2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
61 #include <linux/sysctl.h>
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
69 #include <net/net_namespace.h>
73 #include <net/6lowpan.h>
74 #include <net/firewire.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
96 /* Set to 3 to get tracing... */
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
105 #define INFINITY_LIFE_TIME 0xFFFFFFFF
107 #define IPV6_MAX_STRLEN \
108 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
110 static inline u32 cstamp_delta(unsigned long cstamp)
112 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
116 static int addrconf_sysctl_register(struct inet6_dev *idev);
117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
131 static void ipv6_regen_rndid(unsigned long data);
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135 static int ipv6_generate_stable_address(struct in6_addr *addr,
137 const struct inet6_dev *idev);
140 * Configured unicast address hash table
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
145 static void addrconf_verify(void);
146 static void addrconf_verify_rtnl(void);
147 static void addrconf_verify_work(struct work_struct *);
149 static struct workqueue_struct *addrconf_wq;
150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
155 static void addrconf_type_change(struct net_device *dev,
156 unsigned long event);
157 static int addrconf_ifdown(struct net_device *dev, int how);
159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
161 const struct net_device *dev,
162 u32 flags, u32 noflags);
164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
165 static void addrconf_dad_work(struct work_struct *w);
166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
167 static void addrconf_dad_run(struct inet6_dev *idev);
168 static void addrconf_rs_timer(unsigned long data);
169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
173 struct prefix_info *pinfo);
174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
175 struct net_device *dev);
177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
179 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
180 .mtu6 = IPV6_MIN_MTU,
182 .accept_redirects = 1,
184 .force_mld_version = 0,
185 .mldv1_unsolicited_report_interval = 10 * HZ,
186 .mldv2_unsolicited_report_interval = HZ,
188 .rtr_solicits = MAX_RTR_SOLICITATIONS,
189 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
190 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
192 .temp_valid_lft = TEMP_VALID_LIFETIME,
193 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
194 .regen_max_retry = REGEN_MAX_RETRY,
195 .max_desync_factor = MAX_DESYNC_FACTOR,
196 .max_addresses = IPV6_MAX_ADDRESSES,
197 .accept_ra_defrtr = 1,
198 .accept_ra_from_local = 0,
199 .accept_ra_min_hop_limit= 1,
200 .accept_ra_pinfo = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202 .accept_ra_rtr_pref = 1,
203 .rtr_probe_interval = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205 .accept_ra_rt_info_max_plen = 0,
209 .accept_source_route = 0, /* we do not accept RH0 by default. */
212 .suppress_frag_ndisc = 1,
215 .initialized = false,
217 .use_oif_addrs_only = 0,
218 .ignore_routes_with_linkdown = 0,
219 .keep_addr_on_down = 0,
222 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
224 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
225 .mtu6 = IPV6_MIN_MTU,
227 .accept_redirects = 1,
229 .force_mld_version = 0,
230 .mldv1_unsolicited_report_interval = 10 * HZ,
231 .mldv2_unsolicited_report_interval = HZ,
233 .rtr_solicits = MAX_RTR_SOLICITATIONS,
234 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
235 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
237 .temp_valid_lft = TEMP_VALID_LIFETIME,
238 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
239 .regen_max_retry = REGEN_MAX_RETRY,
240 .max_desync_factor = MAX_DESYNC_FACTOR,
241 .max_addresses = IPV6_MAX_ADDRESSES,
242 .accept_ra_defrtr = 1,
243 .accept_ra_from_local = 0,
244 .accept_ra_min_hop_limit= 1,
245 .accept_ra_pinfo = 1,
246 #ifdef CONFIG_IPV6_ROUTER_PREF
247 .accept_ra_rtr_pref = 1,
248 .rtr_probe_interval = 60 * HZ,
249 #ifdef CONFIG_IPV6_ROUTE_INFO
250 .accept_ra_rt_info_max_plen = 0,
254 .accept_source_route = 0, /* we do not accept RH0 by default. */
257 .suppress_frag_ndisc = 1,
260 .initialized = false,
262 .use_oif_addrs_only = 0,
263 .ignore_routes_with_linkdown = 0,
264 .keep_addr_on_down = 0,
267 /* Check if a valid qdisc is available */
268 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
270 return !qdisc_tx_is_noop(dev);
273 static void addrconf_del_rs_timer(struct inet6_dev *idev)
275 if (del_timer(&idev->rs_timer))
279 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
281 if (cancel_delayed_work(&ifp->dad_work))
285 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
288 if (!timer_pending(&idev->rs_timer))
290 mod_timer(&idev->rs_timer, jiffies + when);
293 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
296 if (!delayed_work_pending(&ifp->dad_work))
298 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
301 static int snmp6_alloc_dev(struct inet6_dev *idev)
305 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
306 if (!idev->stats.ipv6)
309 for_each_possible_cpu(i) {
310 struct ipstats_mib *addrconf_stats;
311 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
312 u64_stats_init(&addrconf_stats->syncp);
316 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
318 if (!idev->stats.icmpv6dev)
320 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
322 if (!idev->stats.icmpv6msgdev)
328 kfree(idev->stats.icmpv6dev);
330 free_percpu(idev->stats.ipv6);
335 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
337 struct inet6_dev *ndev;
342 if (dev->mtu < IPV6_MIN_MTU)
343 return ERR_PTR(-EINVAL);
345 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
349 rwlock_init(&ndev->lock);
351 INIT_LIST_HEAD(&ndev->addr_list);
352 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
353 (unsigned long)ndev);
354 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
356 if (ndev->cnf.stable_secret.initialized)
357 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
359 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
361 ndev->cnf.mtu6 = dev->mtu;
362 ndev->cnf.sysctl = NULL;
363 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 if (!ndev->nd_parms) {
368 if (ndev->cnf.forwarding)
369 dev_disable_lro(dev);
370 /* We refer to the device */
373 if (snmp6_alloc_dev(ndev) < 0) {
375 "%s: cannot allocate memory for statistics; dev=%s.\n",
376 __func__, dev->name);
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
383 if (snmp6_register_dev(ndev) < 0) {
385 "%s: cannot create /proc/net/dev_snmp6/%s\n",
386 __func__, dev->name);
390 /* One reference from device. We must do this before
391 * we invoke __ipv6_regen_rndid().
395 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
396 ndev->cnf.accept_dad = -1;
398 #if IS_ENABLED(CONFIG_IPV6_SIT)
399 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
400 pr_info("%s: Disabled Multicast RS\n", dev->name);
401 ndev->cnf.rtr_solicits = 0;
405 INIT_LIST_HEAD(&ndev->tempaddr_list);
406 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
407 if ((dev->flags&IFF_LOOPBACK) ||
408 dev->type == ARPHRD_TUNNEL ||
409 dev->type == ARPHRD_TUNNEL6 ||
410 dev->type == ARPHRD_SIT ||
411 dev->type == ARPHRD_NONE) {
412 ndev->cnf.use_tempaddr = -1;
415 ipv6_regen_rndid((unsigned long) ndev);
418 ndev->token = in6addr_any;
420 if (netif_running(dev) && addrconf_qdisc_ok(dev))
421 ndev->if_flags |= IF_READY;
423 ipv6_mc_init_dev(ndev);
424 ndev->tstamp = jiffies;
425 err = addrconf_sysctl_register(ndev);
427 ipv6_mc_destroy_dev(ndev);
428 del_timer(&ndev->regen_timer);
429 snmp6_unregister_dev(ndev);
432 /* protected by rtnl_lock */
433 rcu_assign_pointer(dev->ip6_ptr, ndev);
435 /* Join interface-local all-node multicast group */
436 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
438 /* Join all-node multicast group */
439 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
441 /* Join all-router multicast group if forwarding is set */
442 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
443 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
448 neigh_parms_release(&nd_tbl, ndev->nd_parms);
450 in6_dev_finish_destroy(ndev);
454 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
456 struct inet6_dev *idev;
460 idev = __in6_dev_get(dev);
462 idev = ipv6_add_dev(dev);
467 if (dev->flags&IFF_UP)
472 static int inet6_netconf_msgsize_devconf(int type)
474 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
475 + nla_total_size(4); /* NETCONFA_IFINDEX */
477 /* type -1 is used for ALL */
478 if (type == -1 || type == NETCONFA_FORWARDING)
479 size += nla_total_size(4);
480 #ifdef CONFIG_IPV6_MROUTE
481 if (type == -1 || type == NETCONFA_MC_FORWARDING)
482 size += nla_total_size(4);
484 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
485 size += nla_total_size(4);
487 if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
488 size += nla_total_size(4);
493 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
494 struct ipv6_devconf *devconf, u32 portid,
495 u32 seq, int event, unsigned int flags,
498 struct nlmsghdr *nlh;
499 struct netconfmsg *ncm;
501 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
506 ncm = nlmsg_data(nlh);
507 ncm->ncm_family = AF_INET6;
509 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
510 goto nla_put_failure;
512 /* type -1 is used for ALL */
513 if ((type == -1 || type == NETCONFA_FORWARDING) &&
514 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
515 goto nla_put_failure;
516 #ifdef CONFIG_IPV6_MROUTE
517 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
518 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
519 devconf->mc_forwarding) < 0)
520 goto nla_put_failure;
522 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
523 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
524 goto nla_put_failure;
526 if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
527 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
528 devconf->ignore_routes_with_linkdown) < 0)
529 goto nla_put_failure;
535 nlmsg_cancel(skb, nlh);
539 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
540 struct ipv6_devconf *devconf)
545 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
549 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
550 RTM_NEWNETCONF, 0, type);
552 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
553 WARN_ON(err == -EMSGSIZE);
557 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
560 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
563 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
564 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
565 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
566 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
567 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
570 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
571 struct nlmsghdr *nlh)
573 struct net *net = sock_net(in_skb->sk);
574 struct nlattr *tb[NETCONFA_MAX+1];
575 struct netconfmsg *ncm;
577 struct ipv6_devconf *devconf;
578 struct inet6_dev *in6_dev;
579 struct net_device *dev;
583 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
584 devconf_ipv6_policy);
589 if (!tb[NETCONFA_IFINDEX])
592 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
594 case NETCONFA_IFINDEX_ALL:
595 devconf = net->ipv6.devconf_all;
597 case NETCONFA_IFINDEX_DEFAULT:
598 devconf = net->ipv6.devconf_dflt;
601 dev = __dev_get_by_index(net, ifindex);
604 in6_dev = __in6_dev_get(dev);
607 devconf = &in6_dev->cnf;
612 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
616 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
617 NETLINK_CB(in_skb).portid,
618 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
621 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
622 WARN_ON(err == -EMSGSIZE);
626 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
631 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
632 struct netlink_callback *cb)
634 struct net *net = sock_net(skb->sk);
637 struct net_device *dev;
638 struct inet6_dev *idev;
639 struct hlist_head *head;
642 s_idx = idx = cb->args[1];
644 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
646 head = &net->dev_index_head[h];
648 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
650 hlist_for_each_entry_rcu(dev, head, index_hlist) {
653 idev = __in6_dev_get(dev);
657 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
659 NETLINK_CB(cb->skb).portid,
667 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
673 if (h == NETDEV_HASHENTRIES) {
674 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
675 net->ipv6.devconf_all,
676 NETLINK_CB(cb->skb).portid,
678 RTM_NEWNETCONF, NLM_F_MULTI,
684 if (h == NETDEV_HASHENTRIES + 1) {
685 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
686 net->ipv6.devconf_dflt,
687 NETLINK_CB(cb->skb).portid,
689 RTM_NEWNETCONF, NLM_F_MULTI,
703 static void dev_forward_change(struct inet6_dev *idev)
705 struct net_device *dev;
706 struct inet6_ifaddr *ifa;
711 if (idev->cnf.forwarding)
712 dev_disable_lro(dev);
713 if (dev->flags & IFF_MULTICAST) {
714 if (idev->cnf.forwarding) {
715 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
716 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
717 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
719 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
720 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
721 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
725 list_for_each_entry(ifa, &idev->addr_list, if_list) {
726 if (ifa->flags&IFA_F_TENTATIVE)
728 if (idev->cnf.forwarding)
729 addrconf_join_anycast(ifa);
731 addrconf_leave_anycast(ifa);
733 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
734 dev->ifindex, &idev->cnf);
738 static void addrconf_forward_change(struct net *net, __s32 newf)
740 struct net_device *dev;
741 struct inet6_dev *idev;
743 for_each_netdev(net, dev) {
744 idev = __in6_dev_get(dev);
746 int changed = (!idev->cnf.forwarding) ^ (!newf);
747 idev->cnf.forwarding = newf;
749 dev_forward_change(idev);
754 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
760 return restart_syscall();
762 net = (struct net *)table->extra2;
766 if (p == &net->ipv6.devconf_dflt->forwarding) {
767 if ((!newf) ^ (!old))
768 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
769 NETCONFA_IFINDEX_DEFAULT,
770 net->ipv6.devconf_dflt);
775 if (p == &net->ipv6.devconf_all->forwarding) {
776 net->ipv6.devconf_dflt->forwarding = newf;
777 addrconf_forward_change(net, newf);
778 if ((!newf) ^ (!old))
779 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
780 NETCONFA_IFINDEX_ALL,
781 net->ipv6.devconf_all);
782 } else if ((!newf) ^ (!old))
783 dev_forward_change((struct inet6_dev *)table->extra1);
787 rt6_purge_dflt_routers(net);
791 static void addrconf_linkdown_change(struct net *net, __s32 newf)
793 struct net_device *dev;
794 struct inet6_dev *idev;
796 for_each_netdev(net, dev) {
797 idev = __in6_dev_get(dev);
799 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
801 idev->cnf.ignore_routes_with_linkdown = newf;
803 inet6_netconf_notify_devconf(dev_net(dev),
804 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
811 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
817 return restart_syscall();
819 net = (struct net *)table->extra2;
823 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
824 if ((!newf) ^ (!old))
825 inet6_netconf_notify_devconf(net,
826 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
827 NETCONFA_IFINDEX_DEFAULT,
828 net->ipv6.devconf_dflt);
833 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
834 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
835 addrconf_linkdown_change(net, newf);
836 if ((!newf) ^ (!old))
837 inet6_netconf_notify_devconf(net,
838 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
839 NETCONFA_IFINDEX_ALL,
840 net->ipv6.devconf_all);
849 /* Nobody refers to this ifaddr, destroy it */
850 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
852 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
854 #ifdef NET_REFCNT_DEBUG
855 pr_debug("%s\n", __func__);
858 in6_dev_put(ifp->idev);
860 if (cancel_delayed_work(&ifp->dad_work))
861 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
864 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
865 pr_warn("Freeing alive inet6 address %p\n", ifp);
874 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
877 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
880 * Each device address list is sorted in order of scope -
881 * global before linklocal.
