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(net, idev, addr, false, gfp_flags);
1047 if (net->ipv6.devconf_all->disable_policy ||
1048 idev->cnf.disable_policy)
1049 rt->dst_nopolicy = true;
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);
1190 ip6_del_rt(dev_net(ifp->idev->dev), rt);
1192 if (!(rt->rt6i_flags & RTF_EXPIRES))
1193 fib6_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, gfp_t gfp_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,
2334 .fc_type = RTN_UNICAST,
2339 /* Prevent useless cloning on PtP SIT.
2340 This thing is done here expecting that the whole
2341 class of non-broadcast devices need not cloning.
2343 #if IS_ENABLED(CONFIG_IPV6_SIT)
2344 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2345 cfg.fc_flags |= RTF_NONEXTHOP;
2348 ip6_route_add(&cfg, gfp_flags, NULL);
2352 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2354 const struct net_device *dev,
2355 u32 flags, u32 noflags)
2357 struct fib6_node *fn;
2358 struct rt6_info *rt = NULL;
2359 struct fib6_table *table;
2360 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2362 table = fib6_get_table(dev_net(dev), tb_id);
2367 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2371 for_each_fib6_node_rt_rcu(fn) {
2372 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2374 if ((rt->rt6i_flags & flags) != flags)
2376 if ((rt->rt6i_flags & noflags) != 0)
2378 if (!dst_hold_safe(&rt->dst))
2388 /* Create "default" multicast route to the interface */
2390 static void addrconf_add_mroute(struct net_device *dev)
2392 struct fib6_config cfg = {
2393 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2394 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2395 .fc_ifindex = dev->ifindex,
2398 .fc_type = RTN_UNICAST,
2399 .fc_nlinfo.nl_net = dev_net(dev),
2402 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2404 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
2407 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2409 struct inet6_dev *idev;
2413 idev = ipv6_find_idev(dev);
2415 return ERR_PTR(-ENOBUFS);
2417 if (idev->cnf.disable_ipv6)
2418 return ERR_PTR(-EACCES);
2420 /* Add default multicast route */
2421 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2422 addrconf_add_mroute(dev);
2427 static void manage_tempaddrs(struct inet6_dev *idev,
2428 struct inet6_ifaddr *ifp,
2429 __u32 valid_lft, __u32 prefered_lft,
2430 bool create, unsigned long now)
2433 struct inet6_ifaddr *ift;
2435 read_lock_bh(&idev->lock);
2436 /* update all temporary addresses in the list */
2437 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2438 int age, max_valid, max_prefered;
2440 if (ifp != ift->ifpub)
2443 /* RFC 4941 section 3.3:
2444 * If a received option will extend the lifetime of a public
2445 * address, the lifetimes of temporary addresses should
2446 * be extended, subject to the overall constraint that no
2447 * temporary addresses should ever remain "valid" or "preferred"
2448 * for a time longer than (TEMP_VALID_LIFETIME) or
2449 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2451 age = (now - ift->cstamp) / HZ;
2452 max_valid = idev->cnf.temp_valid_lft - age;
2456 max_prefered = idev->cnf.temp_prefered_lft -
2457 idev->desync_factor - age;
2458 if (max_prefered < 0)
2461 if (valid_lft > max_valid)
2462 valid_lft = max_valid;
2464 if (prefered_lft > max_prefered)
2465 prefered_lft = max_prefered;
2467 spin_lock(&ift->lock);
2469 ift->valid_lft = valid_lft;
2470 ift->prefered_lft = prefered_lft;
2472 if (prefered_lft > 0)
2473 ift->flags &= ~IFA_F_DEPRECATED;
2475 spin_unlock(&ift->lock);
2476 if (!(flags&IFA_F_TENTATIVE))
2477 ipv6_ifa_notify(0, ift);
2480 if ((create || list_empty(&idev->tempaddr_list)) &&
2481 idev->cnf.use_tempaddr > 0) {
2482 /* When a new public address is created as described
2483 * in [ADDRCONF], also create a new temporary address.
2484 * Also create a temporary address if it's enabled but
2485 * no temporary address currently exists.
2487 read_unlock_bh(&idev->lock);
2488 ipv6_create_tempaddr(ifp, NULL, false);
2490 read_unlock_bh(&idev->lock);
2494 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2496 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2497 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2500 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2501 const struct prefix_info *pinfo,
2502 struct inet6_dev *in6_dev,
2503 const struct in6_addr *addr, int addr_type,
2504 u32 addr_flags, bool sllao, bool tokenized,
2505 __u32 valid_lft, u32 prefered_lft)
2507 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2508 int create = 0, update_lft = 0;
2510 if (!ifp && valid_lft) {
2511 int max_addresses = in6_dev->cnf.max_addresses;
2513 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2514 if ((net->ipv6.devconf_all->optimistic_dad ||
2515 in6_dev->cnf.optimistic_dad) &&
2516 !net->ipv6.devconf_all->forwarding && sllao)
2517 addr_flags |= IFA_F_OPTIMISTIC;
2520 /* Do not allow to create too much of autoconfigured
2521 * addresses; this would be too easy way to crash kernel.
2523 if (!max_addresses ||
2524 ipv6_count_addresses(in6_dev) < max_addresses)
2525 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2527 addr_type&IPV6_ADDR_SCOPE_MASK,
2528 addr_flags, valid_lft,
2529 prefered_lft, false, NULL);
2531 if (IS_ERR_OR_NULL(ifp))
2535 spin_lock_bh(&ifp->lock);
2536 ifp->flags |= IFA_F_MANAGETEMPADDR;
2537 ifp->cstamp = jiffies;
2538 ifp->tokenized = tokenized;
2539 spin_unlock_bh(&ifp->lock);
2540 addrconf_dad_start(ifp);
2548 /* update lifetime (RFC2462 5.5.3 e) */
2549 spin_lock_bh(&ifp->lock);
2551 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2552 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2555 if (!create && stored_lft) {
2556 const u32 minimum_lft = min_t(u32,
2557 stored_lft, MIN_VALID_LIFETIME);
2558 valid_lft = max(valid_lft, minimum_lft);
2560 /* RFC4862 Section 5.5.3e:
2561 * "Note that the preferred lifetime of the
2562 * corresponding address is always reset to
2563 * the Preferred Lifetime in the received
2564 * Prefix Information option, regardless of
2565 * whether the valid lifetime is also reset or
2568 * So we should always update prefered_lft here.
2574 ifp->valid_lft = valid_lft;
2575 ifp->prefered_lft = prefered_lft;
2578 ifp->flags &= ~IFA_F_DEPRECATED;
2579 spin_unlock_bh(&ifp->lock);
2581 if (!(flags&IFA_F_TENTATIVE))
2582 ipv6_ifa_notify(0, ifp);
2584 spin_unlock_bh(&ifp->lock);
2586 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2595 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2597 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2599 struct prefix_info *pinfo;
2604 struct inet6_dev *in6_dev;
2605 struct net *net = dev_net(dev);
2607 pinfo = (struct prefix_info *) opt;
2609 if (len < sizeof(struct prefix_info)) {
2610 netdev_dbg(dev, "addrconf: prefix option too short\n");
2615 * Validation checks ([ADDRCONF], page 19)
2618 addr_type = ipv6_addr_type(&pinfo->prefix);
2620 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2623 valid_lft = ntohl(pinfo->valid);
2624 prefered_lft = ntohl(pinfo->prefered);
2626 if (prefered_lft > valid_lft) {
2627 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2631 in6_dev = in6_dev_get(dev);
2634 net_dbg_ratelimited("addrconf: device %s not configured\n",
2640 * Two things going on here:
2641 * 1) Add routes for on-link prefixes
2642 * 2) Configure prefixes with the auto flag set
2645 if (pinfo->onlink) {
2646 struct rt6_info *rt;
2647 unsigned long rt_expires;
2649 /* Avoid arithmetic overflow. Really, we could
2650 * save rt_expires in seconds, likely valid_lft,
2651 * but it would require division in fib gc, that it
2655 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2657 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2659 if (addrconf_finite_timeout(rt_expires))
2662 rt = addrconf_get_prefix_route(&pinfo->prefix,
2665 RTF_ADDRCONF | RTF_PREFIX_RT,
2666 RTF_GATEWAY | RTF_DEFAULT);
2669 /* Autoconf prefix route */
2670 if (valid_lft == 0) {
2671 ip6_del_rt(net, rt);
2673 } else if (addrconf_finite_timeout(rt_expires)) {
2675 fib6_set_expires(rt, jiffies + rt_expires);
2677 fib6_clean_expires(rt);
2679 } else if (valid_lft) {
2680 clock_t expires = 0;
2681 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2682 if (addrconf_finite_timeout(rt_expires)) {
2684 flags |= RTF_EXPIRES;
2685 expires = jiffies_to_clock_t(rt_expires);
2687 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2688 dev, expires, flags, GFP_ATOMIC);
2693 /* Try to figure out our local address for this prefix */
2695 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2696 struct in6_addr addr;
2697 bool tokenized = false, dev_addr_generated = false;
2699 if (pinfo->prefix_len == 64) {
2700 memcpy(&addr, &pinfo->prefix, 8);
2702 if (!ipv6_addr_any(&in6_dev->token)) {
2703 read_lock_bh(&in6_dev->lock);
2704 memcpy(addr.s6_addr + 8,
2705 in6_dev->token.s6_addr + 8, 8);
2706 read_unlock_bh(&in6_dev->lock);
2708 } else if (is_addr_mode_generate_stable(in6_dev) &&
2709 !ipv6_generate_stable_address(&addr, 0,
2711 addr_flags |= IFA_F_STABLE_PRIVACY;
2713 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2714 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2717 dev_addr_generated = true;
2721 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2726 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2729 tokenized, valid_lft,
2734 /* Ignore error case here because previous prefix add addr was
2735 * successful which will be notified.
2737 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2738 addr_type, addr_flags, sllao,
2739 tokenized, valid_lft,
2741 dev_addr_generated);
2743 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2745 in6_dev_put(in6_dev);
2749 * Set destination address.
