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/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
62 #include <linux/sysctl.h>
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
70 #include <net/net_namespace.h>
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
97 #define INFINITY_LIFE_TIME 0xFFFFFFFF
99 #define IPV6_MAX_STRLEN \
100 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
102 static inline u32 cstamp_delta(unsigned long cstamp)
104 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
109 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
110 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
111 do_div(tmp, 1000000);
115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
117 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
118 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
119 do_div(tmp, 1000000);
120 if ((s32)tmp > mrt) {
121 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
122 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
123 do_div(tmp, 1000000);
129 static int addrconf_sysctl_register(struct inet6_dev *idev);
130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
142 static void ipv6_regen_rndid(struct inet6_dev *idev);
143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
146 static int ipv6_count_addresses(const struct inet6_dev *idev);
147 static int ipv6_generate_stable_address(struct in6_addr *addr,
149 const struct inet6_dev *idev);
151 #define IN6_ADDR_HSIZE_SHIFT 8
152 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
154 * Configured unicast address hash table
156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
157 static DEFINE_SPINLOCK(addrconf_hash_lock);
159 static void addrconf_verify(void);
160 static void addrconf_verify_rtnl(void);
161 static void addrconf_verify_work(struct work_struct *);
163 static struct workqueue_struct *addrconf_wq;
164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
169 static void addrconf_type_change(struct net_device *dev,
170 unsigned long event);
171 static int addrconf_ifdown(struct net_device *dev, int how);
173 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
175 const struct net_device *dev,
176 u32 flags, u32 noflags);
178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
179 static void addrconf_dad_work(struct work_struct *w);
180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
182 static void addrconf_dad_run(struct inet6_dev *idev);
183 static void addrconf_rs_timer(struct timer_list *t);
184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
188 struct prefix_info *pinfo);
190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
192 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
193 .mtu6 = IPV6_MIN_MTU,
195 .accept_redirects = 1,
197 .force_mld_version = 0,
198 .mldv1_unsolicited_report_interval = 10 * HZ,
199 .mldv2_unsolicited_report_interval = HZ,
201 .rtr_solicits = MAX_RTR_SOLICITATIONS,
202 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
203 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
204 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
206 .temp_valid_lft = TEMP_VALID_LIFETIME,
207 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
208 .regen_max_retry = REGEN_MAX_RETRY,
209 .max_desync_factor = MAX_DESYNC_FACTOR,
210 .max_addresses = IPV6_MAX_ADDRESSES,
211 .accept_ra_defrtr = 1,
212 .accept_ra_from_local = 0,
213 .accept_ra_min_hop_limit= 1,
214 .accept_ra_pinfo = 1,
215 #ifdef CONFIG_IPV6_ROUTER_PREF
216 .accept_ra_rtr_pref = 1,
217 .rtr_probe_interval = 60 * HZ,
218 #ifdef CONFIG_IPV6_ROUTE_INFO
219 .accept_ra_rt_info_min_plen = 0,
220 .accept_ra_rt_info_max_plen = 0,
224 .accept_source_route = 0, /* we do not accept RH0 by default. */
227 .suppress_frag_ndisc = 1,
230 .initialized = false,
232 .use_oif_addrs_only = 0,
233 .ignore_routes_with_linkdown = 0,
234 .keep_addr_on_down = 0,
236 #ifdef CONFIG_IPV6_SEG6_HMAC
237 .seg6_require_hmac = 0,
240 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
246 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
247 .mtu6 = IPV6_MIN_MTU,
249 .accept_redirects = 1,
251 .force_mld_version = 0,
252 .mldv1_unsolicited_report_interval = 10 * HZ,
253 .mldv2_unsolicited_report_interval = HZ,
255 .rtr_solicits = MAX_RTR_SOLICITATIONS,
256 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
257 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
258 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
260 .temp_valid_lft = TEMP_VALID_LIFETIME,
261 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
262 .regen_max_retry = REGEN_MAX_RETRY,
263 .max_desync_factor = MAX_DESYNC_FACTOR,
264 .max_addresses = IPV6_MAX_ADDRESSES,
265 .accept_ra_defrtr = 1,
266 .accept_ra_from_local = 0,
267 .accept_ra_min_hop_limit= 1,
268 .accept_ra_pinfo = 1,
269 #ifdef CONFIG_IPV6_ROUTER_PREF
270 .accept_ra_rtr_pref = 1,
271 .rtr_probe_interval = 60 * HZ,
272 #ifdef CONFIG_IPV6_ROUTE_INFO
273 .accept_ra_rt_info_min_plen = 0,
274 .accept_ra_rt_info_max_plen = 0,
278 .accept_source_route = 0, /* we do not accept RH0 by default. */
281 .suppress_frag_ndisc = 1,
284 .initialized = false,
286 .use_oif_addrs_only = 0,
287 .ignore_routes_with_linkdown = 0,
288 .keep_addr_on_down = 0,
290 #ifdef CONFIG_IPV6_SEG6_HMAC
291 .seg6_require_hmac = 0,
294 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
298 /* Check if link is ready: is it up and is a valid qdisc available */
299 static inline bool addrconf_link_ready(const struct net_device *dev)
301 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
306 if (del_timer(&idev->rs_timer))
310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
312 if (cancel_delayed_work(&ifp->dad_work))
316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
319 if (!timer_pending(&idev->rs_timer))
321 mod_timer(&idev->rs_timer, jiffies + when);
324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
328 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
332 static int snmp6_alloc_dev(struct inet6_dev *idev)
336 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
337 if (!idev->stats.ipv6)
340 for_each_possible_cpu(i) {
341 struct ipstats_mib *addrconf_stats;
342 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
343 u64_stats_init(&addrconf_stats->syncp);
347 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
349 if (!idev->stats.icmpv6dev)
351 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
353 if (!idev->stats.icmpv6msgdev)
359 kfree(idev->stats.icmpv6dev);
361 free_percpu(idev->stats.ipv6);
366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
368 struct inet6_dev *ndev;
373 if (dev->mtu < IPV6_MIN_MTU)
374 return ERR_PTR(-EINVAL);
376 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
380 rwlock_init(&ndev->lock);
382 INIT_LIST_HEAD(&ndev->addr_list);
383 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
384 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
386 if (ndev->cnf.stable_secret.initialized)
387 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
389 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
391 ndev->cnf.mtu6 = dev->mtu;
392 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
393 if (!ndev->nd_parms) {
397 if (ndev->cnf.forwarding)
398 dev_disable_lro(dev);
399 /* We refer to the device */
402 if (snmp6_alloc_dev(ndev) < 0) {
403 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
405 neigh_parms_release(&nd_tbl, ndev->nd_parms);
411 if (snmp6_register_dev(ndev) < 0) {
412 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
413 __func__, dev->name);
417 /* One reference from device. */
418 refcount_set(&ndev->refcnt, 1);
420 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
421 ndev->cnf.accept_dad = -1;
423 #if IS_ENABLED(CONFIG_IPV6_SIT)
424 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
425 pr_info("%s: Disabled Multicast RS\n", dev->name);
426 ndev->cnf.rtr_solicits = 0;
430 INIT_LIST_HEAD(&ndev->tempaddr_list);
431 ndev->desync_factor = U32_MAX;
432 if ((dev->flags&IFF_LOOPBACK) ||
433 dev->type == ARPHRD_TUNNEL ||
434 dev->type == ARPHRD_TUNNEL6 ||
435 dev->type == ARPHRD_SIT ||
436 dev->type == ARPHRD_NONE) {
437 ndev->cnf.use_tempaddr = -1;
439 ipv6_regen_rndid(ndev);
441 ndev->token = in6addr_any;
443 if (netif_running(dev) && addrconf_link_ready(dev))
444 ndev->if_flags |= IF_READY;
446 ipv6_mc_init_dev(ndev);
447 ndev->tstamp = jiffies;
448 err = addrconf_sysctl_register(ndev);
450 ipv6_mc_destroy_dev(ndev);
451 snmp6_unregister_dev(ndev);
454 /* protected by rtnl_lock */
455 rcu_assign_pointer(dev->ip6_ptr, ndev);
457 /* Join interface-local all-node multicast group */
458 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
460 /* Join all-node multicast group */
461 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
463 /* Join all-router multicast group if forwarding is set */
464 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
465 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
470 neigh_parms_release(&nd_tbl, ndev->nd_parms);
472 in6_dev_finish_destroy(ndev);
476 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
478 struct inet6_dev *idev;
482 idev = __in6_dev_get(dev);
484 idev = ipv6_add_dev(dev);
489 if (dev->flags&IFF_UP)
494 static int inet6_netconf_msgsize_devconf(int type)
496 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
497 + nla_total_size(4); /* NETCONFA_IFINDEX */
500 if (type == NETCONFA_ALL)
503 if (all || type == NETCONFA_FORWARDING)
504 size += nla_total_size(4);
505 #ifdef CONFIG_IPV6_MROUTE
506 if (all || type == NETCONFA_MC_FORWARDING)
507 size += nla_total_size(4);
509 if (all || type == NETCONFA_PROXY_NEIGH)
510 size += nla_total_size(4);
512 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
513 size += nla_total_size(4);
518 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
519 struct ipv6_devconf *devconf, u32 portid,
520 u32 seq, int event, unsigned int flags,
523 struct nlmsghdr *nlh;
524 struct netconfmsg *ncm;
527 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
532 if (type == NETCONFA_ALL)
535 ncm = nlmsg_data(nlh);
536 ncm->ncm_family = AF_INET6;
538 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
539 goto nla_put_failure;
544 if ((all || type == NETCONFA_FORWARDING) &&
545 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
546 goto nla_put_failure;
547 #ifdef CONFIG_IPV6_MROUTE
548 if ((all || type == NETCONFA_MC_FORWARDING) &&
549 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
550 devconf->mc_forwarding) < 0)
551 goto nla_put_failure;
553 if ((all || type == NETCONFA_PROXY_NEIGH) &&
554 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
555 goto nla_put_failure;
557 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
558 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
559 devconf->ignore_routes_with_linkdown) < 0)
560 goto nla_put_failure;
567 nlmsg_cancel(skb, nlh);
571 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
572 int ifindex, struct ipv6_devconf *devconf)
577 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
581 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
584 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
585 WARN_ON(err == -EMSGSIZE);
589 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
592 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
595 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
596 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
597 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
598 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
599 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
602 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
603 struct nlmsghdr *nlh,
604 struct netlink_ext_ack *extack)
606 struct net *net = sock_net(in_skb->sk);
607 struct nlattr *tb[NETCONFA_MAX+1];
608 struct inet6_dev *in6_dev = NULL;
609 struct net_device *dev = NULL;
610 struct netconfmsg *ncm;
612 struct ipv6_devconf *devconf;
616 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
617 devconf_ipv6_policy, extack);
621 if (!tb[NETCONFA_IFINDEX])
625 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
627 case NETCONFA_IFINDEX_ALL:
628 devconf = net->ipv6.devconf_all;
630 case NETCONFA_IFINDEX_DEFAULT:
631 devconf = net->ipv6.devconf_dflt;
634 dev = dev_get_by_index(net, ifindex);
637 in6_dev = in6_dev_get(dev);
640 devconf = &in6_dev->cnf;
645 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
649 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
650 NETLINK_CB(in_skb).portid,
651 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
654 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
655 WARN_ON(err == -EMSGSIZE);
659 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
662 in6_dev_put(in6_dev);
668 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
669 struct netlink_callback *cb)
671 struct net *net = sock_net(skb->sk);
674 struct net_device *dev;
675 struct inet6_dev *idev;
676 struct hlist_head *head;
679 s_idx = idx = cb->args[1];
681 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
683 head = &net->dev_index_head[h];
685 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
687 hlist_for_each_entry_rcu(dev, head, index_hlist) {
690 idev = __in6_dev_get(dev);
694 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
696 NETLINK_CB(cb->skb).portid,
704 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
710 if (h == NETDEV_HASHENTRIES) {
711 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
712 net->ipv6.devconf_all,
713 NETLINK_CB(cb->skb).portid,
715 RTM_NEWNETCONF, NLM_F_MULTI,
721 if (h == NETDEV_HASHENTRIES + 1) {
722 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
723 net->ipv6.devconf_dflt,
724 NETLINK_CB(cb->skb).portid,
726 RTM_NEWNETCONF, NLM_F_MULTI,
740 static void dev_forward_change(struct inet6_dev *idev)
742 struct net_device *dev;
743 struct inet6_ifaddr *ifa;
748 if (idev->cnf.forwarding)
749 dev_disable_lro(dev);
750 if (dev->flags & IFF_MULTICAST) {
751 if (idev->cnf.forwarding) {
752 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
753 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
754 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
756 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
757 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
758 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
762 list_for_each_entry(ifa, &idev->addr_list, if_list) {
763 if (ifa->flags&IFA_F_TENTATIVE)
765 if (idev->cnf.forwarding)
766 addrconf_join_anycast(ifa);
768 addrconf_leave_anycast(ifa);
770 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
772 dev->ifindex, &idev->cnf);
776 static void addrconf_forward_change(struct net *net, __s32 newf)
778 struct net_device *dev;
779 struct inet6_dev *idev;
781 for_each_netdev(net, dev) {
782 idev = __in6_dev_get(dev);
784 int changed = (!idev->cnf.forwarding) ^ (!newf);
785 idev->cnf.forwarding = newf;
787 dev_forward_change(idev);
792 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
798 return restart_syscall();
800 net = (struct net *)table->extra2;
804 if (p == &net->ipv6.devconf_dflt->forwarding) {
805 if ((!newf) ^ (!old))
806 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
808 NETCONFA_IFINDEX_DEFAULT,
809 net->ipv6.devconf_dflt);
814 if (p == &net->ipv6.devconf_all->forwarding) {
815 int old_dflt = net->ipv6.devconf_dflt->forwarding;
817 net->ipv6.devconf_dflt->forwarding = newf;
818 if ((!newf) ^ (!old_dflt))
819 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
821 NETCONFA_IFINDEX_DEFAULT,
822 net->ipv6.devconf_dflt);
824 addrconf_forward_change(net, newf);
825 if ((!newf) ^ (!old))
826 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
828 NETCONFA_IFINDEX_ALL,
829 net->ipv6.devconf_all);
830 } else if ((!newf) ^ (!old))
831 dev_forward_change((struct inet6_dev *)table->extra1);
835 rt6_purge_dflt_routers(net);
839 static void addrconf_linkdown_change(struct net *net, __s32 newf)
841 struct net_device *dev;
842 struct inet6_dev *idev;
844 for_each_netdev(net, dev) {
845 idev = __in6_dev_get(dev);
847 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
849 idev->cnf.ignore_routes_with_linkdown = newf;
851 inet6_netconf_notify_devconf(dev_net(dev),
853 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
860 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
866 return restart_syscall();
868 net = (struct net *)table->extra2;
872 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
873 if ((!newf) ^ (!old))
874 inet6_netconf_notify_devconf(net,
876 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
877 NETCONFA_IFINDEX_DEFAULT,
878 net->ipv6.devconf_dflt);
883 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
884 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
885 addrconf_linkdown_change(net, newf);
886 if ((!newf) ^ (!old))
887 inet6_netconf_notify_devconf(net,
889 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
890 NETCONFA_IFINDEX_ALL,
891 net->ipv6.devconf_all);
900 /* Nobody refers to this ifaddr, destroy it */
901 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
903 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
905 #ifdef NET_REFCNT_DEBUG
906 pr_debug("%s\n", __func__);
909 in6_dev_put(ifp->idev);
911 if (cancel_delayed_work(&ifp->dad_work))
912 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
915 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
916 pr_warn("Freeing alive inet6 address %p\n", ifp);
925 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
928 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
931 * Each device address list is sorted in order of scope -
932 * global before linklocal.