883 list_for_each(p, &idev->addr_list) {
884 struct inet6_ifaddr *ifa
885 = list_entry(p, struct inet6_ifaddr, if_list);
886 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
890 list_add_tail(&ifp->if_list, p);
893 static u32 inet6_addr_hash(const struct in6_addr *addr)
895 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
898 /* On success it returns ifp with increased reference count */
900 static struct inet6_ifaddr *
901 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
902 const struct in6_addr *peer_addr, int pfxlen,
903 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
905 struct inet6_ifaddr *ifa = NULL;
909 int addr_type = ipv6_addr_type(addr);
911 if (addr_type == IPV6_ADDR_ANY ||
912 addr_type & IPV6_ADDR_MULTICAST ||
913 (!(idev->dev->flags & IFF_LOOPBACK) &&
914 addr_type & IPV6_ADDR_LOOPBACK))
915 return ERR_PTR(-EADDRNOTAVAIL);
919 err = -ENODEV; /*XXX*/
923 if (idev->cnf.disable_ipv6) {
928 spin_lock(&addrconf_hash_lock);
930 /* Ignore adding duplicate addresses on an interface */
931 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
932 ADBG("ipv6_add_addr: already assigned\n");
937 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
940 ADBG("ipv6_add_addr: malloc failed\n");
945 rt = addrconf_dst_alloc(idev, addr, false);
951 neigh_parms_data_state_setall(idev->nd_parms);
955 ifa->peer_addr = *peer_addr;
957 spin_lock_init(&ifa->lock);
958 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
959 INIT_HLIST_NODE(&ifa->addr_lst);
961 ifa->prefix_len = pfxlen;
962 ifa->flags = flags | IFA_F_TENTATIVE;
963 ifa->valid_lft = valid_lft;
964 ifa->prefered_lft = prefered_lft;
965 ifa->cstamp = ifa->tstamp = jiffies;
966 ifa->tokenized = false;
975 /* Add to big hash table */
976 hash = inet6_addr_hash(addr);
978 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
979 spin_unlock(&addrconf_hash_lock);
981 write_lock(&idev->lock);
982 /* Add to inet6_dev unicast addr list. */
983 ipv6_link_dev_addr(idev, ifa);
985 if (ifa->flags&IFA_F_TEMPORARY) {
986 list_add(&ifa->tmp_list, &idev->tempaddr_list);
991 write_unlock(&idev->lock);
993 rcu_read_unlock_bh();
995 if (likely(err == 0))
996 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1004 spin_unlock(&addrconf_hash_lock);
1008 enum cleanup_prefix_rt_t {
1009 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1010 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1011 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1015 * Check, whether the prefix for ifp would still need a prefix route
1016 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1019 * 1) we don't purge prefix if address was not permanent.
1020 * prefix is managed by its own lifetime.
1021 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1022 * 3) if there are no addresses, delete prefix.
1023 * 4) if there are still other permanent address(es),
1024 * corresponding prefix is still permanent.
1025 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1026 * don't purge the prefix, assume user space is managing it.
1027 * 6) otherwise, update prefix lifetime to the
1028 * longest valid lifetime among the corresponding
1029 * addresses on the device.
1030 * Note: subsequent RA will update lifetime.
1032 static enum cleanup_prefix_rt_t
1033 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1035 struct inet6_ifaddr *ifa;
1036 struct inet6_dev *idev = ifp->idev;
1037 unsigned long lifetime;
1038 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1042 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1045 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1048 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1049 return CLEANUP_PREFIX_RT_NOP;
1051 action = CLEANUP_PREFIX_RT_EXPIRE;
1053 spin_lock(&ifa->lock);
1055 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1057 * Note: Because this address is
1058 * not permanent, lifetime <
1059 * LONG_MAX / HZ here.
1061 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1062 *expires = ifa->tstamp + lifetime * HZ;
1063 spin_unlock(&ifa->lock);
1070 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1072 struct rt6_info *rt;
1074 rt = addrconf_get_prefix_route(&ifp->addr,
1077 0, RTF_GATEWAY | RTF_DEFAULT);
1082 if (!(rt->rt6i_flags & RTF_EXPIRES))
1083 rt6_set_expires(rt, expires);
1090 /* This function wants to get referenced ifp and releases it before return */
1092 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1095 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1096 unsigned long expires;
1100 spin_lock_bh(&ifp->lock);
1102 ifp->state = INET6_IFADDR_STATE_DEAD;
1103 spin_unlock_bh(&ifp->lock);
1105 if (state == INET6_IFADDR_STATE_DEAD)
1108 spin_lock_bh(&addrconf_hash_lock);
1109 hlist_del_init_rcu(&ifp->addr_lst);
1110 spin_unlock_bh(&addrconf_hash_lock);
1112 write_lock_bh(&ifp->idev->lock);
1114 if (ifp->flags&IFA_F_TEMPORARY) {
1115 list_del(&ifp->tmp_list);
1117 in6_ifa_put(ifp->ifpub);
1123 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1124 action = check_cleanup_prefix_route(ifp, &expires);
1126 list_del_init(&ifp->if_list);
1129 write_unlock_bh(&ifp->idev->lock);
1131 addrconf_del_dad_work(ifp);
1133 ipv6_ifa_notify(RTM_DELADDR, ifp);
1135 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1137 if (action != CLEANUP_PREFIX_RT_NOP) {
1138 cleanup_prefix_route(ifp, expires,
1139 action == CLEANUP_PREFIX_RT_DEL);
1142 /* clean up prefsrc entries */
1143 rt6_remove_prefsrc(ifp);
1148 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1150 struct inet6_dev *idev = ifp->idev;
1151 struct in6_addr addr, *tmpaddr;
1152 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1153 unsigned long regen_advance;
1157 unsigned long now = jiffies;
1159 write_lock_bh(&idev->lock);
1161 spin_lock_bh(&ift->lock);
1162 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1163 spin_unlock_bh(&ift->lock);
1170 if (idev->cnf.use_tempaddr <= 0) {
1171 write_unlock_bh(&idev->lock);
1172 pr_info("%s: use_tempaddr is disabled\n", __func__);
1177 spin_lock_bh(&ifp->lock);
1178 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1179 idev->cnf.use_tempaddr = -1; /*XXX*/
1180 spin_unlock_bh(&ifp->lock);
1181 write_unlock_bh(&idev->lock);
1182 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1189 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1190 __ipv6_try_regen_rndid(idev, tmpaddr);
1191 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1192 age = (now - ifp->tstamp) / HZ;
1193 tmp_valid_lft = min_t(__u32,
1195 idev->cnf.temp_valid_lft + age);
1196 tmp_prefered_lft = min_t(__u32,
1198 idev->cnf.temp_prefered_lft + age -
1199 idev->cnf.max_desync_factor);
1200 tmp_plen = ifp->prefix_len;
1201 tmp_tstamp = ifp->tstamp;
1202 spin_unlock_bh(&ifp->lock);
1204 regen_advance = idev->cnf.regen_max_retry *
1205 idev->cnf.dad_transmits *
1206 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1207 write_unlock_bh(&idev->lock);
1209 /* A temporary address is created only if this calculated Preferred
1210 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1211 * an implementation must not create a temporary address with a zero
1212 * Preferred Lifetime.
1213 * Use age calculation as in addrconf_verify to avoid unnecessary
1214 * temporary addresses being generated.
1216 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1217 if (tmp_prefered_lft <= regen_advance + age) {
1224 addr_flags = IFA_F_TEMPORARY;
1225 /* set in addrconf_prefix_rcv() */
1226 if (ifp->flags & IFA_F_OPTIMISTIC)
1227 addr_flags |= IFA_F_OPTIMISTIC;
1229 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1230 ipv6_addr_scope(&addr), addr_flags,
1231 tmp_valid_lft, tmp_prefered_lft);
1235 pr_info("%s: retry temporary address regeneration\n", __func__);
1237 write_lock_bh(&idev->lock);
1241 spin_lock_bh(&ift->lock);
1244 ift->tstamp = tmp_tstamp;
1245 spin_unlock_bh(&ift->lock);
1247 addrconf_dad_start(ift);
1255 * Choose an appropriate source address (RFC3484)
1258 IPV6_SADDR_RULE_INIT = 0,
1259 IPV6_SADDR_RULE_LOCAL,
1260 IPV6_SADDR_RULE_SCOPE,
1261 IPV6_SADDR_RULE_PREFERRED,
1262 #ifdef CONFIG_IPV6_MIP6
1263 IPV6_SADDR_RULE_HOA,
1265 IPV6_SADDR_RULE_OIF,
1266 IPV6_SADDR_RULE_LABEL,
1267 IPV6_SADDR_RULE_PRIVACY,
1268 IPV6_SADDR_RULE_ORCHID,
1269 IPV6_SADDR_RULE_PREFIX,
1270 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1271 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1276 struct ipv6_saddr_score {
1279 struct inet6_ifaddr *ifa;
1280 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1285 struct ipv6_saddr_dst {
1286 const struct in6_addr *addr;
1293 static inline int ipv6_saddr_preferred(int type)
1295 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1300 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1302 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1303 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1309 static int ipv6_get_saddr_eval(struct net *net,
1310 struct ipv6_saddr_score *score,
1311 struct ipv6_saddr_dst *dst,
1316 if (i <= score->rule) {
1318 case IPV6_SADDR_RULE_SCOPE:
1319 ret = score->scopedist;
1321 case IPV6_SADDR_RULE_PREFIX:
1322 ret = score->matchlen;
1325 ret = !!test_bit(i, score->scorebits);
1331 case IPV6_SADDR_RULE_INIT:
1332 /* Rule 0: remember if hiscore is not ready yet */
1335 case IPV6_SADDR_RULE_LOCAL:
1336 /* Rule 1: Prefer same address */
1337 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1339 case IPV6_SADDR_RULE_SCOPE:
1340 /* Rule 2: Prefer appropriate scope
1345 * ---+--+-+---> scope
1347 * | d is scope of the destination.
1349 * | \ <- smaller scope is better if
1350 * B-15 | \ if scope is enough for destination.
1351 * | ret = B - scope (-1 <= scope >= d <= 15).
1353 * |/ <- greater is better
1354 * -C / if scope is not enough for destination.
1355 * /| ret = scope - C (-1 <= d < scope <= 15).
1357 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1358 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1359 * Assume B = 0 and we get C > 29.
1361 ret = __ipv6_addr_src_scope(score->addr_type);
1362 if (ret >= dst->scope)
1365 ret -= 128; /* 30 is enough */
1366 score->scopedist = ret;
1368 case IPV6_SADDR_RULE_PREFERRED:
1370 /* Rule 3: Avoid deprecated and optimistic addresses */
1371 u8 avoid = IFA_F_DEPRECATED;
1373 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1374 avoid |= IFA_F_OPTIMISTIC;
1375 ret = ipv6_saddr_preferred(score->addr_type) ||
1376 !(score->ifa->flags & avoid);
1379 #ifdef CONFIG_IPV6_MIP6
1380 case IPV6_SADDR_RULE_HOA:
1382 /* Rule 4: Prefer home address */
1383 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1384 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1388 case IPV6_SADDR_RULE_OIF:
1389 /* Rule 5: Prefer outgoing interface */
1390 ret = (!dst->ifindex ||
1391 dst->ifindex == score->ifa->idev->dev->ifindex);
1393 case IPV6_SADDR_RULE_LABEL:
1394 /* Rule 6: Prefer matching label */
1395 ret = ipv6_addr_label(net,
1396 &score->ifa->addr, score->addr_type,
1397 score->ifa->idev->dev->ifindex) == dst->label;
1399 case IPV6_SADDR_RULE_PRIVACY:
1401 /* Rule 7: Prefer public address
1402 * Note: prefer temporary address if use_tempaddr >= 2
1404 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1405 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1406 score->ifa->idev->cnf.use_tempaddr >= 2;
1407 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1410 case IPV6_SADDR_RULE_ORCHID:
1411 /* Rule 8-: Prefer ORCHID vs ORCHID or
1412 * non-ORCHID vs non-ORCHID
1414 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1415 ipv6_addr_orchid(dst->addr));
1417 case IPV6_SADDR_RULE_PREFIX:
1418 /* Rule 8: Use longest matching prefix */
1419 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1420 if (ret > score->ifa->prefix_len)
1421 ret = score->ifa->prefix_len;
1422 score->matchlen = ret;
1424 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1425 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1426 /* Optimistic addresses still have lower precedence than other
1427 * preferred addresses.
1429 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1437 __set_bit(i, score->scorebits);
1443 static int __ipv6_dev_get_saddr(struct net *net,
1444 struct ipv6_saddr_dst *dst,
1445 struct inet6_dev *idev,
1446 struct ipv6_saddr_score *scores,
1449 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1451 read_lock_bh(&idev->lock);
1452 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1456 * - Tentative Address (RFC2462 section 5.4)
1457 * - A tentative address is not considered
1458 * "assigned to an interface" in the traditional
1459 * sense, unless it is also flagged as optimistic.
1460 * - Candidate Source Address (section 4)
1461 * - In any case, anycast addresses, multicast
1462 * addresses, and the unspecified address MUST
1463 * NOT be included in a candidate set.
1465 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1466 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1469 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1471 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1472 score->addr_type & IPV6_ADDR_MULTICAST)) {
1473 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1479 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1481 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1482 int minihiscore, miniscore;
1484 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1485 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1487 if (minihiscore > miniscore) {
1488 if (i == IPV6_SADDR_RULE_SCOPE &&
1489 score->scopedist > 0) {
1492 * each remaining entry
1493 * has too small (not enough)
1494 * scope, because ifa entries
1495 * are sorted by their scope
1501 } else if (minihiscore < miniscore) {
1503 in6_ifa_put(hiscore->ifa);
1505 in6_ifa_hold(score->ifa);
1507 swap(hiscore, score);
1508 hiscore_idx = 1 - hiscore_idx;
1510 /* restore our iterator */
1511 score->ifa = hiscore->ifa;
1518 read_unlock_bh(&idev->lock);
1522 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1523 const struct in6_addr *daddr, unsigned int prefs,
1524 struct in6_addr *saddr)
1526 struct ipv6_saddr_score scores[2], *hiscore;
1527 struct ipv6_saddr_dst dst;
1528 struct inet6_dev *idev;
1529 struct net_device *dev;
1531 bool use_oif_addr = false;
1532 int hiscore_idx = 0;
1534 dst_type = __ipv6_addr_type(daddr);
1536 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1537 dst.scope = __ipv6_addr_src_scope(dst_type);
1538 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1541 scores[hiscore_idx].rule = -1;
1542 scores[hiscore_idx].ifa = NULL;
1546 /* Candidate Source Address (section 4)
1547 * - multicast and link-local destination address,
1548 * the set of candidate source address MUST only
1549 * include addresses assigned to interfaces
1550 * belonging to the same link as the outgoing
1552 * (- For site-local destination addresses, the
1553 * set of candidate source addresses MUST only
1554 * include addresses assigned to interfaces
1555 * belonging to the same site as the outgoing
1557 * - "It is RECOMMENDED that the candidate source addresses
1558 * be the set of unicast addresses assigned to the
1559 * interface that will be used to send to the destination
1560 * (the 'outgoing' interface)." (RFC 6724)
1563 idev = __in6_dev_get(dst_dev);
1564 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1565 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1566 (idev && idev->cnf.use_oif_addrs_only)) {
1567 use_oif_addr = true;
1573 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1575 for_each_netdev_rcu(net, dev) {
1576 idev = __in6_dev_get(dev);
1579 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1584 hiscore = &scores[hiscore_idx];
1586 return -EADDRNOTAVAIL;
1588 *saddr = hiscore->ifa->addr;
1589 in6_ifa_put(hiscore->ifa);
1592 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1594 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1597 struct inet6_ifaddr *ifp;
1598 int err = -EADDRNOTAVAIL;
1600 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1601 if (ifp->scope > IFA_LINK)
1603 if (ifp->scope == IFA_LINK &&
1604 !(ifp->flags & banned_flags)) {
1613 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1616 struct inet6_dev *idev;
1617 int err = -EADDRNOTAVAIL;
1620 idev = __in6_dev_get(dev);
1622 read_lock_bh(&idev->lock);
1623 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1624 read_unlock_bh(&idev->lock);
1630 static int ipv6_count_addresses(struct inet6_dev *idev)
1633 struct inet6_ifaddr *ifp;
1635 read_lock_bh(&idev->lock);
1636 list_for_each_entry(ifp, &idev->addr_list, if_list)
1638 read_unlock_bh(&idev->lock);
1642 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1643 const struct net_device *dev, int strict)
1645 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1647 EXPORT_SYMBOL(ipv6_chk_addr);
1649 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1650 const struct net_device *dev, int strict,
1653 struct inet6_ifaddr *ifp;
1654 unsigned int hash = inet6_addr_hash(addr);
1658 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1659 if (!net_eq(dev_net(ifp->idev->dev), net))
1661 /* Decouple optimistic from tentative for evaluation here.
