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 /* Set to 3 to get tracing... */
101 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
103 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
106 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108 #define IPV6_MAX_STRLEN \
109 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
111 static inline u32 cstamp_delta(unsigned long cstamp)
113 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
116 static inline s32 rfc3315_s14_backoff_init(s32 irt)
118 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
119 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
120 do_div(tmp, 1000000);
124 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
126 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
127 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
128 do_div(tmp, 1000000);
129 if ((s32)tmp > mrt) {
130 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
131 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
132 do_div(tmp, 1000000);
138 static int addrconf_sysctl_register(struct inet6_dev *idev);
139 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
141 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
146 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
151 static void ipv6_regen_rndid(struct inet6_dev *idev);
152 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
154 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
155 static int ipv6_count_addresses(struct inet6_dev *idev);
156 static int ipv6_generate_stable_address(struct in6_addr *addr,
158 const struct inet6_dev *idev);
161 * Configured unicast address hash table
163 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
164 static DEFINE_SPINLOCK(addrconf_hash_lock);
166 static void addrconf_verify(void);
167 static void addrconf_verify_rtnl(void);
168 static void addrconf_verify_work(struct work_struct *);
170 static struct workqueue_struct *addrconf_wq;
171 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
173 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
174 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
176 static void addrconf_type_change(struct net_device *dev,
177 unsigned long event);
178 static int addrconf_ifdown(struct net_device *dev, int how);
180 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
182 const struct net_device *dev,
183 u32 flags, u32 noflags);
185 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
186 static void addrconf_dad_work(struct work_struct *w);
187 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id);
188 static void addrconf_dad_run(struct inet6_dev *idev);
189 static void addrconf_rs_timer(unsigned long data);
190 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
191 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
193 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
194 struct prefix_info *pinfo);
195 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
196 struct net_device *dev);
198 static struct ipv6_devconf ipv6_devconf __read_mostly = {
200 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
201 .mtu6 = IPV6_MIN_MTU,
203 .accept_redirects = 1,
205 .force_mld_version = 0,
206 .mldv1_unsolicited_report_interval = 10 * HZ,
207 .mldv2_unsolicited_report_interval = HZ,
209 .rtr_solicits = MAX_RTR_SOLICITATIONS,
210 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
211 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
212 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
214 .temp_valid_lft = TEMP_VALID_LIFETIME,
215 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
216 .regen_max_retry = REGEN_MAX_RETRY,
217 .max_desync_factor = MAX_DESYNC_FACTOR,
218 .max_addresses = IPV6_MAX_ADDRESSES,
219 .accept_ra_defrtr = 1,
220 .accept_ra_from_local = 0,
221 .accept_ra_min_hop_limit= 1,
222 .accept_ra_pinfo = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 .accept_ra_rtr_pref = 1,
225 .rtr_probe_interval = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 .accept_ra_rt_info_min_plen = 0,
228 .accept_ra_rt_info_max_plen = 0,
232 .accept_source_route = 0, /* we do not accept RH0 by default. */
235 .suppress_frag_ndisc = 1,
238 .initialized = false,
240 .use_oif_addrs_only = 0,
241 .ignore_routes_with_linkdown = 0,
242 .keep_addr_on_down = 0,
244 #ifdef CONFIG_IPV6_SEG6_HMAC
245 .seg6_require_hmac = 0,
248 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
252 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
254 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
255 .mtu6 = IPV6_MIN_MTU,
257 .accept_redirects = 1,
259 .force_mld_version = 0,
260 .mldv1_unsolicited_report_interval = 10 * HZ,
261 .mldv2_unsolicited_report_interval = HZ,
263 .rtr_solicits = MAX_RTR_SOLICITATIONS,
264 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
265 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
266 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
268 .temp_valid_lft = TEMP_VALID_LIFETIME,
269 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
270 .regen_max_retry = REGEN_MAX_RETRY,
271 .max_desync_factor = MAX_DESYNC_FACTOR,
272 .max_addresses = IPV6_MAX_ADDRESSES,
273 .accept_ra_defrtr = 1,
274 .accept_ra_from_local = 0,
275 .accept_ra_min_hop_limit= 1,
276 .accept_ra_pinfo = 1,
277 #ifdef CONFIG_IPV6_ROUTER_PREF
278 .accept_ra_rtr_pref = 1,
279 .rtr_probe_interval = 60 * HZ,
280 #ifdef CONFIG_IPV6_ROUTE_INFO
281 .accept_ra_rt_info_min_plen = 0,
282 .accept_ra_rt_info_max_plen = 0,
286 .accept_source_route = 0, /* we do not accept RH0 by default. */
289 .suppress_frag_ndisc = 1,
292 .initialized = false,
294 .use_oif_addrs_only = 0,
295 .ignore_routes_with_linkdown = 0,
296 .keep_addr_on_down = 0,
298 #ifdef CONFIG_IPV6_SEG6_HMAC
299 .seg6_require_hmac = 0,
302 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
306 /* Check if a valid qdisc is available */
307 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
309 return !qdisc_tx_is_noop(dev);
312 static void addrconf_del_rs_timer(struct inet6_dev *idev)
314 if (del_timer(&idev->rs_timer))
318 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
320 if (cancel_delayed_work(&ifp->dad_work))
324 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
327 if (!timer_pending(&idev->rs_timer))
329 mod_timer(&idev->rs_timer, jiffies + when);
332 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
335 if (!delayed_work_pending(&ifp->dad_work))
337 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
340 static int snmp6_alloc_dev(struct inet6_dev *idev)
344 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
345 if (!idev->stats.ipv6)
348 for_each_possible_cpu(i) {
349 struct ipstats_mib *addrconf_stats;
350 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
351 u64_stats_init(&addrconf_stats->syncp);
355 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
357 if (!idev->stats.icmpv6dev)
359 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
361 if (!idev->stats.icmpv6msgdev)
367 kfree(idev->stats.icmpv6dev);
369 free_percpu(idev->stats.ipv6);
374 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
376 struct inet6_dev *ndev;
381 if (dev->mtu < IPV6_MIN_MTU)
382 return ERR_PTR(-EINVAL);
384 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
388 rwlock_init(&ndev->lock);
390 INIT_LIST_HEAD(&ndev->addr_list);
391 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
392 (unsigned long)ndev);
393 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
395 if (ndev->cnf.stable_secret.initialized)
396 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
398 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
400 ndev->cnf.mtu6 = dev->mtu;
401 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
402 if (!ndev->nd_parms) {
406 if (ndev->cnf.forwarding)
407 dev_disable_lro(dev);
408 /* We refer to the device */
411 if (snmp6_alloc_dev(ndev) < 0) {
413 "%s: cannot allocate memory for statistics; dev=%s.\n",
414 __func__, dev->name);
415 neigh_parms_release(&nd_tbl, ndev->nd_parms);
421 if (snmp6_register_dev(ndev) < 0) {
423 "%s: cannot create /proc/net/dev_snmp6/%s\n",
424 __func__, dev->name);
428 /* One reference from device. */
431 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
432 ndev->cnf.accept_dad = -1;
434 #if IS_ENABLED(CONFIG_IPV6_SIT)
435 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
436 pr_info("%s: Disabled Multicast RS\n", dev->name);
437 ndev->cnf.rtr_solicits = 0;
441 INIT_LIST_HEAD(&ndev->tempaddr_list);
442 ndev->desync_factor = U32_MAX;
443 if ((dev->flags&IFF_LOOPBACK) ||
444 dev->type == ARPHRD_TUNNEL ||
445 dev->type == ARPHRD_TUNNEL6 ||
446 dev->type == ARPHRD_SIT ||
447 dev->type == ARPHRD_NONE) {
448 ndev->cnf.use_tempaddr = -1;
450 ipv6_regen_rndid(ndev);
452 ndev->token = in6addr_any;
454 if (netif_running(dev) && addrconf_qdisc_ok(dev))
455 ndev->if_flags |= IF_READY;
457 ipv6_mc_init_dev(ndev);
458 ndev->tstamp = jiffies;
459 err = addrconf_sysctl_register(ndev);
461 ipv6_mc_destroy_dev(ndev);
462 snmp6_unregister_dev(ndev);
465 /* protected by rtnl_lock */
466 rcu_assign_pointer(dev->ip6_ptr, ndev);
468 /* Join interface-local all-node multicast group */
469 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
471 /* Join all-node multicast group */
472 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
474 /* Join all-router multicast group if forwarding is set */
475 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
476 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
481 neigh_parms_release(&nd_tbl, ndev->nd_parms);
483 in6_dev_finish_destroy(ndev);
487 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
489 struct inet6_dev *idev;
493 idev = __in6_dev_get(dev);
495 idev = ipv6_add_dev(dev);
500 if (dev->flags&IFF_UP)
505 static int inet6_netconf_msgsize_devconf(int type)
507 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
508 + nla_total_size(4); /* NETCONFA_IFINDEX */
511 if (type == NETCONFA_ALL)
514 if (all || type == NETCONFA_FORWARDING)
515 size += nla_total_size(4);
516 #ifdef CONFIG_IPV6_MROUTE
517 if (all || type == NETCONFA_MC_FORWARDING)
518 size += nla_total_size(4);
520 if (all || type == NETCONFA_PROXY_NEIGH)
521 size += nla_total_size(4);
523 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
524 size += nla_total_size(4);
529 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
530 struct ipv6_devconf *devconf, u32 portid,
531 u32 seq, int event, unsigned int flags,
534 struct nlmsghdr *nlh;
535 struct netconfmsg *ncm;
538 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
543 if (type == NETCONFA_ALL)
546 ncm = nlmsg_data(nlh);
547 ncm->ncm_family = AF_INET6;
549 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
550 goto nla_put_failure;
555 if ((all || type == NETCONFA_FORWARDING) &&
556 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
557 goto nla_put_failure;
558 #ifdef CONFIG_IPV6_MROUTE
559 if ((all || type == NETCONFA_MC_FORWARDING) &&
560 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
561 devconf->mc_forwarding) < 0)
562 goto nla_put_failure;
564 if ((all || type == NETCONFA_PROXY_NEIGH) &&
565 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
566 goto nla_put_failure;
568 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
569 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
570 devconf->ignore_routes_with_linkdown) < 0)
571 goto nla_put_failure;
578 nlmsg_cancel(skb, nlh);
582 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
583 int ifindex, struct ipv6_devconf *devconf)
588 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
592 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
595 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
596 WARN_ON(err == -EMSGSIZE);
600 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
603 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
606 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
607 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
608 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
609 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
610 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
613 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
614 struct nlmsghdr *nlh,
615 struct netlink_ext_ack *extack)
617 struct net *net = sock_net(in_skb->sk);
618 struct nlattr *tb[NETCONFA_MAX+1];
619 struct netconfmsg *ncm;
621 struct ipv6_devconf *devconf;
622 struct inet6_dev *in6_dev;
623 struct net_device *dev;
627 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
628 devconf_ipv6_policy, extack);
633 if (!tb[NETCONFA_IFINDEX])
636 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
638 case NETCONFA_IFINDEX_ALL:
639 devconf = net->ipv6.devconf_all;
641 case NETCONFA_IFINDEX_DEFAULT:
642 devconf = net->ipv6.devconf_dflt;
645 dev = __dev_get_by_index(net, ifindex);
648 in6_dev = __in6_dev_get(dev);
651 devconf = &in6_dev->cnf;
656 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
660 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
661 NETLINK_CB(in_skb).portid,
662 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