934 list_for_each(p, &idev->addr_list) {
935 struct inet6_ifaddr *ifa
936 = list_entry(p, struct inet6_ifaddr, if_list);
937 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
941 list_add_tail_rcu(&ifp->if_list, p);
944 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
946 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
948 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
951 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
952 struct net_device *dev, unsigned int hash)
954 struct inet6_ifaddr *ifp;
956 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
957 if (!net_eq(dev_net(ifp->idev->dev), net))
959 if (ipv6_addr_equal(&ifp->addr, addr)) {
960 if (!dev || ifp->idev->dev == dev)
967 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
969 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
972 spin_lock(&addrconf_hash_lock);
974 /* Ignore adding duplicate addresses on an interface */
975 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
976 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
979 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
982 spin_unlock(&addrconf_hash_lock);
987 /* On success it returns ifp with increased reference count */
989 static struct inet6_ifaddr *
990 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
991 const struct in6_addr *peer_addr, int pfxlen,
992 int scope, u32 flags, u32 valid_lft, u32 prefered_lft,
993 bool can_block, struct netlink_ext_ack *extack)
995 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
996 struct net *net = dev_net(idev->dev);
997 struct inet6_ifaddr *ifa = NULL;
998 struct rt6_info *rt = NULL;
1000 int addr_type = ipv6_addr_type(addr);
1002 if (addr_type == IPV6_ADDR_ANY ||
1003 addr_type & IPV6_ADDR_MULTICAST ||
1004 (!(idev->dev->flags & IFF_LOOPBACK) &&
1005 addr_type & IPV6_ADDR_LOOPBACK))
1006 return ERR_PTR(-EADDRNOTAVAIL);
1009 err = -ENODEV; /*XXX*/
1013 if (idev->cnf.disable_ipv6) {
1018 /* validator notifier needs to be blocking;
1019 * do not call in atomic context
1022 struct in6_validator_info i6vi = {
1028 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1029 err = notifier_to_errno(err);
1034 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1040 rt = addrconf_dst_alloc(idev, addr, false);
1047 if (net->ipv6.devconf_all->disable_policy ||
1048 idev->cnf.disable_policy)
1049 rt->dst.flags |= DST_NOPOLICY;
1051 neigh_parms_data_state_setall(idev->nd_parms);
1055 ifa->peer_addr = *peer_addr;
1057 spin_lock_init(&ifa->lock);
1058 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1059 INIT_HLIST_NODE(&ifa->addr_lst);
1061 ifa->prefix_len = pfxlen;
1063 /* No need to add the TENTATIVE flag for addresses with NODAD */
1064 if (!(flags & IFA_F_NODAD))
1065 ifa->flags |= IFA_F_TENTATIVE;
1066 ifa->valid_lft = valid_lft;
1067 ifa->prefered_lft = prefered_lft;
1068 ifa->cstamp = ifa->tstamp = jiffies;
1069 ifa->tokenized = false;
1077 refcount_set(&ifa->refcnt, 1);
1081 err = ipv6_add_addr_hash(idev->dev, ifa);
1083 rcu_read_unlock_bh();
1087 write_lock(&idev->lock);
1089 /* Add to inet6_dev unicast addr list. */
1090 ipv6_link_dev_addr(idev, ifa);
1092 if (ifa->flags&IFA_F_TEMPORARY) {
1093 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1098 write_unlock(&idev->lock);
1100 rcu_read_unlock_bh();
1102 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1104 if (unlikely(err < 0)) {
1109 in6_dev_put(ifa->idev);
1118 enum cleanup_prefix_rt_t {
1119 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1120 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1121 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1125 * Check, whether the prefix for ifp would still need a prefix route
1126 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1129 * 1) we don't purge prefix if address was not permanent.
1130 * prefix is managed by its own lifetime.
1131 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1132 * 3) if there are no addresses, delete prefix.
1133 * 4) if there are still other permanent address(es),
1134 * corresponding prefix is still permanent.
1135 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1136 * don't purge the prefix, assume user space is managing it.
1137 * 6) otherwise, update prefix lifetime to the
1138 * longest valid lifetime among the corresponding
1139 * addresses on the device.
1140 * Note: subsequent RA will update lifetime.
1142 static enum cleanup_prefix_rt_t
1143 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1145 struct inet6_ifaddr *ifa;
1146 struct inet6_dev *idev = ifp->idev;
1147 unsigned long lifetime;
1148 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1152 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1155 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1158 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1159 return CLEANUP_PREFIX_RT_NOP;
1161 action = CLEANUP_PREFIX_RT_EXPIRE;
1163 spin_lock(&ifa->lock);
1165 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1167 * Note: Because this address is
1168 * not permanent, lifetime <
1169 * LONG_MAX / HZ here.
1171 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1172 *expires = ifa->tstamp + lifetime * HZ;
1173 spin_unlock(&ifa->lock);
1180 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1182 struct rt6_info *rt;
1184 rt = addrconf_get_prefix_route(&ifp->addr,
1187 0, RTF_GATEWAY | RTF_DEFAULT);
1192 if (!(rt->rt6i_flags & RTF_EXPIRES))
1193 rt6_set_expires(rt, expires);
1200 /* This function wants to get referenced ifp and releases it before return */
1202 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1205 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1206 unsigned long expires;
1210 spin_lock_bh(&ifp->lock);
1212 ifp->state = INET6_IFADDR_STATE_DEAD;
1213 spin_unlock_bh(&ifp->lock);
1215 if (state == INET6_IFADDR_STATE_DEAD)
1218 spin_lock_bh(&addrconf_hash_lock);
1219 hlist_del_init_rcu(&ifp->addr_lst);
1220 spin_unlock_bh(&addrconf_hash_lock);
1222 write_lock_bh(&ifp->idev->lock);
1224 if (ifp->flags&IFA_F_TEMPORARY) {
1225 list_del(&ifp->tmp_list);
1227 in6_ifa_put(ifp->ifpub);
1233 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1234 action = check_cleanup_prefix_route(ifp, &expires);
1236 list_del_rcu(&ifp->if_list);
1239 write_unlock_bh(&ifp->idev->lock);
1241 addrconf_del_dad_work(ifp);
1243 ipv6_ifa_notify(RTM_DELADDR, ifp);
1245 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1247 if (action != CLEANUP_PREFIX_RT_NOP) {
1248 cleanup_prefix_route(ifp, expires,
1249 action == CLEANUP_PREFIX_RT_DEL);
1252 /* clean up prefsrc entries */
1253 rt6_remove_prefsrc(ifp);
1258 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1259 struct inet6_ifaddr *ift,
1262 struct inet6_dev *idev = ifp->idev;
1263 struct in6_addr addr, *tmpaddr;
1264 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1265 unsigned long regen_advance;
1269 unsigned long now = jiffies;
1270 long max_desync_factor;
1271 s32 cnf_temp_preferred_lft;
1273 write_lock_bh(&idev->lock);
1275 spin_lock_bh(&ift->lock);
1276 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1277 spin_unlock_bh(&ift->lock);
1284 if (idev->cnf.use_tempaddr <= 0) {
1285 write_unlock_bh(&idev->lock);
1286 pr_info("%s: use_tempaddr is disabled\n", __func__);
1291 spin_lock_bh(&ifp->lock);
1292 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1293 idev->cnf.use_tempaddr = -1; /*XXX*/
1294 spin_unlock_bh(&ifp->lock);
1295 write_unlock_bh(&idev->lock);
1296 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1303 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1304 ipv6_try_regen_rndid(idev, tmpaddr);
1305 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1306 age = (now - ifp->tstamp) / HZ;
1308 regen_advance = idev->cnf.regen_max_retry *
1309 idev->cnf.dad_transmits *
1310 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1312 /* recalculate max_desync_factor each time and update
1313 * idev->desync_factor if it's larger
1315 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1316 max_desync_factor = min_t(__u32,
1317 idev->cnf.max_desync_factor,
1318 cnf_temp_preferred_lft - regen_advance);
1320 if (unlikely(idev->desync_factor > max_desync_factor)) {
1321 if (max_desync_factor > 0) {
1322 get_random_bytes(&idev->desync_factor,
1323 sizeof(idev->desync_factor));
1324 idev->desync_factor %= max_desync_factor;
1326 idev->desync_factor = 0;
1330 tmp_valid_lft = min_t(__u32,
1332 idev->cnf.temp_valid_lft + age);
1333 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1334 idev->desync_factor;
1335 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1336 tmp_plen = ifp->prefix_len;
1337 tmp_tstamp = ifp->tstamp;
1338 spin_unlock_bh(&ifp->lock);
1340 write_unlock_bh(&idev->lock);
1342 /* A temporary address is created only if this calculated Preferred
1343 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1344 * an implementation must not create a temporary address with a zero
1345 * Preferred Lifetime.
1346 * Use age calculation as in addrconf_verify to avoid unnecessary
1347 * temporary addresses being generated.
1349 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1350 if (tmp_prefered_lft <= regen_advance + age) {
1357 addr_flags = IFA_F_TEMPORARY;
1358 /* set in addrconf_prefix_rcv() */
1359 if (ifp->flags & IFA_F_OPTIMISTIC)
1360 addr_flags |= IFA_F_OPTIMISTIC;
1362 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1363 ipv6_addr_scope(&addr), addr_flags,
1364 tmp_valid_lft, tmp_prefered_lft, block, NULL);
1368 pr_info("%s: retry temporary address regeneration\n", __func__);
1370 write_lock_bh(&idev->lock);
1374 spin_lock_bh(&ift->lock);
1377 ift->tstamp = tmp_tstamp;
1378 spin_unlock_bh(&ift->lock);
1380 addrconf_dad_start(ift);
1388 * Choose an appropriate source address (RFC3484)
1391 IPV6_SADDR_RULE_INIT = 0,
1392 IPV6_SADDR_RULE_LOCAL,
1393 IPV6_SADDR_RULE_SCOPE,
1394 IPV6_SADDR_RULE_PREFERRED,
1395 #ifdef CONFIG_IPV6_MIP6
1396 IPV6_SADDR_RULE_HOA,
1398 IPV6_SADDR_RULE_OIF,
1399 IPV6_SADDR_RULE_LABEL,
1400 IPV6_SADDR_RULE_PRIVACY,
1401 IPV6_SADDR_RULE_ORCHID,
1402 IPV6_SADDR_RULE_PREFIX,
1403 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1404 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1409 struct ipv6_saddr_score {
1412 struct inet6_ifaddr *ifa;
1413 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1418 struct ipv6_saddr_dst {
1419 const struct in6_addr *addr;
1426 static inline int ipv6_saddr_preferred(int type)
1428 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1433 static bool ipv6_use_optimistic_addr(struct net *net,
1434 struct inet6_dev *idev)
1436 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1439 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1441 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1450 static bool ipv6_allow_optimistic_dad(struct net *net,
1451 struct inet6_dev *idev)
1453 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1456 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1465 static int ipv6_get_saddr_eval(struct net *net,
1466 struct ipv6_saddr_score *score,
1467 struct ipv6_saddr_dst *dst,
1472 if (i <= score->rule) {
1474 case IPV6_SADDR_RULE_SCOPE:
1475 ret = score->scopedist;
1477 case IPV6_SADDR_RULE_PREFIX:
1478 ret = score->matchlen;
1481 ret = !!test_bit(i, score->scorebits);
1487 case IPV6_SADDR_RULE_INIT:
1488 /* Rule 0: remember if hiscore is not ready yet */
1491 case IPV6_SADDR_RULE_LOCAL:
1492 /* Rule 1: Prefer same address */
1493 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1495 case IPV6_SADDR_RULE_SCOPE:
1496 /* Rule 2: Prefer appropriate scope
1501 * ---+--+-+---> scope
1503 * | d is scope of the destination.
1505 * | \ <- smaller scope is better if
1506 * B-15 | \ if scope is enough for destination.
1507 * | ret = B - scope (-1 <= scope >= d <= 15).
1509 * |/ <- greater is better
1510 * -C / if scope is not enough for destination.
1511 * /| ret = scope - C (-1 <= d < scope <= 15).
1513 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1514 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1515 * Assume B = 0 and we get C > 29.
1517 ret = __ipv6_addr_src_scope(score->addr_type);
1518 if (ret >= dst->scope)
1521 ret -= 128; /* 30 is enough */
1522 score->scopedist = ret;
1524 case IPV6_SADDR_RULE_PREFERRED:
1526 /* Rule 3: Avoid deprecated and optimistic addresses */
1527 u8 avoid = IFA_F_DEPRECATED;
1529 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1530 avoid |= IFA_F_OPTIMISTIC;
1531 ret = ipv6_saddr_preferred(score->addr_type) ||
1532 !(score->ifa->flags & avoid);
1535 #ifdef CONFIG_IPV6_MIP6
1536 case IPV6_SADDR_RULE_HOA:
1538 /* Rule 4: Prefer home address */
1539 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1540 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1544 case IPV6_SADDR_RULE_OIF:
1545 /* Rule 5: Prefer outgoing interface */
1546 ret = (!dst->ifindex ||
1547 dst->ifindex == score->ifa->idev->dev->ifindex);
1549 case IPV6_SADDR_RULE_LABEL:
1550 /* Rule 6: Prefer matching label */
1551 ret = ipv6_addr_label(net,
1552 &score->ifa->addr, score->addr_type,
1553 score->ifa->idev->dev->ifindex) == dst->label;
1555 case IPV6_SADDR_RULE_PRIVACY:
1557 /* Rule 7: Prefer public address
1558 * Note: prefer temporary address if use_tempaddr >= 2
1560 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1561 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1562 score->ifa->idev->cnf.use_tempaddr >= 2;
1563 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1566 case IPV6_SADDR_RULE_ORCHID:
1567 /* Rule 8-: Prefer ORCHID vs ORCHID or
1568 * non-ORCHID vs non-ORCHID
1570 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1571 ipv6_addr_orchid(dst->addr));
1573 case IPV6_SADDR_RULE_PREFIX:
1574 /* Rule 8: Use longest matching prefix */
1575 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1576 if (ret > score->ifa->prefix_len)
1577 ret = score->ifa->prefix_len;
1578 score->matchlen = ret;
1580 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1581 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1582 /* Optimistic addresses still have lower precedence than other
1583 * preferred addresses.
1585 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1593 __set_bit(i, score->scorebits);
1599 static int __ipv6_dev_get_saddr(struct net *net,
1600 struct ipv6_saddr_dst *dst,
1601 struct inet6_dev *idev,
1602 struct ipv6_saddr_score *scores,
1605 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1607 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1611 * - Tentative Address (RFC2462 section 5.4)
1612 * - A tentative address is not considered
1613 * "assigned to an interface" in the traditional
1614 * sense, unless it is also flagged as optimistic.
1615 * - Candidate Source Address (section 4)
1616 * - In any case, anycast addresses, multicast
1617 * addresses, and the unspecified address MUST
1618 * NOT be included in a candidate set.
1620 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1621 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1624 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1626 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1627 score->addr_type & IPV6_ADDR_MULTICAST)) {
1628 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1634 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1636 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1637 int minihiscore, miniscore;
1639 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1640 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1642 if (minihiscore > miniscore) {
1643 if (i == IPV6_SADDR_RULE_SCOPE &&
1644 score->scopedist > 0) {
1647 * each remaining entry
1648 * has too small (not enough)
1649 * scope, because ifa entries
1650 * are sorted by their scope
1656 } else if (minihiscore < miniscore) {
1657 swap(hiscore, score);
1658 hiscore_idx = 1 - hiscore_idx;
1660 /* restore our iterator */
1661 score->ifa = hiscore->ifa;
1671 static int ipv6_get_saddr_master(struct net *net,
1672 const struct net_device *dst_dev,
1673 const struct net_device *master,
1674 struct ipv6_saddr_dst *dst,
1675 struct ipv6_saddr_score *scores,
1678 struct inet6_dev *idev;
1680 idev = __in6_dev_get(dst_dev);
1682 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1683 scores, hiscore_idx);
1685 idev = __in6_dev_get(master);
1687 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1688 scores, hiscore_idx);
1693 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1694 const struct in6_addr *daddr, unsigned int prefs,
1695 struct in6_addr *saddr)
1697 struct ipv6_saddr_score scores[2], *hiscore;
1698 struct ipv6_saddr_dst dst;
1699 struct inet6_dev *idev;
1700 struct net_device *dev;
1702 bool use_oif_addr = false;
1703 int hiscore_idx = 0;
1706 dst_type = __ipv6_addr_type(daddr);
1708 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1709 dst.scope = __ipv6_addr_src_scope(dst_type);
1710 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1713 scores[hiscore_idx].rule = -1;
1714 scores[hiscore_idx].ifa = NULL;
1718 /* Candidate Source Address (section 4)
1719 * - multicast and link-local destination address,
1720 * the set of candidate source address MUST only
1721 * include addresses assigned to interfaces
1722 * belonging to the same link as the outgoing
1724 * (- For site-local destination addresses, the
1725 * set of candidate source addresses MUST only
1726 * include addresses assigned to interfaces
1727 * belonging to the same site as the outgoing
1729 * - "It is RECOMMENDED that the candidate source addresses
1730 * be the set of unicast addresses assigned to the
1731 * interface that will be used to send to the destination
1732 * (the 'outgoing' interface)." (RFC 6724)
1735 idev = __in6_dev_get(dst_dev);
1736 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1737 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1738 (idev && idev->cnf.use_oif_addrs_only)) {
1739 use_oif_addr = true;
1745 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1747 const struct net_device *master;
1750 /* if dst_dev exists and is enslaved to an L3 device, then
1751 * prefer addresses from dst_dev and then the master over
1752 * any other enslaved devices in the L3 domain.