2750 * Special case for SIT interfaces where we create a new "virtual"
2753 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2755 struct in6_ifreq ireq;
2756 struct net_device *dev;
2762 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2765 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2771 #if IS_ENABLED(CONFIG_IPV6_SIT)
2772 if (dev->type == ARPHRD_SIT) {
2773 const struct net_device_ops *ops = dev->netdev_ops;
2775 struct ip_tunnel_parm p;
2777 err = -EADDRNOTAVAIL;
2778 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2781 memset(&p, 0, sizeof(p));
2782 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2786 p.iph.protocol = IPPROTO_IPV6;
2788 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2790 if (ops->ndo_do_ioctl) {
2791 mm_segment_t oldfs = get_fs();
2794 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2801 dev = __dev_get_by_name(net, p.name);
2804 err = dev_open(dev);
2814 static int ipv6_mc_config(struct sock *sk, bool join,
2815 const struct in6_addr *addr, int ifindex)
2823 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2825 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2832 * Manual configuration of address on an interface
2834 static int inet6_addr_add(struct net *net, int ifindex,
2835 const struct in6_addr *pfx,
2836 const struct in6_addr *peer_pfx,
2837 unsigned int plen, __u32 ifa_flags,
2838 __u32 prefered_lft, __u32 valid_lft,
2839 struct netlink_ext_ack *extack)
2841 struct inet6_ifaddr *ifp;
2842 struct inet6_dev *idev;
2843 struct net_device *dev;
2844 unsigned long timeout;
2854 /* check the lifetime */
2855 if (!valid_lft || prefered_lft > valid_lft)
2858 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2861 dev = __dev_get_by_index(net, ifindex);
2865 idev = addrconf_add_dev(dev);
2867 return PTR_ERR(idev);
2869 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2870 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2871 true, pfx, ifindex);
2877 scope = ipv6_addr_scope(pfx);
2879 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2880 if (addrconf_finite_timeout(timeout)) {
2881 expires = jiffies_to_clock_t(timeout * HZ);
2882 valid_lft = timeout;
2883 flags = RTF_EXPIRES;
2887 ifa_flags |= IFA_F_PERMANENT;
2890 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2891 if (addrconf_finite_timeout(timeout)) {
2893 ifa_flags |= IFA_F_DEPRECATED;
2894 prefered_lft = timeout;
2897 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2898 valid_lft, prefered_lft, true, extack);
2901 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2902 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2903 expires, flags, GFP_KERNEL);
2906 /* Send a netlink notification if DAD is enabled and
2907 * optimistic flag is not set
2909 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2910 ipv6_ifa_notify(0, ifp);
2912 * Note that section 3.1 of RFC 4429 indicates
2913 * that the Optimistic flag should not be set for
2914 * manually configured addresses
2916 addrconf_dad_start(ifp);
2917 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2918 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2921 addrconf_verify_rtnl();
2923 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2924 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2925 false, pfx, ifindex);
2928 return PTR_ERR(ifp);
2931 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2932 const struct in6_addr *pfx, unsigned int plen)
2934 struct inet6_ifaddr *ifp;
2935 struct inet6_dev *idev;
2936 struct net_device *dev;
2941 dev = __dev_get_by_index(net, ifindex);
2945 idev = __in6_dev_get(dev);
2949 read_lock_bh(&idev->lock);
2950 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2951 if (ifp->prefix_len == plen &&
2952 ipv6_addr_equal(pfx, &ifp->addr)) {
2954 read_unlock_bh(&idev->lock);
2956 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2957 (ifa_flags & IFA_F_MANAGETEMPADDR))
2958 manage_tempaddrs(idev, ifp, 0, 0, false,
2961 addrconf_verify_rtnl();
2962 if (ipv6_addr_is_multicast(pfx)) {
2963 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2964 false, pfx, dev->ifindex);
2969 read_unlock_bh(&idev->lock);
2970 return -EADDRNOTAVAIL;
2974 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2976 struct in6_ifreq ireq;
2979 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2982 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2986 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2987 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2988 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2993 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2995 struct in6_ifreq ireq;
2998 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3001 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3005 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3006 ireq.ifr6_prefixlen);
3011 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3012 int plen, int scope)
3014 struct inet6_ifaddr *ifp;
3016 ifp = ipv6_add_addr(idev, addr, NULL, plen,
3017 scope, IFA_F_PERMANENT,
3018 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
3021 spin_lock_bh(&ifp->lock);
3022 ifp->flags &= ~IFA_F_TENTATIVE;
3023 spin_unlock_bh(&ifp->lock);
3024 rt_genid_bump_ipv6(dev_net(idev->dev));
3025 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3030 #if IS_ENABLED(CONFIG_IPV6_SIT)
3031 static void sit_add_v4_addrs(struct inet6_dev *idev)
3033 struct in6_addr addr;
3034 struct net_device *dev;
3035 struct net *net = dev_net(idev->dev);
3041 memset(&addr, 0, sizeof(struct in6_addr));
3042 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3044 if (idev->dev->flags&IFF_POINTOPOINT) {
3045 addr.s6_addr32[0] = htonl(0xfe800000);
3049 scope = IPV6_ADDR_COMPATv4;
3051 pflags |= RTF_NONEXTHOP;
3054 if (addr.s6_addr32[3]) {
3055 add_addr(idev, &addr, plen, scope);
3056 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags,
3061 for_each_netdev(net, dev) {
3062 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3063 if (in_dev && (dev->flags & IFF_UP)) {
3064 struct in_ifaddr *ifa;
3068 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3070 addr.s6_addr32[3] = ifa->ifa_local;
3072 if (ifa->ifa_scope == RT_SCOPE_LINK)
3074 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3075 if (idev->dev->flags&IFF_POINTOPOINT)
3080 add_addr(idev, &addr, plen, flag);
3081 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3082 pflags, GFP_ATOMIC);
3089 static void init_loopback(struct net_device *dev)
3091 struct inet6_dev *idev;
3097 idev = ipv6_find_idev(dev);
3099 pr_debug("%s: add_dev failed\n", __func__);
3103 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3106 void addrconf_add_linklocal(struct inet6_dev *idev,
3107 const struct in6_addr *addr, u32 flags)
3109 struct inet6_ifaddr *ifp;
3110 u32 addr_flags = flags | IFA_F_PERMANENT;
3112 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3113 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3114 idev->cnf.optimistic_dad) &&
3115 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3116 addr_flags |= IFA_F_OPTIMISTIC;
3119 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3120 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3122 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev,
3124 addrconf_dad_start(ifp);
3128 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3130 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3132 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3135 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3136 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3139 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3140 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3146 static int ipv6_generate_stable_address(struct in6_addr *address,
3148 const struct inet6_dev *idev)
3150 static DEFINE_SPINLOCK(lock);
3151 static __u32 digest[SHA_DIGEST_WORDS];
3152 static __u32 workspace[SHA_WORKSPACE_WORDS];
3155 char __data[SHA_MESSAGE_BYTES];
3157 struct in6_addr secret;
3159 unsigned char hwaddr[MAX_ADDR_LEN];
3164 struct in6_addr secret;
3165 struct in6_addr temp;
3166 struct net *net = dev_net(idev->dev);
3168 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3170 if (idev->cnf.stable_secret.initialized)
3171 secret = idev->cnf.stable_secret.secret;
3172 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3173 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3178 spin_lock_bh(&lock);
3181 memset(&data, 0, sizeof(data));
3182 memset(workspace, 0, sizeof(workspace));
3183 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3184 data.prefix[0] = address->s6_addr32[0];
3185 data.prefix[1] = address->s6_addr32[1];
3186 data.secret = secret;
3187 data.dad_count = dad_count;
3189 sha_transform(digest, data.__data, workspace);
3192 temp.s6_addr32[2] = (__force __be32)digest[0];
3193 temp.s6_addr32[3] = (__force __be32)digest[1];
3195 spin_unlock_bh(&lock);
3197 if (ipv6_reserved_interfaceid(temp)) {
3199 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3208 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3210 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3214 s = &idev->cnf.stable_secret;
3215 get_random_bytes(&s->secret, sizeof(s->secret));
3216 s->initialized = true;
3219 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3221 struct in6_addr addr;
3223 /* no link local addresses on L3 master devices */
3224 if (netif_is_l3_master(idev->dev))
3227 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3229 switch (idev->cnf.addr_gen_mode) {
3230 case IN6_ADDR_GEN_MODE_RANDOM:
3231 ipv6_gen_mode_random_init(idev);
3233 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3234 if (!ipv6_generate_stable_address(&addr, 0, idev))
3235 addrconf_add_linklocal(idev, &addr,
3236 IFA_F_STABLE_PRIVACY);
3237 else if (prefix_route)
3238 addrconf_prefix_route(&addr, 64, idev->dev,
3241 case IN6_ADDR_GEN_MODE_EUI64:
3242 /* addrconf_add_linklocal also adds a prefix_route and we
3243 * only need to care about prefix routes if ipv6_generate_eui64
3244 * couldn't generate one.
3246 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3247 addrconf_add_linklocal(idev, &addr, 0);
3248 else if (prefix_route)
3249 addrconf_prefix_route(&addr, 64, idev->dev,
3252 case IN6_ADDR_GEN_MODE_NONE:
3254 /* will not add any link local address */
3259 static void addrconf_dev_config(struct net_device *dev)
3261 struct inet6_dev *idev;
3265 if ((dev->type != ARPHRD_ETHER) &&
3266 (dev->type != ARPHRD_FDDI) &&
3267 (dev->type != ARPHRD_ARCNET) &&
3268 (dev->type != ARPHRD_INFINIBAND) &&
3269 (dev->type != ARPHRD_IEEE1394) &&
3270 (dev->type != ARPHRD_TUNNEL6) &&
3271 (dev->type != ARPHRD_6LOWPAN) &&
3272 (dev->type != ARPHRD_IP6GRE) &&
3273 (dev->type != ARPHRD_IPGRE) &&
3274 (dev->type != ARPHRD_TUNNEL) &&
3275 (dev->type != ARPHRD_NONE)) {
3276 /* Alas, we support only Ethernet autoconfiguration. */
3280 idev = addrconf_add_dev(dev);
3284 /* this device type has no EUI support */
3285 if (dev->type == ARPHRD_NONE &&
3286 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3287 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3289 addrconf_addr_gen(idev, false);
3292 #if IS_ENABLED(CONFIG_IPV6_SIT)
3293 static void addrconf_sit_config(struct net_device *dev)
3295 struct inet6_dev *idev;
3300 * Configure the tunnel with one of our IPv4
3301 * addresses... we should configure all of
3302 * our v4 addrs in the tunnel
3305 idev = ipv6_find_idev(dev);
3307 pr_debug("%s: add_dev failed\n", __func__);
3311 if (dev->priv_flags & IFF_ISATAP) {
3312 addrconf_addr_gen(idev, false);
3316 sit_add_v4_addrs(idev);
3318 if (dev->flags&IFF_POINTOPOINT)
3319 addrconf_add_mroute(dev);
3323 #if IS_ENABLED(CONFIG_NET_IPGRE)
3324 static void addrconf_gre_config(struct net_device *dev)
3326 struct inet6_dev *idev;
3330 idev = ipv6_find_idev(dev);
3332 pr_debug("%s: add_dev failed\n", __func__);
3336 addrconf_addr_gen(idev, true);
3337 if (dev->flags & IFF_POINTOPOINT)
3338 addrconf_add_mroute(dev);
3342 static int fixup_permanent_addr(struct net *net,
3343 struct inet6_dev *idev,
3344 struct inet6_ifaddr *ifp)
3346 /* !rt6i_node means the host route was removed from the
3347 * FIB, for example, if 'lo' device is taken down. In that
3348 * case regenerate the host route.