665 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
666 WARN_ON(err == -EMSGSIZE);
670 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
675 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
676 struct netlink_callback *cb)
678 struct net *net = sock_net(skb->sk);
681 struct net_device *dev;
682 struct inet6_dev *idev;
683 struct hlist_head *head;
686 s_idx = idx = cb->args[1];
688 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
690 head = &net->dev_index_head[h];
692 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
694 hlist_for_each_entry_rcu(dev, head, index_hlist) {
697 idev = __in6_dev_get(dev);
701 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
703 NETLINK_CB(cb->skb).portid,
711 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
717 if (h == NETDEV_HASHENTRIES) {
718 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
719 net->ipv6.devconf_all,
720 NETLINK_CB(cb->skb).portid,
722 RTM_NEWNETCONF, NLM_F_MULTI,
728 if (h == NETDEV_HASHENTRIES + 1) {
729 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
730 net->ipv6.devconf_dflt,
731 NETLINK_CB(cb->skb).portid,
733 RTM_NEWNETCONF, NLM_F_MULTI,
747 static void dev_forward_change(struct inet6_dev *idev)
749 struct net_device *dev;
750 struct inet6_ifaddr *ifa;
755 if (idev->cnf.forwarding)
756 dev_disable_lro(dev);
757 if (dev->flags & IFF_MULTICAST) {
758 if (idev->cnf.forwarding) {
759 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
760 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
761 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
763 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
764 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
765 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
769 list_for_each_entry(ifa, &idev->addr_list, if_list) {
770 if (ifa->flags&IFA_F_TENTATIVE)
772 if (idev->cnf.forwarding)
773 addrconf_join_anycast(ifa);
775 addrconf_leave_anycast(ifa);
777 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
779 dev->ifindex, &idev->cnf);
783 static void addrconf_forward_change(struct net *net, __s32 newf)
785 struct net_device *dev;
786 struct inet6_dev *idev;
788 for_each_netdev(net, dev) {
789 idev = __in6_dev_get(dev);
791 int changed = (!idev->cnf.forwarding) ^ (!newf);
792 idev->cnf.forwarding = newf;
794 dev_forward_change(idev);
799 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
805 return restart_syscall();
807 net = (struct net *)table->extra2;
811 if (p == &net->ipv6.devconf_dflt->forwarding) {
812 if ((!newf) ^ (!old))
813 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
815 NETCONFA_IFINDEX_DEFAULT,
816 net->ipv6.devconf_dflt);
821 if (p == &net->ipv6.devconf_all->forwarding) {
822 int old_dflt = net->ipv6.devconf_dflt->forwarding;
824 net->ipv6.devconf_dflt->forwarding = newf;
825 if ((!newf) ^ (!old_dflt))
826 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
828 NETCONFA_IFINDEX_DEFAULT,
829 net->ipv6.devconf_dflt);
831 addrconf_forward_change(net, newf);
832 if ((!newf) ^ (!old))
833 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
835 NETCONFA_IFINDEX_ALL,
836 net->ipv6.devconf_all);
837 } else if ((!newf) ^ (!old))
838 dev_forward_change((struct inet6_dev *)table->extra1);
842 rt6_purge_dflt_routers(net);
846 static void addrconf_linkdown_change(struct net *net, __s32 newf)
848 struct net_device *dev;
849 struct inet6_dev *idev;
851 for_each_netdev(net, dev) {
852 idev = __in6_dev_get(dev);
854 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
856 idev->cnf.ignore_routes_with_linkdown = newf;
858 inet6_netconf_notify_devconf(dev_net(dev),
860 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
867 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
873 return restart_syscall();
875 net = (struct net *)table->extra2;
879 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
880 if ((!newf) ^ (!old))
881 inet6_netconf_notify_devconf(net,
883 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
884 NETCONFA_IFINDEX_DEFAULT,
885 net->ipv6.devconf_dflt);
890 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
891 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
892 addrconf_linkdown_change(net, newf);
893 if ((!newf) ^ (!old))
894 inet6_netconf_notify_devconf(net,
896 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
897 NETCONFA_IFINDEX_ALL,
898 net->ipv6.devconf_all);
907 /* Nobody refers to this ifaddr, destroy it */
908 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
910 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
912 #ifdef NET_REFCNT_DEBUG
913 pr_debug("%s\n", __func__);
916 in6_dev_put(ifp->idev);
918 if (cancel_delayed_work(&ifp->dad_work))
919 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
922 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
923 pr_warn("Freeing alive inet6 address %p\n", ifp);
932 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
935 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
938 * Each device address list is sorted in order of scope -
939 * global before linklocal.
941 list_for_each(p, &idev->addr_list) {
942 struct inet6_ifaddr *ifa
943 = list_entry(p, struct inet6_ifaddr, if_list);
944 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
948 list_add_tail(&ifp->if_list, p);
951 static u32 inet6_addr_hash(const struct in6_addr *addr)
953 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
956 /* On success it returns ifp with increased reference count */
958 static struct inet6_ifaddr *
959 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
960 const struct in6_addr *peer_addr, int pfxlen,
961 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
963 struct net *net = dev_net(idev->dev);
964 struct inet6_ifaddr *ifa = NULL;
968 int addr_type = ipv6_addr_type(addr);
970 if (addr_type == IPV6_ADDR_ANY ||
971 addr_type & IPV6_ADDR_MULTICAST ||
972 (!(idev->dev->flags & IFF_LOOPBACK) &&
973 addr_type & IPV6_ADDR_LOOPBACK))
974 return ERR_PTR(-EADDRNOTAVAIL);
978 err = -ENODEV; /*XXX*/
982 if (idev->cnf.disable_ipv6) {
987 spin_lock(&addrconf_hash_lock);
989 /* Ignore adding duplicate addresses on an interface */
990 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
991 ADBG("ipv6_add_addr: already assigned\n");
996 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
999 ADBG("ipv6_add_addr: malloc failed\n");
1004 rt = addrconf_dst_alloc(idev, addr, false);
1010 if (net->ipv6.devconf_all->disable_policy ||
1011 idev->cnf.disable_policy)
1012 rt->dst.flags |= DST_NOPOLICY;
1014 neigh_parms_data_state_setall(idev->nd_parms);
1018 ifa->peer_addr = *peer_addr;
1020 spin_lock_init(&ifa->lock);
1021 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1022 INIT_HLIST_NODE(&ifa->addr_lst);
1024 ifa->prefix_len = pfxlen;
1025 ifa->flags = flags | IFA_F_TENTATIVE;
1026 ifa->valid_lft = valid_lft;
1027 ifa->prefered_lft = prefered_lft;
1028 ifa->cstamp = ifa->tstamp = jiffies;
1029 ifa->tokenized = false;
1038 /* Add to big hash table */
1039 hash = inet6_addr_hash(addr);
1041 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1042 spin_unlock(&addrconf_hash_lock);
1044 write_lock(&idev->lock);
1045 /* Add to inet6_dev unicast addr list. */
1046 ipv6_link_dev_addr(idev, ifa);
1048 if (ifa->flags&IFA_F_TEMPORARY) {
1049 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1054 write_unlock(&idev->lock);
1056 rcu_read_unlock_bh();
1058 if (likely(err == 0))
1059 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1067 spin_unlock(&addrconf_hash_lock);
1071 enum cleanup_prefix_rt_t {
1072 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1073 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1074 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1078 * Check, whether the prefix for ifp would still need a prefix route
1079 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1082 * 1) we don't purge prefix if address was not permanent.
1083 * prefix is managed by its own lifetime.
1084 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1085 * 3) if there are no addresses, delete prefix.
1086 * 4) if there are still other permanent address(es),
1087 * corresponding prefix is still permanent.
1088 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1089 * don't purge the prefix, assume user space is managing it.
1090 * 6) otherwise, update prefix lifetime to the
1091 * longest valid lifetime among the corresponding
1092 * addresses on the device.
1093 * Note: subsequent RA will update lifetime.
1095 static enum cleanup_prefix_rt_t
1096 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1098 struct inet6_ifaddr *ifa;
1099 struct inet6_dev *idev = ifp->idev;
1100 unsigned long lifetime;
1101 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1105 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1108 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1111 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1112 return CLEANUP_PREFIX_RT_NOP;
1114 action = CLEANUP_PREFIX_RT_EXPIRE;
1116 spin_lock(&ifa->lock);
1118 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1120 * Note: Because this address is
1121 * not permanent, lifetime <
1122 * LONG_MAX / HZ here.
1124 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1125 *expires = ifa->tstamp + lifetime * HZ;
1126 spin_unlock(&ifa->lock);
1133 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1135 struct rt6_info *rt;
1137 rt = addrconf_get_prefix_route(&ifp->addr,
1140 0, RTF_GATEWAY | RTF_DEFAULT);
1145 if (!(rt->rt6i_flags & RTF_EXPIRES))
1146 rt6_set_expires(rt, expires);
1153 /* This function wants to get referenced ifp and releases it before return */
1155 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1158 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1159 unsigned long expires;
1163 spin_lock_bh(&ifp->lock);
1165 ifp->state = INET6_IFADDR_STATE_DEAD;
1166 spin_unlock_bh(&ifp->lock);
1168 if (state == INET6_IFADDR_STATE_DEAD)
1171 spin_lock_bh(&addrconf_hash_lock);
1172 hlist_del_init_rcu(&ifp->addr_lst);
1173 spin_unlock_bh(&addrconf_hash_lock);
1175 write_lock_bh(&ifp->idev->lock);
1177 if (ifp->flags&IFA_F_TEMPORARY) {
1178 list_del(&ifp->tmp_list);
1180 in6_ifa_put(ifp->ifpub);
1186 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1187 action = check_cleanup_prefix_route(ifp, &expires);
1189 list_del_init(&ifp->if_list);
1192 write_unlock_bh(&ifp->idev->lock);
1194 addrconf_del_dad_work(ifp);
1196 ipv6_ifa_notify(RTM_DELADDR, ifp);
1198 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1200 if (action != CLEANUP_PREFIX_RT_NOP) {
1201 cleanup_prefix_route(ifp, expires,
1202 action == CLEANUP_PREFIX_RT_DEL);
1205 /* clean up prefsrc entries */
1206 rt6_remove_prefsrc(ifp);
1211 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1213 struct inet6_dev *idev = ifp->idev;
1214 struct in6_addr addr, *tmpaddr;
1215 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1216 unsigned long regen_advance;
1220 unsigned long now = jiffies;
1221 long max_desync_factor;
1222 s32 cnf_temp_preferred_lft;
1224 write_lock_bh(&idev->lock);
1226 spin_lock_bh(&ift->lock);
1227 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1228 spin_unlock_bh(&ift->lock);
1235 if (idev->cnf.use_tempaddr <= 0) {
1236 write_unlock_bh(&idev->lock);
1237 pr_info("%s: use_tempaddr is disabled\n", __func__);
1242 spin_lock_bh(&ifp->lock);
1243 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1244 idev->cnf.use_tempaddr = -1; /*XXX*/
1245 spin_unlock_bh(&ifp->lock);
1246 write_unlock_bh(&idev->lock);
1247 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1254 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1255 ipv6_try_regen_rndid(idev, tmpaddr);
1256 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1257 age = (now - ifp->tstamp) / HZ;
1259 regen_advance = idev->cnf.regen_max_retry *
1260 idev->cnf.dad_transmits *
1261 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1263 /* recalculate max_desync_factor each time and update
1264 * idev->desync_factor if it's larger
1266 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1267 max_desync_factor = min_t(__u32,
1268 idev->cnf.max_desync_factor,
1269 cnf_temp_preferred_lft - regen_advance);
1271 if (unlikely(idev->desync_factor > max_desync_factor)) {
1272 if (max_desync_factor > 0) {
1273 get_random_bytes(&idev->desync_factor,
1274 sizeof(idev->desync_factor));
1275 idev->desync_factor %= max_desync_factor;
1277 idev->desync_factor = 0;
1281 tmp_valid_lft = min_t(__u32,
1283 idev->cnf.temp_valid_lft + age);
1284 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1285 idev->desync_factor;
1286 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1287 tmp_plen = ifp->prefix_len;
1288 tmp_tstamp = ifp->tstamp;
1289 spin_unlock_bh(&ifp->lock);
1291 write_unlock_bh(&idev->lock);
1293 /* A temporary address is created only if this calculated Preferred
1294 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1295 * an implementation must not create a temporary address with a zero
1296 * Preferred Lifetime.
1297 * Use age calculation as in addrconf_verify to avoid unnecessary
1298 * temporary addresses being generated.