1754 master = l3mdev_master_dev_rcu(dst_dev);
1756 master_idx = master->ifindex;
1758 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1760 scores, hiscore_idx);
1762 if (scores[hiscore_idx].ifa)
1766 for_each_netdev_rcu(net, dev) {
1767 /* only consider addresses on devices in the
1770 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1772 idev = __in6_dev_get(dev);
1775 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1780 hiscore = &scores[hiscore_idx];
1782 ret = -EADDRNOTAVAIL;
1784 *saddr = hiscore->ifa->addr;
1789 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1791 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1794 struct inet6_ifaddr *ifp;
1795 int err = -EADDRNOTAVAIL;
1797 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1798 if (ifp->scope > IFA_LINK)
1800 if (ifp->scope == IFA_LINK &&
1801 !(ifp->flags & banned_flags)) {
1810 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1813 struct inet6_dev *idev;
1814 int err = -EADDRNOTAVAIL;
1817 idev = __in6_dev_get(dev);
1819 read_lock_bh(&idev->lock);
1820 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1821 read_unlock_bh(&idev->lock);
1827 static int ipv6_count_addresses(const struct inet6_dev *idev)
1829 const struct inet6_ifaddr *ifp;
1833 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1839 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1840 const struct net_device *dev, int strict)
1842 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1843 strict, IFA_F_TENTATIVE);
1845 EXPORT_SYMBOL(ipv6_chk_addr);
1847 /* device argument is used to find the L3 domain of interest. If
1848 * skip_dev_check is set, then the ifp device is not checked against
1849 * the passed in dev argument. So the 2 cases for addresses checks are:
1850 * 1. does the address exist in the L3 domain that dev is part of
1851 * (skip_dev_check = true), or
1853 * 2. does the address exist on the specific device
1854 * (skip_dev_check = false)
1856 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1857 const struct net_device *dev, bool skip_dev_check,
1858 int strict, u32 banned_flags)
1860 unsigned int hash = inet6_addr_hash(net, addr);
1861 const struct net_device *l3mdev;
1862 struct inet6_ifaddr *ifp;
1867 l3mdev = l3mdev_master_dev_rcu(dev);
1871 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1872 if (!net_eq(dev_net(ifp->idev->dev), net))
1875 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1878 /* Decouple optimistic from tentative for evaluation here.
1879 * Ban optimistic addresses explicitly, when required.
1881 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1882 ? (ifp->flags&~IFA_F_TENTATIVE)
1884 if (ipv6_addr_equal(&ifp->addr, addr) &&
1885 !(ifp_flags&banned_flags) &&
1886 (!dev || ifp->idev->dev == dev ||
1887 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1896 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1899 /* Compares an address/prefix_len with addresses on device @dev.
1900 * If one is found it returns true.
1902 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1903 const unsigned int prefix_len, struct net_device *dev)
1905 const struct inet6_ifaddr *ifa;
1906 const struct inet6_dev *idev;
1910 idev = __in6_dev_get(dev);
1912 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1913 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1922 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1924 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1926 const struct inet6_ifaddr *ifa;
1927 const struct inet6_dev *idev;
1932 idev = __in6_dev_get(dev);
1934 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1935 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1944 EXPORT_SYMBOL(ipv6_chk_prefix);
1946 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1947 struct net_device *dev, int strict)
1949 unsigned int hash = inet6_addr_hash(net, addr);
1950 struct inet6_ifaddr *ifp, *result = NULL;
1953 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1954 if (!net_eq(dev_net(ifp->idev->dev), net))
1956 if (ipv6_addr_equal(&ifp->addr, addr)) {
1957 if (!dev || ifp->idev->dev == dev ||
1958 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1970 /* Gets referenced address, destroys ifaddr */
1972 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1975 ifp->flags |= IFA_F_DADFAILED;
1977 if (ifp->flags&IFA_F_TEMPORARY) {
1978 struct inet6_ifaddr *ifpub;
1979 spin_lock_bh(&ifp->lock);
1982 in6_ifa_hold(ifpub);
1983 spin_unlock_bh(&ifp->lock);
1984 ipv6_create_tempaddr(ifpub, ifp, true);
1987 spin_unlock_bh(&ifp->lock);
1990 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1991 spin_lock_bh(&ifp->lock);
1992 addrconf_del_dad_work(ifp);
1993 ifp->flags |= IFA_F_TENTATIVE;
1995 ifp->flags &= ~IFA_F_OPTIMISTIC;
1996 spin_unlock_bh(&ifp->lock);
1998 ipv6_ifa_notify(0, ifp);
2005 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2009 spin_lock_bh(&ifp->lock);
2010 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2011 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2014 spin_unlock_bh(&ifp->lock);
2019 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2021 struct inet6_dev *idev = ifp->idev;
2022 struct net *net = dev_net(ifp->idev->dev);
2024 if (addrconf_dad_end(ifp)) {
2029 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2030 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2032 spin_lock_bh(&ifp->lock);
2034 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2035 int scope = ifp->scope;
2036 u32 flags = ifp->flags;
2037 struct in6_addr new_addr;
2038 struct inet6_ifaddr *ifp2;
2039 u32 valid_lft, preferred_lft;
2040 int pfxlen = ifp->prefix_len;
2041 int retries = ifp->stable_privacy_retry + 1;
2043 if (retries > net->ipv6.sysctl.idgen_retries) {
2044 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2045 ifp->idev->dev->name);
2049 new_addr = ifp->addr;
2050 if (ipv6_generate_stable_address(&new_addr, retries,
2054 valid_lft = ifp->valid_lft;
2055 preferred_lft = ifp->prefered_lft;
2057 spin_unlock_bh(&ifp->lock);
2059 if (idev->cnf.max_addresses &&
2060 ipv6_count_addresses(idev) >=
2061 idev->cnf.max_addresses)
2064 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2065 ifp->idev->dev->name);
2067 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2068 scope, flags, valid_lft,
2069 preferred_lft, false, NULL);
2073 spin_lock_bh(&ifp2->lock);
2074 ifp2->stable_privacy_retry = retries;
2075 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2076 spin_unlock_bh(&ifp2->lock);
2078 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2081 spin_lock_bh(&ifp->lock);
2085 /* transition from _POSTDAD to _ERRDAD */
2086 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2087 spin_unlock_bh(&ifp->lock);
2089 addrconf_mod_dad_work(ifp, 0);
2093 /* Join to solicited addr multicast group.
2094 * caller must hold RTNL */
2095 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2097 struct in6_addr maddr;
2099 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2102 addrconf_addr_solict_mult(addr, &maddr);
2103 ipv6_dev_mc_inc(dev, &maddr);
2106 /* caller must hold RTNL */
2107 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2109 struct in6_addr maddr;
2111 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2114 addrconf_addr_solict_mult(addr, &maddr);
2115 __ipv6_dev_mc_dec(idev, &maddr);
2118 /* caller must hold RTNL */
2119 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2121 struct in6_addr addr;
2123 if (ifp->prefix_len >= 127) /* RFC 6164 */
2125 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2126 if (ipv6_addr_any(&addr))
2128 __ipv6_dev_ac_inc(ifp->idev, &addr);
2131 /* caller must hold RTNL */
2132 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2134 struct in6_addr addr;
2136 if (ifp->prefix_len >= 127) /* RFC 6164 */
2138 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2139 if (ipv6_addr_any(&addr))
2141 __ipv6_dev_ac_dec(ifp->idev, &addr);
2144 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2146 switch (dev->addr_len) {
2148 memcpy(eui, dev->dev_addr, 3);
2151 memcpy(eui + 5, dev->dev_addr + 3, 3);
2153 case EUI64_ADDR_LEN:
2154 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2164 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2166 union fwnet_hwaddr *ha;
2168 if (dev->addr_len != FWNET_ALEN)
2171 ha = (union fwnet_hwaddr *)dev->dev_addr;
2173 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2178 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2180 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2181 if (dev->addr_len != ARCNET_ALEN)
2184 eui[7] = *(u8 *)dev->dev_addr;
2188 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2190 if (dev->addr_len != INFINIBAND_ALEN)
2192 memcpy(eui, dev->dev_addr + 12, 8);
2197 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2201 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2202 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2203 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2204 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2205 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2206 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2210 memcpy(eui + 4, &addr, 4);
2214 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2216 if (dev->priv_flags & IFF_ISATAP)
2217 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2221 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2223 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2226 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2228 memcpy(eui, dev->perm_addr, 3);
2229 memcpy(eui + 5, dev->perm_addr + 3, 3);
2236 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2238 switch (dev->type) {
2241 return addrconf_ifid_eui48(eui, dev);
2243 return addrconf_ifid_arcnet(eui, dev);
2244 case ARPHRD_INFINIBAND:
2245 return addrconf_ifid_infiniband(eui, dev);
2247 return addrconf_ifid_sit(eui, dev);
2250 return addrconf_ifid_gre(eui, dev);
2251 case ARPHRD_6LOWPAN:
2252 return addrconf_ifid_6lowpan(eui, dev);
2253 case ARPHRD_IEEE1394:
2254 return addrconf_ifid_ieee1394(eui, dev);
2255 case ARPHRD_TUNNEL6:
2257 return addrconf_ifid_ip6tnl(eui, dev);
2262 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2265 struct inet6_ifaddr *ifp;
2267 read_lock_bh(&idev->lock);
2268 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2269 if (ifp->scope > IFA_LINK)
2271 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2272 memcpy(eui, ifp->addr.s6_addr+8, 8);
2277 read_unlock_bh(&idev->lock);
2281 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2282 static void ipv6_regen_rndid(struct inet6_dev *idev)
2285 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2286 idev->rndid[0] &= ~0x02;
2289 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2290 * check if generated address is not inappropriate
2292 * - Reserved subnet anycast (RFC 2526)
2293 * 11111101 11....11 1xxxxxxx
2294 * - ISATAP (RFC4214) 6.1
2295 * 00-00-5E-FE-xx-xx-xx-xx
2297 * - XXX: already assigned to an address on the device
2299 if (idev->rndid[0] == 0xfd &&
2300 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2301 (idev->rndid[7]&0x80))
2303 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2304 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2306 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2311 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2313 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2314 ipv6_regen_rndid(idev);
2322 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2323 unsigned long expires, u32 flags)
2325 struct fib6_config cfg = {
2326 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2327 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2328 .fc_ifindex = dev->ifindex,
2329 .fc_expires = expires,
2331 .fc_flags = RTF_UP | flags,
2332 .fc_nlinfo.nl_net = dev_net(dev),
2333 .fc_protocol = RTPROT_KERNEL,
2338 /* Prevent useless cloning on PtP SIT.
2339 This thing is done here expecting that the whole
2340 class of non-broadcast devices need not cloning.
2342 #if IS_ENABLED(CONFIG_IPV6_SIT)
2343 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2344 cfg.fc_flags |= RTF_NONEXTHOP;
2347 ip6_route_add(&cfg, NULL);
2351 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2353 const struct net_device *dev,
2354 u32 flags, u32 noflags)
2356 struct fib6_node *fn;
2357 struct rt6_info *rt = NULL;
2358 struct fib6_table *table;
2359 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2361 table = fib6_get_table(dev_net(dev), tb_id);
2366 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2370 for_each_fib6_node_rt_rcu(fn) {
2371 if (rt->dst.dev->ifindex != dev->ifindex)
2373 if ((rt->rt6i_flags & flags) != flags)
2375 if ((rt->rt6i_flags & noflags) != 0)
2377 if (!dst_hold_safe(&rt->dst))
2387 /* Create "default" multicast route to the interface */
2389 static void addrconf_add_mroute(struct net_device *dev)
2391 struct fib6_config cfg = {
2392 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2393 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2394 .fc_ifindex = dev->ifindex,
2397 .fc_nlinfo.nl_net = dev_net(dev),
2400 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2402 ip6_route_add(&cfg, NULL);
2405 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2407 struct inet6_dev *idev;
2411 idev = ipv6_find_idev(dev);
2413 return ERR_PTR(-ENOBUFS);
2415 if (idev->cnf.disable_ipv6)
2416 return ERR_PTR(-EACCES);
2418 /* Add default multicast route */
2419 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2420 addrconf_add_mroute(dev);
2425 static void manage_tempaddrs(struct inet6_dev *idev,
2426 struct inet6_ifaddr *ifp,
2427 __u32 valid_lft, __u32 prefered_lft,
2428 bool create, unsigned long now)
2431 struct inet6_ifaddr *ift;
2433 read_lock_bh(&idev->lock);
2434 /* update all temporary addresses in the list */
2435 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2436 int age, max_valid, max_prefered;
2438 if (ifp != ift->ifpub)
2441 /* RFC 4941 section 3.3:
2442 * If a received option will extend the lifetime of a public
2443 * address, the lifetimes of temporary addresses should
2444 * be extended, subject to the overall constraint that no
2445 * temporary addresses should ever remain "valid" or "preferred"
2446 * for a time longer than (TEMP_VALID_LIFETIME) or
2447 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2449 age = (now - ift->cstamp) / HZ;
2450 max_valid = idev->cnf.temp_valid_lft - age;
2454 max_prefered = idev->cnf.temp_prefered_lft -
2455 idev->desync_factor - age;
2456 if (max_prefered < 0)
2459 if (valid_lft > max_valid)
2460 valid_lft = max_valid;
2462 if (prefered_lft > max_prefered)
2463 prefered_lft = max_prefered;
2465 spin_lock(&ift->lock);
2467 ift->valid_lft = valid_lft;
2468 ift->prefered_lft = prefered_lft;
2470 if (prefered_lft > 0)
2471 ift->flags &= ~IFA_F_DEPRECATED;
2473 spin_unlock(&ift->lock);
2474 if (!(flags&IFA_F_TENTATIVE))
2475 ipv6_ifa_notify(0, ift);
2478 if ((create || list_empty(&idev->tempaddr_list)) &&
2479 idev->cnf.use_tempaddr > 0) {
2480 /* When a new public address is created as described
2481 * in [ADDRCONF], also create a new temporary address.
2482 * Also create a temporary address if it's enabled but
2483 * no temporary address currently exists.
2485 read_unlock_bh(&idev->lock);
2486 ipv6_create_tempaddr(ifp, NULL, false);
2488 read_unlock_bh(&idev->lock);
2492 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2494 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2495 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2498 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2499 const struct prefix_info *pinfo,
2500 struct inet6_dev *in6_dev,
2501 const struct in6_addr *addr, int addr_type,
2502 u32 addr_flags, bool sllao, bool tokenized,
2503 __u32 valid_lft, u32 prefered_lft)
2505 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2506 int create = 0, update_lft = 0;
2508 if (!ifp && valid_lft) {
2509 int max_addresses = in6_dev->cnf.max_addresses;
2511 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2512 if ((net->ipv6.devconf_all->optimistic_dad ||
2513 in6_dev->cnf.optimistic_dad) &&
2514 !net->ipv6.devconf_all->forwarding && sllao)
2515 addr_flags |= IFA_F_OPTIMISTIC;
2518 /* Do not allow to create too much of autoconfigured
2519 * addresses; this would be too easy way to crash kernel.
2521 if (!max_addresses ||
2522 ipv6_count_addresses(in6_dev) < max_addresses)
2523 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2525 addr_type&IPV6_ADDR_SCOPE_MASK,
2526 addr_flags, valid_lft,
2527 prefered_lft, false, NULL);
2529 if (IS_ERR_OR_NULL(ifp))
2533 spin_lock_bh(&ifp->lock);
2534 ifp->flags |= IFA_F_MANAGETEMPADDR;
2535 ifp->cstamp = jiffies;
2536 ifp->tokenized = tokenized;
2537 spin_unlock_bh(&ifp->lock);
2538 addrconf_dad_start(ifp);
2546 /* update lifetime (RFC2462 5.5.3 e) */
2547 spin_lock_bh(&ifp->lock);
2549 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2550 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2553 if (!create && stored_lft) {
2554 const u32 minimum_lft = min_t(u32,
2555 stored_lft, MIN_VALID_LIFETIME);
2556 valid_lft = max(valid_lft, minimum_lft);
2558 /* RFC4862 Section 5.5.3e:
2559 * "Note that the preferred lifetime of the
2560 * corresponding address is always reset to
2561 * the Preferred Lifetime in the received
2562 * Prefix Information option, regardless of
2563 * whether the valid lifetime is also reset or
2566 * So we should always update prefered_lft here.
