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 fib6_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, bool restart);
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.mtu6 = dev->mtu;
390 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
391 if (!ndev->nd_parms) {
395 if (ndev->cnf.forwarding)
396 dev_disable_lro(dev);
397 /* We refer to the device */
400 if (snmp6_alloc_dev(ndev) < 0) {
401 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
403 neigh_parms_release(&nd_tbl, ndev->nd_parms);
409 if (snmp6_register_dev(ndev) < 0) {
410 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
411 __func__, dev->name);
415 /* One reference from device. */
416 refcount_set(&ndev->refcnt, 1);
418 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
419 ndev->cnf.accept_dad = -1;
421 #if IS_ENABLED(CONFIG_IPV6_SIT)
422 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
423 pr_info("%s: Disabled Multicast RS\n", dev->name);
424 ndev->cnf.rtr_solicits = 0;
428 INIT_LIST_HEAD(&ndev->tempaddr_list);
429 ndev->desync_factor = U32_MAX;
430 if ((dev->flags&IFF_LOOPBACK) ||
431 dev->type == ARPHRD_TUNNEL ||
432 dev->type == ARPHRD_TUNNEL6 ||
433 dev->type == ARPHRD_SIT ||
434 dev->type == ARPHRD_NONE) {
435 ndev->cnf.use_tempaddr = -1;
437 ipv6_regen_rndid(ndev);
439 ndev->token = in6addr_any;
441 if (netif_running(dev) && addrconf_link_ready(dev))
442 ndev->if_flags |= IF_READY;
444 ipv6_mc_init_dev(ndev);
445 ndev->tstamp = jiffies;
446 err = addrconf_sysctl_register(ndev);
448 ipv6_mc_destroy_dev(ndev);
449 snmp6_unregister_dev(ndev);
452 /* protected by rtnl_lock */
453 rcu_assign_pointer(dev->ip6_ptr, ndev);
455 /* Join interface-local all-node multicast group */
456 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
458 /* Join all-node multicast group */
459 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
461 /* Join all-router multicast group if forwarding is set */
462 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
463 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
468 neigh_parms_release(&nd_tbl, ndev->nd_parms);
470 in6_dev_finish_destroy(ndev);
474 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
476 struct inet6_dev *idev;
480 idev = __in6_dev_get(dev);
482 idev = ipv6_add_dev(dev);
487 if (dev->flags&IFF_UP)
492 static int inet6_netconf_msgsize_devconf(int type)
494 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
495 + nla_total_size(4); /* NETCONFA_IFINDEX */
498 if (type == NETCONFA_ALL)
501 if (all || type == NETCONFA_FORWARDING)
502 size += nla_total_size(4);
503 #ifdef CONFIG_IPV6_MROUTE
504 if (all || type == NETCONFA_MC_FORWARDING)
505 size += nla_total_size(4);
507 if (all || type == NETCONFA_PROXY_NEIGH)
508 size += nla_total_size(4);
510 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
511 size += nla_total_size(4);
516 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
517 struct ipv6_devconf *devconf, u32 portid,
518 u32 seq, int event, unsigned int flags,
521 struct nlmsghdr *nlh;
522 struct netconfmsg *ncm;
525 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
530 if (type == NETCONFA_ALL)
533 ncm = nlmsg_data(nlh);
534 ncm->ncm_family = AF_INET6;
536 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
537 goto nla_put_failure;
542 if ((all || type == NETCONFA_FORWARDING) &&
543 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
544 goto nla_put_failure;
545 #ifdef CONFIG_IPV6_MROUTE
546 if ((all || type == NETCONFA_MC_FORWARDING) &&
547 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
548 devconf->mc_forwarding) < 0)
549 goto nla_put_failure;
551 if ((all || type == NETCONFA_PROXY_NEIGH) &&
552 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
553 goto nla_put_failure;
555 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
556 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
557 devconf->ignore_routes_with_linkdown) < 0)
558 goto nla_put_failure;
565 nlmsg_cancel(skb, nlh);
569 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
570 int ifindex, struct ipv6_devconf *devconf)
575 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
579 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
582 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
583 WARN_ON(err == -EMSGSIZE);
587 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
590 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
593 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
594 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
595 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
596 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
597 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
600 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
601 struct nlmsghdr *nlh,
602 struct netlink_ext_ack *extack)
604 struct net *net = sock_net(in_skb->sk);
605 struct nlattr *tb[NETCONFA_MAX+1];
606 struct inet6_dev *in6_dev = NULL;
607 struct net_device *dev = NULL;
608 struct netconfmsg *ncm;
610 struct ipv6_devconf *devconf;
614 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
615 devconf_ipv6_policy, extack);
619 if (!tb[NETCONFA_IFINDEX])
623 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
625 case NETCONFA_IFINDEX_ALL:
626 devconf = net->ipv6.devconf_all;
628 case NETCONFA_IFINDEX_DEFAULT:
629 devconf = net->ipv6.devconf_dflt;
632 dev = dev_get_by_index(net, ifindex);
635 in6_dev = in6_dev_get(dev);
638 devconf = &in6_dev->cnf;
643 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
647 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
648 NETLINK_CB(in_skb).portid,
649 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
652 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
653 WARN_ON(err == -EMSGSIZE);
657 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
660 in6_dev_put(in6_dev);
666 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
667 struct netlink_callback *cb)
669 struct net *net = sock_net(skb->sk);
672 struct net_device *dev;
673 struct inet6_dev *idev;
674 struct hlist_head *head;
677 s_idx = idx = cb->args[1];
679 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
681 head = &net->dev_index_head[h];
683 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
685 hlist_for_each_entry_rcu(dev, head, index_hlist) {
688 idev = __in6_dev_get(dev);
692 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
694 NETLINK_CB(cb->skb).portid,
702 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
708 if (h == NETDEV_HASHENTRIES) {
709 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
710 net->ipv6.devconf_all,
711 NETLINK_CB(cb->skb).portid,
713 RTM_NEWNETCONF, NLM_F_MULTI,
719 if (h == NETDEV_HASHENTRIES + 1) {
720 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
721 net->ipv6.devconf_dflt,
722 NETLINK_CB(cb->skb).portid,
724 RTM_NEWNETCONF, NLM_F_MULTI,
738 static void dev_forward_change(struct inet6_dev *idev)
740 struct net_device *dev;
741 struct inet6_ifaddr *ifa;
746 if (idev->cnf.forwarding)
747 dev_disable_lro(dev);
748 if (dev->flags & IFF_MULTICAST) {
749 if (idev->cnf.forwarding) {
750 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
751 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
752 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
754 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
755 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
756 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
760 list_for_each_entry(ifa, &idev->addr_list, if_list) {
761 if (ifa->flags&IFA_F_TENTATIVE)
763 if (idev->cnf.forwarding)
764 addrconf_join_anycast(ifa);
766 addrconf_leave_anycast(ifa);
768 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
770 dev->ifindex, &idev->cnf);
774 static void addrconf_forward_change(struct net *net, __s32 newf)
776 struct net_device *dev;
777 struct inet6_dev *idev;
779 for_each_netdev(net, dev) {
780 idev = __in6_dev_get(dev);
782 int changed = (!idev->cnf.forwarding) ^ (!newf);
783 idev->cnf.forwarding = newf;
785 dev_forward_change(idev);
790 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
796 return restart_syscall();
798 net = (struct net *)table->extra2;
802 if (p == &net->ipv6.devconf_dflt->forwarding) {
803 if ((!newf) ^ (!old))
804 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
806 NETCONFA_IFINDEX_DEFAULT,
807 net->ipv6.devconf_dflt);
812 if (p == &net->ipv6.devconf_all->forwarding) {
813 int old_dflt = net->ipv6.devconf_dflt->forwarding;
815 net->ipv6.devconf_dflt->forwarding = newf;
816 if ((!newf) ^ (!old_dflt))
817 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
819 NETCONFA_IFINDEX_DEFAULT,
820 net->ipv6.devconf_dflt);
822 addrconf_forward_change(net, newf);
823 if ((!newf) ^ (!old))
824 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
826 NETCONFA_IFINDEX_ALL,
827 net->ipv6.devconf_all);
828 } else if ((!newf) ^ (!old))
829 dev_forward_change((struct inet6_dev *)table->extra1);
833 rt6_purge_dflt_routers(net);
837 static void addrconf_linkdown_change(struct net *net, __s32 newf)
839 struct net_device *dev;
840 struct inet6_dev *idev;
842 for_each_netdev(net, dev) {
843 idev = __in6_dev_get(dev);
845 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
847 idev->cnf.ignore_routes_with_linkdown = newf;
849 inet6_netconf_notify_devconf(dev_net(dev),
851 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
858 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
864 return restart_syscall();
866 net = (struct net *)table->extra2;
870 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
871 if ((!newf) ^ (!old))
872 inet6_netconf_notify_devconf(net,
874 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
875 NETCONFA_IFINDEX_DEFAULT,
876 net->ipv6.devconf_dflt);
881 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
882 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
883 addrconf_linkdown_change(net, newf);
884 if ((!newf) ^ (!old))
885 inet6_netconf_notify_devconf(net,
887 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
888 NETCONFA_IFINDEX_ALL,
889 net->ipv6.devconf_all);
898 /* Nobody refers to this ifaddr, destroy it */
899 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
901 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
903 #ifdef NET_REFCNT_DEBUG
904 pr_debug("%s\n", __func__);
907 in6_dev_put(ifp->idev);
909 if (cancel_delayed_work(&ifp->dad_work))
910 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
913 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
914 pr_warn("Freeing alive inet6 address %p\n", ifp);
922 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
925 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
928 * Each device address list is sorted in order of scope -
929 * global before linklocal.
931 list_for_each(p, &idev->addr_list) {
932 struct inet6_ifaddr *ifa
933 = list_entry(p, struct inet6_ifaddr, if_list);
934 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
938 list_add_tail_rcu(&ifp->if_list, p);
941 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
943 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
945 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
948 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
949 struct net_device *dev, unsigned int hash)
951 struct inet6_ifaddr *ifp;
953 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
954 if (!net_eq(dev_net(ifp->idev->dev), net))
956 if (ipv6_addr_equal(&ifp->addr, addr)) {
957 if (!dev || ifp->idev->dev == dev)
964 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
966 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
969 spin_lock(&addrconf_hash_lock);
971 /* Ignore adding duplicate addresses on an interface */
972 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
973 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
976 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
979 spin_unlock(&addrconf_hash_lock);
984 /* On success it returns ifp with increased reference count */
986 static struct inet6_ifaddr *
987 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
988 bool can_block, struct netlink_ext_ack *extack)
990 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
991 int addr_type = ipv6_addr_type(cfg->pfx);
992 struct net *net = dev_net(idev->dev);
993 struct inet6_ifaddr *ifa = NULL;
994 struct fib6_info *f6i = NULL;
997 if (addr_type == IPV6_ADDR_ANY ||
998 (addr_type & IPV6_ADDR_MULTICAST &&
999 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1000 (!(idev->dev->flags & IFF_LOOPBACK) &&
1001 addr_type & IPV6_ADDR_LOOPBACK))
1002 return ERR_PTR(-EADDRNOTAVAIL);
1005 err = -ENODEV; /*XXX*/
1009 if (idev->cnf.disable_ipv6) {
1014 /* validator notifier needs to be blocking;
1015 * do not call in atomic context
1018 struct in6_validator_info i6vi = {
1019 .i6vi_addr = *cfg->pfx,
1024 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1025 err = notifier_to_errno(err);
1030 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1036 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1043 if (net->ipv6.devconf_all->disable_policy ||
1044 idev->cnf.disable_policy)
1045 f6i->dst_nopolicy = true;
1047 neigh_parms_data_state_setall(idev->nd_parms);
1049 ifa->addr = *cfg->pfx;
1051 ifa->peer_addr = *cfg->peer_pfx;
1053 spin_lock_init(&ifa->lock);
1054 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1055 INIT_HLIST_NODE(&ifa->addr_lst);
1056 ifa->scope = cfg->scope;
1057 ifa->prefix_len = cfg->plen;
1058 ifa->rt_priority = cfg->rt_priority;
1059 ifa->flags = cfg->ifa_flags;
1060 /* No need to add the TENTATIVE flag for addresses with NODAD */
1061 if (!(cfg->ifa_flags & IFA_F_NODAD))
1062 ifa->flags |= IFA_F_TENTATIVE;
1063 ifa->valid_lft = cfg->valid_lft;
1064 ifa->prefered_lft = cfg->preferred_lft;
1065 ifa->cstamp = ifa->tstamp = jiffies;
1066 ifa->tokenized = false;
1074 refcount_set(&ifa->refcnt, 1);
1078 err = ipv6_add_addr_hash(idev->dev, ifa);
1080 rcu_read_unlock_bh();
1084 write_lock(&idev->lock);
1086 /* Add to inet6_dev unicast addr list. */
1087 ipv6_link_dev_addr(idev, ifa);
1089 if (ifa->flags&IFA_F_TEMPORARY) {
1090 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1095 write_unlock(&idev->lock);
1097 rcu_read_unlock_bh();
1099 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1101 if (unlikely(err < 0)) {
1102 fib6_info_release(f6i);
1106 in6_dev_put(ifa->idev);
1115 enum cleanup_prefix_rt_t {
1116 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1117 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1118 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1122 * Check, whether the prefix for ifp would still need a prefix route
1123 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1126 * 1) we don't purge prefix if address was not permanent.
1127 * prefix is managed by its own lifetime.
1128 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1129 * 3) if there are no addresses, delete prefix.
1130 * 4) if there are still other permanent address(es),
1131 * corresponding prefix is still permanent.
1132 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1133 * don't purge the prefix, assume user space is managing it.
1134 * 6) otherwise, update prefix lifetime to the
1135 * longest valid lifetime among the corresponding
1136 * addresses on the device.
1137 * Note: subsequent RA will update lifetime.
1139 static enum cleanup_prefix_rt_t
1140 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1142 struct inet6_ifaddr *ifa;
1143 struct inet6_dev *idev = ifp->idev;
1144 unsigned long lifetime;
1145 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1149 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1152 if (ifa->prefix_len != ifp->prefix_len ||
1153 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1156 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1157 return CLEANUP_PREFIX_RT_NOP;
1159 action = CLEANUP_PREFIX_RT_EXPIRE;
1161 spin_lock(&ifa->lock);
1163 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1165 * Note: Because this address is
1166 * not permanent, lifetime <
1167 * LONG_MAX / HZ here.