1300 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1301 if (tmp_prefered_lft <= regen_advance + age) {
1308 addr_flags = IFA_F_TEMPORARY;
1309 /* set in addrconf_prefix_rcv() */
1310 if (ifp->flags & IFA_F_OPTIMISTIC)
1311 addr_flags |= IFA_F_OPTIMISTIC;
1313 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1314 ipv6_addr_scope(&addr), addr_flags,
1315 tmp_valid_lft, tmp_prefered_lft);
1319 pr_info("%s: retry temporary address regeneration\n", __func__);
1321 write_lock_bh(&idev->lock);
1325 spin_lock_bh(&ift->lock);
1328 ift->tstamp = tmp_tstamp;
1329 spin_unlock_bh(&ift->lock);
1331 addrconf_dad_start(ift);
1339 * Choose an appropriate source address (RFC3484)
1342 IPV6_SADDR_RULE_INIT = 0,
1343 IPV6_SADDR_RULE_LOCAL,
1344 IPV6_SADDR_RULE_SCOPE,
1345 IPV6_SADDR_RULE_PREFERRED,
1346 #ifdef CONFIG_IPV6_MIP6
1347 IPV6_SADDR_RULE_HOA,
1349 IPV6_SADDR_RULE_OIF,
1350 IPV6_SADDR_RULE_LABEL,
1351 IPV6_SADDR_RULE_PRIVACY,
1352 IPV6_SADDR_RULE_ORCHID,
1353 IPV6_SADDR_RULE_PREFIX,
1354 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1355 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1360 struct ipv6_saddr_score {
1363 struct inet6_ifaddr *ifa;
1364 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1369 struct ipv6_saddr_dst {
1370 const struct in6_addr *addr;
1377 static inline int ipv6_saddr_preferred(int type)
1379 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1384 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1386 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1387 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1393 static int ipv6_get_saddr_eval(struct net *net,
1394 struct ipv6_saddr_score *score,
1395 struct ipv6_saddr_dst *dst,
1400 if (i <= score->rule) {
1402 case IPV6_SADDR_RULE_SCOPE:
1403 ret = score->scopedist;
1405 case IPV6_SADDR_RULE_PREFIX:
1406 ret = score->matchlen;
1409 ret = !!test_bit(i, score->scorebits);
1415 case IPV6_SADDR_RULE_INIT:
1416 /* Rule 0: remember if hiscore is not ready yet */
1419 case IPV6_SADDR_RULE_LOCAL:
1420 /* Rule 1: Prefer same address */
1421 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1423 case IPV6_SADDR_RULE_SCOPE:
1424 /* Rule 2: Prefer appropriate scope
1429 * ---+--+-+---> scope
1431 * | d is scope of the destination.
1433 * | \ <- smaller scope is better if
1434 * B-15 | \ if scope is enough for destination.
1435 * | ret = B - scope (-1 <= scope >= d <= 15).
1437 * |/ <- greater is better
1438 * -C / if scope is not enough for destination.
1439 * /| ret = scope - C (-1 <= d < scope <= 15).
1441 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1442 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1443 * Assume B = 0 and we get C > 29.
1445 ret = __ipv6_addr_src_scope(score->addr_type);
1446 if (ret >= dst->scope)
1449 ret -= 128; /* 30 is enough */
1450 score->scopedist = ret;
1452 case IPV6_SADDR_RULE_PREFERRED:
1454 /* Rule 3: Avoid deprecated and optimistic addresses */
1455 u8 avoid = IFA_F_DEPRECATED;
1457 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1458 avoid |= IFA_F_OPTIMISTIC;
1459 ret = ipv6_saddr_preferred(score->addr_type) ||
1460 !(score->ifa->flags & avoid);
1463 #ifdef CONFIG_IPV6_MIP6
1464 case IPV6_SADDR_RULE_HOA:
1466 /* Rule 4: Prefer home address */
1467 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1468 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1472 case IPV6_SADDR_RULE_OIF:
1473 /* Rule 5: Prefer outgoing interface */
1474 ret = (!dst->ifindex ||
1475 dst->ifindex == score->ifa->idev->dev->ifindex);
1477 case IPV6_SADDR_RULE_LABEL:
1478 /* Rule 6: Prefer matching label */
1479 ret = ipv6_addr_label(net,
1480 &score->ifa->addr, score->addr_type,
1481 score->ifa->idev->dev->ifindex) == dst->label;
1483 case IPV6_SADDR_RULE_PRIVACY:
1485 /* Rule 7: Prefer public address
1486 * Note: prefer temporary address if use_tempaddr >= 2
1488 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1489 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1490 score->ifa->idev->cnf.use_tempaddr >= 2;
1491 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1494 case IPV6_SADDR_RULE_ORCHID:
1495 /* Rule 8-: Prefer ORCHID vs ORCHID or
1496 * non-ORCHID vs non-ORCHID
1498 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1499 ipv6_addr_orchid(dst->addr));
1501 case IPV6_SADDR_RULE_PREFIX:
1502 /* Rule 8: Use longest matching prefix */
1503 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1504 if (ret > score->ifa->prefix_len)
1505 ret = score->ifa->prefix_len;
1506 score->matchlen = ret;
1508 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1509 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1510 /* Optimistic addresses still have lower precedence than other
1511 * preferred addresses.
1513 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1521 __set_bit(i, score->scorebits);
1527 static int __ipv6_dev_get_saddr(struct net *net,
1528 struct ipv6_saddr_dst *dst,
1529 struct inet6_dev *idev,
1530 struct ipv6_saddr_score *scores,
1533 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1535 read_lock_bh(&idev->lock);
1536 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1540 * - Tentative Address (RFC2462 section 5.4)
1541 * - A tentative address is not considered
1542 * "assigned to an interface" in the traditional
1543 * sense, unless it is also flagged as optimistic.
1544 * - Candidate Source Address (section 4)
1545 * - In any case, anycast addresses, multicast
1546 * addresses, and the unspecified address MUST
1547 * NOT be included in a candidate set.
1549 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1550 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1553 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1555 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1556 score->addr_type & IPV6_ADDR_MULTICAST)) {
1557 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1563 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1565 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1566 int minihiscore, miniscore;
1568 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1569 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1571 if (minihiscore > miniscore) {
1572 if (i == IPV6_SADDR_RULE_SCOPE &&
1573 score->scopedist > 0) {
1576 * each remaining entry
1577 * has too small (not enough)
1578 * scope, because ifa entries
1579 * are sorted by their scope
1585 } else if (minihiscore < miniscore) {
1587 in6_ifa_put(hiscore->ifa);
1589 in6_ifa_hold(score->ifa);
1591 swap(hiscore, score);
1592 hiscore_idx = 1 - hiscore_idx;
1594 /* restore our iterator */
1595 score->ifa = hiscore->ifa;
1602 read_unlock_bh(&idev->lock);
1606 static int ipv6_get_saddr_master(struct net *net,
1607 const struct net_device *dst_dev,
1608 const struct net_device *master,
1609 struct ipv6_saddr_dst *dst,
1610 struct ipv6_saddr_score *scores,
1613 struct inet6_dev *idev;
1615 idev = __in6_dev_get(dst_dev);
1617 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1618 scores, hiscore_idx);
1620 idev = __in6_dev_get(master);
1622 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1623 scores, hiscore_idx);
1628 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1629 const struct in6_addr *daddr, unsigned int prefs,
1630 struct in6_addr *saddr)
1632 struct ipv6_saddr_score scores[2], *hiscore;
1633 struct ipv6_saddr_dst dst;
1634 struct inet6_dev *idev;
1635 struct net_device *dev;
1637 bool use_oif_addr = false;
1638 int hiscore_idx = 0;
1640 dst_type = __ipv6_addr_type(daddr);
1642 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1643 dst.scope = __ipv6_addr_src_scope(dst_type);
1644 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1647 scores[hiscore_idx].rule = -1;
1648 scores[hiscore_idx].ifa = NULL;
1652 /* Candidate Source Address (section 4)
1653 * - multicast and link-local destination address,
1654 * the set of candidate source address MUST only
1655 * include addresses assigned to interfaces
1656 * belonging to the same link as the outgoing
1658 * (- For site-local destination addresses, the
1659 * set of candidate source addresses MUST only
1660 * include addresses assigned to interfaces
1661 * belonging to the same site as the outgoing
1663 * - "It is RECOMMENDED that the candidate source addresses
1664 * be the set of unicast addresses assigned to the
1665 * interface that will be used to send to the destination
1666 * (the 'outgoing' interface)." (RFC 6724)
1669 idev = __in6_dev_get(dst_dev);
1670 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1671 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1672 (idev && idev->cnf.use_oif_addrs_only)) {
1673 use_oif_addr = true;
1679 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1681 const struct net_device *master;
1684 /* if dst_dev exists and is enslaved to an L3 device, then
1685 * prefer addresses from dst_dev and then the master over
1686 * any other enslaved devices in the L3 domain.
1688 master = l3mdev_master_dev_rcu(dst_dev);
1690 master_idx = master->ifindex;
1692 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1694 scores, hiscore_idx);
1696 if (scores[hiscore_idx].ifa)
1700 for_each_netdev_rcu(net, dev) {
1701 /* only consider addresses on devices in the
1704 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1706 idev = __in6_dev_get(dev);
1709 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1716 hiscore = &scores[hiscore_idx];
1718 return -EADDRNOTAVAIL;
1720 *saddr = hiscore->ifa->addr;
1721 in6_ifa_put(hiscore->ifa);
1724 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1726 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1729 struct inet6_ifaddr *ifp;
1730 int err = -EADDRNOTAVAIL;
1732 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1733 if (ifp->scope > IFA_LINK)
1735 if (ifp->scope == IFA_LINK &&
1736 !(ifp->flags & banned_flags)) {
1745 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1748 struct inet6_dev *idev;
1749 int err = -EADDRNOTAVAIL;
1752 idev = __in6_dev_get(dev);
1754 read_lock_bh(&idev->lock);
1755 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1756 read_unlock_bh(&idev->lock);
1762 static int ipv6_count_addresses(struct inet6_dev *idev)
1765 struct inet6_ifaddr *ifp;
1767 read_lock_bh(&idev->lock);
1768 list_for_each_entry(ifp, &idev->addr_list, if_list)
1770 read_unlock_bh(&idev->lock);
1774 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1775 const struct net_device *dev, int strict)
1777 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1779 EXPORT_SYMBOL(ipv6_chk_addr);
1781 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1782 const struct net_device *dev, int strict,
1785 struct inet6_ifaddr *ifp;
1786 unsigned int hash = inet6_addr_hash(addr);
1790 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1791 if (!net_eq(dev_net(ifp->idev->dev), net))
1793 /* Decouple optimistic from tentative for evaluation here.
1794 * Ban optimistic addresses explicitly, when required.
1796 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1797 ? (ifp->flags&~IFA_F_TENTATIVE)
1799 if (ipv6_addr_equal(&ifp->addr, addr) &&
1800 !(ifp_flags&banned_flags) &&
1801 (!dev || ifp->idev->dev == dev ||
1802 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1803 rcu_read_unlock_bh();
1808 rcu_read_unlock_bh();
1811 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1813 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1814 struct net_device *dev)
1816 unsigned int hash = inet6_addr_hash(addr);
1817 struct inet6_ifaddr *ifp;
1819 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1820 if (!net_eq(dev_net(ifp->idev->dev), net))
1822 if (ipv6_addr_equal(&ifp->addr, addr)) {
1823 if (!dev || ifp->idev->dev == dev)
1830 /* Compares an address/prefix_len with addresses on device @dev.
1831 * If one is found it returns true.