1662 * Ban optimistic addresses explicitly, when required.
1664 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1665 ? (ifp->flags&~IFA_F_TENTATIVE)
1667 if (ipv6_addr_equal(&ifp->addr, addr) &&
1668 !(ifp_flags&banned_flags) &&
1669 (!dev || ifp->idev->dev == dev ||
1670 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1671 rcu_read_unlock_bh();
1676 rcu_read_unlock_bh();
1679 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1681 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1682 struct net_device *dev)
1684 unsigned int hash = inet6_addr_hash(addr);
1685 struct inet6_ifaddr *ifp;
1687 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1688 if (!net_eq(dev_net(ifp->idev->dev), net))
1690 if (ipv6_addr_equal(&ifp->addr, addr)) {
1691 if (!dev || ifp->idev->dev == dev)
1698 /* Compares an address/prefix_len with addresses on device @dev.
1699 * If one is found it returns true.
1701 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1702 const unsigned int prefix_len, struct net_device *dev)
1704 struct inet6_dev *idev;
1705 struct inet6_ifaddr *ifa;
1709 idev = __in6_dev_get(dev);
1711 read_lock_bh(&idev->lock);
1712 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1713 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1717 read_unlock_bh(&idev->lock);
1723 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1725 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1727 struct inet6_dev *idev;
1728 struct inet6_ifaddr *ifa;
1733 idev = __in6_dev_get(dev);
1735 read_lock_bh(&idev->lock);
1736 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1737 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1742 read_unlock_bh(&idev->lock);
1747 EXPORT_SYMBOL(ipv6_chk_prefix);
1749 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1750 struct net_device *dev, int strict)
1752 struct inet6_ifaddr *ifp, *result = NULL;
1753 unsigned int hash = inet6_addr_hash(addr);
1756 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1757 if (!net_eq(dev_net(ifp->idev->dev), net))
1759 if (ipv6_addr_equal(&ifp->addr, addr)) {
1760 if (!dev || ifp->idev->dev == dev ||
1761 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1768 rcu_read_unlock_bh();
1773 /* Gets referenced address, destroys ifaddr */
1775 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1778 ifp->flags |= IFA_F_DADFAILED;
1780 if (ifp->flags&IFA_F_PERMANENT) {
1781 spin_lock_bh(&ifp->lock);
1782 addrconf_del_dad_work(ifp);
1783 ifp->flags |= IFA_F_TENTATIVE;
1784 spin_unlock_bh(&ifp->lock);
1786 ipv6_ifa_notify(0, ifp);
1788 } else if (ifp->flags&IFA_F_TEMPORARY) {
1789 struct inet6_ifaddr *ifpub;
1790 spin_lock_bh(&ifp->lock);
1793 in6_ifa_hold(ifpub);
1794 spin_unlock_bh(&ifp->lock);
1795 ipv6_create_tempaddr(ifpub, ifp);
1798 spin_unlock_bh(&ifp->lock);
1806 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1810 spin_lock_bh(&ifp->lock);
1811 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1812 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1815 spin_unlock_bh(&ifp->lock);
1820 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1822 struct in6_addr addr;
1823 struct inet6_dev *idev = ifp->idev;
1824 struct net *net = dev_net(ifp->idev->dev);
1826 if (addrconf_dad_end(ifp)) {
1831 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1832 ifp->idev->dev->name, &ifp->addr);
1834 spin_lock_bh(&ifp->lock);
1836 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1837 int scope = ifp->scope;
1838 u32 flags = ifp->flags;
1839 struct in6_addr new_addr;
1840 struct inet6_ifaddr *ifp2;
1841 u32 valid_lft, preferred_lft;
1842 int pfxlen = ifp->prefix_len;
1843 int retries = ifp->stable_privacy_retry + 1;
1845 if (retries > net->ipv6.sysctl.idgen_retries) {
1846 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1847 ifp->idev->dev->name);
1851 new_addr = ifp->addr;
1852 if (ipv6_generate_stable_address(&new_addr, retries,
1856 valid_lft = ifp->valid_lft;
1857 preferred_lft = ifp->prefered_lft;
1859 spin_unlock_bh(&ifp->lock);
1861 if (idev->cnf.max_addresses &&
1862 ipv6_count_addresses(idev) >=
1863 idev->cnf.max_addresses)
1866 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1867 ifp->idev->dev->name);
1869 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1870 scope, flags, valid_lft,
1875 spin_lock_bh(&ifp2->lock);
1876 ifp2->stable_privacy_retry = retries;
1877 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1878 spin_unlock_bh(&ifp2->lock);
1880 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1883 spin_lock_bh(&ifp->lock);
1884 } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1885 addr.s6_addr32[0] = htonl(0xfe800000);
1886 addr.s6_addr32[1] = 0;
1888 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1889 ipv6_addr_equal(&ifp->addr, &addr)) {
1890 /* DAD failed for link-local based on MAC address */
1891 idev->cnf.disable_ipv6 = 1;
1893 pr_info("%s: IPv6 being disabled!\n",
1894 ifp->idev->dev->name);
1899 /* transition from _POSTDAD to _ERRDAD */
1900 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1901 spin_unlock_bh(&ifp->lock);
1903 addrconf_mod_dad_work(ifp, 0);
1906 /* Join to solicited addr multicast group.
1907 * caller must hold RTNL */
1908 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1910 struct in6_addr maddr;
1912 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1915 addrconf_addr_solict_mult(addr, &maddr);
1916 ipv6_dev_mc_inc(dev, &maddr);
1919 /* caller must hold RTNL */
1920 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1922 struct in6_addr maddr;
1924 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1927 addrconf_addr_solict_mult(addr, &maddr);
1928 __ipv6_dev_mc_dec(idev, &maddr);
1931 /* caller must hold RTNL */
1932 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1934 struct in6_addr addr;
1936 if (ifp->prefix_len >= 127) /* RFC 6164 */
1938 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1939 if (ipv6_addr_any(&addr))
1941 __ipv6_dev_ac_inc(ifp->idev, &addr);
1944 /* caller must hold RTNL */
1945 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1947 struct in6_addr addr;
1949 if (ifp->prefix_len >= 127) /* RFC 6164 */
1951 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1952 if (ipv6_addr_any(&addr))
1954 __ipv6_dev_ac_dec(ifp->idev, &addr);
1957 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1959 if (dev->addr_len != EUI64_ADDR_LEN)
1961 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
1966 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1968 union fwnet_hwaddr *ha;
1970 if (dev->addr_len != FWNET_ALEN)
1973 ha = (union fwnet_hwaddr *)dev->dev_addr;
1975 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1980 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1982 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1983 if (dev->addr_len != ARCNET_ALEN)
1986 eui[7] = *(u8 *)dev->dev_addr;
1990 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1992 if (dev->addr_len != INFINIBAND_ALEN)
1994 memcpy(eui, dev->dev_addr + 12, 8);
1999 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2003 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2004 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2005 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2006 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2007 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2008 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2012 memcpy(eui + 4, &addr, 4);
2016 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2018 if (dev->priv_flags & IFF_ISATAP)
2019 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2023 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2025 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2028 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2030 memcpy(eui, dev->perm_addr, 3);
2031 memcpy(eui + 5, dev->perm_addr + 3, 3);
2038 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2040 switch (dev->type) {
2043 return addrconf_ifid_eui48(eui, dev);
2045 return addrconf_ifid_arcnet(eui, dev);
2046 case ARPHRD_INFINIBAND:
2047 return addrconf_ifid_infiniband(eui, dev);
2049 return addrconf_ifid_sit(eui, dev);
2051 return addrconf_ifid_gre(eui, dev);
2052 case ARPHRD_6LOWPAN:
2053 return addrconf_ifid_eui64(eui, dev);
2054 case ARPHRD_IEEE1394:
2055 return addrconf_ifid_ieee1394(eui, dev);
2056 case ARPHRD_TUNNEL6:
2057 return addrconf_ifid_ip6tnl(eui, dev);
2062 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2065 struct inet6_ifaddr *ifp;
2067 read_lock_bh(&idev->lock);
2068 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2069 if (ifp->scope > IFA_LINK)
2071 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2072 memcpy(eui, ifp->addr.s6_addr+8, 8);
2077 read_unlock_bh(&idev->lock);
2081 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2082 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2085 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2086 idev->rndid[0] &= ~0x02;
2089 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2090 * check if generated address is not inappropriate
2092 * - Reserved subnet anycast (RFC 2526)
2093 * 11111101 11....11 1xxxxxxx
2094 * - ISATAP (RFC4214) 6.1
2095 * 00-00-5E-FE-xx-xx-xx-xx
2097 * - XXX: already assigned to an address on the device
2099 if (idev->rndid[0] == 0xfd &&
2100 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2101 (idev->rndid[7]&0x80))
2103 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2104 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2106 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2111 static void ipv6_regen_rndid(unsigned long data)
2113 struct inet6_dev *idev = (struct inet6_dev *) data;
2114 unsigned long expires;
2117 write_lock_bh(&idev->lock);
2122 __ipv6_regen_rndid(idev);
2125 idev->cnf.temp_prefered_lft * HZ -
2126 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2127 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2128 idev->cnf.max_desync_factor * HZ;
2129 if (time_before(expires, jiffies)) {
2130 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2131 __func__, idev->dev->name);
2135 if (!mod_timer(&idev->regen_timer, expires))
2139 write_unlock_bh(&idev->lock);
2140 rcu_read_unlock_bh();
2144 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2146 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2147 __ipv6_regen_rndid(idev);
2155 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2156 unsigned long expires, u32 flags)
2158 struct fib6_config cfg = {
2159 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2160 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2161 .fc_ifindex = dev->ifindex,
2162 .fc_expires = expires,
2164 .fc_flags = RTF_UP | flags,
2165 .fc_nlinfo.nl_net = dev_net(dev),
2166 .fc_protocol = RTPROT_KERNEL,
2171 /* Prevent useless cloning on PtP SIT.
2172 This thing is done here expecting that the whole
2173 class of non-broadcast devices need not cloning.
2175 #if IS_ENABLED(CONFIG_IPV6_SIT)
2176 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2177 cfg.fc_flags |= RTF_NONEXTHOP;
2180 ip6_route_add(&cfg);
2184 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2186 const struct net_device *dev,
2187 u32 flags, u32 noflags)
2189 struct fib6_node *fn;
2190 struct rt6_info *rt = NULL;
2191 struct fib6_table *table;
2192 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2194 table = fib6_get_table(dev_net(dev), tb_id);
2198 read_lock_bh(&table->tb6_lock);
2199 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2203 noflags |= RTF_CACHE;
2204 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2205 if (rt->dst.dev->ifindex != dev->ifindex)
2207 if ((rt->rt6i_flags & flags) != flags)
2209 if ((rt->rt6i_flags & noflags) != 0)
2215 read_unlock_bh(&table->tb6_lock);
2220 /* Create "default" multicast route to the interface */
2222 static void addrconf_add_mroute(struct net_device *dev)
2224 struct fib6_config cfg = {
2225 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2226 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2227 .fc_ifindex = dev->ifindex,
2230 .fc_nlinfo.nl_net = dev_net(dev),
2233 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2235 ip6_route_add(&cfg);
2238 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2240 struct inet6_dev *idev;
2244 idev = ipv6_find_idev(dev);
2246 return ERR_PTR(-ENOBUFS);
2248 if (idev->cnf.disable_ipv6)
2249 return ERR_PTR(-EACCES);
2251 /* Add default multicast route */
2252 if (!(dev->flags & IFF_LOOPBACK))
2253 addrconf_add_mroute(dev);
2258 static void manage_tempaddrs(struct inet6_dev *idev,
2259 struct inet6_ifaddr *ifp,
2260 __u32 valid_lft, __u32 prefered_lft,
2261 bool create, unsigned long now)
2264 struct inet6_ifaddr *ift;
2266 read_lock_bh(&idev->lock);
2267 /* update all temporary addresses in the list */
2268 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2269 int age, max_valid, max_prefered;
2271 if (ifp != ift->ifpub)
2274 /* RFC 4941 section 3.3:
2275 * If a received option will extend the lifetime of a public
2276 * address, the lifetimes of temporary addresses should
2277 * be extended, subject to the overall constraint that no
2278 * temporary addresses should ever remain "valid" or "preferred"
2279 * for a time longer than (TEMP_VALID_LIFETIME) or
2280 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2282 age = (now - ift->cstamp) / HZ;
2283 max_valid = idev->cnf.temp_valid_lft - age;
2287 max_prefered = idev->cnf.temp_prefered_lft -
2288 idev->cnf.max_desync_factor - age;
2289 if (max_prefered < 0)
2292 if (valid_lft > max_valid)
2293 valid_lft = max_valid;
2295 if (prefered_lft > max_prefered)
2296 prefered_lft = max_prefered;
2298 spin_lock(&ift->lock);
2300 ift->valid_lft = valid_lft;
2301 ift->prefered_lft = prefered_lft;
2303 if (prefered_lft > 0)
2304 ift->flags &= ~IFA_F_DEPRECATED;
2306 spin_unlock(&ift->lock);
2307 if (!(flags&IFA_F_TENTATIVE))
2308 ipv6_ifa_notify(0, ift);
2311 if ((create || list_empty(&idev->tempaddr_list)) &&
2312 idev->cnf.use_tempaddr > 0) {
2313 /* When a new public address is created as described
2314 * in [ADDRCONF], also create a new temporary address.
2315 * Also create a temporary address if it's enabled but
2316 * no temporary address currently exists.
