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/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
60 #include <linux/sysctl.h>
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
68 #include <net/net_namespace.h>
72 #include <net/af_ieee802154.h>
73 #include <net/firewire.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86 #include <linux/random.h>
87 #include <linux/uaccess.h>
88 #include <asm/unaligned.h>
90 #include <linux/proc_fs.h>
91 #include <linux/seq_file.h>
92 #include <linux/export.h>
94 /* Set to 3 to get tracing... */
98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #define INFINITY_LIFE_TIME 0xFFFFFFFF
105 static inline u32 cstamp_delta(unsigned long cstamp)
107 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 static void addrconf_sysctl_register(struct inet6_dev *idev);
112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
125 static void ipv6_regen_rndid(unsigned long data);
127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
128 static int ipv6_count_addresses(struct inet6_dev *idev);
131 * Configured unicast address hash table
133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
134 static DEFINE_SPINLOCK(addrconf_hash_lock);
136 static void addrconf_verify(void);
137 static void addrconf_verify_rtnl(void);
138 static void addrconf_verify_work(struct work_struct *);
140 static struct workqueue_struct *addrconf_wq;
141 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
143 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
144 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_type_change(struct net_device *dev,
147 unsigned long event);
148 static int addrconf_ifdown(struct net_device *dev, int how);
150 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
152 const struct net_device *dev,
153 u32 flags, u32 noflags);
155 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
156 static void addrconf_dad_work(struct work_struct *w);
157 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
158 static void addrconf_dad_run(struct inet6_dev *idev);
159 static void addrconf_rs_timer(unsigned long data);
160 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
161 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
164 struct prefix_info *pinfo);
165 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
166 struct net_device *dev);
168 static struct ipv6_devconf ipv6_devconf __read_mostly = {
170 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
171 .mtu6 = IPV6_MIN_MTU,
173 .accept_redirects = 1,
175 .force_mld_version = 0,
176 .mldv1_unsolicited_report_interval = 10 * HZ,
177 .mldv2_unsolicited_report_interval = HZ,
179 .rtr_solicits = MAX_RTR_SOLICITATIONS,
180 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
181 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
183 .temp_valid_lft = TEMP_VALID_LIFETIME,
184 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
185 .regen_max_retry = REGEN_MAX_RETRY,
186 .max_desync_factor = MAX_DESYNC_FACTOR,
187 .max_addresses = IPV6_MAX_ADDRESSES,
188 .accept_ra_defrtr = 1,
189 .accept_ra_pinfo = 1,
190 #ifdef CONFIG_IPV6_ROUTER_PREF
191 .accept_ra_rtr_pref = 1,
192 .rtr_probe_interval = 60 * HZ,
193 #ifdef CONFIG_IPV6_ROUTE_INFO
194 .accept_ra_rt_info_max_plen = 0,
198 .accept_source_route = 0, /* we do not accept RH0 by default. */
201 .suppress_frag_ndisc = 1,
204 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
206 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
207 .mtu6 = IPV6_MIN_MTU,
209 .accept_redirects = 1,
211 .force_mld_version = 0,
212 .mldv1_unsolicited_report_interval = 10 * HZ,
213 .mldv2_unsolicited_report_interval = HZ,
215 .rtr_solicits = MAX_RTR_SOLICITATIONS,
216 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
217 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
219 .temp_valid_lft = TEMP_VALID_LIFETIME,
220 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
221 .regen_max_retry = REGEN_MAX_RETRY,
222 .max_desync_factor = MAX_DESYNC_FACTOR,
223 .max_addresses = IPV6_MAX_ADDRESSES,
224 .accept_ra_defrtr = 1,
225 .accept_ra_pinfo = 1,
226 #ifdef CONFIG_IPV6_ROUTER_PREF
227 .accept_ra_rtr_pref = 1,
228 .rtr_probe_interval = 60 * HZ,
229 #ifdef CONFIG_IPV6_ROUTE_INFO
230 .accept_ra_rt_info_max_plen = 0,
234 .accept_source_route = 0, /* we do not accept RH0 by default. */
237 .suppress_frag_ndisc = 1,
240 /* Check if a valid qdisc is available */
241 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
243 return !qdisc_tx_is_noop(dev);
246 static void addrconf_del_rs_timer(struct inet6_dev *idev)
248 if (del_timer(&idev->rs_timer))
252 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
254 if (cancel_delayed_work(&ifp->dad_work))
258 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
261 if (!timer_pending(&idev->rs_timer))
263 mod_timer(&idev->rs_timer, jiffies + when);
266 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
269 if (!delayed_work_pending(&ifp->dad_work))
271 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
274 static int snmp6_alloc_dev(struct inet6_dev *idev)
278 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
279 sizeof(struct ipstats_mib),
280 __alignof__(struct ipstats_mib)) < 0)
283 for_each_possible_cpu(i) {
284 struct ipstats_mib *addrconf_stats;
285 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i);
286 u64_stats_init(&addrconf_stats->syncp);
287 #if SNMP_ARRAY_SZ == 2
288 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i);
289 u64_stats_init(&addrconf_stats->syncp);
294 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
296 if (!idev->stats.icmpv6dev)
298 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
300 if (!idev->stats.icmpv6msgdev)
306 kfree(idev->stats.icmpv6dev);
308 snmp_mib_free((void __percpu **)idev->stats.ipv6);
313 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
315 struct inet6_dev *ndev;
319 if (dev->mtu < IPV6_MIN_MTU)
322 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
327 rwlock_init(&ndev->lock);
329 INIT_LIST_HEAD(&ndev->addr_list);
330 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
331 (unsigned long)ndev);
332 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
333 ndev->cnf.mtu6 = dev->mtu;
334 ndev->cnf.sysctl = NULL;
335 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
336 if (ndev->nd_parms == NULL) {
340 if (ndev->cnf.forwarding)
341 dev_disable_lro(dev);
342 /* We refer to the device */
345 if (snmp6_alloc_dev(ndev) < 0) {
347 "%s: cannot allocate memory for statistics; dev=%s.\n",
348 __func__, dev->name);
349 neigh_parms_release(&nd_tbl, ndev->nd_parms);
355 if (snmp6_register_dev(ndev) < 0) {
357 "%s: cannot create /proc/net/dev_snmp6/%s\n",
358 __func__, dev->name);
359 neigh_parms_release(&nd_tbl, ndev->nd_parms);
361 in6_dev_finish_destroy(ndev);
365 /* One reference from device. We must do this before
366 * we invoke __ipv6_regen_rndid().
370 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
371 ndev->cnf.accept_dad = -1;
373 #if IS_ENABLED(CONFIG_IPV6_SIT)
374 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
375 pr_info("%s: Disabled Multicast RS\n", dev->name);
376 ndev->cnf.rtr_solicits = 0;
380 INIT_LIST_HEAD(&ndev->tempaddr_list);
381 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
382 if ((dev->flags&IFF_LOOPBACK) ||
383 dev->type == ARPHRD_TUNNEL ||
384 dev->type == ARPHRD_TUNNEL6 ||
385 dev->type == ARPHRD_SIT ||
386 dev->type == ARPHRD_NONE) {
387 ndev->cnf.use_tempaddr = -1;
390 ipv6_regen_rndid((unsigned long) ndev);
393 ndev->token = in6addr_any;
395 if (netif_running(dev) && addrconf_qdisc_ok(dev))
396 ndev->if_flags |= IF_READY;
398 ipv6_mc_init_dev(ndev);
399 ndev->tstamp = jiffies;
400 addrconf_sysctl_register(ndev);
401 /* protected by rtnl_lock */
402 rcu_assign_pointer(dev->ip6_ptr, ndev);
404 /* Join interface-local all-node multicast group */
405 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
407 /* Join all-node multicast group */
408 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
410 /* Join all-router multicast group if forwarding is set */
411 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
412 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
417 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
419 struct inet6_dev *idev;
423 idev = __in6_dev_get(dev);
425 idev = ipv6_add_dev(dev);
430 if (dev->flags&IFF_UP)
435 static int inet6_netconf_msgsize_devconf(int type)
437 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
438 + nla_total_size(4); /* NETCONFA_IFINDEX */
440 /* type -1 is used for ALL */
441 if (type == -1 || type == NETCONFA_FORWARDING)
442 size += nla_total_size(4);
443 #ifdef CONFIG_IPV6_MROUTE
444 if (type == -1 || type == NETCONFA_MC_FORWARDING)
445 size += nla_total_size(4);
447 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
448 size += nla_total_size(4);
453 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
454 struct ipv6_devconf *devconf, u32 portid,
455 u32 seq, int event, unsigned int flags,
458 struct nlmsghdr *nlh;
459 struct netconfmsg *ncm;
461 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
466 ncm = nlmsg_data(nlh);
467 ncm->ncm_family = AF_INET6;
469 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
470 goto nla_put_failure;
472 /* type -1 is used for ALL */
473 if ((type == -1 || type == NETCONFA_FORWARDING) &&
474 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
475 goto nla_put_failure;
476 #ifdef CONFIG_IPV6_MROUTE
477 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
478 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
479 devconf->mc_forwarding) < 0)
480 goto nla_put_failure;
482 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
483 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
484 goto nla_put_failure;
486 return nlmsg_end(skb, nlh);
489 nlmsg_cancel(skb, nlh);
493 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
494 struct ipv6_devconf *devconf)
499 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
503 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
504 RTM_NEWNETCONF, 0, type);
506 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
507 WARN_ON(err == -EMSGSIZE);
511 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
514 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
517 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
518 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
519 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
520 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
523 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
524 struct nlmsghdr *nlh)
526 struct net *net = sock_net(in_skb->sk);
527 struct nlattr *tb[NETCONFA_MAX+1];
528 struct netconfmsg *ncm;
530 struct ipv6_devconf *devconf;
531 struct inet6_dev *in6_dev;
532 struct net_device *dev;
536 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
537 devconf_ipv6_policy);
542 if (!tb[NETCONFA_IFINDEX])
545 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
547 case NETCONFA_IFINDEX_ALL:
548 devconf = net->ipv6.devconf_all;
550 case NETCONFA_IFINDEX_DEFAULT:
551 devconf = net->ipv6.devconf_dflt;
554 dev = __dev_get_by_index(net, ifindex);
557 in6_dev = __in6_dev_get(dev);
560 devconf = &in6_dev->cnf;
565 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
569 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
570 NETLINK_CB(in_skb).portid,
571 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
574 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
575 WARN_ON(err == -EMSGSIZE);
579 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
584 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
585 struct netlink_callback *cb)
587 struct net *net = sock_net(skb->sk);
590 struct net_device *dev;
591 struct inet6_dev *idev;
592 struct hlist_head *head;
595 s_idx = idx = cb->args[1];
597 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
599 head = &net->dev_index_head[h];
601 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
603 hlist_for_each_entry_rcu(dev, head, index_hlist) {
606 idev = __in6_dev_get(dev);
610 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
612 NETLINK_CB(cb->skb).portid,
620 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
626 if (h == NETDEV_HASHENTRIES) {
627 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
628 net->ipv6.devconf_all,
629 NETLINK_CB(cb->skb).portid,
631 RTM_NEWNETCONF, NLM_F_MULTI,
637 if (h == NETDEV_HASHENTRIES + 1) {
638 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
639 net->ipv6.devconf_dflt,
640 NETLINK_CB(cb->skb).portid,
642 RTM_NEWNETCONF, NLM_F_MULTI,
656 static void dev_forward_change(struct inet6_dev *idev)
658 struct net_device *dev;
659 struct inet6_ifaddr *ifa;
664 if (idev->cnf.forwarding)
665 dev_disable_lro(dev);
666 if (dev->flags & IFF_MULTICAST) {
667 if (idev->cnf.forwarding) {
668 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
669 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
670 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
672 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
673 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
674 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
678 list_for_each_entry(ifa, &idev->addr_list, if_list) {
679 if (ifa->flags&IFA_F_TENTATIVE)
681 if (idev->cnf.forwarding)
682 addrconf_join_anycast(ifa);
684 addrconf_leave_anycast(ifa);
686 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
687 dev->ifindex, &idev->cnf);
691 static void addrconf_forward_change(struct net *net, __s32 newf)
693 struct net_device *dev;
694 struct inet6_dev *idev;
696 for_each_netdev(net, dev) {
697 idev = __in6_dev_get(dev);
699 int changed = (!idev->cnf.forwarding) ^ (!newf);
700 idev->cnf.forwarding = newf;
702 dev_forward_change(idev);
707 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
713 return restart_syscall();
715 net = (struct net *)table->extra2;
719 if (p == &net->ipv6.devconf_dflt->forwarding) {
720 if ((!newf) ^ (!old))
721 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
722 NETCONFA_IFINDEX_DEFAULT,
723 net->ipv6.devconf_dflt);
728 if (p == &net->ipv6.devconf_all->forwarding) {
729 net->ipv6.devconf_dflt->forwarding = newf;
730 addrconf_forward_change(net, newf);
731 if ((!newf) ^ (!old))
732 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
733 NETCONFA_IFINDEX_ALL,
734 net->ipv6.devconf_all);
735 } else if ((!newf) ^ (!old))
736 dev_forward_change((struct inet6_dev *)table->extra1);
740 rt6_purge_dflt_routers(net);
745 /* Nobody refers to this ifaddr, destroy it */
746 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
748 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
750 #ifdef NET_REFCNT_DEBUG
751 pr_debug("%s\n", __func__);
754 in6_dev_put(ifp->idev);
756 if (cancel_delayed_work(&ifp->dad_work))
757 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
760 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
761 pr_warn("Freeing alive inet6 address %p\n", ifp);
770 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
773 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
776 * Each device address list is sorted in order of scope -
777 * global before linklocal.