3350 if (!ifp->rt || !ifp->rt->rt6i_node) {
3351 struct rt6_info *rt, *prev;
3353 rt = addrconf_dst_alloc(net, idev, &ifp->addr, false,
3358 /* ifp->rt can be accessed outside of rtnl */
3359 spin_lock(&ifp->lock);
3362 spin_unlock(&ifp->lock);
3367 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3368 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3369 idev->dev, 0, 0, GFP_ATOMIC);
3372 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3373 addrconf_dad_start(ifp);
3378 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3380 struct inet6_ifaddr *ifp, *tmp;
3381 struct inet6_dev *idev;
3383 idev = __in6_dev_get(dev);
3387 write_lock_bh(&idev->lock);
3389 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3390 if ((ifp->flags & IFA_F_PERMANENT) &&
3391 fixup_permanent_addr(net, idev, ifp) < 0) {
3392 write_unlock_bh(&idev->lock);
3395 write_lock_bh(&idev->lock);
3397 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3398 idev->dev->name, &ifp->addr);
3402 write_unlock_bh(&idev->lock);
3405 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3408 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3409 struct netdev_notifier_changeupper_info *info;
3410 struct inet6_dev *idev = __in6_dev_get(dev);
3411 struct net *net = dev_net(dev);
3412 int run_pending = 0;
3416 case NETDEV_REGISTER:
3417 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3418 idev = ipv6_add_dev(dev);
3420 return notifier_from_errno(PTR_ERR(idev));
3424 case NETDEV_CHANGEMTU:
3425 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3426 if (dev->mtu < IPV6_MIN_MTU) {
3427 addrconf_ifdown(dev, dev != net->loopback_dev);
3432 rt6_mtu_change(dev, dev->mtu);
3433 idev->cnf.mtu6 = dev->mtu;
3437 /* allocate new idev */
3438 idev = ipv6_add_dev(dev);
3442 /* device is still not ready */
3443 if (!(idev->if_flags & IF_READY))
3452 if (dev->flags & IFF_SLAVE)
3455 if (idev && idev->cnf.disable_ipv6)
3458 if (event == NETDEV_UP) {
3459 /* restore routes for permanent addresses */
3460 addrconf_permanent_addr(net, dev);
3462 if (!addrconf_link_ready(dev)) {
3463 /* device is not ready yet. */
3464 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3469 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3470 idev = ipv6_add_dev(dev);
3472 if (!IS_ERR_OR_NULL(idev)) {
3473 idev->if_flags |= IF_READY;
3476 } else if (event == NETDEV_CHANGE) {
3477 if (!addrconf_link_ready(dev)) {
3478 /* device is still not ready. */
3479 rt6_sync_down_dev(dev, event);
3484 if (idev->if_flags & IF_READY) {
3485 /* device is already configured -
3486 * but resend MLD reports, we might
3487 * have roamed and need to update
3488 * multicast snooping switches
3491 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3494 idev->if_flags |= IF_READY;
3497 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3503 switch (dev->type) {
3504 #if IS_ENABLED(CONFIG_IPV6_SIT)
3506 addrconf_sit_config(dev);
3509 #if IS_ENABLED(CONFIG_NET_IPGRE)
3511 addrconf_gre_config(dev);
3514 case ARPHRD_LOOPBACK:
3519 addrconf_dev_config(dev);
3523 if (!IS_ERR_OR_NULL(idev)) {
3525 addrconf_dad_run(idev);
3527 /* Device has an address by now */
3528 rt6_sync_up(dev, RTNH_F_DEAD);
3531 * If the MTU changed during the interface down,
3532 * when the interface up, the changed MTU must be
3533 * reflected in the idev as well as routers.
3535 if (idev->cnf.mtu6 != dev->mtu &&
3536 dev->mtu >= IPV6_MIN_MTU) {
3537 rt6_mtu_change(dev, dev->mtu);
3538 idev->cnf.mtu6 = dev->mtu;
3540 idev->tstamp = jiffies;
3541 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3544 * If the changed mtu during down is lower than
3545 * IPV6_MIN_MTU stop IPv6 on this interface.
3547 if (dev->mtu < IPV6_MIN_MTU)
3548 addrconf_ifdown(dev, dev != net->loopback_dev);
3553 case NETDEV_UNREGISTER:
3555 * Remove all addresses from this interface.
3557 addrconf_ifdown(dev, event != NETDEV_DOWN);
3560 case NETDEV_CHANGENAME:
3562 snmp6_unregister_dev(idev);
3563 addrconf_sysctl_unregister(idev);
3564 err = addrconf_sysctl_register(idev);
3566 return notifier_from_errno(err);
3567 err = snmp6_register_dev(idev);
3569 addrconf_sysctl_unregister(idev);
3570 return notifier_from_errno(err);
3575 case NETDEV_PRE_TYPE_CHANGE:
3576 case NETDEV_POST_TYPE_CHANGE:
3578 addrconf_type_change(dev, event);
3581 case NETDEV_CHANGEUPPER:
3584 /* flush all routes if dev is linked to or unlinked from
3585 * an L3 master device (e.g., VRF)
3587 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3588 addrconf_ifdown(dev, 0);
3595 * addrconf module should be notified of a device going up
3597 static struct notifier_block ipv6_dev_notf = {
3598 .notifier_call = addrconf_notify,
3599 .priority = ADDRCONF_NOTIFY_PRIORITY,
3602 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3604 struct inet6_dev *idev;
3607 idev = __in6_dev_get(dev);
3609 if (event == NETDEV_POST_TYPE_CHANGE)
3610 ipv6_mc_remap(idev);
3611 else if (event == NETDEV_PRE_TYPE_CHANGE)
3612 ipv6_mc_unmap(idev);
3615 static bool addr_is_local(const struct in6_addr *addr)
3617 return ipv6_addr_type(addr) &
3618 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3621 static int addrconf_ifdown(struct net_device *dev, int how)
3623 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3624 struct net *net = dev_net(dev);
3625 struct inet6_dev *idev;
3626 struct inet6_ifaddr *ifa, *tmp;
3633 rt6_disable_ip(dev, event);
3635 idev = __in6_dev_get(dev);
3640 * Step 1: remove reference to ipv6 device from parent device.
3646 /* protected by rtnl_lock */
3647 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3649 /* Step 1.5: remove snmp6 entry */
3650 snmp6_unregister_dev(idev);
3654 /* aggregate the system setting and interface setting */
3655 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3657 _keep_addr = idev->cnf.keep_addr_on_down;
3659 /* combine the user config with event to determine if permanent
3660 * addresses are to be removed from address hash table
3662 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3664 /* Step 2: clear hash table */
3665 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3666 struct hlist_head *h = &inet6_addr_lst[i];
3668 spin_lock_bh(&addrconf_hash_lock);
3670 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3671 if (ifa->idev == idev) {
3672 addrconf_del_dad_work(ifa);
3673 /* combined flag + permanent flag decide if
3674 * address is retained on a down event
3677 !(ifa->flags & IFA_F_PERMANENT) ||
3678 addr_is_local(&ifa->addr)) {
3679 hlist_del_init_rcu(&ifa->addr_lst);
3684 spin_unlock_bh(&addrconf_hash_lock);
3687 write_lock_bh(&idev->lock);
3689 addrconf_del_rs_timer(idev);
3691 /* Step 2: clear flags for stateless addrconf */
3693 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3695 /* Step 3: clear tempaddr list */
3696 while (!list_empty(&idev->tempaddr_list)) {
3697 ifa = list_first_entry(&idev->tempaddr_list,
3698 struct inet6_ifaddr, tmp_list);
3699 list_del(&ifa->tmp_list);
3700 write_unlock_bh(&idev->lock);
3701 spin_lock_bh(&ifa->lock);
3704 in6_ifa_put(ifa->ifpub);
3707 spin_unlock_bh(&ifa->lock);
3709 write_lock_bh(&idev->lock);
3712 /* re-combine the user config with event to determine if permanent
3713 * addresses are to be removed from the interface list
3715 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3717 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3718 struct rt6_info *rt = NULL;
3721 addrconf_del_dad_work(ifa);
3723 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3724 !addr_is_local(&ifa->addr);
3726 write_unlock_bh(&idev->lock);
3727 spin_lock_bh(&ifa->lock);
3730 /* set state to skip the notifier below */
3731 state = INET6_IFADDR_STATE_DEAD;
3732 ifa->state = INET6_IFADDR_STATE_PREDAD;
3733 if (!(ifa->flags & IFA_F_NODAD))
3734 ifa->flags |= IFA_F_TENTATIVE;
3740 ifa->state = INET6_IFADDR_STATE_DEAD;
3743 spin_unlock_bh(&ifa->lock);
3746 ip6_del_rt(net, rt);
3748 if (state != INET6_IFADDR_STATE_DEAD) {
3749 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3750 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3752 if (idev->cnf.forwarding)
3753 addrconf_leave_anycast(ifa);
3754 addrconf_leave_solict(ifa->idev, &ifa->addr);
3757 write_lock_bh(&idev->lock);
3759 list_del_rcu(&ifa->if_list);
3764 write_unlock_bh(&idev->lock);
3766 /* Step 5: Discard anycast and multicast list */
3768 ipv6_ac_destroy_dev(idev);
3769 ipv6_mc_destroy_dev(idev);
3774 idev->tstamp = jiffies;
3776 /* Last: Shot the device (if unregistered) */
3778 addrconf_sysctl_unregister(idev);
3779 neigh_parms_release(&nd_tbl, idev->nd_parms);
3780 neigh_ifdown(&nd_tbl, dev);
3786 static void addrconf_rs_timer(struct timer_list *t)
3788 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3789 struct net_device *dev = idev->dev;
3790 struct in6_addr lladdr;
3792 write_lock(&idev->lock);
3793 if (idev->dead || !(idev->if_flags & IF_READY))
3796 if (!ipv6_accept_ra(idev))
3799 /* Announcement received after solicitation was sent */
3800 if (idev->if_flags & IF_RA_RCVD)
3803 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3804 write_unlock(&idev->lock);
3805 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3806 ndisc_send_rs(dev, &lladdr,
3807 &in6addr_linklocal_allrouters);
3811 write_lock(&idev->lock);
3812 idev->rs_interval = rfc3315_s14_backoff_update(
3813 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3814 /* The wait after the last probe can be shorter */
3815 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3816 idev->cnf.rtr_solicits) ?
3817 idev->cnf.rtr_solicit_delay :
3821 * Note: we do not support deprecated "all on-link"
3822 * assumption any longer.
3824 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3828 write_unlock(&idev->lock);
3834 * Duplicate Address Detection
3836 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3838 unsigned long rand_num;
3839 struct inet6_dev *idev = ifp->idev;
3842 if (ifp->flags & IFA_F_OPTIMISTIC)
3845 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3848 if (idev->cnf.enhanced_dad ||
3849 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3851 get_random_bytes(&nonce, 6);
3854 ifp->dad_nonce = nonce;
3855 ifp->dad_probes = idev->cnf.dad_transmits;
3856 addrconf_mod_dad_work(ifp, rand_num);
3859 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3861 struct inet6_dev *idev = ifp->idev;
3862 struct net_device *dev = idev->dev;
3863 bool bump_id, notify = false;
3866 addrconf_join_solict(dev, &ifp->addr);
3868 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3870 read_lock_bh(&idev->lock);
3871 spin_lock(&ifp->lock);
3872 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3876 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3877 (net->ipv6.devconf_all->accept_dad < 1 &&
3878 idev->cnf.accept_dad < 1) ||
3879 !(ifp->flags&IFA_F_TENTATIVE) ||
3880 ifp->flags & IFA_F_NODAD) {
3881 bool send_na = false;
3883 if (ifp->flags & IFA_F_TENTATIVE &&
3884 !(ifp->flags & IFA_F_OPTIMISTIC))
3886 bump_id = ifp->flags & IFA_F_TENTATIVE;
3887 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3888 spin_unlock(&ifp->lock);
3889 read_unlock_bh(&idev->lock);
3891 addrconf_dad_completed(ifp, bump_id, send_na);
3895 if (!(idev->if_flags & IF_READY)) {
3896 spin_unlock(&ifp->lock);
3897 read_unlock_bh(&idev->lock);
3899 * If the device is not ready:
3900 * - keep it tentative if it is a permanent address.
3901 * - otherwise, kill it.