2318 read_unlock_bh(&idev->lock);
2319 ipv6_create_tempaddr(ifp, NULL);
2321 read_unlock_bh(&idev->lock);
2325 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2327 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2328 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2331 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2333 struct prefix_info *pinfo;
2338 struct inet6_dev *in6_dev;
2339 struct net *net = dev_net(dev);
2341 pinfo = (struct prefix_info *) opt;
2343 if (len < sizeof(struct prefix_info)) {
2344 ADBG("addrconf: prefix option too short\n");
2349 * Validation checks ([ADDRCONF], page 19)
2352 addr_type = ipv6_addr_type(&pinfo->prefix);
2354 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2357 valid_lft = ntohl(pinfo->valid);
2358 prefered_lft = ntohl(pinfo->prefered);
2360 if (prefered_lft > valid_lft) {
2361 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2365 in6_dev = in6_dev_get(dev);
2368 net_dbg_ratelimited("addrconf: device %s not configured\n",
2374 * Two things going on here:
2375 * 1) Add routes for on-link prefixes
2376 * 2) Configure prefixes with the auto flag set
2379 if (pinfo->onlink) {
2380 struct rt6_info *rt;
2381 unsigned long rt_expires;
2383 /* Avoid arithmetic overflow. Really, we could
2384 * save rt_expires in seconds, likely valid_lft,
2385 * but it would require division in fib gc, that it
2389 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2391 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2393 if (addrconf_finite_timeout(rt_expires))
2396 rt = addrconf_get_prefix_route(&pinfo->prefix,
2399 RTF_ADDRCONF | RTF_PREFIX_RT,
2400 RTF_GATEWAY | RTF_DEFAULT);
2403 /* Autoconf prefix route */
2404 if (valid_lft == 0) {
2407 } else if (addrconf_finite_timeout(rt_expires)) {
2409 rt6_set_expires(rt, jiffies + rt_expires);
2411 rt6_clean_expires(rt);
2413 } else if (valid_lft) {
2414 clock_t expires = 0;
2415 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2416 if (addrconf_finite_timeout(rt_expires)) {
2418 flags |= RTF_EXPIRES;
2419 expires = jiffies_to_clock_t(rt_expires);
2421 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2422 dev, expires, flags);
2427 /* Try to figure out our local address for this prefix */
2429 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2430 struct inet6_ifaddr *ifp;
2431 struct in6_addr addr;
2432 int create = 0, update_lft = 0;
2433 bool tokenized = false;
2435 if (pinfo->prefix_len == 64) {
2436 memcpy(&addr, &pinfo->prefix, 8);
2438 if (!ipv6_addr_any(&in6_dev->token)) {
2439 read_lock_bh(&in6_dev->lock);
2440 memcpy(addr.s6_addr + 8,
2441 in6_dev->token.s6_addr + 8, 8);
2442 read_unlock_bh(&in6_dev->lock);
2444 } else if (is_addr_mode_generate_stable(in6_dev) &&
2445 !ipv6_generate_stable_address(&addr, 0,
2447 addr_flags |= IFA_F_STABLE_PRIVACY;
2449 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2450 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2451 in6_dev_put(in6_dev);
2456 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2458 in6_dev_put(in6_dev);
2463 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2465 if (!ifp && valid_lft) {
2466 int max_addresses = in6_dev->cnf.max_addresses;
2468 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2469 if (in6_dev->cnf.optimistic_dad &&
2470 !net->ipv6.devconf_all->forwarding && sllao)
2471 addr_flags |= IFA_F_OPTIMISTIC;
2474 /* Do not allow to create too much of autoconfigured
2475 * addresses; this would be too easy way to crash kernel.
2477 if (!max_addresses ||
2478 ipv6_count_addresses(in6_dev) < max_addresses)
2479 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2481 addr_type&IPV6_ADDR_SCOPE_MASK,
2482 addr_flags, valid_lft,
2485 if (IS_ERR_OR_NULL(ifp)) {
2486 in6_dev_put(in6_dev);
2492 spin_lock_bh(&ifp->lock);
2493 ifp->flags |= IFA_F_MANAGETEMPADDR;
2494 ifp->cstamp = jiffies;
2495 ifp->tokenized = tokenized;
2496 spin_unlock_bh(&ifp->lock);
2497 addrconf_dad_start(ifp);
2505 /* update lifetime (RFC2462 5.5.3 e) */
2506 spin_lock_bh(&ifp->lock);
2508 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2509 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2512 if (!update_lft && !create && stored_lft) {
2513 const u32 minimum_lft = min_t(u32,
2514 stored_lft, MIN_VALID_LIFETIME);
2515 valid_lft = max(valid_lft, minimum_lft);
2517 /* RFC4862 Section 5.5.3e:
2518 * "Note that the preferred lifetime of the
2519 * corresponding address is always reset to
2520 * the Preferred Lifetime in the received
2521 * Prefix Information option, regardless of
2522 * whether the valid lifetime is also reset or
2525 * So we should always update prefered_lft here.
2531 ifp->valid_lft = valid_lft;
2532 ifp->prefered_lft = prefered_lft;
2535 ifp->flags &= ~IFA_F_DEPRECATED;
2536 spin_unlock_bh(&ifp->lock);
2538 if (!(flags&IFA_F_TENTATIVE))
2539 ipv6_ifa_notify(0, ifp);
2541 spin_unlock_bh(&ifp->lock);
2543 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2550 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2551 in6_dev_put(in6_dev);
2555 * Set destination address.
2556 * Special case for SIT interfaces where we create a new "virtual"
2559 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2561 struct in6_ifreq ireq;
2562 struct net_device *dev;
2568 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2571 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2577 #if IS_ENABLED(CONFIG_IPV6_SIT)
2578 if (dev->type == ARPHRD_SIT) {
2579 const struct net_device_ops *ops = dev->netdev_ops;
2581 struct ip_tunnel_parm p;
2583 err = -EADDRNOTAVAIL;
2584 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2587 memset(&p, 0, sizeof(p));
2588 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2592 p.iph.protocol = IPPROTO_IPV6;
2594 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2596 if (ops->ndo_do_ioctl) {
2597 mm_segment_t oldfs = get_fs();
2600 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2607 dev = __dev_get_by_name(net, p.name);
2610 err = dev_open(dev);
2620 static int ipv6_mc_config(struct sock *sk, bool join,
2621 const struct in6_addr *addr, int ifindex)
2629 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2631 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2638 * Manual configuration of address on an interface
2640 static int inet6_addr_add(struct net *net, int ifindex,
2641 const struct in6_addr *pfx,
2642 const struct in6_addr *peer_pfx,
2643 unsigned int plen, __u32 ifa_flags,
2644 __u32 prefered_lft, __u32 valid_lft)
2646 struct inet6_ifaddr *ifp;
2647 struct inet6_dev *idev;
2648 struct net_device *dev;
2649 unsigned long timeout;
2659 /* check the lifetime */
2660 if (!valid_lft || prefered_lft > valid_lft)
2663 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2666 dev = __dev_get_by_index(net, ifindex);
2670 idev = addrconf_add_dev(dev);
2672 return PTR_ERR(idev);
2674 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2675 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2676 true, pfx, ifindex);
2682 scope = ipv6_addr_scope(pfx);
2684 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2685 if (addrconf_finite_timeout(timeout)) {
2686 expires = jiffies_to_clock_t(timeout * HZ);
2687 valid_lft = timeout;
2688 flags = RTF_EXPIRES;
2692 ifa_flags |= IFA_F_PERMANENT;
2695 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2696 if (addrconf_finite_timeout(timeout)) {
2698 ifa_flags |= IFA_F_DEPRECATED;
2699 prefered_lft = timeout;
2702 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2703 valid_lft, prefered_lft);
2706 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2707 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2712 * Note that section 3.1 of RFC 4429 indicates
2713 * that the Optimistic flag should not be set for
2714 * manually configured addresses
2716 addrconf_dad_start(ifp);
2717 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2718 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2721 addrconf_verify_rtnl();
2723 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2724 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2725 false, pfx, ifindex);
2728 return PTR_ERR(ifp);
2731 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2732 const struct in6_addr *pfx, unsigned int plen)
2734 struct inet6_ifaddr *ifp;
2735 struct inet6_dev *idev;
2736 struct net_device *dev;
2741 dev = __dev_get_by_index(net, ifindex);
2745 idev = __in6_dev_get(dev);
2749 read_lock_bh(&idev->lock);
2750 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2751 if (ifp->prefix_len == plen &&
2752 ipv6_addr_equal(pfx, &ifp->addr)) {
2754 read_unlock_bh(&idev->lock);
2756 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2757 (ifa_flags & IFA_F_MANAGETEMPADDR))
2758 manage_tempaddrs(idev, ifp, 0, 0, false,
2761 addrconf_verify_rtnl();
2762 if (ipv6_addr_is_multicast(pfx)) {
2763 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2764 false, pfx, dev->ifindex);
2769 read_unlock_bh(&idev->lock);
2770 return -EADDRNOTAVAIL;
2774 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2776 struct in6_ifreq ireq;
2779 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2782 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2786 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2787 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2788 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2793 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2795 struct in6_ifreq ireq;
2798 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2801 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2805 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2806 ireq.ifr6_prefixlen);
2811 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2812 int plen, int scope)
2814 struct inet6_ifaddr *ifp;
2816 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2817 scope, IFA_F_PERMANENT,
2818 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2820 spin_lock_bh(&ifp->lock);
2821 ifp->flags &= ~IFA_F_TENTATIVE;
2822 spin_unlock_bh(&ifp->lock);
2823 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2828 #if IS_ENABLED(CONFIG_IPV6_SIT)
2829 static void sit_add_v4_addrs(struct inet6_dev *idev)
2831 struct in6_addr addr;
2832 struct net_device *dev;
2833 struct net *net = dev_net(idev->dev);
2839 memset(&addr, 0, sizeof(struct in6_addr));
2840 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2842 if (idev->dev->flags&IFF_POINTOPOINT) {
2843 addr.s6_addr32[0] = htonl(0xfe800000);
2847 scope = IPV6_ADDR_COMPATv4;
2849 pflags |= RTF_NONEXTHOP;
2852 if (addr.s6_addr32[3]) {
2853 add_addr(idev, &addr, plen, scope);
2854 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2858 for_each_netdev(net, dev) {
2859 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2860 if (in_dev && (dev->flags & IFF_UP)) {
2861 struct in_ifaddr *ifa;
2865 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2867 addr.s6_addr32[3] = ifa->ifa_local;
2869 if (ifa->ifa_scope == RT_SCOPE_LINK)
2871 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2872 if (idev->dev->flags&IFF_POINTOPOINT)
2877 add_addr(idev, &addr, plen, flag);
2878 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2886 static void init_loopback(struct net_device *dev)
2888 struct inet6_dev *idev;
2889 struct net_device *sp_dev;
2890 struct inet6_ifaddr *sp_ifa;
2891 struct rt6_info *sp_rt;
2897 idev = ipv6_find_idev(dev);
2899 pr_debug("%s: add_dev failed\n", __func__);
2903 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2905 /* Add routes to other interface's IPv6 addresses */
2906 for_each_netdev(dev_net(dev), sp_dev) {
2907 if (!strcmp(sp_dev->name, dev->name))
2910 idev = __in6_dev_get(sp_dev);
2914 read_lock_bh(&idev->lock);
2915 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2917 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2921 /* This dst has been added to garbage list when
2922 * lo device down, release this obsolete dst and
2923 * reallocate a new router for ifa.
2925 if (sp_ifa->rt->dst.obsolete > 0) {
2926 ip6_rt_put(sp_ifa->rt);
2933 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2935 /* Failure cases are ignored */
2936 if (!IS_ERR(sp_rt)) {
2941 read_unlock_bh(&idev->lock);
2945 static void addrconf_add_linklocal(struct inet6_dev *idev,
2946 const struct in6_addr *addr, u32 flags)
2948 struct inet6_ifaddr *ifp;
2949 u32 addr_flags = flags | IFA_F_PERMANENT;
2951 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2952 if (idev->cnf.optimistic_dad &&
2953 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2954 addr_flags |= IFA_F_OPTIMISTIC;
2957 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2958 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2960 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2961 addrconf_dad_start(ifp);
2966 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2968 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2971 if (address.s6_addr32[2] == htonl(0x02005eff) &&
2972 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2975 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2976 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2982 static int ipv6_generate_stable_address(struct in6_addr *address,
2984 const struct inet6_dev *idev)
2986 static DEFINE_SPINLOCK(lock);
2987 static __u32 digest[SHA_DIGEST_WORDS];
2988 static __u32 workspace[SHA_WORKSPACE_WORDS];
2991 char __data[SHA_MESSAGE_BYTES];
2993 struct in6_addr secret;
2995 unsigned char hwaddr[MAX_ADDR_LEN];
3000 struct in6_addr secret;
3001 struct in6_addr temp;
3002 struct net *net = dev_net(idev->dev);
3004 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3006 if (idev->cnf.stable_secret.initialized)
3007 secret = idev->cnf.stable_secret.secret;
3008 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3009 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3014 spin_lock_bh(&lock);
3017 memset(&data, 0, sizeof(data));
3018 memset(workspace, 0, sizeof(workspace));
3019 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3020 data.prefix[0] = address->s6_addr32[0];
3021 data.prefix[1] = address->s6_addr32[1];
3022 data.secret = secret;
3023 data.dad_count = dad_count;
3025 sha_transform(digest, data.__data, workspace);
3028 temp.s6_addr32[2] = (__force __be32)digest[0];
3029 temp.s6_addr32[3] = (__force __be32)digest[1];
3031 spin_unlock_bh(&lock);
3033 if (ipv6_reserved_interfaceid(temp)) {
3035 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3044 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3046 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3050 s = &idev->cnf.stable_secret;
3051 get_random_bytes(&s->secret, sizeof(s->secret));
3052 s->initialized = true;
3055 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3057 struct in6_addr addr;
3059 /* no link local addresses on L3 master devices */
3060 if (netif_is_l3_master(idev->dev))
3063 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3065 switch (idev->addr_gen_mode) {
3066 case IN6_ADDR_GEN_MODE_RANDOM:
3067 ipv6_gen_mode_random_init(idev);
3069 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3070 if (!ipv6_generate_stable_address(&addr, 0, idev))
3071 addrconf_add_linklocal(idev, &addr,
3072 IFA_F_STABLE_PRIVACY);
3073 else if (prefix_route)
3074 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3076 case IN6_ADDR_GEN_MODE_EUI64:
3077 /* addrconf_add_linklocal also adds a prefix_route and we
3078 * only need to care about prefix routes if ipv6_generate_eui64
3079 * couldn't generate one.