2572 ifp->valid_lft = valid_lft;
2573 ifp->prefered_lft = prefered_lft;
2576 ifp->flags &= ~IFA_F_DEPRECATED;
2577 spin_unlock_bh(&ifp->lock);
2579 if (!(flags&IFA_F_TENTATIVE))
2580 ipv6_ifa_notify(0, ifp);
2582 spin_unlock_bh(&ifp->lock);
2584 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2593 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2595 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2597 struct prefix_info *pinfo;
2602 struct inet6_dev *in6_dev;
2603 struct net *net = dev_net(dev);
2605 pinfo = (struct prefix_info *) opt;
2607 if (len < sizeof(struct prefix_info)) {
2608 netdev_dbg(dev, "addrconf: prefix option too short\n");
2613 * Validation checks ([ADDRCONF], page 19)
2616 addr_type = ipv6_addr_type(&pinfo->prefix);
2618 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2621 valid_lft = ntohl(pinfo->valid);
2622 prefered_lft = ntohl(pinfo->prefered);
2624 if (prefered_lft > valid_lft) {
2625 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2629 in6_dev = in6_dev_get(dev);
2632 net_dbg_ratelimited("addrconf: device %s not configured\n",
2638 * Two things going on here:
2639 * 1) Add routes for on-link prefixes
2640 * 2) Configure prefixes with the auto flag set
2643 if (pinfo->onlink) {
2644 struct rt6_info *rt;
2645 unsigned long rt_expires;
2647 /* Avoid arithmetic overflow. Really, we could
2648 * save rt_expires in seconds, likely valid_lft,
2649 * but it would require division in fib gc, that it
2653 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2655 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2657 if (addrconf_finite_timeout(rt_expires))
2660 rt = addrconf_get_prefix_route(&pinfo->prefix,
2663 RTF_ADDRCONF | RTF_PREFIX_RT,
2664 RTF_GATEWAY | RTF_DEFAULT);
2667 /* Autoconf prefix route */
2668 if (valid_lft == 0) {
2671 } else if (addrconf_finite_timeout(rt_expires)) {
2673 rt6_set_expires(rt, jiffies + rt_expires);
2675 rt6_clean_expires(rt);
2677 } else if (valid_lft) {
2678 clock_t expires = 0;
2679 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2680 if (addrconf_finite_timeout(rt_expires)) {
2682 flags |= RTF_EXPIRES;
2683 expires = jiffies_to_clock_t(rt_expires);
2685 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2686 dev, expires, flags);
2691 /* Try to figure out our local address for this prefix */
2693 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2694 struct in6_addr addr;
2695 bool tokenized = false, dev_addr_generated = false;
2697 if (pinfo->prefix_len == 64) {
2698 memcpy(&addr, &pinfo->prefix, 8);
2700 if (!ipv6_addr_any(&in6_dev->token)) {
2701 read_lock_bh(&in6_dev->lock);
2702 memcpy(addr.s6_addr + 8,
2703 in6_dev->token.s6_addr + 8, 8);
2704 read_unlock_bh(&in6_dev->lock);
2706 } else if (is_addr_mode_generate_stable(in6_dev) &&
2707 !ipv6_generate_stable_address(&addr, 0,
2709 addr_flags |= IFA_F_STABLE_PRIVACY;
2711 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2712 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2715 dev_addr_generated = true;
2719 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2724 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2727 tokenized, valid_lft,
2732 /* Ignore error case here because previous prefix add addr was
2733 * successful which will be notified.
2735 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2736 addr_type, addr_flags, sllao,
2737 tokenized, valid_lft,
2739 dev_addr_generated);
2741 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2743 in6_dev_put(in6_dev);
2747 * Set destination address.
2748 * Special case for SIT interfaces where we create a new "virtual"
2751 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2753 struct in6_ifreq ireq;
2754 struct net_device *dev;
2760 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2763 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2769 #if IS_ENABLED(CONFIG_IPV6_SIT)
2770 if (dev->type == ARPHRD_SIT) {
2771 const struct net_device_ops *ops = dev->netdev_ops;
2773 struct ip_tunnel_parm p;
2775 err = -EADDRNOTAVAIL;
2776 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2779 memset(&p, 0, sizeof(p));
2780 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2784 p.iph.protocol = IPPROTO_IPV6;
2786 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2788 if (ops->ndo_do_ioctl) {
2789 mm_segment_t oldfs = get_fs();
2792 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2799 dev = __dev_get_by_name(net, p.name);
2802 err = dev_open(dev);
2812 static int ipv6_mc_config(struct sock *sk, bool join,
2813 const struct in6_addr *addr, int ifindex)
2821 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2823 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2830 * Manual configuration of address on an interface
2832 static int inet6_addr_add(struct net *net, int ifindex,
2833 const struct in6_addr *pfx,
2834 const struct in6_addr *peer_pfx,
2835 unsigned int plen, __u32 ifa_flags,
2836 __u32 prefered_lft, __u32 valid_lft,
2837 struct netlink_ext_ack *extack)
2839 struct inet6_ifaddr *ifp;
2840 struct inet6_dev *idev;
2841 struct net_device *dev;
2842 unsigned long timeout;
2852 /* check the lifetime */
2853 if (!valid_lft || prefered_lft > valid_lft)
2856 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2859 dev = __dev_get_by_index(net, ifindex);
2863 idev = addrconf_add_dev(dev);
2865 return PTR_ERR(idev);
2867 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2868 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2869 true, pfx, ifindex);
2875 scope = ipv6_addr_scope(pfx);
2877 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2878 if (addrconf_finite_timeout(timeout)) {
2879 expires = jiffies_to_clock_t(timeout * HZ);
2880 valid_lft = timeout;
2881 flags = RTF_EXPIRES;
2885 ifa_flags |= IFA_F_PERMANENT;
2888 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2889 if (addrconf_finite_timeout(timeout)) {
2891 ifa_flags |= IFA_F_DEPRECATED;
2892 prefered_lft = timeout;
2895 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2896 valid_lft, prefered_lft, true, extack);
2899 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2900 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2904 /* Send a netlink notification if DAD is enabled and
2905 * optimistic flag is not set
2907 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2908 ipv6_ifa_notify(0, ifp);
2910 * Note that section 3.1 of RFC 4429 indicates
2911 * that the Optimistic flag should not be set for
2912 * manually configured addresses
2914 addrconf_dad_start(ifp);
2915 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2916 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2919 addrconf_verify_rtnl();
2921 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2922 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2923 false, pfx, ifindex);
2926 return PTR_ERR(ifp);
2929 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2930 const struct in6_addr *pfx, unsigned int plen)
2932 struct inet6_ifaddr *ifp;
2933 struct inet6_dev *idev;
2934 struct net_device *dev;
2939 dev = __dev_get_by_index(net, ifindex);
2943 idev = __in6_dev_get(dev);
2947 read_lock_bh(&idev->lock);
2948 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2949 if (ifp->prefix_len == plen &&
2950 ipv6_addr_equal(pfx, &ifp->addr)) {
2952 read_unlock_bh(&idev->lock);
2954 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2955 (ifa_flags & IFA_F_MANAGETEMPADDR))
2956 manage_tempaddrs(idev, ifp, 0, 0, false,
2959 addrconf_verify_rtnl();
2960 if (ipv6_addr_is_multicast(pfx)) {
2961 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2962 false, pfx, dev->ifindex);
2967 read_unlock_bh(&idev->lock);
2968 return -EADDRNOTAVAIL;
2972 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2974 struct in6_ifreq ireq;
2977 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2980 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2984 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2985 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2986 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2991 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2993 struct in6_ifreq ireq;
2996 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2999 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3003 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3004 ireq.ifr6_prefixlen);
3009 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3010 int plen, int scope)
3012 struct inet6_ifaddr *ifp;
3014 ifp = ipv6_add_addr(idev, addr, NULL, plen,
3015 scope, IFA_F_PERMANENT,
3016 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
3019 spin_lock_bh(&ifp->lock);
3020 ifp->flags &= ~IFA_F_TENTATIVE;
3021 spin_unlock_bh(&ifp->lock);
3022 rt_genid_bump_ipv6(dev_net(idev->dev));
3023 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3028 #if IS_ENABLED(CONFIG_IPV6_SIT)
3029 static void sit_add_v4_addrs(struct inet6_dev *idev)
3031 struct in6_addr addr;
3032 struct net_device *dev;
3033 struct net *net = dev_net(idev->dev);
3039 memset(&addr, 0, sizeof(struct in6_addr));
3040 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3042 if (idev->dev->flags&IFF_POINTOPOINT) {
3043 addr.s6_addr32[0] = htonl(0xfe800000);
3047 scope = IPV6_ADDR_COMPATv4;
3049 pflags |= RTF_NONEXTHOP;
3052 if (addr.s6_addr32[3]) {
3053 add_addr(idev, &addr, plen, scope);
3054 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3058 for_each_netdev(net, dev) {
3059 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3060 if (in_dev && (dev->flags & IFF_UP)) {
3061 struct in_ifaddr *ifa;
3065 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3067 addr.s6_addr32[3] = ifa->ifa_local;
3069 if (ifa->ifa_scope == RT_SCOPE_LINK)
3071 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3072 if (idev->dev->flags&IFF_POINTOPOINT)
3077 add_addr(idev, &addr, plen, flag);
3078 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3086 static void init_loopback(struct net_device *dev)
3088 struct inet6_dev *idev;
3094 idev = ipv6_find_idev(dev);
3096 pr_debug("%s: add_dev failed\n", __func__);
3100 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3103 void addrconf_add_linklocal(struct inet6_dev *idev,
3104 const struct in6_addr *addr, u32 flags)
3106 struct inet6_ifaddr *ifp;
3107 u32 addr_flags = flags | IFA_F_PERMANENT;
3109 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3110 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3111 idev->cnf.optimistic_dad) &&
3112 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3113 addr_flags |= IFA_F_OPTIMISTIC;
3116 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3117 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3119 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3120 addrconf_dad_start(ifp);
3124 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3126 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3128 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3131 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3132 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3135 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3136 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3142 static int ipv6_generate_stable_address(struct in6_addr *address,
3144 const struct inet6_dev *idev)
3146 static DEFINE_SPINLOCK(lock);
3147 static __u32 digest[SHA_DIGEST_WORDS];
3148 static __u32 workspace[SHA_WORKSPACE_WORDS];
3151 char __data[SHA_MESSAGE_BYTES];
3153 struct in6_addr secret;
3155 unsigned char hwaddr[MAX_ADDR_LEN];
3160 struct in6_addr secret;
3161 struct in6_addr temp;
3162 struct net *net = dev_net(idev->dev);
3164 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3166 if (idev->cnf.stable_secret.initialized)
3167 secret = idev->cnf.stable_secret.secret;
3168 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3169 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3174 spin_lock_bh(&lock);
3177 memset(&data, 0, sizeof(data));
3178 memset(workspace, 0, sizeof(workspace));
3179 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3180 data.prefix[0] = address->s6_addr32[0];
3181 data.prefix[1] = address->s6_addr32[1];
3182 data.secret = secret;
3183 data.dad_count = dad_count;
3185 sha_transform(digest, data.__data, workspace);
3188 temp.s6_addr32[2] = (__force __be32)digest[0];
3189 temp.s6_addr32[3] = (__force __be32)digest[1];
3191 spin_unlock_bh(&lock);
3193 if (ipv6_reserved_interfaceid(temp)) {
3195 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3204 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3206 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3210 s = &idev->cnf.stable_secret;
3211 get_random_bytes(&s->secret, sizeof(s->secret));
3212 s->initialized = true;
3215 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3217 struct in6_addr addr;
3219 /* no link local addresses on L3 master devices */
3220 if (netif_is_l3_master(idev->dev))
3223 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3225 switch (idev->cnf.addr_gen_mode) {
3226 case IN6_ADDR_GEN_MODE_RANDOM:
3227 ipv6_gen_mode_random_init(idev);
3229 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3230 if (!ipv6_generate_stable_address(&addr, 0, idev))
3231 addrconf_add_linklocal(idev, &addr,
3232 IFA_F_STABLE_PRIVACY);
3233 else if (prefix_route)
3234 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3236 case IN6_ADDR_GEN_MODE_EUI64:
3237 /* addrconf_add_linklocal also adds a prefix_route and we
3238 * only need to care about prefix routes if ipv6_generate_eui64
3239 * couldn't generate one.
3241 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3242 addrconf_add_linklocal(idev, &addr, 0);
3243 else if (prefix_route)
3244 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3246 case IN6_ADDR_GEN_MODE_NONE:
3248 /* will not add any link local address */
3253 static void addrconf_dev_config(struct net_device *dev)
3255 struct inet6_dev *idev;
3259 if ((dev->type != ARPHRD_ETHER) &&
3260 (dev->type != ARPHRD_FDDI) &&
3261 (dev->type != ARPHRD_ARCNET) &&
3262 (dev->type != ARPHRD_INFINIBAND) &&
3263 (dev->type != ARPHRD_IEEE1394) &&
3264 (dev->type != ARPHRD_TUNNEL6) &&
3265 (dev->type != ARPHRD_6LOWPAN) &&
3266 (dev->type != ARPHRD_IP6GRE) &&
3267 (dev->type != ARPHRD_IPGRE) &&
3268 (dev->type != ARPHRD_TUNNEL) &&
3269 (dev->type != ARPHRD_NONE)) {
3270 /* Alas, we support only Ethernet autoconfiguration. */
3274 idev = addrconf_add_dev(dev);
3278 /* this device type has no EUI support */
3279 if (dev->type == ARPHRD_NONE &&
3280 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3281 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3283 addrconf_addr_gen(idev, false);
3286 #if IS_ENABLED(CONFIG_IPV6_SIT)
3287 static void addrconf_sit_config(struct net_device *dev)
3289 struct inet6_dev *idev;
3294 * Configure the tunnel with one of our IPv4
3295 * addresses... we should configure all of
3296 * our v4 addrs in the tunnel
3299 idev = ipv6_find_idev(dev);
3301 pr_debug("%s: add_dev failed\n", __func__);
3305 if (dev->priv_flags & IFF_ISATAP) {
3306 addrconf_addr_gen(idev, false);
3310 sit_add_v4_addrs(idev);
3312 if (dev->flags&IFF_POINTOPOINT)
3313 addrconf_add_mroute(dev);
3317 #if IS_ENABLED(CONFIG_NET_IPGRE)
3318 static void addrconf_gre_config(struct net_device *dev)
3320 struct inet6_dev *idev;
3324 idev = ipv6_find_idev(dev);
3326 pr_debug("%s: add_dev failed\n", __func__);
3330 addrconf_addr_gen(idev, true);
3331 if (dev->flags & IFF_POINTOPOINT)
3332 addrconf_add_mroute(dev);
3336 static int fixup_permanent_addr(struct inet6_dev *idev,
3337 struct inet6_ifaddr *ifp)
3339 /* !rt6i_node means the host route was removed from the
3340 * FIB, for example, if 'lo' device is taken down. In that
3341 * case regenerate the host route.
3343 if (!ifp->rt || !ifp->rt->rt6i_node) {
3344 struct rt6_info *rt, *prev;
3346 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3350 /* ifp->rt can be accessed outside of rtnl */
3351 spin_lock(&ifp->lock);
3354 spin_unlock(&ifp->lock);
3359 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3360 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3364 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3365 addrconf_dad_start(ifp);
3370 static void addrconf_permanent_addr(struct net_device *dev)
3372 struct inet6_ifaddr *ifp, *tmp;
3373 struct inet6_dev *idev;
3375 idev = __in6_dev_get(dev);
3379 write_lock_bh(&idev->lock);
3381 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3382 if ((ifp->flags & IFA_F_PERMANENT) &&
3383 fixup_permanent_addr(idev, ifp) < 0) {
3384 write_unlock_bh(&idev->lock);
3387 write_lock_bh(&idev->lock);
3389 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3390 idev->dev->name, &ifp->addr);
3394 write_unlock_bh(&idev->lock);
3397 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3400 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3401 struct netdev_notifier_changeupper_info *info;
3402 struct inet6_dev *idev = __in6_dev_get(dev);
3403 struct net *net = dev_net(dev);
3404 int run_pending = 0;
3408 case NETDEV_REGISTER:
3409 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3410 idev = ipv6_add_dev(dev);
3412 return notifier_from_errno(PTR_ERR(idev));
3416 case NETDEV_CHANGEMTU:
3417 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3418 if (dev->mtu < IPV6_MIN_MTU) {
3419 addrconf_ifdown(dev, dev != net->loopback_dev);
3424 rt6_mtu_change(dev, dev->mtu);
3425 idev->cnf.mtu6 = dev->mtu;
3429 /* allocate new idev */
3430 idev = ipv6_add_dev(dev);
3434 /* device is still not ready */
3435 if (!(idev->if_flags & IF_READY))
3444 if (dev->flags & IFF_SLAVE)
3447 if (idev && idev->cnf.disable_ipv6)
3450 if (event == NETDEV_UP) {
3451 /* restore routes for permanent addresses */
3452 addrconf_permanent_addr(dev);
3454 if (!addrconf_link_ready(dev)) {
3455 /* device is not ready yet. */
3456 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3461 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3462 idev = ipv6_add_dev(dev);
3464 if (!IS_ERR_OR_NULL(idev)) {
3465 idev->if_flags |= IF_READY;
3468 } else if (event == NETDEV_CHANGE) {
3469 if (!addrconf_link_ready(dev)) {
3470 /* device is still not ready. */
3471 rt6_sync_down_dev(dev, event);
3476 if (idev->if_flags & IF_READY) {
3477 /* device is already configured -
3478 * but resend MLD reports, we might
3479 * have roamed and need to update
3480 * multicast snooping switches
3483 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3486 idev->if_flags |= IF_READY;
3489 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3495 switch (dev->type) {
3496 #if IS_ENABLED(CONFIG_IPV6_SIT)
3498 addrconf_sit_config(dev);
3501 #if IS_ENABLED(CONFIG_NET_IPGRE)
3503 addrconf_gre_config(dev);
3506 case ARPHRD_LOOPBACK:
3511 addrconf_dev_config(dev);
3515 if (!IS_ERR_OR_NULL(idev)) {
3517 addrconf_dad_run(idev);
3519 /* Device has an address by now */
3520 rt6_sync_up(dev, RTNH_F_DEAD);
3523 * If the MTU changed during the interface down,
3524 * when the interface up, the changed MTU must be
3525 * reflected in the idev as well as routers.