3904 addrconf_dad_stop(ifp, 0);
3909 * Optimistic nodes can start receiving
3912 if (ifp->flags & IFA_F_OPTIMISTIC) {
3913 ip6_ins_rt(net, ifp->rt);
3914 if (ipv6_use_optimistic_addr(net, idev)) {
3915 /* Because optimistic nodes can use this address,
3916 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3922 addrconf_dad_kick(ifp);
3924 spin_unlock(&ifp->lock);
3925 read_unlock_bh(&idev->lock);
3927 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3930 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3932 bool begin_dad = false;
3934 spin_lock_bh(&ifp->lock);
3935 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3936 ifp->state = INET6_IFADDR_STATE_PREDAD;
3939 spin_unlock_bh(&ifp->lock);
3942 addrconf_mod_dad_work(ifp, 0);
3945 static void addrconf_dad_work(struct work_struct *w)
3947 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3948 struct inet6_ifaddr,
3950 struct inet6_dev *idev = ifp->idev;
3951 bool bump_id, disable_ipv6 = false;
3952 struct in6_addr mcaddr;
3958 } action = DAD_PROCESS;
3962 spin_lock_bh(&ifp->lock);
3963 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3965 ifp->state = INET6_IFADDR_STATE_DAD;
3966 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3968 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3970 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3971 idev->cnf.accept_dad > 1) &&
3972 !idev->cnf.disable_ipv6 &&
3973 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3974 struct in6_addr addr;
3976 addr.s6_addr32[0] = htonl(0xfe800000);
3977 addr.s6_addr32[1] = 0;
3979 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3980 ipv6_addr_equal(&ifp->addr, &addr)) {
3981 /* DAD failed for link-local based on MAC */
3982 idev->cnf.disable_ipv6 = 1;
3984 pr_info("%s: IPv6 being disabled!\n",
3985 ifp->idev->dev->name);
3986 disable_ipv6 = true;
3990 spin_unlock_bh(&ifp->lock);
3992 if (action == DAD_BEGIN) {
3993 addrconf_dad_begin(ifp);
3995 } else if (action == DAD_ABORT) {
3997 addrconf_dad_stop(ifp, 1);
3999 addrconf_ifdown(idev->dev, 0);
4003 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4006 write_lock_bh(&idev->lock);
4007 if (idev->dead || !(idev->if_flags & IF_READY)) {
4008 write_unlock_bh(&idev->lock);
4012 spin_lock(&ifp->lock);
4013 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4014 spin_unlock(&ifp->lock);
4015 write_unlock_bh(&idev->lock);
4019 if (ifp->dad_probes == 0) {
4020 bool send_na = false;
4023 * DAD was successful
4026 if (ifp->flags & IFA_F_TENTATIVE &&
4027 !(ifp->flags & IFA_F_OPTIMISTIC))
4029 bump_id = ifp->flags & IFA_F_TENTATIVE;
4030 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4031 spin_unlock(&ifp->lock);
4032 write_unlock_bh(&idev->lock);
4034 addrconf_dad_completed(ifp, bump_id, send_na);
4040 addrconf_mod_dad_work(ifp,
4041 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4042 spin_unlock(&ifp->lock);
4043 write_unlock_bh(&idev->lock);
4045 /* send a neighbour solicitation for our addr */
4046 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4047 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4054 /* ifp->idev must be at least read locked */
4055 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4057 struct inet6_ifaddr *ifpiter;
4058 struct inet6_dev *idev = ifp->idev;
4060 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4061 if (ifpiter->scope > IFA_LINK)
4063 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4064 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4065 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4072 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4075 struct net_device *dev = ifp->idev->dev;
4076 struct in6_addr lladdr;
4077 bool send_rs, send_mld;
4079 addrconf_del_dad_work(ifp);
4082 * Configure the address for reception. Now it is valid.
4085 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4087 /* If added prefix is link local and we are prepared to process
4088 router advertisements, start sending router solicitations.
4091 read_lock_bh(&ifp->idev->lock);
4092 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4093 send_rs = send_mld &&
4094 ipv6_accept_ra(ifp->idev) &&
4095 ifp->idev->cnf.rtr_solicits != 0 &&
4096 (dev->flags&IFF_LOOPBACK) == 0;
4097 read_unlock_bh(&ifp->idev->lock);
4099 /* While dad is in progress mld report's source address is in6_addrany.
4100 * Resend with proper ll now.
4103 ipv6_mc_dad_complete(ifp->idev);
4105 /* send unsolicited NA if enabled */
4107 (ifp->idev->cnf.ndisc_notify ||
4108 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4109 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4110 /*router=*/ !!ifp->idev->cnf.forwarding,
4111 /*solicited=*/ false, /*override=*/ true,
4117 * If a host as already performed a random delay
4118 * [...] as part of DAD [...] there is no need
4119 * to delay again before sending the first RS
4121 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4123 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4125 write_lock_bh(&ifp->idev->lock);
4126 spin_lock(&ifp->lock);
4127 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4128 ifp->idev->cnf.rtr_solicit_interval);
4129 ifp->idev->rs_probes = 1;
4130 ifp->idev->if_flags |= IF_RS_SENT;
4131 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4132 spin_unlock(&ifp->lock);
4133 write_unlock_bh(&ifp->idev->lock);
4137 rt_genid_bump_ipv6(dev_net(dev));
4139 /* Make sure that a new temporary address will be created
4140 * before this temporary address becomes deprecated.
4142 if (ifp->flags & IFA_F_TEMPORARY)
4143 addrconf_verify_rtnl();
4146 static void addrconf_dad_run(struct inet6_dev *idev)
4148 struct inet6_ifaddr *ifp;
4150 read_lock_bh(&idev->lock);
4151 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4152 spin_lock(&ifp->lock);
4153 if (ifp->flags & IFA_F_TENTATIVE &&
4154 ifp->state == INET6_IFADDR_STATE_DAD)
4155 addrconf_dad_kick(ifp);
4156 spin_unlock(&ifp->lock);
4158 read_unlock_bh(&idev->lock);
4161 #ifdef CONFIG_PROC_FS
4162 struct if6_iter_state {
4163 struct seq_net_private p;
4168 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4170 struct if6_iter_state *state = seq->private;
4171 struct net *net = seq_file_net(seq);
4172 struct inet6_ifaddr *ifa = NULL;
4175 /* initial bucket if pos is 0 */
4181 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4182 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4184 if (!net_eq(dev_net(ifa->idev->dev), net))
4186 /* sync with offset */
4187 if (p < state->offset) {
4195 /* prepare for next bucket */
4202 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4203 struct inet6_ifaddr *ifa)
4205 struct if6_iter_state *state = seq->private;
4206 struct net *net = seq_file_net(seq);
4208 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4209 if (!net_eq(dev_net(ifa->idev->dev), net))
4215 while (++state->bucket < IN6_ADDR_HSIZE) {
4217 hlist_for_each_entry_rcu(ifa,
4218 &inet6_addr_lst[state->bucket], addr_lst) {
4219 if (!net_eq(dev_net(ifa->idev->dev), net))
4229 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4233 return if6_get_first(seq, *pos);
4236 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4238 struct inet6_ifaddr *ifa;
4240 ifa = if6_get_next(seq, v);
4245 static void if6_seq_stop(struct seq_file *seq, void *v)
4251 static int if6_seq_show(struct seq_file *seq, void *v)
4253 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4254 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4256 ifp->idev->dev->ifindex,
4260 ifp->idev->dev->name);
4264 static const struct seq_operations if6_seq_ops = {
4265 .start = if6_seq_start,
4266 .next = if6_seq_next,
4267 .show = if6_seq_show,
4268 .stop = if6_seq_stop,
4271 static int if6_seq_open(struct inode *inode, struct file *file)
4273 return seq_open_net(inode, file, &if6_seq_ops,
4274 sizeof(struct if6_iter_state));
4277 static const struct file_operations if6_fops = {
4278 .open = if6_seq_open,
4280 .llseek = seq_lseek,
4281 .release = seq_release_net,
4284 static int __net_init if6_proc_net_init(struct net *net)
4286 if (!proc_create("if_inet6", 0444, net->proc_net, &if6_fops))
4291 static void __net_exit if6_proc_net_exit(struct net *net)
4293 remove_proc_entry("if_inet6", net->proc_net);
4296 static struct pernet_operations if6_proc_net_ops = {
4297 .init = if6_proc_net_init,
4298 .exit = if6_proc_net_exit,
4301 int __init if6_proc_init(void)
4303 return register_pernet_subsys(&if6_proc_net_ops);
4306 void if6_proc_exit(void)
4308 unregister_pernet_subsys(&if6_proc_net_ops);
4310 #endif /* CONFIG_PROC_FS */
4312 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4313 /* Check if address is a home address configured on any interface. */
4314 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4316 unsigned int hash = inet6_addr_hash(net, addr);
4317 struct inet6_ifaddr *ifp = NULL;
4321 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4322 if (!net_eq(dev_net(ifp->idev->dev), net))
4324 if (ipv6_addr_equal(&ifp->addr, addr) &&
4325 (ifp->flags & IFA_F_HOMEADDRESS)) {
4336 * Periodic address status verification
4339 static void addrconf_verify_rtnl(void)
4341 unsigned long now, next, next_sec, next_sched;
4342 struct inet6_ifaddr *ifp;
4349 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4351 cancel_delayed_work(&addr_chk_work);
4353 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4355 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4358 /* When setting preferred_lft to a value not zero or
4359 * infinity, while valid_lft is infinity
4360 * IFA_F_PERMANENT has a non-infinity life time.