779 list_for_each(p, &idev->addr_list) {
780 struct inet6_ifaddr *ifa
781 = list_entry(p, struct inet6_ifaddr, if_list);
782 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
786 list_add_tail(&ifp->if_list, p);
789 static u32 inet6_addr_hash(const struct in6_addr *addr)
791 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
794 /* On success it returns ifp with increased reference count */
796 static struct inet6_ifaddr *
797 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
798 const struct in6_addr *peer_addr, int pfxlen,
799 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
801 struct inet6_ifaddr *ifa = NULL;
805 int addr_type = ipv6_addr_type(addr);
807 if (addr_type == IPV6_ADDR_ANY ||
808 addr_type & IPV6_ADDR_MULTICAST ||
809 (!(idev->dev->flags & IFF_LOOPBACK) &&
810 addr_type & IPV6_ADDR_LOOPBACK))
811 return ERR_PTR(-EADDRNOTAVAIL);
815 err = -ENODEV; /*XXX*/
819 if (idev->cnf.disable_ipv6) {
824 spin_lock(&addrconf_hash_lock);
826 /* Ignore adding duplicate addresses on an interface */
827 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
828 ADBG("ipv6_add_addr: already assigned\n");
833 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
836 ADBG("ipv6_add_addr: malloc failed\n");
841 rt = addrconf_dst_alloc(idev, addr, false);
847 neigh_parms_data_state_setall(idev->nd_parms);
851 ifa->peer_addr = *peer_addr;
853 spin_lock_init(&ifa->lock);
854 spin_lock_init(&ifa->state_lock);
855 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
856 INIT_HLIST_NODE(&ifa->addr_lst);
858 ifa->prefix_len = pfxlen;
859 ifa->flags = flags | IFA_F_TENTATIVE;
860 ifa->valid_lft = valid_lft;
861 ifa->prefered_lft = prefered_lft;
862 ifa->cstamp = ifa->tstamp = jiffies;
863 ifa->tokenized = false;
872 /* Add to big hash table */
873 hash = inet6_addr_hash(addr);
875 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
876 spin_unlock(&addrconf_hash_lock);
878 write_lock(&idev->lock);
879 /* Add to inet6_dev unicast addr list. */
880 ipv6_link_dev_addr(idev, ifa);
882 if (ifa->flags&IFA_F_TEMPORARY) {
883 list_add(&ifa->tmp_list, &idev->tempaddr_list);
888 write_unlock(&idev->lock);
890 rcu_read_unlock_bh();
892 if (likely(err == 0))
893 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
901 spin_unlock(&addrconf_hash_lock);
905 enum cleanup_prefix_rt_t {
906 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
907 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
908 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
912 * Check, whether the prefix for ifp would still need a prefix route
913 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
916 * 1) we don't purge prefix if address was not permanent.
917 * prefix is managed by its own lifetime.
918 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
919 * 3) if there are no addresses, delete prefix.
920 * 4) if there are still other permanent address(es),
921 * corresponding prefix is still permanent.
922 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
923 * don't purge the prefix, assume user space is managing it.
924 * 6) otherwise, update prefix lifetime to the
925 * longest valid lifetime among the corresponding
926 * addresses on the device.
927 * Note: subsequent RA will update lifetime.
929 static enum cleanup_prefix_rt_t
930 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
932 struct inet6_ifaddr *ifa;
933 struct inet6_dev *idev = ifp->idev;
934 unsigned long lifetime;
935 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
939 list_for_each_entry(ifa, &idev->addr_list, if_list) {
942 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
945 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
946 return CLEANUP_PREFIX_RT_NOP;
948 action = CLEANUP_PREFIX_RT_EXPIRE;
950 spin_lock(&ifa->lock);
952 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
954 * Note: Because this address is
955 * not permanent, lifetime <
956 * LONG_MAX / HZ here.
958 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
959 *expires = ifa->tstamp + lifetime * HZ;
960 spin_unlock(&ifa->lock);
967 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
971 rt = addrconf_get_prefix_route(&ifp->addr,
974 0, RTF_GATEWAY | RTF_DEFAULT);
979 if (!(rt->rt6i_flags & RTF_EXPIRES))
980 rt6_set_expires(rt, expires);
987 /* This function wants to get referenced ifp and releases it before return */
989 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
992 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
993 unsigned long expires;
997 spin_lock_bh(&ifp->state_lock);
999 ifp->state = INET6_IFADDR_STATE_DEAD;
1000 spin_unlock_bh(&ifp->state_lock);
1002 if (state == INET6_IFADDR_STATE_DEAD)
1005 spin_lock_bh(&addrconf_hash_lock);
1006 hlist_del_init_rcu(&ifp->addr_lst);
1007 spin_unlock_bh(&addrconf_hash_lock);
1009 write_lock_bh(&ifp->idev->lock);
1011 if (ifp->flags&IFA_F_TEMPORARY) {
1012 list_del(&ifp->tmp_list);
1014 in6_ifa_put(ifp->ifpub);
1020 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1021 action = check_cleanup_prefix_route(ifp, &expires);
1023 list_del_init(&ifp->if_list);
1026 write_unlock_bh(&ifp->idev->lock);
1028 addrconf_del_dad_work(ifp);
1030 ipv6_ifa_notify(RTM_DELADDR, ifp);
1032 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1034 if (action != CLEANUP_PREFIX_RT_NOP) {
1035 cleanup_prefix_route(ifp, expires,
1036 action == CLEANUP_PREFIX_RT_DEL);
1039 /* clean up prefsrc entries */
1040 rt6_remove_prefsrc(ifp);
1045 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1047 struct inet6_dev *idev = ifp->idev;
1048 struct in6_addr addr, *tmpaddr;
1049 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1050 unsigned long regen_advance;
1054 unsigned long now = jiffies;
1056 write_lock_bh(&idev->lock);
1058 spin_lock_bh(&ift->lock);
1059 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1060 spin_unlock_bh(&ift->lock);
1067 if (idev->cnf.use_tempaddr <= 0) {
1068 write_unlock_bh(&idev->lock);
1069 pr_info("%s: use_tempaddr is disabled\n", __func__);
1074 spin_lock_bh(&ifp->lock);
1075 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1076 idev->cnf.use_tempaddr = -1; /*XXX*/
1077 spin_unlock_bh(&ifp->lock);
1078 write_unlock_bh(&idev->lock);
1079 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1086 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1087 __ipv6_try_regen_rndid(idev, tmpaddr);
1088 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1089 age = (now - ifp->tstamp) / HZ;
1090 tmp_valid_lft = min_t(__u32,
1092 idev->cnf.temp_valid_lft + age);
1093 tmp_prefered_lft = min_t(__u32,
1095 idev->cnf.temp_prefered_lft + age -
1096 idev->cnf.max_desync_factor);
1097 tmp_plen = ifp->prefix_len;
1098 tmp_tstamp = ifp->tstamp;
1099 spin_unlock_bh(&ifp->lock);
1101 regen_advance = idev->cnf.regen_max_retry *
1102 idev->cnf.dad_transmits *
1103 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1104 write_unlock_bh(&idev->lock);
1106 /* A temporary address is created only if this calculated Preferred
1107 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1108 * an implementation must not create a temporary address with a zero
1109 * Preferred Lifetime.
1110 * Use age calculation as in addrconf_verify to avoid unnecessary
1111 * temporary addresses being generated.
1113 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1114 if (tmp_prefered_lft <= regen_advance + age) {
1121 addr_flags = IFA_F_TEMPORARY;
1122 /* set in addrconf_prefix_rcv() */
1123 if (ifp->flags & IFA_F_OPTIMISTIC)
1124 addr_flags |= IFA_F_OPTIMISTIC;
1126 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1127 ipv6_addr_scope(&addr), addr_flags,
1128 tmp_valid_lft, tmp_prefered_lft);
1132 pr_info("%s: retry temporary address regeneration\n", __func__);
1134 write_lock_bh(&idev->lock);
1138 spin_lock_bh(&ift->lock);
1141 ift->tstamp = tmp_tstamp;
1142 spin_unlock_bh(&ift->lock);
1144 addrconf_dad_start(ift);
1152 * Choose an appropriate source address (RFC3484)
1155 IPV6_SADDR_RULE_INIT = 0,
1156 IPV6_SADDR_RULE_LOCAL,
1157 IPV6_SADDR_RULE_SCOPE,
1158 IPV6_SADDR_RULE_PREFERRED,
1159 #ifdef CONFIG_IPV6_MIP6
1160 IPV6_SADDR_RULE_HOA,
1162 IPV6_SADDR_RULE_OIF,
1163 IPV6_SADDR_RULE_LABEL,
1164 IPV6_SADDR_RULE_PRIVACY,
1165 IPV6_SADDR_RULE_ORCHID,
1166 IPV6_SADDR_RULE_PREFIX,
1170 struct ipv6_saddr_score {
1173 struct inet6_ifaddr *ifa;
1174 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1179 struct ipv6_saddr_dst {
1180 const struct in6_addr *addr;
1187 static inline int ipv6_saddr_preferred(int type)
1189 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1194 static int ipv6_get_saddr_eval(struct net *net,
1195 struct ipv6_saddr_score *score,
1196 struct ipv6_saddr_dst *dst,
1201 if (i <= score->rule) {
1203 case IPV6_SADDR_RULE_SCOPE:
1204 ret = score->scopedist;
1206 case IPV6_SADDR_RULE_PREFIX:
1207 ret = score->matchlen;
1210 ret = !!test_bit(i, score->scorebits);
1216 case IPV6_SADDR_RULE_INIT:
1217 /* Rule 0: remember if hiscore is not ready yet */
1220 case IPV6_SADDR_RULE_LOCAL:
1221 /* Rule 1: Prefer same address */
1222 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1224 case IPV6_SADDR_RULE_SCOPE:
1225 /* Rule 2: Prefer appropriate scope
1230 * ---+--+-+---> scope
1232 * | d is scope of the destination.
1234 * | \ <- smaller scope is better if
1235 * B-15 | \ if scope is enough for destination.
1236 * | ret = B - scope (-1 <= scope >= d <= 15).
1238 * |/ <- greater is better
1239 * -C / if scope is not enough for destination.
1240 * /| ret = scope - C (-1 <= d < scope <= 15).
1242 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1243 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1244 * Assume B = 0 and we get C > 29.
1246 ret = __ipv6_addr_src_scope(score->addr_type);
1247 if (ret >= dst->scope)
1250 ret -= 128; /* 30 is enough */
1251 score->scopedist = ret;
1253 case IPV6_SADDR_RULE_PREFERRED:
1254 /* Rule 3: Avoid deprecated and optimistic addresses */
1255 ret = ipv6_saddr_preferred(score->addr_type) ||
1256 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1258 #ifdef CONFIG_IPV6_MIP6
1259 case IPV6_SADDR_RULE_HOA:
1261 /* Rule 4: Prefer home address */
1262 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1263 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1267 case IPV6_SADDR_RULE_OIF:
1268 /* Rule 5: Prefer outgoing interface */
1269 ret = (!dst->ifindex ||
1270 dst->ifindex == score->ifa->idev->dev->ifindex);
1272 case IPV6_SADDR_RULE_LABEL:
1273 /* Rule 6: Prefer matching label */
1274 ret = ipv6_addr_label(net,
1275 &score->ifa->addr, score->addr_type,
1276 score->ifa->idev->dev->ifindex) == dst->label;
1278 case IPV6_SADDR_RULE_PRIVACY:
1280 /* Rule 7: Prefer public address
1281 * Note: prefer temporary address if use_tempaddr >= 2
1283 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1284 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1285 score->ifa->idev->cnf.use_tempaddr >= 2;
1286 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1289 case IPV6_SADDR_RULE_ORCHID:
1290 /* Rule 8-: Prefer ORCHID vs ORCHID or
1291 * non-ORCHID vs non-ORCHID
1293 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1294 ipv6_addr_orchid(dst->addr));
1296 case IPV6_SADDR_RULE_PREFIX:
1297 /* Rule 8: Use longest matching prefix */
1298 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1299 if (ret > score->ifa->prefix_len)
1300 ret = score->ifa->prefix_len;
1301 score->matchlen = ret;
1308 __set_bit(i, score->scorebits);
1314 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1315 const struct in6_addr *daddr, unsigned int prefs,
1316 struct in6_addr *saddr)
1318 struct ipv6_saddr_score scores[2],
1319 *score = &scores[0], *hiscore = &scores[1];
1320 struct ipv6_saddr_dst dst;
1321 struct net_device *dev;
1324 dst_type = __ipv6_addr_type(daddr);
1326 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1327 dst.scope = __ipv6_addr_src_scope(dst_type);
1328 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1332 hiscore->ifa = NULL;
1336 for_each_netdev_rcu(net, dev) {
1337 struct inet6_dev *idev;
1339 /* Candidate Source Address (section 4)
1340 * - multicast and link-local destination address,
1341 * the set of candidate source address MUST only
1342 * include addresses assigned to interfaces
1343 * belonging to the same link as the outgoing
1345 * (- For site-local destination addresses, the
1346 * set of candidate source addresses MUST only
1347 * include addresses assigned to interfaces
1348 * belonging to the same site as the outgoing
1351 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1352 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1353 dst.ifindex && dev->ifindex != dst.ifindex)
1356 idev = __in6_dev_get(dev);
1360 read_lock_bh(&idev->lock);
1361 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1365 * - Tentative Address (RFC2462 section 5.4)
1366 * - A tentative address is not considered
1367 * "assigned to an interface" in the traditional
1368 * sense, unless it is also flagged as optimistic.