3081 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3082 addrconf_add_linklocal(idev, &addr, 0);
3083 else if (prefix_route)
3084 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3086 case IN6_ADDR_GEN_MODE_NONE:
3088 /* will not add any link local address */
3093 static void addrconf_dev_config(struct net_device *dev)
3095 struct inet6_dev *idev;
3099 if ((dev->type != ARPHRD_ETHER) &&
3100 (dev->type != ARPHRD_FDDI) &&
3101 (dev->type != ARPHRD_ARCNET) &&
3102 (dev->type != ARPHRD_INFINIBAND) &&
3103 (dev->type != ARPHRD_IEEE1394) &&
3104 (dev->type != ARPHRD_TUNNEL6) &&
3105 (dev->type != ARPHRD_6LOWPAN) &&
3106 (dev->type != ARPHRD_NONE)) {
3107 /* Alas, we support only Ethernet autoconfiguration. */
3111 idev = addrconf_add_dev(dev);
3115 /* this device type has no EUI support */
3116 if (dev->type == ARPHRD_NONE &&
3117 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3118 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3120 addrconf_addr_gen(idev, false);
3123 #if IS_ENABLED(CONFIG_IPV6_SIT)
3124 static void addrconf_sit_config(struct net_device *dev)
3126 struct inet6_dev *idev;
3131 * Configure the tunnel with one of our IPv4
3132 * addresses... we should configure all of
3133 * our v4 addrs in the tunnel
3136 idev = ipv6_find_idev(dev);
3138 pr_debug("%s: add_dev failed\n", __func__);
3142 if (dev->priv_flags & IFF_ISATAP) {
3143 addrconf_addr_gen(idev, false);
3147 sit_add_v4_addrs(idev);
3149 if (dev->flags&IFF_POINTOPOINT)
3150 addrconf_add_mroute(dev);
3154 #if IS_ENABLED(CONFIG_NET_IPGRE)
3155 static void addrconf_gre_config(struct net_device *dev)
3157 struct inet6_dev *idev;
3161 idev = ipv6_find_idev(dev);
3163 pr_debug("%s: add_dev failed\n", __func__);
3167 addrconf_addr_gen(idev, true);
3168 if (dev->flags & IFF_POINTOPOINT)
3169 addrconf_add_mroute(dev);
3173 static int fixup_permanent_addr(struct inet6_dev *idev,
3174 struct inet6_ifaddr *ifp)
3177 struct rt6_info *rt;
3179 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3180 if (unlikely(IS_ERR(rt)))
3186 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3187 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3191 addrconf_dad_start(ifp);
3196 static void addrconf_permanent_addr(struct net_device *dev)
3198 struct inet6_ifaddr *ifp, *tmp;
3199 struct inet6_dev *idev;
3201 idev = __in6_dev_get(dev);
3205 write_lock_bh(&idev->lock);
3207 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3208 if ((ifp->flags & IFA_F_PERMANENT) &&
3209 fixup_permanent_addr(idev, ifp) < 0) {
3210 write_unlock_bh(&idev->lock);
3212 write_lock_bh(&idev->lock);
3214 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3215 idev->dev->name, &ifp->addr);
3219 write_unlock_bh(&idev->lock);
3222 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3225 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3226 struct inet6_dev *idev = __in6_dev_get(dev);
3227 int run_pending = 0;
3231 case NETDEV_REGISTER:
3232 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3233 idev = ipv6_add_dev(dev);
3235 return notifier_from_errno(PTR_ERR(idev));
3239 case NETDEV_CHANGEMTU:
3240 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3241 if (dev->mtu < IPV6_MIN_MTU) {
3242 addrconf_ifdown(dev, 1);
3247 rt6_mtu_change(dev, dev->mtu);
3248 idev->cnf.mtu6 = dev->mtu;
3252 /* allocate new idev */
3253 idev = ipv6_add_dev(dev);
3257 /* device is still not ready */
3258 if (!(idev->if_flags & IF_READY))
3267 if (dev->flags & IFF_SLAVE)
3270 if (idev && idev->cnf.disable_ipv6)
3273 if (event == NETDEV_UP) {
3274 if (!addrconf_qdisc_ok(dev)) {
3275 /* device is not ready yet. */
3276 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3281 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3282 idev = ipv6_add_dev(dev);
3284 if (!IS_ERR_OR_NULL(idev)) {
3285 idev->if_flags |= IF_READY;
3288 } else if (event == NETDEV_CHANGE) {
3289 if (!addrconf_qdisc_ok(dev)) {
3290 /* device is still not ready. */
3295 if (idev->if_flags & IF_READY)
3296 /* device is already configured. */
3298 idev->if_flags |= IF_READY;
3301 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3307 /* restore routes for permanent addresses */
3308 addrconf_permanent_addr(dev);
3310 switch (dev->type) {
3311 #if IS_ENABLED(CONFIG_IPV6_SIT)
3313 addrconf_sit_config(dev);
3316 #if IS_ENABLED(CONFIG_NET_IPGRE)
3318 addrconf_gre_config(dev);
3321 case ARPHRD_LOOPBACK:
3326 addrconf_dev_config(dev);
3330 if (!IS_ERR_OR_NULL(idev)) {
3332 addrconf_dad_run(idev);
3335 * If the MTU changed during the interface down,
3336 * when the interface up, the changed MTU must be
3337 * reflected in the idev as well as routers.
3339 if (idev->cnf.mtu6 != dev->mtu &&
3340 dev->mtu >= IPV6_MIN_MTU) {
3341 rt6_mtu_change(dev, dev->mtu);
3342 idev->cnf.mtu6 = dev->mtu;
3344 idev->tstamp = jiffies;
3345 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3348 * If the changed mtu during down is lower than
3349 * IPV6_MIN_MTU stop IPv6 on this interface.
3351 if (dev->mtu < IPV6_MIN_MTU)
3352 addrconf_ifdown(dev, 1);
3357 case NETDEV_UNREGISTER:
3359 * Remove all addresses from this interface.
3361 addrconf_ifdown(dev, event != NETDEV_DOWN);
3364 case NETDEV_CHANGENAME:
3366 snmp6_unregister_dev(idev);
3367 addrconf_sysctl_unregister(idev);
3368 err = addrconf_sysctl_register(idev);
3370 return notifier_from_errno(err);
3371 err = snmp6_register_dev(idev);
3373 addrconf_sysctl_unregister(idev);
3374 return notifier_from_errno(err);
3379 case NETDEV_PRE_TYPE_CHANGE:
3380 case NETDEV_POST_TYPE_CHANGE:
3382 addrconf_type_change(dev, event);
3390 * addrconf module should be notified of a device going up
3392 static struct notifier_block ipv6_dev_notf = {
3393 .notifier_call = addrconf_notify,
3396 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3398 struct inet6_dev *idev;
3401 idev = __in6_dev_get(dev);
3403 if (event == NETDEV_POST_TYPE_CHANGE)
3404 ipv6_mc_remap(idev);
3405 else if (event == NETDEV_PRE_TYPE_CHANGE)
3406 ipv6_mc_unmap(idev);
3409 static int addrconf_ifdown(struct net_device *dev, int how)
3411 struct net *net = dev_net(dev);
3412 struct inet6_dev *idev;
3413 struct inet6_ifaddr *ifa, *tmp;
3414 struct list_head del_list;
3421 rt6_ifdown(net, dev);
3422 neigh_ifdown(&nd_tbl, dev);
3424 idev = __in6_dev_get(dev);
3429 * Step 1: remove reference to ipv6 device from parent device.
3435 /* protected by rtnl_lock */
3436 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3438 /* Step 1.5: remove snmp6 entry */
3439 snmp6_unregister_dev(idev);
3443 /* aggregate the system setting and interface setting */
3444 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3446 _keep_addr = idev->cnf.keep_addr_on_down;
3448 /* combine the user config with event to determine if permanent
3449 * addresses are to be removed from address hash table
3451 keep_addr = !(how || _keep_addr <= 0);
3453 /* Step 2: clear hash table */
3454 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3455 struct hlist_head *h = &inet6_addr_lst[i];
3457 spin_lock_bh(&addrconf_hash_lock);
3459 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3460 if (ifa->idev == idev) {
3461 addrconf_del_dad_work(ifa);
3462 /* combined flag + permanent flag decide if
3463 * address is retained on a down event
3466 !(ifa->flags & IFA_F_PERMANENT)) {
3467 hlist_del_init_rcu(&ifa->addr_lst);
3472 spin_unlock_bh(&addrconf_hash_lock);
3475 write_lock_bh(&idev->lock);
3477 addrconf_del_rs_timer(idev);
3479 /* Step 2: clear flags for stateless addrconf */
3481 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3483 if (how && del_timer(&idev->regen_timer))
3486 /* Step 3: clear tempaddr list */
3487 while (!list_empty(&idev->tempaddr_list)) {
3488 ifa = list_first_entry(&idev->tempaddr_list,
3489 struct inet6_ifaddr, tmp_list);
3490 list_del(&ifa->tmp_list);
3491 write_unlock_bh(&idev->lock);
3492 spin_lock_bh(&ifa->lock);
3495 in6_ifa_put(ifa->ifpub);
3498 spin_unlock_bh(&ifa->lock);
3500 write_lock_bh(&idev->lock);
3503 /* re-combine the user config with event to determine if permanent
3504 * addresses are to be removed from the interface list
3506 keep_addr = (!how && _keep_addr > 0);
3508 INIT_LIST_HEAD(&del_list);
3509 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3510 addrconf_del_dad_work(ifa);
3512 write_unlock_bh(&idev->lock);
3513 spin_lock_bh(&ifa->lock);
3515 if (keep_addr && (ifa->flags & IFA_F_PERMANENT)) {
3516 /* set state to skip the notifier below */
3517 state = INET6_IFADDR_STATE_DEAD;
3519 if (!(ifa->flags & IFA_F_NODAD))
3520 ifa->flags |= IFA_F_TENTATIVE;
3523 ifa->state = INET6_IFADDR_STATE_DEAD;
3525 list_del(&ifa->if_list);
3526 list_add(&ifa->if_list, &del_list);
3529 spin_unlock_bh(&ifa->lock);
3531 if (state != INET6_IFADDR_STATE_DEAD) {
3532 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3533 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3536 write_lock_bh(&idev->lock);
3539 write_unlock_bh(&idev->lock);
3541 /* now clean up addresses to be removed */
3542 while (!list_empty(&del_list)) {
3543 ifa = list_first_entry(&del_list,
3544 struct inet6_ifaddr, if_list);
3545 list_del(&ifa->if_list);
3550 /* Step 5: Discard anycast and multicast list */
3552 ipv6_ac_destroy_dev(idev);
3553 ipv6_mc_destroy_dev(idev);
3558 idev->tstamp = jiffies;
3560 /* Last: Shot the device (if unregistered) */
3562 addrconf_sysctl_unregister(idev);
3563 neigh_parms_release(&nd_tbl, idev->nd_parms);
3564 neigh_ifdown(&nd_tbl, dev);
3570 static void addrconf_rs_timer(unsigned long data)
3572 struct inet6_dev *idev = (struct inet6_dev *)data;
3573 struct net_device *dev = idev->dev;
3574 struct in6_addr lladdr;
3576 write_lock(&idev->lock);
3577 if (idev->dead || !(idev->if_flags & IF_READY))
3580 if (!ipv6_accept_ra(idev))
3583 /* Announcement received after solicitation was sent */
3584 if (idev->if_flags & IF_RA_RCVD)
3587 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3588 write_unlock(&idev->lock);
3589 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3590 ndisc_send_rs(dev, &lladdr,
3591 &in6addr_linklocal_allrouters);
3595 write_lock(&idev->lock);
3596 /* The wait after the last probe can be shorter */
3597 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3598 idev->cnf.rtr_solicits) ?
3599 idev->cnf.rtr_solicit_delay :
3600 idev->cnf.rtr_solicit_interval);
3603 * Note: we do not support deprecated "all on-link"
3604 * assumption any longer.
3606 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3610 write_unlock(&idev->lock);
3616 * Duplicate Address Detection
3618 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3620 unsigned long rand_num;
3621 struct inet6_dev *idev = ifp->idev;
3623 if (ifp->flags & IFA_F_OPTIMISTIC)
3626 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3628 ifp->dad_probes = idev->cnf.dad_transmits;
3629 addrconf_mod_dad_work(ifp, rand_num);
3632 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3634 struct inet6_dev *idev = ifp->idev;
3635 struct net_device *dev = idev->dev;
3636 bool notify = false;
3638 addrconf_join_solict(dev, &ifp->addr);
3640 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3642 read_lock_bh(&idev->lock);
3643 spin_lock(&ifp->lock);
3644 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3647 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3648 idev->cnf.accept_dad < 1 ||
3649 !(ifp->flags&IFA_F_TENTATIVE) ||
3650 ifp->flags & IFA_F_NODAD) {
3651 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3652 spin_unlock(&ifp->lock);
3653 read_unlock_bh(&idev->lock);
3655 addrconf_dad_completed(ifp);
3659 if (!(idev->if_flags & IF_READY)) {
3660 spin_unlock(&ifp->lock);
3661 read_unlock_bh(&idev->lock);
3663 * If the device is not ready:
3664 * - keep it tentative if it is a permanent address.
3665 * - otherwise, kill it.
3668 addrconf_dad_stop(ifp, 0);
3673 * Optimistic nodes can start receiving
3676 if (ifp->flags & IFA_F_OPTIMISTIC) {
3677 ip6_ins_rt(ifp->rt);
3678 if (ipv6_use_optimistic_addr(idev)) {
3679 /* Because optimistic nodes can use this address,
3680 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3686 addrconf_dad_kick(ifp);
3688 spin_unlock(&ifp->lock);
3689 read_unlock_bh(&idev->lock);
3691 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3694 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3696 bool begin_dad = false;
3698 spin_lock_bh(&ifp->lock);
3699 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3700 ifp->state = INET6_IFADDR_STATE_PREDAD;
3703 spin_unlock_bh(&ifp->lock);
3706 addrconf_mod_dad_work(ifp, 0);
3709 static void addrconf_dad_work(struct work_struct *w)
3711 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3712 struct inet6_ifaddr,
3714 struct inet6_dev *idev = ifp->idev;
3715 struct in6_addr mcaddr;
3721 } action = DAD_PROCESS;
3725 spin_lock_bh(&ifp->lock);
3726 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3728 ifp->state = INET6_IFADDR_STATE_DAD;
3729 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3731 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3733 spin_unlock_bh(&ifp->lock);
3735 if (action == DAD_BEGIN) {
3736 addrconf_dad_begin(ifp);
3738 } else if (action == DAD_ABORT) {
3739 addrconf_dad_stop(ifp, 1);
3743 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3746 write_lock_bh(&idev->lock);
3747 if (idev->dead || !(idev->if_flags & IF_READY)) {
3748 write_unlock_bh(&idev->lock);
3752 spin_lock(&ifp->lock);
3753 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3754 spin_unlock(&ifp->lock);
3755 write_unlock_bh(&idev->lock);
3759 if (ifp->dad_probes == 0) {
3761 * DAD was successful
3764 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3765 spin_unlock(&ifp->lock);
3766 write_unlock_bh(&idev->lock);
3768 addrconf_dad_completed(ifp);
3774 addrconf_mod_dad_work(ifp,
3775 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3776 spin_unlock(&ifp->lock);
3777 write_unlock_bh(&idev->lock);
3779 /* send a neighbour solicitation for our addr */
3780 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3781 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3787 /* ifp->idev must be at least read locked */
3788 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3790 struct inet6_ifaddr *ifpiter;
3791 struct inet6_dev *idev = ifp->idev;
3793 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3794 if (ifpiter->scope > IFA_LINK)
3796 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3797 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3798 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3805 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3807 struct net_device *dev = ifp->idev->dev;
3808 struct in6_addr lladdr;
3809 bool send_rs, send_mld;
3811 addrconf_del_dad_work(ifp);
3814 * Configure the address for reception. Now it is valid.
3817 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3819 /* If added prefix is link local and we are prepared to process
3820 router advertisements, start sending router solicitations.
3823 read_lock_bh(&ifp->idev->lock);
3824 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3825 send_rs = send_mld &&
3826 ipv6_accept_ra(ifp->idev) &&
3827 ifp->idev->cnf.rtr_solicits > 0 &&
3828 (dev->flags&IFF_LOOPBACK) == 0;
3829 read_unlock_bh(&ifp->idev->lock);
3831 /* While dad is in progress mld report's source address is in6_addrany.
3832 * Resend with proper ll now.