1169 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1170 *expires = ifa->tstamp + lifetime * HZ;
1171 spin_unlock(&ifa->lock);
1178 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1179 bool del_rt, bool del_peer)
1181 struct fib6_info *f6i;
1183 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1186 0, RTF_GATEWAY | RTF_DEFAULT);
1189 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1191 if (!(f6i->fib6_flags & RTF_EXPIRES))
1192 fib6_set_expires(f6i, expires);
1193 fib6_info_release(f6i);
1199 /* This function wants to get referenced ifp and releases it before return */
1201 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1204 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1205 unsigned long expires;
1209 spin_lock_bh(&ifp->lock);
1211 ifp->state = INET6_IFADDR_STATE_DEAD;
1212 spin_unlock_bh(&ifp->lock);
1214 if (state == INET6_IFADDR_STATE_DEAD)
1217 spin_lock_bh(&addrconf_hash_lock);
1218 hlist_del_init_rcu(&ifp->addr_lst);
1219 spin_unlock_bh(&addrconf_hash_lock);
1221 write_lock_bh(&ifp->idev->lock);
1223 if (ifp->flags&IFA_F_TEMPORARY) {
1224 list_del(&ifp->tmp_list);
1226 in6_ifa_put(ifp->ifpub);
1232 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1233 action = check_cleanup_prefix_route(ifp, &expires);
1235 list_del_rcu(&ifp->if_list);
1238 write_unlock_bh(&ifp->idev->lock);
1240 addrconf_del_dad_work(ifp);
1242 ipv6_ifa_notify(RTM_DELADDR, ifp);
1244 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1246 if (action != CLEANUP_PREFIX_RT_NOP) {
1247 cleanup_prefix_route(ifp, expires,
1248 action == CLEANUP_PREFIX_RT_DEL, false);
1251 /* clean up prefsrc entries */
1252 rt6_remove_prefsrc(ifp);
1257 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1258 struct inet6_ifaddr *ift,
1261 struct inet6_dev *idev = ifp->idev;
1262 struct in6_addr addr, *tmpaddr;
1263 unsigned long tmp_tstamp, age;
1264 unsigned long regen_advance;
1265 struct ifa6_config cfg;
1267 unsigned long now = jiffies;
1268 long max_desync_factor;
1269 s32 cnf_temp_preferred_lft;
1271 write_lock_bh(&idev->lock);
1273 spin_lock_bh(&ift->lock);
1274 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1275 spin_unlock_bh(&ift->lock);
1282 if (idev->cnf.use_tempaddr <= 0) {
1283 write_unlock_bh(&idev->lock);
1284 pr_info("%s: use_tempaddr is disabled\n", __func__);
1289 spin_lock_bh(&ifp->lock);
1290 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1291 idev->cnf.use_tempaddr = -1; /*XXX*/
1292 spin_unlock_bh(&ifp->lock);
1293 write_unlock_bh(&idev->lock);
1294 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1301 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1302 ipv6_try_regen_rndid(idev, tmpaddr);
1303 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1304 age = (now - ifp->tstamp) / HZ;
1306 regen_advance = idev->cnf.regen_max_retry *
1307 idev->cnf.dad_transmits *
1308 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1310 /* recalculate max_desync_factor each time and update
1311 * idev->desync_factor if it's larger
1313 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1314 max_desync_factor = min_t(__u32,
1315 idev->cnf.max_desync_factor,
1316 cnf_temp_preferred_lft - regen_advance);
1318 if (unlikely(idev->desync_factor > max_desync_factor)) {
1319 if (max_desync_factor > 0) {
1320 get_random_bytes(&idev->desync_factor,
1321 sizeof(idev->desync_factor));
1322 idev->desync_factor %= max_desync_factor;
1324 idev->desync_factor = 0;
1328 memset(&cfg, 0, sizeof(cfg));
1329 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1330 idev->cnf.temp_valid_lft + age);
1331 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1332 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1334 cfg.plen = ifp->prefix_len;
1335 tmp_tstamp = ifp->tstamp;
1336 spin_unlock_bh(&ifp->lock);
1338 write_unlock_bh(&idev->lock);
1340 /* A temporary address is created only if this calculated Preferred
1341 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1342 * an implementation must not create a temporary address with a zero
1343 * Preferred Lifetime.
1344 * Use age calculation as in addrconf_verify to avoid unnecessary
1345 * temporary addresses being generated.
1347 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1348 if (cfg.preferred_lft <= regen_advance + age) {
1355 cfg.ifa_flags = IFA_F_TEMPORARY;
1356 /* set in addrconf_prefix_rcv() */
1357 if (ifp->flags & IFA_F_OPTIMISTIC)
1358 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1361 cfg.scope = ipv6_addr_scope(cfg.pfx);
1363 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1367 pr_info("%s: retry temporary address regeneration\n", __func__);
1369 write_lock_bh(&idev->lock);
1373 spin_lock_bh(&ift->lock);
1376 ift->tstamp = tmp_tstamp;
1377 spin_unlock_bh(&ift->lock);
1379 addrconf_dad_start(ift);
1387 * Choose an appropriate source address (RFC3484)
1390 IPV6_SADDR_RULE_INIT = 0,
1391 IPV6_SADDR_RULE_LOCAL,
1392 IPV6_SADDR_RULE_SCOPE,
1393 IPV6_SADDR_RULE_PREFERRED,
1394 #ifdef CONFIG_IPV6_MIP6
1395 IPV6_SADDR_RULE_HOA,
1397 IPV6_SADDR_RULE_OIF,
1398 IPV6_SADDR_RULE_LABEL,
1399 IPV6_SADDR_RULE_PRIVACY,
1400 IPV6_SADDR_RULE_ORCHID,
1401 IPV6_SADDR_RULE_PREFIX,
1402 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1403 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1408 struct ipv6_saddr_score {
1411 struct inet6_ifaddr *ifa;
1412 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1417 struct ipv6_saddr_dst {
1418 const struct in6_addr *addr;
1425 static inline int ipv6_saddr_preferred(int type)
1427 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1432 static bool ipv6_use_optimistic_addr(struct net *net,
1433 struct inet6_dev *idev)
1435 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1438 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1440 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1449 static bool ipv6_allow_optimistic_dad(struct net *net,
1450 struct inet6_dev *idev)
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1455 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1464 static int ipv6_get_saddr_eval(struct net *net,
1465 struct ipv6_saddr_score *score,
1466 struct ipv6_saddr_dst *dst,
1471 if (i <= score->rule) {
1473 case IPV6_SADDR_RULE_SCOPE:
1474 ret = score->scopedist;
1476 case IPV6_SADDR_RULE_PREFIX:
1477 ret = score->matchlen;
1480 ret = !!test_bit(i, score->scorebits);
1486 case IPV6_SADDR_RULE_INIT:
1487 /* Rule 0: remember if hiscore is not ready yet */
1490 case IPV6_SADDR_RULE_LOCAL:
1491 /* Rule 1: Prefer same address */
1492 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1494 case IPV6_SADDR_RULE_SCOPE:
1495 /* Rule 2: Prefer appropriate scope
1500 * ---+--+-+---> scope
1502 * | d is scope of the destination.
1504 * | \ <- smaller scope is better if
1505 * B-15 | \ if scope is enough for destination.
1506 * | ret = B - scope (-1 <= scope >= d <= 15).
1508 * |/ <- greater is better
1509 * -C / if scope is not enough for destination.
1510 * /| ret = scope - C (-1 <= d < scope <= 15).
1512 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1513 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1514 * Assume B = 0 and we get C > 29.
1516 ret = __ipv6_addr_src_scope(score->addr_type);
1517 if (ret >= dst->scope)
1520 ret -= 128; /* 30 is enough */
1521 score->scopedist = ret;
1523 case IPV6_SADDR_RULE_PREFERRED:
1525 /* Rule 3: Avoid deprecated and optimistic addresses */
1526 u8 avoid = IFA_F_DEPRECATED;
1528 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1529 avoid |= IFA_F_OPTIMISTIC;
1530 ret = ipv6_saddr_preferred(score->addr_type) ||
1531 !(score->ifa->flags & avoid);
1534 #ifdef CONFIG_IPV6_MIP6
1535 case IPV6_SADDR_RULE_HOA:
1537 /* Rule 4: Prefer home address */
1538 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1539 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1543 case IPV6_SADDR_RULE_OIF:
1544 /* Rule 5: Prefer outgoing interface */
1545 ret = (!dst->ifindex ||
1546 dst->ifindex == score->ifa->idev->dev->ifindex);
1548 case IPV6_SADDR_RULE_LABEL:
1549 /* Rule 6: Prefer matching label */
1550 ret = ipv6_addr_label(net,
1551 &score->ifa->addr, score->addr_type,
1552 score->ifa->idev->dev->ifindex) == dst->label;
1554 case IPV6_SADDR_RULE_PRIVACY:
1556 /* Rule 7: Prefer public address
1557 * Note: prefer temporary address if use_tempaddr >= 2
1559 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1560 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1561 score->ifa->idev->cnf.use_tempaddr >= 2;
1562 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1565 case IPV6_SADDR_RULE_ORCHID:
1566 /* Rule 8-: Prefer ORCHID vs ORCHID or
1567 * non-ORCHID vs non-ORCHID
1569 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1570 ipv6_addr_orchid(dst->addr));
1572 case IPV6_SADDR_RULE_PREFIX:
1573 /* Rule 8: Use longest matching prefix */
1574 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1575 if (ret > score->ifa->prefix_len)
1576 ret = score->ifa->prefix_len;
1577 score->matchlen = ret;
1579 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1580 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1581 /* Optimistic addresses still have lower precedence than other
1582 * preferred addresses.
1584 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1592 __set_bit(i, score->scorebits);
1598 static int __ipv6_dev_get_saddr(struct net *net,
1599 struct ipv6_saddr_dst *dst,
1600 struct inet6_dev *idev,
1601 struct ipv6_saddr_score *scores,
1604 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1606 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1610 * - Tentative Address (RFC2462 section 5.4)
1611 * - A tentative address is not considered
1612 * "assigned to an interface" in the traditional
1613 * sense, unless it is also flagged as optimistic.
1614 * - Candidate Source Address (section 4)
1615 * - In any case, anycast addresses, multicast
1616 * addresses, and the unspecified address MUST
1617 * NOT be included in a candidate set.
1619 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1620 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1623 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1625 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1626 score->addr_type & IPV6_ADDR_MULTICAST)) {
1627 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1633 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1635 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1636 int minihiscore, miniscore;
1638 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1639 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1641 if (minihiscore > miniscore) {
1642 if (i == IPV6_SADDR_RULE_SCOPE &&
1643 score->scopedist > 0) {
1646 * each remaining entry
1647 * has too small (not enough)
1648 * scope, because ifa entries
1649 * are sorted by their scope
1655 } else if (minihiscore < miniscore) {
1656 swap(hiscore, score);
1657 hiscore_idx = 1 - hiscore_idx;
1659 /* restore our iterator */
1660 score->ifa = hiscore->ifa;
1670 static int ipv6_get_saddr_master(struct net *net,
1671 const struct net_device *dst_dev,
1672 const struct net_device *master,
1673 struct ipv6_saddr_dst *dst,
1674 struct ipv6_saddr_score *scores,
1677 struct inet6_dev *idev;
1679 idev = __in6_dev_get(dst_dev);
1681 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1682 scores, hiscore_idx);
1684 idev = __in6_dev_get(master);
1686 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1687 scores, hiscore_idx);
1692 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1693 const struct in6_addr *daddr, unsigned int prefs,
1694 struct in6_addr *saddr)
1696 struct ipv6_saddr_score scores[2], *hiscore;
1697 struct ipv6_saddr_dst dst;
1698 struct inet6_dev *idev;
1699 struct net_device *dev;
1701 bool use_oif_addr = false;
1702 int hiscore_idx = 0;
1705 dst_type = __ipv6_addr_type(daddr);
1707 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1708 dst.scope = __ipv6_addr_src_scope(dst_type);
1709 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1712 scores[hiscore_idx].rule = -1;
1713 scores[hiscore_idx].ifa = NULL;
1717 /* Candidate Source Address (section 4)
1718 * - multicast and link-local destination address,
1719 * the set of candidate source address MUST only
1720 * include addresses assigned to interfaces
1721 * belonging to the same link as the outgoing
1723 * (- For site-local destination addresses, the
1724 * set of candidate source addresses MUST only
1725 * include addresses assigned to interfaces
1726 * belonging to the same site as the outgoing
1728 * - "It is RECOMMENDED that the candidate source addresses
1729 * be the set of unicast addresses assigned to the
1730 * interface that will be used to send to the destination
1731 * (the 'outgoing' interface)." (RFC 6724)
1734 idev = __in6_dev_get(dst_dev);
1735 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1736 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1737 (idev && idev->cnf.use_oif_addrs_only)) {
1738 use_oif_addr = true;
1744 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1746 const struct net_device *master;
1749 /* if dst_dev exists and is enslaved to an L3 device, then
1750 * prefer addresses from dst_dev and then the master over
1751 * any other enslaved devices in the L3 domain.
1753 master = l3mdev_master_dev_rcu(dst_dev);
1755 master_idx = master->ifindex;
1757 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1759 scores, hiscore_idx);
1761 if (scores[hiscore_idx].ifa)
1765 for_each_netdev_rcu(net, dev) {
1766 /* only consider addresses on devices in the
1769 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1771 idev = __in6_dev_get(dev);
1774 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1779 hiscore = &scores[hiscore_idx];
1781 ret = -EADDRNOTAVAIL;
1783 *saddr = hiscore->ifa->addr;
1788 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1790 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1793 struct inet6_ifaddr *ifp;
1794 int err = -EADDRNOTAVAIL;
1796 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1797 if (ifp->scope > IFA_LINK)
1799 if (ifp->scope == IFA_LINK &&
1800 !(ifp->flags & banned_flags)) {
1809 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1812 struct inet6_dev *idev;
1813 int err = -EADDRNOTAVAIL;
1816 idev = __in6_dev_get(dev);
1818 read_lock_bh(&idev->lock);
1819 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1820 read_unlock_bh(&idev->lock);
1826 static int ipv6_count_addresses(const struct inet6_dev *idev)
1828 const struct inet6_ifaddr *ifp;
1832 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1838 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1839 const struct net_device *dev, int strict)
1841 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1842 strict, IFA_F_TENTATIVE);
1844 EXPORT_SYMBOL(ipv6_chk_addr);
1846 /* device argument is used to find the L3 domain of interest. If
1847 * skip_dev_check is set, then the ifp device is not checked against
1848 * the passed in dev argument. So the 2 cases for addresses checks are:
1849 * 1. does the address exist in the L3 domain that dev is part of
1850 * (skip_dev_check = true), or
1852 * 2. does the address exist on the specific device
1853 * (skip_dev_check = false)
1855 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1856 const struct net_device *dev, bool skip_dev_check,
1857 int strict, u32 banned_flags)
1859 unsigned int hash = inet6_addr_hash(net, addr);
1860 const struct net_device *l3mdev;
1861 struct inet6_ifaddr *ifp;
1866 l3mdev = l3mdev_master_dev_rcu(dev);
1870 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1871 if (!net_eq(dev_net(ifp->idev->dev), net))
1874 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1877 /* Decouple optimistic from tentative for evaluation here.
1878 * Ban optimistic addresses explicitly, when required.
1880 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1881 ? (ifp->flags&~IFA_F_TENTATIVE)
1883 if (ipv6_addr_equal(&ifp->addr, addr) &&
1884 !(ifp_flags&banned_flags) &&
1885 (!dev || ifp->idev->dev == dev ||
1886 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1895 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1898 /* Compares an address/prefix_len with addresses on device @dev.
1899 * If one is found it returns true.
1901 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1902 const unsigned int prefix_len, struct net_device *dev)
1904 const struct inet6_ifaddr *ifa;
1905 const struct inet6_dev *idev;
1909 idev = __in6_dev_get(dev);
1911 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1912 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1921 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1923 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1925 const struct inet6_ifaddr *ifa;
1926 const struct inet6_dev *idev;
1931 idev = __in6_dev_get(dev);
1933 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1934 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1943 EXPORT_SYMBOL(ipv6_chk_prefix);
1945 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1946 struct net_device *dev, int strict)
1948 unsigned int hash = inet6_addr_hash(net, addr);
1949 struct inet6_ifaddr *ifp, *result = NULL;
1952 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1953 if (!net_eq(dev_net(ifp->idev->dev), net))
1955 if (ipv6_addr_equal(&ifp->addr, addr)) {
1956 if (!dev || ifp->idev->dev == dev ||
1957 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1969 /* Gets referenced address, destroys ifaddr */
1971 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1974 ifp->flags |= IFA_F_DADFAILED;
1976 if (ifp->flags&IFA_F_TEMPORARY) {
1977 struct inet6_ifaddr *ifpub;
1978 spin_lock_bh(&ifp->lock);
1981 in6_ifa_hold(ifpub);
1982 spin_unlock_bh(&ifp->lock);
1983 ipv6_create_tempaddr(ifpub, ifp, true);
1986 spin_unlock_bh(&ifp->lock);
1989 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1990 spin_lock_bh(&ifp->lock);
1991 addrconf_del_dad_work(ifp);
1992 ifp->flags |= IFA_F_TENTATIVE;
1994 ifp->flags &= ~IFA_F_OPTIMISTIC;
1995 spin_unlock_bh(&ifp->lock);
1997 ipv6_ifa_notify(0, ifp);
2004 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2008 spin_lock_bh(&ifp->lock);
2009 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2010 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2013 spin_unlock_bh(&ifp->lock);
2018 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2020 struct inet6_dev *idev = ifp->idev;
2021 struct net *net = dev_net(ifp->idev->dev);
2023 if (addrconf_dad_end(ifp)) {
2028 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2029 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2031 spin_lock_bh(&ifp->lock);
2033 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2034 struct in6_addr new_addr;
2035 struct inet6_ifaddr *ifp2;
2036 int retries = ifp->stable_privacy_retry + 1;
2037 struct ifa6_config cfg = {
2039 .plen = ifp->prefix_len,
2040 .ifa_flags = ifp->flags,
2041 .valid_lft = ifp->valid_lft,
2042 .preferred_lft = ifp->prefered_lft,
2043 .scope = ifp->scope,
2046 if (retries > net->ipv6.sysctl.idgen_retries) {
2047 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2048 ifp->idev->dev->name);
2052 new_addr = ifp->addr;
2053 if (ipv6_generate_stable_address(&new_addr, retries,
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, &cfg, false, NULL);
2071 spin_lock_bh(&ifp2->lock);
2072 ifp2->stable_privacy_retry = retries;
2073 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2074 spin_unlock_bh(&ifp2->lock);
2076 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2079 spin_lock_bh(&ifp->lock);
2083 /* transition from _POSTDAD to _ERRDAD */
2084 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2085 spin_unlock_bh(&ifp->lock);
2087 addrconf_mod_dad_work(ifp, 0);
2091 /* Join to solicited addr multicast group.
