pos);
4120 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4122 struct inet6_ifaddr *ifa;
4124 ifa = if6_get_next(seq, v);
4129 static void if6_seq_stop(struct seq_file *seq, void *v)
4132 rcu_read_unlock_bh();
4135 static int if6_seq_show(struct seq_file *seq, void *v)
4137 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4138 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4140 ifp->idev->dev->ifindex,
4144 ifp->idev->dev->name);
4148 static const struct seq_operations if6_seq_ops = {
4149 .start = if6_seq_start,
4150 .next = if6_seq_next,
4151 .show = if6_seq_show,
4152 .stop = if6_seq_stop,
4155 static int if6_seq_open(struct inode *inode, struct file *file)
4157 return seq_open_net(inode, file, &if6_seq_ops,
4158 sizeof(struct if6_iter_state));
4161 static const struct file_operations if6_fops = {
4162 .owner = THIS_MODULE,
4163 .open = if6_seq_open,
4165 .llseek = seq_lseek,
4166 .release = seq_release_net,
4169 static int __net_init if6_proc_net_init(struct net *net)
4171 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4176 static void __net_exit if6_proc_net_exit(struct net *net)
4178 remove_proc_entry("if_inet6", net->proc_net);
4181 static struct pernet_operations if6_proc_net_ops = {
4182 .init = if6_proc_net_init,
4183 .exit = if6_proc_net_exit,
4186 int __init if6_proc_init(void)
4188 return register_pernet_subsys(&if6_proc_net_ops);
4191 void if6_proc_exit(void)
4193 unregister_pernet_subsys(&if6_proc_net_ops);
4195 #endif /* CONFIG_PROC_FS */
4197 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4198 /* Check if address is a home address configured on any interface. */
4199 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4202 struct inet6_ifaddr *ifp = NULL;
4203 unsigned int hash = inet6_addr_hash(addr);
4206 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4207 if (!net_eq(dev_net(ifp->idev->dev), net))
4209 if (ipv6_addr_equal(&ifp->addr, addr) &&
4210 (ifp->flags & IFA_F_HOMEADDRESS)) {
4215 rcu_read_unlock_bh();
4221 * Periodic address status verification
4224 static void addrconf_verify_rtnl(void)
4226 unsigned long now, next, next_sec, next_sched;
4227 struct inet6_ifaddr *ifp;
4234 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4236 cancel_delayed_work(&addr_chk_work);
4238 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4240 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4243 /* When setting preferred_lft to a value not zero or
4244 * infinity, while valid_lft is infinity
4245 * IFA_F_PERMANENT has a non-infinity life time.
4247 if ((ifp->flags & IFA_F_PERMANENT) &&
4248 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4251 spin_lock(&ifp->lock);
4252 /* We try to batch several events at once. */
4253 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4255 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4256 age >= ifp->valid_lft) {
4257 spin_unlock(&ifp->lock);
4261 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4262 spin_unlock(&ifp->lock);
4264 } else if (age >= ifp->prefered_lft) {
4265 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4268 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4270 ifp->flags |= IFA_F_DEPRECATED;
4273 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4274 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4275 next = ifp->tstamp + ifp->valid_lft * HZ;
4277 spin_unlock(&ifp->lock);
4282 ipv6_ifa_notify(0, ifp);
4286 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4287 !(ifp->flags&IFA_F_TENTATIVE)) {
4288 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4289 ifp->idev->cnf.dad_transmits *
4290 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4292 if (age >= ifp->prefered_lft - regen_advance) {
4293 struct inet6_ifaddr *ifpub = ifp->ifpub;
4294 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4295 next = ifp->tstamp + ifp->prefered_lft * HZ;
4296 if (!ifp->regen_count && ifpub) {
4299 in6_ifa_hold(ifpub);
4300 spin_unlock(&ifp->lock);
4302 spin_lock(&ifpub->lock);
4303 ifpub->regen_count = 0;
4304 spin_unlock(&ifpub->lock);
4305 ipv6_create_tempaddr(ifpub, ifp);
4310 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4311 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4312 spin_unlock(&ifp->lock);
4314 /* ifp->prefered_lft <= ifp->valid_lft */
4315 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4316 next = ifp->tstamp + ifp->prefered_lft * HZ;
4317 spin_unlock(&ifp->lock);
4322 next_sec = round_jiffies_up(next);
4325 /* If rounded timeout is accurate enough, accept it. */
4326 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4327 next_sched = next_sec;
4329 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4330 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4331 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4333 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4334 now, next, next_sec, next_sched);
4335 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4336 rcu_read_unlock_bh();
4339 static void addrconf_verify_work(struct work_struct *w)
4342 addrconf_verify_rtnl();
4346 static void addrconf_verify(void)
4348 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4351 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4352 struct in6_addr **peer_pfx)
4354 struct in6_addr *pfx = NULL;
4359 pfx = nla_data(addr);
4362 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4364 pfx = nla_data(local);
4370 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4371 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4372 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4373 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4374 [IFA_FLAGS] = { .len = sizeof(u32) },
4378 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4380 struct net *net = sock_net(skb->sk);
4381 struct ifaddrmsg *ifm;
4382 struct nlattr *tb[IFA_MAX+1];
4383 struct in6_addr *pfx, *peer_pfx;
4387 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4391 ifm = nlmsg_data(nlh);
4392 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4396 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4398 /* We ignore other flags so far. */
4399 ifa_flags &= IFA_F_MANAGETEMPADDR;
4401 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4402 ifm->ifa_prefixlen);
4405 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4406 u32 prefered_lft, u32 valid_lft)
4410 unsigned long timeout;
4411 bool was_managetempaddr;
4412 bool had_prefixroute;
4416 if (!valid_lft || (prefered_lft > valid_lft))
4419 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4420 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4423 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4424 if (addrconf_finite_timeout(timeout)) {
4425 expires = jiffies_to_clock_t(timeout * HZ);
4426 valid_lft = timeout;
4427 flags = RTF_EXPIRES;