3527 if (idev->cnf.mtu6 != dev->mtu &&
3528 dev->mtu >= IPV6_MIN_MTU) {
3529 rt6_mtu_change(dev, dev->mtu);
3530 idev->cnf.mtu6 = dev->mtu;
3532 idev->tstamp = jiffies;
3533 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3536 * If the changed mtu during down is lower than
3537 * IPV6_MIN_MTU stop IPv6 on this interface.
3539 if (dev->mtu < IPV6_MIN_MTU)
3540 addrconf_ifdown(dev, dev != net->loopback_dev);
3545 case NETDEV_UNREGISTER:
3547 * Remove all addresses from this interface.
3549 addrconf_ifdown(dev, event != NETDEV_DOWN);
3552 case NETDEV_CHANGENAME:
3554 snmp6_unregister_dev(idev);
3555 addrconf_sysctl_unregister(idev);
3556 err = addrconf_sysctl_register(idev);
3558 return notifier_from_errno(err);
3559 err = snmp6_register_dev(idev);
3561 addrconf_sysctl_unregister(idev);
3562 return notifier_from_errno(err);
3567 case NETDEV_PRE_TYPE_CHANGE:
3568 case NETDEV_POST_TYPE_CHANGE:
3570 addrconf_type_change(dev, event);
3573 case NETDEV_CHANGEUPPER:
3576 /* flush all routes if dev is linked to or unlinked from
3577 * an L3 master device (e.g., VRF)
3579 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3580 addrconf_ifdown(dev, 0);
3587 * addrconf module should be notified of a device going up
3589 static struct notifier_block ipv6_dev_notf = {
3590 .notifier_call = addrconf_notify,
3591 .priority = ADDRCONF_NOTIFY_PRIORITY,
3594 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3596 struct inet6_dev *idev;
3599 idev = __in6_dev_get(dev);
3601 if (event == NETDEV_POST_TYPE_CHANGE)
3602 ipv6_mc_remap(idev);
3603 else if (event == NETDEV_PRE_TYPE_CHANGE)
3604 ipv6_mc_unmap(idev);
3607 static bool addr_is_local(const struct in6_addr *addr)
3609 return ipv6_addr_type(addr) &
3610 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3613 static int addrconf_ifdown(struct net_device *dev, int how)
3615 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3616 struct net *net = dev_net(dev);
3617 struct inet6_dev *idev;
3618 struct inet6_ifaddr *ifa, *tmp;
3625 rt6_disable_ip(dev, event);
3627 idev = __in6_dev_get(dev);
3632 * Step 1: remove reference to ipv6 device from parent device.
3638 /* protected by rtnl_lock */
3639 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3641 /* Step 1.5: remove snmp6 entry */
3642 snmp6_unregister_dev(idev);
3646 /* aggregate the system setting and interface setting */
3647 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3649 _keep_addr = idev->cnf.keep_addr_on_down;
3651 /* combine the user config with event to determine if permanent
3652 * addresses are to be removed from address hash table
3654 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3656 /* Step 2: clear hash table */
3657 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3658 struct hlist_head *h = &inet6_addr_lst[i];
3660 spin_lock_bh(&addrconf_hash_lock);
3662 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3663 if (ifa->idev == idev) {
3664 addrconf_del_dad_work(ifa);
3665 /* combined flag + permanent flag decide if
3666 * address is retained on a down event
3669 !(ifa->flags & IFA_F_PERMANENT) ||
3670 addr_is_local(&ifa->addr)) {
3671 hlist_del_init_rcu(&ifa->addr_lst);
3676 spin_unlock_bh(&addrconf_hash_lock);
3679 write_lock_bh(&idev->lock);
3681 addrconf_del_rs_timer(idev);
3683 /* Step 2: clear flags for stateless addrconf */
3685 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3687 /* Step 3: clear tempaddr list */
3688 while (!list_empty(&idev->tempaddr_list)) {
3689 ifa = list_first_entry(&idev->tempaddr_list,
3690 struct inet6_ifaddr, tmp_list);
3691 list_del(&ifa->tmp_list);
3692 write_unlock_bh(&idev->lock);
3693 spin_lock_bh(&ifa->lock);
3696 in6_ifa_put(ifa->ifpub);
3699 spin_unlock_bh(&ifa->lock);
3701 write_lock_bh(&idev->lock);
3704 /* re-combine the user config with event to determine if permanent
3705 * addresses are to be removed from the interface list
3707 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3709 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3710 struct rt6_info *rt = NULL;
3713 addrconf_del_dad_work(ifa);
3715 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3716 !addr_is_local(&ifa->addr);
3718 write_unlock_bh(&idev->lock);
3719 spin_lock_bh(&ifa->lock);
3722 /* set state to skip the notifier below */
3723 state = INET6_IFADDR_STATE_DEAD;
3724 ifa->state = INET6_IFADDR_STATE_PREDAD;
3725 if (!(ifa->flags & IFA_F_NODAD))
3726 ifa->flags |= IFA_F_TENTATIVE;
3732 ifa->state = INET6_IFADDR_STATE_DEAD;
3735 spin_unlock_bh(&ifa->lock);
3740 if (state != INET6_IFADDR_STATE_DEAD) {
3741 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3742 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3744 if (idev->cnf.forwarding)
3745 addrconf_leave_anycast(ifa);
3746 addrconf_leave_solict(ifa->idev, &ifa->addr);
3749 write_lock_bh(&idev->lock);
3751 list_del_rcu(&ifa->if_list);
3756 write_unlock_bh(&idev->lock);
3758 /* Step 5: Discard anycast and multicast list */
3760 ipv6_ac_destroy_dev(idev);
3761 ipv6_mc_destroy_dev(idev);
3766 idev->tstamp = jiffies;
3768 /* Last: Shot the device (if unregistered) */
3770 addrconf_sysctl_unregister(idev);
3771 neigh_parms_release(&nd_tbl, idev->nd_parms);
3772 neigh_ifdown(&nd_tbl, dev);
3778 static void addrconf_rs_timer(struct timer_list *t)
3780 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3781 struct net_device *dev = idev->dev;
3782 struct in6_addr lladdr;
3784 write_lock(&idev->lock);
3785 if (idev->dead || !(idev->if_flags & IF_READY))
3788 if (!ipv6_accept_ra(idev))
3791 /* Announcement received after solicitation was sent */
3792 if (idev->if_flags & IF_RA_RCVD)
3795 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3796 write_unlock(&idev->lock);
3797 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3798 ndisc_send_rs(dev, &lladdr,
3799 &in6addr_linklocal_allrouters);
3803 write_lock(&idev->lock);
3804 idev->rs_interval = rfc3315_s14_backoff_update(
3805 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3806 /* The wait after the last probe can be shorter */
3807 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3808 idev->cnf.rtr_solicits) ?
3809 idev->cnf.rtr_solicit_delay :
3813 * Note: we do not support deprecated "all on-link"
3814 * assumption any longer.
3816 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3820 write_unlock(&idev->lock);
3826 * Duplicate Address Detection
3828 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3830 unsigned long rand_num;
3831 struct inet6_dev *idev = ifp->idev;
3834 if (ifp->flags & IFA_F_OPTIMISTIC)
3837 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3840 if (idev->cnf.enhanced_dad ||
3841 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3843 get_random_bytes(&nonce, 6);
3846 ifp->dad_nonce = nonce;
3847 ifp->dad_probes = idev->cnf.dad_transmits;
3848 addrconf_mod_dad_work(ifp, rand_num);
3851 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3853 struct inet6_dev *idev = ifp->idev;
3854 struct net_device *dev = idev->dev;
3855 bool bump_id, notify = false;
3857 addrconf_join_solict(dev, &ifp->addr);
3859 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3861 read_lock_bh(&idev->lock);
3862 spin_lock(&ifp->lock);
3863 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3866 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3867 (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3868 idev->cnf.accept_dad < 1) ||
3869 !(ifp->flags&IFA_F_TENTATIVE) ||
3870 ifp->flags & IFA_F_NODAD) {
3871 bool send_na = false;
3873 if (ifp->flags & IFA_F_TENTATIVE &&
3874 !(ifp->flags & IFA_F_OPTIMISTIC))
3876 bump_id = ifp->flags & IFA_F_TENTATIVE;
3877 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3878 spin_unlock(&ifp->lock);
3879 read_unlock_bh(&idev->lock);
3881 addrconf_dad_completed(ifp, bump_id, send_na);
3885 if (!(idev->if_flags & IF_READY)) {
3886 spin_unlock(&ifp->lock);
3887 read_unlock_bh(&idev->lock);
3889 * If the device is not ready:
3890 * - keep it tentative if it is a permanent address.
3891 * - otherwise, kill it.
3894 addrconf_dad_stop(ifp, 0);
3899 * Optimistic nodes can start receiving
3902 if (ifp->flags & IFA_F_OPTIMISTIC) {
3903 ip6_ins_rt(ifp->rt);
3904 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3905 /* Because optimistic nodes can use this address,
3906 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3912 addrconf_dad_kick(ifp);
3914 spin_unlock(&ifp->lock);
3915 read_unlock_bh(&idev->lock);
3917 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3920 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3922 bool begin_dad = false;
3924 spin_lock_bh(&ifp->lock);
3925 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3926 ifp->state = INET6_IFADDR_STATE_PREDAD;
3929 spin_unlock_bh(&ifp->lock);
3932 addrconf_mod_dad_work(ifp, 0);
3935 static void addrconf_dad_work(struct work_struct *w)
3937 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3938 struct inet6_ifaddr,
3940 struct inet6_dev *idev = ifp->idev;
3941 bool bump_id, disable_ipv6 = false;
3942 struct in6_addr mcaddr;
3948 } action = DAD_PROCESS;
3952 spin_lock_bh(&ifp->lock);
3953 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3955 ifp->state = INET6_IFADDR_STATE_DAD;
3956 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3958 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3960 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3961 idev->cnf.accept_dad > 1) &&
3962 !idev->cnf.disable_ipv6 &&
3963 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3964 struct in6_addr addr;
3966 addr.s6_addr32[0] = htonl(0xfe800000);
3967 addr.s6_addr32[1] = 0;
3969 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3970 ipv6_addr_equal(&ifp->addr, &addr)) {
3971 /* DAD failed for link-local based on MAC */
3972 idev->cnf.disable_ipv6 = 1;
3974 pr_info("%s: IPv6 being disabled!\n",
3975 ifp->idev->dev->name);
3976 disable_ipv6 = true;
3980 spin_unlock_bh(&ifp->lock);
3982 if (action == DAD_BEGIN) {
3983 addrconf_dad_begin(ifp);
3985 } else if (action == DAD_ABORT) {
3987 addrconf_dad_stop(ifp, 1);
3989 addrconf_ifdown(idev->dev, 0);
3993 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3996 write_lock_bh(&idev->lock);
3997 if (idev->dead || !(idev->if_flags & IF_READY)) {
3998 write_unlock_bh(&idev->lock);
4002 spin_lock(&ifp->lock);
4003 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4004 spin_unlock(&ifp->lock);
4005 write_unlock_bh(&idev->lock);
4009 if (ifp->dad_probes == 0) {
4010 bool send_na = false;
4013 * DAD was successful
4016 if (ifp->flags & IFA_F_TENTATIVE &&
4017 !(ifp->flags & IFA_F_OPTIMISTIC))
4019 bump_id = ifp->flags & IFA_F_TENTATIVE;
4020 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4021 spin_unlock(&ifp->lock);
4022 write_unlock_bh(&idev->lock);
4024 addrconf_dad_completed(ifp, bump_id, send_na);
4030 addrconf_mod_dad_work(ifp,
4031 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4032 spin_unlock(&ifp->lock);
4033 write_unlock_bh(&idev->lock);
4035 /* send a neighbour solicitation for our addr */
4036 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4037 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4044 /* ifp->idev must be at least read locked */
4045 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4047 struct inet6_ifaddr *ifpiter;
4048 struct inet6_dev *idev = ifp->idev;
4050 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4051 if (ifpiter->scope > IFA_LINK)
4053 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4054 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4055 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4062 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4065 struct net_device *dev = ifp->idev->dev;
4066 struct in6_addr lladdr;
4067 bool send_rs, send_mld;
4069 addrconf_del_dad_work(ifp);
4072 * Configure the address for reception. Now it is valid.
4075 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4077 /* If added prefix is link local and we are prepared to process
4078 router advertisements, start sending router solicitations.
4081 read_lock_bh(&ifp->idev->lock);
4082 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4083 send_rs = send_mld &&
4084 ipv6_accept_ra(ifp->idev) &&
4085 ifp->idev->cnf.rtr_solicits != 0 &&
4086 (dev->flags&IFF_LOOPBACK) == 0;
4087 read_unlock_bh(&ifp->idev->lock);
4089 /* While dad is in progress mld report's source address is in6_addrany.
4090 * Resend with proper ll now.
4093 ipv6_mc_dad_complete(ifp->idev);
4095 /* send unsolicited NA if enabled */
4097 (ifp->idev->cnf.ndisc_notify ||
4098 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4099 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4100 /*router=*/ !!ifp->idev->cnf.forwarding,
4101 /*solicited=*/ false, /*override=*/ true,
4107 * If a host as already performed a random delay
4108 * [...] as part of DAD [...] there is no need
4109 * to delay again before sending the first RS
4111 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4113 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4115 write_lock_bh(&ifp->idev->lock);
4116 spin_lock(&ifp->lock);
4117 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4118 ifp->idev->cnf.rtr_solicit_interval);
4119 ifp->idev->rs_probes = 1;
4120 ifp->idev->if_flags |= IF_RS_SENT;
4121 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4122 spin_unlock(&ifp->lock);
4123 write_unlock_bh(&ifp->idev->lock);
4127 rt_genid_bump_ipv6(dev_net(dev));
4129 /* Make sure that a new temporary address will be created
4130 * before this temporary address becomes deprecated.