2092 * caller must hold RTNL */
2093 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2095 struct in6_addr maddr;
2097 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2100 addrconf_addr_solict_mult(addr, &maddr);
2101 ipv6_dev_mc_inc(dev, &maddr);
2104 /* caller must hold RTNL */
2105 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2107 struct in6_addr maddr;
2109 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2112 addrconf_addr_solict_mult(addr, &maddr);
2113 __ipv6_dev_mc_dec(idev, &maddr);
2116 /* caller must hold RTNL */
2117 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2119 struct in6_addr addr;
2121 if (ifp->prefix_len >= 127) /* RFC 6164 */
2123 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2124 if (ipv6_addr_any(&addr))
2126 __ipv6_dev_ac_inc(ifp->idev, &addr);
2129 /* caller must hold RTNL */
2130 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2132 struct in6_addr addr;
2134 if (ifp->prefix_len >= 127) /* RFC 6164 */
2136 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2137 if (ipv6_addr_any(&addr))
2139 __ipv6_dev_ac_dec(ifp->idev, &addr);
2142 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2144 switch (dev->addr_len) {
2146 memcpy(eui, dev->dev_addr, 3);
2149 memcpy(eui + 5, dev->dev_addr + 3, 3);
2151 case EUI64_ADDR_LEN:
2152 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2162 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2164 union fwnet_hwaddr *ha;
2166 if (dev->addr_len != FWNET_ALEN)
2169 ha = (union fwnet_hwaddr *)dev->dev_addr;
2171 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2176 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2178 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2179 if (dev->addr_len != ARCNET_ALEN)
2182 eui[7] = *(u8 *)dev->dev_addr;
2186 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2188 if (dev->addr_len != INFINIBAND_ALEN)
2190 memcpy(eui, dev->dev_addr + 12, 8);
2195 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2199 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2200 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2201 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2202 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2203 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2204 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2208 memcpy(eui + 4, &addr, 4);
2212 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2214 if (dev->priv_flags & IFF_ISATAP)
2215 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2219 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2221 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2224 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2226 memcpy(eui, dev->perm_addr, 3);
2227 memcpy(eui + 5, dev->perm_addr + 3, 3);
2234 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2236 switch (dev->type) {
2239 return addrconf_ifid_eui48(eui, dev);
2241 return addrconf_ifid_arcnet(eui, dev);
2242 case ARPHRD_INFINIBAND:
2243 return addrconf_ifid_infiniband(eui, dev);
2245 return addrconf_ifid_sit(eui, dev);
2248 return addrconf_ifid_gre(eui, dev);
2249 case ARPHRD_6LOWPAN:
2250 return addrconf_ifid_6lowpan(eui, dev);
2251 case ARPHRD_IEEE1394:
2252 return addrconf_ifid_ieee1394(eui, dev);
2253 case ARPHRD_TUNNEL6:
2256 return addrconf_ifid_ip6tnl(eui, dev);
2261 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2264 struct inet6_ifaddr *ifp;
2266 read_lock_bh(&idev->lock);
2267 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2268 if (ifp->scope > IFA_LINK)
2270 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2271 memcpy(eui, ifp->addr.s6_addr+8, 8);
2276 read_unlock_bh(&idev->lock);
2280 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2281 static void ipv6_regen_rndid(struct inet6_dev *idev)
2284 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2285 idev->rndid[0] &= ~0x02;
2288 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2289 * check if generated address is not inappropriate
2291 * - Reserved subnet anycast (RFC 2526)
2292 * 11111101 11....11 1xxxxxxx
2293 * - ISATAP (RFC4214) 6.1
2294 * 00-00-5E-FE-xx-xx-xx-xx
2296 * - XXX: already assigned to an address on the device
2298 if (idev->rndid[0] == 0xfd &&
2299 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2300 (idev->rndid[7]&0x80))
2302 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2303 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2305 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2310 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2312 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2313 ipv6_regen_rndid(idev);
2321 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2322 struct net_device *dev, unsigned long expires,
2323 u32 flags, gfp_t gfp_flags)
2325 struct fib6_config cfg = {
2326 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2327 .fc_metric = 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 fib6_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 fib6_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->fib6_flags & flags) != flags)
2376 if ((rt->fib6_flags & noflags) != 0)
2378 if (!fib6_info_hold_safe(rt))
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_MULTICAST,
2399 .fc_nlinfo.nl_net = dev_net(dev),
2400 .fc_protocol = RTPROT_KERNEL,
2403 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2405 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2408 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2410 struct inet6_dev *idev;
2414 idev = ipv6_find_idev(dev);
2416 return ERR_PTR(-ENOBUFS);
2418 if (idev->cnf.disable_ipv6)
2419 return ERR_PTR(-EACCES);
2421 /* Add default multicast route */
2422 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2423 addrconf_add_mroute(dev);
2428 static void manage_tempaddrs(struct inet6_dev *idev,
2429 struct inet6_ifaddr *ifp,
2430 __u32 valid_lft, __u32 prefered_lft,
2431 bool create, unsigned long now)
2434 struct inet6_ifaddr *ift;
2436 read_lock_bh(&idev->lock);
2437 /* update all temporary addresses in the list */
2438 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2439 int age, max_valid, max_prefered;
2441 if (ifp != ift->ifpub)
2444 /* RFC 4941 section 3.3:
2445 * If a received option will extend the lifetime of a public
2446 * address, the lifetimes of temporary addresses should
2447 * be extended, subject to the overall constraint that no
2448 * temporary addresses should ever remain "valid" or "preferred"
2449 * for a time longer than (TEMP_VALID_LIFETIME) or
2450 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2452 age = (now - ift->cstamp) / HZ;
2453 max_valid = idev->cnf.temp_valid_lft - age;
2457 max_prefered = idev->cnf.temp_prefered_lft -
2458 idev->desync_factor - age;
2459 if (max_prefered < 0)
2462 if (valid_lft > max_valid)
2463 valid_lft = max_valid;
2465 if (prefered_lft > max_prefered)
2466 prefered_lft = max_prefered;
2468 spin_lock(&ift->lock);
2470 ift->valid_lft = valid_lft;
2471 ift->prefered_lft = prefered_lft;
2473 if (prefered_lft > 0)
2474 ift->flags &= ~IFA_F_DEPRECATED;
2476 spin_unlock(&ift->lock);
2477 if (!(flags&IFA_F_TENTATIVE))
2478 ipv6_ifa_notify(0, ift);
2481 if ((create || list_empty(&idev->tempaddr_list)) &&
2482 idev->cnf.use_tempaddr > 0) {
2483 /* When a new public address is created as described
2484 * in [ADDRCONF], also create a new temporary address.
2485 * Also create a temporary address if it's enabled but
2486 * no temporary address currently exists.
2488 read_unlock_bh(&idev->lock);
2489 ipv6_create_tempaddr(ifp, NULL, false);
2491 read_unlock_bh(&idev->lock);
2495 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2497 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2498 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2501 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2502 const struct prefix_info *pinfo,
2503 struct inet6_dev *in6_dev,
2504 const struct in6_addr *addr, int addr_type,
2505 u32 addr_flags, bool sllao, bool tokenized,
2506 __u32 valid_lft, u32 prefered_lft)
2508 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2509 int create = 0, update_lft = 0;
2511 if (!ifp && valid_lft) {
2512 int max_addresses = in6_dev->cnf.max_addresses;
2513 struct ifa6_config cfg = {
2515 .plen = pinfo->prefix_len,
2516 .ifa_flags = addr_flags,
2517 .valid_lft = valid_lft,
2518 .preferred_lft = prefered_lft,
2519 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2522 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2523 if ((net->ipv6.devconf_all->optimistic_dad ||
2524 in6_dev->cnf.optimistic_dad) &&
2525 !net->ipv6.devconf_all->forwarding && sllao)
2526 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2529 /* Do not allow to create too much of autoconfigured
2530 * addresses; this would be too easy way to crash kernel.
2532 if (!max_addresses ||
2533 ipv6_count_addresses(in6_dev) < max_addresses)
2534 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2536 if (IS_ERR_OR_NULL(ifp))
2540 spin_lock_bh(&ifp->lock);
2541 ifp->flags |= IFA_F_MANAGETEMPADDR;
2542 ifp->cstamp = jiffies;
2543 ifp->tokenized = tokenized;
2544 spin_unlock_bh(&ifp->lock);
2545 addrconf_dad_start(ifp);
2553 /* update lifetime (RFC2462 5.5.3 e) */
2554 spin_lock_bh(&ifp->lock);
2556 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2557 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2560 if (!create && stored_lft) {
2561 const u32 minimum_lft = min_t(u32,
2562 stored_lft, MIN_VALID_LIFETIME);
2563 valid_lft = max(valid_lft, minimum_lft);
2565 /* RFC4862 Section 5.5.3e:
2566 * "Note that the preferred lifetime of the
2567 * corresponding address is always reset to
2568 * the Preferred Lifetime in the received
2569 * Prefix Information option, regardless of
2570 * whether the valid lifetime is also reset or
2573 * So we should always update prefered_lft here.
2579 ifp->valid_lft = valid_lft;
2580 ifp->prefered_lft = prefered_lft;
2583 ifp->flags &= ~IFA_F_DEPRECATED;
2584 spin_unlock_bh(&ifp->lock);
2586 if (!(flags&IFA_F_TENTATIVE))
2587 ipv6_ifa_notify(0, ifp);
2589 spin_unlock_bh(&ifp->lock);
2591 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2600 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2602 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2604 struct prefix_info *pinfo;
2609 struct inet6_dev *in6_dev;
2610 struct net *net = dev_net(dev);
2612 pinfo = (struct prefix_info *) opt;
2614 if (len < sizeof(struct prefix_info)) {
2615 netdev_dbg(dev, "addrconf: prefix option too short\n");
2620 * Validation checks ([ADDRCONF], page 19)
2623 addr_type = ipv6_addr_type(&pinfo->prefix);
2625 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2628 valid_lft = ntohl(pinfo->valid);
2629 prefered_lft = ntohl(pinfo->prefered);
2631 if (prefered_lft > valid_lft) {
2632 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2636 in6_dev = in6_dev_get(dev);
2639 net_dbg_ratelimited("addrconf: device %s not configured\n",
2645 * Two things going on here:
2646 * 1) Add routes for on-link prefixes
2647 * 2) Configure prefixes with the auto flag set
2650 if (pinfo->onlink) {
2651 struct fib6_info *rt;
2652 unsigned long rt_expires;
2654 /* Avoid arithmetic overflow. Really, we could
2655 * save rt_expires in seconds, likely valid_lft,
2656 * but it would require division in fib gc, that it
2660 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2662 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2664 if (addrconf_finite_timeout(rt_expires))
2667 rt = addrconf_get_prefix_route(&pinfo->prefix,
2670 RTF_ADDRCONF | RTF_PREFIX_RT,
2671 RTF_GATEWAY | RTF_DEFAULT);
2674 /* Autoconf prefix route */
2675 if (valid_lft == 0) {
2676 ip6_del_rt(net, rt);
2678 } else if (addrconf_finite_timeout(rt_expires)) {
2680 fib6_set_expires(rt, jiffies + rt_expires);
2682 fib6_clean_expires(rt);
2684 } else if (valid_lft) {
2685 clock_t expires = 0;
2686 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2687 if (addrconf_finite_timeout(rt_expires)) {
2689 flags |= RTF_EXPIRES;
2690 expires = jiffies_to_clock_t(rt_expires);
2692 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2693 0, dev, expires, flags,
2696 fib6_info_release(rt);
2699 /* Try to figure out our local address for this prefix */
2701 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2702 struct in6_addr addr;
2703 bool tokenized = false, dev_addr_generated = false;
2705 if (pinfo->prefix_len == 64) {
2706 memcpy(&addr, &pinfo->prefix, 8);
2708 if (!ipv6_addr_any(&in6_dev->token)) {
2709 read_lock_bh(&in6_dev->lock);
2710 memcpy(addr.s6_addr + 8,
2711 in6_dev->token.s6_addr + 8, 8);
2712 read_unlock_bh(&in6_dev->lock);
2714 } else if (is_addr_mode_generate_stable(in6_dev) &&
2715 !ipv6_generate_stable_address(&addr, 0,
2717 addr_flags |= IFA_F_STABLE_PRIVACY;
2719 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2720 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2723 dev_addr_generated = true;
2727 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2732 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2735 tokenized, valid_lft,
2740 /* Ignore error case here because previous prefix add addr was
2741 * successful which will be notified.
2743 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2744 addr_type, addr_flags, sllao,
2745 tokenized, valid_lft,
2747 dev_addr_generated);
2749 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2751 in6_dev_put(in6_dev);
2755 * Set destination address.