3835 ipv6_mc_dad_complete(ifp->idev);
3839 * If a host as already performed a random delay
3840 * [...] as part of DAD [...] there is no need
3841 * to delay again before sending the first RS
3843 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3845 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3847 write_lock_bh(&ifp->idev->lock);
3848 spin_lock(&ifp->lock);
3849 ifp->idev->rs_probes = 1;
3850 ifp->idev->if_flags |= IF_RS_SENT;
3851 addrconf_mod_rs_timer(ifp->idev,
3852 ifp->idev->cnf.rtr_solicit_interval);
3853 spin_unlock(&ifp->lock);
3854 write_unlock_bh(&ifp->idev->lock);
3858 static void addrconf_dad_run(struct inet6_dev *idev)
3860 struct inet6_ifaddr *ifp;
3862 read_lock_bh(&idev->lock);
3863 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3864 spin_lock(&ifp->lock);
3865 if (ifp->flags & IFA_F_TENTATIVE &&
3866 ifp->state == INET6_IFADDR_STATE_DAD)
3867 addrconf_dad_kick(ifp);
3868 spin_unlock(&ifp->lock);
3870 read_unlock_bh(&idev->lock);
3873 #ifdef CONFIG_PROC_FS
3874 struct if6_iter_state {
3875 struct seq_net_private p;
3880 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3882 struct inet6_ifaddr *ifa = NULL;
3883 struct if6_iter_state *state = seq->private;
3884 struct net *net = seq_file_net(seq);
3887 /* initial bucket if pos is 0 */
3893 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3894 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3896 if (!net_eq(dev_net(ifa->idev->dev), net))
3898 /* sync with offset */
3899 if (p < state->offset) {
3907 /* prepare for next bucket */
3914 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3915 struct inet6_ifaddr *ifa)
3917 struct if6_iter_state *state = seq->private;
3918 struct net *net = seq_file_net(seq);
3920 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3921 if (!net_eq(dev_net(ifa->idev->dev), net))
3927 while (++state->bucket < IN6_ADDR_HSIZE) {
3929 hlist_for_each_entry_rcu_bh(ifa,
3930 &inet6_addr_lst[state->bucket], addr_lst) {
3931 if (!net_eq(dev_net(ifa->idev->dev), net))
3941 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3945 return if6_get_first(seq, *pos);
3948 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3950 struct inet6_ifaddr *ifa;
3952 ifa = if6_get_next(seq, v);
3957 static void if6_seq_stop(struct seq_file *seq, void *v)
3960 rcu_read_unlock_bh();
3963 static int if6_seq_show(struct seq_file *seq, void *v)
3965 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3966 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3968 ifp->idev->dev->ifindex,
3972 ifp->idev->dev->name);
3976 static const struct seq_operations if6_seq_ops = {
3977 .start = if6_seq_start,
3978 .next = if6_seq_next,
3979 .show = if6_seq_show,
3980 .stop = if6_seq_stop,
3983 static int if6_seq_open(struct inode *inode, struct file *file)
3985 return seq_open_net(inode, file, &if6_seq_ops,
3986 sizeof(struct if6_iter_state));
3989 static const struct file_operations if6_fops = {
3990 .owner = THIS_MODULE,
3991 .open = if6_seq_open,
3993 .llseek = seq_lseek,
3994 .release = seq_release_net,
3997 static int __net_init if6_proc_net_init(struct net *net)
3999 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4004 static void __net_exit if6_proc_net_exit(struct net *net)
4006 remove_proc_entry("if_inet6", net->proc_net);
4009 static struct pernet_operations if6_proc_net_ops = {
4010 .init = if6_proc_net_init,
4011 .exit = if6_proc_net_exit,
4014 int __init if6_proc_init(void)
4016 return register_pernet_subsys(&if6_proc_net_ops);
4019 void if6_proc_exit(void)
4021 unregister_pernet_subsys(&if6_proc_net_ops);
4023 #endif /* CONFIG_PROC_FS */
4025 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4026 /* Check if address is a home address configured on any interface. */
4027 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4030 struct inet6_ifaddr *ifp = NULL;
4031 unsigned int hash = inet6_addr_hash(addr);
4034 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4035 if (!net_eq(dev_net(ifp->idev->dev), net))
4037 if (ipv6_addr_equal(&ifp->addr, addr) &&
4038 (ifp->flags & IFA_F_HOMEADDRESS)) {
4043 rcu_read_unlock_bh();
4049 * Periodic address status verification
4052 static void addrconf_verify_rtnl(void)
4054 unsigned long now, next, next_sec, next_sched;
4055 struct inet6_ifaddr *ifp;
4062 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4064 cancel_delayed_work(&addr_chk_work);
4066 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4068 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4071 /* When setting preferred_lft to a value not zero or
4072 * infinity, while valid_lft is infinity
4073 * IFA_F_PERMANENT has a non-infinity life time.
4075 if ((ifp->flags & IFA_F_PERMANENT) &&
4076 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4079 spin_lock(&ifp->lock);
4080 /* We try to batch several events at once. */
4081 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4083 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4084 age >= ifp->valid_lft) {
4085 spin_unlock(&ifp->lock);
4089 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4090 spin_unlock(&ifp->lock);
4092 } else if (age >= ifp->prefered_lft) {
4093 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4096 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4098 ifp->flags |= IFA_F_DEPRECATED;
4101 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4102 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4103 next = ifp->tstamp + ifp->valid_lft * HZ;
4105 spin_unlock(&ifp->lock);
4110 ipv6_ifa_notify(0, ifp);
4114 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4115 !(ifp->flags&IFA_F_TENTATIVE)) {
4116 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4117 ifp->idev->cnf.dad_transmits *
4118 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4120 if (age >= ifp->prefered_lft - regen_advance) {
4121 struct inet6_ifaddr *ifpub = ifp->ifpub;
4122 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4123 next = ifp->tstamp + ifp->prefered_lft * HZ;
4124 if (!ifp->regen_count && ifpub) {
4127 in6_ifa_hold(ifpub);
4128 spin_unlock(&ifp->lock);
4130 spin_lock(&ifpub->lock);
4131 ifpub->regen_count = 0;
4132 spin_unlock(&ifpub->lock);
4133 ipv6_create_tempaddr(ifpub, ifp);
4138 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4139 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4140 spin_unlock(&ifp->lock);
4142 /* ifp->prefered_lft <= ifp->valid_lft */
4143 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4144 next = ifp->tstamp + ifp->prefered_lft * HZ;
4145 spin_unlock(&ifp->lock);
4150 next_sec = round_jiffies_up(next);
4153 /* If rounded timeout is accurate enough, accept it. */
4154 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4155 next_sched = next_sec;
4157 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4158 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4159 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4161 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4162 now, next, next_sec, next_sched);
4163 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4164 rcu_read_unlock_bh();
4167 static void addrconf_verify_work(struct work_struct *w)
4170 addrconf_verify_rtnl();
4174 static void addrconf_verify(void)
4176 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4179 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4180 struct in6_addr **peer_pfx)
4182 struct in6_addr *pfx = NULL;
4187 pfx = nla_data(addr);
4190 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4192 pfx = nla_data(local);
4198 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4199 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4200 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4201 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4202 [IFA_FLAGS] = { .len = sizeof(u32) },
4206 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4208 struct net *net = sock_net(skb->sk);
4209 struct ifaddrmsg *ifm;
4210 struct nlattr *tb[IFA_MAX+1];
4211 struct in6_addr *pfx, *peer_pfx;
4215 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4219 ifm = nlmsg_data(nlh);
4220 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4224 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4226 /* We ignore other flags so far. */
4227 ifa_flags &= IFA_F_MANAGETEMPADDR;
4229 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4230 ifm->ifa_prefixlen);
4233 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4234 u32 prefered_lft, u32 valid_lft)
4238 unsigned long timeout;
4239 bool was_managetempaddr;
4240 bool had_prefixroute;
4244 if (!valid_lft || (prefered_lft > valid_lft))
4247 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4248 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4251 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4252 if (addrconf_finite_timeout(timeout)) {
4253 expires = jiffies_to_clock_t(timeout * HZ);
4254 valid_lft = timeout;
4255 flags = RTF_EXPIRES;
4259 ifa_flags |= IFA_F_PERMANENT;
4262 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4263 if (addrconf_finite_timeout(timeout)) {
4265 ifa_flags |= IFA_F_DEPRECATED;
4266 prefered_lft = timeout;
4269 spin_lock_bh(&ifp->lock);
4270 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4271 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4272 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4273 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4274 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4275 IFA_F_NOPREFIXROUTE);
4276 ifp->flags |= ifa_flags;
4277 ifp->tstamp = jiffies;
4278 ifp->valid_lft = valid_lft;
4279 ifp->prefered_lft = prefered_lft;
4281 spin_unlock_bh(&ifp->lock);
4282 if (!(ifp->flags&IFA_F_TENTATIVE))
4283 ipv6_ifa_notify(0, ifp);
4285 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4286 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4288 } else if (had_prefixroute) {
4289 enum cleanup_prefix_rt_t action;
4290 unsigned long rt_expires;
4292 write_lock_bh(&ifp->idev->lock);
4293 action = check_cleanup_prefix_route(ifp, &rt_expires);
4294 write_unlock_bh(&ifp->idev->lock);
4296 if (action != CLEANUP_PREFIX_RT_NOP) {
4297 cleanup_prefix_route(ifp, rt_expires,
4298 action == CLEANUP_PREFIX_RT_DEL);
4302 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4303 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4304 valid_lft = prefered_lft = 0;
4305 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4306 !was_managetempaddr, jiffies);
4309 addrconf_verify_rtnl();
4315 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4317 struct net *net = sock_net(skb->sk);
4318 struct ifaddrmsg *ifm;
4319 struct nlattr *tb[IFA_MAX+1];
4320 struct in6_addr *pfx, *peer_pfx;
4321 struct inet6_ifaddr *ifa;
4322 struct net_device *dev;
4323 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4327 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4331 ifm = nlmsg_data(nlh);
4332 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4336 if (tb[IFA_CACHEINFO]) {
4337 struct ifa_cacheinfo *ci;
4339 ci = nla_data(tb[IFA_CACHEINFO]);
4340 valid_lft = ci->ifa_valid;
4341 preferred_lft = ci->ifa_prefered;
4343 preferred_lft = INFINITY_LIFE_TIME;
4344 valid_lft = INFINITY_LIFE_TIME;
4347 dev = __dev_get_by_index(net, ifm->ifa_index);
4351 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4353 /* We ignore other flags so far. */
4354 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4355 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4357 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4360 * It would be best to check for !NLM_F_CREATE here but
4361 * userspace already relies on not having to provide this.
4363 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4364 ifm->ifa_prefixlen, ifa_flags,
4365 preferred_lft, valid_lft);
4368 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4369 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4372 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4379 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4380 u8 scope, int ifindex)
4382 struct ifaddrmsg *ifm;
4384 ifm = nlmsg_data(nlh);
4385 ifm->ifa_family = AF_INET6;
4386 ifm->ifa_prefixlen = prefixlen;
4387 ifm->ifa_flags = flags;
4388 ifm->ifa_scope = scope;
4389 ifm->ifa_index = ifindex;
4392 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4393 unsigned long tstamp, u32 preferred, u32 valid)
4395 struct ifa_cacheinfo ci;
4397 ci.cstamp = cstamp_delta(cstamp);
4398 ci.tstamp = cstamp_delta(tstamp);
4399 ci.ifa_prefered = preferred;
4400 ci.ifa_valid = valid;
4402 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4405 static inline int rt_scope(int ifa_scope)
4407 if (ifa_scope & IFA_HOST)
4408 return RT_SCOPE_HOST;
4409 else if (ifa_scope & IFA_LINK)
4410 return RT_SCOPE_LINK;
4411 else if (ifa_scope & IFA_SITE)
4412 return RT_SCOPE_SITE;
4414 return RT_SCOPE_UNIVERSE;
4417 static inline int inet6_ifaddr_msgsize(void)
4419 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4420 + nla_total_size(16) /* IFA_LOCAL */
4421 + nla_total_size(16) /* IFA_ADDRESS */
4422 + nla_total_size(sizeof(struct ifa_cacheinfo))
4423 + nla_total_size(4) /* IFA_FLAGS */;
4426 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4427 u32 portid, u32 seq, int event, unsigned int flags)
4429 struct nlmsghdr *nlh;
4430 u32 preferred, valid;
4432 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4436 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4437 ifa->idev->dev->ifindex);
4439 if (!((ifa->flags&IFA_F_PERMANENT) &&
4440 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4441 preferred = ifa->prefered_lft;
4442 valid = ifa->valid_lft;
4443 if (preferred != INFINITY_LIFE_TIME) {
4444 long tval = (jiffies - ifa->tstamp)/HZ;
4445 if (preferred > tval)
4449 if (valid != INFINITY_LIFE_TIME) {
4457 preferred = INFINITY_LIFE_TIME;
4458 valid = INFINITY_LIFE_TIME;
4461 if (!ipv6_addr_any(&ifa->peer_addr)) {
4462 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4463 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4466 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4469 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4472 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4475 nlmsg_end(skb, nlh);
4479 nlmsg_cancel(skb, nlh);
4483 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4484 u32 portid, u32 seq, int event, u16 flags)
4486 struct nlmsghdr *nlh;
4487 u8 scope = RT_SCOPE_UNIVERSE;
4488 int ifindex = ifmca->idev->dev->ifindex;
4490 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4491 scope = RT_SCOPE_SITE;
4493 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4497 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4498 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4499 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4500 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4501 nlmsg_cancel(skb, nlh);
4505 nlmsg_end(skb, nlh);
4509 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4510 u32 portid, u32 seq, int event, unsigned int flags)
4512 struct nlmsghdr *nlh;
4513 u8 scope = RT_SCOPE_UNIVERSE;
4514 int ifindex = ifaca->aca_idev->dev->ifindex;
4516 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4517 scope = RT_SCOPE_SITE;
4519 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4523 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4524 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4525 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4526 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4527 nlmsg_cancel(skb, nlh);
4531 nlmsg_end(skb, nlh);
4541 /* called with rcu_read_lock() */
4542 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4543 struct netlink_callback *cb, enum addr_type_t type,
4544 int s_ip_idx, int *p_ip_idx)
4546 struct ifmcaddr6 *ifmca;
4547 struct ifacaddr6 *ifaca;
4549 int ip_idx = *p_ip_idx;
4551 read_lock_bh(&idev->lock);
4553 case UNICAST_ADDR: {
4554 struct inet6_ifaddr *ifa;
4556 /* unicast address incl. temp addr */
4557 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4558 if (++ip_idx < s_ip_idx)
4560 err = inet6_fill_ifaddr(skb, ifa,
4561 NETLINK_CB(cb->skb).portid,
4567 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4571 case MULTICAST_ADDR:
4572 /* multicast address */
4573 for (ifmca = idev->mc_list; ifmca;
4574 ifmca = ifmca->next, ip_idx++) {
4575 if (ip_idx < s_ip_idx)
4577 err = inet6_fill_ifmcaddr(skb, ifmca,
4578 NETLINK_CB(cb->skb).portid,
4587 /* anycast address */
4588 for (ifaca = idev->ac_list; ifaca;
4589 ifaca = ifaca->aca_next, ip_idx++) {
4590 if (ip_idx < s_ip_idx)
4592 err = inet6_fill_ifacaddr(skb, ifaca,
4593 NETLINK_CB(cb->skb).portid,
4604 read_unlock_bh(&idev->lock);
4609 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4610 enum addr_type_t type)
4612 struct net *net = sock_net(skb->sk);
4615 int s_idx, s_ip_idx;
4616 struct net_device *dev;
4617 struct inet6_dev *idev;
4618 struct hlist_head *head;
4621 s_idx = idx = cb->args[1];
4622 s_ip_idx = ip_idx = cb->args[2];
4625 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4626 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4628 head = &net->dev_index_head[h];
4629 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4632 if (h > s_h || idx > s_idx)
4635 idev = __in6_dev_get(dev);
4639 if (in6_dump_addrs(idev, skb, cb, type,
4640 s_ip_idx, &ip_idx) < 0)
4650 cb->args[2] = ip_idx;
4655 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4657 enum addr_type_t type = UNICAST_ADDR;
4659 return inet6_dump_addr(skb, cb, type);
4662 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4664 enum addr_type_t type = MULTICAST_ADDR;
4666 return inet6_dump_addr(skb, cb, type);
4670 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4672 enum addr_type_t type = ANYCAST_ADDR;
4674 return inet6_dump_addr(skb, cb, type);
4677 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4679 struct net *net = sock_net(in_skb->sk);
4680 struct ifaddrmsg *ifm;