4132 if (ifp->flags & IFA_F_TEMPORARY)
4133 addrconf_verify_rtnl();
4136 static void addrconf_dad_run(struct inet6_dev *idev)
4138 struct inet6_ifaddr *ifp;
4140 read_lock_bh(&idev->lock);
4141 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4142 spin_lock(&ifp->lock);
4143 if (ifp->flags & IFA_F_TENTATIVE &&
4144 ifp->state == INET6_IFADDR_STATE_DAD)
4145 addrconf_dad_kick(ifp);
4146 spin_unlock(&ifp->lock);
4148 read_unlock_bh(&idev->lock);
4151 #ifdef CONFIG_PROC_FS
4152 struct if6_iter_state {
4153 struct seq_net_private p;
4158 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4160 struct if6_iter_state *state = seq->private;
4161 struct net *net = seq_file_net(seq);
4162 struct inet6_ifaddr *ifa = NULL;
4165 /* initial bucket if pos is 0 */
4171 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4172 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4174 if (!net_eq(dev_net(ifa->idev->dev), net))
4176 /* sync with offset */
4177 if (p < state->offset) {
4185 /* prepare for next bucket */
4192 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4193 struct inet6_ifaddr *ifa)
4195 struct if6_iter_state *state = seq->private;
4196 struct net *net = seq_file_net(seq);
4198 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4199 if (!net_eq(dev_net(ifa->idev->dev), net))
4205 while (++state->bucket < IN6_ADDR_HSIZE) {
4207 hlist_for_each_entry_rcu(ifa,
4208 &inet6_addr_lst[state->bucket], addr_lst) {
4209 if (!net_eq(dev_net(ifa->idev->dev), net))
4219 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4223 return if6_get_first(seq, *pos);
4226 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4228 struct inet6_ifaddr *ifa;
4230 ifa = if6_get_next(seq, v);
4235 static void if6_seq_stop(struct seq_file *seq, void *v)
4241 static int if6_seq_show(struct seq_file *seq, void *v)
4243 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4244 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4246 ifp->idev->dev->ifindex,
4250 ifp->idev->dev->name);
4254 static const struct seq_operations if6_seq_ops = {
4255 .start = if6_seq_start,
4256 .next = if6_seq_next,
4257 .show = if6_seq_show,
4258 .stop = if6_seq_stop,
4261 static int if6_seq_open(struct inode *inode, struct file *file)
4263 return seq_open_net(inode, file, &if6_seq_ops,
4264 sizeof(struct if6_iter_state));
4267 static const struct file_operations if6_fops = {
4268 .open = if6_seq_open,
4270 .llseek = seq_lseek,
4271 .release = seq_release_net,
4274 static int __net_init if6_proc_net_init(struct net *net)
4276 if (!proc_create("if_inet6", 0444, net->proc_net, &if6_fops))
4281 static void __net_exit if6_proc_net_exit(struct net *net)
4283 remove_proc_entry("if_inet6", net->proc_net);
4286 static struct pernet_operations if6_proc_net_ops = {
4287 .init = if6_proc_net_init,
4288 .exit = if6_proc_net_exit,
4291 int __init if6_proc_init(void)
4293 return register_pernet_subsys(&if6_proc_net_ops);
4296 void if6_proc_exit(void)
4298 unregister_pernet_subsys(&if6_proc_net_ops);
4300 #endif /* CONFIG_PROC_FS */
4302 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4303 /* Check if address is a home address configured on any interface. */
4304 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4306 unsigned int hash = inet6_addr_hash(net, addr);
4307 struct inet6_ifaddr *ifp = NULL;
4311 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4312 if (!net_eq(dev_net(ifp->idev->dev), net))
4314 if (ipv6_addr_equal(&ifp->addr, addr) &&
4315 (ifp->flags & IFA_F_HOMEADDRESS)) {
4326 * Periodic address status verification
4329 static void addrconf_verify_rtnl(void)
4331 unsigned long now, next, next_sec, next_sched;
4332 struct inet6_ifaddr *ifp;
4339 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4341 cancel_delayed_work(&addr_chk_work);
4343 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4345 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4348 /* When setting preferred_lft to a value not zero or
4349 * infinity, while valid_lft is infinity
4350 * IFA_F_PERMANENT has a non-infinity life time.
4352 if ((ifp->flags & IFA_F_PERMANENT) &&
4353 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4356 spin_lock(&ifp->lock);
4357 /* We try to batch several events at once. */
4358 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4360 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4361 age >= ifp->valid_lft) {
4362 spin_unlock(&ifp->lock);
4366 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4367 spin_unlock(&ifp->lock);
4369 } else if (age >= ifp->prefered_lft) {
4370 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4373 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4375 ifp->flags |= IFA_F_DEPRECATED;
4378 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4379 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4380 next = ifp->tstamp + ifp->valid_lft * HZ;
4382 spin_unlock(&ifp->lock);
4387 ipv6_ifa_notify(0, ifp);
4391 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4392 !(ifp->flags&IFA_F_TENTATIVE)) {
4393 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4394 ifp->idev->cnf.dad_transmits *
4395 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4397 if (age >= ifp->prefered_lft - regen_advance) {
4398 struct inet6_ifaddr *ifpub = ifp->ifpub;
4399 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4400 next = ifp->tstamp + ifp->prefered_lft * HZ;
4401 if (!ifp->regen_count && ifpub) {
4404 in6_ifa_hold(ifpub);
4405 spin_unlock(&ifp->lock);
4407 spin_lock(&ifpub->lock);
4408 ifpub->regen_count = 0;
4409 spin_unlock(&ifpub->lock);
4410 rcu_read_unlock_bh();
4411 ipv6_create_tempaddr(ifpub, ifp, true);
4417 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4418 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4419 spin_unlock(&ifp->lock);
4421 /* ifp->prefered_lft <= ifp->valid_lft */
4422 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4423 next = ifp->tstamp + ifp->prefered_lft * HZ;
4424 spin_unlock(&ifp->lock);
4429 next_sec = round_jiffies_up(next);
4432 /* If rounded timeout is accurate enough, accept it. */
4433 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4434 next_sched = next_sec;
4436 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4437 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4438 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4440 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4441 now, next, next_sec, next_sched);
4442 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4443 rcu_read_unlock_bh();
4446 static void addrconf_verify_work(struct work_struct *w)
4449 addrconf_verify_rtnl();
4453 static void addrconf_verify(void)
4455 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4458 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4459 struct in6_addr **peer_pfx)
4461 struct in6_addr *pfx = NULL;
4466 pfx = nla_data(addr);
4469 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4471 pfx = nla_data(local);
4477 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4478 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4479 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4480 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4481 [IFA_FLAGS] = { .len = sizeof(u32) },
4485 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4486 struct netlink_ext_ack *extack)
4488 struct net *net = sock_net(skb->sk);
4489 struct ifaddrmsg *ifm;
4490 struct nlattr *tb[IFA_MAX+1];
4491 struct in6_addr *pfx, *peer_pfx;
4495 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4500 ifm = nlmsg_data(nlh);
4501 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4505 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4507 /* We ignore other flags so far. */
4508 ifa_flags &= IFA_F_MANAGETEMPADDR;
4510 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4511 ifm->ifa_prefixlen);
4514 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4515 u32 prefered_lft, u32 valid_lft)
4519 unsigned long timeout;
4520 bool was_managetempaddr;
4521 bool had_prefixroute;
4525 if (!valid_lft || (prefered_lft > valid_lft))
4528 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4529 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4532 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4533 ifa_flags &= ~IFA_F_OPTIMISTIC;
4535 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4536 if (addrconf_finite_timeout(timeout)) {
4537 expires = jiffies_to_clock_t(timeout * HZ);
4538 valid_lft = timeout;
4539 flags = RTF_EXPIRES;
4543 ifa_flags |= IFA_F_PERMANENT;
4546 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4547 if (addrconf_finite_timeout(timeout)) {
4549 ifa_flags |= IFA_F_DEPRECATED;
4550 prefered_lft = timeout;
4553 spin_lock_bh(&ifp->lock);
4554 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4555 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4556 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4557 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4558 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4559 IFA_F_NOPREFIXROUTE);
4560 ifp->flags |= ifa_flags;
4561 ifp->tstamp = jiffies;
4562 ifp->valid_lft = valid_lft;
4563 ifp->prefered_lft = prefered_lft;
4565 spin_unlock_bh(&ifp->lock);
4566 if (!(ifp->flags&IFA_F_TENTATIVE))
4567 ipv6_ifa_notify(0, ifp);
4569 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4570 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4572 } else if (had_prefixroute) {
4573 enum cleanup_prefix_rt_t action;
4574 unsigned long rt_expires;
4576 write_lock_bh(&ifp->idev->lock);
4577 action = check_cleanup_prefix_route(ifp, &rt_expires);
4578 write_unlock_bh(&ifp->idev->lock);
4580 if (action != CLEANUP_PREFIX_RT_NOP) {
4581 cleanup_prefix_route(ifp, rt_expires,
4582 action == CLEANUP_PREFIX_RT_DEL);
4586 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4587 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4588 valid_lft = prefered_lft = 0;
4589 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4590 !was_managetempaddr, jiffies);
4593 addrconf_verify_rtnl();
4599 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4600 struct netlink_ext_ack *extack)
4602 struct net *net = sock_net(skb->sk);
4603 struct ifaddrmsg *ifm;
4604 struct nlattr *tb[IFA_MAX+1];
4605 struct in6_addr *pfx, *peer_pfx;
4606 struct inet6_ifaddr *ifa;
4607 struct net_device *dev;
4608 struct inet6_dev *idev;
4609 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4613 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4618 ifm = nlmsg_data(nlh);
4619 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4623 if (tb[IFA_CACHEINFO]) {
4624 struct ifa_cacheinfo *ci;
4626 ci = nla_data(tb[IFA_CACHEINFO]);
4627 valid_lft = ci->ifa_valid;
4628 preferred_lft = ci->ifa_prefered;
4630 preferred_lft = INFINITY_LIFE_TIME;
4631 valid_lft = INFINITY_LIFE_TIME;
4634 dev = __dev_get_by_index(net, ifm->ifa_index);
4638 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4640 /* We ignore other flags so far. */
4641 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4642 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4644 idev = ipv6_find_idev(dev);
4646 return PTR_ERR(idev);
4648 if (!ipv6_allow_optimistic_dad(net, idev))
4649 ifa_flags &= ~IFA_F_OPTIMISTIC;
4651 if (ifa_flags & IFA_F_NODAD && ifa_flags & IFA_F_OPTIMISTIC) {
4652 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4656 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4659 * It would be best to check for !NLM_F_CREATE here but
4660 * userspace already relies on not having to provide this.
4662 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4663 ifm->ifa_prefixlen, ifa_flags,
4664 preferred_lft, valid_lft, extack);
4667 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4668 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4671 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4678 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4679 u8 scope, int ifindex)
4681 struct ifaddrmsg *ifm;
4683 ifm = nlmsg_data(nlh);
4684 ifm->ifa_family = AF_INET6;
4685 ifm->ifa_prefixlen = prefixlen;
4686 ifm->ifa_flags = flags;
4687 ifm->ifa_scope = scope;
4688 ifm->ifa_index = ifindex;
4691 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4692 unsigned long tstamp, u32 preferred, u32 valid)
4694 struct ifa_cacheinfo ci;
4696 ci.cstamp = cstamp_delta(cstamp);
4697 ci.tstamp = cstamp_delta(tstamp);
4698 ci.ifa_prefered = preferred;
4699 ci.ifa_valid = valid;
4701 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4704 static inline int rt_scope(int ifa_scope)
4706 if (ifa_scope & IFA_HOST)
4707 return RT_SCOPE_HOST;
4708 else if (ifa_scope & IFA_LINK)
4709 return RT_SCOPE_LINK;
4710 else if (ifa_scope & IFA_SITE)
4711 return RT_SCOPE_SITE;
4713 return RT_SCOPE_UNIVERSE;
4716 static inline int inet6_ifaddr_msgsize(void)
4718 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4719 + nla_total_size(16) /* IFA_LOCAL */
4720 + nla_total_size(16) /* IFA_ADDRESS */
4721 + nla_total_size(sizeof(struct ifa_cacheinfo))
4722 + nla_total_size(4) /* IFA_FLAGS */;
4725 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4726 u32 portid, u32 seq, int event, unsigned int flags)
4728 struct nlmsghdr *nlh;
4729 u32 preferred, valid;
4731 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4735 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4736 ifa->idev->dev->ifindex);
4738 if (!((ifa->flags&IFA_F_PERMANENT) &&
4739 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4740 preferred = ifa->prefered_lft;
4741 valid = ifa->valid_lft;
4742 if (preferred != INFINITY_LIFE_TIME) {
4743 long tval = (jiffies - ifa->tstamp)/HZ;
4744 if (preferred > tval)
4748 if (valid != INFINITY_LIFE_TIME) {
4756 preferred = INFINITY_LIFE_TIME;
4757 valid = INFINITY_LIFE_TIME;
4760 if (!ipv6_addr_any(&ifa->peer_addr)) {
4761 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4762 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4765 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4768 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4771 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4774 nlmsg_end(skb, nlh);
4778 nlmsg_cancel(skb, nlh);
4782 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4783 u32 portid, u32 seq, int event, u16 flags)
4785 struct nlmsghdr *nlh;
4786 u8 scope = RT_SCOPE_UNIVERSE;
4787 int ifindex = ifmca->idev->dev->ifindex;
4789 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4790 scope = RT_SCOPE_SITE;
4792 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4796 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4797 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4798 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4799 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4800 nlmsg_cancel(skb, nlh);
4804 nlmsg_end(skb, nlh);
4808 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4809 u32 portid, u32 seq, int event, unsigned int flags)
4811 struct nlmsghdr *nlh;
4812 u8 scope = RT_SCOPE_UNIVERSE;
4813 int ifindex = ifaca->aca_idev->dev->ifindex;
4815 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4816 scope = RT_SCOPE_SITE;
4818 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4822 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4823 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4824 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4825 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4826 nlmsg_cancel(skb, nlh);
4830 nlmsg_end(skb, nlh);
4840 /* called with rcu_read_lock() */
4841 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4842 struct netlink_callback *cb, enum addr_type_t type,
4843 int s_ip_idx, int *p_ip_idx)
4845 struct ifmcaddr6 *ifmca;
4846 struct ifacaddr6 *ifaca;
4848 int ip_idx = *p_ip_idx;
4850 read_lock_bh(&idev->lock);
4852 case UNICAST_ADDR: {
4853 struct inet6_ifaddr *ifa;
4855 /* unicast address incl. temp addr */
4856 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4857 if (++ip_idx < s_ip_idx)
4859 err = inet6_fill_ifaddr(skb, ifa,
4860 NETLINK_CB(cb->skb).portid,
4866 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4870 case MULTICAST_ADDR:
4871 /* multicast address */
4872 for (ifmca = idev->mc_list; ifmca;
4873 ifmca = ifmca->next, ip_idx++) {
4874 if (ip_idx < s_ip_idx)
4876 err = inet6_fill_ifmcaddr(skb, ifmca,
4877 NETLINK_CB(cb->skb).portid,
4886 /* anycast address */
4887 for (ifaca = idev->ac_list; ifaca;
4888 ifaca = ifaca->aca_next, ip_idx++) {
4889 if (ip_idx < s_ip_idx)
4891 err = inet6_fill_ifacaddr(skb, ifaca,
4892 NETLINK_CB(cb->skb).portid,
4903 read_unlock_bh(&idev->lock);
4908 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4909 enum addr_type_t type)
4911 struct net *net = sock_net(skb->sk);
4914 int s_idx, s_ip_idx;
4915 struct net_device *dev;
4916 struct inet6_dev *idev;
4917 struct hlist_head *head;
4920 s_idx = idx = cb->args[1];
4921 s_ip_idx = ip_idx = cb->args[2];
4924 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4925 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4927 head = &net->dev_index_head[h];
4928 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4931 if (h > s_h || idx > s_idx)
4934 idev = __in6_dev_get(dev);
4938 if (in6_dump_addrs(idev, skb, cb, type,
4939 s_ip_idx, &ip_idx) < 0)
4949 cb->args[2] = ip_idx;
4954 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4956 enum addr_type_t type = UNICAST_ADDR;
4958 return inet6_dump_addr(skb, cb, type);
4961 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4963 enum addr_type_t type = MULTICAST_ADDR;
4965 return inet6_dump_addr(skb, cb, type);
4969 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4971 enum addr_type_t type = ANYCAST_ADDR;
4973 return inet6_dump_addr(skb, cb, type);
4976 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4977 struct netlink_ext_ack *extack)
4979 struct net *net = sock_net(in_skb->sk);
4980 struct ifaddrmsg *ifm;
4981 struct nlattr *tb[IFA_MAX+1];
4982 struct in6_addr *addr = NULL, *peer;
4983 struct net_device *dev = NULL;
4984 struct inet6_ifaddr *ifa;
4985 struct sk_buff *skb;
4988 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4993 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4997 ifm = nlmsg_data(nlh);
4999 dev = dev_get_by_index(net, ifm->ifa_index);
5001 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5003 err = -EADDRNOTAVAIL;
5007 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5013 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5014 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5016 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5017 WARN_ON(err == -EMSGSIZE);
5021 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5030 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5032 struct sk_buff *skb;
5033 struct net *net = dev_net(ifa->idev->dev);
5036 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5040 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5042 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5043 WARN_ON(err == -EMSGSIZE);
5047 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5051 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5054 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5055 __s32 *array, int bytes)
5057 BUG_ON(bytes < (DEVCONF_MAX * 4));
5059 memset(array, 0, bytes);
5060 array[DEVCONF_FORWARDING] = cnf->forwarding;
5061 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5062 array[DEVCONF_MTU6] = cnf->mtu6;
5063 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5064 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5065 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5066 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5067 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5068 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5069 jiffies_to_msecs(cnf->rtr_solicit_interval);
5070 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5071 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5072 array[DEVCONF_RTR_SOLICIT_DELAY] =
5073 jiffies_to_msecs(cnf->rtr_solicit_delay);
5074 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5075 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5076 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5077 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5078 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5079 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5080 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5081 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5082 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5083 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5084 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5085 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5086 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5087 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5088 #ifdef CONFIG_IPV6_ROUTER_PREF
5089 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5090 array[DEVCONF_RTR_PROBE_INTERVAL] =
5091 jiffies_to_msecs(cnf->rtr_probe_interval);
5092 #ifdef CONFIG_IPV6_ROUTE_INFO
5093 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5094 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5097 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5098 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5099 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5100 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5101 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5103 #ifdef CONFIG_IPV6_MROUTE
5104 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5106 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5107 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5108 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5109 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5110 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5111 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5112 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5113 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5114 /* we omit DEVCONF_STABLE_SECRET for now */
5115 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5116 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5117 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5118 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5119 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5120 #ifdef CONFIG_IPV6_SEG6_HMAC
5121 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5123 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5124 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5125 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5126 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5129 static inline size_t inet6_ifla6_size(void)
5131 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5132 + nla_total_size(sizeof(struct ifla_cacheinfo))
5133 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5134 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5135 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5136 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5139 static inline size_t inet6_if_nlmsg_size(void)
5141 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5142 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5143 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5144 + nla_total_size(4) /* IFLA_MTU */
5145 + nla_total_size(4) /* IFLA_LINK */
5146 + nla_total_size(1) /* IFLA_OPERSTATE */
5147 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5150 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5154 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5157 /* Use put_unaligned() because stats may not be aligned for u64. */
5158 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5159 for (i = 1; i < ICMP6_MIB_MAX; i++)
5160 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5162 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5165 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5166 int bytes, size_t syncpoff)
5169 u64 buff[IPSTATS_MIB_MAX];
5170 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5174 memset(buff, 0, sizeof(buff));
5175 buff[0] = IPSTATS_MIB_MAX;
5177 for_each_possible_cpu(c) {
5178 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5179 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5182 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5183 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5186 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5190 case IFLA_INET6_STATS:
5191 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5192 offsetof(struct ipstats_mib, syncp));
5194 case IFLA_INET6_ICMP6STATS:
5195 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5200 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5201 u32 ext_filter_mask)
5204 struct ifla_cacheinfo ci;
5206 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5207 goto nla_put_failure;
5208 ci.max_reasm_len = IPV6_MAXPLEN;
5209 ci.tstamp = cstamp_delta(idev->tstamp);
5210 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5211 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5212 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5213 goto nla_put_failure;
5214 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5216 goto nla_put_failure;
5217 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5219 /* XXX - MC not implemented */
5221 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5224 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5226 goto nla_put_failure;
5227 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5229 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5231 goto nla_put_failure;
5232 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5234 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5236 goto nla_put_failure;
5238 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5239 goto nla_put_failure;
5241 read_lock_bh(&idev->lock);
5242 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5243 read_unlock_bh(&idev->lock);
5251 static size_t inet6_get_link_af_size(const struct net_device *dev,
5252 u32 ext_filter_mask)
5254 if (!__in6_dev_get(dev))
5257 return inet6_ifla6_size();
5260 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5261 u32 ext_filter_mask)
5263 struct inet6_dev *idev = __in6_dev_get(dev);
5268 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5274 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5276 struct inet6_ifaddr *ifp;
5277 struct net_device *dev = idev->dev;
5278 bool clear_token, update_rs = false;
5279 struct in6_addr ll_addr;
5285 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5287 if (!ipv6_accept_ra(idev))
5289 if (idev->cnf.rtr_solicits == 0)
5292 write_lock_bh(&idev->lock);
5294 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5295 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5297 write_unlock_bh(&idev->lock);
5299 clear_token = ipv6_addr_any(token);
5303 if (!idev->dead && (idev->if_flags & IF_READY) &&
5304 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5305 IFA_F_OPTIMISTIC)) {
5306 /* If we're not ready, then normal ifup will take care
5307 * of this. Otherwise, we need to request our rs here.