2756 * Special case for SIT interfaces where we create a new "virtual"
2759 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2761 struct in6_ifreq ireq;
2762 struct net_device *dev;
2768 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2771 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2777 #if IS_ENABLED(CONFIG_IPV6_SIT)
2778 if (dev->type == ARPHRD_SIT) {
2779 const struct net_device_ops *ops = dev->netdev_ops;
2781 struct ip_tunnel_parm p;
2783 err = -EADDRNOTAVAIL;
2784 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2787 memset(&p, 0, sizeof(p));
2788 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2792 p.iph.protocol = IPPROTO_IPV6;
2794 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2796 if (ops->ndo_do_ioctl) {
2797 mm_segment_t oldfs = get_fs();
2800 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2807 dev = __dev_get_by_name(net, p.name);
2810 err = dev_open(dev);
2820 static int ipv6_mc_config(struct sock *sk, bool join,
2821 const struct in6_addr *addr, int ifindex)
2829 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2831 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2838 * Manual configuration of address on an interface
2840 static int inet6_addr_add(struct net *net, int ifindex,
2841 struct ifa6_config *cfg,
2842 struct netlink_ext_ack *extack)
2844 struct inet6_ifaddr *ifp;
2845 struct inet6_dev *idev;
2846 struct net_device *dev;
2847 unsigned long timeout;
2853 if (cfg->plen > 128)
2856 /* check the lifetime */
2857 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2860 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2863 dev = __dev_get_by_index(net, ifindex);
2867 idev = addrconf_add_dev(dev);
2869 return PTR_ERR(idev);
2871 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2872 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2873 true, cfg->pfx, ifindex);
2879 cfg->scope = ipv6_addr_scope(cfg->pfx);
2881 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2882 if (addrconf_finite_timeout(timeout)) {
2883 expires = jiffies_to_clock_t(timeout * HZ);
2884 cfg->valid_lft = timeout;
2885 flags = RTF_EXPIRES;
2889 cfg->ifa_flags |= IFA_F_PERMANENT;
2892 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2893 if (addrconf_finite_timeout(timeout)) {
2895 cfg->ifa_flags |= IFA_F_DEPRECATED;
2896 cfg->preferred_lft = timeout;
2899 ifp = ipv6_add_addr(idev, cfg, true, extack);
2901 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2902 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2903 ifp->rt_priority, dev, expires,
2907 /* Send a netlink notification if DAD is enabled and
2908 * optimistic flag is not set
2910 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2911 ipv6_ifa_notify(0, ifp);
2913 * Note that section 3.1 of RFC 4429 indicates
2914 * that the Optimistic flag should not be set for
2915 * manually configured addresses
2917 addrconf_dad_start(ifp);
2918 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2919 manage_tempaddrs(idev, ifp, cfg->valid_lft,
2920 cfg->preferred_lft, true, jiffies);
2922 addrconf_verify_rtnl();
2924 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2925 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2929 return PTR_ERR(ifp);
2932 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2933 const struct in6_addr *pfx, unsigned int plen)
2935 struct inet6_ifaddr *ifp;
2936 struct inet6_dev *idev;
2937 struct net_device *dev;
2942 dev = __dev_get_by_index(net, ifindex);
2946 idev = __in6_dev_get(dev);
2950 read_lock_bh(&idev->lock);
2951 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2952 if (ifp->prefix_len == plen &&
2953 ipv6_addr_equal(pfx, &ifp->addr)) {
2955 read_unlock_bh(&idev->lock);
2957 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2958 (ifa_flags & IFA_F_MANAGETEMPADDR))
2959 manage_tempaddrs(idev, ifp, 0, 0, false,
2962 addrconf_verify_rtnl();
2963 if (ipv6_addr_is_multicast(pfx)) {
2964 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2965 false, pfx, dev->ifindex);
2970 read_unlock_bh(&idev->lock);
2971 return -EADDRNOTAVAIL;
2975 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2977 struct ifa6_config cfg = {
2978 .ifa_flags = IFA_F_PERMANENT,
2979 .preferred_lft = INFINITY_LIFE_TIME,
2980 .valid_lft = INFINITY_LIFE_TIME,
2982 struct in6_ifreq ireq;
2985 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2988 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2991 cfg.pfx = &ireq.ifr6_addr;
2992 cfg.plen = ireq.ifr6_prefixlen;
2995 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3000 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3002 struct in6_ifreq ireq;
3005 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3008 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3012 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3013 ireq.ifr6_prefixlen);
3018 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3019 int plen, int scope)
3021 struct inet6_ifaddr *ifp;
3022 struct ifa6_config cfg = {
3025 .ifa_flags = IFA_F_PERMANENT,
3026 .valid_lft = INFINITY_LIFE_TIME,
3027 .preferred_lft = INFINITY_LIFE_TIME,
3031 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3033 spin_lock_bh(&ifp->lock);
3034 ifp->flags &= ~IFA_F_TENTATIVE;
3035 spin_unlock_bh(&ifp->lock);
3036 rt_genid_bump_ipv6(dev_net(idev->dev));
3037 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3042 #if IS_ENABLED(CONFIG_IPV6_SIT)
3043 static void sit_add_v4_addrs(struct inet6_dev *idev)
3045 struct in6_addr addr;
3046 struct net_device *dev;
3047 struct net *net = dev_net(idev->dev);
3053 memset(&addr, 0, sizeof(struct in6_addr));
3054 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3056 if (idev->dev->flags&IFF_POINTOPOINT) {
3057 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3060 addr.s6_addr32[0] = htonl(0xfe800000);
3064 scope = IPV6_ADDR_COMPATv4;
3066 pflags |= RTF_NONEXTHOP;
3069 if (addr.s6_addr32[3]) {
3070 add_addr(idev, &addr, plen, scope);
3071 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3076 for_each_netdev(net, dev) {
3077 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3078 if (in_dev && (dev->flags & IFF_UP)) {
3079 struct in_ifaddr *ifa;
3083 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3085 addr.s6_addr32[3] = ifa->ifa_local;
3087 if (ifa->ifa_scope == RT_SCOPE_LINK)
3089 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3090 if (idev->dev->flags&IFF_POINTOPOINT)
3095 add_addr(idev, &addr, plen, flag);
3096 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3097 0, pflags, GFP_KERNEL);
3104 static void init_loopback(struct net_device *dev)
3106 struct inet6_dev *idev;
3112 idev = ipv6_find_idev(dev);
3114 pr_debug("%s: add_dev failed\n", __func__);
3118 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3121 void addrconf_add_linklocal(struct inet6_dev *idev,
3122 const struct in6_addr *addr, u32 flags)
3124 struct ifa6_config cfg = {
3127 .ifa_flags = flags | IFA_F_PERMANENT,
3128 .valid_lft = INFINITY_LIFE_TIME,
3129 .preferred_lft = INFINITY_LIFE_TIME,
3132 struct inet6_ifaddr *ifp;
3134 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3135 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3136 idev->cnf.optimistic_dad) &&
3137 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3138 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3141 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3143 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3145 addrconf_dad_start(ifp);
3149 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3151 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3153 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3156 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3157 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3160 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3161 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3167 static int ipv6_generate_stable_address(struct in6_addr *address,
3169 const struct inet6_dev *idev)
3171 static DEFINE_SPINLOCK(lock);
3172 static __u32 digest[SHA_DIGEST_WORDS];
3173 static __u32 workspace[SHA_WORKSPACE_WORDS];
3176 char __data[SHA_MESSAGE_BYTES];
3178 struct in6_addr secret;
3180 unsigned char hwaddr[MAX_ADDR_LEN];
3185 struct in6_addr secret;
3186 struct in6_addr temp;
3187 struct net *net = dev_net(idev->dev);
3189 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3191 if (idev->cnf.stable_secret.initialized)
3192 secret = idev->cnf.stable_secret.secret;
3193 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3194 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3199 spin_lock_bh(&lock);
3202 memset(&data, 0, sizeof(data));
3203 memset(workspace, 0, sizeof(workspace));
3204 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3205 data.prefix[0] = address->s6_addr32[0];
3206 data.prefix[1] = address->s6_addr32[1];
3207 data.secret = secret;
3208 data.dad_count = dad_count;
3210 sha_transform(digest, data.__data, workspace);
3213 temp.s6_addr32[2] = (__force __be32)digest[0];
3214 temp.s6_addr32[3] = (__force __be32)digest[1];
3216 spin_unlock_bh(&lock);
3218 if (ipv6_reserved_interfaceid(temp)) {
3220 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3229 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3231 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3235 s = &idev->cnf.stable_secret;
3236 get_random_bytes(&s->secret, sizeof(s->secret));
3237 s->initialized = true;
3240 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3242 struct in6_addr addr;
3244 /* no link local addresses on L3 master devices */
3245 if (netif_is_l3_master(idev->dev))
3248 /* no link local addresses on devices flagged as slaves */
3249 if (idev->dev->flags & IFF_SLAVE)
3252 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3254 switch (idev->cnf.addr_gen_mode) {
3255 case IN6_ADDR_GEN_MODE_RANDOM:
3256 ipv6_gen_mode_random_init(idev);
3258 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3259 if (!ipv6_generate_stable_address(&addr, 0, idev))
3260 addrconf_add_linklocal(idev, &addr,
3261 IFA_F_STABLE_PRIVACY);
3262 else if (prefix_route)
3263 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3266 case IN6_ADDR_GEN_MODE_EUI64:
3267 /* addrconf_add_linklocal also adds a prefix_route and we
3268 * only need to care about prefix routes if ipv6_generate_eui64
3269 * couldn't generate one.
3271 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3272 addrconf_add_linklocal(idev, &addr, 0);
3273 else if (prefix_route)
3274 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3277 case IN6_ADDR_GEN_MODE_NONE:
3279 /* will not add any link local address */
3284 static void addrconf_dev_config(struct net_device *dev)
3286 struct inet6_dev *idev;
3290 if ((dev->type != ARPHRD_ETHER) &&
3291 (dev->type != ARPHRD_FDDI) &&
3292 (dev->type != ARPHRD_ARCNET) &&
3293 (dev->type != ARPHRD_INFINIBAND) &&
3294 (dev->type != ARPHRD_IEEE1394) &&
3295 (dev->type != ARPHRD_TUNNEL6) &&
3296 (dev->type != ARPHRD_6LOWPAN) &&
3297 (dev->type != ARPHRD_IP6GRE) &&
3298 (dev->type != ARPHRD_IPGRE) &&
3299 (dev->type != ARPHRD_TUNNEL) &&
3300 (dev->type != ARPHRD_NONE) &&
3301 (dev->type != ARPHRD_RAWIP)) {
3302 /* Alas, we support only Ethernet autoconfiguration. */
3303 idev = __in6_dev_get(dev);
3304 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3305 dev->flags & IFF_MULTICAST)
3310 idev = addrconf_add_dev(dev);
3314 /* this device type has no EUI support */
3315 if (dev->type == ARPHRD_NONE &&
3316 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3317 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3319 addrconf_addr_gen(idev, false);
3322 #if IS_ENABLED(CONFIG_IPV6_SIT)
3323 static void addrconf_sit_config(struct net_device *dev)
3325 struct inet6_dev *idev;
3330 * Configure the tunnel with one of our IPv4
3331 * addresses... we should configure all of
3332 * our v4 addrs in the tunnel
3335 idev = ipv6_find_idev(dev);
3337 pr_debug("%s: add_dev failed\n", __func__);
3341 if (dev->priv_flags & IFF_ISATAP) {
3342 addrconf_addr_gen(idev, false);
3346 sit_add_v4_addrs(idev);
3348 if (dev->flags&IFF_POINTOPOINT)
3349 addrconf_add_mroute(dev);
3353 #if IS_ENABLED(CONFIG_NET_IPGRE)
3354 static void addrconf_gre_config(struct net_device *dev)
3356 struct inet6_dev *idev;
3360 idev = ipv6_find_idev(dev);
3362 pr_debug("%s: add_dev failed\n", __func__);
3366 addrconf_addr_gen(idev, true);
3367 if (dev->flags & IFF_POINTOPOINT)
3368 addrconf_add_mroute(dev);
3372 static int fixup_permanent_addr(struct net *net,
3373 struct inet6_dev *idev,
3374 struct inet6_ifaddr *ifp)
3376 /* !fib6_node means the host route was removed from the
3377 * FIB, for example, if 'lo' device is taken down. In that
3378 * case regenerate the host route.
3380 if (!ifp->rt || !ifp->rt->fib6_node) {
3381 struct fib6_info *f6i, *prev;
3383 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3386 return PTR_ERR(f6i);
3388 /* ifp->rt can be accessed outside of rtnl */
3389 spin_lock(&ifp->lock);
3392 spin_unlock(&ifp->lock);
3394 fib6_info_release(prev);
3397 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3398 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3399 ifp->rt_priority, idev->dev, 0, 0,
3403 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3404 addrconf_dad_start(ifp);
3409 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3411 struct inet6_ifaddr *ifp, *tmp;
3412 struct inet6_dev *idev;
3414 idev = __in6_dev_get(dev);
3418 write_lock_bh(&idev->lock);
3420 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3421 if ((ifp->flags & IFA_F_PERMANENT) &&
3422 fixup_permanent_addr(net, idev, ifp) < 0) {
3423 write_unlock_bh(&idev->lock);
3426 write_lock_bh(&idev->lock);
3428 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3429 idev->dev->name, &ifp->addr);
3433 write_unlock_bh(&idev->lock);
3436 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3439 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3440 struct netdev_notifier_change_info *change_info;
3441 struct netdev_notifier_changeupper_info *info;
3442 struct inet6_dev *idev = __in6_dev_get(dev);
3443 struct net *net = dev_net(dev);
3444 int run_pending = 0;
3448 case NETDEV_REGISTER:
3449 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3450 idev = ipv6_add_dev(dev);
3452 return notifier_from_errno(PTR_ERR(idev));
3456 case NETDEV_CHANGEMTU:
3457 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3458 if (dev->mtu < IPV6_MIN_MTU) {
3459 addrconf_ifdown(dev, dev != net->loopback_dev);
3464 rt6_mtu_change(dev, dev->mtu);
3465 idev->cnf.mtu6 = dev->mtu;
3469 /* allocate new idev */
3470 idev = ipv6_add_dev(dev);
3474 /* device is still not ready */
3475 if (!(idev->if_flags & IF_READY))
3484 if (dev->flags & IFF_SLAVE)
3487 if (idev && idev->cnf.disable_ipv6)
3490 if (event == NETDEV_UP) {
3491 /* restore routes for permanent addresses */
3492 addrconf_permanent_addr(net, dev);
3494 if (!addrconf_link_ready(dev)) {
3495 /* device is not ready yet. */
3496 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3501 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3502 idev = ipv6_add_dev(dev);
3504 if (!IS_ERR_OR_NULL(idev)) {
3505 idev->if_flags |= IF_READY;
3508 } else if (event == NETDEV_CHANGE) {
3509 if (!addrconf_link_ready(dev)) {
3510 /* device is still not ready. */
3511 rt6_sync_down_dev(dev, event);
3515 if (!IS_ERR_OR_NULL(idev)) {
3516 if (idev->if_flags & IF_READY) {
3517 /* device is already configured -
3518 * but resend MLD reports, we might
3519 * have roamed and need to update
3520 * multicast snooping switches
3524 if (change_info->flags_changed & IFF_NOARP)
3525 addrconf_dad_run(idev, true);
3526 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3529 idev->if_flags |= IF_READY;
3532 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3538 switch (dev->type) {
3539 #if IS_ENABLED(CONFIG_IPV6_SIT)
3541 addrconf_sit_config(dev);
3544 #if IS_ENABLED(CONFIG_NET_IPGRE)
3546 addrconf_gre_config(dev);
3549 case ARPHRD_LOOPBACK:
3554 addrconf_dev_config(dev);
3558 if (!IS_ERR_OR_NULL(idev)) {
3560 addrconf_dad_run(idev, false);
3562 /* Device has an address by now */
3563 rt6_sync_up(dev, RTNH_F_DEAD);
3566 * If the MTU changed during the interface down,
3567 * when the interface up, the changed MTU must be
3568 * reflected in the idev as well as routers.
3570 if (idev->cnf.mtu6 != dev->mtu &&
3571 dev->mtu >= IPV6_MIN_MTU) {
3572 rt6_mtu_change(dev, dev->mtu);
3573 idev->cnf.mtu6 = dev->mtu;
3575 idev->tstamp = jiffies;
3576 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3579 * If the changed mtu during down is lower than
3580 * IPV6_MIN_MTU stop IPv6 on this interface.
3582 if (dev->mtu < IPV6_MIN_MTU)
3583 addrconf_ifdown(dev, dev != net->loopback_dev);
3588 case NETDEV_UNREGISTER:
3590 * Remove all addresses from this interface.
3592 addrconf_ifdown(dev, event != NETDEV_DOWN);
3595 case NETDEV_CHANGENAME:
3597 snmp6_unregister_dev(idev);
3598 addrconf_sysctl_unregister(idev);
3599 err = addrconf_sysctl_register(idev);
3601 return notifier_from_errno(err);
3602 err = snmp6_register_dev(idev);
3604 addrconf_sysctl_unregister(idev);
3605 return notifier_from_errno(err);
3610 case NETDEV_PRE_TYPE_CHANGE:
3611 case NETDEV_POST_TYPE_CHANGE:
3613 addrconf_type_change(dev, event);
3616 case NETDEV_CHANGEUPPER:
3619 /* flush all routes if dev is linked to or unlinked from
3620 * an L3 master device (e.g., VRF)
3622 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3623 addrconf_ifdown(dev, 0);
3630 * addrconf module should be notified of a device going up
3632 static struct notifier_block ipv6_dev_notf = {
3633 .notifier_call = addrconf_notify,
3634 .priority = ADDRCONF_NOTIFY_PRIORITY,
3637 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3639 struct inet6_dev *idev;
3642 idev = __in6_dev_get(dev);
3644 if (event == NETDEV_POST_TYPE_CHANGE)
3645 ipv6_mc_remap(idev);
3646 else if (event == NETDEV_PRE_TYPE_CHANGE)
3647 ipv6_mc_unmap(idev);
3650 static bool addr_is_local(const struct in6_addr *addr)
3652 return ipv6_addr_type(addr) &
3653 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3656 static int addrconf_ifdown(struct net_device *dev, int how)
3658 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3659 struct net *net = dev_net(dev);
3660 struct inet6_dev *idev;
3661 struct inet6_ifaddr *ifa, *tmp;
3662 bool keep_addr = false;
3667 rt6_disable_ip(dev, event);
3669 idev = __in6_dev_get(dev);
3674 * Step 1: remove reference to ipv6 device from parent device.