1369 * - Candidate Source Address (section 4)
1370 * - In any case, anycast addresses, multicast
1371 * addresses, and the unspecified address MUST
1372 * NOT be included in a candidate set.
1374 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1375 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1378 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1380 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1381 score->addr_type & IPV6_ADDR_MULTICAST)) {
1382 LIMIT_NETDEBUG(KERN_DEBUG
1383 "ADDRCONF: unspecified / multicast address "
1384 "assigned as unicast address on %s",
1390 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1392 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1393 int minihiscore, miniscore;
1395 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1396 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1398 if (minihiscore > miniscore) {
1399 if (i == IPV6_SADDR_RULE_SCOPE &&
1400 score->scopedist > 0) {
1403 * each remaining entry
1404 * has too small (not enough)
1405 * scope, because ifa entries
1406 * are sorted by their scope
1412 } else if (minihiscore < miniscore) {
1414 in6_ifa_put(hiscore->ifa);
1416 in6_ifa_hold(score->ifa);
1418 swap(hiscore, score);
1420 /* restore our iterator */
1421 score->ifa = hiscore->ifa;
1428 read_unlock_bh(&idev->lock);
1433 return -EADDRNOTAVAIL;
1435 *saddr = hiscore->ifa->addr;
1436 in6_ifa_put(hiscore->ifa);
1439 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1441 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1444 struct inet6_ifaddr *ifp;
1445 int err = -EADDRNOTAVAIL;
1447 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1448 if (ifp->scope > IFA_LINK)
1450 if (ifp->scope == IFA_LINK &&
1451 !(ifp->flags & banned_flags)) {
1460 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1463 struct inet6_dev *idev;
1464 int err = -EADDRNOTAVAIL;
1467 idev = __in6_dev_get(dev);
1469 read_lock_bh(&idev->lock);
1470 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1471 read_unlock_bh(&idev->lock);
1477 static int ipv6_count_addresses(struct inet6_dev *idev)
1480 struct inet6_ifaddr *ifp;
1482 read_lock_bh(&idev->lock);
1483 list_for_each_entry(ifp, &idev->addr_list, if_list)
1485 read_unlock_bh(&idev->lock);
1489 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1490 const struct net_device *dev, int strict)
1492 struct inet6_ifaddr *ifp;
1493 unsigned int hash = inet6_addr_hash(addr);
1496 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1497 if (!net_eq(dev_net(ifp->idev->dev), net))
1499 if (ipv6_addr_equal(&ifp->addr, addr) &&
1500 !(ifp->flags&IFA_F_TENTATIVE) &&
1501 (dev == NULL || ifp->idev->dev == dev ||
1502 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1503 rcu_read_unlock_bh();
1508 rcu_read_unlock_bh();
1511 EXPORT_SYMBOL(ipv6_chk_addr);
1513 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1514 struct net_device *dev)
1516 unsigned int hash = inet6_addr_hash(addr);
1517 struct inet6_ifaddr *ifp;
1519 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1520 if (!net_eq(dev_net(ifp->idev->dev), net))
1522 if (ipv6_addr_equal(&ifp->addr, addr)) {
1523 if (dev == NULL || ifp->idev->dev == dev)
1530 /* Compares an address/prefix_len with addresses on device @dev.
1531 * If one is found it returns true.
1533 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1534 const unsigned int prefix_len, struct net_device *dev)
1536 struct inet6_dev *idev;
1537 struct inet6_ifaddr *ifa;
1541 idev = __in6_dev_get(dev);
1543 read_lock_bh(&idev->lock);
1544 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1545 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1549 read_unlock_bh(&idev->lock);
1555 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1557 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1559 struct inet6_dev *idev;
1560 struct inet6_ifaddr *ifa;
1565 idev = __in6_dev_get(dev);
1567 read_lock_bh(&idev->lock);
1568 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1569 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1574 read_unlock_bh(&idev->lock);
1579 EXPORT_SYMBOL(ipv6_chk_prefix);
1581 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1582 struct net_device *dev, int strict)
1584 struct inet6_ifaddr *ifp, *result = NULL;
1585 unsigned int hash = inet6_addr_hash(addr);
1588 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1589 if (!net_eq(dev_net(ifp->idev->dev), net))
1591 if (ipv6_addr_equal(&ifp->addr, addr)) {
1592 if (dev == NULL || ifp->idev->dev == dev ||
1593 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1600 rcu_read_unlock_bh();
1605 /* Gets referenced address, destroys ifaddr */
1607 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1609 if (ifp->flags&IFA_F_PERMANENT) {
1610 spin_lock_bh(&ifp->lock);
1611 addrconf_del_dad_work(ifp);
1612 ifp->flags |= IFA_F_TENTATIVE;
1614 ifp->flags |= IFA_F_DADFAILED;
1615 spin_unlock_bh(&ifp->lock);
1617 ipv6_ifa_notify(0, ifp);
1619 } else if (ifp->flags&IFA_F_TEMPORARY) {
1620 struct inet6_ifaddr *ifpub;
1621 spin_lock_bh(&ifp->lock);
1624 in6_ifa_hold(ifpub);
1625 spin_unlock_bh(&ifp->lock);
1626 ipv6_create_tempaddr(ifpub, ifp);
1629 spin_unlock_bh(&ifp->lock);
1637 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1641 spin_lock_bh(&ifp->state_lock);
1642 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1643 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1646 spin_unlock_bh(&ifp->state_lock);
1651 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1653 struct inet6_dev *idev = ifp->idev;
1655 if (addrconf_dad_end(ifp)) {
1660 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1661 ifp->idev->dev->name, &ifp->addr);
1663 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1664 struct in6_addr addr;
1666 addr.s6_addr32[0] = htonl(0xfe800000);
1667 addr.s6_addr32[1] = 0;
1669 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1670 ipv6_addr_equal(&ifp->addr, &addr)) {
1671 /* DAD failed for link-local based on MAC address */
1672 idev->cnf.disable_ipv6 = 1;
1674 pr_info("%s: IPv6 being disabled!\n",
1675 ifp->idev->dev->name);
1679 spin_lock_bh(&ifp->state_lock);
1680 /* transition from _POSTDAD to _ERRDAD */
1681 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1682 spin_unlock_bh(&ifp->state_lock);
1684 addrconf_mod_dad_work(ifp, 0);
1687 /* Join to solicited addr multicast group.
1688 * caller must hold RTNL */
1689 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1691 struct in6_addr maddr;
1693 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1696 addrconf_addr_solict_mult(addr, &maddr);
1697 ipv6_dev_mc_inc(dev, &maddr);
1700 /* caller must hold RTNL */
1701 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1703 struct in6_addr maddr;
1705 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1708 addrconf_addr_solict_mult(addr, &maddr);
1709 __ipv6_dev_mc_dec(idev, &maddr);
1712 /* caller must hold RTNL */
1713 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1715 struct in6_addr addr;
1717 if (ifp->prefix_len >= 127) /* RFC 6164 */
1719 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1720 if (ipv6_addr_any(&addr))
1722 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1725 /* caller must hold RTNL */
1726 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1728 struct in6_addr addr;
1730 if (ifp->prefix_len >= 127) /* RFC 6164 */
1732 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1733 if (ipv6_addr_any(&addr))
1735 __ipv6_dev_ac_dec(ifp->idev, &addr);
1738 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1740 if (dev->addr_len != ETH_ALEN)
1742 memcpy(eui, dev->dev_addr, 3);
1743 memcpy(eui + 5, dev->dev_addr + 3, 3);
1746 * The zSeries OSA network cards can be shared among various
1747 * OS instances, but the OSA cards have only one MAC address.
1748 * This leads to duplicate address conflicts in conjunction
1749 * with IPv6 if more than one instance uses the same card.
1751 * The driver for these cards can deliver a unique 16-bit
1752 * identifier for each instance sharing the same card. It is
1753 * placed instead of 0xFFFE in the interface identifier. The
1754 * "u" bit of the interface identifier is not inverted in this
1755 * case. Hence the resulting interface identifier has local
1756 * scope according to RFC2373.
1759 eui[3] = (dev->dev_id >> 8) & 0xFF;
1760 eui[4] = dev->dev_id & 0xFF;
1769 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1771 if (dev->addr_len != IEEE802154_ADDR_LEN)
1773 memcpy(eui, dev->dev_addr, 8);
1778 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1780 union fwnet_hwaddr *ha;
1782 if (dev->addr_len != FWNET_ALEN)
1785 ha = (union fwnet_hwaddr *)dev->dev_addr;
1787 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1792 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1794 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1795 if (dev->addr_len != ARCNET_ALEN)
1798 eui[7] = *(u8 *)dev->dev_addr;
1802 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1804 if (dev->addr_len != INFINIBAND_ALEN)
1806 memcpy(eui, dev->dev_addr + 12, 8);
1811 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1815 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1816 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1817 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1818 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1819 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1820 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1824 memcpy(eui + 4, &addr, 4);
1828 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1830 if (dev->priv_flags & IFF_ISATAP)
1831 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1835 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1837 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1840 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1842 memcpy(eui, dev->perm_addr, 3);
1843 memcpy(eui + 5, dev->perm_addr + 3, 3);
1850 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1852 switch (dev->type) {
1855 return addrconf_ifid_eui48(eui, dev);
1857 return addrconf_ifid_arcnet(eui, dev);
1858 case ARPHRD_INFINIBAND:
1859 return addrconf_ifid_infiniband(eui, dev);
1861 return addrconf_ifid_sit(eui, dev);
1863 return addrconf_ifid_gre(eui, dev);
1864 case ARPHRD_6LOWPAN:
1865 case ARPHRD_IEEE802154:
1866 return addrconf_ifid_eui64(eui, dev);
1867 case ARPHRD_IEEE1394:
1868 return addrconf_ifid_ieee1394(eui, dev);
1869 case ARPHRD_TUNNEL6:
1870 return addrconf_ifid_ip6tnl(eui, dev);
1875 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1878 struct inet6_ifaddr *ifp;
1880 read_lock_bh(&idev->lock);
1881 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1882 if (ifp->scope > IFA_LINK)
1884 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1885 memcpy(eui, ifp->addr.s6_addr+8, 8);
1890 read_unlock_bh(&idev->lock);
1894 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1895 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1898 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1899 idev->rndid[0] &= ~0x02;
1902 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1903 * check if generated address is not inappropriate
1905 * - Reserved subnet anycast (RFC 2526)
1906 * 11111101 11....11 1xxxxxxx
1907 * - ISATAP (RFC4214) 6.1
1908 * 00-00-5E-FE-xx-xx-xx-xx
1910 * - XXX: already assigned to an address on the device
1912 if (idev->rndid[0] == 0xfd &&
1913 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1914 (idev->rndid[7]&0x80))
1916 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1917 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1919 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1924 static void ipv6_regen_rndid(unsigned long data)
1926 struct inet6_dev *idev = (struct inet6_dev *) data;
1927 unsigned long expires;
1930 write_lock_bh(&idev->lock);
1935 __ipv6_regen_rndid(idev);
1938 idev->cnf.temp_prefered_lft * HZ -
1939 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1940 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1941 idev->cnf.max_desync_factor * HZ;
1942 if (time_before(expires, jiffies)) {
1943 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1944 __func__, idev->dev->name);
1948 if (!mod_timer(&idev->regen_timer, expires))
1952 write_unlock_bh(&idev->lock);
1953 rcu_read_unlock_bh();
1957 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1959 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1960 __ipv6_regen_rndid(idev);
1968 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1969 unsigned long expires, u32 flags)
1971 struct fib6_config cfg = {
1972 .fc_table = RT6_TABLE_PREFIX,
1973 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1974 .fc_ifindex = dev->ifindex,
1975 .fc_expires = expires,
1977 .fc_flags = RTF_UP | flags,
1978 .fc_nlinfo.nl_net = dev_net(dev),
1979 .fc_protocol = RTPROT_KERNEL,
1984 /* Prevent useless cloning on PtP SIT.
1985 This thing is done here expecting that the whole
1986 class of non-broadcast devices need not cloning.