4431 ifa_flags |= IFA_F_PERMANENT;
4434 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4435 if (addrconf_finite_timeout(timeout)) {
4437 ifa_flags |= IFA_F_DEPRECATED;
4438 prefered_lft = timeout;
4441 spin_lock_bh(&ifp->lock);
4442 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4443 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4444 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4445 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4446 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4447 IFA_F_NOPREFIXROUTE);
4448 ifp->flags |= ifa_flags;
4449 ifp->tstamp = jiffies;
4450 ifp->valid_lft = valid_lft;
4451 ifp->prefered_lft = prefered_lft;
4453 spin_unlock_bh(&ifp->lock);
4454 if (!(ifp->flags&IFA_F_TENTATIVE))
4455 ipv6_ifa_notify(0, ifp);
4457 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4458 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4460 } else if (had_prefixroute) {
4461 enum cleanup_prefix_rt_t action;
4462 unsigned long rt_expires;
4464 write_lock_bh(&ifp->idev->lock);
4465 action = check_cleanup_prefix_route(ifp, &rt_expires);
4466 write_unlock_bh(&ifp->idev->lock);
4468 if (action != CLEANUP_PREFIX_RT_NOP) {
4469 cleanup_prefix_route(ifp, rt_expires,
4470 action == CLEANUP_PREFIX_RT_DEL);
4474 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4475 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4476 valid_lft = prefered_lft = 0;
4477 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4478 !was_managetempaddr, jiffies);
4481 addrconf_verify_rtnl();
4487 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4489 struct net *net = sock_net(skb->sk);
4490 struct ifaddrmsg *ifm;
4491 struct nlattr *tb[IFA_MAX+1];
4492 struct in6_addr *pfx, *peer_pfx;
4493 struct inet6_ifaddr *ifa;
4494 struct net_device *dev;
4495 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4499 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4503 ifm = nlmsg_data(nlh);
4504 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4508 if (tb[IFA_CACHEINFO]) {
4509 struct ifa_cacheinfo *ci;
4511 ci = nla_data(tb[IFA_CACHEINFO]);
4512 valid_lft = ci->ifa_valid;
4513 preferred_lft = ci->ifa_prefered;
4515 preferred_lft = INFINITY_LIFE_TIME;
4516 valid_lft = INFINITY_LIFE_TIME;
4519 dev = __dev_get_by_index(net, ifm->ifa_index);
4523 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4525 /* We ignore other flags so far. */
4526 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4527 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4529 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4532 * It would be best to check for !NLM_F_CREATE here but
4533 * userspace already relies on not having to provide this.
4535 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4536 ifm->ifa_prefixlen, ifa_flags,
4537 preferred_lft, valid_lft);
4540 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4541 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4544 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4551 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4552 u8 scope, int ifindex)
4554 struct ifaddrmsg *ifm;
4556 ifm = nlmsg_data(nlh);
4557 ifm->ifa_family = AF_INET6;
4558 ifm->ifa_prefixlen = prefixlen;
4559 ifm->ifa_flags = flags;
4560 ifm->ifa_scope = scope;
4561 ifm->ifa_index = ifindex;
4564 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4565 unsigned long tstamp, u32 preferred, u32 valid)
4567 struct ifa_cacheinfo ci;
4569 ci.cstamp = cstamp_delta(cstamp);
4570 ci.tstamp = cstamp_delta(tstamp);
4571 ci.ifa_prefered = preferred;
4572 ci.ifa_valid = valid;
4574 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4577 static inline int rt_scope(int ifa_scope)
4579 if (ifa_scope & IFA_HOST)
4580 return RT_SCOPE_HOST;
4581 else if (ifa_scope & IFA_LINK)
4582 return RT_SCOPE_LINK;
4583 else if (ifa_scope & IFA_SITE)
4584 return RT_SCOPE_SITE;
4586 return RT_SCOPE_UNIVERSE;
4589 static inline int inet6_ifaddr_msgsize(void)
4591 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4592 + nla_total_size(16) /* IFA_LOCAL */
4593 + nla_total_size(16) /* IFA_ADDRESS */
4594 + nla_total_size(sizeof(struct ifa_cacheinfo))
4595 + nla_total_size(4) /* IFA_FLAGS */;
4598 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4599 u32 portid, u32 seq, int event, unsigned int flags)
4601 struct nlmsghdr *nlh;
4602 u32 preferred, valid;
4604 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4608 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4609 ifa->idev->dev->ifindex);
4611 if (!((ifa->flags&IFA_F_PERMANENT) &&
4612 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4613 preferred = ifa->prefered_lft;
4614 valid = ifa->valid_lft;
4615 if (preferred != INFINITY_LIFE_TIME) {
4616 long tval = (jiffies - ifa->tstamp)/HZ;
4617 if (preferred > tval)
4621 if (valid != INFINITY_LIFE_TIME) {
4629 preferred = INFINITY_LIFE_TIME;
4630 valid = INFINITY_LIFE_TIME;
4633 if (!ipv6_addr_any(&ifa->peer_addr)) {
4634 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4635 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4638 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4641 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4644 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4647 nlmsg_end(skb, nlh);
4651 nlmsg_cancel(skb, nlh);
4655 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4656 u32 portid, u32 seq, int event, u16 flags)
4658 struct nlmsghdr *nlh;
4659 u8 scope = RT_SCOPE_UNIVERSE;
4660 int ifindex = ifmca->idev->dev->ifindex;
4662 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4663 scope = RT_SCOPE_SITE;
4665 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4669 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4670 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4671 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4672 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4673 nlmsg_cancel(skb, nlh);
4677 nlmsg_end(skb, nlh);
4681 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4682 u32 portid, u32 seq, int event, unsigned int flags)
4684 struct nlmsghdr *nlh;
4685 u8 scope = RT_SCOPE_UNIVERSE;
4686 int ifindex = ifaca->aca_idev->dev->ifindex;
4688 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4689 scope = RT_SCOPE_SITE;
4691 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4695 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4696 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4697 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4698 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4699 nlmsg_cancel(skb, nlh);
4703 nlmsg_end(skb, nlh);
4713 /* called with rcu_read_lock() */
4714 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4715 struct netlink_callback *cb, enum addr_type_t type,
4716 int s_ip_idx, int *p_ip_idx)
4718 struct ifmcaddr6 *ifmca;