4362 if ((ifp->flags & IFA_F_PERMANENT) &&
4363 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4366 spin_lock(&ifp->lock);
4367 /* We try to batch several events at once. */
4368 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4370 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4371 age >= ifp->valid_lft) {
4372 spin_unlock(&ifp->lock);
4376 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4377 spin_unlock(&ifp->lock);
4379 } else if (age >= ifp->prefered_lft) {
4380 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4383 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4385 ifp->flags |= IFA_F_DEPRECATED;
4388 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4389 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4390 next = ifp->tstamp + ifp->valid_lft * HZ;
4392 spin_unlock(&ifp->lock);
4397 ipv6_ifa_notify(0, ifp);
4401 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4402 !(ifp->flags&IFA_F_TENTATIVE)) {
4403 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4404 ifp->idev->cnf.dad_transmits *
4405 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4407 if (age >= ifp->prefered_lft - regen_advance) {
4408 struct inet6_ifaddr *ifpub = ifp->ifpub;
4409 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4410 next = ifp->tstamp + ifp->prefered_lft * HZ;
4411 if (!ifp->regen_count && ifpub) {
4414 in6_ifa_hold(ifpub);
4415 spin_unlock(&ifp->lock);
4417 spin_lock(&ifpub->lock);
4418 ifpub->regen_count = 0;
4419 spin_unlock(&ifpub->lock);
4420 rcu_read_unlock_bh();
4421 ipv6_create_tempaddr(ifpub, ifp, true);
4427 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4428 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4429 spin_unlock(&ifp->lock);
4431 /* ifp->prefered_lft <= ifp->valid_lft */
4432 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4433 next = ifp->tstamp + ifp->prefered_lft * HZ;
4434 spin_unlock(&ifp->lock);
4439 next_sec = round_jiffies_up(next);
4442 /* If rounded timeout is accurate enough, accept it. */
4443 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4444 next_sched = next_sec;
4446 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4447 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4448 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4450 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4451 now, next, next_sec, next_sched);
4452 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4453 rcu_read_unlock_bh();
4456 static void addrconf_verify_work(struct work_struct *w)
4459 addrconf_verify_rtnl();
4463 static void addrconf_verify(void)
4465 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4468 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4469 struct in6_addr **peer_pfx)
4471 struct in6_addr *pfx = NULL;
4476 pfx = nla_data(addr);
4479 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4481 pfx = nla_data(local);
4487 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4488 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4489 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4490 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4491 [IFA_FLAGS] = { .len = sizeof(u32) },
4495 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4496 struct netlink_ext_ack *extack)
4498 struct net *net = sock_net(skb->sk);
4499 struct ifaddrmsg *ifm;
4500 struct nlattr *tb[IFA_MAX+1];
4501 struct in6_addr *pfx, *peer_pfx;
4505 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4510 ifm = nlmsg_data(nlh);
4511 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4515 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4517 /* We ignore other flags so far. */
4518 ifa_flags &= IFA_F_MANAGETEMPADDR;
4520 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4521 ifm->ifa_prefixlen);
4524 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4525 u32 prefered_lft, u32 valid_lft)
4529 unsigned long timeout;
4530 bool was_managetempaddr;
4531 bool had_prefixroute;
4535 if (!valid_lft || (prefered_lft > valid_lft))
4538 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4539 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4542 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4543 ifa_flags &= ~IFA_F_OPTIMISTIC;
4545 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4546 if (addrconf_finite_timeout(timeout)) {
4547 expires = jiffies_to_clock_t(timeout * HZ);
4548 valid_lft = timeout;
4549 flags = RTF_EXPIRES;
4553 ifa_flags |= IFA_F_PERMANENT;
4556 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4557 if (addrconf_finite_timeout(timeout)) {
4559 ifa_flags |= IFA_F_DEPRECATED;
4560 prefered_lft = timeout;
4563 spin_lock_bh(&ifp->lock);
4564 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4565 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4566 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4567 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4568 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4569 IFA_F_NOPREFIXROUTE);
4570 ifp->flags |= ifa_flags;
4571 ifp->tstamp = jiffies;
4572 ifp->valid_lft = valid_lft;
4573 ifp->prefered_lft = prefered_lft;
4575 spin_unlock_bh(&ifp->lock);
4576 if (!(ifp->flags&IFA_F_TENTATIVE))
4577 ipv6_ifa_notify(0, ifp);
4579 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4580 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4581 ifp->idev->dev, expires, flags,
4583 } else if (had_prefixroute) {
4584 enum cleanup_prefix_rt_t action;
4585 unsigned long rt_expires;
4587 write_lock_bh(&ifp->idev->lock);
4588 action = check_cleanup_prefix_route(ifp, &rt_expires);
4589 write_unlock_bh(&ifp->idev->lock);
4591 if (action != CLEANUP_PREFIX_RT_NOP) {
4592 cleanup_prefix_route(ifp, rt_expires,
4593 action == CLEANUP_PREFIX_RT_DEL);
4597 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4598 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4599 valid_lft = prefered_lft = 0;
4600 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4601 !was_managetempaddr, jiffies);
4604 addrconf_verify_rtnl();
4610 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4611 struct netlink_ext_ack *extack)
4613 struct net *net = sock_net(skb->sk);
4614 struct ifaddrmsg *ifm;
4615 struct nlattr *tb[IFA_MAX+1];
4616 struct in6_addr *pfx, *peer_pfx;
4617 struct inet6_ifaddr *ifa;
4618 struct net_device *dev;
4619 struct inet6_dev *idev;
4620 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4624 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4629 ifm = nlmsg_data(nlh);
4630 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4634 if (tb[IFA_CACHEINFO]) {
4635 struct ifa_cacheinfo *ci;
4637 ci = nla_data(tb[IFA_CACHEINFO]);
4638 valid_lft = ci->ifa_valid;
4639 preferred_lft = ci->ifa_prefered;
4641 preferred_lft = INFINITY_LIFE_TIME;
4642 valid_lft = INFINITY_LIFE_TIME;
4645 dev = __dev_get_by_index(net, ifm->ifa_index);
4649 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4651 /* We ignore other flags so far. */
4652 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4653 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4655 idev = ipv6_find_idev(dev);
4657 return PTR_ERR(idev);
4659 if (!ipv6_allow_optimistic_dad(net, idev))
4660 ifa_flags &= ~IFA_F_OPTIMISTIC;
4662 if (ifa_flags & IFA_F_NODAD && ifa_flags & IFA_F_OPTIMISTIC) {
4663 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4667 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4670 * It would be best to check for !NLM_F_CREATE here but
4671 * userspace already relies on not having to provide this.
4673 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4674 ifm->ifa_prefixlen, ifa_flags,
4675 preferred_lft, valid_lft, extack);
4678 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4679 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4682 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4689 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4690 u8 scope, int ifindex)
4692 struct ifaddrmsg *ifm;
4694 ifm = nlmsg_data(nlh);
4695 ifm->ifa_family = AF_INET6;
4696 ifm->ifa_prefixlen = prefixlen;
4697 ifm->ifa_flags = flags;
4698 ifm->ifa_scope = scope;
4699 ifm->ifa_index = ifindex;
4702 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4703 unsigned long tstamp, u32 preferred, u32 valid)
4705 struct ifa_cacheinfo ci;
4707 ci.cstamp = cstamp_delta(cstamp);
4708 ci.tstamp = cstamp_delta(tstamp);
4709 ci.ifa_prefered = preferred;
4710 ci.ifa_valid = valid;
4712 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4715 static inline int rt_scope(int ifa_scope)
4717 if (ifa_scope & IFA_HOST)
4718 return RT_SCOPE_HOST;
4719 else if (ifa_scope & IFA_LINK)
4720 return RT_SCOPE_LINK;
4721 else if (ifa_scope & IFA_SITE)
4722 return RT_SCOPE_SITE;
4724 return RT_SCOPE_UNIVERSE;
4727 static inline int inet6_ifaddr_msgsize(void)
4729 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4730 + nla_total_size(16) /* IFA_LOCAL */
4731 + nla_total_size(16) /* IFA_ADDRESS */
4732 + nla_total_size(sizeof(struct ifa_cacheinfo))
4733 + nla_total_size(4) /* IFA_FLAGS */;
4736 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4737 u32 portid, u32 seq, int event, unsigned int flags)
4739 struct nlmsghdr *nlh;
4740 u32 preferred, valid;
4742 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4746 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4747 ifa->idev->dev->ifindex);
4749 if (!((ifa->flags&IFA_F_PERMANENT) &&
4750 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4751 preferred = ifa->prefered_lft;
4752 valid = ifa->valid_lft;
4753 if (preferred != INFINITY_LIFE_TIME) {
4754 long tval = (jiffies - ifa->tstamp)/HZ;
4755 if (preferred > tval)
4759 if (valid != INFINITY_LIFE_TIME) {
4767 preferred = INFINITY_LIFE_TIME;
4768 valid = INFINITY_LIFE_TIME;
4771 if (!ipv6_addr_any(&ifa->peer_addr)) {
4772 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4773 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4776 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4779 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4782 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4785 nlmsg_end(skb, nlh);
4789 nlmsg_cancel(skb, nlh);
4793 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4794 u32 portid, u32 seq, int event, u16 flags)
4796 struct nlmsghdr *nlh;
4797 u8 scope = RT_SCOPE_UNIVERSE;
4798 int ifindex = ifmca->idev->dev->ifindex;
4800 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4801 scope = RT_SCOPE_SITE;
4803 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4807 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4808 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4809 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4810 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4811 nlmsg_cancel(skb, nlh);
4815 nlmsg_end(skb, nlh);
4819 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4820 u32 portid, u32 seq, int event, unsigned int flags)
4822 struct nlmsghdr *nlh;
4823 u8 scope = RT_SCOPE_UNIVERSE;
4824 int ifindex = ifaca->aca_idev->dev->ifindex;
4826 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4827 scope = RT_SCOPE_SITE;
4829 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4833 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4834 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4835 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4836 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4837 nlmsg_cancel(skb, nlh);
4841 nlmsg_end(skb, nlh);
4851 /* called with rcu_read_lock() */
4852 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4853 struct netlink_callback *cb, enum addr_type_t type,
4854 int s_ip_idx, int *p_ip_idx)
4856 struct ifmcaddr6 *ifmca;
4857 struct ifacaddr6 *ifaca;
4859 int ip_idx = *p_ip_idx;
4861 read_lock_bh(&idev->lock);
4863 case UNICAST_ADDR: {
4864 struct inet6_ifaddr *ifa;
4866 /* unicast address incl. temp addr */
4867 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4868 if (++ip_idx < s_ip_idx)
4870 err = inet6_fill_ifaddr(skb, ifa,
4871 NETLINK_CB(cb->skb).portid,
4877 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4881 case MULTICAST_ADDR:
4882 /* multicast address */
4883 for (ifmca = idev->mc_list; ifmca;
4884 ifmca = ifmca->next, ip_idx++) {
4885 if (ip_idx < s_ip_idx)
4887 err = inet6_fill_ifmcaddr(skb, ifmca,
4888 NETLINK_CB(cb->skb).portid,
4897 /* anycast address */
4898 for (ifaca = idev->ac_list; ifaca;
4899 ifaca = ifaca->aca_next, ip_idx++) {
4900 if (ip_idx < s_ip_idx)
4902 err = inet6_fill_ifacaddr(skb, ifaca,
4903 NETLINK_CB(cb->skb).portid,
4914 read_unlock_bh(&idev->lock);
4919 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4920 enum addr_type_t type)
4922 struct net *net = sock_net(skb->sk);
4925 int s_idx, s_ip_idx;
4926 struct net_device *dev;
4927 struct inet6_dev *idev;
4928 struct hlist_head *head;
4931 s_idx = idx = cb->args[1];
4932 s_ip_idx = ip_idx = cb->args[2];
4935 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4936 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4938 head = &net->dev_index_head[h];
4939 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4942 if (h > s_h || idx > s_idx)
4945 idev = __in6_dev_get(dev);
4949 if (in6_dump_addrs(idev, skb, cb, type,
4950 s_ip_idx, &ip_idx) < 0)
4960 cb->args[2] = ip_idx;
4965 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4967 enum addr_type_t type = UNICAST_ADDR;
4969 return inet6_dump_addr(skb, cb, type);
4972 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4974 enum addr_type_t type = MULTICAST_ADDR;
4976 return inet6_dump_addr(skb, cb, type);
4980 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4982 enum addr_type_t type = ANYCAST_ADDR;
4984 return inet6_dump_addr(skb, cb, type);
4987 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4988 struct netlink_ext_ack *extack)
4990 struct net *net = sock_net(in_skb->sk);
4991 struct ifaddrmsg *ifm;
4992 struct nlattr *tb[IFA_MAX+1];
4993 struct in6_addr *addr = NULL, *peer;
4994 struct net_device *dev = NULL;
4995 struct inet6_ifaddr *ifa;
4996 struct sk_buff *skb;
4999 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5004 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5008 ifm = nlmsg_data(nlh);
5010 dev = dev_get_by_index(net, ifm->ifa_index);
5012 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5014 err = -EADDRNOTAVAIL;
5018 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5024 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5025 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5027 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5028 WARN_ON(err == -EMSGSIZE);
5032 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5041 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5043 struct sk_buff *skb;
5044 struct net *net = dev_net(ifa->idev->dev);
5047 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5051 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5053 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5054 WARN_ON(err == -EMSGSIZE);
5058 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5062 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5065 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5066 __s32 *array, int bytes)
5068 BUG_ON(bytes < (DEVCONF_MAX * 4));
5070 memset(array, 0, bytes);
5071 array[DEVCONF_FORWARDING] = cnf->forwarding;