3680 /* protected by rtnl_lock */
3681 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3683 /* Step 1.5: remove snmp6 entry */
3684 snmp6_unregister_dev(idev);
3688 /* combine the user config with event to determine if permanent
3689 * addresses are to be removed from address hash table
3691 if (!how && !idev->cnf.disable_ipv6) {
3692 /* aggregate the system setting and interface setting */
3693 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3696 _keep_addr = idev->cnf.keep_addr_on_down;
3698 keep_addr = (_keep_addr > 0);
3701 /* Step 2: clear hash table */
3702 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3703 struct hlist_head *h = &inet6_addr_lst[i];
3705 spin_lock_bh(&addrconf_hash_lock);
3707 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3708 if (ifa->idev == idev) {
3709 addrconf_del_dad_work(ifa);
3710 /* combined flag + permanent flag decide if
3711 * address is retained on a down event
3714 !(ifa->flags & IFA_F_PERMANENT) ||
3715 addr_is_local(&ifa->addr)) {
3716 hlist_del_init_rcu(&ifa->addr_lst);
3721 spin_unlock_bh(&addrconf_hash_lock);
3724 write_lock_bh(&idev->lock);
3726 addrconf_del_rs_timer(idev);
3728 /* Step 2: clear flags for stateless addrconf */
3730 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3732 /* Step 3: clear tempaddr list */
3733 while (!list_empty(&idev->tempaddr_list)) {
3734 ifa = list_first_entry(&idev->tempaddr_list,
3735 struct inet6_ifaddr, tmp_list);
3736 list_del(&ifa->tmp_list);
3737 write_unlock_bh(&idev->lock);
3738 spin_lock_bh(&ifa->lock);
3741 in6_ifa_put(ifa->ifpub);
3744 spin_unlock_bh(&ifa->lock);
3746 write_lock_bh(&idev->lock);
3749 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3750 struct fib6_info *rt = NULL;
3753 addrconf_del_dad_work(ifa);
3755 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3756 !addr_is_local(&ifa->addr);
3758 write_unlock_bh(&idev->lock);
3759 spin_lock_bh(&ifa->lock);
3762 /* set state to skip the notifier below */
3763 state = INET6_IFADDR_STATE_DEAD;
3764 ifa->state = INET6_IFADDR_STATE_PREDAD;
3765 if (!(ifa->flags & IFA_F_NODAD))
3766 ifa->flags |= IFA_F_TENTATIVE;
3772 ifa->state = INET6_IFADDR_STATE_DEAD;
3775 spin_unlock_bh(&ifa->lock);
3778 ip6_del_rt(net, rt);
3780 if (state != INET6_IFADDR_STATE_DEAD) {
3781 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3782 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3784 if (idev->cnf.forwarding)
3785 addrconf_leave_anycast(ifa);
3786 addrconf_leave_solict(ifa->idev, &ifa->addr);
3789 write_lock_bh(&idev->lock);
3791 list_del_rcu(&ifa->if_list);
3796 write_unlock_bh(&idev->lock);
3798 /* Step 5: Discard anycast and multicast list */
3800 ipv6_ac_destroy_dev(idev);
3801 ipv6_mc_destroy_dev(idev);
3806 idev->tstamp = jiffies;
3808 /* Last: Shot the device (if unregistered) */
3810 addrconf_sysctl_unregister(idev);
3811 neigh_parms_release(&nd_tbl, idev->nd_parms);
3812 neigh_ifdown(&nd_tbl, dev);
3818 static void addrconf_rs_timer(struct timer_list *t)
3820 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3821 struct net_device *dev = idev->dev;
3822 struct in6_addr lladdr;
3824 write_lock(&idev->lock);
3825 if (idev->dead || !(idev->if_flags & IF_READY))
3828 if (!ipv6_accept_ra(idev))
3831 /* Announcement received after solicitation was sent */
3832 if (idev->if_flags & IF_RA_RCVD)
3835 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3836 write_unlock(&idev->lock);
3837 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3838 ndisc_send_rs(dev, &lladdr,
3839 &in6addr_linklocal_allrouters);
3843 write_lock(&idev->lock);
3844 idev->rs_interval = rfc3315_s14_backoff_update(
3845 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3846 /* The wait after the last probe can be shorter */
3847 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3848 idev->cnf.rtr_solicits) ?
3849 idev->cnf.rtr_solicit_delay :
3853 * Note: we do not support deprecated "all on-link"
3854 * assumption any longer.
3856 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3860 write_unlock(&idev->lock);
3866 * Duplicate Address Detection
3868 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3870 unsigned long rand_num;
3871 struct inet6_dev *idev = ifp->idev;
3874 if (ifp->flags & IFA_F_OPTIMISTIC)
3877 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3880 if (idev->cnf.enhanced_dad ||
3881 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3883 get_random_bytes(&nonce, 6);
3886 ifp->dad_nonce = nonce;
3887 ifp->dad_probes = idev->cnf.dad_transmits;
3888 addrconf_mod_dad_work(ifp, rand_num);
3891 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3893 struct inet6_dev *idev = ifp->idev;
3894 struct net_device *dev = idev->dev;
3895 bool bump_id, notify = false;
3898 addrconf_join_solict(dev, &ifp->addr);
3900 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3902 read_lock_bh(&idev->lock);
3903 spin_lock(&ifp->lock);
3904 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3908 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3909 (net->ipv6.devconf_all->accept_dad < 1 &&
3910 idev->cnf.accept_dad < 1) ||
3911 !(ifp->flags&IFA_F_TENTATIVE) ||
3912 ifp->flags & IFA_F_NODAD) {
3913 bool send_na = false;
3915 if (ifp->flags & IFA_F_TENTATIVE &&
3916 !(ifp->flags & IFA_F_OPTIMISTIC))
3918 bump_id = ifp->flags & IFA_F_TENTATIVE;
3919 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3920 spin_unlock(&ifp->lock);
3921 read_unlock_bh(&idev->lock);
3923 addrconf_dad_completed(ifp, bump_id, send_na);
3927 if (!(idev->if_flags & IF_READY)) {
3928 spin_unlock(&ifp->lock);
3929 read_unlock_bh(&idev->lock);
3931 * If the device is not ready:
3932 * - keep it tentative if it is a permanent address.
3933 * - otherwise, kill it.
3936 addrconf_dad_stop(ifp, 0);
3941 * Optimistic nodes can start receiving
3944 if (ifp->flags & IFA_F_OPTIMISTIC) {
3945 ip6_ins_rt(net, ifp->rt);
3946 if (ipv6_use_optimistic_addr(net, idev)) {
3947 /* Because optimistic nodes can use this address,
3948 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3954 addrconf_dad_kick(ifp);
3956 spin_unlock(&ifp->lock);
3957 read_unlock_bh(&idev->lock);
3959 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3962 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3964 bool begin_dad = false;
3966 spin_lock_bh(&ifp->lock);
3967 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3968 ifp->state = INET6_IFADDR_STATE_PREDAD;
3971 spin_unlock_bh(&ifp->lock);
3974 addrconf_mod_dad_work(ifp, 0);
3977 static void addrconf_dad_work(struct work_struct *w)
3979 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3980 struct inet6_ifaddr,
3982 struct inet6_dev *idev = ifp->idev;
3983 bool bump_id, disable_ipv6 = false;
3984 struct in6_addr mcaddr;
3990 } action = DAD_PROCESS;
3994 spin_lock_bh(&ifp->lock);
3995 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3997 ifp->state = INET6_IFADDR_STATE_DAD;
3998 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4000 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4002 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4003 idev->cnf.accept_dad > 1) &&
4004 !idev->cnf.disable_ipv6 &&
4005 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4006 struct in6_addr addr;
4008 addr.s6_addr32[0] = htonl(0xfe800000);
4009 addr.s6_addr32[1] = 0;
4011 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4012 ipv6_addr_equal(&ifp->addr, &addr)) {
4013 /* DAD failed for link-local based on MAC */
4014 idev->cnf.disable_ipv6 = 1;
4016 pr_info("%s: IPv6 being disabled!\n",
4017 ifp->idev->dev->name);
4018 disable_ipv6 = true;
4022 spin_unlock_bh(&ifp->lock);
4024 if (action == DAD_BEGIN) {
4025 addrconf_dad_begin(ifp);
4027 } else if (action == DAD_ABORT) {
4029 addrconf_dad_stop(ifp, 1);
4031 addrconf_ifdown(idev->dev, 0);
4035 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4038 write_lock_bh(&idev->lock);
4039 if (idev->dead || !(idev->if_flags & IF_READY)) {
4040 write_unlock_bh(&idev->lock);
4044 spin_lock(&ifp->lock);
4045 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4046 spin_unlock(&ifp->lock);
4047 write_unlock_bh(&idev->lock);
4051 if (ifp->dad_probes == 0) {
4052 bool send_na = false;
4055 * DAD was successful
4058 if (ifp->flags & IFA_F_TENTATIVE &&
4059 !(ifp->flags & IFA_F_OPTIMISTIC))
4061 bump_id = ifp->flags & IFA_F_TENTATIVE;
4062 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4063 spin_unlock(&ifp->lock);
4064 write_unlock_bh(&idev->lock);
4066 addrconf_dad_completed(ifp, bump_id, send_na);
4072 addrconf_mod_dad_work(ifp,
4073 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4074 spin_unlock(&ifp->lock);
4075 write_unlock_bh(&idev->lock);
4077 /* send a neighbour solicitation for our addr */
4078 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4079 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4086 /* ifp->idev must be at least read locked */
4087 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4089 struct inet6_ifaddr *ifpiter;
4090 struct inet6_dev *idev = ifp->idev;
4092 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4093 if (ifpiter->scope > IFA_LINK)
4095 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4096 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4097 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4104 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4107 struct net_device *dev = ifp->idev->dev;
4108 struct in6_addr lladdr;
4109 bool send_rs, send_mld;
4111 addrconf_del_dad_work(ifp);
4114 * Configure the address for reception. Now it is valid.
4117 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4119 /* If added prefix is link local and we are prepared to process
4120 router advertisements, start sending router solicitations.
4123 read_lock_bh(&ifp->idev->lock);
4124 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4125 send_rs = send_mld &&
4126 ipv6_accept_ra(ifp->idev) &&
4127 ifp->idev->cnf.rtr_solicits != 0 &&
4128 (dev->flags&IFF_LOOPBACK) == 0;
4129 read_unlock_bh(&ifp->idev->lock);
4131 /* While dad is in progress mld report's source address is in6_addrany.
4132 * Resend with proper ll now.
4135 ipv6_mc_dad_complete(ifp->idev);
4137 /* send unsolicited NA if enabled */
4139 (ifp->idev->cnf.ndisc_notify ||
4140 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4141 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4142 /*router=*/ !!ifp->idev->cnf.forwarding,
4143 /*solicited=*/ false, /*override=*/ true,
4149 * If a host as already performed a random delay
4150 * [...] as part of DAD [...] there is no need
4151 * to delay again before sending the first RS
4153 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4155 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4157 write_lock_bh(&ifp->idev->lock);
4158 spin_lock(&ifp->lock);
4159 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4160 ifp->idev->cnf.rtr_solicit_interval);
4161 ifp->idev->rs_probes = 1;
4162 ifp->idev->if_flags |= IF_RS_SENT;
4163 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4164 spin_unlock(&ifp->lock);
4165 write_unlock_bh(&ifp->idev->lock);
4169 rt_genid_bump_ipv6(dev_net(dev));
4171 /* Make sure that a new temporary address will be created
4172 * before this temporary address becomes deprecated.
4174 if (ifp->flags & IFA_F_TEMPORARY)
4175 addrconf_verify_rtnl();
4178 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4180 struct inet6_ifaddr *ifp;
4182 read_lock_bh(&idev->lock);
4183 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4184 spin_lock(&ifp->lock);
4185 if ((ifp->flags & IFA_F_TENTATIVE &&
4186 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4188 ifp->state = INET6_IFADDR_STATE_PREDAD;
4189 addrconf_dad_kick(ifp);
4191 spin_unlock(&ifp->lock);
4193 read_unlock_bh(&idev->lock);
4196 #ifdef CONFIG_PROC_FS
4197 struct if6_iter_state {
4198 struct seq_net_private p;
4203 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4205 struct if6_iter_state *state = seq->private;
4206 struct net *net = seq_file_net(seq);
4207 struct inet6_ifaddr *ifa = NULL;
4210 /* initial bucket if pos is 0 */
4216 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4217 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4219 if (!net_eq(dev_net(ifa->idev->dev), net))
4221 /* sync with offset */
4222 if (p < state->offset) {
4229 /* prepare for next bucket */
4236 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4237 struct inet6_ifaddr *ifa)
4239 struct if6_iter_state *state = seq->private;
4240 struct net *net = seq_file_net(seq);
4242 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4243 if (!net_eq(dev_net(ifa->idev->dev), net))
4250 while (++state->bucket < IN6_ADDR_HSIZE) {
4251 hlist_for_each_entry_rcu(ifa,
4252 &inet6_addr_lst[state->bucket], addr_lst) {
4253 if (!net_eq(dev_net(ifa->idev->dev), net))
4262 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4266 return if6_get_first(seq, *pos);
4269 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4271 struct inet6_ifaddr *ifa;
4273 ifa = if6_get_next(seq, v);
4278 static void if6_seq_stop(struct seq_file *seq, void *v)
4284 static int if6_seq_show(struct seq_file *seq, void *v)
4286 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4287 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4289 ifp->idev->dev->ifindex,
4293 ifp->idev->dev->name);
4297 static const struct seq_operations if6_seq_ops = {
4298 .start = if6_seq_start,
4299 .next = if6_seq_next,
4300 .show = if6_seq_show,
4301 .stop = if6_seq_stop,
4304 static int __net_init if6_proc_net_init(struct net *net)
4306 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4307 sizeof(struct if6_iter_state)))
4312 static void __net_exit if6_proc_net_exit(struct net *net)
4314 remove_proc_entry("if_inet6", net->proc_net);
4317 static struct pernet_operations if6_proc_net_ops = {
4318 .init = if6_proc_net_init,
4319 .exit = if6_proc_net_exit,
4322 int __init if6_proc_init(void)
4324 return register_pernet_subsys(&if6_proc_net_ops);
4327 void if6_proc_exit(void)
4329 unregister_pernet_subsys(&if6_proc_net_ops);
4331 #endif /* CONFIG_PROC_FS */
4333 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4334 /* Check if address is a home address configured on any interface. */
4335 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4337 unsigned int hash = inet6_addr_hash(net, addr);
4338 struct inet6_ifaddr *ifp = NULL;
4342 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4343 if (!net_eq(dev_net(ifp->idev->dev), net))
4345 if (ipv6_addr_equal(&ifp->addr, addr) &&
4346 (ifp->flags & IFA_F_HOMEADDRESS)) {
4357 * Periodic address status verification
4360 static void addrconf_verify_rtnl(void)
4362 unsigned long now, next, next_sec, next_sched;
4363 struct inet6_ifaddr *ifp;
4370 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4372 cancel_delayed_work(&addr_chk_work);
4374 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4376 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4379 /* When setting preferred_lft to a value not zero or
4380 * infinity, while valid_lft is infinity
4381 * IFA_F_PERMANENT has a non-infinity life time.