4719 struct ifacaddr6 *ifaca;
4721 int ip_idx = *p_ip_idx;
4723 read_lock_bh(&idev->lock);
4725 case UNICAST_ADDR: {
4726 struct inet6_ifaddr *ifa;
4728 /* unicast address incl. temp addr */
4729 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4730 if (ip_idx < s_ip_idx)
4732 err = inet6_fill_ifaddr(skb, ifa,
4733 NETLINK_CB(cb->skb).portid,
4739 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4745 case MULTICAST_ADDR:
4746 /* multicast address */
4747 for (ifmca = idev->mc_list; ifmca;
4748 ifmca = ifmca->next, ip_idx++) {
4749 if (ip_idx < s_ip_idx)
4751 err = inet6_fill_ifmcaddr(skb, ifmca,
4752 NETLINK_CB(cb->skb).portid,
4761 /* anycast address */
4762 for (ifaca = idev->ac_list; ifaca;
4763 ifaca = ifaca->aca_next, ip_idx++) {
4764 if (ip_idx < s_ip_idx)
4766 err = inet6_fill_ifacaddr(skb, ifaca,
4767 NETLINK_CB(cb->skb).portid,
4778 read_unlock_bh(&idev->lock);
4783 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4784 enum addr_type_t type)
4786 struct net *net = sock_net(skb->sk);
4789 int s_idx, s_ip_idx;
4790 struct net_device *dev;
4791 struct inet6_dev *idev;
4792 struct hlist_head *head;
4795 s_idx = idx = cb->args[1];
4796 s_ip_idx = ip_idx = cb->args[2];
4799 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4800 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4802 head = &net->dev_index_head[h];
4803 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4806 if (h > s_h || idx > s_idx)
4809 idev = __in6_dev_get(dev);
4813 if (in6_dump_addrs(idev, skb, cb, type,
4814 s_ip_idx, &ip_idx) < 0)
4824 cb->args[2] = ip_idx;
4829 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4831 enum addr_type_t type = UNICAST_ADDR;
4833 return inet6_dump_addr(skb, cb, type);
4836 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4838 enum addr_type_t type = MULTICAST_ADDR;
4840 return inet6_dump_addr(skb, cb, type);
4844 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4846 enum addr_type_t type = ANYCAST_ADDR;
4848 return inet6_dump_addr(skb, cb, type);
4851 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4853 struct net *net = sock_net(in_skb->sk);
4854 struct ifaddrmsg *ifm;
4855 struct nlattr *tb[IFA_MAX+1];
4856 struct in6_addr *addr = NULL, *peer;
4857 struct net_device *dev = NULL;
4858 struct inet6_ifaddr *ifa;
4859 struct sk_buff *skb;
4862 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4866 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4872 ifm = nlmsg_data(nlh);
4874 dev = __dev_get_by_index(net, ifm->ifa_index);
4876 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4878 err = -EADDRNOTAVAIL;
4882 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4888 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4889 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4891 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4892 WARN_ON(err == -EMSGSIZE);
4896 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4903 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4905 struct sk_buff *skb;
4906 struct net *net = dev_net(ifa->idev->dev);
4909 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4913 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4915 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4916 WARN_ON(err == -EMSGSIZE);
4920 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4924 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4927 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4928 __s32 *array, int bytes)
4930 BUG_ON(bytes < (DEVCONF_MAX * 4));
4932 memset(array, 0, bytes);
4933 array[DEVCONF_FORWARDING] = cnf->forwarding;
4934 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4935 array[DEVCONF_MTU6] = cnf->mtu6;
4936 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4937 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4938 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4939 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4940 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4941 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4942 jiffies_to_msecs(cnf->rtr_solicit_interval);
4943 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4944 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4945 array[DEVCONF_RTR_SOLICIT_DELAY] =
4946 jiffies_to_msecs(cnf->rtr_solicit_delay);
4947 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4948 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4949 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4950 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4951 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4952 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4953 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4954 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4955 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4956 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4957 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4958 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4959 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4960 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4961 #ifdef CONFIG_IPV6_ROUTER_PREF
4962 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4963 array[DEVCONF_RTR_PROBE_INTERVAL] =
4964 jiffies_to_msecs(cnf->rtr_probe_interval);
4965 #ifdef CONFIG_IPV6_ROUTE_INFO
4966 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4969 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4970 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4971 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4972 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4973 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4975 #ifdef CONFIG_IPV6_MROUTE
4976 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4978 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4979 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4980 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4981 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4982 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4983 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4984 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4985 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4986 /* we omit DEVCONF_STABLE_SECRET for now */
4987 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4988 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4989 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4990 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4993 static inline size_t inet6_ifla6_size(void)
4995 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4996 + nla_total_size(sizeof(struct ifla_cacheinfo))