4681 struct nlattr *tb[IFA_MAX+1];
4682 struct in6_addr *addr = NULL, *peer;
4683 struct net_device *dev = NULL;
4684 struct inet6_ifaddr *ifa;
4685 struct sk_buff *skb;
4688 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4692 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4698 ifm = nlmsg_data(nlh);
4700 dev = __dev_get_by_index(net, ifm->ifa_index);
4702 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4704 err = -EADDRNOTAVAIL;
4708 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4714 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4715 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4717 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4718 WARN_ON(err == -EMSGSIZE);
4722 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4729 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4731 struct sk_buff *skb;
4732 struct net *net = dev_net(ifa->idev->dev);
4735 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4739 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4741 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4742 WARN_ON(err == -EMSGSIZE);
4746 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4750 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4753 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4754 __s32 *array, int bytes)
4756 BUG_ON(bytes < (DEVCONF_MAX * 4));
4758 memset(array, 0, bytes);
4759 array[DEVCONF_FORWARDING] = cnf->forwarding;
4760 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4761 array[DEVCONF_MTU6] = cnf->mtu6;
4762 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4763 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4764 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4765 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4766 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4767 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4768 jiffies_to_msecs(cnf->rtr_solicit_interval);
4769 array[DEVCONF_RTR_SOLICIT_DELAY] =
4770 jiffies_to_msecs(cnf->rtr_solicit_delay);
4771 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4772 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4773 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4774 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4775 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4776 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4777 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4778 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4779 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4780 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4781 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4782 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4783 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4784 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4785 #ifdef CONFIG_IPV6_ROUTER_PREF
4786 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4787 array[DEVCONF_RTR_PROBE_INTERVAL] =
4788 jiffies_to_msecs(cnf->rtr_probe_interval);
4789 #ifdef CONFIG_IPV6_ROUTE_INFO
4790 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4793 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4794 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4795 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4796 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4797 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4799 #ifdef CONFIG_IPV6_MROUTE
4800 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4802 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4803 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4804 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4805 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4806 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4807 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4808 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4809 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4810 /* we omit DEVCONF_STABLE_SECRET for now */
4811 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4812 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4813 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4814 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4817 static inline size_t inet6_ifla6_size(void)
4819 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4820 + nla_total_size(sizeof(struct ifla_cacheinfo))
4821 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4822 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4823 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4824 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4827 static inline size_t inet6_if_nlmsg_size(void)
4829 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4830 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4831 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4832 + nla_total_size(4) /* IFLA_MTU */
4833 + nla_total_size(4) /* IFLA_LINK */
4834 + nla_total_size(1) /* IFLA_OPERSTATE */
4835 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4838 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4839 int items, int bytes)
4842 int pad = bytes - sizeof(u64) * items;
4845 /* Use put_unaligned() because stats may not be aligned for u64. */
4846 put_unaligned(items, &stats[0]);
4847 for (i = 1; i < items; i++)
4848 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4850 memset(&stats[items], 0, pad);
4853 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4854 int bytes, size_t syncpoff)
4857 u64 buff[IPSTATS_MIB_MAX];
4858 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4862 memset(buff, 0, sizeof(buff));
4863 buff[0] = IPSTATS_MIB_MAX;
4865 for_each_possible_cpu(c) {
4866 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4867 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4870 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4871 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4874 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4878 case IFLA_INET6_STATS:
4879 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4880 offsetof(struct ipstats_mib, syncp));
4882 case IFLA_INET6_ICMP6STATS:
4883 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4888 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4889 u32 ext_filter_mask)
4892 struct ifla_cacheinfo ci;
4894 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4895 goto nla_put_failure;
4896 ci.max_reasm_len = IPV6_MAXPLEN;
4897 ci.tstamp = cstamp_delta(idev->tstamp);
4898 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4899 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4900 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4901 goto nla_put_failure;
4902 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4904 goto nla_put_failure;
4905 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4907 /* XXX - MC not implemented */
4909 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4912 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4914 goto nla_put_failure;
4915 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4917 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4919 goto nla_put_failure;
4920 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4922 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4924 goto nla_put_failure;
4926 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4927 goto nla_put_failure;
4929 read_lock_bh(&idev->lock);
4930 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4931 read_unlock_bh(&idev->lock);
4939 static size_t inet6_get_link_af_size(const struct net_device *dev,
4940 u32 ext_filter_mask)
4942 if (!__in6_dev_get(dev))
4945 return inet6_ifla6_size();
4948 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4949 u32 ext_filter_mask)
4951 struct inet6_dev *idev = __in6_dev_get(dev);
4956 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4962 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4964 struct inet6_ifaddr *ifp;
4965 struct net_device *dev = idev->dev;
4966 bool update_rs = false;
4967 struct in6_addr ll_addr;
4973 if (ipv6_addr_any(token))
4975 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4977 if (!ipv6_accept_ra(idev))
4979 if (idev->cnf.rtr_solicits <= 0)
4982 write_lock_bh(&idev->lock);
4984 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4985 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4987 write_unlock_bh(&idev->lock);
4989 if (!idev->dead && (idev->if_flags & IF_READY) &&
4990 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4991 IFA_F_OPTIMISTIC)) {
4993 /* If we're not ready, then normal ifup will take care
4994 * of this. Otherwise, we need to request our rs here.
4996 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5000 write_lock_bh(&idev->lock);
5003 idev->if_flags |= IF_RS_SENT;
5004 idev->rs_probes = 1;
5005 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
5008 /* Well, that's kinda nasty ... */
5009 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5010 spin_lock(&ifp->lock);
5011 if (ifp->tokenized) {
5013 ifp->prefered_lft = 0;
5015 spin_unlock(&ifp->lock);
5018 write_unlock_bh(&idev->lock);
5019 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5020 addrconf_verify_rtnl();
5024 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5025 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5026 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5029 static int inet6_validate_link_af(const struct net_device *dev,
5030 const struct nlattr *nla)
5032 struct nlattr *tb[IFLA_INET6_MAX + 1];
5034 if (dev && !__in6_dev_get(dev))
5035 return -EAFNOSUPPORT;
5037 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5040 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5043 struct inet6_dev *idev = __in6_dev_get(dev);
5044 struct nlattr *tb[IFLA_INET6_MAX + 1];
5047 return -EAFNOSUPPORT;
5049 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5052 if (tb[IFLA_INET6_TOKEN]) {
5053 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5058 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5059 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5061 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5062 mode != IN6_ADDR_GEN_MODE_NONE &&
5063 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5064 mode != IN6_ADDR_GEN_MODE_RANDOM)
5067 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5068 !idev->cnf.stable_secret.initialized &&
5069 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5072 idev->addr_gen_mode = mode;
5079 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5080 u32 portid, u32 seq, int event, unsigned int flags)
5082 struct net_device *dev = idev->dev;
5083 struct ifinfomsg *hdr;
5084 struct nlmsghdr *nlh;
5087 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5091 hdr = nlmsg_data(nlh);
5092 hdr->ifi_family = AF_INET6;
5094 hdr->ifi_type = dev->type;
5095 hdr->ifi_index = dev->ifindex;
5096 hdr->ifi_flags = dev_get_flags(dev);
5097 hdr->ifi_change = 0;
5099 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5101 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5102 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5103 (dev->ifindex != dev_get_iflink(dev) &&
5104 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5105 nla_put_u8(skb, IFLA_OPERSTATE,
5106 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5107 goto nla_put_failure;
5108 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5110 goto nla_put_failure;
5112 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5113 goto nla_put_failure;
5115 nla_nest_end(skb, protoinfo);
5116 nlmsg_end(skb, nlh);
5120 nlmsg_cancel(skb, nlh);
5124 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5126 struct net *net = sock_net(skb->sk);
5129 struct net_device *dev;
5130 struct inet6_dev *idev;
5131 struct hlist_head *head;
5134 s_idx = cb->args[1];
5137 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5139 head = &net->dev_index_head[h];
5140 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5143 idev = __in6_dev_get(dev);
5146 if (inet6_fill_ifinfo(skb, idev,
5147 NETLINK_CB(cb->skb).portid,
5149 RTM_NEWLINK, NLM_F_MULTI) < 0)
5163 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5165 struct sk_buff *skb;
5166 struct net *net = dev_net(idev->dev);
5169 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5173 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5175 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5176 WARN_ON(err == -EMSGSIZE);
5180 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5184 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5187 static inline size_t inet6_prefix_nlmsg_size(void)
5189 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5190 + nla_total_size(sizeof(struct in6_addr))
5191 + nla_total_size(sizeof(struct prefix_cacheinfo));
5194 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5195 struct prefix_info *pinfo, u32 portid, u32 seq,
5196 int event, unsigned int flags)
5198 struct prefixmsg *pmsg;
5199 struct nlmsghdr *nlh;
5200 struct prefix_cacheinfo ci;
5202 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5206 pmsg = nlmsg_data(nlh);
5207 pmsg->prefix_family = AF_INET6;
5208 pmsg->prefix_pad1 = 0;
5209 pmsg->prefix_pad2 = 0;
5210 pmsg->prefix_ifindex = idev->dev->ifindex;
5211 pmsg->prefix_len = pinfo->prefix_len;
5212 pmsg->prefix_type = pinfo->type;
5213 pmsg->prefix_pad3 = 0;
5214 pmsg->prefix_flags = 0;
5216 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5217 if (pinfo->autoconf)
5218 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5220 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5221 goto nla_put_failure;
5222 ci.preferred_time = ntohl(pinfo->prefered);
5223 ci.valid_time = ntohl(pinfo->valid);
5224 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5225 goto nla_put_failure;
5226 nlmsg_end(skb, nlh);
5230 nlmsg_cancel(skb, nlh);
5234 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5235 struct prefix_info *pinfo)
5237 struct sk_buff *skb;
5238 struct net *net = dev_net(idev->dev);
5241 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5245 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5247 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5248 WARN_ON(err == -EMSGSIZE);
5252 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5256 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5259 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5261 struct net *net = dev_net(ifp->idev->dev);
5266 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5271 * If the address was optimistic
5272 * we inserted the route at the start of
5273 * our DAD process, so we don't need
5276 if (!(ifp->rt->rt6i_node))
5277 ip6_ins_rt(ifp->rt);
5278 if (ifp->idev->cnf.forwarding)
5279 addrconf_join_anycast(ifp);
5280 if (!ipv6_addr_any(&ifp->peer_addr))
5281 addrconf_prefix_route(&ifp->peer_addr, 128,
5282 ifp->idev->dev, 0, 0);
5285 if (ifp->idev->cnf.forwarding)
5286 addrconf_leave_anycast(ifp);
5287 addrconf_leave_solict(ifp->idev, &ifp->addr);
5288 if (!ipv6_addr_any(&ifp->peer_addr)) {
5289 struct rt6_info *rt;
5291 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5292 ifp->idev->dev, 0, 0);
5297 dst_hold(&ifp->rt->dst);
5299 ip6_del_rt(ifp->rt);
5302 rt_genid_bump_ipv6(net);
5305 atomic_inc(&net->ipv6.dev_addr_genid);
5308 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5311 if (likely(ifp->idev->dead == 0))
5312 __ipv6_ifa_notify(event, ifp);
5313 rcu_read_unlock_bh();
5316 #ifdef CONFIG_SYSCTL
5319 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5320 void __user *buffer, size_t *lenp, loff_t *ppos)
5322 int *valp = ctl->data;
5325 struct ctl_table lctl;
5329 * ctl->data points to idev->cnf.forwarding, we should
5330 * not modify it until we get the rtnl lock.
5335 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5338 ret = addrconf_fixup_forwarding(ctl, valp, val);
5345 int addrconf_sysctl_hop_limit(struct ctl_table *ctl, int write,
5346 void __user *buffer, size_t *lenp, loff_t *ppos)
5348 struct ctl_table lctl;
5349 int min_hl = 1, max_hl = 255;
5352 lctl.extra1 = &min_hl;
5353 lctl.extra2 = &max_hl;
5355 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5359 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5360 void __user *buffer, size_t *lenp, loff_t *ppos)
5362 struct inet6_dev *idev = ctl->extra1;
5363 int min_mtu = IPV6_MIN_MTU;
5364 struct ctl_table lctl;
5367 lctl.extra1 = &min_mtu;
5368 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5370 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5373 static void dev_disable_change(struct inet6_dev *idev)
5375 struct netdev_notifier_info info;
5377 if (!idev || !idev->dev)
5380 netdev_notifier_info_init(&info, idev->dev);
5381 if (idev->cnf.disable_ipv6)
5382 addrconf_notify(NULL, NETDEV_DOWN, &info);
5384 addrconf_notify(NULL, NETDEV_UP, &info);
5387 static void addrconf_disable_change(struct net *net, __s32 newf)
5389 struct net_device *dev;
5390 struct inet6_dev *idev;
5393 for_each_netdev_rcu(net, dev) {
5394 idev = __in6_dev_get(dev);
5396 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5397 idev->cnf.disable_ipv6 = newf;
5399 dev_disable_change(idev);
5405 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5410 if (!rtnl_trylock())
5411 return restart_syscall();
5413 net = (struct net *)table->extra2;
5417 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5422 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5423 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5424 addrconf_disable_change(net, newf);
5425 } else if ((!newf) ^ (!old))
5426 dev_disable_change((struct inet6_dev *)table->extra1);
5433 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5434 void __user *buffer, size_t *lenp, loff_t *ppos)
5436 int *valp = ctl->data;
5439 struct ctl_table lctl;
5443 * ctl->data points to idev->cnf.disable_ipv6, we should
5444 * not modify it until we get the rtnl lock.