1988 #if IS_ENABLED(CONFIG_IPV6_SIT)
1989 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1990 cfg.fc_flags |= RTF_NONEXTHOP;
1993 ip6_route_add(&cfg);
1997 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1999 const struct net_device *dev,
2000 u32 flags, u32 noflags)
2002 struct fib6_node *fn;
2003 struct rt6_info *rt = NULL;
2004 struct fib6_table *table;
2006 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2010 read_lock_bh(&table->tb6_lock);
2011 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2014 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2015 if (rt->dst.dev->ifindex != dev->ifindex)
2017 if ((rt->rt6i_flags & flags) != flags)
2019 if ((rt->rt6i_flags & noflags) != 0)
2025 read_unlock_bh(&table->tb6_lock);
2030 /* Create "default" multicast route to the interface */
2032 static void addrconf_add_mroute(struct net_device *dev)
2034 struct fib6_config cfg = {
2035 .fc_table = RT6_TABLE_LOCAL,
2036 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2037 .fc_ifindex = dev->ifindex,
2040 .fc_nlinfo.nl_net = dev_net(dev),
2043 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2045 ip6_route_add(&cfg);
2048 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2050 struct inet6_dev *idev;
2054 idev = ipv6_find_idev(dev);
2056 return ERR_PTR(-ENOBUFS);
2058 if (idev->cnf.disable_ipv6)
2059 return ERR_PTR(-EACCES);
2061 /* Add default multicast route */
2062 if (!(dev->flags & IFF_LOOPBACK))
2063 addrconf_add_mroute(dev);
2068 static void manage_tempaddrs(struct inet6_dev *idev,
2069 struct inet6_ifaddr *ifp,
2070 __u32 valid_lft, __u32 prefered_lft,
2071 bool create, unsigned long now)
2074 struct inet6_ifaddr *ift;
2076 read_lock_bh(&idev->lock);
2077 /* update all temporary addresses in the list */
2078 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2079 int age, max_valid, max_prefered;
2081 if (ifp != ift->ifpub)
2084 /* RFC 4941 section 3.3:
2085 * If a received option will extend the lifetime of a public
2086 * address, the lifetimes of temporary addresses should
2087 * be extended, subject to the overall constraint that no
2088 * temporary addresses should ever remain "valid" or "preferred"
2089 * for a time longer than (TEMP_VALID_LIFETIME) or
2090 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2092 age = (now - ift->cstamp) / HZ;
2093 max_valid = idev->cnf.temp_valid_lft - age;
2097 max_prefered = idev->cnf.temp_prefered_lft -
2098 idev->cnf.max_desync_factor - age;
2099 if (max_prefered < 0)
2102 if (valid_lft > max_valid)
2103 valid_lft = max_valid;
2105 if (prefered_lft > max_prefered)
2106 prefered_lft = max_prefered;
2108 spin_lock(&ift->lock);
2110 ift->valid_lft = valid_lft;
2111 ift->prefered_lft = prefered_lft;
2113 if (prefered_lft > 0)
2114 ift->flags &= ~IFA_F_DEPRECATED;
2116 spin_unlock(&ift->lock);
2117 if (!(flags&IFA_F_TENTATIVE))
2118 ipv6_ifa_notify(0, ift);
2121 if ((create || list_empty(&idev->tempaddr_list)) &&
2122 idev->cnf.use_tempaddr > 0) {
2123 /* When a new public address is created as described
2124 * in [ADDRCONF], also create a new temporary address.
2125 * Also create a temporary address if it's enabled but
2126 * no temporary address currently exists.
2128 read_unlock_bh(&idev->lock);
2129 ipv6_create_tempaddr(ifp, NULL);
2131 read_unlock_bh(&idev->lock);
2135 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2137 struct prefix_info *pinfo;
2141 struct inet6_dev *in6_dev;
2142 struct net *net = dev_net(dev);
2144 pinfo = (struct prefix_info *) opt;
2146 if (len < sizeof(struct prefix_info)) {
2147 ADBG("addrconf: prefix option too short\n");
2152 * Validation checks ([ADDRCONF], page 19)
2155 addr_type = ipv6_addr_type(&pinfo->prefix);
2157 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2160 valid_lft = ntohl(pinfo->valid);
2161 prefered_lft = ntohl(pinfo->prefered);
2163 if (prefered_lft > valid_lft) {
2164 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2168 in6_dev = in6_dev_get(dev);
2170 if (in6_dev == NULL) {
2171 net_dbg_ratelimited("addrconf: device %s not configured\n",
2177 * Two things going on here:
2178 * 1) Add routes for on-link prefixes
2179 * 2) Configure prefixes with the auto flag set
2182 if (pinfo->onlink) {
2183 struct rt6_info *rt;
2184 unsigned long rt_expires;
2186 /* Avoid arithmetic overflow. Really, we could
2187 * save rt_expires in seconds, likely valid_lft,
2188 * but it would require division in fib gc, that it
2192 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2194 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2196 if (addrconf_finite_timeout(rt_expires))
2199 rt = addrconf_get_prefix_route(&pinfo->prefix,
2202 RTF_ADDRCONF | RTF_PREFIX_RT,
2203 RTF_GATEWAY | RTF_DEFAULT);
2206 /* Autoconf prefix route */
2207 if (valid_lft == 0) {
2210 } else if (addrconf_finite_timeout(rt_expires)) {
2212 rt6_set_expires(rt, jiffies + rt_expires);
2214 rt6_clean_expires(rt);
2216 } else if (valid_lft) {
2217 clock_t expires = 0;
2218 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2219 if (addrconf_finite_timeout(rt_expires)) {
2221 flags |= RTF_EXPIRES;
2222 expires = jiffies_to_clock_t(rt_expires);
2224 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2225 dev, expires, flags);
2230 /* Try to figure out our local address for this prefix */
2232 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2233 struct inet6_ifaddr *ifp;
2234 struct in6_addr addr;
2235 int create = 0, update_lft = 0;
2236 bool tokenized = false;
2238 if (pinfo->prefix_len == 64) {
2239 memcpy(&addr, &pinfo->prefix, 8);
2241 if (!ipv6_addr_any(&in6_dev->token)) {
2242 read_lock_bh(&in6_dev->lock);
2243 memcpy(addr.s6_addr + 8,
2244 in6_dev->token.s6_addr + 8, 8);
2245 read_unlock_bh(&in6_dev->lock);
2247 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2248 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2249 in6_dev_put(in6_dev);
2254 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2256 in6_dev_put(in6_dev);
2261 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2263 if (ifp == NULL && valid_lft) {
2264 int max_addresses = in6_dev->cnf.max_addresses;
2267 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2268 if (in6_dev->cnf.optimistic_dad &&
2269 !net->ipv6.devconf_all->forwarding && sllao)
2270 addr_flags = IFA_F_OPTIMISTIC;
2273 /* Do not allow to create too much of autoconfigured
2274 * addresses; this would be too easy way to crash kernel.
2276 if (!max_addresses ||
2277 ipv6_count_addresses(in6_dev) < max_addresses)
2278 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2280 addr_type&IPV6_ADDR_SCOPE_MASK,
2281 addr_flags, valid_lft,
2284 if (IS_ERR_OR_NULL(ifp)) {
2285 in6_dev_put(in6_dev);
2291 spin_lock_bh(&ifp->lock);
2292 ifp->flags |= IFA_F_MANAGETEMPADDR;
2293 ifp->cstamp = jiffies;
2294 ifp->tokenized = tokenized;
2295 spin_unlock_bh(&ifp->lock);
2296 addrconf_dad_start(ifp);
2304 /* update lifetime (RFC2462 5.5.3 e) */
2305 spin_lock(&ifp->lock);
2307 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2308 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2311 if (!update_lft && !create && stored_lft) {
2312 const u32 minimum_lft = min(
2313 stored_lft, (u32)MIN_VALID_LIFETIME);
2314 valid_lft = max(valid_lft, minimum_lft);
2316 /* RFC4862 Section 5.5.3e:
2317 * "Note that the preferred lifetime of the
2318 * corresponding address is always reset to
2319 * the Preferred Lifetime in the received
2320 * Prefix Information option, regardless of
2321 * whether the valid lifetime is also reset or
2324 * So we should always update prefered_lft here.
2330 ifp->valid_lft = valid_lft;
2331 ifp->prefered_lft = prefered_lft;
2334 ifp->flags &= ~IFA_F_DEPRECATED;
2335 spin_unlock(&ifp->lock);
2337 if (!(flags&IFA_F_TENTATIVE))
2338 ipv6_ifa_notify(0, ifp);
2340 spin_unlock(&ifp->lock);
2342 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2349 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2350 in6_dev_put(in6_dev);
2354 * Set destination address.
2355 * Special case for SIT interfaces where we create a new "virtual"
2358 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2360 struct in6_ifreq ireq;
2361 struct net_device *dev;
2367 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2370 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2376 #if IS_ENABLED(CONFIG_IPV6_SIT)
2377 if (dev->type == ARPHRD_SIT) {
2378 const struct net_device_ops *ops = dev->netdev_ops;
2380 struct ip_tunnel_parm p;
2382 err = -EADDRNOTAVAIL;
2383 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2386 memset(&p, 0, sizeof(p));
2387 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2391 p.iph.protocol = IPPROTO_IPV6;
2393 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2395 if (ops->ndo_do_ioctl) {
2396 mm_segment_t oldfs = get_fs();
2399 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2406 dev = __dev_get_by_name(net, p.name);
2409 err = dev_open(dev);
2420 * Manual configuration of address on an interface
2422 static int inet6_addr_add(struct net *net, int ifindex,
2423 const struct in6_addr *pfx,
2424 const struct in6_addr *peer_pfx,
2425 unsigned int plen, __u32 ifa_flags,
2426 __u32 prefered_lft, __u32 valid_lft)
2428 struct inet6_ifaddr *ifp;
2429 struct inet6_dev *idev;
2430 struct net_device *dev;
2434 unsigned long timeout;
2441 /* check the lifetime */
2442 if (!valid_lft || prefered_lft > valid_lft)
2445 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2448 dev = __dev_get_by_index(net, ifindex);
2452 idev = addrconf_add_dev(dev);
2454 return PTR_ERR(idev);
2456 scope = ipv6_addr_scope(pfx);
2458 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2459 if (addrconf_finite_timeout(timeout)) {
2460 expires = jiffies_to_clock_t(timeout * HZ);
2461 valid_lft = timeout;
2462 flags = RTF_EXPIRES;
2466 ifa_flags |= IFA_F_PERMANENT;
2469 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2470 if (addrconf_finite_timeout(timeout)) {
2472 ifa_flags |= IFA_F_DEPRECATED;
2473 prefered_lft = timeout;
2476 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2477 valid_lft, prefered_lft);
2480 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2481 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2486 * Note that section 3.1 of RFC 4429 indicates
2487 * that the Optimistic flag should not be set for
2488 * manually configured addresses
2490 addrconf_dad_start(ifp);
2491 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2492 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2495 addrconf_verify_rtnl();
2499 return PTR_ERR(ifp);
2502 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2505 struct inet6_ifaddr *ifp;
2506 struct inet6_dev *idev;
2507 struct net_device *dev;
2512 dev = __dev_get_by_index(net, ifindex);
2516 if ((idev = __in6_dev_get(dev)) == NULL)
2519 read_lock_bh(&idev->lock);
2520 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2521 if (ifp->prefix_len == plen &&
2522 ipv6_addr_equal(pfx, &ifp->addr)) {
2524 read_unlock_bh(&idev->lock);
2530 read_unlock_bh(&idev->lock);
2531 return -EADDRNOTAVAIL;
2535 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2537 struct in6_ifreq ireq;
2540 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2543 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2547 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2548 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2549 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2554 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2556 struct in6_ifreq ireq;
2559 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2562 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2566 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2567 ireq.ifr6_prefixlen);
2572 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2573 int plen, int scope)
2575 struct inet6_ifaddr *ifp;
2577 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2578 scope, IFA_F_PERMANENT,
2579 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2581 spin_lock_bh(&ifp->lock);
2582 ifp->flags &= ~IFA_F_TENTATIVE;
2583 spin_unlock_bh(&ifp->lock);
2584 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2589 #if IS_ENABLED(CONFIG_IPV6_SIT)
2590 static void sit_add_v4_addrs(struct inet6_dev *idev)
2592 struct in6_addr addr;
2593 struct net_device *dev;
2594 struct net *net = dev_net(idev->dev);
2600 memset(&addr, 0, sizeof(struct in6_addr));
2601 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2603 if (idev->dev->flags&IFF_POINTOPOINT) {
2604 addr.s6_addr32[0] = htonl(0xfe800000);
2608 scope = IPV6_ADDR_COMPATv4;
2610 pflags |= RTF_NONEXTHOP;
2613 if (addr.s6_addr32[3]) {
2614 add_addr(idev, &addr, plen, scope);
2615 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2619 for_each_netdev(net, dev) {
2620 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2621 if (in_dev && (dev->flags & IFF_UP)) {
2622 struct in_ifaddr *ifa;
2626 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2628 addr.s6_addr32[3] = ifa->ifa_local;
2630 if (ifa->ifa_scope == RT_SCOPE_LINK)
2632 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2633 if (idev->dev->flags&IFF_POINTOPOINT)
2638 add_addr(idev, &addr, plen, flag);
2639 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2647 static void init_loopback(struct net_device *dev)
2649 struct inet6_dev *idev;
2650 struct net_device *sp_dev;
2651 struct inet6_ifaddr *sp_ifa;
2652 struct rt6_info *sp_rt;
2658 if ((idev = ipv6_find_idev(dev)) == NULL) {
2659 pr_debug("%s: add_dev failed\n", __func__);
2663 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2665 /* Add routes to other interface's IPv6 addresses */
2666 for_each_netdev(dev_net(dev), sp_dev) {
2667 if (!strcmp(sp_dev->name, dev->name))
2670 idev = __in6_dev_get(sp_dev);
2674 read_lock_bh(&idev->lock);
2675 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2677 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2681 /* This dst has been added to garbage list when
2682 * lo device down, release this obsolete dst and
2683 * reallocate a new router for ifa.