4383 if ((ifp->flags & IFA_F_PERMANENT) &&
4384 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4387 spin_lock(&ifp->lock);
4388 /* We try to batch several events at once. */
4389 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4391 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4392 age >= ifp->valid_lft) {
4393 spin_unlock(&ifp->lock);
4397 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4398 spin_unlock(&ifp->lock);
4400 } else if (age >= ifp->prefered_lft) {
4401 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4404 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4406 ifp->flags |= IFA_F_DEPRECATED;
4409 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4410 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4411 next = ifp->tstamp + ifp->valid_lft * HZ;
4413 spin_unlock(&ifp->lock);
4418 ipv6_ifa_notify(0, ifp);
4422 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4423 !(ifp->flags&IFA_F_TENTATIVE)) {
4424 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4425 ifp->idev->cnf.dad_transmits *
4426 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4428 if (age >= ifp->prefered_lft - regen_advance) {
4429 struct inet6_ifaddr *ifpub = ifp->ifpub;
4430 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4431 next = ifp->tstamp + ifp->prefered_lft * HZ;
4432 if (!ifp->regen_count && ifpub) {
4435 in6_ifa_hold(ifpub);
4436 spin_unlock(&ifp->lock);
4438 spin_lock(&ifpub->lock);
4439 ifpub->regen_count = 0;
4440 spin_unlock(&ifpub->lock);
4441 rcu_read_unlock_bh();
4442 ipv6_create_tempaddr(ifpub, ifp, true);
4448 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4449 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4450 spin_unlock(&ifp->lock);
4452 /* ifp->prefered_lft <= ifp->valid_lft */
4453 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4454 next = ifp->tstamp + ifp->prefered_lft * HZ;
4455 spin_unlock(&ifp->lock);
4460 next_sec = round_jiffies_up(next);
4463 /* If rounded timeout is accurate enough, accept it. */
4464 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4465 next_sched = next_sec;
4467 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4468 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4469 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4471 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4472 now, next, next_sec, next_sched);
4473 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4474 rcu_read_unlock_bh();
4477 static void addrconf_verify_work(struct work_struct *w)
4480 addrconf_verify_rtnl();
4484 static void addrconf_verify(void)
4486 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4489 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4490 struct in6_addr **peer_pfx)
4492 struct in6_addr *pfx = NULL;
4497 pfx = nla_data(addr);
4500 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4502 pfx = nla_data(local);
4508 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4509 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4510 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4511 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4512 [IFA_FLAGS] = { .len = sizeof(u32) },
4513 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4517 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4518 struct netlink_ext_ack *extack)
4520 struct net *net = sock_net(skb->sk);
4521 struct ifaddrmsg *ifm;
4522 struct nlattr *tb[IFA_MAX+1];
4523 struct in6_addr *pfx, *peer_pfx;
4527 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4532 ifm = nlmsg_data(nlh);
4533 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4537 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4539 /* We ignore other flags so far. */
4540 ifa_flags &= IFA_F_MANAGETEMPADDR;
4542 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4543 ifm->ifa_prefixlen);
4546 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4547 unsigned long expires, u32 flags,
4550 struct fib6_info *f6i;
4553 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4556 0, RTF_GATEWAY | RTF_DEFAULT);
4560 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4561 if (f6i->fib6_metric != prio) {
4562 /* delete old one */
4563 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4566 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4568 ifp->rt_priority, ifp->idev->dev,
4569 expires, flags, GFP_KERNEL);
4572 fib6_clean_expires(f6i);
4574 fib6_set_expires(f6i, expires);
4576 fib6_info_release(f6i);
4582 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4586 unsigned long timeout;
4587 bool was_managetempaddr;
4588 bool had_prefixroute;
4589 bool new_peer = false;
4593 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4596 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4597 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4600 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4601 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4603 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4604 if (addrconf_finite_timeout(timeout)) {
4605 expires = jiffies_to_clock_t(timeout * HZ);
4606 cfg->valid_lft = timeout;
4607 flags = RTF_EXPIRES;
4611 cfg->ifa_flags |= IFA_F_PERMANENT;
4614 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4615 if (addrconf_finite_timeout(timeout)) {
4617 cfg->ifa_flags |= IFA_F_DEPRECATED;
4618 cfg->preferred_lft = timeout;
4621 if (cfg->peer_pfx &&
4622 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4623 if (!ipv6_addr_any(&ifp->peer_addr))
4624 cleanup_prefix_route(ifp, expires, true, true);
4628 spin_lock_bh(&ifp->lock);
4629 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4630 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4631 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4632 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4633 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4634 IFA_F_NOPREFIXROUTE);
4635 ifp->flags |= cfg->ifa_flags;
4636 ifp->tstamp = jiffies;
4637 ifp->valid_lft = cfg->valid_lft;
4638 ifp->prefered_lft = cfg->preferred_lft;
4640 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4641 ifp->rt_priority = cfg->rt_priority;
4644 ifp->peer_addr = *cfg->peer_pfx;
4646 spin_unlock_bh(&ifp->lock);
4647 if (!(ifp->flags&IFA_F_TENTATIVE))
4648 ipv6_ifa_notify(0, ifp);
4650 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4653 if (had_prefixroute)
4654 rc = modify_prefix_route(ifp, expires, flags, false);
4656 /* prefix route could have been deleted; if so restore it */
4657 if (rc == -ENOENT) {
4658 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4659 ifp->rt_priority, ifp->idev->dev,
4660 expires, flags, GFP_KERNEL);
4663 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4664 rc = modify_prefix_route(ifp, expires, flags, true);
4666 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4667 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4668 ifp->rt_priority, ifp->idev->dev,
4669 expires, flags, GFP_KERNEL);
4671 } else if (had_prefixroute) {
4672 enum cleanup_prefix_rt_t action;
4673 unsigned long rt_expires;
4675 write_lock_bh(&ifp->idev->lock);
4676 action = check_cleanup_prefix_route(ifp, &rt_expires);
4677 write_unlock_bh(&ifp->idev->lock);
4679 if (action != CLEANUP_PREFIX_RT_NOP) {
4680 cleanup_prefix_route(ifp, rt_expires,
4681 action == CLEANUP_PREFIX_RT_DEL, false);
4685 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4686 if (was_managetempaddr &&
4687 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4689 cfg->preferred_lft = 0;
4691 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4692 cfg->preferred_lft, !was_managetempaddr,
4696 addrconf_verify_rtnl();
4702 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4703 struct netlink_ext_ack *extack)
4705 struct net *net = sock_net(skb->sk);
4706 struct ifaddrmsg *ifm;
4707 struct nlattr *tb[IFA_MAX+1];
4708 struct in6_addr *peer_pfx;
4709 struct inet6_ifaddr *ifa;
4710 struct net_device *dev;
4711 struct inet6_dev *idev;
4712 struct ifa6_config cfg;
4715 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4720 memset(&cfg, 0, sizeof(cfg));
4722 ifm = nlmsg_data(nlh);
4723 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4727 cfg.peer_pfx = peer_pfx;
4728 cfg.plen = ifm->ifa_prefixlen;
4729 if (tb[IFA_RT_PRIORITY])
4730 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4732 cfg.valid_lft = INFINITY_LIFE_TIME;
4733 cfg.preferred_lft = INFINITY_LIFE_TIME;
4735 if (tb[IFA_CACHEINFO]) {
4736 struct ifa_cacheinfo *ci;
4738 ci = nla_data(tb[IFA_CACHEINFO]);
4739 cfg.valid_lft = ci->ifa_valid;
4740 cfg.preferred_lft = ci->ifa_prefered;
4743 dev = __dev_get_by_index(net, ifm->ifa_index);
4748 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4750 cfg.ifa_flags = ifm->ifa_flags;
4752 /* We ignore other flags so far. */
4753 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4754 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4755 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4757 idev = ipv6_find_idev(dev);
4761 if (!ipv6_allow_optimistic_dad(net, idev))
4762 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4764 if (cfg.ifa_flags & IFA_F_NODAD &&
4765 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4766 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4770 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4773 * It would be best to check for !NLM_F_CREATE here but
4774 * userspace already relies on not having to provide this.
4776 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4779 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4780 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4783 err = inet6_addr_modify(ifa, &cfg);
4790 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4791 u8 scope, int ifindex)
4793 struct ifaddrmsg *ifm;
4795 ifm = nlmsg_data(nlh);
4796 ifm->ifa_family = AF_INET6;
4797 ifm->ifa_prefixlen = prefixlen;
4798 ifm->ifa_flags = flags;
4799 ifm->ifa_scope = scope;
4800 ifm->ifa_index = ifindex;
4803 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4804 unsigned long tstamp, u32 preferred, u32 valid)
4806 struct ifa_cacheinfo ci;
4808 ci.cstamp = cstamp_delta(cstamp);
4809 ci.tstamp = cstamp_delta(tstamp);
4810 ci.ifa_prefered = preferred;
4811 ci.ifa_valid = valid;
4813 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4816 static inline int rt_scope(int ifa_scope)
4818 if (ifa_scope & IFA_HOST)
4819 return RT_SCOPE_HOST;
4820 else if (ifa_scope & IFA_LINK)
4821 return RT_SCOPE_LINK;
4822 else if (ifa_scope & IFA_SITE)
4823 return RT_SCOPE_SITE;
4825 return RT_SCOPE_UNIVERSE;
4828 static inline int inet6_ifaddr_msgsize(void)
4830 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4831 + nla_total_size(16) /* IFA_LOCAL */
4832 + nla_total_size(16) /* IFA_ADDRESS */
4833 + nla_total_size(sizeof(struct ifa_cacheinfo))
4834 + nla_total_size(4) /* IFA_FLAGS */
4835 + nla_total_size(4) /* IFA_RT_PRIORITY */;
4838 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4839 u32 portid, u32 seq, int event, unsigned int flags)
4841 struct nlmsghdr *nlh;
4842 u32 preferred, valid;
4844 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4848 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4849 ifa->idev->dev->ifindex);
4851 if (!((ifa->flags&IFA_F_PERMANENT) &&
4852 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4853 preferred = ifa->prefered_lft;
4854 valid = ifa->valid_lft;
4855 if (preferred != INFINITY_LIFE_TIME) {
4856 long tval = (jiffies - ifa->tstamp)/HZ;
4857 if (preferred > tval)
4861 if (valid != INFINITY_LIFE_TIME) {
4869 preferred = INFINITY_LIFE_TIME;
4870 valid = INFINITY_LIFE_TIME;
4873 if (!ipv6_addr_any(&ifa->peer_addr)) {
4874 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4875 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4878 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4881 if (ifa->rt_priority &&
4882 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4885 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4888 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4891 nlmsg_end(skb, nlh);
4895 nlmsg_cancel(skb, nlh);
4899 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4900 u32 portid, u32 seq, int event, u16 flags)
4902 struct nlmsghdr *nlh;
4903 u8 scope = RT_SCOPE_UNIVERSE;
4904 int ifindex = ifmca->idev->dev->ifindex;
4906 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4907 scope = RT_SCOPE_SITE;
4909 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4913 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4914 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4915 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4916 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4917 nlmsg_cancel(skb, nlh);
4921 nlmsg_end(skb, nlh);
4925 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4926 u32 portid, u32 seq, int event, unsigned int flags)
4928 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4929 int ifindex = dev ? dev->ifindex : 1;
4930 struct nlmsghdr *nlh;
4931 u8 scope = RT_SCOPE_UNIVERSE;
4933 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4934 scope = RT_SCOPE_SITE;
4936 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4940 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4941 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4942 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4943 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4944 nlmsg_cancel(skb, nlh);
4948 nlmsg_end(skb, nlh);
4958 /* called with rcu_read_lock() */
4959 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4960 struct netlink_callback *cb, enum addr_type_t type,
4961 int s_ip_idx, int *p_ip_idx)
4963 struct ifmcaddr6 *ifmca;
4964 struct ifacaddr6 *ifaca;
4966 int ip_idx = *p_ip_idx;
4968 read_lock_bh(&idev->lock);
4970 case UNICAST_ADDR: {
4971 struct inet6_ifaddr *ifa;
4973 /* unicast address incl. temp addr */
4974 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4975 if (ip_idx < s_ip_idx)
4977 err = inet6_fill_ifaddr(skb, ifa,
4978 NETLINK_CB(cb->skb).portid,
4984 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4990 case MULTICAST_ADDR:
4991 /* multicast address */
4992 for (ifmca = idev->mc_list; ifmca;
4993 ifmca = ifmca->next, ip_idx++) {
4994 if (ip_idx < s_ip_idx)
4996 err = inet6_fill_ifmcaddr(skb, ifmca,
4997 NETLINK_CB(cb->skb).portid,
5006 /* anycast address */
5007 for (ifaca = idev->ac_list; ifaca;
5008 ifaca = ifaca->aca_next, ip_idx++) {
5009 if (ip_idx < s_ip_idx)
5011 err = inet6_fill_ifacaddr(skb, ifaca,
5012 NETLINK_CB(cb->skb).portid,
5023 read_unlock_bh(&idev->lock);
5028 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5029 enum addr_type_t type)
5031 struct net *net = sock_net(skb->sk);
5034 int s_idx, s_ip_idx;
5035 struct net_device *dev;
5036 struct inet6_dev *idev;
5037 struct hlist_head *head;
5040 s_idx = idx = cb->args[1];
5041 s_ip_idx = ip_idx = cb->args[2];
5044 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
5045 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5047 head = &net->dev_index_head[h];
5048 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5051 if (h > s_h || idx > s_idx)
5054 idev = __in6_dev_get(dev);
5058 if (in6_dump_addrs(idev, skb, cb, type,
5059 s_ip_idx, &ip_idx) < 0)
5069 cb->args[2] = ip_idx;
5074 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5076 enum addr_type_t type = UNICAST_ADDR;
5078 return inet6_dump_addr(skb, cb, type);
5081 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5083 enum addr_type_t type = MULTICAST_ADDR;
5085 return inet6_dump_addr(skb, cb, type);
5089 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5091 enum addr_type_t type = ANYCAST_ADDR;
5093 return inet6_dump_addr(skb, cb, type);
5096 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5097 struct netlink_ext_ack *extack)
5099 struct net *net = sock_net(in_skb->sk);
5100 struct ifaddrmsg *ifm;
5101 struct nlattr *tb[IFA_MAX+1];
5102 struct in6_addr *addr = NULL, *peer;
5103 struct net_device *dev = NULL;
5104 struct inet6_ifaddr *ifa;
5105 struct sk_buff *skb;
5108 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5113 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5117 ifm = nlmsg_data(nlh);
5119 dev = dev_get_by_index(net, ifm->ifa_index);
5121 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5123 err = -EADDRNOTAVAIL;
5127 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5133 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5134 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5136 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5137 WARN_ON(err == -EMSGSIZE);
5141 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5150 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5152 struct sk_buff *skb;
5153 struct net *net = dev_net(ifa->idev->dev);
5156 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5160 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5162 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5163 WARN_ON(err == -EMSGSIZE);
5167 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5171 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5174 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5175 __s32 *array, int bytes)
5177 BUG_ON(bytes < (DEVCONF_MAX * 4));
5179 memset(array, 0, bytes);
5180 array[DEVCONF_FORWARDING] = cnf->forwarding;
5181 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5182 array[DEVCONF_MTU6] = cnf->mtu6;
5183 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5184 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5185 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5186 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5187 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5188 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5189 jiffies_to_msecs(cnf->rtr_solicit_interval);
5190 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5191 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5192 array[DEVCONF_RTR_SOLICIT_DELAY] =
5193 jiffies_to_msecs(cnf->rtr_solicit_delay);
5194 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5195 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5196 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5197 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5198 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5199 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5200 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5201 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5202 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5203 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5204 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5205 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5206 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5207 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5208 #ifdef CONFIG_IPV6_ROUTER_PREF
5209 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5210 array[DEVCONF_RTR_PROBE_INTERVAL] =
5211 jiffies_to_msecs(cnf->rtr_probe_interval);
5212 #ifdef CONFIG_IPV6_ROUTE_INFO
5213 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5214 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5217 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5218 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5219 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5220 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5221 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5223 #ifdef CONFIG_IPV6_MROUTE
5224 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5226 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5227 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5228 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5229 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5230 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5231 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5232 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5233 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5234 /* we omit DEVCONF_STABLE_SECRET for now */
5235 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5236 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5237 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5238 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5239 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5240 #ifdef CONFIG_IPV6_SEG6_HMAC
5241 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5243 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5244 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5245 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5246 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5249 static inline size_t inet6_ifla6_size(void)
5251 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5252 + nla_total_size(sizeof(struct ifla_cacheinfo))
5253 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5254 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5255 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5256 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5257 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5261 static inline size_t inet6_if_nlmsg_size(void)
5263 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5264 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5265 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5266 + nla_total_size(4) /* IFLA_MTU */
5267 + nla_total_size(4) /* IFLA_LINK */
5268 + nla_total_size(1) /* IFLA_OPERSTATE */
5269 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5272 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5276 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5279 /* Use put_unaligned() because stats may not be aligned for u64. */
5280 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5281 for (i = 1; i < ICMP6_MIB_MAX; i++)
5282 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5284 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5287 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5288 int bytes, size_t syncpoff)
5291 u64 buff[IPSTATS_MIB_MAX];
5292 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5296 memset(buff, 0, sizeof(buff));
5297 buff[0] = IPSTATS_MIB_MAX;
5299 for_each_possible_cpu(c) {
5300 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5301 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5304 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5305 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5308 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5312 case IFLA_INET6_STATS:
5313 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5314 offsetof(struct ipstats_mib, syncp));
5316 case IFLA_INET6_ICMP6STATS:
5317 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5322 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5323 u32 ext_filter_mask)
5326 struct ifla_cacheinfo ci;
5328 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5329 goto nla_put_failure;
5330 ci.max_reasm_len = IPV6_MAXPLEN;
5331 ci.tstamp = cstamp_delta(idev->tstamp);
5332 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5333 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5334 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5335 goto nla_put_failure;
5336 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5338 goto nla_put_failure;
5339 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5341 /* XXX - MC not implemented */
5343 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5346 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5348 goto nla_put_failure;
5349 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5351 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5353 goto nla_put_failure;
5354 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5356 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5358 goto nla_put_failure;
5360 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5361 goto nla_put_failure;
5363 read_lock_bh(&idev->lock);
5364 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5365 read_unlock_bh(&idev->lock);
5373 static size_t inet6_get_link_af_size(const struct net_device *dev,
5374 u32 ext_filter_mask)
5376 if (!__in6_dev_get(dev))
5379 return inet6_ifla6_size();
5382 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5383 u32 ext_filter_mask)
5385 struct inet6_dev *idev = __in6_dev_get(dev);
5390 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5396 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5398 struct inet6_ifaddr *ifp;
5399 struct net_device *dev = idev->dev;
5400 bool clear_token, update_rs = false;
5401 struct in6_addr ll_addr;
5407 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5409 if (!ipv6_accept_ra(idev))
5411 if (idev->cnf.rtr_solicits == 0)
5414 write_lock_bh(&idev->lock);
5416 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5417 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5419 write_unlock_bh(&idev->lock);
5421 clear_token = ipv6_addr_any(token);
5425 if (!idev->dead && (idev->if_flags & IF_READY) &&
5426 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5427 IFA_F_OPTIMISTIC)) {
5428 /* If we're not ready, then normal ifup will take care
5429 * of this. Otherwise, we need to request our rs here.