1833 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1834 const unsigned int prefix_len, struct net_device *dev)
1836 struct inet6_dev *idev;
1837 struct inet6_ifaddr *ifa;
1841 idev = __in6_dev_get(dev);
1843 read_lock_bh(&idev->lock);
1844 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1845 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1849 read_unlock_bh(&idev->lock);
1855 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1857 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1859 struct inet6_dev *idev;
1860 struct inet6_ifaddr *ifa;
1865 idev = __in6_dev_get(dev);
1867 read_lock_bh(&idev->lock);
1868 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1869 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1874 read_unlock_bh(&idev->lock);
1879 EXPORT_SYMBOL(ipv6_chk_prefix);
1881 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1882 struct net_device *dev, int strict)
1884 struct inet6_ifaddr *ifp, *result = NULL;
1885 unsigned int hash = inet6_addr_hash(addr);
1888 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1889 if (!net_eq(dev_net(ifp->idev->dev), net))
1891 if (ipv6_addr_equal(&ifp->addr, addr)) {
1892 if (!dev || ifp->idev->dev == dev ||
1893 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1900 rcu_read_unlock_bh();
1905 /* Gets referenced address, destroys ifaddr */
1907 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1910 ifp->flags |= IFA_F_DADFAILED;
1912 if (ifp->flags&IFA_F_PERMANENT) {
1913 spin_lock_bh(&ifp->lock);
1914 addrconf_del_dad_work(ifp);
1915 ifp->flags |= IFA_F_TENTATIVE;
1916 spin_unlock_bh(&ifp->lock);
1918 ipv6_ifa_notify(0, ifp);
1920 } else if (ifp->flags&IFA_F_TEMPORARY) {
1921 struct inet6_ifaddr *ifpub;
1922 spin_lock_bh(&ifp->lock);
1925 in6_ifa_hold(ifpub);
1926 spin_unlock_bh(&ifp->lock);
1927 ipv6_create_tempaddr(ifpub, ifp);
1930 spin_unlock_bh(&ifp->lock);
1938 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1942 spin_lock_bh(&ifp->lock);
1943 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1944 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1947 spin_unlock_bh(&ifp->lock);
1952 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1954 struct inet6_dev *idev = ifp->idev;
1955 struct net *net = dev_net(ifp->idev->dev);
1957 if (addrconf_dad_end(ifp)) {
1962 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1963 ifp->idev->dev->name, &ifp->addr);
1965 spin_lock_bh(&ifp->lock);
1967 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1968 int scope = ifp->scope;
1969 u32 flags = ifp->flags;
1970 struct in6_addr new_addr;
1971 struct inet6_ifaddr *ifp2;
1972 u32 valid_lft, preferred_lft;
1973 int pfxlen = ifp->prefix_len;
1974 int retries = ifp->stable_privacy_retry + 1;
1976 if (retries > net->ipv6.sysctl.idgen_retries) {
1977 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1978 ifp->idev->dev->name);
1982 new_addr = ifp->addr;
1983 if (ipv6_generate_stable_address(&new_addr, retries,
1987 valid_lft = ifp->valid_lft;
1988 preferred_lft = ifp->prefered_lft;
1990 spin_unlock_bh(&ifp->lock);
1992 if (idev->cnf.max_addresses &&
1993 ipv6_count_addresses(idev) >=
1994 idev->cnf.max_addresses)
1997 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1998 ifp->idev->dev->name);
2000 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2001 scope, flags, valid_lft,
2006 spin_lock_bh(&ifp2->lock);
2007 ifp2->stable_privacy_retry = retries;
2008 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2009 spin_unlock_bh(&ifp2->lock);
2011 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2014 spin_lock_bh(&ifp->lock);
2018 /* transition from _POSTDAD to _ERRDAD */
2019 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2020 spin_unlock_bh(&ifp->lock);
2022 addrconf_mod_dad_work(ifp, 0);
2026 /* Join to solicited addr multicast group.
2027 * caller must hold RTNL */
2028 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2030 struct in6_addr maddr;
2032 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2035 addrconf_addr_solict_mult(addr, &maddr);
2036 ipv6_dev_mc_inc(dev, &maddr);
2039 /* caller must hold RTNL */
2040 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2042 struct in6_addr maddr;
2044 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2047 addrconf_addr_solict_mult(addr, &maddr);
2048 __ipv6_dev_mc_dec(idev, &maddr);
2051 /* caller must hold RTNL */
2052 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2054 struct in6_addr addr;
2056 if (ifp->prefix_len >= 127) /* RFC 6164 */
2058 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2059 if (ipv6_addr_any(&addr))
2061 __ipv6_dev_ac_inc(ifp->idev, &addr);
2064 /* caller must hold RTNL */
2065 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2067 struct in6_addr addr;
2069 if (ifp->prefix_len >= 127) /* RFC 6164 */
2071 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2072 if (ipv6_addr_any(&addr))
2074 __ipv6_dev_ac_dec(ifp->idev, &addr);
2077 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2079 switch (dev->addr_len) {
2081 memcpy(eui, dev->dev_addr, 3);
2084 memcpy(eui + 5, dev->dev_addr + 3, 3);
2086 case EUI64_ADDR_LEN:
2087 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2097 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2099 union fwnet_hwaddr *ha;
2101 if (dev->addr_len != FWNET_ALEN)
2104 ha = (union fwnet_hwaddr *)dev->dev_addr;
2106 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2111 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2113 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2114 if (dev->addr_len != ARCNET_ALEN)
2117 eui[7] = *(u8 *)dev->dev_addr;
2121 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2123 if (dev->addr_len != INFINIBAND_ALEN)
2125 memcpy(eui, dev->dev_addr + 12, 8);
2130 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2134 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2135 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2136 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2137 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2138 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2139 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2143 memcpy(eui + 4, &addr, 4);
2147 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2149 if (dev->priv_flags & IFF_ISATAP)
2150 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2154 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2156 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2159 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2161 memcpy(eui, dev->perm_addr, 3);
2162 memcpy(eui + 5, dev->perm_addr + 3, 3);
2169 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2171 switch (dev->type) {
2174 return addrconf_ifid_eui48(eui, dev);
2176 return addrconf_ifid_arcnet(eui, dev);
2177 case ARPHRD_INFINIBAND:
2178 return addrconf_ifid_infiniband(eui, dev);
2180 return addrconf_ifid_sit(eui, dev);
2183 return addrconf_ifid_gre(eui, dev);
2184 case ARPHRD_6LOWPAN:
2185 return addrconf_ifid_6lowpan(eui, dev);
2186 case ARPHRD_IEEE1394:
2187 return addrconf_ifid_ieee1394(eui, dev);
2188 case ARPHRD_TUNNEL6:
2190 return addrconf_ifid_ip6tnl(eui, dev);
2195 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2198 struct inet6_ifaddr *ifp;
2200 read_lock_bh(&idev->lock);
2201 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2202 if (ifp->scope > IFA_LINK)
2204 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2205 memcpy(eui, ifp->addr.s6_addr+8, 8);
2210 read_unlock_bh(&idev->lock);
2214 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2215 static void ipv6_regen_rndid(struct inet6_dev *idev)
2218 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2219 idev->rndid[0] &= ~0x02;
2222 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2223 * check if generated address is not inappropriate
2225 * - Reserved subnet anycast (RFC 2526)
2226 * 11111101 11....11 1xxxxxxx
2227 * - ISATAP (RFC4214) 6.1
2228 * 00-00-5E-FE-xx-xx-xx-xx
2230 * - XXX: already assigned to an address on the device
2232 if (idev->rndid[0] == 0xfd &&
2233 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2234 (idev->rndid[7]&0x80))
2236 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2237 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2239 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2244 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2246 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2247 ipv6_regen_rndid(idev);
2255 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2256 unsigned long expires, u32 flags)
2258 struct fib6_config cfg = {
2259 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2260 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2261 .fc_ifindex = dev->ifindex,
2262 .fc_expires = expires,
2264 .fc_flags = RTF_UP | flags,
2265 .fc_nlinfo.nl_net = dev_net(dev),
2266 .fc_protocol = RTPROT_KERNEL,
2271 /* Prevent useless cloning on PtP SIT.
2272 This thing is done here expecting that the whole
2273 class of non-broadcast devices need not cloning.
2275 #if IS_ENABLED(CONFIG_IPV6_SIT)
2276 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2277 cfg.fc_flags |= RTF_NONEXTHOP;
2280 ip6_route_add(&cfg);
2284 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2286 const struct net_device *dev,
2287 u32 flags, u32 noflags)
2289 struct fib6_node *fn;
2290 struct rt6_info *rt = NULL;
2291 struct fib6_table *table;
2292 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2294 table = fib6_get_table(dev_net(dev), tb_id);
2298 read_lock_bh(&table->tb6_lock);
2299 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2303 noflags |= RTF_CACHE;
2304 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2305 if (rt->dst.dev->ifindex != dev->ifindex)
2307 if ((rt->rt6i_flags & flags) != flags)
2309 if ((rt->rt6i_flags & noflags) != 0)
2315 read_unlock_bh(&table->tb6_lock);
2320 /* Create "default" multicast route to the interface */
2322 static void addrconf_add_mroute(struct net_device *dev)
2324 struct fib6_config cfg = {
2325 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2326 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2327 .fc_ifindex = dev->ifindex,
2330 .fc_nlinfo.nl_net = dev_net(dev),
2333 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2335 ip6_route_add(&cfg);
2338 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2340 struct inet6_dev *idev;
2344 idev = ipv6_find_idev(dev);
2346 return ERR_PTR(-ENOBUFS);
2348 if (idev->cnf.disable_ipv6)
2349 return ERR_PTR(-EACCES);
2351 /* Add default multicast route */
2352 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2353 addrconf_add_mroute(dev);
2358 static void manage_tempaddrs(struct inet6_dev *idev,
2359 struct inet6_ifaddr *ifp,
2360 __u32 valid_lft, __u32 prefered_lft,
2361 bool create, unsigned long now)
2364 struct inet6_ifaddr *ift;
2366 read_lock_bh(&idev->lock);
2367 /* update all temporary addresses in the list */
2368 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2369 int age, max_valid, max_prefered;
2371 if (ifp != ift->ifpub)
2374 /* RFC 4941 section 3.3:
2375 * If a received option will extend the lifetime of a public
2376 * address, the lifetimes of temporary addresses should
2377 * be extended, subject to the overall constraint that no
2378 * temporary addresses should ever remain "valid" or "preferred"
2379 * for a time longer than (TEMP_VALID_LIFETIME) or
2380 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2382 age = (now - ift->cstamp) / HZ;
2383 max_valid = idev->cnf.temp_valid_lft - age;
2387 max_prefered = idev->cnf.temp_prefered_lft -
2388 idev->desync_factor - age;
2389 if (max_prefered < 0)
2392 if (valid_lft > max_valid)
2393 valid_lft = max_valid;
2395 if (prefered_lft > max_prefered)
2396 prefered_lft = max_prefered;
2398 spin_lock(&ift->lock);
2400 ift->valid_lft = valid_lft;
2401 ift->prefered_lft = prefered_lft;
2403 if (prefered_lft > 0)
2404 ift->flags &= ~IFA_F_DEPRECATED;
2406 spin_unlock(&ift->lock);
2407 if (!(flags&IFA_F_TENTATIVE))
2408 ipv6_ifa_notify(0, ift);
2411 if ((create || list_empty(&idev->tempaddr_list)) &&
2412 idev->cnf.use_tempaddr > 0) {
2413 /* When a new public address is created as described
2414 * in [ADDRCONF], also create a new temporary address.
2415 * Also create a temporary address if it's enabled but
2416 * no temporary address currently exists.
2418 read_unlock_bh(&idev->lock);
2419 ipv6_create_tempaddr(ifp, NULL);
2421 read_unlock_bh(&idev->lock);
2425 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2427 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2428 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2431 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2432 const struct prefix_info *pinfo,
2433 struct inet6_dev *in6_dev,
2434 const struct in6_addr *addr, int addr_type,
2435 u32 addr_flags, bool sllao, bool tokenized,
2436 __u32 valid_lft, u32 prefered_lft)
2438 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2439 int create = 0, update_lft = 0;
2441 if (!ifp && valid_lft) {
2442 int max_addresses = in6_dev->cnf.max_addresses;
2444 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2445 if (in6_dev->cnf.optimistic_dad &&
2446 !net->ipv6.devconf_all->forwarding && sllao)
2447 addr_flags |= IFA_F_OPTIMISTIC;
2450 /* Do not allow to create too much of autoconfigured
2451 * addresses; this would be too easy way to crash kernel.
2453 if (!max_addresses ||
2454 ipv6_count_addresses(in6_dev) < max_addresses)
2455 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2457 addr_type&IPV6_ADDR_SCOPE_MASK,
2458 addr_flags, valid_lft,
2461 if (IS_ERR_OR_NULL(ifp))
2466 spin_lock_bh(&ifp->lock);
2467 ifp->flags |= IFA_F_MANAGETEMPADDR;
2468 ifp->cstamp = jiffies;
2469 ifp->tokenized = tokenized;
2470 spin_unlock_bh(&ifp->lock);
2471 addrconf_dad_start(ifp);
2479 /* update lifetime (RFC2462 5.5.3 e) */
2480 spin_lock_bh(&ifp->lock);
2482 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2483 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2486 if (!update_lft && !create && stored_lft) {
2487 const u32 minimum_lft = min_t(u32,
2488 stored_lft, MIN_VALID_LIFETIME);
2489 valid_lft = max(valid_lft, minimum_lft);
2491 /* RFC4862 Section 5.5.3e:
2492 * "Note that the preferred lifetime of the
2493 * corresponding address is always reset to
2494 * the Preferred Lifetime in the received
2495 * Prefix Information option, regardless of
2496 * whether the valid lifetime is also reset or
2499 * So we should always update prefered_lft here.