5072 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5073 array[DEVCONF_MTU6] = cnf->mtu6;
5074 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5075 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5076 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5077 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5078 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5079 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5080 jiffies_to_msecs(cnf->rtr_solicit_interval);
5081 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5082 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5083 array[DEVCONF_RTR_SOLICIT_DELAY] =
5084 jiffies_to_msecs(cnf->rtr_solicit_delay);
5085 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5086 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5087 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5088 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5089 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5090 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5091 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5092 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5093 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5094 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5095 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5096 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5097 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5098 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5099 #ifdef CONFIG_IPV6_ROUTER_PREF
5100 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5101 array[DEVCONF_RTR_PROBE_INTERVAL] =
5102 jiffies_to_msecs(cnf->rtr_probe_interval);
5103 #ifdef CONFIG_IPV6_ROUTE_INFO
5104 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5105 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5108 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5109 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5110 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5111 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5112 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5114 #ifdef CONFIG_IPV6_MROUTE
5115 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5117 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5118 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5119 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5120 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5121 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5122 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5123 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5124 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5125 /* we omit DEVCONF_STABLE_SECRET for now */
5126 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5127 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5128 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5129 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5130 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5131 #ifdef CONFIG_IPV6_SEG6_HMAC
5132 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5134 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5135 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5136 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5137 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5140 static inline size_t inet6_ifla6_size(void)
5142 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5143 + nla_total_size(sizeof(struct ifla_cacheinfo))
5144 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5145 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5146 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5147 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5150 static inline size_t inet6_if_nlmsg_size(void)
5152 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5153 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5154 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5155 + nla_total_size(4) /* IFLA_MTU */
5156 + nla_total_size(4) /* IFLA_LINK */
5157 + nla_total_size(1) /* IFLA_OPERSTATE */
5158 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5161 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5165 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5168 /* Use put_unaligned() because stats may not be aligned for u64. */
5169 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5170 for (i = 1; i < ICMP6_MIB_MAX; i++)
5171 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5173 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5176 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5177 int bytes, size_t syncpoff)
5180 u64 buff[IPSTATS_MIB_MAX];
5181 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5185 memset(buff, 0, sizeof(buff));
5186 buff[0] = IPSTATS_MIB_MAX;
5188 for_each_possible_cpu(c) {
5189 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5190 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5193 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5194 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5197 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5201 case IFLA_INET6_STATS:
5202 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5203 offsetof(struct ipstats_mib, syncp));
5205 case IFLA_INET6_ICMP6STATS:
5206 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5211 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5212 u32 ext_filter_mask)
5215 struct ifla_cacheinfo ci;
5217 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5218 goto nla_put_failure;
5219 ci.max_reasm_len = IPV6_MAXPLEN;
5220 ci.tstamp = cstamp_delta(idev->tstamp);
5221 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5222 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5223 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5224 goto nla_put_failure;
5225 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5227 goto nla_put_failure;
5228 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5230 /* XXX - MC not implemented */
5232 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5235 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5237 goto nla_put_failure;
5238 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5240 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5242 goto nla_put_failure;
5243 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5245 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5247 goto nla_put_failure;
5249 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5250 goto nla_put_failure;
5252 read_lock_bh(&idev->lock);
5253 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5254 read_unlock_bh(&idev->lock);
5262 static size_t inet6_get_link_af_size(const struct net_device *dev,
5263 u32 ext_filter_mask)
5265 if (!__in6_dev_get(dev))
5268 return inet6_ifla6_size();
5271 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5272 u32 ext_filter_mask)
5274 struct inet6_dev *idev = __in6_dev_get(dev);
5279 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5285 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5287 struct inet6_ifaddr *ifp;
5288 struct net_device *dev = idev->dev;
5289 bool clear_token, update_rs = false;
5290 struct in6_addr ll_addr;
5296 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5298 if (!ipv6_accept_ra(idev))
5300 if (idev->cnf.rtr_solicits == 0)
5303 write_lock_bh(&idev->lock);
5305 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5306 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5308 write_unlock_bh(&idev->lock);
5310 clear_token = ipv6_addr_any(token);
5314 if (!idev->dead && (idev->if_flags & IF_READY) &&
5315 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5316 IFA_F_OPTIMISTIC)) {
5317 /* If we're not ready, then normal ifup will take care
5318 * of this. Otherwise, we need to request our rs here.
5320 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5325 write_lock_bh(&idev->lock);
5328 idev->if_flags |= IF_RS_SENT;
5329 idev->rs_interval = rfc3315_s14_backoff_init(
5330 idev->cnf.rtr_solicit_interval);
5331 idev->rs_probes = 1;
5332 addrconf_mod_rs_timer(idev, idev->rs_interval);
5335 /* Well, that's kinda nasty ... */
5336 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5337 spin_lock(&ifp->lock);
5338 if (ifp->tokenized) {
5340 ifp->prefered_lft = 0;
5342 spin_unlock(&ifp->lock);
5345 write_unlock_bh(&idev->lock);
5346 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5347 addrconf_verify_rtnl();
5351 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5352 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5353 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5356 static int inet6_validate_link_af(const struct net_device *dev,
5357 const struct nlattr *nla)
5359 struct nlattr *tb[IFLA_INET6_MAX + 1];
5361 if (dev && !__in6_dev_get(dev))
5362 return -EAFNOSUPPORT;
5364 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5368 static int check_addr_gen_mode(int mode)
5370 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5371 mode != IN6_ADDR_GEN_MODE_NONE &&
5372 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5373 mode != IN6_ADDR_GEN_MODE_RANDOM)
5378 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5381 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5382 !idev->cnf.stable_secret.initialized &&
5383 !net->ipv6.devconf_dflt->stable_secret.initialized)
5388 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5391 struct inet6_dev *idev = __in6_dev_get(dev);
5392 struct nlattr *tb[IFLA_INET6_MAX + 1];
5395 return -EAFNOSUPPORT;
5397 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5400 if (tb[IFLA_INET6_TOKEN]) {
5401 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5406 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5407 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5409 if (check_addr_gen_mode(mode) < 0 ||
5410 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5413 idev->cnf.addr_gen_mode = mode;
5420 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5421 u32 portid, u32 seq, int event, unsigned int flags)
5423 struct net_device *dev = idev->dev;
5424 struct ifinfomsg *hdr;
5425 struct nlmsghdr *nlh;
5428 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5432 hdr = nlmsg_data(nlh);
5433 hdr->ifi_family = AF_INET6;
5435 hdr->ifi_type = dev->type;
5436 hdr->ifi_index = dev->ifindex;
5437 hdr->ifi_flags = dev_get_flags(dev);
5438 hdr->ifi_change = 0;
5440 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5442 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5443 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5444 (dev->ifindex != dev_get_iflink(dev) &&
5445 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5446 nla_put_u8(skb, IFLA_OPERSTATE,
5447 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5448 goto nla_put_failure;
5449 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5451 goto nla_put_failure;
5453 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5454 goto nla_put_failure;
5456 nla_nest_end(skb, protoinfo);
5457 nlmsg_end(skb, nlh);
5461 nlmsg_cancel(skb, nlh);
5465 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5467 struct net *net = sock_net(skb->sk);
5470 struct net_device *dev;
5471 struct inet6_dev *idev;
5472 struct hlist_head *head;
5475 s_idx = cb->args[1];
5478 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5480 head = &net->dev_index_head[h];
5481 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5484 idev = __in6_dev_get(dev);
5487 if (inet6_fill_ifinfo(skb, idev,
5488 NETLINK_CB(cb->skb).portid,
5490 RTM_NEWLINK, NLM_F_MULTI) < 0)
5504 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5506 struct sk_buff *skb;
5507 struct net *net = dev_net(idev->dev);
5510 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5514 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5516 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5517 WARN_ON(err == -EMSGSIZE);
5521 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5525 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5528 static inline size_t inet6_prefix_nlmsg_size(void)
5530 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5531 + nla_total_size(sizeof(struct in6_addr))
5532 + nla_total_size(sizeof(struct prefix_cacheinfo));
5535 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5536 struct prefix_info *pinfo, u32 portid, u32 seq,
5537 int event, unsigned int flags)
5539 struct prefixmsg *pmsg;
5540 struct nlmsghdr *nlh;
5541 struct prefix_cacheinfo ci;
5543 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5547 pmsg = nlmsg_data(nlh);
5548 pmsg->prefix_family = AF_INET6;
5549 pmsg->prefix_pad1 = 0;
5550 pmsg->prefix_pad2 = 0;
5551 pmsg->prefix_ifindex = idev->dev->ifindex;
5552 pmsg->prefix_len = pinfo->prefix_len;
5553 pmsg->prefix_type = pinfo->type;
5554 pmsg->prefix_pad3 = 0;
5555 pmsg->prefix_flags = 0;
5557 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5558 if (pinfo->autoconf)
5559 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5561 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5562 goto nla_put_failure;
5563 ci.preferred_time = ntohl(pinfo->prefered);
5564 ci.valid_time = ntohl(pinfo->valid);
5565 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5566 goto nla_put_failure;
5567 nlmsg_end(skb, nlh);
5571 nlmsg_cancel(skb, nlh);
5575 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5576 struct prefix_info *pinfo)
5578 struct sk_buff *skb;
5579 struct net *net = dev_net(idev->dev);
5582 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5586 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5588 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5589 WARN_ON(err == -EMSGSIZE);
5593 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5597 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5600 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5602 struct net *net = dev_net(ifp->idev->dev);
5607 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5612 * If the address was optimistic
5613 * we inserted the route at the start of
5614 * our DAD process, so we don't need
5617 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5618 ip6_ins_rt(net, ifp->rt);
5619 if (ifp->idev->cnf.forwarding)
5620 addrconf_join_anycast(ifp);
5621 if (!ipv6_addr_any(&ifp->peer_addr))
5622 addrconf_prefix_route(&ifp->peer_addr, 128,
5623 ifp->idev->dev, 0, 0,
5627 if (ifp->idev->cnf.forwarding)
5628 addrconf_leave_anycast(ifp);
5629 addrconf_leave_solict(ifp->idev, &ifp->addr);
5630 if (!ipv6_addr_any(&ifp->peer_addr)) {
5631 struct rt6_info *rt;
5633 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5634 ifp->idev->dev, 0, 0);
5636 ip6_del_rt(net, rt);
5639 if (dst_hold_safe(&ifp->rt->dst))
5640 ip6_del_rt(net, ifp->rt);
5642 rt_genid_bump_ipv6(net);
5645 atomic_inc(&net->ipv6.dev_addr_genid);
5648 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5651 if (likely(ifp->idev->dead == 0))
5652 __ipv6_ifa_notify(event, ifp);
5653 rcu_read_unlock_bh();
5656 #ifdef CONFIG_SYSCTL
5659 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5660 void __user *buffer, size_t *lenp, loff_t *ppos)
5662 int *valp = ctl->data;
5665 struct ctl_table lctl;
5669 * ctl->data points to idev->cnf.forwarding, we should
5670 * not modify it until we get the rtnl lock.