5431 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5436 write_lock_bh(&idev->lock);
5439 idev->if_flags |= IF_RS_SENT;
5440 idev->rs_interval = rfc3315_s14_backoff_init(
5441 idev->cnf.rtr_solicit_interval);
5442 idev->rs_probes = 1;
5443 addrconf_mod_rs_timer(idev, idev->rs_interval);
5446 /* Well, that's kinda nasty ... */
5447 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5448 spin_lock(&ifp->lock);
5449 if (ifp->tokenized) {
5451 ifp->prefered_lft = 0;
5453 spin_unlock(&ifp->lock);
5456 write_unlock_bh(&idev->lock);
5457 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5458 addrconf_verify_rtnl();
5462 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5463 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5464 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5467 static int inet6_validate_link_af(const struct net_device *dev,
5468 const struct nlattr *nla)
5470 struct nlattr *tb[IFLA_INET6_MAX + 1];
5472 if (dev && !__in6_dev_get(dev))
5473 return -EAFNOSUPPORT;
5475 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5479 static int check_addr_gen_mode(int mode)
5481 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5482 mode != IN6_ADDR_GEN_MODE_NONE &&
5483 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5484 mode != IN6_ADDR_GEN_MODE_RANDOM)
5489 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5492 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5493 !idev->cnf.stable_secret.initialized &&
5494 !net->ipv6.devconf_dflt->stable_secret.initialized)
5499 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5502 struct inet6_dev *idev = __in6_dev_get(dev);
5503 struct nlattr *tb[IFLA_INET6_MAX + 1];
5506 return -EAFNOSUPPORT;
5508 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5511 if (tb[IFLA_INET6_TOKEN]) {
5512 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5517 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5518 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5520 if (check_addr_gen_mode(mode) < 0 ||
5521 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5524 idev->cnf.addr_gen_mode = mode;
5531 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5532 u32 portid, u32 seq, int event, unsigned int flags)
5534 struct net_device *dev = idev->dev;
5535 struct ifinfomsg *hdr;
5536 struct nlmsghdr *nlh;
5539 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5543 hdr = nlmsg_data(nlh);
5544 hdr->ifi_family = AF_INET6;
5546 hdr->ifi_type = dev->type;
5547 hdr->ifi_index = dev->ifindex;
5548 hdr->ifi_flags = dev_get_flags(dev);
5549 hdr->ifi_change = 0;
5551 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5553 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5554 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5555 (dev->ifindex != dev_get_iflink(dev) &&
5556 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5557 nla_put_u8(skb, IFLA_OPERSTATE,
5558 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5559 goto nla_put_failure;
5560 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5562 goto nla_put_failure;
5564 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5565 goto nla_put_failure;
5567 nla_nest_end(skb, protoinfo);
5568 nlmsg_end(skb, nlh);
5572 nlmsg_cancel(skb, nlh);
5576 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5578 struct net *net = sock_net(skb->sk);
5581 struct net_device *dev;
5582 struct inet6_dev *idev;
5583 struct hlist_head *head;
5586 s_idx = cb->args[1];
5589 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5591 head = &net->dev_index_head[h];
5592 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5595 idev = __in6_dev_get(dev);
5598 if (inet6_fill_ifinfo(skb, idev,
5599 NETLINK_CB(cb->skb).portid,
5601 RTM_NEWLINK, NLM_F_MULTI) < 0)
5615 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5617 struct sk_buff *skb;
5618 struct net *net = dev_net(idev->dev);
5621 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5625 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5627 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5628 WARN_ON(err == -EMSGSIZE);
5632 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5636 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5639 static inline size_t inet6_prefix_nlmsg_size(void)
5641 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5642 + nla_total_size(sizeof(struct in6_addr))
5643 + nla_total_size(sizeof(struct prefix_cacheinfo));
5646 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5647 struct prefix_info *pinfo, u32 portid, u32 seq,
5648 int event, unsigned int flags)
5650 struct prefixmsg *pmsg;
5651 struct nlmsghdr *nlh;
5652 struct prefix_cacheinfo ci;
5654 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5658 pmsg = nlmsg_data(nlh);
5659 pmsg->prefix_family = AF_INET6;
5660 pmsg->prefix_pad1 = 0;
5661 pmsg->prefix_pad2 = 0;
5662 pmsg->prefix_ifindex = idev->dev->ifindex;
5663 pmsg->prefix_len = pinfo->prefix_len;
5664 pmsg->prefix_type = pinfo->type;
5665 pmsg->prefix_pad3 = 0;
5666 pmsg->prefix_flags = 0;
5668 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5669 if (pinfo->autoconf)
5670 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5672 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5673 goto nla_put_failure;
5674 ci.preferred_time = ntohl(pinfo->prefered);
5675 ci.valid_time = ntohl(pinfo->valid);
5676 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5677 goto nla_put_failure;
5678 nlmsg_end(skb, nlh);
5682 nlmsg_cancel(skb, nlh);
5686 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5687 struct prefix_info *pinfo)
5689 struct sk_buff *skb;
5690 struct net *net = dev_net(idev->dev);
5693 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5697 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5699 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5700 WARN_ON(err == -EMSGSIZE);
5704 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5708 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5711 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5713 struct net *net = dev_net(ifp->idev->dev);
5718 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5723 * If the address was optimistic we inserted the route at the
5724 * start of our DAD process, so we don't need to do it again.
5725 * If the device was taken down in the middle of the DAD
5726 * cycle there is a race where we could get here without a
5727 * host route, so nothing to insert. That will be fixed when
5728 * the device is brought up.
5730 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
5731 ip6_ins_rt(net, ifp->rt);
5732 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
5733 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
5734 &ifp->addr, ifp->idev->dev->name);
5737 if (ifp->idev->cnf.forwarding)
5738 addrconf_join_anycast(ifp);
5739 if (!ipv6_addr_any(&ifp->peer_addr))
5740 addrconf_prefix_route(&ifp->peer_addr, 128,
5741 ifp->rt_priority, ifp->idev->dev,
5745 if (ifp->idev->cnf.forwarding)
5746 addrconf_leave_anycast(ifp);
5747 addrconf_leave_solict(ifp->idev, &ifp->addr);
5748 if (!ipv6_addr_any(&ifp->peer_addr)) {
5749 struct fib6_info *rt;
5751 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5752 ifp->idev->dev, 0, 0);
5754 ip6_del_rt(net, rt);
5757 ip6_del_rt(net, ifp->rt);
5760 rt_genid_bump_ipv6(net);
5763 atomic_inc(&net->ipv6.dev_addr_genid);
5766 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5769 if (likely(ifp->idev->dead == 0))
5770 __ipv6_ifa_notify(event, ifp);
5771 rcu_read_unlock_bh();
5774 #ifdef CONFIG_SYSCTL
5777 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5778 void __user *buffer, size_t *lenp, loff_t *ppos)
5780 int *valp = ctl->data;
5783 struct ctl_table lctl;
5787 * ctl->data points to idev->cnf.forwarding, we should
5788 * not modify it until we get the rtnl lock.
5793 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5796 ret = addrconf_fixup_forwarding(ctl, valp, val);
5803 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5804 void __user *buffer, size_t *lenp, loff_t *ppos)
5806 struct inet6_dev *idev = ctl->extra1;
5807 int min_mtu = IPV6_MIN_MTU;
5808 struct ctl_table lctl;
5811 lctl.extra1 = &min_mtu;
5812 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5814 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5817 static void dev_disable_change(struct inet6_dev *idev)
5819 struct netdev_notifier_info info;
5821 if (!idev || !idev->dev)
5824 netdev_notifier_info_init(&info, idev->dev);
5825 if (idev->cnf.disable_ipv6)
5826 addrconf_notify(NULL, NETDEV_DOWN, &info);
5828 addrconf_notify(NULL, NETDEV_UP, &info);
5831 static void addrconf_disable_change(struct net *net, __s32 newf)
5833 struct net_device *dev;
5834 struct inet6_dev *idev;
5836 for_each_netdev(net, dev) {
5837 idev = __in6_dev_get(dev);
5839 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5840 idev->cnf.disable_ipv6 = newf;
5842 dev_disable_change(idev);
5847 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5852 if (!rtnl_trylock())
5853 return restart_syscall();
5855 net = (struct net *)table->extra2;
5859 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5864 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5865 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5866 addrconf_disable_change(net, newf);
5867 } else if ((!newf) ^ (!old))
5868 dev_disable_change((struct inet6_dev *)table->extra1);
5875 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5876 void __user *buffer, size_t *lenp, loff_t *ppos)
5878 int *valp = ctl->data;
5881 struct ctl_table lctl;
5885 * ctl->data points to idev->cnf.disable_ipv6, we should
5886 * not modify it until we get the rtnl lock.
5891 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5894 ret = addrconf_disable_ipv6(ctl, valp, val);
5901 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5902 void __user *buffer, size_t *lenp, loff_t *ppos)
5904 int *valp = ctl->data;
5909 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5912 if (write && old != new) {
5913 struct net *net = ctl->extra2;
5915 if (!rtnl_trylock())
5916 return restart_syscall();
5918 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5919 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5920 NETCONFA_PROXY_NEIGH,
5921 NETCONFA_IFINDEX_DEFAULT,
5922 net->ipv6.devconf_dflt);
5923 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5924 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5925 NETCONFA_PROXY_NEIGH,
5926 NETCONFA_IFINDEX_ALL,
5927 net->ipv6.devconf_all);
5929 struct inet6_dev *idev = ctl->extra1;
5931 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5932 NETCONFA_PROXY_NEIGH,
5942 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5943 void __user *buffer, size_t *lenp,
5948 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5949 struct net *net = (struct net *)ctl->extra2;
5950 struct ctl_table tmp = {
5952 .maxlen = sizeof(new_val),
5956 if (!rtnl_trylock())
5957 return restart_syscall();
5959 new_val = *((u32 *)ctl->data);
5961 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
5966 if (check_addr_gen_mode(new_val) < 0) {
5972 if (check_stable_privacy(idev, net, new_val) < 0) {
5977 if (idev->cnf.addr_gen_mode != new_val) {
5978 idev->cnf.addr_gen_mode = new_val;
5979 addrconf_dev_config(idev->dev);
5981 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
5982 struct net_device *dev;
5984 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
5985 for_each_netdev(net, dev) {
5986 idev = __in6_dev_get(dev);
5988 idev->cnf.addr_gen_mode != new_val) {
5989 idev->cnf.addr_gen_mode = new_val;
5990 addrconf_dev_config(idev->dev);
5995 *((u32 *)ctl->data) = new_val;
6004 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6005 void __user *buffer, size_t *lenp,
6009 struct in6_addr addr;
6010 char str[IPV6_MAX_STRLEN];
6011 struct ctl_table lctl = *ctl;
6012 struct net *net = ctl->extra2;
6013 struct ipv6_stable_secret *secret = ctl->data;
6015 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6018 lctl.maxlen = IPV6_MAX_STRLEN;
6021 if (!rtnl_trylock())
6022 return restart_syscall();
6024 if (!write && !secret->initialized) {
6029 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6030 if (err >= sizeof(str)) {
6035 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6039 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6044 secret->initialized = true;
6045 secret->secret = addr;
6047 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6048 struct net_device *dev;
6050 for_each_netdev(net, dev) {
6051 struct inet6_dev *idev = __in6_dev_get(dev);
6054 idev->cnf.addr_gen_mode =
6055 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6059 struct inet6_dev *idev = ctl->extra1;
6061 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6071 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6073 void __user *buffer,
6077 int *valp = ctl->data;
6080 struct ctl_table lctl;
6083 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6084 * we should not modify it until we get the rtnl lock.