4997 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4998 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4999 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5000 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5003 static inline size_t inet6_if_nlmsg_size(void)
5005 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5006 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5007 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5008 + nla_total_size(4) /* IFLA_MTU */
5009 + nla_total_size(4) /* IFLA_LINK */
5010 + nla_total_size(1) /* IFLA_OPERSTATE */
5011 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5014 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5018 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5021 /* Use put_unaligned() because stats may not be aligned for u64. */
5022 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5023 for (i = 1; i < ICMP6_MIB_MAX; i++)
5024 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5026 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5029 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5030 int bytes, size_t syncpoff)
5033 u64 buff[IPSTATS_MIB_MAX];
5034 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5038 memset(buff, 0, sizeof(buff));
5039 buff[0] = IPSTATS_MIB_MAX;
5041 for_each_possible_cpu(c) {
5042 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5043 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5046 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5047 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5050 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5054 case IFLA_INET6_STATS:
5055 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5056 offsetof(struct ipstats_mib, syncp));
5058 case IFLA_INET6_ICMP6STATS:
5059 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5064 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5065 u32 ext_filter_mask)
5068 struct ifla_cacheinfo ci;
5070 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5071 goto nla_put_failure;
5072 ci.max_reasm_len = IPV6_MAXPLEN;
5073 ci.tstamp = cstamp_delta(idev->tstamp);
5074 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5075 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5076 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5077 goto nla_put_failure;
5078 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5080 goto nla_put_failure;
5081 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5083 /* XXX - MC not implemented */
5085 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5088 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5090 goto nla_put_failure;
5091 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5093 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5095 goto nla_put_failure;
5096 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5098 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5100 goto nla_put_failure;
5102 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5103 goto nla_put_failure;
5105 read_lock_bh(&idev->lock);
5106 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5107 read_unlock_bh(&idev->lock);
5115 static size_t inet6_get_link_af_size(const struct net_device *dev,
5116 u32 ext_filter_mask)
5118 if (!__in6_dev_get(dev))
5121 return inet6_ifla6_size();
5124 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5125 u32 ext_filter_mask)
5127 struct inet6_dev *idev = __in6_dev_get(dev);
5132 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5138 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5140 struct inet6_ifaddr *ifp;
5141 struct net_device *dev = idev->dev;
5142 bool clear_token, update_rs = false;
5143 struct in6_addr ll_addr;
5149 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5151 if (!ipv6_accept_ra(idev))
5153 if (idev->cnf.rtr_solicits == 0)
5156 write_lock_bh(&idev->lock);
5158 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5159 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5161 write_unlock_bh(&idev->lock);
5163 clear_token = ipv6_addr_any(token);
5167 if (!idev->dead && (idev->if_flags & IF_READY) &&
5168 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5169 IFA_F_OPTIMISTIC)) {
5170 /* If we're not ready, then normal ifup will take care
5171 * of this. Otherwise, we need to request our rs here.
5173 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5178 write_lock_bh(&idev->lock);
5181 idev->if_flags |= IF_RS_SENT;
5182 idev->rs_interval = rfc3315_s14_backoff_init(
5183 idev->cnf.rtr_solicit_interval);
5184 idev->rs_probes = 1;
5185 addrconf_mod_rs_timer(idev, idev->rs_interval);
5188 /* Well, that's kinda nasty ... */
5189 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5190 spin_lock(&ifp->lock);
5191 if (ifp->tokenized) {
5193 ifp->prefered_lft = 0;
5195 spin_unlock(&ifp->lock);
5198 write_unlock_bh(&idev->lock);
5199 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5200 addrconf_verify_rtnl();
5204 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5205 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5206 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5209 static int inet6_validate_link_af(const struct net_device *dev,
5210 const struct nlattr *nla)
5212 struct nlattr *tb[IFLA_INET6_MAX + 1];
5214 if (dev && !__in6_dev_get(dev))
5215 return -EAFNOSUPPORT;
5217 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5220 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5223 struct inet6_dev *idev = __in6_dev_get(dev);
5224 struct nlattr *tb[IFLA_INET6_MAX + 1];
5227 return -EAFNOSUPPORT;
5229 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5232 if (tb[IFLA_INET6_TOKEN]) {
5233 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5238 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5239 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5241 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5242 mode != IN6_ADDR_GEN_MODE_NONE &&
5243 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5244 mode != IN6_ADDR_GEN_MODE_RANDOM)
5247 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5248 !idev->cnf.stable_secret.initialized &&
5249 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5252 idev->addr_gen_mode = mode;
5259 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5260 u32 portid, u32 seq, int event, unsigned int flags)
5262 struct net_device *dev = idev->dev;
5263 struct ifinfomsg *hdr;
5264 struct nlmsghdr *nlh;
5267 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5271 hdr = nlmsg_data(nlh);
5272 hdr->ifi_family = AF_INET6;
5274 hdr->ifi_type = dev->type;
5275 hdr->ifi_index = dev->ifindex;
5276 hdr->ifi_flags = dev_get_flags(dev);
5277 hdr->ifi_change = 0;
5279 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5281 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5282 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5283 (dev->ifindex != dev_get_iflink(dev) &&