5449 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5452 ret = addrconf_disable_ipv6(ctl, valp, val);
5459 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5460 void __user *buffer, size_t *lenp, loff_t *ppos)
5462 int *valp = ctl->data;
5467 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5470 if (write && old != new) {
5471 struct net *net = ctl->extra2;
5473 if (!rtnl_trylock())
5474 return restart_syscall();
5476 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5477 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5478 NETCONFA_IFINDEX_DEFAULT,
5479 net->ipv6.devconf_dflt);
5480 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5481 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5482 NETCONFA_IFINDEX_ALL,
5483 net->ipv6.devconf_all);
5485 struct inet6_dev *idev = ctl->extra1;
5487 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5497 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5498 void __user *buffer, size_t *lenp,
5502 struct in6_addr addr;
5503 char str[IPV6_MAX_STRLEN];
5504 struct ctl_table lctl = *ctl;
5505 struct net *net = ctl->extra2;
5506 struct ipv6_stable_secret *secret = ctl->data;
5508 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5511 lctl.maxlen = IPV6_MAX_STRLEN;
5514 if (!rtnl_trylock())
5515 return restart_syscall();
5517 if (!write && !secret->initialized) {
5522 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5523 if (err >= sizeof(str)) {
5528 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5532 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5537 secret->initialized = true;
5538 secret->secret = addr;
5540 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5541 struct net_device *dev;
5543 for_each_netdev(net, dev) {
5544 struct inet6_dev *idev = __in6_dev_get(dev);
5547 idev->addr_gen_mode =
5548 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5552 struct inet6_dev *idev = ctl->extra1;
5554 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5564 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5566 void __user *buffer,
5570 int *valp = ctl->data;
5573 struct ctl_table lctl;
5576 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5577 * we should not modify it until we get the rtnl lock.
5582 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5585 ret = addrconf_fixup_linkdown(ctl, valp, val);
5591 static struct addrconf_sysctl_table
5593 struct ctl_table_header *sysctl_header;
5594 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5595 } addrconf_sysctl __read_mostly = {
5596 .sysctl_header = NULL,
5599 .procname = "forwarding",
5600 .data = &ipv6_devconf.forwarding,
5601 .maxlen = sizeof(int),
5603 .proc_handler = addrconf_sysctl_forward,
5606 .procname = "hop_limit",
5607 .data = &ipv6_devconf.hop_limit,
5608 .maxlen = sizeof(int),
5610 .proc_handler = addrconf_sysctl_hop_limit,
5614 .data = &ipv6_devconf.mtu6,
5615 .maxlen = sizeof(int),
5617 .proc_handler = addrconf_sysctl_mtu,
5620 .procname = "accept_ra",
5621 .data = &ipv6_devconf.accept_ra,
5622 .maxlen = sizeof(int),
5624 .proc_handler = proc_dointvec,
5627 .procname = "accept_redirects",
5628 .data = &ipv6_devconf.accept_redirects,
5629 .maxlen = sizeof(int),
5631 .proc_handler = proc_dointvec,
5634 .procname = "autoconf",
5635 .data = &ipv6_devconf.autoconf,
5636 .maxlen = sizeof(int),
5638 .proc_handler = proc_dointvec,
5641 .procname = "dad_transmits",
5642 .data = &ipv6_devconf.dad_transmits,
5643 .maxlen = sizeof(int),
5645 .proc_handler = proc_dointvec,
5648 .procname = "router_solicitations",
5649 .data = &ipv6_devconf.rtr_solicits,
5650 .maxlen = sizeof(int),
5652 .proc_handler = proc_dointvec,
5655 .procname = "router_solicitation_interval",
5656 .data = &ipv6_devconf.rtr_solicit_interval,
5657 .maxlen = sizeof(int),
5659 .proc_handler = proc_dointvec_jiffies,
5662 .procname = "router_solicitation_delay",
5663 .data = &ipv6_devconf.rtr_solicit_delay,
5664 .maxlen = sizeof(int),
5666 .proc_handler = proc_dointvec_jiffies,
5669 .procname = "force_mld_version",
5670 .data = &ipv6_devconf.force_mld_version,
5671 .maxlen = sizeof(int),
5673 .proc_handler = proc_dointvec,
5676 .procname = "mldv1_unsolicited_report_interval",
5678 &ipv6_devconf.mldv1_unsolicited_report_interval,
5679 .maxlen = sizeof(int),
5681 .proc_handler = proc_dointvec_ms_jiffies,
5684 .procname = "mldv2_unsolicited_report_interval",
5686 &ipv6_devconf.mldv2_unsolicited_report_interval,
5687 .maxlen = sizeof(int),
5689 .proc_handler = proc_dointvec_ms_jiffies,
5692 .procname = "use_tempaddr",
5693 .data = &ipv6_devconf.use_tempaddr,
5694 .maxlen = sizeof(int),
5696 .proc_handler = proc_dointvec,
5699 .procname = "temp_valid_lft",
5700 .data = &ipv6_devconf.temp_valid_lft,
5701 .maxlen = sizeof(int),
5703 .proc_handler = proc_dointvec,
5706 .procname = "temp_prefered_lft",
5707 .data = &ipv6_devconf.temp_prefered_lft,
5708 .maxlen = sizeof(int),
5710 .proc_handler = proc_dointvec,
5713 .procname = "regen_max_retry",
5714 .data = &ipv6_devconf.regen_max_retry,
5715 .maxlen = sizeof(int),
5717 .proc_handler = proc_dointvec,
5720 .procname = "max_desync_factor",
5721 .data = &ipv6_devconf.max_desync_factor,
5722 .maxlen = sizeof(int),
5724 .proc_handler = proc_dointvec,
5727 .procname = "max_addresses",
5728 .data = &ipv6_devconf.max_addresses,
5729 .maxlen = sizeof(int),
5731 .proc_handler = proc_dointvec,
5734 .procname = "accept_ra_defrtr",
5735 .data = &ipv6_devconf.accept_ra_defrtr,
5736 .maxlen = sizeof(int),
5738 .proc_handler = proc_dointvec,
5741 .procname = "accept_ra_min_hop_limit",
5742 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5743 .maxlen = sizeof(int),
5745 .proc_handler = proc_dointvec,
5748 .procname = "accept_ra_pinfo",
5749 .data = &ipv6_devconf.accept_ra_pinfo,
5750 .maxlen = sizeof(int),
5752 .proc_handler = proc_dointvec,
5754 #ifdef CONFIG_IPV6_ROUTER_PREF
5756 .procname = "accept_ra_rtr_pref",
5757 .data = &ipv6_devconf.accept_ra_rtr_pref,
5758 .maxlen = sizeof(int),
5760 .proc_handler = proc_dointvec,
5763 .procname = "router_probe_interval",
5764 .data = &ipv6_devconf.rtr_probe_interval,
5765 .maxlen = sizeof(int),
5767 .proc_handler = proc_dointvec_jiffies,
5769 #ifdef CONFIG_IPV6_ROUTE_INFO
5771 .procname = "accept_ra_rt_info_max_plen",
5772 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5773 .maxlen = sizeof(int),
5775 .proc_handler = proc_dointvec,
5780 .procname = "proxy_ndp",
5781 .data = &ipv6_devconf.proxy_ndp,
5782 .maxlen = sizeof(int),
5784 .proc_handler = addrconf_sysctl_proxy_ndp,
5787 .procname = "accept_source_route",
5788 .data = &ipv6_devconf.accept_source_route,
5789 .maxlen = sizeof(int),
5791 .proc_handler = proc_dointvec,
5793 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5795 .procname = "optimistic_dad",
5796 .data = &ipv6_devconf.optimistic_dad,
5797 .maxlen = sizeof(int),
5799 .proc_handler = proc_dointvec,
5803 .procname = "use_optimistic",
5804 .data = &ipv6_devconf.use_optimistic,
5805 .maxlen = sizeof(int),
5807 .proc_handler = proc_dointvec,
5811 #ifdef CONFIG_IPV6_MROUTE
5813 .procname = "mc_forwarding",
5814 .data = &ipv6_devconf.mc_forwarding,
5815 .maxlen = sizeof(int),
5817 .proc_handler = proc_dointvec,
5821 .procname = "disable_ipv6",
5822 .data = &ipv6_devconf.disable_ipv6,
5823 .maxlen = sizeof(int),
5825 .proc_handler = addrconf_sysctl_disable,
5828 .procname = "accept_dad",
5829 .data = &ipv6_devconf.accept_dad,
5830 .maxlen = sizeof(int),
5832 .proc_handler = proc_dointvec,
5835 .procname = "force_tllao",
5836 .data = &ipv6_devconf.force_tllao,
5837 .maxlen = sizeof(int),
5839 .proc_handler = proc_dointvec
5842 .procname = "ndisc_notify",
5843 .data = &ipv6_devconf.ndisc_notify,
5844 .maxlen = sizeof(int),
5846 .proc_handler = proc_dointvec
5849 .procname = "suppress_frag_ndisc",
5850 .data = &ipv6_devconf.suppress_frag_ndisc,
5851 .maxlen = sizeof(int),
5853 .proc_handler = proc_dointvec
5856 .procname = "accept_ra_from_local",
5857 .data = &ipv6_devconf.accept_ra_from_local,
5858 .maxlen = sizeof(int),
5860 .proc_handler = proc_dointvec,
5863 .procname = "accept_ra_mtu",
5864 .data = &ipv6_devconf.accept_ra_mtu,
5865 .maxlen = sizeof(int),
5867 .proc_handler = proc_dointvec,
5870 .procname = "stable_secret",
5871 .data = &ipv6_devconf.stable_secret,
5872 .maxlen = IPV6_MAX_STRLEN,
5874 .proc_handler = addrconf_sysctl_stable_secret,
5877 .procname = "use_oif_addrs_only",
5878 .data = &ipv6_devconf.use_oif_addrs_only,
5879 .maxlen = sizeof(int),
5881 .proc_handler = proc_dointvec,
5884 .procname = "ignore_routes_with_linkdown",
5885 .data = &ipv6_devconf.ignore_routes_with_linkdown,
5886 .maxlen = sizeof(int),
5888 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
5891 .procname = "drop_unicast_in_l2_multicast",
5892 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
5893 .maxlen = sizeof(int),
5895 .proc_handler = proc_dointvec,
5898 .procname = "drop_unsolicited_na",
5899 .data = &ipv6_devconf.drop_unsolicited_na,
5900 .maxlen = sizeof(int),
5902 .proc_handler = proc_dointvec,
5905 .procname = "keep_addr_on_down",
5906 .data = &ipv6_devconf.keep_addr_on_down,
5907 .maxlen = sizeof(int),
5909 .proc_handler = proc_dointvec,
5918 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5919 struct inet6_dev *idev, struct ipv6_devconf *p)
5922 struct addrconf_sysctl_table *t;
5923 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5925 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5929 for (i = 0; t->addrconf_vars[i].data; i++) {
5930 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5931 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5932 t->addrconf_vars[i].extra2 = net;
5935 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5937 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5938 if (!t->sysctl_header)
5950 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5952 struct addrconf_sysctl_table *t;
5959 unregister_net_sysctl_table(t->sysctl_header);
5963 static int addrconf_sysctl_register(struct inet6_dev *idev)
5967 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5970 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5971 &ndisc_ifinfo_sysctl_change);
5974 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5977 neigh_sysctl_unregister(idev->nd_parms);
5982 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5984 __addrconf_sysctl_unregister(&idev->cnf);
5985 neigh_sysctl_unregister(idev->nd_parms);
5991 static int __net_init addrconf_init_net(struct net *net)
5994 struct ipv6_devconf *all, *dflt;
5996 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6000 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6002 goto err_alloc_dflt;
6004 /* these will be inherited by all namespaces */
6005 dflt->autoconf = ipv6_defaults.autoconf;
6006 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6008 dflt->stable_secret.initialized = false;
6009 all->stable_secret.initialized = false;
6011 net->ipv6.devconf_all = all;
6012 net->ipv6.devconf_dflt = dflt;
6014 #ifdef CONFIG_SYSCTL
6015 err = __addrconf_sysctl_register(net, "all", NULL, all);
6019 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6025 #ifdef CONFIG_SYSCTL
6027 __addrconf_sysctl_unregister(all);
6037 static void __net_exit addrconf_exit_net(struct net *net)
6039 #ifdef CONFIG_SYSCTL
6040 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6041 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6043 kfree(net->ipv6.devconf_dflt);
6044 kfree(net->ipv6.devconf_all);
6047 static struct pernet_operations addrconf_ops = {
6048 .init = addrconf_init_net,
6049 .exit = addrconf_exit_net,
6052 static struct rtnl_af_ops inet6_ops __read_mostly = {
6054 .fill_link_af = inet6_fill_link_af,
6055 .get_link_af_size = inet6_get_link_af_size,
6056 .validate_link_af = inet6_validate_link_af,
6057 .set_link_af = inet6_set_link_af,
6061 * Init / cleanup code
6064 int __init addrconf_init(void)
6066 struct inet6_dev *idev;
6069 err = ipv6_addr_label_init();
6071 pr_crit("%s: cannot initialize default policy table: %d\n",
6076 err = register_pernet_subsys(&addrconf_ops);
6080 addrconf_wq = create_workqueue("ipv6_addrconf");
6086 /* The addrconf netdev notifier requires that loopback_dev
6087 * has it's ipv6 private information allocated and setup
6088 * before it can bring up and give link-local addresses
6089 * to other devices which are up.
6091 * Unfortunately, loopback_dev is not necessarily the first
6092 * entry in the global dev_base list of net devices. In fact,
6093 * it is likely to be the very last entry on that list.
6094 * So this causes the notifier registry below to try and
6095 * give link-local addresses to all devices besides loopback_dev
6096 * first, then loopback_dev, which cases all the non-loopback_dev
6097 * devices to fail to get a link-local address.
6099 * So, as a temporary fix, allocate the ipv6 structure for
6100 * loopback_dev first by hand.
6101 * Longer term, all of the dependencies ipv6 has upon the loopback
6102 * device and it being up should be removed.
6105 idev = ipv6_add_dev(init_net.loopback_dev);
6108 err = PTR_ERR(idev);
6112 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6113 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6115 register_netdevice_notifier(&ipv6_dev_notf);
6119 rtnl_af_register(&inet6_ops);
6121 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6126 /* Only the first call to __rtnl_register can fail */
6127 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6128 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6129 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6130 inet6_dump_ifaddr, NULL);
6131 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6132 inet6_dump_ifmcaddr, NULL);
6133 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6134 inet6_dump_ifacaddr, NULL);
6135 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6136 inet6_netconf_dump_devconf, NULL);
6138 ipv6_addr_label_rtnl_register();
6142 rtnl_af_unregister(&inet6_ops);
6143 unregister_netdevice_notifier(&ipv6_dev_notf);
6145 destroy_workqueue(addrconf_wq);
6147 unregister_pernet_subsys(&addrconf_ops);
6149 ipv6_addr_label_cleanup();
6154 void addrconf_cleanup(void)
6156 struct net_device *dev;
6159 unregister_netdevice_notifier(&ipv6_dev_notf);
6160 unregister_pernet_subsys(&addrconf_ops);
6161 ipv6_addr_label_cleanup();
6165 __rtnl_af_unregister(&inet6_ops);
6167 /* clean dev list */
6168 for_each_netdev(&init_net, dev) {
6169 if (__in6_dev_get(dev) == NULL)
6171 addrconf_ifdown(dev, 1);
6173 addrconf_ifdown(init_net.loopback_dev, 2);
6178 spin_lock_bh(&addrconf_hash_lock);
6179 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6180 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6181 spin_unlock_bh(&addrconf_hash_lock);
6182 cancel_delayed_work(&addr_chk_work);
6185 destroy_workqueue(addrconf_wq);