2505 ifp->valid_lft = valid_lft;
2506 ifp->prefered_lft = prefered_lft;
2509 ifp->flags &= ~IFA_F_DEPRECATED;
2510 spin_unlock_bh(&ifp->lock);
2512 if (!(flags&IFA_F_TENTATIVE))
2513 ipv6_ifa_notify(0, ifp);
2515 spin_unlock_bh(&ifp->lock);
2517 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2526 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2528 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2530 struct prefix_info *pinfo;
2535 struct inet6_dev *in6_dev;
2536 struct net *net = dev_net(dev);
2538 pinfo = (struct prefix_info *) opt;
2540 if (len < sizeof(struct prefix_info)) {
2541 ADBG("addrconf: prefix option too short\n");
2546 * Validation checks ([ADDRCONF], page 19)
2549 addr_type = ipv6_addr_type(&pinfo->prefix);
2551 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2554 valid_lft = ntohl(pinfo->valid);
2555 prefered_lft = ntohl(pinfo->prefered);
2557 if (prefered_lft > valid_lft) {
2558 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2562 in6_dev = in6_dev_get(dev);
2565 net_dbg_ratelimited("addrconf: device %s not configured\n",
2571 * Two things going on here:
2572 * 1) Add routes for on-link prefixes
2573 * 2) Configure prefixes with the auto flag set
2576 if (pinfo->onlink) {
2577 struct rt6_info *rt;
2578 unsigned long rt_expires;
2580 /* Avoid arithmetic overflow. Really, we could
2581 * save rt_expires in seconds, likely valid_lft,
2582 * but it would require division in fib gc, that it
2586 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2588 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2590 if (addrconf_finite_timeout(rt_expires))
2593 rt = addrconf_get_prefix_route(&pinfo->prefix,
2596 RTF_ADDRCONF | RTF_PREFIX_RT,
2597 RTF_GATEWAY | RTF_DEFAULT);
2600 /* Autoconf prefix route */
2601 if (valid_lft == 0) {
2604 } else if (addrconf_finite_timeout(rt_expires)) {
2606 rt6_set_expires(rt, jiffies + rt_expires);
2608 rt6_clean_expires(rt);
2610 } else if (valid_lft) {
2611 clock_t expires = 0;
2612 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2613 if (addrconf_finite_timeout(rt_expires)) {
2615 flags |= RTF_EXPIRES;
2616 expires = jiffies_to_clock_t(rt_expires);
2618 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2619 dev, expires, flags);
2624 /* Try to figure out our local address for this prefix */
2626 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2627 struct in6_addr addr;
2628 bool tokenized = false, dev_addr_generated = false;
2630 if (pinfo->prefix_len == 64) {
2631 memcpy(&addr, &pinfo->prefix, 8);
2633 if (!ipv6_addr_any(&in6_dev->token)) {
2634 read_lock_bh(&in6_dev->lock);
2635 memcpy(addr.s6_addr + 8,
2636 in6_dev->token.s6_addr + 8, 8);
2637 read_unlock_bh(&in6_dev->lock);
2639 } else if (is_addr_mode_generate_stable(in6_dev) &&
2640 !ipv6_generate_stable_address(&addr, 0,
2642 addr_flags |= IFA_F_STABLE_PRIVACY;
2644 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2645 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2648 dev_addr_generated = true;
2652 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2657 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2660 tokenized, valid_lft,
2665 /* Ignore error case here because previous prefix add addr was
2666 * successful which will be notified.
2668 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2669 addr_type, addr_flags, sllao,
2670 tokenized, valid_lft,
2672 dev_addr_generated);
2674 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2676 in6_dev_put(in6_dev);
2680 * Set destination address.
2681 * Special case for SIT interfaces where we create a new "virtual"
2684 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2686 struct in6_ifreq ireq;
2687 struct net_device *dev;
2693 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2696 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2702 #if IS_ENABLED(CONFIG_IPV6_SIT)
2703 if (dev->type == ARPHRD_SIT) {
2704 const struct net_device_ops *ops = dev->netdev_ops;
2706 struct ip_tunnel_parm p;
2708 err = -EADDRNOTAVAIL;
2709 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2712 memset(&p, 0, sizeof(p));
2713 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2717 p.iph.protocol = IPPROTO_IPV6;
2719 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2721 if (ops->ndo_do_ioctl) {
2722 mm_segment_t oldfs = get_fs();
2725 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2732 dev = __dev_get_by_name(net, p.name);
2735 err = dev_open(dev);
2745 static int ipv6_mc_config(struct sock *sk, bool join,
2746 const struct in6_addr *addr, int ifindex)
2754 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2756 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2763 * Manual configuration of address on an interface
2765 static int inet6_addr_add(struct net *net, int ifindex,
2766 const struct in6_addr *pfx,
2767 const struct in6_addr *peer_pfx,
2768 unsigned int plen, __u32 ifa_flags,
2769 __u32 prefered_lft, __u32 valid_lft)
2771 struct inet6_ifaddr *ifp;
2772 struct inet6_dev *idev;
2773 struct net_device *dev;
2774 unsigned long timeout;
2784 /* check the lifetime */
2785 if (!valid_lft || prefered_lft > valid_lft)
2788 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2791 dev = __dev_get_by_index(net, ifindex);
2795 idev = addrconf_add_dev(dev);
2797 return PTR_ERR(idev);
2799 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2800 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2801 true, pfx, ifindex);
2807 scope = ipv6_addr_scope(pfx);
2809 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2810 if (addrconf_finite_timeout(timeout)) {
2811 expires = jiffies_to_clock_t(timeout * HZ);
2812 valid_lft = timeout;
2813 flags = RTF_EXPIRES;
2817 ifa_flags |= IFA_F_PERMANENT;
2820 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2821 if (addrconf_finite_timeout(timeout)) {
2823 ifa_flags |= IFA_F_DEPRECATED;
2824 prefered_lft = timeout;
2827 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2828 valid_lft, prefered_lft);
2831 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2832 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2837 * Note that section 3.1 of RFC 4429 indicates
2838 * that the Optimistic flag should not be set for
2839 * manually configured addresses
2841 addrconf_dad_start(ifp);
2842 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2843 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2846 addrconf_verify_rtnl();
2848 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2849 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2850 false, pfx, ifindex);
2853 return PTR_ERR(ifp);
2856 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2857 const struct in6_addr *pfx, unsigned int plen)
2859 struct inet6_ifaddr *ifp;
2860 struct inet6_dev *idev;
2861 struct net_device *dev;
2866 dev = __dev_get_by_index(net, ifindex);
2870 idev = __in6_dev_get(dev);
2874 read_lock_bh(&idev->lock);
2875 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2876 if (ifp->prefix_len == plen &&
2877 ipv6_addr_equal(pfx, &ifp->addr)) {
2879 read_unlock_bh(&idev->lock);
2881 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2882 (ifa_flags & IFA_F_MANAGETEMPADDR))
2883 manage_tempaddrs(idev, ifp, 0, 0, false,
2886 addrconf_verify_rtnl();
2887 if (ipv6_addr_is_multicast(pfx)) {
2888 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2889 false, pfx, dev->ifindex);
2894 read_unlock_bh(&idev->lock);
2895 return -EADDRNOTAVAIL;
2899 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2901 struct in6_ifreq ireq;
2904 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2907 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2911 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2912 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2913 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2918 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2920 struct in6_ifreq ireq;
2923 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2926 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2930 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2931 ireq.ifr6_prefixlen);
2936 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2937 int plen, int scope)
2939 struct inet6_ifaddr *ifp;
2941 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2942 scope, IFA_F_PERMANENT,
2943 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2945 spin_lock_bh(&ifp->lock);
2946 ifp->flags &= ~IFA_F_TENTATIVE;
2947 spin_unlock_bh(&ifp->lock);
2948 rt_genid_bump_ipv6(dev_net(idev->dev));
2949 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2954 #if IS_ENABLED(CONFIG_IPV6_SIT)
2955 static void sit_add_v4_addrs(struct inet6_dev *idev)
2957 struct in6_addr addr;
2958 struct net_device *dev;
2959 struct net *net = dev_net(idev->dev);
2965 memset(&addr, 0, sizeof(struct in6_addr));
2966 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2968 if (idev->dev->flags&IFF_POINTOPOINT) {
2969 addr.s6_addr32[0] = htonl(0xfe800000);
2973 scope = IPV6_ADDR_COMPATv4;
2975 pflags |= RTF_NONEXTHOP;
2978 if (addr.s6_addr32[3]) {
2979 add_addr(idev, &addr, plen, scope);
2980 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2984 for_each_netdev(net, dev) {
2985 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2986 if (in_dev && (dev->flags & IFF_UP)) {
2987 struct in_ifaddr *ifa;
2991 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2993 addr.s6_addr32[3] = ifa->ifa_local;
2995 if (ifa->ifa_scope == RT_SCOPE_LINK)
2997 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2998 if (idev->dev->flags&IFF_POINTOPOINT)
3003 add_addr(idev, &addr, plen, flag);
3004 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3012 static void init_loopback(struct net_device *dev)
3014 struct inet6_dev *idev;
3015 struct net_device *sp_dev;
3016 struct inet6_ifaddr *sp_ifa;
3017 struct rt6_info *sp_rt;
3023 idev = ipv6_find_idev(dev);
3025 pr_debug("%s: add_dev failed\n", __func__);
3029 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3031 /* Add routes to other interface's IPv6 addresses */
3032 for_each_netdev(dev_net(dev), sp_dev) {
3033 if (!strcmp(sp_dev->name, dev->name))
3036 idev = __in6_dev_get(sp_dev);
3040 read_lock_bh(&idev->lock);
3041 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
3043 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
3047 /* This dst has been added to garbage list when
3048 * lo device down, release this obsolete dst and
3049 * reallocate a new router for ifa.
3051 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3052 ip6_rt_put(sp_ifa->rt);
3059 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3061 /* Failure cases are ignored */
3062 if (!IS_ERR(sp_rt)) {
3067 read_unlock_bh(&idev->lock);
3071 void addrconf_add_linklocal(struct inet6_dev *idev,
3072 const struct in6_addr *addr, u32 flags)
3074 struct inet6_ifaddr *ifp;
3075 u32 addr_flags = flags | IFA_F_PERMANENT;
3077 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3078 if (idev->cnf.optimistic_dad &&
3079 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3080 addr_flags |= IFA_F_OPTIMISTIC;
3083 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3084 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3086 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3087 addrconf_dad_start(ifp);
3091 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3093 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3095 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3098 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3099 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3102 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3103 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3109 static int ipv6_generate_stable_address(struct in6_addr *address,
3111 const struct inet6_dev *idev)
3113 static DEFINE_SPINLOCK(lock);
3114 static __u32 digest[SHA_DIGEST_WORDS];
3115 static __u32 workspace[SHA_WORKSPACE_WORDS];
3118 char __data[SHA_MESSAGE_BYTES];
3120 struct in6_addr secret;
3122 unsigned char hwaddr[MAX_ADDR_LEN];
3127 struct in6_addr secret;
3128 struct in6_addr temp;
3129 struct net *net = dev_net(idev->dev);
3131 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3133 if (idev->cnf.stable_secret.initialized)
3134 secret = idev->cnf.stable_secret.secret;
3135 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3136 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3141 spin_lock_bh(&lock);
3144 memset(&data, 0, sizeof(data));
3145 memset(workspace, 0, sizeof(workspace));
3146 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3147 data.prefix[0] = address->s6_addr32[0];
3148 data.prefix[1] = address->s6_addr32[1];
3149 data.secret = secret;
3150 data.dad_count = dad_count;
3152 sha_transform(digest, data.__data, workspace);
3155 temp.s6_addr32[2] = (__force __be32)digest[0];
3156 temp.s6_addr32[3] = (__force __be32)digest[1];
3158 spin_unlock_bh(&lock);
3160 if (ipv6_reserved_interfaceid(temp)) {
3162 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3171 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3173 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3177 s = &idev->cnf.stable_secret;
3178 get_random_bytes(&s->secret, sizeof(s->secret));
3179 s->initialized = true;
3182 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3184 struct in6_addr addr;
3186 /* no link local addresses on L3 master devices */
3187 if (netif_is_l3_master(idev->dev))
3190 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3192 switch (idev->cnf.addr_gen_mode) {
3193 case IN6_ADDR_GEN_MODE_RANDOM:
3194 ipv6_gen_mode_random_init(idev);
3196 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3197 if (!ipv6_generate_stable_address(&addr, 0, idev))
3198 addrconf_add_linklocal(idev, &addr,
3199 IFA_F_STABLE_PRIVACY);
3200 else if (prefix_route)
3201 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3203 case IN6_ADDR_GEN_MODE_EUI64:
3204 /* addrconf_add_linklocal also adds a prefix_route and we
3205 * only need to care about prefix routes if ipv6_generate_eui64
3206 * couldn't generate one.