5284 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5285 nla_put_u8(skb, IFLA_OPERSTATE,
5286 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5287 goto nla_put_failure;
5288 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5290 goto nla_put_failure;
5292 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5293 goto nla_put_failure;
5295 nla_nest_end(skb, protoinfo);
5296 nlmsg_end(skb, nlh);
5300 nlmsg_cancel(skb, nlh);
5304 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5306 struct net *net = sock_net(skb->sk);
5309 struct net_device *dev;
5310 struct inet6_dev *idev;
5311 struct hlist_head *head;
5314 s_idx = cb->args[1];
5317 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5319 head = &net->dev_index_head[h];
5320 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5323 idev = __in6_dev_get(dev);
5326 if (inet6_fill_ifinfo(skb, idev,
5327 NETLINK_CB(cb->skb).portid,
5329 RTM_NEWLINK, NLM_F_MULTI) < 0)
5343 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5345 struct sk_buff *skb;
5346 struct net *net = dev_net(idev->dev);
5349 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5353 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5355 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5356 WARN_ON(err == -EMSGSIZE);
5360 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5364 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5367 static inline size_t inet6_prefix_nlmsg_size(void)
5369 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5370 + nla_total_size(sizeof(struct in6_addr))
5371 + nla_total_size(sizeof(struct prefix_cacheinfo));
5374 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5375 struct prefix_info *pinfo, u32 portid, u32 seq,
5376 int event, unsigned int flags)
5378 struct prefixmsg *pmsg;
5379 struct nlmsghdr *nlh;
5380 struct prefix_cacheinfo ci;
5382 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5386 pmsg = nlmsg_data(nlh);
5387 pmsg->prefix_family = AF_INET6;
5388 pmsg->prefix_pad1 = 0;
5389 pmsg->prefix_pad2 = 0;
5390 pmsg->prefix_ifindex = idev->dev->ifindex;
5391 pmsg->prefix_len = pinfo->prefix_len;
5392 pmsg->prefix_type = pinfo->type;
5393 pmsg->prefix_pad3 = 0;
5394 pmsg->prefix_flags = 0;
5396 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5397 if (pinfo->autoconf)
5398 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5400 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5401 goto nla_put_failure;
5402 ci.preferred_time = ntohl(pinfo->prefered);
5403 ci.valid_time = ntohl(pinfo->valid);
5404 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5405 goto nla_put_failure;
5406 nlmsg_end(skb, nlh);
5410 nlmsg_cancel(skb, nlh);
5414 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5415 struct prefix_info *pinfo)
5417 struct sk_buff *skb;
5418 struct net *net = dev_net(idev->dev);
5421 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5425 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5427 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5428 WARN_ON(err == -EMSGSIZE);
5432 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5436 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5439 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5441 struct net *net = dev_net(ifp->idev->dev);
5446 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5451 * If the address was optimistic we inserted the route at the
5452 * start of our DAD process, so we don't need to do it again.
5453 * If the device was taken down in the middle of the DAD
5454 * cycle there is a race where we could get here without a
5455 * host route, so nothing to insert. That will be fixed when
5456 * the device is brought up.
5458 if (ifp->rt && !rcu_access_pointer(ifp->rt->rt6i_node)) {
5459 ip6_ins_rt(ifp->rt);
5460 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
5461 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
5462 &ifp->addr, ifp->idev->dev->name);
5465 if (ifp->idev->cnf.forwarding)
5466 addrconf_join_anycast(ifp);
5467 if (!ipv6_addr_any(&ifp->peer_addr))
5468 addrconf_prefix_route(&ifp->peer_addr, 128,
5469 ifp->idev->dev, 0, 0);
5472 if (ifp->idev->cnf.forwarding)
5473 addrconf_leave_anycast(ifp);
5474 addrconf_leave_solict(ifp->idev, &ifp->addr);
5475 if (!ipv6_addr_any(&ifp->peer_addr)) {
5476 struct rt6_info *rt;
5478 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5479 ifp->idev->dev, 0, 0);
5484 dst_hold(&ifp->rt->dst);
5485 ip6_del_rt(ifp->rt);
5487 rt_genid_bump_ipv6(net);
5490 atomic_inc(&net->ipv6.dev_addr_genid);
5493 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5496 if (likely(ifp->idev->dead == 0))
5497 __ipv6_ifa_notify(event, ifp);
5498 rcu_read_unlock_bh();
5501 #ifdef CONFIG_SYSCTL
5504 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5505 void __user *buffer, size_t *lenp, loff_t *ppos)
5507 int *valp = ctl->data;
5510 struct ctl_table lctl;
5514 * ctl->data points to idev->cnf.forwarding, we should
5515 * not modify it until we get the rtnl lock.
5520 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5523 ret = addrconf_fixup_forwarding(ctl, valp, val);
5530 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5531 void __user *buffer, size_t *lenp, loff_t *ppos)
5533 struct inet6_dev *idev = ctl->extra1;
5534 int min_mtu = IPV6_MIN_MTU;
5535 struct ctl_table lctl;
5538 lctl.extra1 = &min_mtu;
5539 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5541 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5544 static void dev_disable_change(struct inet6_dev *idev)
5546 struct netdev_notifier_info info;
5548 if (!idev || !idev->dev)
5551 netdev_notifier_info_init(&info, idev->dev);
5552 if (idev->cnf.disable_ipv6)
5553 addrconf_notify(NULL, NETDEV_DOWN, &info);
5555 addrconf_notify(NULL, NETDEV_UP, &info);
5558 static void addrconf_disable_change(struct net *net, __s32 newf)
5560 struct net_device *dev;
5561 struct inet6_dev *idev;
5563 for_each_netdev(net, dev) {
5564 idev = __in6_dev_get(dev);
5566 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5567 idev->cnf.disable_ipv6 = newf;
5569 dev_disable_change(idev);
5574 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5579 if (!rtnl_trylock())
5580 return restart_syscall();
5582 net = (struct net *)table->extra2;
5586 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5591 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5592 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5593 addrconf_disable_change(net, newf);
5594 } else if ((!newf) ^ (!old))
5595 dev_disable_change((struct inet6_dev *)table->extra1);
5602 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5603 void __user *buffer, size_t *lenp, loff_t *ppos)
5605 int *valp = ctl->data;
5608 struct ctl_table lctl;
5612 * ctl->data points to idev->cnf.disable_ipv6, we should
5613 * not modify it until we get the rtnl lock.