2685 if (sp_ifa->rt->dst.obsolete > 0) {
2686 ip6_rt_put(sp_ifa->rt);
2693 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2695 /* Failure cases are ignored */
2696 if (!IS_ERR(sp_rt)) {
2701 read_unlock_bh(&idev->lock);
2705 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2707 struct inet6_ifaddr *ifp;
2708 u32 addr_flags = IFA_F_PERMANENT;
2710 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2711 if (idev->cnf.optimistic_dad &&
2712 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2713 addr_flags |= IFA_F_OPTIMISTIC;
2717 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2718 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2720 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2721 addrconf_dad_start(ifp);
2726 static void addrconf_dev_config(struct net_device *dev)
2728 struct in6_addr addr;
2729 struct inet6_dev *idev;
2733 if ((dev->type != ARPHRD_ETHER) &&
2734 (dev->type != ARPHRD_FDDI) &&
2735 (dev->type != ARPHRD_ARCNET) &&
2736 (dev->type != ARPHRD_INFINIBAND) &&
2737 (dev->type != ARPHRD_IEEE802154) &&
2738 (dev->type != ARPHRD_IEEE1394) &&
2739 (dev->type != ARPHRD_TUNNEL6) &&
2740 (dev->type != ARPHRD_6LOWPAN)) {
2741 /* Alas, we support only Ethernet autoconfiguration. */
2745 idev = addrconf_add_dev(dev);
2749 memset(&addr, 0, sizeof(struct in6_addr));
2750 addr.s6_addr32[0] = htonl(0xFE800000);
2752 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2753 addrconf_add_linklocal(idev, &addr);
2756 #if IS_ENABLED(CONFIG_IPV6_SIT)
2757 static void addrconf_sit_config(struct net_device *dev)
2759 struct inet6_dev *idev;
2764 * Configure the tunnel with one of our IPv4
2765 * addresses... we should configure all of
2766 * our v4 addrs in the tunnel
2769 if ((idev = ipv6_find_idev(dev)) == NULL) {
2770 pr_debug("%s: add_dev failed\n", __func__);
2774 if (dev->priv_flags & IFF_ISATAP) {
2775 struct in6_addr addr;
2777 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2778 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2779 addrconf_add_linklocal(idev, &addr);
2783 sit_add_v4_addrs(idev);
2785 if (dev->flags&IFF_POINTOPOINT)
2786 addrconf_add_mroute(dev);
2790 #if IS_ENABLED(CONFIG_NET_IPGRE)
2791 static void addrconf_gre_config(struct net_device *dev)
2793 struct inet6_dev *idev;
2794 struct in6_addr addr;
2798 if ((idev = ipv6_find_idev(dev)) == NULL) {
2799 pr_debug("%s: add_dev failed\n", __func__);
2803 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2804 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2805 addrconf_add_linklocal(idev, &addr);
2807 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2812 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2814 struct in6_addr lladdr;
2816 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2817 addrconf_add_linklocal(idev, &lladdr);
2823 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2826 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2827 struct inet6_dev *idev = __in6_dev_get(dev);
2828 int run_pending = 0;
2832 case NETDEV_REGISTER:
2833 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2834 idev = ipv6_add_dev(dev);
2836 return notifier_from_errno(-ENOMEM);
2842 if (dev->flags & IFF_SLAVE)
2845 if (event == NETDEV_UP) {
2846 if (!addrconf_qdisc_ok(dev)) {
2847 /* device is not ready yet. */
2848 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2853 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2854 idev = ipv6_add_dev(dev);
2857 idev->if_flags |= IF_READY;
2861 if (!addrconf_qdisc_ok(dev)) {
2862 /* device is still not ready. */
2867 if (idev->if_flags & IF_READY)
2868 /* device is already configured. */
2870 idev->if_flags |= IF_READY;
2873 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2879 switch (dev->type) {
2880 #if IS_ENABLED(CONFIG_IPV6_SIT)
2882 addrconf_sit_config(dev);
2885 #if IS_ENABLED(CONFIG_NET_IPGRE)
2887 addrconf_gre_config(dev);
2890 case ARPHRD_LOOPBACK:
2895 addrconf_dev_config(dev);
2901 addrconf_dad_run(idev);
2904 * If the MTU changed during the interface down,
2905 * when the interface up, the changed MTU must be
2906 * reflected in the idev as well as routers.
2908 if (idev->cnf.mtu6 != dev->mtu &&
2909 dev->mtu >= IPV6_MIN_MTU) {
2910 rt6_mtu_change(dev, dev->mtu);
2911 idev->cnf.mtu6 = dev->mtu;
2913 idev->tstamp = jiffies;
2914 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2917 * If the changed mtu during down is lower than
2918 * IPV6_MIN_MTU stop IPv6 on this interface.
2920 if (dev->mtu < IPV6_MIN_MTU)
2921 addrconf_ifdown(dev, 1);
2925 case NETDEV_CHANGEMTU:
2926 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2927 rt6_mtu_change(dev, dev->mtu);
2928 idev->cnf.mtu6 = dev->mtu;
2932 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2933 idev = ipv6_add_dev(dev);
2939 * if MTU under IPV6_MIN_MTU.
2940 * Stop IPv6 on this interface.
2944 case NETDEV_UNREGISTER:
2946 * Remove all addresses from this interface.
2948 addrconf_ifdown(dev, event != NETDEV_DOWN);
2951 case NETDEV_CHANGENAME:
2953 snmp6_unregister_dev(idev);
2954 addrconf_sysctl_unregister(idev);
2955 addrconf_sysctl_register(idev);
2956 err = snmp6_register_dev(idev);
2958 return notifier_from_errno(err);
2962 case NETDEV_PRE_TYPE_CHANGE:
2963 case NETDEV_POST_TYPE_CHANGE:
2964 addrconf_type_change(dev, event);
2972 * addrconf module should be notified of a device going up
2974 static struct notifier_block ipv6_dev_notf = {
2975 .notifier_call = addrconf_notify,
2978 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2980 struct inet6_dev *idev;
2983 idev = __in6_dev_get(dev);
2985 if (event == NETDEV_POST_TYPE_CHANGE)
2986 ipv6_mc_remap(idev);
2987 else if (event == NETDEV_PRE_TYPE_CHANGE)
2988 ipv6_mc_unmap(idev);
2991 static int addrconf_ifdown(struct net_device *dev, int how)
2993 struct net *net = dev_net(dev);
2994 struct inet6_dev *idev;
2995 struct inet6_ifaddr *ifa;
3000 rt6_ifdown(net, dev);
3001 neigh_ifdown(&nd_tbl, dev);
3003 idev = __in6_dev_get(dev);
3008 * Step 1: remove reference to ipv6 device from parent device.
3014 /* protected by rtnl_lock */
3015 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3017 /* Step 1.5: remove snmp6 entry */
3018 snmp6_unregister_dev(idev);
3022 /* Step 2: clear hash table */
3023 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3024 struct hlist_head *h = &inet6_addr_lst[i];
3026 spin_lock_bh(&addrconf_hash_lock);
3028 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3029 if (ifa->idev == idev) {
3030 hlist_del_init_rcu(&ifa->addr_lst);
3031 addrconf_del_dad_work(ifa);
3035 spin_unlock_bh(&addrconf_hash_lock);
3038 write_lock_bh(&idev->lock);
3040 addrconf_del_rs_timer(idev);
3042 /* Step 2: clear flags for stateless addrconf */
3044 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3046 if (how && del_timer(&idev->regen_timer))
3049 /* Step 3: clear tempaddr list */
3050 while (!list_empty(&idev->tempaddr_list)) {
3051 ifa = list_first_entry(&idev->tempaddr_list,
3052 struct inet6_ifaddr, tmp_list);
3053 list_del(&ifa->tmp_list);
3054 write_unlock_bh(&idev->lock);
3055 spin_lock_bh(&ifa->lock);
3058 in6_ifa_put(ifa->ifpub);
3061 spin_unlock_bh(&ifa->lock);
3063 write_lock_bh(&idev->lock);
3066 while (!list_empty(&idev->addr_list)) {
3067 ifa = list_first_entry(&idev->addr_list,
3068 struct inet6_ifaddr, if_list);
3069 addrconf_del_dad_work(ifa);
3071 list_del(&ifa->if_list);
3073 write_unlock_bh(&idev->lock);
3075 spin_lock_bh(&ifa->state_lock);
3077 ifa->state = INET6_IFADDR_STATE_DEAD;
3078 spin_unlock_bh(&ifa->state_lock);
3080 if (state != INET6_IFADDR_STATE_DEAD) {
3081 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3082 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3086 write_lock_bh(&idev->lock);
3089 write_unlock_bh(&idev->lock);
3091 /* Step 5: Discard multicast list */
3093 ipv6_mc_destroy_dev(idev);
3097 idev->tstamp = jiffies;
3099 /* Last: Shot the device (if unregistered) */
3101 addrconf_sysctl_unregister(idev);
3102 neigh_parms_release(&nd_tbl, idev->nd_parms);
3103 neigh_ifdown(&nd_tbl, dev);
3109 static void addrconf_rs_timer(unsigned long data)
3111 struct inet6_dev *idev = (struct inet6_dev *)data;
3112 struct net_device *dev = idev->dev;
3113 struct in6_addr lladdr;
3115 write_lock(&idev->lock);
3116 if (idev->dead || !(idev->if_flags & IF_READY))
3119 if (!ipv6_accept_ra(idev))
3122 /* Announcement received after solicitation was sent */
3123 if (idev->if_flags & IF_RA_RCVD)
3126 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3127 write_unlock(&idev->lock);
3128 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3129 ndisc_send_rs(dev, &lladdr,
3130 &in6addr_linklocal_allrouters);
3134 write_lock(&idev->lock);
3135 /* The wait after the last probe can be shorter */
3136 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3137 idev->cnf.rtr_solicits) ?
3138 idev->cnf.rtr_solicit_delay :
3139 idev->cnf.rtr_solicit_interval);
3142 * Note: we do not support deprecated "all on-link"
3143 * assumption any longer.
3145 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3149 write_unlock(&idev->lock);
3155 * Duplicate Address Detection
3157 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3159 unsigned long rand_num;
3160 struct inet6_dev *idev = ifp->idev;
3162 if (ifp->flags & IFA_F_OPTIMISTIC)
3165 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3167 ifp->dad_probes = idev->cnf.dad_transmits;
3168 addrconf_mod_dad_work(ifp, rand_num);
3171 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3173 struct inet6_dev *idev = ifp->idev;
3174 struct net_device *dev = idev->dev;
3176 addrconf_join_solict(dev, &ifp->addr);
3178 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3180 read_lock_bh(&idev->lock);
3181 spin_lock(&ifp->lock);
3182 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3185 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3186 idev->cnf.accept_dad < 1 ||
3187 !(ifp->flags&IFA_F_TENTATIVE) ||
3188 ifp->flags & IFA_F_NODAD) {
3189 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3190 spin_unlock(&ifp->lock);
3191 read_unlock_bh(&idev->lock);
3193 addrconf_dad_completed(ifp);
3197 if (!(idev->if_flags & IF_READY)) {
3198 spin_unlock(&ifp->lock);
3199 read_unlock_bh(&idev->lock);
3201 * If the device is not ready:
3202 * - keep it tentative if it is a permanent address.
3203 * - otherwise, kill it.
3206 addrconf_dad_stop(ifp, 0);
3211 * Optimistic nodes can start receiving
3214 if (ifp->flags & IFA_F_OPTIMISTIC)
3215 ip6_ins_rt(ifp->rt);
3217 addrconf_dad_kick(ifp);
3219 spin_unlock(&ifp->lock);
3220 read_unlock_bh(&idev->lock);
3223 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3225 bool begin_dad = false;
3227 spin_lock_bh(&ifp->state_lock);
3228 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3229 ifp->state = INET6_IFADDR_STATE_PREDAD;
3232 spin_unlock_bh(&ifp->state_lock);
3235 addrconf_mod_dad_work(ifp, 0);
3238 static void addrconf_dad_work(struct work_struct *w)
3240 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3241 struct inet6_ifaddr,
3243 struct inet6_dev *idev = ifp->idev;
3244 struct in6_addr mcaddr;
3250 } action = DAD_PROCESS;
3254 spin_lock_bh(&ifp->state_lock);
3255 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3257 ifp->state = INET6_IFADDR_STATE_DAD;
3258 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3260 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3262 spin_unlock_bh(&ifp->state_lock);
3264 if (action == DAD_BEGIN) {
3265 addrconf_dad_begin(ifp);
3267 } else if (action == DAD_ABORT) {
3268 addrconf_dad_stop(ifp, 1);
3272 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3275 write_lock_bh(&idev->lock);
3276 if (idev->dead || !(idev->if_flags & IF_READY)) {
3277 write_unlock_bh(&idev->lock);
3281 spin_lock(&ifp->lock);
3282 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3283 spin_unlock(&ifp->lock);
3284 write_unlock_bh(&idev->lock);
3288 if (ifp->dad_probes == 0) {
3290 * DAD was successful
3293 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3294 spin_unlock(&ifp->lock);
3295 write_unlock_bh(&idev->lock);
3297 addrconf_dad_completed(ifp);
3303 addrconf_mod_dad_work(ifp,
3304 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3305 spin_unlock(&ifp->lock);
3306 write_unlock_bh(&idev->lock);
3308 /* send a neighbour solicitation for our addr */
3309 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3310 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3316 /* ifp->idev must be at least read locked */
3317 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3319 struct inet6_ifaddr *ifpiter;
3320 struct inet6_dev *idev = ifp->idev;
3322 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3323 if (ifpiter->scope > IFA_LINK)
3325 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3326 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3327 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3334 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3336 struct net_device *dev = ifp->idev->dev;
3337 struct in6_addr lladdr;
3338 bool send_rs, send_mld;
3340 addrconf_del_dad_work(ifp);
3343 * Configure the address for reception. Now it is valid.
3346 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3348 /* If added prefix is link local and we are prepared to process
3349 router advertisements, start sending router solicitations.
3352 read_lock_bh(&ifp->idev->lock);
3353 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3354 send_rs = send_mld &&
3355 ipv6_accept_ra(ifp->idev) &&
3356 ifp->idev->cnf.rtr_solicits > 0 &&
3357 (dev->flags&IFF_LOOPBACK) == 0;
3358 read_unlock_bh(&ifp->idev->lock);
3360 /* While dad is in progress mld report's source address is in6_addrany.
3361 * Resend with proper ll now.