5675 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5678 ret = addrconf_fixup_forwarding(ctl, valp, val);
5685 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5686 void __user *buffer, size_t *lenp, loff_t *ppos)
5688 struct inet6_dev *idev = ctl->extra1;
5689 int min_mtu = IPV6_MIN_MTU;
5690 struct ctl_table lctl;
5693 lctl.extra1 = &min_mtu;
5694 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5696 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5699 static void dev_disable_change(struct inet6_dev *idev)
5701 struct netdev_notifier_info info;
5703 if (!idev || !idev->dev)
5706 netdev_notifier_info_init(&info, idev->dev);
5707 if (idev->cnf.disable_ipv6)
5708 addrconf_notify(NULL, NETDEV_DOWN, &info);
5710 addrconf_notify(NULL, NETDEV_UP, &info);
5713 static void addrconf_disable_change(struct net *net, __s32 newf)
5715 struct net_device *dev;
5716 struct inet6_dev *idev;
5718 for_each_netdev(net, dev) {
5719 idev = __in6_dev_get(dev);
5721 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5722 idev->cnf.disable_ipv6 = newf;
5724 dev_disable_change(idev);
5729 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5734 if (!rtnl_trylock())
5735 return restart_syscall();
5737 net = (struct net *)table->extra2;
5741 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5746 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5747 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5748 addrconf_disable_change(net, newf);
5749 } else if ((!newf) ^ (!old))
5750 dev_disable_change((struct inet6_dev *)table->extra1);
5757 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5758 void __user *buffer, size_t *lenp, loff_t *ppos)
5760 int *valp = ctl->data;
5763 struct ctl_table lctl;
5767 * ctl->data points to idev->cnf.disable_ipv6, we should
5768 * not modify it until we get the rtnl lock.
5773 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5776 ret = addrconf_disable_ipv6(ctl, valp, val);
5783 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5784 void __user *buffer, size_t *lenp, loff_t *ppos)
5786 int *valp = ctl->data;
5791 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5794 if (write && old != new) {
5795 struct net *net = ctl->extra2;
5797 if (!rtnl_trylock())
5798 return restart_syscall();
5800 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5801 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5802 NETCONFA_PROXY_NEIGH,
5803 NETCONFA_IFINDEX_DEFAULT,
5804 net->ipv6.devconf_dflt);
5805 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5806 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5807 NETCONFA_PROXY_NEIGH,
5808 NETCONFA_IFINDEX_ALL,
5809 net->ipv6.devconf_all);
5811 struct inet6_dev *idev = ctl->extra1;
5813 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5814 NETCONFA_PROXY_NEIGH,
5824 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5825 void __user *buffer, size_t *lenp,
5830 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5831 struct net *net = (struct net *)ctl->extra2;
5833 if (!rtnl_trylock())
5834 return restart_syscall();
5836 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5839 new_val = *((int *)ctl->data);
5841 if (check_addr_gen_mode(new_val) < 0) {
5846 /* request for default */
5847 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5848 ipv6_devconf_dflt.addr_gen_mode = new_val;
5850 /* request for individual net device */
5855 if (check_stable_privacy(idev, net, new_val) < 0) {
5860 if (idev->cnf.addr_gen_mode != new_val) {
5861 idev->cnf.addr_gen_mode = new_val;
5862 addrconf_dev_config(idev->dev);
5873 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5874 void __user *buffer, size_t *lenp,
5878 struct in6_addr addr;
5879 char str[IPV6_MAX_STRLEN];
5880 struct ctl_table lctl = *ctl;
5881 struct net *net = ctl->extra2;
5882 struct ipv6_stable_secret *secret = ctl->data;
5884 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5887 lctl.maxlen = IPV6_MAX_STRLEN;
5890 if (!rtnl_trylock())
5891 return restart_syscall();
5893 if (!write && !secret->initialized) {
5898 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5899 if (err >= sizeof(str)) {
5904 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5908 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5913 secret->initialized = true;
5914 secret->secret = addr;
5916 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5917 struct net_device *dev;
5919 for_each_netdev(net, dev) {
5920 struct inet6_dev *idev = __in6_dev_get(dev);
5923 idev->cnf.addr_gen_mode =
5924 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5928 struct inet6_dev *idev = ctl->extra1;
5930 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5940 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5942 void __user *buffer,
5946 int *valp = ctl->data;
5949 struct ctl_table lctl;
5952 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5953 * we should not modify it until we get the rtnl lock.
5958 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5961 ret = addrconf_fixup_linkdown(ctl, valp, val);
5968 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5972 rt->dst.flags |= DST_NOPOLICY;
5974 rt->dst.flags &= ~DST_NOPOLICY;
5979 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5981 struct inet6_ifaddr *ifa;
5983 read_lock_bh(&idev->lock);
5984 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5985 spin_lock(&ifa->lock);
5987 struct rt6_info *rt = ifa->rt;
5991 ifa->rt->dst_nopolicy = val ? true : false;
5992 if (rt->rt6i_pcpu) {
5993 for_each_possible_cpu(cpu) {
5994 struct rt6_info **rtp;
5996 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5997 addrconf_set_nopolicy(*rtp, val);
6002 spin_unlock(&ifa->lock);
6004 read_unlock_bh(&idev->lock);
6008 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6010 struct inet6_dev *idev;
6013 if (!rtnl_trylock())
6014 return restart_syscall();
6018 net = (struct net *)ctl->extra2;
6019 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6024 if (valp == &net->ipv6.devconf_all->disable_policy) {
6025 struct net_device *dev;
6027 for_each_netdev(net, dev) {
6028 idev = __in6_dev_get(dev);
6030 addrconf_disable_policy_idev(idev, val);
6033 idev = (struct inet6_dev *)ctl->extra1;
6034 addrconf_disable_policy_idev(idev, val);
6042 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6043 void __user *buffer, size_t *lenp,
6046 int *valp = ctl->data;
6049 struct ctl_table lctl;
6054 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6056 if (write && (*valp != val))
6057 ret = addrconf_disable_policy(ctl, valp, val);
6065 static int minus_one = -1;
6066 static const int zero = 0;
6067 static const int one = 1;
6068 static const int two_five_five = 255;
6070 static const struct ctl_table addrconf_sysctl[] = {
6072 .procname = "forwarding",
6073 .data = &ipv6_devconf.forwarding,
6074 .maxlen = sizeof(int),
6076 .proc_handler = addrconf_sysctl_forward,
6079 .procname = "hop_limit",
6080 .data = &ipv6_devconf.hop_limit,
6081 .maxlen = sizeof(int),
6083 .proc_handler = proc_dointvec_minmax,
6084 .extra1 = (void *)&one,
6085 .extra2 = (void *)&two_five_five,
6089 .data = &ipv6_devconf.mtu6,
6090 .maxlen = sizeof(int),
6092 .proc_handler = addrconf_sysctl_mtu,
6095 .procname = "accept_ra",
6096 .data = &ipv6_devconf.accept_ra,
6097 .maxlen = sizeof(int),
6099 .proc_handler = proc_dointvec,
6102 .procname = "accept_redirects",
6103 .data = &ipv6_devconf.accept_redirects,
6104 .maxlen = sizeof(int),
6106 .proc_handler = proc_dointvec,
6109 .procname = "autoconf",
6110 .data = &ipv6_devconf.autoconf,
6111 .maxlen = sizeof(int),
6113 .proc_handler = proc_dointvec,
6116 .procname = "dad_transmits",
6117 .data = &ipv6_devconf.dad_transmits,
6118 .maxlen = sizeof(int),
6120 .proc_handler = proc_dointvec,
6123 .procname = "router_solicitations",
6124 .data = &ipv6_devconf.rtr_solicits,
6125 .maxlen = sizeof(int),
6127 .proc_handler = proc_dointvec_minmax,
6128 .extra1 = &minus_one,
6131 .procname = "router_solicitation_interval",
6132 .data = &ipv6_devconf.rtr_solicit_interval,
6133 .maxlen = sizeof(int),
6135 .proc_handler = proc_dointvec_jiffies,
6138 .procname = "router_solicitation_max_interval",
6139 .data = &ipv6_devconf.rtr_solicit_max_interval,
6140 .maxlen = sizeof(int),
6142 .proc_handler = proc_dointvec_jiffies,
6145 .procname = "router_solicitation_delay",
6146 .data = &ipv6_devconf.rtr_solicit_delay,
6147 .maxlen = sizeof(int),
6149 .proc_handler = proc_dointvec_jiffies,
6152 .procname = "force_mld_version",
6153 .data = &ipv6_devconf.force_mld_version,
6154 .maxlen = sizeof(int),
6156 .proc_handler = proc_dointvec,
6159 .procname = "mldv1_unsolicited_report_interval",
6161 &ipv6_devconf.mldv1_unsolicited_report_interval,
6162 .maxlen = sizeof(int),
6164 .proc_handler = proc_dointvec_ms_jiffies,
6167 .procname = "mldv2_unsolicited_report_interval",
6169 &ipv6_devconf.mldv2_unsolicited_report_interval,
6170 .maxlen = sizeof(int),
6172 .proc_handler = proc_dointvec_ms_jiffies,
6175 .procname = "use_tempaddr",
6176 .data = &ipv6_devconf.use_tempaddr,
6177 .maxlen = sizeof(int),
6179 .proc_handler = proc_dointvec,
6182 .procname = "temp_valid_lft",
6183 .data = &ipv6_devconf.temp_valid_lft,
6184 .maxlen = sizeof(int),
6186 .proc_handler = proc_dointvec,
6189 .procname = "temp_prefered_lft",
6190 .data = &ipv6_devconf.temp_prefered_lft,
6191 .maxlen = sizeof(int),
6193 .proc_handler = proc_dointvec,
6196 .procname = "regen_max_retry",
6197 .data = &ipv6_devconf.regen_max_retry,
6198 .maxlen = sizeof(int),
6200 .proc_handler = proc_dointvec,
6203 .procname = "max_desync_factor",
6204 .data = &ipv6_devconf.max_desync_factor,
6205 .maxlen = sizeof(int),
6207 .proc_handler = proc_dointvec,
6210 .procname = "max_addresses",
6211 .data = &ipv6_devconf.max_addresses,
6212 .maxlen = sizeof(int),
6214 .proc_handler = proc_dointvec,
6217 .procname = "accept_ra_defrtr",
6218 .data = &ipv6_devconf.accept_ra_defrtr,
6219 .maxlen = sizeof(int),
6221 .proc_handler = proc_dointvec,
6224 .procname = "accept_ra_min_hop_limit",
6225 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6226 .maxlen = sizeof(int),
6228 .proc_handler = proc_dointvec,
6231 .procname = "accept_ra_pinfo",
6232 .data = &ipv6_devconf.accept_ra_pinfo,
6233 .maxlen = sizeof(int),
6235 .proc_handler = proc_dointvec,
6237 #ifdef CONFIG_IPV6_ROUTER_PREF
6239 .procname = "accept_ra_rtr_pref",
6240 .data = &ipv6_devconf.accept_ra_rtr_pref,
6241 .maxlen = sizeof(int),
6243 .proc_handler = proc_dointvec,
6246 .procname = "router_probe_interval",
6247 .data = &ipv6_devconf.rtr_probe_interval,
6248 .maxlen = sizeof(int),
6250 .proc_handler = proc_dointvec_jiffies,
6252 #ifdef CONFIG_IPV6_ROUTE_INFO