3208 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3209 addrconf_add_linklocal(idev, &addr, 0);
3210 else if (prefix_route)
3211 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3213 case IN6_ADDR_GEN_MODE_NONE:
3215 /* will not add any link local address */
3220 static void addrconf_dev_config(struct net_device *dev)
3222 struct inet6_dev *idev;
3226 if ((dev->type != ARPHRD_ETHER) &&
3227 (dev->type != ARPHRD_FDDI) &&
3228 (dev->type != ARPHRD_ARCNET) &&
3229 (dev->type != ARPHRD_INFINIBAND) &&
3230 (dev->type != ARPHRD_IEEE1394) &&
3231 (dev->type != ARPHRD_TUNNEL6) &&
3232 (dev->type != ARPHRD_6LOWPAN) &&
3233 (dev->type != ARPHRD_IP6GRE) &&
3234 (dev->type != ARPHRD_IPGRE) &&
3235 (dev->type != ARPHRD_TUNNEL) &&
3236 (dev->type != ARPHRD_NONE)) {
3237 /* Alas, we support only Ethernet autoconfiguration. */
3241 idev = addrconf_add_dev(dev);
3245 /* this device type has no EUI support */
3246 if (dev->type == ARPHRD_NONE &&
3247 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3248 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3250 addrconf_addr_gen(idev, false);
3253 #if IS_ENABLED(CONFIG_IPV6_SIT)
3254 static void addrconf_sit_config(struct net_device *dev)
3256 struct inet6_dev *idev;
3261 * Configure the tunnel with one of our IPv4
3262 * addresses... we should configure all of
3263 * our v4 addrs in the tunnel
3266 idev = ipv6_find_idev(dev);
3268 pr_debug("%s: add_dev failed\n", __func__);
3272 if (dev->priv_flags & IFF_ISATAP) {
3273 addrconf_addr_gen(idev, false);
3277 sit_add_v4_addrs(idev);
3279 if (dev->flags&IFF_POINTOPOINT)
3280 addrconf_add_mroute(dev);
3284 #if IS_ENABLED(CONFIG_NET_IPGRE)
3285 static void addrconf_gre_config(struct net_device *dev)
3287 struct inet6_dev *idev;
3291 idev = ipv6_find_idev(dev);
3293 pr_debug("%s: add_dev failed\n", __func__);
3297 addrconf_addr_gen(idev, true);
3298 if (dev->flags & IFF_POINTOPOINT)
3299 addrconf_add_mroute(dev);
3303 static int fixup_permanent_addr(struct inet6_dev *idev,
3304 struct inet6_ifaddr *ifp)
3306 /* rt6i_ref == 0 means the host route was removed from the
3307 * FIB, for example, if 'lo' device is taken down. In that
3308 * case regenerate the host route.
3310 if (!ifp->rt || !atomic_read(&ifp->rt->rt6i_ref)) {
3311 struct rt6_info *rt, *prev;
3313 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3314 if (unlikely(IS_ERR(rt)))
3317 /* ifp->rt can be accessed outside of rtnl */
3318 spin_lock(&ifp->lock);
3321 spin_unlock(&ifp->lock);
3326 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3327 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3331 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3332 addrconf_dad_start(ifp);
3337 static void addrconf_permanent_addr(struct net_device *dev)
3339 struct inet6_ifaddr *ifp, *tmp;
3340 struct inet6_dev *idev;
3342 idev = __in6_dev_get(dev);
3346 write_lock_bh(&idev->lock);
3348 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3349 if ((ifp->flags & IFA_F_PERMANENT) &&
3350 fixup_permanent_addr(idev, ifp) < 0) {
3351 write_unlock_bh(&idev->lock);
3353 write_lock_bh(&idev->lock);
3355 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3356 idev->dev->name, &ifp->addr);
3360 write_unlock_bh(&idev->lock);
3363 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3366 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3367 struct netdev_notifier_changeupper_info *info;
3368 struct inet6_dev *idev = __in6_dev_get(dev);
3369 int run_pending = 0;
3373 case NETDEV_REGISTER:
3374 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3375 idev = ipv6_add_dev(dev);
3377 return notifier_from_errno(PTR_ERR(idev));
3381 case NETDEV_CHANGEMTU:
3382 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3383 if (dev->mtu < IPV6_MIN_MTU) {
3384 addrconf_ifdown(dev, 1);
3389 rt6_mtu_change(dev, dev->mtu);
3390 idev->cnf.mtu6 = dev->mtu;
3394 /* allocate new idev */
3395 idev = ipv6_add_dev(dev);
3399 /* device is still not ready */
3400 if (!(idev->if_flags & IF_READY))
3409 if (dev->flags & IFF_SLAVE)
3412 if (idev && idev->cnf.disable_ipv6)
3415 if (event == NETDEV_UP) {
3416 /* restore routes for permanent addresses */
3417 addrconf_permanent_addr(dev);
3419 if (!addrconf_qdisc_ok(dev)) {
3420 /* device is not ready yet. */
3421 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3426 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3427 idev = ipv6_add_dev(dev);
3429 if (!IS_ERR_OR_NULL(idev)) {
3430 idev->if_flags |= IF_READY;
3433 } else if (event == NETDEV_CHANGE) {
3434 if (!addrconf_qdisc_ok(dev)) {
3435 /* device is still not ready. */
3440 if (idev->if_flags & IF_READY) {
3441 /* device is already configured -
3442 * but resend MLD reports, we might
3443 * have roamed and need to update
3444 * multicast snooping switches
3449 idev->if_flags |= IF_READY;
3452 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3458 switch (dev->type) {
3459 #if IS_ENABLED(CONFIG_IPV6_SIT)
3461 addrconf_sit_config(dev);
3464 #if IS_ENABLED(CONFIG_NET_IPGRE)
3466 addrconf_gre_config(dev);
3469 case ARPHRD_LOOPBACK:
3474 addrconf_dev_config(dev);
3478 if (!IS_ERR_OR_NULL(idev)) {
3480 addrconf_dad_run(idev);
3483 * If the MTU changed during the interface down,
3484 * when the interface up, the changed MTU must be
3485 * reflected in the idev as well as routers.
3487 if (idev->cnf.mtu6 != dev->mtu &&
3488 dev->mtu >= IPV6_MIN_MTU) {
3489 rt6_mtu_change(dev, dev->mtu);
3490 idev->cnf.mtu6 = dev->mtu;
3492 idev->tstamp = jiffies;
3493 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3496 * If the changed mtu during down is lower than
3497 * IPV6_MIN_MTU stop IPv6 on this interface.
3499 if (dev->mtu < IPV6_MIN_MTU)
3500 addrconf_ifdown(dev, 1);
3505 case NETDEV_UNREGISTER:
3507 * Remove all addresses from this interface.
3509 addrconf_ifdown(dev, event != NETDEV_DOWN);
3512 case NETDEV_CHANGENAME:
3514 snmp6_unregister_dev(idev);
3515 addrconf_sysctl_unregister(idev);
3516 err = addrconf_sysctl_register(idev);
3518 return notifier_from_errno(err);
3519 err = snmp6_register_dev(idev);
3521 addrconf_sysctl_unregister(idev);
3522 return notifier_from_errno(err);
3527 case NETDEV_PRE_TYPE_CHANGE:
3528 case NETDEV_POST_TYPE_CHANGE:
3530 addrconf_type_change(dev, event);
3533 case NETDEV_CHANGEUPPER:
3536 /* flush all routes if dev is linked to or unlinked from
3537 * an L3 master device (e.g., VRF)
3539 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3540 addrconf_ifdown(dev, 0);
3547 * addrconf module should be notified of a device going up
3549 static struct notifier_block ipv6_dev_notf = {
3550 .notifier_call = addrconf_notify,
3553 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3555 struct inet6_dev *idev;
3558 idev = __in6_dev_get(dev);
3560 if (event == NETDEV_POST_TYPE_CHANGE)
3561 ipv6_mc_remap(idev);
3562 else if (event == NETDEV_PRE_TYPE_CHANGE)
3563 ipv6_mc_unmap(idev);
3566 static bool addr_is_local(const struct in6_addr *addr)
3568 return ipv6_addr_type(addr) &
3569 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3572 static int addrconf_ifdown(struct net_device *dev, int how)
3574 struct net *net = dev_net(dev);
3575 struct inet6_dev *idev;
3576 struct inet6_ifaddr *ifa, *tmp;
3577 struct list_head del_list;
3584 rt6_ifdown(net, dev);
3585 neigh_ifdown(&nd_tbl, dev);
3587 idev = __in6_dev_get(dev);
3592 * Step 1: remove reference to ipv6 device from parent device.
3598 /* protected by rtnl_lock */
3599 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3601 /* Step 1.5: remove snmp6 entry */
3602 snmp6_unregister_dev(idev);
3606 /* aggregate the system setting and interface setting */
3607 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3609 _keep_addr = idev->cnf.keep_addr_on_down;
3611 /* combine the user config with event to determine if permanent
3612 * addresses are to be removed from address hash table
3614 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3616 /* Step 2: clear hash table */
3617 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3618 struct hlist_head *h = &inet6_addr_lst[i];
3620 spin_lock_bh(&addrconf_hash_lock);
3622 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3623 if (ifa->idev == idev) {
3624 addrconf_del_dad_work(ifa);
3625 /* combined flag + permanent flag decide if
3626 * address is retained on a down event
3629 !(ifa->flags & IFA_F_PERMANENT) ||
3630 addr_is_local(&ifa->addr)) {
3631 hlist_del_init_rcu(&ifa->addr_lst);
3636 spin_unlock_bh(&addrconf_hash_lock);
3639 write_lock_bh(&idev->lock);
3641 addrconf_del_rs_timer(idev);
3643 /* Step 2: clear flags for stateless addrconf */
3645 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3647 /* Step 3: clear tempaddr list */
3648 while (!list_empty(&idev->tempaddr_list)) {
3649 ifa = list_first_entry(&idev->tempaddr_list,
3650 struct inet6_ifaddr, tmp_list);
3651 list_del(&ifa->tmp_list);
3652 write_unlock_bh(&idev->lock);
3653 spin_lock_bh(&ifa->lock);
3656 in6_ifa_put(ifa->ifpub);
3659 spin_unlock_bh(&ifa->lock);
3661 write_lock_bh(&idev->lock);
3664 /* re-combine the user config with event to determine if permanent
3665 * addresses are to be removed from the interface list
3667 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3669 INIT_LIST_HEAD(&del_list);
3670 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3671 struct rt6_info *rt = NULL;
3674 addrconf_del_dad_work(ifa);
3676 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3677 !addr_is_local(&ifa->addr);
3679 list_move(&ifa->if_list, &del_list);
3681 write_unlock_bh(&idev->lock);
3682 spin_lock_bh(&ifa->lock);
3685 /* set state to skip the notifier below */
3686 state = INET6_IFADDR_STATE_DEAD;
3687 ifa->state = INET6_IFADDR_STATE_PREDAD;
3688 if (!(ifa->flags & IFA_F_NODAD))
3689 ifa->flags |= IFA_F_TENTATIVE;
3695 ifa->state = INET6_IFADDR_STATE_DEAD;
3698 spin_unlock_bh(&ifa->lock);
3703 if (state != INET6_IFADDR_STATE_DEAD) {
3704 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3705 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3707 if (idev->cnf.forwarding)
3708 addrconf_leave_anycast(ifa);
3709 addrconf_leave_solict(ifa->idev, &ifa->addr);
3712 write_lock_bh(&idev->lock);
3715 write_unlock_bh(&idev->lock);
3717 /* now clean up addresses to be removed */
3718 while (!list_empty(&del_list)) {
3719 ifa = list_first_entry(&del_list,
3720 struct inet6_ifaddr, if_list);
3721 list_del(&ifa->if_list);
3726 /* Step 5: Discard anycast and multicast list */
3728 ipv6_ac_destroy_dev(idev);
3729 ipv6_mc_destroy_dev(idev);
3734 idev->tstamp = jiffies;
3736 /* Last: Shot the device (if unregistered) */
3738 addrconf_sysctl_unregister(idev);
3739 neigh_parms_release(&nd_tbl, idev->nd_parms);
3740 neigh_ifdown(&nd_tbl, dev);
3746 static void addrconf_rs_timer(unsigned long data)
3748 struct inet6_dev *idev = (struct inet6_dev *)data;
3749 struct net_device *dev = idev->dev;
3750 struct in6_addr lladdr;
3752 write_lock(&idev->lock);
3753 if (idev->dead || !(idev->if_flags & IF_READY))
3756 if (!ipv6_accept_ra(idev))
3759 /* Announcement received after solicitation was sent */
3760 if (idev->if_flags & IF_RA_RCVD)
3763 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3764 write_unlock(&idev->lock);
3765 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3766 ndisc_send_rs(dev, &lladdr,
3767 &in6addr_linklocal_allrouters);
3771 write_lock(&idev->lock);
3772 idev->rs_interval = rfc3315_s14_backoff_update(
3773 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3774 /* The wait after the last probe can be shorter */
3775 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3776 idev->cnf.rtr_solicits) ?