3364 ipv6_mc_dad_complete(ifp->idev);
3368 * If a host as already performed a random delay
3369 * [...] as part of DAD [...] there is no need
3370 * to delay again before sending the first RS
3372 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3374 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3376 write_lock_bh(&ifp->idev->lock);
3377 spin_lock(&ifp->lock);
3378 ifp->idev->rs_probes = 1;
3379 ifp->idev->if_flags |= IF_RS_SENT;
3380 addrconf_mod_rs_timer(ifp->idev,
3381 ifp->idev->cnf.rtr_solicit_interval);
3382 spin_unlock(&ifp->lock);
3383 write_unlock_bh(&ifp->idev->lock);
3387 static void addrconf_dad_run(struct inet6_dev *idev)
3389 struct inet6_ifaddr *ifp;
3391 read_lock_bh(&idev->lock);
3392 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3393 spin_lock(&ifp->lock);
3394 if (ifp->flags & IFA_F_TENTATIVE &&
3395 ifp->state == INET6_IFADDR_STATE_DAD)
3396 addrconf_dad_kick(ifp);
3397 spin_unlock(&ifp->lock);
3399 read_unlock_bh(&idev->lock);
3402 #ifdef CONFIG_PROC_FS
3403 struct if6_iter_state {
3404 struct seq_net_private p;
3409 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3411 struct inet6_ifaddr *ifa = NULL;
3412 struct if6_iter_state *state = seq->private;
3413 struct net *net = seq_file_net(seq);
3416 /* initial bucket if pos is 0 */
3422 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3423 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3425 if (!net_eq(dev_net(ifa->idev->dev), net))
3427 /* sync with offset */
3428 if (p < state->offset) {
3436 /* prepare for next bucket */
3443 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3444 struct inet6_ifaddr *ifa)
3446 struct if6_iter_state *state = seq->private;
3447 struct net *net = seq_file_net(seq);
3449 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3450 if (!net_eq(dev_net(ifa->idev->dev), net))
3456 while (++state->bucket < IN6_ADDR_HSIZE) {
3458 hlist_for_each_entry_rcu_bh(ifa,
3459 &inet6_addr_lst[state->bucket], addr_lst) {
3460 if (!net_eq(dev_net(ifa->idev->dev), net))
3470 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3474 return if6_get_first(seq, *pos);
3477 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3479 struct inet6_ifaddr *ifa;
3481 ifa = if6_get_next(seq, v);
3486 static void if6_seq_stop(struct seq_file *seq, void *v)
3489 rcu_read_unlock_bh();
3492 static int if6_seq_show(struct seq_file *seq, void *v)
3494 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3495 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3497 ifp->idev->dev->ifindex,
3501 ifp->idev->dev->name);
3505 static const struct seq_operations if6_seq_ops = {
3506 .start = if6_seq_start,
3507 .next = if6_seq_next,
3508 .show = if6_seq_show,
3509 .stop = if6_seq_stop,
3512 static int if6_seq_open(struct inode *inode, struct file *file)
3514 return seq_open_net(inode, file, &if6_seq_ops,
3515 sizeof(struct if6_iter_state));
3518 static const struct file_operations if6_fops = {
3519 .owner = THIS_MODULE,
3520 .open = if6_seq_open,
3522 .llseek = seq_lseek,
3523 .release = seq_release_net,
3526 static int __net_init if6_proc_net_init(struct net *net)
3528 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3533 static void __net_exit if6_proc_net_exit(struct net *net)
3535 remove_proc_entry("if_inet6", net->proc_net);
3538 static struct pernet_operations if6_proc_net_ops = {
3539 .init = if6_proc_net_init,
3540 .exit = if6_proc_net_exit,
3543 int __init if6_proc_init(void)
3545 return register_pernet_subsys(&if6_proc_net_ops);
3548 void if6_proc_exit(void)
3550 unregister_pernet_subsys(&if6_proc_net_ops);
3552 #endif /* CONFIG_PROC_FS */
3554 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3555 /* Check if address is a home address configured on any interface. */
3556 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3559 struct inet6_ifaddr *ifp = NULL;
3560 unsigned int hash = inet6_addr_hash(addr);
3563 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3564 if (!net_eq(dev_net(ifp->idev->dev), net))
3566 if (ipv6_addr_equal(&ifp->addr, addr) &&
3567 (ifp->flags & IFA_F_HOMEADDRESS)) {
3572 rcu_read_unlock_bh();
3578 * Periodic address status verification
3581 static void addrconf_verify_rtnl(void)
3583 unsigned long now, next, next_sec, next_sched;
3584 struct inet6_ifaddr *ifp;
3591 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3593 cancel_delayed_work(&addr_chk_work);
3595 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3597 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3600 /* When setting preferred_lft to a value not zero or
3601 * infinity, while valid_lft is infinity
3602 * IFA_F_PERMANENT has a non-infinity life time.
3604 if ((ifp->flags & IFA_F_PERMANENT) &&
3605 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3608 spin_lock(&ifp->lock);
3609 /* We try to batch several events at once. */
3610 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3612 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3613 age >= ifp->valid_lft) {
3614 spin_unlock(&ifp->lock);
3618 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3619 spin_unlock(&ifp->lock);
3621 } else if (age >= ifp->prefered_lft) {
3622 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3625 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3627 ifp->flags |= IFA_F_DEPRECATED;
3630 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3631 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3632 next = ifp->tstamp + ifp->valid_lft * HZ;
3634 spin_unlock(&ifp->lock);
3639 ipv6_ifa_notify(0, ifp);
3643 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3644 !(ifp->flags&IFA_F_TENTATIVE)) {
3645 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3646 ifp->idev->cnf.dad_transmits *
3647 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3649 if (age >= ifp->prefered_lft - regen_advance) {
3650 struct inet6_ifaddr *ifpub = ifp->ifpub;
3651 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3652 next = ifp->tstamp + ifp->prefered_lft * HZ;
3653 if (!ifp->regen_count && ifpub) {
3656 in6_ifa_hold(ifpub);
3657 spin_unlock(&ifp->lock);
3659 spin_lock(&ifpub->lock);
3660 ifpub->regen_count = 0;
3661 spin_unlock(&ifpub->lock);
3662 ipv6_create_tempaddr(ifpub, ifp);
3667 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3668 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3669 spin_unlock(&ifp->lock);
3671 /* ifp->prefered_lft <= ifp->valid_lft */
3672 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3673 next = ifp->tstamp + ifp->prefered_lft * HZ;
3674 spin_unlock(&ifp->lock);
3679 next_sec = round_jiffies_up(next);
3682 /* If rounded timeout is accurate enough, accept it. */
3683 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3684 next_sched = next_sec;
3686 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3687 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3688 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3690 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3691 now, next, next_sec, next_sched);
3692 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3693 rcu_read_unlock_bh();
3696 static void addrconf_verify_work(struct work_struct *w)
3699 addrconf_verify_rtnl();
3703 static void addrconf_verify(void)
3705 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3708 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3709 struct in6_addr **peer_pfx)
3711 struct in6_addr *pfx = NULL;
3716 pfx = nla_data(addr);
3719 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3721 pfx = nla_data(local);
3727 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3728 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3729 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3730 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3731 [IFA_FLAGS] = { .len = sizeof(u32) },
3735 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3737 struct net *net = sock_net(skb->sk);
3738 struct ifaddrmsg *ifm;
3739 struct nlattr *tb[IFA_MAX+1];
3740 struct in6_addr *pfx, *peer_pfx;
3743 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3747 ifm = nlmsg_data(nlh);
3748 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3752 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3755 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3756 u32 prefered_lft, u32 valid_lft)
3760 unsigned long timeout;
3761 bool was_managetempaddr;
3762 bool had_prefixroute;
3766 if (!valid_lft || (prefered_lft > valid_lft))
3769 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3770 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3773 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3774 if (addrconf_finite_timeout(timeout)) {
3775 expires = jiffies_to_clock_t(timeout * HZ);
3776 valid_lft = timeout;
3777 flags = RTF_EXPIRES;
3781 ifa_flags |= IFA_F_PERMANENT;
3784 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3785 if (addrconf_finite_timeout(timeout)) {
3787 ifa_flags |= IFA_F_DEPRECATED;
3788 prefered_lft = timeout;
3791 spin_lock_bh(&ifp->lock);
3792 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3793 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3794 !(ifp->flags & IFA_F_NOPREFIXROUTE);
3795 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3796 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3797 IFA_F_NOPREFIXROUTE);
3798 ifp->flags |= ifa_flags;
3799 ifp->tstamp = jiffies;
3800 ifp->valid_lft = valid_lft;
3801 ifp->prefered_lft = prefered_lft;
3803 spin_unlock_bh(&ifp->lock);
3804 if (!(ifp->flags&IFA_F_TENTATIVE))
3805 ipv6_ifa_notify(0, ifp);
3807 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3808 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3810 } else if (had_prefixroute) {
3811 enum cleanup_prefix_rt_t action;
3812 unsigned long rt_expires;
3814 write_lock_bh(&ifp->idev->lock);
3815 action = check_cleanup_prefix_route(ifp, &rt_expires);
3816 write_unlock_bh(&ifp->idev->lock);
3818 if (action != CLEANUP_PREFIX_RT_NOP) {
3819 cleanup_prefix_route(ifp, rt_expires,
3820 action == CLEANUP_PREFIX_RT_DEL);
3824 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3825 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3826 valid_lft = prefered_lft = 0;
3827 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3828 !was_managetempaddr, jiffies);
3831 addrconf_verify_rtnl();
3837 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3839 struct net *net = sock_net(skb->sk);
3840 struct ifaddrmsg *ifm;
3841 struct nlattr *tb[IFA_MAX+1];
3842 struct in6_addr *pfx, *peer_pfx;
3843 struct inet6_ifaddr *ifa;
3844 struct net_device *dev;
3845 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3849 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3853 ifm = nlmsg_data(nlh);
3854 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3858 if (tb[IFA_CACHEINFO]) {
3859 struct ifa_cacheinfo *ci;
3861 ci = nla_data(tb[IFA_CACHEINFO]);
3862 valid_lft = ci->ifa_valid;
3863 preferred_lft = ci->ifa_prefered;
3865 preferred_lft = INFINITY_LIFE_TIME;
3866 valid_lft = INFINITY_LIFE_TIME;
3869 dev = __dev_get_by_index(net, ifm->ifa_index);
3873 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3875 /* We ignore other flags so far. */
3876 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3877 IFA_F_NOPREFIXROUTE;
3879 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3882 * It would be best to check for !NLM_F_CREATE here but
3883 * userspace already relies on not having to provide this.