6254 .procname = "accept_ra_rt_info_min_plen",
6255 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6256 .maxlen = sizeof(int),
6258 .proc_handler = proc_dointvec,
6261 .procname = "accept_ra_rt_info_max_plen",
6262 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6263 .maxlen = sizeof(int),
6265 .proc_handler = proc_dointvec,
6270 .procname = "proxy_ndp",
6271 .data = &ipv6_devconf.proxy_ndp,
6272 .maxlen = sizeof(int),
6274 .proc_handler = addrconf_sysctl_proxy_ndp,
6277 .procname = "accept_source_route",
6278 .data = &ipv6_devconf.accept_source_route,
6279 .maxlen = sizeof(int),
6281 .proc_handler = proc_dointvec,
6283 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6285 .procname = "optimistic_dad",
6286 .data = &ipv6_devconf.optimistic_dad,
6287 .maxlen = sizeof(int),
6289 .proc_handler = proc_dointvec,
6292 .procname = "use_optimistic",
6293 .data = &ipv6_devconf.use_optimistic,
6294 .maxlen = sizeof(int),
6296 .proc_handler = proc_dointvec,
6299 #ifdef CONFIG_IPV6_MROUTE
6301 .procname = "mc_forwarding",
6302 .data = &ipv6_devconf.mc_forwarding,
6303 .maxlen = sizeof(int),
6305 .proc_handler = proc_dointvec,
6309 .procname = "disable_ipv6",
6310 .data = &ipv6_devconf.disable_ipv6,
6311 .maxlen = sizeof(int),
6313 .proc_handler = addrconf_sysctl_disable,
6316 .procname = "accept_dad",
6317 .data = &ipv6_devconf.accept_dad,
6318 .maxlen = sizeof(int),
6320 .proc_handler = proc_dointvec,
6323 .procname = "force_tllao",
6324 .data = &ipv6_devconf.force_tllao,
6325 .maxlen = sizeof(int),
6327 .proc_handler = proc_dointvec
6330 .procname = "ndisc_notify",
6331 .data = &ipv6_devconf.ndisc_notify,
6332 .maxlen = sizeof(int),
6334 .proc_handler = proc_dointvec
6337 .procname = "suppress_frag_ndisc",
6338 .data = &ipv6_devconf.suppress_frag_ndisc,
6339 .maxlen = sizeof(int),
6341 .proc_handler = proc_dointvec
6344 .procname = "accept_ra_from_local",
6345 .data = &ipv6_devconf.accept_ra_from_local,
6346 .maxlen = sizeof(int),
6348 .proc_handler = proc_dointvec,
6351 .procname = "accept_ra_mtu",
6352 .data = &ipv6_devconf.accept_ra_mtu,
6353 .maxlen = sizeof(int),
6355 .proc_handler = proc_dointvec,
6358 .procname = "stable_secret",
6359 .data = &ipv6_devconf.stable_secret,
6360 .maxlen = IPV6_MAX_STRLEN,
6362 .proc_handler = addrconf_sysctl_stable_secret,
6365 .procname = "use_oif_addrs_only",
6366 .data = &ipv6_devconf.use_oif_addrs_only,
6367 .maxlen = sizeof(int),
6369 .proc_handler = proc_dointvec,
6372 .procname = "ignore_routes_with_linkdown",
6373 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6374 .maxlen = sizeof(int),
6376 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6379 .procname = "drop_unicast_in_l2_multicast",
6380 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6381 .maxlen = sizeof(int),
6383 .proc_handler = proc_dointvec,
6386 .procname = "drop_unsolicited_na",
6387 .data = &ipv6_devconf.drop_unsolicited_na,
6388 .maxlen = sizeof(int),
6390 .proc_handler = proc_dointvec,
6393 .procname = "keep_addr_on_down",
6394 .data = &ipv6_devconf.keep_addr_on_down,
6395 .maxlen = sizeof(int),
6397 .proc_handler = proc_dointvec,
6401 .procname = "seg6_enabled",
6402 .data = &ipv6_devconf.seg6_enabled,
6403 .maxlen = sizeof(int),
6405 .proc_handler = proc_dointvec,
6407 #ifdef CONFIG_IPV6_SEG6_HMAC
6409 .procname = "seg6_require_hmac",
6410 .data = &ipv6_devconf.seg6_require_hmac,
6411 .maxlen = sizeof(int),
6413 .proc_handler = proc_dointvec,
6417 .procname = "enhanced_dad",
6418 .data = &ipv6_devconf.enhanced_dad,
6419 .maxlen = sizeof(int),
6421 .proc_handler = proc_dointvec,
6424 .procname = "addr_gen_mode",
6425 .data = &ipv6_devconf.addr_gen_mode,
6426 .maxlen = sizeof(int),
6428 .proc_handler = addrconf_sysctl_addr_gen_mode,
6431 .procname = "disable_policy",
6432 .data = &ipv6_devconf.disable_policy,
6433 .maxlen = sizeof(int),
6435 .proc_handler = addrconf_sysctl_disable_policy,
6438 .procname = "ndisc_tclass",
6439 .data = &ipv6_devconf.ndisc_tclass,
6440 .maxlen = sizeof(int),
6442 .proc_handler = proc_dointvec_minmax,
6443 .extra1 = (void *)&zero,
6444 .extra2 = (void *)&two_five_five,
6451 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6452 struct inet6_dev *idev, struct ipv6_devconf *p)
6455 struct ctl_table *table;
6456 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6458 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6462 for (i = 0; table[i].data; i++) {
6463 table[i].data += (char *)p - (char *)&ipv6_devconf;
6464 /* If one of these is already set, then it is not safe to
6465 * overwrite either of them: this makes proc_dointvec_minmax
6468 if (!table[i].extra1 && !table[i].extra2) {
6469 table[i].extra1 = idev; /* embedded; no ref */
6470 table[i].extra2 = net;
6474 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6476 p->sysctl_header = register_net_sysctl(net, path, table);
6477 if (!p->sysctl_header)
6480 if (!strcmp(dev_name, "all"))
6481 ifindex = NETCONFA_IFINDEX_ALL;
6482 else if (!strcmp(dev_name, "default"))
6483 ifindex = NETCONFA_IFINDEX_DEFAULT;
6485 ifindex = idev->dev->ifindex;
6486 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6496 static void __addrconf_sysctl_unregister(struct net *net,
6497 struct ipv6_devconf *p, int ifindex)
6499 struct ctl_table *table;
6501 if (!p->sysctl_header)
6504 table = p->sysctl_header->ctl_table_arg;
6505 unregister_net_sysctl_table(p->sysctl_header);
6506 p->sysctl_header = NULL;
6509 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6512 static int addrconf_sysctl_register(struct inet6_dev *idev)
6516 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6519 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6520 &ndisc_ifinfo_sysctl_change);
6523 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6526 neigh_sysctl_unregister(idev->nd_parms);
6531 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6533 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6534 idev->dev->ifindex);
6535 neigh_sysctl_unregister(idev->nd_parms);
6541 static int __net_init addrconf_init_net(struct net *net)
6544 struct ipv6_devconf *all, *dflt;
6546 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6550 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6552 goto err_alloc_dflt;
6554 /* these will be inherited by all namespaces */
6555 dflt->autoconf = ipv6_defaults.autoconf;
6556 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6558 dflt->stable_secret.initialized = false;
6559 all->stable_secret.initialized = false;
6561 net->ipv6.devconf_all = all;
6562 net->ipv6.devconf_dflt = dflt;
6564 #ifdef CONFIG_SYSCTL
6565 err = __addrconf_sysctl_register(net, "all", NULL, all);
6569 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6575 #ifdef CONFIG_SYSCTL
6577 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6587 static void __net_exit addrconf_exit_net(struct net *net)
6589 #ifdef CONFIG_SYSCTL
6590 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6591 NETCONFA_IFINDEX_DEFAULT);
6592 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6593 NETCONFA_IFINDEX_ALL);
6595 kfree(net->ipv6.devconf_dflt);
6596 kfree(net->ipv6.devconf_all);
6599 static struct pernet_operations addrconf_ops = {
6600 .init = addrconf_init_net,
6601 .exit = addrconf_exit_net,
6604 static struct rtnl_af_ops inet6_ops __read_mostly = {
6606 .fill_link_af = inet6_fill_link_af,
6607 .get_link_af_size = inet6_get_link_af_size,
6608 .validate_link_af = inet6_validate_link_af,
6609 .set_link_af = inet6_set_link_af,
6613 * Init / cleanup code
6616 int __init addrconf_init(void)
6618 struct inet6_dev *idev;
6621 err = ipv6_addr_label_init();
6623 pr_crit("%s: cannot initialize default policy table: %d\n",
6628 err = register_pernet_subsys(&addrconf_ops);
6632 addrconf_wq = create_workqueue("ipv6_addrconf");
6638 /* The addrconf netdev notifier requires that loopback_dev
6639 * has it's ipv6 private information allocated and setup
6640 * before it can bring up and give link-local addresses
6641 * to other devices which are up.
6643 * Unfortunately, loopback_dev is not necessarily the first
6644 * entry in the global dev_base list of net devices. In fact,
6645 * it is likely to be the very last entry on that list.
6646 * So this causes the notifier registry below to try and
6647 * give link-local addresses to all devices besides loopback_dev
6648 * first, then loopback_dev, which cases all the non-loopback_dev
6649 * devices to fail to get a link-local address.
6651 * So, as a temporary fix, allocate the ipv6 structure for
6652 * loopback_dev first by hand.
6653 * Longer term, all of the dependencies ipv6 has upon the loopback
6654 * device and it being up should be removed.
6657 idev = ipv6_add_dev(init_net.loopback_dev);
6660 err = PTR_ERR(idev);
6664 ip6_route_init_special_entries();
6666 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6667 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6669 register_netdevice_notifier(&ipv6_dev_notf);
6673 rtnl_af_register(&inet6_ops);
6675 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6676 NULL, inet6_dump_ifinfo, 0);
6680 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6681 inet6_rtm_newaddr, NULL, 0);
6684 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6685 inet6_rtm_deladdr, NULL, 0);
6688 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6689 inet6_rtm_getaddr, inet6_dump_ifaddr,
6690 RTNL_FLAG_DOIT_UNLOCKED);
6693 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6694 NULL, inet6_dump_ifmcaddr, 0);
6697 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6698 NULL, inet6_dump_ifacaddr, 0);
6701 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6702 inet6_netconf_get_devconf,
6703 inet6_netconf_dump_devconf,
6704 RTNL_FLAG_DOIT_UNLOCKED);
6707 err = ipv6_addr_label_rtnl_register();
6713 rtnl_unregister_all(PF_INET6);
6714 rtnl_af_unregister(&inet6_ops);
6715 unregister_netdevice_notifier(&ipv6_dev_notf);
6717 destroy_workqueue(addrconf_wq);
6719 unregister_pernet_subsys(&addrconf_ops);
6721 ipv6_addr_label_cleanup();
6726 void addrconf_cleanup(void)
6728 struct net_device *dev;
6731 unregister_netdevice_notifier(&ipv6_dev_notf);
6732 unregister_pernet_subsys(&addrconf_ops);
6733 ipv6_addr_label_cleanup();
6735 rtnl_af_unregister(&inet6_ops);
6739 /* clean dev list */
6740 for_each_netdev(&init_net, dev) {
6741 if (__in6_dev_get(dev) == NULL)
6743 addrconf_ifdown(dev, 1);
6745 addrconf_ifdown(init_net.loopback_dev, 2);
6750 spin_lock_bh(&addrconf_hash_lock);
6751 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6752 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6753 spin_unlock_bh(&addrconf_hash_lock);
6754 cancel_delayed_work(&addr_chk_work);
6757 destroy_workqueue(addrconf_wq);