5309 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5314 write_lock_bh(&idev->lock);
5317 idev->if_flags |= IF_RS_SENT;
5318 idev->rs_interval = rfc3315_s14_backoff_init(
5319 idev->cnf.rtr_solicit_interval);
5320 idev->rs_probes = 1;
5321 addrconf_mod_rs_timer(idev, idev->rs_interval);
5324 /* Well, that's kinda nasty ... */
5325 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5326 spin_lock(&ifp->lock);
5327 if (ifp->tokenized) {
5329 ifp->prefered_lft = 0;
5331 spin_unlock(&ifp->lock);
5334 write_unlock_bh(&idev->lock);
5335 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5336 addrconf_verify_rtnl();
5340 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5341 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5342 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5345 static int inet6_validate_link_af(const struct net_device *dev,
5346 const struct nlattr *nla)
5348 struct nlattr *tb[IFLA_INET6_MAX + 1];
5350 if (dev && !__in6_dev_get(dev))
5351 return -EAFNOSUPPORT;
5353 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5357 static int check_addr_gen_mode(int mode)
5359 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5360 mode != IN6_ADDR_GEN_MODE_NONE &&
5361 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5362 mode != IN6_ADDR_GEN_MODE_RANDOM)
5367 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5370 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5371 !idev->cnf.stable_secret.initialized &&
5372 !net->ipv6.devconf_dflt->stable_secret.initialized)
5377 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5380 struct inet6_dev *idev = __in6_dev_get(dev);
5381 struct nlattr *tb[IFLA_INET6_MAX + 1];
5384 return -EAFNOSUPPORT;
5386 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5389 if (tb[IFLA_INET6_TOKEN]) {
5390 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5395 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5396 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5398 if (check_addr_gen_mode(mode) < 0 ||
5399 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5402 idev->cnf.addr_gen_mode = mode;
5409 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5410 u32 portid, u32 seq, int event, unsigned int flags)
5412 struct net_device *dev = idev->dev;
5413 struct ifinfomsg *hdr;
5414 struct nlmsghdr *nlh;
5417 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5421 hdr = nlmsg_data(nlh);
5422 hdr->ifi_family = AF_INET6;
5424 hdr->ifi_type = dev->type;
5425 hdr->ifi_index = dev->ifindex;
5426 hdr->ifi_flags = dev_get_flags(dev);
5427 hdr->ifi_change = 0;
5429 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5431 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5432 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5433 (dev->ifindex != dev_get_iflink(dev) &&
5434 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5435 nla_put_u8(skb, IFLA_OPERSTATE,
5436 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5437 goto nla_put_failure;
5438 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5440 goto nla_put_failure;
5442 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5443 goto nla_put_failure;
5445 nla_nest_end(skb, protoinfo);
5446 nlmsg_end(skb, nlh);
5450 nlmsg_cancel(skb, nlh);
5454 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5456 struct net *net = sock_net(skb->sk);
5459 struct net_device *dev;
5460 struct inet6_dev *idev;
5461 struct hlist_head *head;
5464 s_idx = cb->args[1];
5467 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5469 head = &net->dev_index_head[h];
5470 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5473 idev = __in6_dev_get(dev);
5476 if (inet6_fill_ifinfo(skb, idev,
5477 NETLINK_CB(cb->skb).portid,
5479 RTM_NEWLINK, NLM_F_MULTI) < 0)
5493 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5495 struct sk_buff *skb;
5496 struct net *net = dev_net(idev->dev);
5499 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5503 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5505 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5506 WARN_ON(err == -EMSGSIZE);
5510 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5514 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5517 static inline size_t inet6_prefix_nlmsg_size(void)
5519 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5520 + nla_total_size(sizeof(struct in6_addr))
5521 + nla_total_size(sizeof(struct prefix_cacheinfo));
5524 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5525 struct prefix_info *pinfo, u32 portid, u32 seq,
5526 int event, unsigned int flags)
5528 struct prefixmsg *pmsg;
5529 struct nlmsghdr *nlh;
5530 struct prefix_cacheinfo ci;
5532 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5536 pmsg = nlmsg_data(nlh);
5537 pmsg->prefix_family = AF_INET6;
5538 pmsg->prefix_pad1 = 0;
5539 pmsg->prefix_pad2 = 0;
5540 pmsg->prefix_ifindex = idev->dev->ifindex;
5541 pmsg->prefix_len = pinfo->prefix_len;
5542 pmsg->prefix_type = pinfo->type;
5543 pmsg->prefix_pad3 = 0;
5544 pmsg->prefix_flags = 0;
5546 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5547 if (pinfo->autoconf)
5548 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5550 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5551 goto nla_put_failure;
5552 ci.preferred_time = ntohl(pinfo->prefered);
5553 ci.valid_time = ntohl(pinfo->valid);
5554 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5555 goto nla_put_failure;
5556 nlmsg_end(skb, nlh);
5560 nlmsg_cancel(skb, nlh);
5564 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5565 struct prefix_info *pinfo)
5567 struct sk_buff *skb;
5568 struct net *net = dev_net(idev->dev);
5571 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5575 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5577 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5578 WARN_ON(err == -EMSGSIZE);
5582 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5586 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5589 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5591 struct net *net = dev_net(ifp->idev->dev);
5596 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5601 * If the address was optimistic
5602 * we inserted the route at the start of
5603 * our DAD process, so we don't need
5606 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5607 ip6_ins_rt(ifp->rt);
5608 if (ifp->idev->cnf.forwarding)
5609 addrconf_join_anycast(ifp);
5610 if (!ipv6_addr_any(&ifp->peer_addr))
5611 addrconf_prefix_route(&ifp->peer_addr, 128,
5612 ifp->idev->dev, 0, 0);
5615 if (ifp->idev->cnf.forwarding)
5616 addrconf_leave_anycast(ifp);
5617 addrconf_leave_solict(ifp->idev, &ifp->addr);
5618 if (!ipv6_addr_any(&ifp->peer_addr)) {
5619 struct rt6_info *rt;
5621 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5622 ifp->idev->dev, 0, 0);
5627 if (dst_hold_safe(&ifp->rt->dst))
5628 ip6_del_rt(ifp->rt);
5630 rt_genid_bump_ipv6(net);
5633 atomic_inc(&net->ipv6.dev_addr_genid);
5636 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5639 if (likely(ifp->idev->dead == 0))
5640 __ipv6_ifa_notify(event, ifp);
5641 rcu_read_unlock_bh();
5644 #ifdef CONFIG_SYSCTL
5647 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5648 void __user *buffer, size_t *lenp, loff_t *ppos)
5650 int *valp = ctl->data;
5653 struct ctl_table lctl;
5657 * ctl->data points to idev->cnf.forwarding, we should
5658 * not modify it until we get the rtnl lock.
5663 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5666 ret = addrconf_fixup_forwarding(ctl, valp, val);
5673 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5674 void __user *buffer, size_t *lenp, loff_t *ppos)
5676 struct inet6_dev *idev = ctl->extra1;
5677 int min_mtu = IPV6_MIN_MTU;
5678 struct ctl_table lctl;
5681 lctl.extra1 = &min_mtu;
5682 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5684 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5687 static void dev_disable_change(struct inet6_dev *idev)
5689 struct netdev_notifier_info info;
5691 if (!idev || !idev->dev)
5694 netdev_notifier_info_init(&info, idev->dev);
5695 if (idev->cnf.disable_ipv6)
5696 addrconf_notify(NULL, NETDEV_DOWN, &info);
5698 addrconf_notify(NULL, NETDEV_UP, &info);
5701 static void addrconf_disable_change(struct net *net, __s32 newf)
5703 struct net_device *dev;
5704 struct inet6_dev *idev;
5706 for_each_netdev(net, dev) {
5707 idev = __in6_dev_get(dev);
5709 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5710 idev->cnf.disable_ipv6 = newf;
5712 dev_disable_change(idev);
5717 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5722 if (!rtnl_trylock())
5723 return restart_syscall();
5725 net = (struct net *)table->extra2;
5729 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5734 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5735 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5736 addrconf_disable_change(net, newf);
5737 } else if ((!newf) ^ (!old))
5738 dev_disable_change((struct inet6_dev *)table->extra1);
5745 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5746 void __user *buffer, size_t *lenp, loff_t *ppos)
5748 int *valp = ctl->data;
5751 struct ctl_table lctl;
5755 * ctl->data points to idev->cnf.disable_ipv6, we should
5756 * not modify it until we get the rtnl lock.
5761 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5764 ret = addrconf_disable_ipv6(ctl, valp, val);
5771 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5772 void __user *buffer, size_t *lenp, loff_t *ppos)
5774 int *valp = ctl->data;
5779 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5782 if (write && old != new) {
5783 struct net *net = ctl->extra2;
5785 if (!rtnl_trylock())
5786 return restart_syscall();
5788 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5789 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5790 NETCONFA_PROXY_NEIGH,
5791 NETCONFA_IFINDEX_DEFAULT,
5792 net->ipv6.devconf_dflt);
5793 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5794 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5795 NETCONFA_PROXY_NEIGH,
5796 NETCONFA_IFINDEX_ALL,
5797 net->ipv6.devconf_all);
5799 struct inet6_dev *idev = ctl->extra1;
5801 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5802 NETCONFA_PROXY_NEIGH,
5812 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5813 void __user *buffer, size_t *lenp,
5818 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5819 struct net *net = (struct net *)ctl->extra2;
5821 if (!rtnl_trylock())
5822 return restart_syscall();
5824 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5827 new_val = *((int *)ctl->data);
5829 if (check_addr_gen_mode(new_val) < 0) {
5834 /* request for default */
5835 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5836 ipv6_devconf_dflt.addr_gen_mode = new_val;
5838 /* request for individual net device */
5843 if (check_stable_privacy(idev, net, new_val) < 0) {
5848 if (idev->cnf.addr_gen_mode != new_val) {
5849 idev->cnf.addr_gen_mode = new_val;
5850 addrconf_dev_config(idev->dev);
5861 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5862 void __user *buffer, size_t *lenp,
5866 struct in6_addr addr;
5867 char str[IPV6_MAX_STRLEN];
5868 struct ctl_table lctl = *ctl;
5869 struct net *net = ctl->extra2;
5870 struct ipv6_stable_secret *secret = ctl->data;
5872 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5875 lctl.maxlen = IPV6_MAX_STRLEN;
5878 if (!rtnl_trylock())
5879 return restart_syscall();
5881 if (!write && !secret->initialized) {
5886 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5887 if (err >= sizeof(str)) {
5892 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5896 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5901 secret->initialized = true;
5902 secret->secret = addr;
5904 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5905 struct net_device *dev;
5907 for_each_netdev(net, dev) {
5908 struct inet6_dev *idev = __in6_dev_get(dev);
5911 idev->cnf.addr_gen_mode =
5912 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5916 struct inet6_dev *idev = ctl->extra1;
5918 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5928 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5930 void __user *buffer,
5934 int *valp = ctl->data;
5937 struct ctl_table lctl;
5940 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5941 * we should not modify it until we get the rtnl lock.