3885 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3886 ifm->ifa_prefixlen, ifa_flags,
3887 preferred_lft, valid_lft);
3890 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3891 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3894 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3901 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3902 u8 scope, int ifindex)
3904 struct ifaddrmsg *ifm;
3906 ifm = nlmsg_data(nlh);
3907 ifm->ifa_family = AF_INET6;
3908 ifm->ifa_prefixlen = prefixlen;
3909 ifm->ifa_flags = flags;
3910 ifm->ifa_scope = scope;
3911 ifm->ifa_index = ifindex;
3914 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3915 unsigned long tstamp, u32 preferred, u32 valid)
3917 struct ifa_cacheinfo ci;
3919 ci.cstamp = cstamp_delta(cstamp);
3920 ci.tstamp = cstamp_delta(tstamp);
3921 ci.ifa_prefered = preferred;
3922 ci.ifa_valid = valid;
3924 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3927 static inline int rt_scope(int ifa_scope)
3929 if (ifa_scope & IFA_HOST)
3930 return RT_SCOPE_HOST;
3931 else if (ifa_scope & IFA_LINK)
3932 return RT_SCOPE_LINK;
3933 else if (ifa_scope & IFA_SITE)
3934 return RT_SCOPE_SITE;
3936 return RT_SCOPE_UNIVERSE;
3939 static inline int inet6_ifaddr_msgsize(void)
3941 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3942 + nla_total_size(16) /* IFA_LOCAL */
3943 + nla_total_size(16) /* IFA_ADDRESS */
3944 + nla_total_size(sizeof(struct ifa_cacheinfo))
3945 + nla_total_size(4) /* IFA_FLAGS */;
3948 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3949 u32 portid, u32 seq, int event, unsigned int flags)
3951 struct nlmsghdr *nlh;
3952 u32 preferred, valid;
3954 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3958 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3959 ifa->idev->dev->ifindex);
3961 if (!((ifa->flags&IFA_F_PERMANENT) &&
3962 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3963 preferred = ifa->prefered_lft;
3964 valid = ifa->valid_lft;
3965 if (preferred != INFINITY_LIFE_TIME) {
3966 long tval = (jiffies - ifa->tstamp)/HZ;
3967 if (preferred > tval)
3971 if (valid != INFINITY_LIFE_TIME) {
3979 preferred = INFINITY_LIFE_TIME;
3980 valid = INFINITY_LIFE_TIME;
3983 if (!ipv6_addr_any(&ifa->peer_addr)) {
3984 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3985 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3988 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3991 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3994 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
3997 return nlmsg_end(skb, nlh);
4000 nlmsg_cancel(skb, nlh);
4004 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4005 u32 portid, u32 seq, int event, u16 flags)
4007 struct nlmsghdr *nlh;
4008 u8 scope = RT_SCOPE_UNIVERSE;
4009 int ifindex = ifmca->idev->dev->ifindex;
4011 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4012 scope = RT_SCOPE_SITE;
4014 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4018 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4019 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4020 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4021 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4022 nlmsg_cancel(skb, nlh);
4026 return nlmsg_end(skb, nlh);
4029 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4030 u32 portid, u32 seq, int event, unsigned int flags)
4032 struct nlmsghdr *nlh;
4033 u8 scope = RT_SCOPE_UNIVERSE;
4034 int ifindex = ifaca->aca_idev->dev->ifindex;
4036 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4037 scope = RT_SCOPE_SITE;
4039 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4043 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4044 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4045 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4046 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4047 nlmsg_cancel(skb, nlh);
4051 return nlmsg_end(skb, nlh);
4060 /* called with rcu_read_lock() */
4061 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4062 struct netlink_callback *cb, enum addr_type_t type,
4063 int s_ip_idx, int *p_ip_idx)
4065 struct ifmcaddr6 *ifmca;
4066 struct ifacaddr6 *ifaca;
4068 int ip_idx = *p_ip_idx;
4070 read_lock_bh(&idev->lock);
4072 case UNICAST_ADDR: {
4073 struct inet6_ifaddr *ifa;
4075 /* unicast address incl. temp addr */
4076 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4077 if (++ip_idx < s_ip_idx)
4079 err = inet6_fill_ifaddr(skb, ifa,
4080 NETLINK_CB(cb->skb).portid,
4086 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4090 case MULTICAST_ADDR:
4091 /* multicast address */
4092 for (ifmca = idev->mc_list; ifmca;
4093 ifmca = ifmca->next, ip_idx++) {
4094 if (ip_idx < s_ip_idx)
4096 err = inet6_fill_ifmcaddr(skb, ifmca,
4097 NETLINK_CB(cb->skb).portid,
4106 /* anycast address */
4107 for (ifaca = idev->ac_list; ifaca;
4108 ifaca = ifaca->aca_next, ip_idx++) {
4109 if (ip_idx < s_ip_idx)
4111 err = inet6_fill_ifacaddr(skb, ifaca,
4112 NETLINK_CB(cb->skb).portid,
4123 read_unlock_bh(&idev->lock);
4128 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4129 enum addr_type_t type)
4131 struct net *net = sock_net(skb->sk);
4134 int s_idx, s_ip_idx;
4135 struct net_device *dev;
4136 struct inet6_dev *idev;
4137 struct hlist_head *head;
4140 s_idx = idx = cb->args[1];
4141 s_ip_idx = ip_idx = cb->args[2];
4144 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4145 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4147 head = &net->dev_index_head[h];
4148 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4151 if (h > s_h || idx > s_idx)
4154 idev = __in6_dev_get(dev);
4158 if (in6_dump_addrs(idev, skb, cb, type,
4159 s_ip_idx, &ip_idx) <= 0)
4169 cb->args[2] = ip_idx;
4174 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4176 enum addr_type_t type = UNICAST_ADDR;
4178 return inet6_dump_addr(skb, cb, type);
4181 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4183 enum addr_type_t type = MULTICAST_ADDR;
4185 return inet6_dump_addr(skb, cb, type);
4189 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4191 enum addr_type_t type = ANYCAST_ADDR;
4193 return inet6_dump_addr(skb, cb, type);
4196 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4198 struct net *net = sock_net(in_skb->sk);
4199 struct ifaddrmsg *ifm;
4200 struct nlattr *tb[IFA_MAX+1];
4201 struct in6_addr *addr = NULL, *peer;
4202 struct net_device *dev = NULL;
4203 struct inet6_ifaddr *ifa;
4204 struct sk_buff *skb;
4207 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4211 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4217 ifm = nlmsg_data(nlh);
4219 dev = __dev_get_by_index(net, ifm->ifa_index);
4221 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4223 err = -EADDRNOTAVAIL;
4227 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4233 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4234 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4236 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4237 WARN_ON(err == -EMSGSIZE);
4241 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4248 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4250 struct sk_buff *skb;
4251 struct net *net = dev_net(ifa->idev->dev);
4254 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4258 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4260 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4261 WARN_ON(err == -EMSGSIZE);
4265 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4269 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4272 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4273 __s32 *array, int bytes)
4275 BUG_ON(bytes < (DEVCONF_MAX * 4));
4277 memset(array, 0, bytes);
4278 array[DEVCONF_FORWARDING] = cnf->forwarding;
4279 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4280 array[DEVCONF_MTU6] = cnf->mtu6;
4281 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4282 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4283 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4284 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4285 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4286 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4287 jiffies_to_msecs(cnf->rtr_solicit_interval);
4288 array[DEVCONF_RTR_SOLICIT_DELAY] =
4289 jiffies_to_msecs(cnf->rtr_solicit_delay);
4290 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4291 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4292 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4293 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4294 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4295 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4296 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4297 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4298 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4299 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4300 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4301 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4302 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4303 #ifdef CONFIG_IPV6_ROUTER_PREF
4304 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4305 array[DEVCONF_RTR_PROBE_INTERVAL] =
4306 jiffies_to_msecs(cnf->rtr_probe_interval);
4307 #ifdef CONFIG_IPV6_ROUTE_INFO
4308 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4311 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4312 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4313 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4314 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4316 #ifdef CONFIG_IPV6_MROUTE
4317 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4319 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4320 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4321 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4322 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4323 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4326 static inline size_t inet6_ifla6_size(void)
4328 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4329 + nla_total_size(sizeof(struct ifla_cacheinfo))
4330 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4331 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4332 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4333 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4336 static inline size_t inet6_if_nlmsg_size(void)
4338 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4339 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4340 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4341 + nla_total_size(4) /* IFLA_MTU */
4342 + nla_total_size(4) /* IFLA_LINK */
4343 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4346 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4347 int items, int bytes)
4350 int pad = bytes - sizeof(u64) * items;
4353 /* Use put_unaligned() because stats may not be aligned for u64. */
4354 put_unaligned(items, &stats[0]);
4355 for (i = 1; i < items; i++)
4356 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4358 memset(&stats[items], 0, pad);
4361 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4362 int items, int bytes, size_t syncpoff)
4365 int pad = bytes - sizeof(u64) * items;
4368 /* Use put_unaligned() because stats may not be aligned for u64. */
4369 put_unaligned(items, &stats[0]);
4370 for (i = 1; i < items; i++)
4371 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4373 memset(&stats[items], 0, pad);
4376 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4380 case IFLA_INET6_STATS:
4381 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4382 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4384 case IFLA_INET6_ICMP6STATS:
4385 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4390 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4393 struct ifla_cacheinfo ci;
4395 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4396 goto nla_put_failure;
4397 ci.max_reasm_len = IPV6_MAXPLEN;
4398 ci.tstamp = cstamp_delta(idev->tstamp);
4399 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4400 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4401 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4402 goto nla_put_failure;
4403 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4405 goto nla_put_failure;
4406 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4408 /* XXX - MC not implemented */
4410 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4412 goto nla_put_failure;
4413 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4415 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4417 goto nla_put_failure;
4418 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4420 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4422 goto nla_put_failure;
4423 read_lock_bh(&idev->lock);
4424 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4425 read_unlock_bh(&idev->lock);
4433 static size_t inet6_get_link_af_size(const struct net_device *dev)
4435 if (!__in6_dev_get(dev))
4438 return inet6_ifla6_size();
4441 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4443 struct inet6_dev *idev = __in6_dev_get(dev);
4448 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4454 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4456 struct inet6_ifaddr *ifp;
4457 struct net_device *dev = idev->dev;
4458 bool update_rs = false;
4459 struct in6_addr ll_addr;
4465 if (ipv6_addr_any(token))
4467 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4469 if (!ipv6_accept_ra(idev))
4471 if (idev->cnf.rtr_solicits <= 0)
4474 write_lock_bh(&idev->lock);
4476 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4477 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4479 write_unlock_bh(&idev->lock);
4481 if (!idev->dead && (idev->if_flags & IF_READY) &&
4482 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4483 IFA_F_OPTIMISTIC)) {
4485 /* If we're not ready, then normal ifup will take care
4486 * of this. Otherwise, we need to request our rs here.
4488 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4492 write_lock_bh(&idev->lock);
4495 idev->if_flags |= IF_RS_SENT;
4496 idev->rs_probes = 1;
4497 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4500 /* Well, that's kinda nasty ... */
4501 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4502 spin_lock(&ifp->lock);
4503 if (ifp->tokenized) {
4505 ifp->prefered_lft = 0;
4507 spin_unlock(&ifp->lock);
4510 write_unlock_bh(&idev->lock);
4511 addrconf_verify_rtnl();
4515 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4518 struct inet6_dev *idev = __in6_dev_get(dev);
4519 struct nlattr *tb[IFLA_INET6_MAX + 1];
4522 return -EAFNOSUPPORT;
4524 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4527 if (tb[IFLA_INET6_TOKEN])
4528 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4533 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4534 u32 portid, u32 seq, int event, unsigned int flags)
4536 struct net_device *dev = idev->dev;
4537 struct ifinfomsg *hdr;
4538 struct nlmsghdr *nlh;
4541 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4545 hdr = nlmsg_data(nlh);
4546 hdr->ifi_family = AF_INET6;
4548 hdr->ifi_type = dev->type;
4549 hdr->ifi_index = dev->ifindex;
4550 hdr->ifi_flags = dev_get_flags(dev);
4551 hdr->ifi_change = 0;
4553 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4555 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4556 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4557 (dev->ifindex != dev->iflink &&
4558 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4559 goto nla_put_failure;
4560 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4561 if (protoinfo == NULL)
4562 goto nla_put_failure;
4564 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4565 goto nla_put_failure;
4567 nla_nest_end(skb, protoinfo);
4568 return nlmsg_end(skb, nlh);
4571 nlmsg_cancel(skb, nlh);
4575 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4577 struct net *net = sock_net(skb->sk);
4580 struct net_device *dev;
4581 struct inet6_dev *idev;
4582 struct hlist_head *head;
4585 s_idx = cb->args[1];
4588 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4590 head = &net->dev_index_head[h];
4591 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4594 idev = __in6_dev_get(dev);
4597 if (inet6_fill_ifinfo(skb, idev,
4598 NETLINK_CB(cb->skb).portid,
4600 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4614 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4616 struct sk_buff *skb;
4617 struct net *net = dev_net(idev->dev);
4620 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4624 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4626 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4627 WARN_ON(err == -EMSGSIZE);
4631 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4635 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4638 static inline size_t inet6_prefix_nlmsg_size(void)
4640 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4641 + nla_total_size(sizeof(struct in6_addr))
4642 + nla_total_size(sizeof(struct prefix_cacheinfo));
4645 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4646 struct prefix_info *pinfo, u32 portid, u32 seq,
4647 int event, unsigned int flags)
4649 struct prefixmsg *pmsg;
4650 struct nlmsghdr *nlh;
4651 struct prefix_cacheinfo ci;
4653 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4657 pmsg = nlmsg_data(nlh);
4658 pmsg->prefix_family = AF_INET6;
4659 pmsg->prefix_pad1 = 0;
4660 pmsg->prefix_pad2 = 0;
4661 pmsg->prefix_ifindex = idev->dev->ifindex;
4662 pmsg->prefix_len = pinfo->prefix_len;
4663 pmsg->prefix_type = pinfo->type;
4664 pmsg->prefix_pad3 = 0;
4665 pmsg->prefix_flags = 0;
4667 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4668 if (pinfo->autoconf)
4669 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4671 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4672 goto nla_put_failure;
4673 ci.preferred_time = ntohl(pinfo->prefered);
4674 ci.valid_time = ntohl(pinfo->valid);
4675 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4676 goto nla_put_failure;
4677 return nlmsg_end(skb, nlh);
4680 nlmsg_cancel(skb, nlh);
4684 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4685 struct prefix_info *pinfo)
4687 struct sk_buff *skb;
4688 struct net *net = dev_net(idev->dev);
4691 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4695 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4697 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4698 WARN_ON(err == -EMSGSIZE);
4702 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4706 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4709 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4711 struct net *net = dev_net(ifp->idev->dev);
4716 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4721 * If the address was optimistic
4722 * we inserted the route at the start of
4723 * our DAD process, so we don't need
4726 if (!(ifp->rt->rt6i_node))
4727 ip6_ins_rt(ifp->rt);
4728 if (ifp->idev->cnf.forwarding)
4729 addrconf_join_anycast(ifp);
4730 if (!ipv6_addr_any(&ifp->peer_addr))
4731 addrconf_prefix_route(&ifp->peer_addr, 128,
4732 ifp->idev->dev, 0, 0);
4735 if (ifp->idev->cnf.forwarding)
4736 addrconf_leave_anycast(ifp);
4737 addrconf_leave_solict(ifp->idev, &ifp->addr);
4738 if (!ipv6_addr_any(&ifp->peer_addr)) {
4739 struct rt6_info *rt;
4740 struct net_device *dev = ifp->idev->dev;
4742 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4750 dst_hold(&ifp->rt->dst);
4752 if (ip6_del_rt(ifp->rt))
4753 dst_free(&ifp->rt->dst);
4756 atomic_inc(&net->ipv6.dev_addr_genid);
4757 rt_genid_bump_ipv6(net);
4760 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4763 if (likely(ifp->idev->dead == 0))
4764 __ipv6_ifa_notify(event, ifp);
4765 rcu_read_unlock_bh();
4768 #ifdef CONFIG_SYSCTL
4771 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4772 void __user *buffer, size_t *lenp, loff_t *ppos)
4774 int *valp = ctl->data;
4777 struct ctl_table lctl;
4781 * ctl->data points to idev->cnf.forwarding, we should
4782 * not modify it until we get the rtnl lock.