5618 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5621 ret = addrconf_disable_ipv6(ctl, valp, val);
5628 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5629 void __user *buffer, size_t *lenp, loff_t *ppos)
5631 int *valp = ctl->data;
5636 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5639 if (write && old != new) {
5640 struct net *net = ctl->extra2;
5642 if (!rtnl_trylock())
5643 return restart_syscall();
5645 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5646 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5647 NETCONFA_IFINDEX_DEFAULT,
5648 net->ipv6.devconf_dflt);
5649 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5650 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5651 NETCONFA_IFINDEX_ALL,
5652 net->ipv6.devconf_all);
5654 struct inet6_dev *idev = ctl->extra1;
5656 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5666 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5667 void __user *buffer, size_t *lenp,
5671 struct in6_addr addr;
5672 char str[IPV6_MAX_STRLEN];
5673 struct ctl_table lctl = *ctl;
5674 struct net *net = ctl->extra2;
5675 struct ipv6_stable_secret *secret = ctl->data;
5677 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5680 lctl.maxlen = IPV6_MAX_STRLEN;
5683 if (!rtnl_trylock())
5684 return restart_syscall();
5686 if (!write && !secret->initialized) {
5691 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5692 if (err >= sizeof(str)) {
5697 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5701 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5706 secret->initialized = true;
5707 secret->secret = addr;
5709 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5710 struct net_device *dev;
5712 for_each_netdev(net, dev) {
5713 struct inet6_dev *idev = __in6_dev_get(dev);
5716 idev->addr_gen_mode =
5717 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5721 struct inet6_dev *idev = ctl->extra1;
5723 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5733 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5735 void __user *buffer,
5739 int *valp = ctl->data;
5742 struct ctl_table lctl;
5745 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5746 * we should not modify it until we get the rtnl lock.
5751 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5754 ret = addrconf_fixup_linkdown(ctl, valp, val);
5760 static int minus_one = -1;
5761 static const int one = 1;
5762 static const int two_five_five = 255;
5764 static const struct ctl_table addrconf_sysctl[] = {
5766 .procname = "forwarding",
5767 .data = &ipv6_devconf.forwarding,
5768 .maxlen = sizeof(int),
5770 .proc_handler = addrconf_sysctl_forward,
5773 .procname = "hop_limit",
5774 .data = &ipv6_devconf.hop_limit,
5775 .maxlen = sizeof(int),
5777 .proc_handler = proc_dointvec_minmax,
5778 .extra1 = (void *)&one,
5779 .extra2 = (void *)&two_five_five,
5783 .data = &ipv6_devconf.mtu6,
5784 .maxlen = sizeof(int),
5786 .proc_handler = addrconf_sysctl_mtu,
5789 .procname = "accept_ra",
5790 .data = &ipv6_devconf.accept_ra,
5791 .maxlen = sizeof(int),
5793 .proc_handler = proc_dointvec,
5796 .procname = "accept_redirects",
5797 .data = &ipv6_devconf.accept_redirects,
5798 .maxlen = sizeof(int),
5800 .proc_handler = proc_dointvec,
5803 .procname = "autoconf",
5804 .data = &ipv6_devconf.autoconf,
5805 .maxlen = sizeof(int),
5807 .proc_handler = proc_dointvec,
5810 .procname = "dad_transmits",
5811 .data = &ipv6_devconf.dad_transmits,
5812 .maxlen = sizeof(int),
5814 .proc_handler = proc_dointvec,
5817 .procname = "router_solicitations",
5818 .data = &ipv6_devconf.rtr_solicits,
5819 .maxlen = sizeof(int),
5821 .proc_handler = proc_dointvec_minmax,
5822 .extra1 = &minus_one,
5825 .procname = "router_solicitation_interval",
5826 .data = &ipv6_devconf.rtr_solicit_interval,
5827 .maxlen = sizeof(int),
5829 .proc_handler = proc_dointvec_jiffies,
5832 .procname = "router_solicitation_max_interval",
5833 .data = &ipv6_devconf.rtr_solicit_max_interval,
5834 .maxlen = sizeof(int),
5836 .proc_handler = proc_dointvec_jiffies,
5839 .procname = "router_solicitation_delay",
5840 .data = &ipv6_devconf.rtr_solicit_delay,
5841 .maxlen = sizeof(int),
5843 .proc_handler = proc_dointvec_jiffies,
5846 .procname = "force_mld_version",
5847 .data = &ipv6_devconf.force_mld_version,
5848 .maxlen = sizeof(int),
5850 .proc_handler = proc_dointvec,
5853 .procname = "mldv1_unsolicited_report_interval",
5855 &ipv6_devconf.mldv1_unsolicited_report_interval,
5856 .maxlen = sizeof(int),
5858 .proc_handler = proc_dointvec_ms_jiffies,
5861 .procname = "mldv2_unsolicited_report_interval",
5863 &ipv6_devconf.mldv2_unsolicited_report_interval,
5864 .maxlen = sizeof(int),
5866 .proc_handler = proc_dointvec_ms_jiffies,
5869 .procname = "use_tempaddr",
5870 .data = &ipv6_devconf.use_tempaddr,
5871 .maxlen = sizeof(int),
5873 .proc_handler = proc_dointvec,
5876 .procname = "temp_valid_lft",
5877 .data = &ipv6_devconf.temp_valid_lft,
5878 .maxlen = sizeof(int),
5880 .proc_handler = proc_dointvec,
5883 .procname = "temp_prefered_lft",
5884 .data = &ipv6_devconf.temp_prefered_lft,
5885 .maxlen = sizeof(int),
5887 .proc_handler = proc_dointvec,
5890 .procname = "regen_max_retry",