5946 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5949 ret = addrconf_fixup_linkdown(ctl, valp, val);
5956 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5960 rt->dst.flags |= DST_NOPOLICY;
5962 rt->dst.flags &= ~DST_NOPOLICY;
5967 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5969 struct inet6_ifaddr *ifa;
5971 read_lock_bh(&idev->lock);
5972 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5973 spin_lock(&ifa->lock);
5975 struct rt6_info *rt = ifa->rt;
5979 addrconf_set_nopolicy(ifa->rt, val);
5980 if (rt->rt6i_pcpu) {
5981 for_each_possible_cpu(cpu) {
5982 struct rt6_info **rtp;
5984 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5985 addrconf_set_nopolicy(*rtp, val);
5990 spin_unlock(&ifa->lock);
5992 read_unlock_bh(&idev->lock);
5996 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5998 struct inet6_dev *idev;
6001 if (!rtnl_trylock())
6002 return restart_syscall();
6006 net = (struct net *)ctl->extra2;
6007 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6012 if (valp == &net->ipv6.devconf_all->disable_policy) {
6013 struct net_device *dev;
6015 for_each_netdev(net, dev) {
6016 idev = __in6_dev_get(dev);
6018 addrconf_disable_policy_idev(idev, val);
6021 idev = (struct inet6_dev *)ctl->extra1;
6022 addrconf_disable_policy_idev(idev, val);
6030 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6031 void __user *buffer, size_t *lenp,
6034 int *valp = ctl->data;
6037 struct ctl_table lctl;
6042 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6044 if (write && (*valp != val))
6045 ret = addrconf_disable_policy(ctl, valp, val);
6053 static int minus_one = -1;
6054 static const int zero = 0;
6055 static const int one = 1;
6056 static const int two_five_five = 255;
6058 static const struct ctl_table addrconf_sysctl[] = {
6060 .procname = "forwarding",
6061 .data = &ipv6_devconf.forwarding,
6062 .maxlen = sizeof(int),
6064 .proc_handler = addrconf_sysctl_forward,
6067 .procname = "hop_limit",
6068 .data = &ipv6_devconf.hop_limit,
6069 .maxlen = sizeof(int),
6071 .proc_handler = proc_dointvec_minmax,
6072 .extra1 = (void *)&one,
6073 .extra2 = (void *)&two_five_five,
6077 .data = &ipv6_devconf.mtu6,
6078 .maxlen = sizeof(int),
6080 .proc_handler = addrconf_sysctl_mtu,
6083 .procname = "accept_ra",
6084 .data = &ipv6_devconf.accept_ra,
6085 .maxlen = sizeof(int),
6087 .proc_handler = proc_dointvec,
6090 .procname = "accept_redirects",
6091 .data = &ipv6_devconf.accept_redirects,
6092 .maxlen = sizeof(int),
6094 .proc_handler = proc_dointvec,
6097 .procname = "autoconf",
6098 .data = &ipv6_devconf.autoconf,
6099 .maxlen = sizeof(int),
6101 .proc_handler = proc_dointvec,
6104 .procname = "dad_transmits",
6105 .data = &ipv6_devconf.dad_transmits,
6106 .maxlen = sizeof(int),
6108 .proc_handler = proc_dointvec,
6111 .procname = "router_solicitations",
6112 .data = &ipv6_devconf.rtr_solicits,
6113 .maxlen = sizeof(int),
6115 .proc_handler = proc_dointvec_minmax,
6116 .extra1 = &minus_one,
6119 .procname = "router_solicitation_interval",
6120 .data = &ipv6_devconf.rtr_solicit_interval,
6121 .maxlen = sizeof(int),
6123 .proc_handler = proc_dointvec_jiffies,
6126 .procname = "router_solicitation_max_interval",
6127 .data = &ipv6_devconf.rtr_solicit_max_interval,
6128 .maxlen = sizeof(int),
6130 .proc_handler = proc_dointvec_jiffies,
6133 .procname = "router_solicitation_delay",
6134 .data = &ipv6_devconf.rtr_solicit_delay,
6135 .maxlen = sizeof(int),
6137 .proc_handler = proc_dointvec_jiffies,
6140 .procname = "force_mld_version",
6141 .data = &ipv6_devconf.force_mld_version,
6142 .maxlen = sizeof(int),
6144 .proc_handler = proc_dointvec,
6147 .procname = "mldv1_unsolicited_report_interval",
6149 &ipv6_devconf.mldv1_unsolicited_report_interval,
6150 .maxlen = sizeof(int),
6152 .proc_handler = proc_dointvec_ms_jiffies,
6155 .procname = "mldv2_unsolicited_report_interval",
6157 &ipv6_devconf.mldv2_unsolicited_report_interval,
6158 .maxlen = sizeof(int),
6160 .proc_handler = proc_dointvec_ms_jiffies,
6163 .procname = "use_tempaddr",
6164 .data = &ipv6_devconf.use_tempaddr,
6165 .maxlen = sizeof(int),
6167 .proc_handler = proc_dointvec,
6170 .procname = "temp_valid_lft",
6171 .data = &ipv6_devconf.temp_valid_lft,
6172 .maxlen = sizeof(int),
6174 .proc_handler = proc_dointvec,
6177 .procname = "temp_prefered_lft",
6178 .data = &ipv6_devconf.temp_prefered_lft,
6179 .maxlen = sizeof(int),
6181 .proc_handler = proc_dointvec,
6184 .procname = "regen_max_retry",
6185 .data = &ipv6_devconf.regen_max_retry,
6186 .maxlen = sizeof(int),
6188 .proc_handler = proc_dointvec,
6191 .procname = "max_desync_factor",
6192 .data = &ipv6_devconf.max_desync_factor,
6193 .maxlen = sizeof(int),
6195 .proc_handler = proc_dointvec,
6198 .procname = "max_addresses",
6199 .data = &ipv6_devconf.max_addresses,
6200 .maxlen = sizeof(int),
6202 .proc_handler = proc_dointvec,
6205 .procname = "accept_ra_defrtr",
6206 .data = &ipv6_devconf.accept_ra_defrtr,
6207 .maxlen = sizeof(int),
6209 .proc_handler = proc_dointvec,
6212 .procname = "accept_ra_min_hop_limit",
6213 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6214 .maxlen = sizeof(int),
6216 .proc_handler = proc_dointvec,
6219 .procname = "accept_ra_pinfo",
6220 .data = &ipv6_devconf.accept_ra_pinfo,
6221 .maxlen = sizeof(int),
6223 .proc_handler = proc_dointvec,
6225 #ifdef CONFIG_IPV6_ROUTER_PREF
6227 .procname = "accept_ra_rtr_pref",
6228 .data = &ipv6_devconf.accept_ra_rtr_pref,
6229 .maxlen = sizeof(int),
6231 .proc_handler = proc_dointvec,
6234 .procname = "router_probe_interval",
6235 .data = &ipv6_devconf.rtr_probe_interval,
6236 .maxlen = sizeof(int),
6238 .proc_handler = proc_dointvec_jiffies,
6240 #ifdef CONFIG_IPV6_ROUTE_INFO
6242 .procname = "accept_ra_rt_info_min_plen",
6243 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6244 .maxlen = sizeof(int),
6246 .proc_handler = proc_dointvec,
6249 .procname = "accept_ra_rt_info_max_plen",
6250 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6251 .maxlen = sizeof(int),
6253 .proc_handler = proc_dointvec,
6258 .procname = "proxy_ndp",
6259 .data = &ipv6_devconf.proxy_ndp,
6260 .maxlen = sizeof(int),
6262 .proc_handler = addrconf_sysctl_proxy_ndp,
6265 .procname = "accept_source_route",
6266 .data = &ipv6_devconf.accept_source_route,
6267 .maxlen = sizeof(int),
6269 .proc_handler = proc_dointvec,
6271 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6273 .procname = "optimistic_dad",
6274 .data = &ipv6_devconf.optimistic_dad,
6275 .maxlen = sizeof(int),
6277 .proc_handler = proc_dointvec,
6280 .procname = "use_optimistic",
6281 .data = &ipv6_devconf.use_optimistic,
6282 .maxlen = sizeof(int),
6284 .proc_handler = proc_dointvec,
6287 #ifdef CONFIG_IPV6_MROUTE
6289 .procname = "mc_forwarding",
6290 .data = &ipv6_devconf.mc_forwarding,
6291 .maxlen = sizeof(int),
6293 .proc_handler = proc_dointvec,
6297 .procname = "disable_ipv6",
6298 .data = &ipv6_devconf.disable_ipv6,
6299 .maxlen = sizeof(int),
6301 .proc_handler = addrconf_sysctl_disable,
6304 .procname = "accept_dad",
6305 .data = &ipv6_devconf.accept_dad,
6306 .maxlen = sizeof(int),
6308 .proc_handler = proc_dointvec,
6311 .procname = "force_tllao",
6312 .data = &ipv6_devconf.force_tllao,
6313 .maxlen = sizeof(int),
6315 .proc_handler = proc_dointvec
6318 .procname = "ndisc_notify",
6319 .data = &ipv6_devconf.ndisc_notify,
6320 .maxlen = sizeof(int),
6322 .proc_handler = proc_dointvec
6325 .procname = "suppress_frag_ndisc",
6326 .data = &ipv6_devconf.suppress_frag_ndisc,
6327 .maxlen = sizeof(int),
6329 .proc_handler = proc_dointvec
6332 .procname = "accept_ra_from_local",
6333 .data = &ipv6_devconf.accept_ra_from_local,
6334 .maxlen = sizeof(int),
6336 .proc_handler = proc_dointvec,
6339 .procname = "accept_ra_mtu",
6340 .data = &ipv6_devconf.accept_ra_mtu,
6341 .maxlen = sizeof(int),
6343 .proc_handler = proc_dointvec,
6346 .procname = "stable_secret",
6347 .data = &ipv6_devconf.stable_secret,
6348 .maxlen = IPV6_MAX_STRLEN,
6350 .proc_handler = addrconf_sysctl_stable_secret,
6353 .procname = "use_oif_addrs_only",
6354 .data = &ipv6_devconf.use_oif_addrs_only,
6355 .maxlen = sizeof(int),
6357 .proc_handler = proc_dointvec,
6360 .procname = "ignore_routes_with_linkdown",
6361 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6362 .maxlen = sizeof(int),
6364 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6367 .procname = "drop_unicast_in_l2_multicast",
6368 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6369 .maxlen = sizeof(int),
6371 .proc_handler = proc_dointvec,
6374 .procname = "drop_unsolicited_na",
6375 .data = &ipv6_devconf.drop_unsolicited_na,
6376 .maxlen = sizeof(int),
6378 .proc_handler = proc_dointvec,
6381 .procname = "keep_addr_on_down",
6382 .data = &ipv6_devconf.keep_addr_on_down,
6383 .maxlen = sizeof(int),
6385 .proc_handler = proc_dointvec,
6389 .procname = "seg6_enabled",
6390 .data = &ipv6_devconf.seg6_enabled,
6391 .maxlen = sizeof(int),
6393 .proc_handler = proc_dointvec,
6395 #ifdef CONFIG_IPV6_SEG6_HMAC
6397 .procname = "seg6_require_hmac",
6398 .data = &ipv6_devconf.seg6_require_hmac,
6399 .maxlen = sizeof(int),
6401 .proc_handler = proc_dointvec,
6405 .procname = "enhanced_dad",
6406 .data = &ipv6_devconf.enhanced_dad,
6407 .maxlen = sizeof(int),
6409 .proc_handler = proc_dointvec,
6412 .procname = "addr_gen_mode",
6413 .data = &ipv6_devconf.addr_gen_mode,
6414 .maxlen = sizeof(int),
6416 .proc_handler = addrconf_sysctl_addr_gen_mode,
6419 .procname = "disable_policy",
6420 .data = &ipv6_devconf.disable_policy,
6421 .maxlen = sizeof(int),
6423 .proc_handler = addrconf_sysctl_disable_policy,
6426 .procname = "ndisc_tclass",
6427 .data = &ipv6_devconf.ndisc_tclass,
6428 .maxlen = sizeof(int),
6430 .proc_handler = proc_dointvec_minmax,
6431 .extra1 = (void *)&zero,
6432 .extra2 = (void *)&two_five_five,
6439 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6440 struct inet6_dev *idev, struct ipv6_devconf *p)
6443 struct ctl_table *table;
6444 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6446 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6450 for (i = 0; table[i].data; i++) {
6451 table[i].data += (char *)p - (char *)&ipv6_devconf;
6452 /* If one of these is already set, then it is not safe to
6453 * overwrite either of them: this makes proc_dointvec_minmax
6456 if (!table[i].extra1 && !table[i].extra2) {
6457 table[i].extra1 = idev; /* embedded; no ref */
6458 table[i].extra2 = net;
6462 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6464 p->sysctl_header = register_net_sysctl(net, path, table);
6465 if (!p->sysctl_header)
6468 if (!strcmp(dev_name, "all"))
6469 ifindex = NETCONFA_IFINDEX_ALL;
6470 else if (!strcmp(dev_name, "default"))
6471 ifindex = NETCONFA_IFINDEX_DEFAULT;
6473 ifindex = idev->dev->ifindex;
6474 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6484 static void __addrconf_sysctl_unregister(struct net *net,
6485 struct ipv6_devconf *p, int ifindex)
6487 struct ctl_table *table;
6489 if (!p->sysctl_header)
6492 table = p->sysctl_header->ctl_table_arg;
6493 unregister_net_sysctl_table(p->sysctl_header);
6494 p->sysctl_header = NULL;
6497 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6500 static int addrconf_sysctl_register(struct inet6_dev *idev)
6504 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6507 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6508 &ndisc_ifinfo_sysctl_change);
6511 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6514 neigh_sysctl_unregister(idev->nd_parms);
6519 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6521 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6522 idev->dev->ifindex);
6523 neigh_sysctl_unregister(idev->nd_parms);
6529 static int __net_init addrconf_init_net(struct net *net)
6532 struct ipv6_devconf *all, *dflt;
6534 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6538 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6540 goto err_alloc_dflt;
6542 /* these will be inherited by all namespaces */
6543 dflt->autoconf = ipv6_defaults.autoconf;
6544 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6546 dflt->stable_secret.initialized = false;
6547 all->stable_secret.initialized = false;
6549 net->ipv6.devconf_all = all;
6550 net->ipv6.devconf_dflt = dflt;
6552 #ifdef CONFIG_SYSCTL
6553 err = __addrconf_sysctl_register(net, "all", NULL, all);
6557 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6563 #ifdef CONFIG_SYSCTL
6565 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6575 static void __net_exit addrconf_exit_net(struct net *net)
6577 #ifdef CONFIG_SYSCTL
6578 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6579 NETCONFA_IFINDEX_DEFAULT);
6580 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6581 NETCONFA_IFINDEX_ALL);
6583 kfree(net->ipv6.devconf_dflt);
6584 kfree(net->ipv6.devconf_all);
6587 static struct pernet_operations addrconf_ops = {
6588 .init = addrconf_init_net,
6589 .exit = addrconf_exit_net,
6592 static struct rtnl_af_ops inet6_ops __read_mostly = {
6594 .fill_link_af = inet6_fill_link_af,
6595 .get_link_af_size = inet6_get_link_af_size,
6596 .validate_link_af = inet6_validate_link_af,
6597 .set_link_af = inet6_set_link_af,
6601 * Init / cleanup code
6604 int __init addrconf_init(void)
6606 struct inet6_dev *idev;
6609 err = ipv6_addr_label_init();
6611 pr_crit("%s: cannot initialize default policy table: %d\n",
6616 err = register_pernet_subsys(&addrconf_ops);
6620 addrconf_wq = create_workqueue("ipv6_addrconf");
6626 /* The addrconf netdev notifier requires that loopback_dev
6627 * has it's ipv6 private information allocated and setup
6628 * before it can bring up and give link-local addresses
6629 * to other devices which are up.
6631 * Unfortunately, loopback_dev is not necessarily the first
6632 * entry in the global dev_base list of net devices. In fact,
6633 * it is likely to be the very last entry on that list.
6634 * So this causes the notifier registry below to try and
6635 * give link-local addresses to all devices besides loopback_dev
6636 * first, then loopback_dev, which cases all the non-loopback_dev
6637 * devices to fail to get a link-local address.
6639 * So, as a temporary fix, allocate the ipv6 structure for
6640 * loopback_dev first by hand.
6641 * Longer term, all of the dependencies ipv6 has upon the loopback
6642 * device and it being up should be removed.
6645 idev = ipv6_add_dev(init_net.loopback_dev);
6648 err = PTR_ERR(idev);
6652 ip6_route_init_special_entries();
6654 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6655 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6657 register_netdevice_notifier(&ipv6_dev_notf);
6661 rtnl_af_register(&inet6_ops);
6663 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6664 NULL, inet6_dump_ifinfo, 0);
6668 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6669 inet6_rtm_newaddr, NULL, 0);
6672 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6673 inet6_rtm_deladdr, NULL, 0);
6676 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6677 inet6_rtm_getaddr, inet6_dump_ifaddr,
6678 RTNL_FLAG_DOIT_UNLOCKED);
6681 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6682 NULL, inet6_dump_ifmcaddr, 0);
6685 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6686 NULL, inet6_dump_ifacaddr, 0);
6689 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6690 inet6_netconf_get_devconf,
6691 inet6_netconf_dump_devconf,
6692 RTNL_FLAG_DOIT_UNLOCKED);
6695 err = ipv6_addr_label_rtnl_register();
6701 rtnl_unregister_all(PF_INET6);
6702 rtnl_af_unregister(&inet6_ops);
6703 unregister_netdevice_notifier(&ipv6_dev_notf);
6705 destroy_workqueue(addrconf_wq);
6707 unregister_pernet_subsys(&addrconf_ops);
6709 ipv6_addr_label_cleanup();
6714 void addrconf_cleanup(void)
6716 struct net_device *dev;
6719 unregister_netdevice_notifier(&ipv6_dev_notf);
6720 unregister_pernet_subsys(&addrconf_ops);
6721 ipv6_addr_label_cleanup();
6723 rtnl_af_unregister(&inet6_ops);
6727 /* clean dev list */
6728 for_each_netdev(&init_net, dev) {
6729 if (__in6_dev_get(dev) == NULL)
6731 addrconf_ifdown(dev, 1);
6733 addrconf_ifdown(init_net.loopback_dev, 2);
6738 spin_lock_bh(&addrconf_hash_lock);
6739 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6740 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6741 spin_unlock_bh(&addrconf_hash_lock);
6742 cancel_delayed_work(&addr_chk_work);
6745 destroy_workqueue(addrconf_wq);