6089 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6092 ret = addrconf_fixup_linkdown(ctl, valp, val);
6099 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6103 rt->dst.flags |= DST_NOPOLICY;
6105 rt->dst.flags &= ~DST_NOPOLICY;
6110 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6112 struct inet6_ifaddr *ifa;
6114 read_lock_bh(&idev->lock);
6115 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6116 spin_lock(&ifa->lock);
6118 struct fib6_info *rt = ifa->rt;
6122 ifa->rt->dst_nopolicy = val ? true : false;
6123 if (rt->rt6i_pcpu) {
6124 for_each_possible_cpu(cpu) {
6125 struct rt6_info **rtp;
6127 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6128 addrconf_set_nopolicy(*rtp, val);
6133 spin_unlock(&ifa->lock);
6135 read_unlock_bh(&idev->lock);
6139 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6141 struct inet6_dev *idev;
6144 if (!rtnl_trylock())
6145 return restart_syscall();
6149 net = (struct net *)ctl->extra2;
6150 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6155 if (valp == &net->ipv6.devconf_all->disable_policy) {
6156 struct net_device *dev;
6158 for_each_netdev(net, dev) {
6159 idev = __in6_dev_get(dev);
6161 addrconf_disable_policy_idev(idev, val);
6164 idev = (struct inet6_dev *)ctl->extra1;
6165 addrconf_disable_policy_idev(idev, val);
6173 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6174 void __user *buffer, size_t *lenp,
6177 int *valp = ctl->data;
6180 struct ctl_table lctl;
6185 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6187 if (write && (*valp != val))
6188 ret = addrconf_disable_policy(ctl, valp, val);
6196 static int minus_one = -1;
6197 static const int zero = 0;
6198 static const int one = 1;
6199 static const int two_five_five = 255;
6201 static const struct ctl_table addrconf_sysctl[] = {
6203 .procname = "forwarding",
6204 .data = &ipv6_devconf.forwarding,
6205 .maxlen = sizeof(int),
6207 .proc_handler = addrconf_sysctl_forward,
6210 .procname = "hop_limit",
6211 .data = &ipv6_devconf.hop_limit,
6212 .maxlen = sizeof(int),
6214 .proc_handler = proc_dointvec_minmax,
6215 .extra1 = (void *)&one,
6216 .extra2 = (void *)&two_five_five,
6220 .data = &ipv6_devconf.mtu6,
6221 .maxlen = sizeof(int),
6223 .proc_handler = addrconf_sysctl_mtu,
6226 .procname = "accept_ra",
6227 .data = &ipv6_devconf.accept_ra,
6228 .maxlen = sizeof(int),
6230 .proc_handler = proc_dointvec,
6233 .procname = "accept_redirects",
6234 .data = &ipv6_devconf.accept_redirects,
6235 .maxlen = sizeof(int),
6237 .proc_handler = proc_dointvec,
6240 .procname = "autoconf",
6241 .data = &ipv6_devconf.autoconf,
6242 .maxlen = sizeof(int),
6244 .proc_handler = proc_dointvec,
6247 .procname = "dad_transmits",
6248 .data = &ipv6_devconf.dad_transmits,
6249 .maxlen = sizeof(int),
6251 .proc_handler = proc_dointvec,
6254 .procname = "router_solicitations",
6255 .data = &ipv6_devconf.rtr_solicits,
6256 .maxlen = sizeof(int),
6258 .proc_handler = proc_dointvec_minmax,
6259 .extra1 = &minus_one,
6262 .procname = "router_solicitation_interval",
6263 .data = &ipv6_devconf.rtr_solicit_interval,
6264 .maxlen = sizeof(int),
6266 .proc_handler = proc_dointvec_jiffies,
6269 .procname = "router_solicitation_max_interval",
6270 .data = &ipv6_devconf.rtr_solicit_max_interval,
6271 .maxlen = sizeof(int),
6273 .proc_handler = proc_dointvec_jiffies,
6276 .procname = "router_solicitation_delay",
6277 .data = &ipv6_devconf.rtr_solicit_delay,
6278 .maxlen = sizeof(int),
6280 .proc_handler = proc_dointvec_jiffies,
6283 .procname = "force_mld_version",
6284 .data = &ipv6_devconf.force_mld_version,
6285 .maxlen = sizeof(int),
6287 .proc_handler = proc_dointvec,
6290 .procname = "mldv1_unsolicited_report_interval",
6292 &ipv6_devconf.mldv1_unsolicited_report_interval,
6293 .maxlen = sizeof(int),
6295 .proc_handler = proc_dointvec_ms_jiffies,
6298 .procname = "mldv2_unsolicited_report_interval",
6300 &ipv6_devconf.mldv2_unsolicited_report_interval,
6301 .maxlen = sizeof(int),
6303 .proc_handler = proc_dointvec_ms_jiffies,
6306 .procname = "use_tempaddr",
6307 .data = &ipv6_devconf.use_tempaddr,
6308 .maxlen = sizeof(int),
6310 .proc_handler = proc_dointvec,
6313 .procname = "temp_valid_lft",
6314 .data = &ipv6_devconf.temp_valid_lft,
6315 .maxlen = sizeof(int),
6317 .proc_handler = proc_dointvec,
6320 .procname = "temp_prefered_lft",
6321 .data = &ipv6_devconf.temp_prefered_lft,
6322 .maxlen = sizeof(int),
6324 .proc_handler = proc_dointvec,
6327 .procname = "regen_max_retry",
6328 .data = &ipv6_devconf.regen_max_retry,
6329 .maxlen = sizeof(int),
6331 .proc_handler = proc_dointvec,
6334 .procname = "max_desync_factor",
6335 .data = &ipv6_devconf.max_desync_factor,
6336 .maxlen = sizeof(int),
6338 .proc_handler = proc_dointvec,
6341 .procname = "max_addresses",
6342 .data = &ipv6_devconf.max_addresses,
6343 .maxlen = sizeof(int),
6345 .proc_handler = proc_dointvec,
6348 .procname = "accept_ra_defrtr",
6349 .data = &ipv6_devconf.accept_ra_defrtr,
6350 .maxlen = sizeof(int),
6352 .proc_handler = proc_dointvec,
6355 .procname = "accept_ra_min_hop_limit",
6356 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6357 .maxlen = sizeof(int),
6359 .proc_handler = proc_dointvec,
6362 .procname = "accept_ra_pinfo",
6363 .data = &ipv6_devconf.accept_ra_pinfo,
6364 .maxlen = sizeof(int),
6366 .proc_handler = proc_dointvec,
6368 #ifdef CONFIG_IPV6_ROUTER_PREF
6370 .procname = "accept_ra_rtr_pref",
6371 .data = &ipv6_devconf.accept_ra_rtr_pref,
6372 .maxlen = sizeof(int),
6374 .proc_handler = proc_dointvec,
6377 .procname = "router_probe_interval",
6378 .data = &ipv6_devconf.rtr_probe_interval,
6379 .maxlen = sizeof(int),
6381 .proc_handler = proc_dointvec_jiffies,
6383 #ifdef CONFIG_IPV6_ROUTE_INFO
6385 .procname = "accept_ra_rt_info_min_plen",
6386 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6387 .maxlen = sizeof(int),
6389 .proc_handler = proc_dointvec,
6392 .procname = "accept_ra_rt_info_max_plen",
6393 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6394 .maxlen = sizeof(int),
6396 .proc_handler = proc_dointvec,
6401 .procname = "proxy_ndp",
6402 .data = &ipv6_devconf.proxy_ndp,
6403 .maxlen = sizeof(int),
6405 .proc_handler = addrconf_sysctl_proxy_ndp,
6408 .procname = "accept_source_route",
6409 .data = &ipv6_devconf.accept_source_route,
6410 .maxlen = sizeof(int),
6412 .proc_handler = proc_dointvec,
6414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6416 .procname = "optimistic_dad",
6417 .data = &ipv6_devconf.optimistic_dad,
6418 .maxlen = sizeof(int),
6420 .proc_handler = proc_dointvec,
6423 .procname = "use_optimistic",
6424 .data = &ipv6_devconf.use_optimistic,
6425 .maxlen = sizeof(int),
6427 .proc_handler = proc_dointvec,
6430 #ifdef CONFIG_IPV6_MROUTE
6432 .procname = "mc_forwarding",
6433 .data = &ipv6_devconf.mc_forwarding,
6434 .maxlen = sizeof(int),
6436 .proc_handler = proc_dointvec,
6440 .procname = "disable_ipv6",
6441 .data = &ipv6_devconf.disable_ipv6,
6442 .maxlen = sizeof(int),
6444 .proc_handler = addrconf_sysctl_disable,
6447 .procname = "accept_dad",
6448 .data = &ipv6_devconf.accept_dad,
6449 .maxlen = sizeof(int),
6451 .proc_handler = proc_dointvec,
6454 .procname = "force_tllao",
6455 .data = &ipv6_devconf.force_tllao,
6456 .maxlen = sizeof(int),
6458 .proc_handler = proc_dointvec
6461 .procname = "ndisc_notify",
6462 .data = &ipv6_devconf.ndisc_notify,
6463 .maxlen = sizeof(int),
6465 .proc_handler = proc_dointvec
6468 .procname = "suppress_frag_ndisc",
6469 .data = &ipv6_devconf.suppress_frag_ndisc,
6470 .maxlen = sizeof(int),
6472 .proc_handler = proc_dointvec
6475 .procname = "accept_ra_from_local",
6476 .data = &ipv6_devconf.accept_ra_from_local,
6477 .maxlen = sizeof(int),
6479 .proc_handler = proc_dointvec,
6482 .procname = "accept_ra_mtu",
6483 .data = &ipv6_devconf.accept_ra_mtu,
6484 .maxlen = sizeof(int),
6486 .proc_handler = proc_dointvec,
6489 .procname = "stable_secret",
6490 .data = &ipv6_devconf.stable_secret,
6491 .maxlen = IPV6_MAX_STRLEN,
6493 .proc_handler = addrconf_sysctl_stable_secret,
6496 .procname = "use_oif_addrs_only",
6497 .data = &ipv6_devconf.use_oif_addrs_only,
6498 .maxlen = sizeof(int),
6500 .proc_handler = proc_dointvec,
6503 .procname = "ignore_routes_with_linkdown",
6504 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6505 .maxlen = sizeof(int),
6507 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6510 .procname = "drop_unicast_in_l2_multicast",
6511 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6512 .maxlen = sizeof(int),
6514 .proc_handler = proc_dointvec,
6517 .procname = "drop_unsolicited_na",
6518 .data = &ipv6_devconf.drop_unsolicited_na,
6519 .maxlen = sizeof(int),
6521 .proc_handler = proc_dointvec,
6524 .procname = "keep_addr_on_down",
6525 .data = &ipv6_devconf.keep_addr_on_down,
6526 .maxlen = sizeof(int),
6528 .proc_handler = proc_dointvec,
6532 .procname = "seg6_enabled",
6533 .data = &ipv6_devconf.seg6_enabled,
6534 .maxlen = sizeof(int),
6536 .proc_handler = proc_dointvec,
6538 #ifdef CONFIG_IPV6_SEG6_HMAC
6540 .procname = "seg6_require_hmac",
6541 .data = &ipv6_devconf.seg6_require_hmac,
6542 .maxlen = sizeof(int),
6544 .proc_handler = proc_dointvec,
6548 .procname = "enhanced_dad",
6549 .data = &ipv6_devconf.enhanced_dad,
6550 .maxlen = sizeof(int),
6552 .proc_handler = proc_dointvec,
6555 .procname = "addr_gen_mode",
6556 .data = &ipv6_devconf.addr_gen_mode,
6557 .maxlen = sizeof(int),
6559 .proc_handler = addrconf_sysctl_addr_gen_mode,
6562 .procname = "disable_policy",
6563 .data = &ipv6_devconf.disable_policy,
6564 .maxlen = sizeof(int),
6566 .proc_handler = addrconf_sysctl_disable_policy,
6569 .procname = "ndisc_tclass",
6570 .data = &ipv6_devconf.ndisc_tclass,
6571 .maxlen = sizeof(int),
6573 .proc_handler = proc_dointvec_minmax,
6574 .extra1 = (void *)&zero,
6575 .extra2 = (void *)&two_five_five,
6582 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6583 struct inet6_dev *idev, struct ipv6_devconf *p)
6586 struct ctl_table *table;
6587 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6589 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6593 for (i = 0; table[i].data; i++) {
6594 table[i].data += (char *)p - (char *)&ipv6_devconf;
6595 /* If one of these is already set, then it is not safe to
6596 * overwrite either of them: this makes proc_dointvec_minmax
6599 if (!table[i].extra1 && !table[i].extra2) {
6600 table[i].extra1 = idev; /* embedded; no ref */
6601 table[i].extra2 = net;
6605 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6607 p->sysctl_header = register_net_sysctl(net, path, table);
6608 if (!p->sysctl_header)
6611 if (!strcmp(dev_name, "all"))
6612 ifindex = NETCONFA_IFINDEX_ALL;
6613 else if (!strcmp(dev_name, "default"))
6614 ifindex = NETCONFA_IFINDEX_DEFAULT;
6616 ifindex = idev->dev->ifindex;
6617 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6627 static void __addrconf_sysctl_unregister(struct net *net,
6628 struct ipv6_devconf *p, int ifindex)
6630 struct ctl_table *table;
6632 if (!p->sysctl_header)
6635 table = p->sysctl_header->ctl_table_arg;
6636 unregister_net_sysctl_table(p->sysctl_header);
6637 p->sysctl_header = NULL;
6640 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6643 static int addrconf_sysctl_register(struct inet6_dev *idev)
6647 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6650 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6651 &ndisc_ifinfo_sysctl_change);
6654 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6657 neigh_sysctl_unregister(idev->nd_parms);
6662 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6664 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6665 idev->dev->ifindex);
6666 neigh_sysctl_unregister(idev->nd_parms);
6672 static int __net_init addrconf_init_net(struct net *net)
6675 struct ipv6_devconf *all, *dflt;
6677 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6681 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6683 goto err_alloc_dflt;
6685 /* these will be inherited by all namespaces */
6686 dflt->autoconf = ipv6_defaults.autoconf;
6687 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6689 dflt->stable_secret.initialized = false;
6690 all->stable_secret.initialized = false;
6692 net->ipv6.devconf_all = all;
6693 net->ipv6.devconf_dflt = dflt;
6695 #ifdef CONFIG_SYSCTL
6696 err = __addrconf_sysctl_register(net, "all", NULL, all);
6700 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6706 #ifdef CONFIG_SYSCTL
6708 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6718 static void __net_exit addrconf_exit_net(struct net *net)
6720 #ifdef CONFIG_SYSCTL
6721 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6722 NETCONFA_IFINDEX_DEFAULT);
6723 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6724 NETCONFA_IFINDEX_ALL);
6726 kfree(net->ipv6.devconf_dflt);
6727 kfree(net->ipv6.devconf_all);
6730 static struct pernet_operations addrconf_ops = {
6731 .init = addrconf_init_net,
6732 .exit = addrconf_exit_net,
6735 static struct rtnl_af_ops inet6_ops __read_mostly = {
6737 .fill_link_af = inet6_fill_link_af,
6738 .get_link_af_size = inet6_get_link_af_size,
6739 .validate_link_af = inet6_validate_link_af,
6740 .set_link_af = inet6_set_link_af,
6744 * Init / cleanup code
6747 int __init addrconf_init(void)
6749 struct inet6_dev *idev;
6752 err = ipv6_addr_label_init();
6754 pr_crit("%s: cannot initialize default policy table: %d\n",
6759 err = register_pernet_subsys(&addrconf_ops);
6763 addrconf_wq = create_workqueue("ipv6_addrconf");
6769 /* The addrconf netdev notifier requires that loopback_dev
6770 * has it's ipv6 private information allocated and setup
6771 * before it can bring up and give link-local addresses
6772 * to other devices which are up.
6774 * Unfortunately, loopback_dev is not necessarily the first
6775 * entry in the global dev_base list of net devices. In fact,
6776 * it is likely to be the very last entry on that list.
6777 * So this causes the notifier registry below to try and
6778 * give link-local addresses to all devices besides loopback_dev
6779 * first, then loopback_dev, which cases all the non-loopback_dev
6780 * devices to fail to get a link-local address.
6782 * So, as a temporary fix, allocate the ipv6 structure for
6783 * loopback_dev first by hand.
6784 * Longer term, all of the dependencies ipv6 has upon the loopback
6785 * device and it being up should be removed.
6788 idev = ipv6_add_dev(init_net.loopback_dev);
6791 err = PTR_ERR(idev);
6795 ip6_route_init_special_entries();
6797 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6798 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6800 register_netdevice_notifier(&ipv6_dev_notf);
6804 rtnl_af_register(&inet6_ops);
6806 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6807 NULL, inet6_dump_ifinfo, 0);
6811 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6812 inet6_rtm_newaddr, NULL, 0);
6815 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6816 inet6_rtm_deladdr, NULL, 0);
6819 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6820 inet6_rtm_getaddr, inet6_dump_ifaddr,
6821 RTNL_FLAG_DOIT_UNLOCKED);
6824 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6825 NULL, inet6_dump_ifmcaddr, 0);
6828 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6829 NULL, inet6_dump_ifacaddr, 0);
6832 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6833 inet6_netconf_get_devconf,
6834 inet6_netconf_dump_devconf,
6835 RTNL_FLAG_DOIT_UNLOCKED);
6838 err = ipv6_addr_label_rtnl_register();
6844 rtnl_unregister_all(PF_INET6);
6845 rtnl_af_unregister(&inet6_ops);
6846 unregister_netdevice_notifier(&ipv6_dev_notf);
6848 destroy_workqueue(addrconf_wq);
6850 unregister_pernet_subsys(&addrconf_ops);
6852 ipv6_addr_label_cleanup();
6857 void addrconf_cleanup(void)
6859 struct net_device *dev;
6862 unregister_netdevice_notifier(&ipv6_dev_notf);
6863 unregister_pernet_subsys(&addrconf_ops);
6864 ipv6_addr_label_cleanup();
6866 rtnl_af_unregister(&inet6_ops);
6870 /* clean dev list */
6871 for_each_netdev(&init_net, dev) {
6872 if (__in6_dev_get(dev) == NULL)
6874 addrconf_ifdown(dev, 1);
6876 addrconf_ifdown(init_net.loopback_dev, 2);
6881 spin_lock_bh(&addrconf_hash_lock);
6882 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6883 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6884 spin_unlock_bh(&addrconf_hash_lock);
6885 cancel_delayed_work(&addr_chk_work);
6888 destroy_workqueue(addrconf_wq);