5891 .data = &ipv6_devconf.regen_max_retry,
5892 .maxlen = sizeof(int),
5894 .proc_handler = proc_dointvec,
5897 .procname = "max_desync_factor",
5898 .data = &ipv6_devconf.max_desync_factor,
5899 .maxlen = sizeof(int),
5901 .proc_handler = proc_dointvec,
5904 .procname = "max_addresses",
5905 .data = &ipv6_devconf.max_addresses,
5906 .maxlen = sizeof(int),
5908 .proc_handler = proc_dointvec,
5911 .procname = "accept_ra_defrtr",
5912 .data = &ipv6_devconf.accept_ra_defrtr,
5913 .maxlen = sizeof(int),
5915 .proc_handler = proc_dointvec,
5918 .procname = "accept_ra_min_hop_limit",
5919 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5920 .maxlen = sizeof(int),
5922 .proc_handler = proc_dointvec,
5925 .procname = "accept_ra_pinfo",
5926 .data = &ipv6_devconf.accept_ra_pinfo,
5927 .maxlen = sizeof(int),
5929 .proc_handler = proc_dointvec,
5931 #ifdef CONFIG_IPV6_ROUTER_PREF
5933 .procname = "accept_ra_rtr_pref",
5934 .data = &ipv6_devconf.accept_ra_rtr_pref,
5935 .maxlen = sizeof(int),
5937 .proc_handler = proc_dointvec,
5940 .procname = "router_probe_interval",
5941 .data = &ipv6_devconf.rtr_probe_interval,
5942 .maxlen = sizeof(int),
5944 .proc_handler = proc_dointvec_jiffies,
5946 #ifdef CONFIG_IPV6_ROUTE_INFO
5948 .procname = "accept_ra_rt_info_max_plen",
5949 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5950 .maxlen = sizeof(int),
5952 .proc_handler = proc_dointvec,
5957 .procname = "proxy_ndp",
5958 .data = &ipv6_devconf.proxy_ndp,
5959 .maxlen = sizeof(int),
5961 .proc_handler = addrconf_sysctl_proxy_ndp,
5964 .procname = "accept_source_route",
5965 .data = &ipv6_devconf.accept_source_route,
5966 .maxlen = sizeof(int),
5968 .proc_handler = proc_dointvec,
5970 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5972 .procname = "optimistic_dad",
5973 .data = &ipv6_devconf.optimistic_dad,
5974 .maxlen = sizeof(int),
5976 .proc_handler = proc_dointvec,
5979 .procname = "use_optimistic",
5980 .data = &ipv6_devconf.use_optimistic,
5981 .maxlen = sizeof(int),
5983 .proc_handler = proc_dointvec,
5986 #ifdef CONFIG_IPV6_MROUTE
5988 .procname = "mc_forwarding",
5989 .data = &ipv6_devconf.mc_forwarding,
5990 .maxlen = sizeof(int),
5992 .proc_handler = proc_dointvec,
5996 .procname = "disable_ipv6",
5997 .data = &ipv6_devconf.disable_ipv6,
5998 .maxlen = sizeof(int),
6000 .proc_handler = addrconf_sysctl_disable,
6003 .procname = "accept_dad",
6004 .data = &ipv6_devconf.accept_dad,
6005 .maxlen = sizeof(int),
6007 .proc_handler = proc_dointvec,
6010 .procname = "force_tllao",
6011 .data = &ipv6_devconf.force_tllao,
6012 .maxlen = sizeof(int),
6014 .proc_handler = proc_dointvec
6017 .procname = "ndisc_notify",
6018 .data = &ipv6_devconf.ndisc_notify,
6019 .maxlen = sizeof(int),
6021 .proc_handler = proc_dointvec
6024 .procname = "suppress_frag_ndisc",
6025 .data = &ipv6_devconf.suppress_frag_ndisc,
6026 .maxlen = sizeof(int),
6028 .proc_handler = proc_dointvec
6031 .procname = "accept_ra_from_local",
6032 .data = &ipv6_devconf.accept_ra_from_local,
6033 .maxlen = sizeof(int),
6035 .proc_handler = proc_dointvec,
6038 .procname = "accept_ra_mtu",
6039 .data = &ipv6_devconf.accept_ra_mtu,
6040 .maxlen = sizeof(int),
6042 .proc_handler = proc_dointvec,
6045 .procname = "stable_secret",
6046 .data = &ipv6_devconf.stable_secret,
6047 .maxlen = IPV6_MAX_STRLEN,
6049 .proc_handler = addrconf_sysctl_stable_secret,
6052 .procname = "use_oif_addrs_only",
6053 .data = &ipv6_devconf.use_oif_addrs_only,
6054 .maxlen = sizeof(int),
6056 .proc_handler = proc_dointvec,
6059 .procname = "ignore_routes_with_linkdown",
6060 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6061 .maxlen = sizeof(int),
6063 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6066 .procname = "drop_unicast_in_l2_multicast",
6067 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6068 .maxlen = sizeof(int),
6070 .proc_handler = proc_dointvec,
6073 .procname = "drop_unsolicited_na",
6074 .data = &ipv6_devconf.drop_unsolicited_na,
6075 .maxlen = sizeof(int),
6077 .proc_handler = proc_dointvec,
6080 .procname = "keep_addr_on_down",
6081 .data = &ipv6_devconf.keep_addr_on_down,
6082 .maxlen = sizeof(int),
6084 .proc_handler = proc_dointvec,
6092 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6093 struct inet6_dev *idev, struct ipv6_devconf *p)
6096 struct ctl_table *table;
6097 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6099 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6103 for (i = 0; table[i].data; i++) {
6104 table[i].data += (char *)p - (char *)&ipv6_devconf;
6105 /* If one of these is already set, then it is not safe to
6106 * overwrite either of them: this makes proc_dointvec_minmax
6109 if (!table[i].extra1 && !table[i].extra2) {
6110 table[i].extra1 = idev; /* embedded; no ref */
6111 table[i].extra2 = net;
6115 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6117 p->sysctl_header = register_net_sysctl(net, path, table);
6118 if (!p->sysctl_header)
6121 if (!strcmp(dev_name, "all"))
6122 ifindex = NETCONFA_IFINDEX_ALL;
6123 else if (!strcmp(dev_name, "default"))
6124 ifindex = NETCONFA_IFINDEX_DEFAULT;
6126 ifindex = idev->dev->ifindex;