3777 idev->cnf.rtr_solicit_delay :
3781 * Note: we do not support deprecated "all on-link"
3782 * assumption any longer.
3784 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3788 write_unlock(&idev->lock);
3794 * Duplicate Address Detection
3796 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3798 unsigned long rand_num;
3799 struct inet6_dev *idev = ifp->idev;
3802 if (ifp->flags & IFA_F_OPTIMISTIC)
3805 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3808 if (idev->cnf.enhanced_dad ||
3809 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3811 get_random_bytes(&nonce, 6);
3814 ifp->dad_nonce = nonce;
3815 ifp->dad_probes = idev->cnf.dad_transmits;
3816 addrconf_mod_dad_work(ifp, rand_num);
3819 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3821 struct inet6_dev *idev = ifp->idev;
3822 struct net_device *dev = idev->dev;
3823 bool bump_id, notify = false;
3825 addrconf_join_solict(dev, &ifp->addr);
3827 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3829 read_lock_bh(&idev->lock);
3830 spin_lock(&ifp->lock);
3831 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3834 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3835 idev->cnf.accept_dad < 1 ||
3836 !(ifp->flags&IFA_F_TENTATIVE) ||
3837 ifp->flags & IFA_F_NODAD) {
3838 bump_id = ifp->flags & IFA_F_TENTATIVE;
3839 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3840 spin_unlock(&ifp->lock);
3841 read_unlock_bh(&idev->lock);
3843 addrconf_dad_completed(ifp, bump_id);
3847 if (!(idev->if_flags & IF_READY)) {
3848 spin_unlock(&ifp->lock);
3849 read_unlock_bh(&idev->lock);
3851 * If the device is not ready:
3852 * - keep it tentative if it is a permanent address.
3853 * - otherwise, kill it.
3856 addrconf_dad_stop(ifp, 0);
3861 * Optimistic nodes can start receiving
3864 if (ifp->flags & IFA_F_OPTIMISTIC) {
3865 ip6_ins_rt(ifp->rt);
3866 if (ipv6_use_optimistic_addr(idev)) {
3867 /* Because optimistic nodes can use this address,
3868 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3874 addrconf_dad_kick(ifp);
3876 spin_unlock(&ifp->lock);
3877 read_unlock_bh(&idev->lock);
3879 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3882 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3884 bool begin_dad = false;
3886 spin_lock_bh(&ifp->lock);
3887 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3888 ifp->state = INET6_IFADDR_STATE_PREDAD;
3891 spin_unlock_bh(&ifp->lock);
3894 addrconf_mod_dad_work(ifp, 0);
3897 static void addrconf_dad_work(struct work_struct *w)
3899 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3900 struct inet6_ifaddr,
3902 struct inet6_dev *idev = ifp->idev;
3903 bool bump_id, disable_ipv6 = false;
3904 struct in6_addr mcaddr;
3910 } action = DAD_PROCESS;
3914 spin_lock_bh(&ifp->lock);
3915 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3917 ifp->state = INET6_IFADDR_STATE_DAD;
3918 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3920 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3922 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3923 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3924 struct in6_addr addr;
3926 addr.s6_addr32[0] = htonl(0xfe800000);
3927 addr.s6_addr32[1] = 0;
3929 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3930 ipv6_addr_equal(&ifp->addr, &addr)) {
3931 /* DAD failed for link-local based on MAC */
3932 idev->cnf.disable_ipv6 = 1;
3934 pr_info("%s: IPv6 being disabled!\n",
3935 ifp->idev->dev->name);
3936 disable_ipv6 = true;
3940 spin_unlock_bh(&ifp->lock);
3942 if (action == DAD_BEGIN) {
3943 addrconf_dad_begin(ifp);
3945 } else if (action == DAD_ABORT) {
3947 addrconf_dad_stop(ifp, 1);
3949 addrconf_ifdown(idev->dev, 0);
3953 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3956 write_lock_bh(&idev->lock);
3957 if (idev->dead || !(idev->if_flags & IF_READY)) {
3958 write_unlock_bh(&idev->lock);
3962 spin_lock(&ifp->lock);
3963 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3964 spin_unlock(&ifp->lock);
3965 write_unlock_bh(&idev->lock);
3969 if (ifp->dad_probes == 0) {
3971 * DAD was successful
3974 bump_id = ifp->flags & IFA_F_TENTATIVE;
3975 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3976 spin_unlock(&ifp->lock);
3977 write_unlock_bh(&idev->lock);
3979 addrconf_dad_completed(ifp, bump_id);
3985 addrconf_mod_dad_work(ifp,
3986 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3987 spin_unlock(&ifp->lock);
3988 write_unlock_bh(&idev->lock);
3990 /* send a neighbour solicitation for our addr */
3991 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3992 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
3999 /* ifp->idev must be at least read locked */
4000 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4002 struct inet6_ifaddr *ifpiter;
4003 struct inet6_dev *idev = ifp->idev;
4005 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4006 if (ifpiter->scope > IFA_LINK)
4008 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4009 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4010 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4017 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
4019 struct net_device *dev = ifp->idev->dev;
4020 struct in6_addr lladdr;
4021 bool send_rs, send_mld;
4023 addrconf_del_dad_work(ifp);
4026 * Configure the address for reception. Now it is valid.
4029 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4031 /* If added prefix is link local and we are prepared to process
4032 router advertisements, start sending router solicitations.
4035 read_lock_bh(&ifp->idev->lock);
4036 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4037 send_rs = send_mld &&
4038 ipv6_accept_ra(ifp->idev) &&
4039 ifp->idev->cnf.rtr_solicits != 0 &&
4040 (dev->flags&IFF_LOOPBACK) == 0;
4041 read_unlock_bh(&ifp->idev->lock);
4043 /* While dad is in progress mld report's source address is in6_addrany.
4044 * Resend with proper ll now.
4047 ipv6_mc_dad_complete(ifp->idev);
4051 * If a host as already performed a random delay
4052 * [...] as part of DAD [...] there is no need
4053 * to delay again before sending the first RS
4055 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4057 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4059 write_lock_bh(&ifp->idev->lock);
4060 spin_lock(&ifp->lock);
4061 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4062 ifp->idev->cnf.rtr_solicit_interval);
4063 ifp->idev->rs_probes = 1;
4064 ifp->idev->if_flags |= IF_RS_SENT;
4065 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4066 spin_unlock(&ifp->lock);
4067 write_unlock_bh(&ifp->idev->lock);
4071 rt_genid_bump_ipv6(dev_net(dev));
4073 /* Make sure that a new temporary address will be created
4074 * before this temporary address becomes deprecated.
4076 if (ifp->flags & IFA_F_TEMPORARY)
4077 addrconf_verify_rtnl();
4080 static void addrconf_dad_run(struct inet6_dev *idev)
4082 struct inet6_ifaddr *ifp;
4084 read_lock_bh(&idev->lock);
4085 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4086 spin_lock(&ifp->lock);
4087 if (ifp->flags & IFA_F_TENTATIVE &&
4088 ifp->state == INET6_IFADDR_STATE_DAD)
4089 addrconf_dad_kick(ifp);
4090 spin_unlock(&ifp->lock);
4092 read_unlock_bh(&idev->lock);
4095 #ifdef CONFIG_PROC_FS
4096 struct if6_iter_state {
4097 struct seq_net_private p;
4102 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4104 struct inet6_ifaddr *ifa = NULL;
4105 struct if6_iter_state *state = seq->private;
4106 struct net *net = seq_file_net(seq);
4109 /* initial bucket if pos is 0 */
4115 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4116 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4118 if (!net_eq(dev_net(ifa->idev->dev), net))
4120 /* sync with offset */
4121 if (p < state->offset) {
4129 /* prepare for next bucket */
4136 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4137 struct inet6_ifaddr *ifa)
4139 struct if6_iter_state *state = seq->private;
4140 struct net *net = seq_file_net(seq);
4142 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4143 if (!net_eq(dev_net(ifa->idev->dev), net))
4149 while (++state->bucket < IN6_ADDR_HSIZE) {
4151 hlist_for_each_entry_rcu_bh(ifa,
4152 &inet6_addr_lst[state->bucket], addr_lst) {
4153 if (!net_eq(dev_net(ifa->idev->dev), net))
4163 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4167 return if6_get_first(seq, *pos);
4170 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4172 struct inet6_ifaddr *ifa;
4174 ifa = if6_get_next(seq, v);
4179 static void if6_seq_stop(struct seq_file *seq, void *v)
4182 rcu_read_unlock_bh();
4185 static int if6_seq_show(struct seq_file *seq, void *v)
4187 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4188 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4190 ifp->idev->dev->ifindex,
4194 ifp->idev->dev->name);
4198 static const struct seq_operations if6_seq_ops = {
4199 .start = if6_seq_start,
4200 .next = if6_seq_next,
4201 .show = if6_seq_show,
4202 .stop = if6_seq_stop,
4205 static int if6_seq_open(struct inode *inode, struct file *file)
4207 return seq_open_net(inode, file, &if6_seq_ops,
4208 sizeof(struct if6_iter_state));
4211 static const struct file_operations if6_fops = {
4212 .owner = THIS_MODULE,
4213 .open = if6_seq_open,
4215 .llseek = seq_lseek,
4216 .release = seq_release_net,
4219 static int __net_init if6_proc_net_init(struct net *net)
4221 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4226 static void __net_exit if6_proc_net_exit(struct net *net)
4228 remove_proc_entry("if_inet6", net->proc_net);
4231 static struct pernet_operations if6_proc_net_ops = {
4232 .init = if6_proc_net_init,
4233 .exit = if6_proc_net_exit,
4236 int __init if6_proc_init(void)
4238 return register_pernet_subsys(&if6_proc_net_ops);
4241 void if6_proc_exit(void)
4243 unregister_pernet_subsys(&if6_proc_net_ops);
4245 #endif /* CONFIG_PROC_FS */
4247 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4248 /* Check if address is a home address configured on any interface. */
4249 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4252 struct inet6_ifaddr *ifp = NULL;
4253 unsigned int hash = inet6_addr_hash(addr);
4256 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4257 if (!net_eq(dev_net(ifp->idev->dev), net))
4259 if (ipv6_addr_equal(&ifp->addr, addr) &&
4260 (ifp->flags & IFA_F_HOMEADDRESS)) {
4265 rcu_read_unlock_bh();
4271 * Periodic address status verification
4274 static void addrconf_verify_rtnl(void)
4276 unsigned long now, next, next_sec, next_sched;
4277 struct inet6_ifaddr *ifp;
4284 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4286 cancel_delayed_work(&addr_chk_work);
4288 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4290 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4293 /* When setting preferred_lft to a value not zero or
4294 * infinity, while valid_lft is infinity
4295 * IFA_F_PERMANENT has a non-infinity life time.