4787 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4790 ret = addrconf_fixup_forwarding(ctl, valp, val);
4796 static void dev_disable_change(struct inet6_dev *idev)
4798 struct netdev_notifier_info info;
4800 if (!idev || !idev->dev)
4803 netdev_notifier_info_init(&info, idev->dev);
4804 if (idev->cnf.disable_ipv6)
4805 addrconf_notify(NULL, NETDEV_DOWN, &info);
4807 addrconf_notify(NULL, NETDEV_UP, &info);
4810 static void addrconf_disable_change(struct net *net, __s32 newf)
4812 struct net_device *dev;
4813 struct inet6_dev *idev;
4816 for_each_netdev_rcu(net, dev) {
4817 idev = __in6_dev_get(dev);
4819 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4820 idev->cnf.disable_ipv6 = newf;
4822 dev_disable_change(idev);
4828 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4833 if (!rtnl_trylock())
4834 return restart_syscall();
4836 net = (struct net *)table->extra2;
4840 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4845 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4846 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4847 addrconf_disable_change(net, newf);
4848 } else if ((!newf) ^ (!old))
4849 dev_disable_change((struct inet6_dev *)table->extra1);
4856 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4857 void __user *buffer, size_t *lenp, loff_t *ppos)
4859 int *valp = ctl->data;
4862 struct ctl_table lctl;
4866 * ctl->data points to idev->cnf.disable_ipv6, we should
4867 * not modify it until we get the rtnl lock.
4872 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4875 ret = addrconf_disable_ipv6(ctl, valp, val);
4882 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4883 void __user *buffer, size_t *lenp, loff_t *ppos)
4885 int *valp = ctl->data;
4890 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4893 if (write && old != new) {
4894 struct net *net = ctl->extra2;
4896 if (!rtnl_trylock())
4897 return restart_syscall();
4899 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4900 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4901 NETCONFA_IFINDEX_DEFAULT,
4902 net->ipv6.devconf_dflt);
4903 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4904 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4905 NETCONFA_IFINDEX_ALL,
4906 net->ipv6.devconf_all);
4908 struct inet6_dev *idev = ctl->extra1;
4910 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4921 static struct addrconf_sysctl_table
4923 struct ctl_table_header *sysctl_header;
4924 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4925 } addrconf_sysctl __read_mostly = {
4926 .sysctl_header = NULL,
4929 .procname = "forwarding",
4930 .data = &ipv6_devconf.forwarding,
4931 .maxlen = sizeof(int),
4933 .proc_handler = addrconf_sysctl_forward,
4936 .procname = "hop_limit",
4937 .data = &ipv6_devconf.hop_limit,
4938 .maxlen = sizeof(int),
4940 .proc_handler = proc_dointvec,
4944 .data = &ipv6_devconf.mtu6,
4945 .maxlen = sizeof(int),
4947 .proc_handler = proc_dointvec,
4950 .procname = "accept_ra",
4951 .data = &ipv6_devconf.accept_ra,
4952 .maxlen = sizeof(int),
4954 .proc_handler = proc_dointvec,
4957 .procname = "accept_redirects",
4958 .data = &ipv6_devconf.accept_redirects,
4959 .maxlen = sizeof(int),
4961 .proc_handler = proc_dointvec,
4964 .procname = "autoconf",
4965 .data = &ipv6_devconf.autoconf,
4966 .maxlen = sizeof(int),
4968 .proc_handler = proc_dointvec,
4971 .procname = "dad_transmits",
4972 .data = &ipv6_devconf.dad_transmits,
4973 .maxlen = sizeof(int),
4975 .proc_handler = proc_dointvec,
4978 .procname = "router_solicitations",
4979 .data = &ipv6_devconf.rtr_solicits,
4980 .maxlen = sizeof(int),
4982 .proc_handler = proc_dointvec,
4985 .procname = "router_solicitation_interval",
4986 .data = &ipv6_devconf.rtr_solicit_interval,
4987 .maxlen = sizeof(int),
4989 .proc_handler = proc_dointvec_jiffies,
4992 .procname = "router_solicitation_delay",
4993 .data = &ipv6_devconf.rtr_solicit_delay,
4994 .maxlen = sizeof(int),
4996 .proc_handler = proc_dointvec_jiffies,
4999 .procname = "force_mld_version",
5000 .data = &ipv6_devconf.force_mld_version,
5001 .maxlen = sizeof(int),
5003 .proc_handler = proc_dointvec,
5006 .procname = "mldv1_unsolicited_report_interval",
5008 &ipv6_devconf.mldv1_unsolicited_report_interval,
5009 .maxlen = sizeof(int),
5011 .proc_handler = proc_dointvec_ms_jiffies,
5014 .procname = "mldv2_unsolicited_report_interval",
5016 &ipv6_devconf.mldv2_unsolicited_report_interval,
5017 .maxlen = sizeof(int),
5019 .proc_handler = proc_dointvec_ms_jiffies,
5022 .procname = "use_tempaddr",
5023 .data = &ipv6_devconf.use_tempaddr,
5024 .maxlen = sizeof(int),
5026 .proc_handler = proc_dointvec,
5029 .procname = "temp_valid_lft",
5030 .data = &ipv6_devconf.temp_valid_lft,
5031 .maxlen = sizeof(int),
5033 .proc_handler = proc_dointvec,
5036 .procname = "temp_prefered_lft",
5037 .data = &ipv6_devconf.temp_prefered_lft,
5038 .maxlen = sizeof(int),
5040 .proc_handler = proc_dointvec,
5043 .procname = "regen_max_retry",
5044 .data = &ipv6_devconf.regen_max_retry,
5045 .maxlen = sizeof(int),
5047 .proc_handler = proc_dointvec,
5050 .procname = "max_desync_factor",
5051 .data = &ipv6_devconf.max_desync_factor,
5052 .maxlen = sizeof(int),
5054 .proc_handler = proc_dointvec,
5057 .procname = "max_addresses",
5058 .data = &ipv6_devconf.max_addresses,
5059 .maxlen = sizeof(int),
5061 .proc_handler = proc_dointvec,
5064 .procname = "accept_ra_defrtr",
5065 .data = &ipv6_devconf.accept_ra_defrtr,
5066 .maxlen = sizeof(int),
5068 .proc_handler = proc_dointvec,
5071 .procname = "accept_ra_pinfo",
5072 .data = &ipv6_devconf.accept_ra_pinfo,
5073 .maxlen = sizeof(int),
5075 .proc_handler = proc_dointvec,
5077 #ifdef CONFIG_IPV6_ROUTER_PREF
5079 .procname = "accept_ra_rtr_pref",
5080 .data = &ipv6_devconf.accept_ra_rtr_pref,
5081 .maxlen = sizeof(int),
5083 .proc_handler = proc_dointvec,
5086 .procname = "router_probe_interval",
5087 .data = &ipv6_devconf.rtr_probe_interval,
5088 .maxlen = sizeof(int),
5090 .proc_handler = proc_dointvec_jiffies,
5092 #ifdef CONFIG_IPV6_ROUTE_INFO
5094 .procname = "accept_ra_rt_info_max_plen",
5095 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5096 .maxlen = sizeof(int),
5098 .proc_handler = proc_dointvec,
5103 .procname = "proxy_ndp",
5104 .data = &ipv6_devconf.proxy_ndp,
5105 .maxlen = sizeof(int),
5107 .proc_handler = addrconf_sysctl_proxy_ndp,
5110 .procname = "accept_source_route",
5111 .data = &ipv6_devconf.accept_source_route,
5112 .maxlen = sizeof(int),
5114 .proc_handler = proc_dointvec,
5116 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5118 .procname = "optimistic_dad",
5119 .data = &ipv6_devconf.optimistic_dad,
5120 .maxlen = sizeof(int),
5122 .proc_handler = proc_dointvec,
5126 #ifdef CONFIG_IPV6_MROUTE
5128 .procname = "mc_forwarding",
5129 .data = &ipv6_devconf.mc_forwarding,
5130 .maxlen = sizeof(int),
5132 .proc_handler = proc_dointvec,
5136 .procname = "disable_ipv6",
5137 .data = &ipv6_devconf.disable_ipv6,
5138 .maxlen = sizeof(int),
5140 .proc_handler = addrconf_sysctl_disable,
5143 .procname = "accept_dad",
5144 .data = &ipv6_devconf.accept_dad,
5145 .maxlen = sizeof(int),
5147 .proc_handler = proc_dointvec,
5150 .procname = "force_tllao",
5151 .data = &ipv6_devconf.force_tllao,
5152 .maxlen = sizeof(int),
5154 .proc_handler = proc_dointvec
5157 .procname = "ndisc_notify",
5158 .data = &ipv6_devconf.ndisc_notify,
5159 .maxlen = sizeof(int),
5161 .proc_handler = proc_dointvec
5164 .procname = "suppress_frag_ndisc",
5165 .data = &ipv6_devconf.suppress_frag_ndisc,
5166 .maxlen = sizeof(int),
5168 .proc_handler = proc_dointvec
5176 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5177 struct inet6_dev *idev, struct ipv6_devconf *p)
5180 struct addrconf_sysctl_table *t;
5181 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5183 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5187 for (i = 0; t->addrconf_vars[i].data; i++) {
5188 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5189 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5190 t->addrconf_vars[i].extra2 = net;
5193 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5195 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5196 if (t->sysctl_header == NULL)
5208 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5210 struct addrconf_sysctl_table *t;
5212 if (p->sysctl == NULL)
5217 unregister_net_sysctl_table(t->sysctl_header);
5221 static void addrconf_sysctl_register(struct inet6_dev *idev)
5223 neigh_sysctl_register(idev->dev, idev->nd_parms,
5224 &ndisc_ifinfo_sysctl_change);
5225 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5229 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5231 __addrconf_sysctl_unregister(&idev->cnf);
5232 neigh_sysctl_unregister(idev->nd_parms);
5238 static int __net_init addrconf_init_net(struct net *net)
5241 struct ipv6_devconf *all, *dflt;
5243 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5247 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5249 goto err_alloc_dflt;
5251 /* these will be inherited by all namespaces */
5252 dflt->autoconf = ipv6_defaults.autoconf;
5253 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5255 net->ipv6.devconf_all = all;
5256 net->ipv6.devconf_dflt = dflt;
5258 #ifdef CONFIG_SYSCTL
5259 err = __addrconf_sysctl_register(net, "all", NULL, all);
5263 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5269 #ifdef CONFIG_SYSCTL
5271 __addrconf_sysctl_unregister(all);
5281 static void __net_exit addrconf_exit_net(struct net *net)
5283 #ifdef CONFIG_SYSCTL
5284 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5285 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5287 if (!net_eq(net, &init_net)) {
5288 kfree(net->ipv6.devconf_dflt);
5289 kfree(net->ipv6.devconf_all);
5293 static struct pernet_operations addrconf_ops = {
5294 .init = addrconf_init_net,
5295 .exit = addrconf_exit_net,
5298 static struct rtnl_af_ops inet6_ops = {
5300 .fill_link_af = inet6_fill_link_af,
5301 .get_link_af_size = inet6_get_link_af_size,
5302 .set_link_af = inet6_set_link_af,
5306 * Init / cleanup code
5309 int __init addrconf_init(void)
5313 err = ipv6_addr_label_init();
5315 pr_crit("%s: cannot initialize default policy table: %d\n",
5320 err = register_pernet_subsys(&addrconf_ops);
5324 addrconf_wq = create_workqueue("ipv6_addrconf");
5330 /* The addrconf netdev notifier requires that loopback_dev
5331 * has it's ipv6 private information allocated and setup
5332 * before it can bring up and give link-local addresses
5333 * to other devices which are up.
5335 * Unfortunately, loopback_dev is not necessarily the first
5336 * entry in the global dev_base list of net devices. In fact,
5337 * it is likely to be the very last entry on that list.
5338 * So this causes the notifier registry below to try and
5339 * give link-local addresses to all devices besides loopback_dev
5340 * first, then loopback_dev, which cases all the non-loopback_dev
5341 * devices to fail to get a link-local address.
5343 * So, as a temporary fix, allocate the ipv6 structure for
5344 * loopback_dev first by hand.
5345 * Longer term, all of the dependencies ipv6 has upon the loopback
5346 * device and it being up should be removed.
5349 if (!ipv6_add_dev(init_net.loopback_dev))
5355 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5356 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5358 register_netdevice_notifier(&ipv6_dev_notf);
5362 rtnl_af_register(&inet6_ops);
5364 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5369 /* Only the first call to __rtnl_register can fail */
5370 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5371 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5372 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5373 inet6_dump_ifaddr, NULL);
5374 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5375 inet6_dump_ifmcaddr, NULL);
5376 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5377 inet6_dump_ifacaddr, NULL);
5378 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5379 inet6_netconf_dump_devconf, NULL);
5381 ipv6_addr_label_rtnl_register();
5385 rtnl_af_unregister(&inet6_ops);
5386 unregister_netdevice_notifier(&ipv6_dev_notf);
5388 destroy_workqueue(addrconf_wq);
5390 unregister_pernet_subsys(&addrconf_ops);
5392 ipv6_addr_label_cleanup();
5397 void addrconf_cleanup(void)
5399 struct net_device *dev;
5402 unregister_netdevice_notifier(&ipv6_dev_notf);
5403 unregister_pernet_subsys(&addrconf_ops);
5404 ipv6_addr_label_cleanup();
5408 __rtnl_af_unregister(&inet6_ops);
5410 /* clean dev list */
5411 for_each_netdev(&init_net, dev) {
5412 if (__in6_dev_get(dev) == NULL)
5414 addrconf_ifdown(dev, 1);
5416 addrconf_ifdown(init_net.loopback_dev, 2);
5421 spin_lock_bh(&addrconf_hash_lock);
5422 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5423 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5424 spin_unlock_bh(&addrconf_hash_lock);
5425 cancel_delayed_work(&addr_chk_work);
5428 destroy_workqueue(addrconf_wq);