6127 inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6136 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6138 struct ctl_table *table;
6140 if (!p->sysctl_header)
6143 table = p->sysctl_header->ctl_table_arg;
6144 unregister_net_sysctl_table(p->sysctl_header);
6145 p->sysctl_header = NULL;
6149 static int addrconf_sysctl_register(struct inet6_dev *idev)
6153 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6156 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6157 &ndisc_ifinfo_sysctl_change);
6160 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6163 neigh_sysctl_unregister(idev->nd_parms);
6168 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6170 __addrconf_sysctl_unregister(&idev->cnf);
6171 neigh_sysctl_unregister(idev->nd_parms);
6177 static int __net_init addrconf_init_net(struct net *net)
6180 struct ipv6_devconf *all, *dflt;
6182 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6186 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6188 goto err_alloc_dflt;
6190 /* these will be inherited by all namespaces */
6191 dflt->autoconf = ipv6_defaults.autoconf;
6192 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6194 dflt->stable_secret.initialized = false;
6195 all->stable_secret.initialized = false;
6197 net->ipv6.devconf_all = all;
6198 net->ipv6.devconf_dflt = dflt;
6200 #ifdef CONFIG_SYSCTL
6201 err = __addrconf_sysctl_register(net, "all", NULL, all);
6205 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6211 #ifdef CONFIG_SYSCTL
6213 __addrconf_sysctl_unregister(all);
6223 static void __net_exit addrconf_exit_net(struct net *net)
6225 #ifdef CONFIG_SYSCTL
6226 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6227 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6229 kfree(net->ipv6.devconf_dflt);
6230 kfree(net->ipv6.devconf_all);
6233 static struct pernet_operations addrconf_ops = {
6234 .init = addrconf_init_net,
6235 .exit = addrconf_exit_net,
6238 static struct rtnl_af_ops inet6_ops __read_mostly = {
6240 .fill_link_af = inet6_fill_link_af,
6241 .get_link_af_size = inet6_get_link_af_size,
6242 .validate_link_af = inet6_validate_link_af,
6243 .set_link_af = inet6_set_link_af,
6247 * Init / cleanup code
6250 int __init addrconf_init(void)
6252 struct inet6_dev *idev;
6255 err = ipv6_addr_label_init();
6257 pr_crit("%s: cannot initialize default policy table: %d\n",
6262 err = register_pernet_subsys(&addrconf_ops);
6266 addrconf_wq = create_workqueue("ipv6_addrconf");
6272 /* The addrconf netdev notifier requires that loopback_dev
6273 * has it's ipv6 private information allocated and setup
6274 * before it can bring up and give link-local addresses
6275 * to other devices which are up.
6277 * Unfortunately, loopback_dev is not necessarily the first
6278 * entry in the global dev_base list of net devices. In fact,
6279 * it is likely to be the very last entry on that list.
6280 * So this causes the notifier registry below to try and
6281 * give link-local addresses to all devices besides loopback_dev
6282 * first, then loopback_dev, which cases all the non-loopback_dev
6283 * devices to fail to get a link-local address.
6285 * So, as a temporary fix, allocate the ipv6 structure for
6286 * loopback_dev first by hand.
6287 * Longer term, all of the dependencies ipv6 has upon the loopback
6288 * device and it being up should be removed.
6291 idev = ipv6_add_dev(init_net.loopback_dev);
6294 err = PTR_ERR(idev);
6298 ip6_route_init_special_entries();
6300 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6301 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6303 register_netdevice_notifier(&ipv6_dev_notf);
6307 rtnl_af_register(&inet6_ops);
6309 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6314 /* Only the first call to __rtnl_register can fail */
6315 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6316 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6317 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6318 inet6_dump_ifaddr, NULL);
6319 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6320 inet6_dump_ifmcaddr, NULL);
6321 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6322 inet6_dump_ifacaddr, NULL);
6323 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6324 inet6_netconf_dump_devconf, NULL);
6326 ipv6_addr_label_rtnl_register();
6330 rtnl_af_unregister(&inet6_ops);
6331 unregister_netdevice_notifier(&ipv6_dev_notf);
6333 destroy_workqueue(addrconf_wq);
6335 unregister_pernet_subsys(&addrconf_ops);
6337 ipv6_addr_label_cleanup();
6342 void addrconf_cleanup(void)
6344 struct net_device *dev;
6347 unregister_netdevice_notifier(&ipv6_dev_notf);
6348 unregister_pernet_subsys(&addrconf_ops);
6349 ipv6_addr_label_cleanup();
6353 __rtnl_af_unregister(&inet6_ops);
6355 /* clean dev list */
6356 for_each_netdev(&init_net, dev) {
6357 if (__in6_dev_get(dev) == NULL)
6359 addrconf_ifdown(dev, 1);
6361 addrconf_ifdown(init_net.loopback_dev, 2);
6366 spin_lock_bh(&addrconf_hash_lock);
6367 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6368 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6369 spin_unlock_bh(&addrconf_hash_lock);
6370 cancel_delayed_work(&addr_chk_work);
6373 destroy_workqueue(addrconf_wq);