4297 if ((ifp->flags & IFA_F_PERMANENT) &&
4298 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4301 spin_lock(&ifp->lock);
4302 /* We try to batch several events at once. */
4303 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4305 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4306 age >= ifp->valid_lft) {
4307 spin_unlock(&ifp->lock);
4311 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4312 spin_unlock(&ifp->lock);
4314 } else if (age >= ifp->prefered_lft) {
4315 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4318 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4320 ifp->flags |= IFA_F_DEPRECATED;
4323 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4324 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4325 next = ifp->tstamp + ifp->valid_lft * HZ;
4327 spin_unlock(&ifp->lock);
4332 ipv6_ifa_notify(0, ifp);
4336 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4337 !(ifp->flags&IFA_F_TENTATIVE)) {
4338 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4339 ifp->idev->cnf.dad_transmits *
4340 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4342 if (age >= ifp->prefered_lft - regen_advance) {
4343 struct inet6_ifaddr *ifpub = ifp->ifpub;
4344 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4345 next = ifp->tstamp + ifp->prefered_lft * HZ;
4346 if (!ifp->regen_count && ifpub) {
4349 in6_ifa_hold(ifpub);
4350 spin_unlock(&ifp->lock);
4352 spin_lock(&ifpub->lock);
4353 ifpub->regen_count = 0;
4354 spin_unlock(&ifpub->lock);
4355 ipv6_create_tempaddr(ifpub, ifp);
4360 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4361 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4362 spin_unlock(&ifp->lock);
4364 /* ifp->prefered_lft <= ifp->valid_lft */
4365 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4366 next = ifp->tstamp + ifp->prefered_lft * HZ;
4367 spin_unlock(&ifp->lock);
4372 next_sec = round_jiffies_up(next);
4375 /* If rounded timeout is accurate enough, accept it. */
4376 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4377 next_sched = next_sec;
4379 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4380 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4381 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4383 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4384 now, next, next_sec, next_sched);
4385 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4386 rcu_read_unlock_bh();
4389 static void addrconf_verify_work(struct work_struct *w)
4392 addrconf_verify_rtnl();
4396 static void addrconf_verify(void)
4398 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4401 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4402 struct in6_addr **peer_pfx)
4404 struct in6_addr *pfx = NULL;
4409 pfx = nla_data(addr);
4412 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4414 pfx = nla_data(local);
4420 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4421 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4422 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4423 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4424 [IFA_FLAGS] = { .len = sizeof(u32) },
4428 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4429 struct netlink_ext_ack *extack)
4431 struct net *net = sock_net(skb->sk);
4432 struct ifaddrmsg *ifm;
4433 struct nlattr *tb[IFA_MAX+1];
4434 struct in6_addr *pfx, *peer_pfx;
4438 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4443 ifm = nlmsg_data(nlh);
4444 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4448 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4450 /* We ignore other flags so far. */
4451 ifa_flags &= IFA_F_MANAGETEMPADDR;
4453 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4454 ifm->ifa_prefixlen);
4457 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4458 u32 prefered_lft, u32 valid_lft)
4462 unsigned long timeout;
4463 bool was_managetempaddr;
4464 bool had_prefixroute;
4468 if (!valid_lft || (prefered_lft > valid_lft))
4471 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4472 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4475 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4476 if (addrconf_finite_timeout(timeout)) {
4477 expires = jiffies_to_clock_t(timeout * HZ);
4478 valid_lft = timeout;
4479 flags = RTF_EXPIRES;
4483 ifa_flags |= IFA_F_PERMANENT;
4486 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4487 if (addrconf_finite_timeout(timeout)) {
4489 ifa_flags |= IFA_F_DEPRECATED;
4490 prefered_lft = timeout;
4493 spin_lock_bh(&ifp->lock);
4494 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4495 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4496 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4497 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4498 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4499 IFA_F_NOPREFIXROUTE);
4500 ifp->flags |= ifa_flags;
4501 ifp->tstamp = jiffies;
4502 ifp->valid_lft = valid_lft;
4503 ifp->prefered_lft = prefered_lft;
4505 spin_unlock_bh(&ifp->lock);
4506 if (!(ifp->flags&IFA_F_TENTATIVE))
4507 ipv6_ifa_notify(0, ifp);
4509 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4510 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4512 } else if (had_prefixroute) {
4513 enum cleanup_prefix_rt_t action;
4514 unsigned long rt_expires;
4516 write_lock_bh(&ifp->idev->lock);
4517 action = check_cleanup_prefix_route(ifp, &rt_expires);
4518 write_unlock_bh(&ifp->idev->lock);
4520 if (action != CLEANUP_PREFIX_RT_NOP) {
4521 cleanup_prefix_route(ifp, rt_expires,
4522 action == CLEANUP_PREFIX_RT_DEL);
4526 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4527 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4528 valid_lft = prefered_lft = 0;
4529 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4530 !was_managetempaddr, jiffies);
4533 addrconf_verify_rtnl();
4539 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4540 struct netlink_ext_ack *extack)
4542 struct net *net = sock_net(skb->sk);
4543 struct ifaddrmsg *ifm;
4544 struct nlattr *tb[IFA_MAX+1];
4545 struct in6_addr *pfx, *peer_pfx;
4546 struct inet6_ifaddr *ifa;
4547 struct net_device *dev;
4548 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4552 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4557 ifm = nlmsg_data(nlh);
4558 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4562 if (tb[IFA_CACHEINFO]) {
4563 struct ifa_cacheinfo *ci;
4565 ci = nla_data(tb[IFA_CACHEINFO]);
4566 valid_lft = ci->ifa_valid;
4567 preferred_lft = ci->ifa_prefered;
4569 preferred_lft = INFINITY_LIFE_TIME;
4570 valid_lft = INFINITY_LIFE_TIME;
4573 dev = __dev_get_by_index(net, ifm->ifa_index);
4577 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4579 /* We ignore other flags so far. */
4580 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4581 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4583 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4586 * It would be best to check for !NLM_F_CREATE here but
4587 * userspace already relies on not having to provide this.
4589 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4590 ifm->ifa_prefixlen, ifa_flags,
4591 preferred_lft, valid_lft);
4594 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4595 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4598 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4605 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4606 u8 scope, int ifindex)
4608 struct ifaddrmsg *ifm;
4610 ifm = nlmsg_data(nlh);
4611 ifm->ifa_family = AF_INET6;
4612 ifm->ifa_prefixlen = prefixlen;
4613 ifm->ifa_flags = flags;
4614 ifm->ifa_scope = scope;
4615 ifm->ifa_index = ifindex;
4618 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4619 unsigned long tstamp, u32 preferred, u32 valid)
4621 struct ifa_cacheinfo ci;
4623 ci.cstamp = cstamp_delta(cstamp);
4624 ci.tstamp = cstamp_delta(tstamp);
4625 ci.ifa_prefered = preferred;
4626 ci.ifa_valid = valid;
4628 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4631 static inline int rt_scope(int ifa_scope)
4633 if (ifa_scope & IFA_HOST)
4634 return RT_SCOPE_HOST;
4635 else if (ifa_scope & IFA_LINK)
4636 return RT_SCOPE_LINK;
4637 else if (ifa_scope & IFA_SITE)
4638 return RT_SCOPE_SITE;
4640 return RT_SCOPE_UNIVERSE;
4643 static inline int inet6_ifaddr_msgsize(void)
4645 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4646 + nla_total_size(16) /* IFA_LOCAL */
4647 + nla_total_size(16) /* IFA_ADDRESS */
4648 + nla_total_size(sizeof(struct ifa_cacheinfo))
4649 + nla_total_size(4) /* IFA_FLAGS */;
4652 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4653 u32 portid, u32 seq, int event, unsigned int flags)
4655 struct nlmsghdr *nlh;
4656 u32 preferred, valid;
4658 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4662 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4663 ifa->idev->dev->ifindex);
4665 if (!((ifa->flags&IFA_F_PERMANENT) &&
4666 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4667 preferred = ifa->prefered_lft;
4668 valid = ifa->valid_lft;
4669 if (preferred != INFINITY_LIFE_TIME) {
4670 long tval = (jiffies - ifa->tstamp)/HZ;
4671 if (preferred > tval)
4675 if (valid != INFINITY_LIFE_TIME) {
4683 preferred = INFINITY_LIFE_TIME;
4684 valid = INFINITY_LIFE_TIME;
4687 if (!ipv6_addr_any(&ifa->peer_addr)) {
4688 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4689 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4692 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4695 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4698 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4701 nlmsg_end(skb, nlh);
4705 nlmsg_cancel(skb, nlh);
4709 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4710 u32 portid, u32 seq, int event, u16 flags)
4712 struct nlmsghdr *nlh;
4713 u8 scope = RT_SCOPE_UNIVERSE;
4714 int ifindex = ifmca->idev->dev->ifindex;
4716 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4717 scope = RT_SCOPE_SITE;
4719 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4723 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4724 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4725 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4726 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4727 nlmsg_cancel(skb, nlh);
4731 nlmsg_end(skb, nlh);
4735 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4736 u32 portid, u32 seq, int event, unsigned int flags)
4738 struct nlmsghdr *nlh;
4739 u8 scope = RT_SCOPE_UNIVERSE;
4740 int ifindex = ifaca->aca_idev->dev->ifindex;
4742 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4743 scope = RT_SCOPE_SITE;
4745 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4749 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4750 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4751 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4752 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4753 nlmsg_cancel(skb, nlh);
4757 nlmsg_end(skb, nlh);
4767 /* called with rcu_read_lock() */
4768 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4769 struct netlink_callback *cb, enum addr_type_t type,
4770 int s_ip_idx, int *p_ip_idx)
4772 struct ifmcaddr6 *ifmca;
4773 struct ifacaddr6 *ifaca;
4775 int ip_idx = *p_ip_idx;
4777 read_lock_bh(&idev->lock);
4779 case UNICAST_ADDR: {
4780 struct inet6_ifaddr *ifa;
4782 /* unicast address incl. temp addr */
4783 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4784 if (++ip_idx < s_ip_idx)
4786 err = inet6_fill_ifaddr(skb, ifa,
4787 NETLINK_CB(cb->skb).portid,
4793 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4797 case MULTICAST_ADDR:
4798 /* multicast address */
4799 for (ifmca = idev->mc_list; ifmca;
4800 ifmca = ifmca->next, ip_idx++) {
4801 if (ip_idx < s_ip_idx)
4803 err = inet6_fill_ifmcaddr(skb, ifmca,
4804 NETLINK_CB(cb->skb).portid,
4813 /* anycast address */
4814 for (ifaca = idev->ac_list; ifaca;
4815 ifaca = ifaca->aca_next, ip_idx++) {
4816 if (ip_idx < s_ip_idx)
4818 err = inet6_fill_ifacaddr(skb, ifaca,
4819 NETLINK_CB(cb->skb).portid,
4830 read_unlock_bh(&idev->lock);
4835 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4836 enum addr_type_t type)
4838 struct net *net = sock_net(skb->sk);
4841 int s_idx, s_ip_idx;
4842 struct net_device *dev;
4843 struct inet6_dev *idev;
4844 struct hlist_head *head;
4847 s_idx = idx = cb->args[1];
4848 s_ip_idx = ip_idx = cb->args[2];
4851 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4852 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4854 head = &net->dev_index_head[h];
4855 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4858 if (h > s_h || idx > s_idx)
4861 idev = __in6_dev_get(dev);
4865 if (in6_dump_addrs(idev, skb, cb, type,
4866 s_ip_idx, &ip_idx) < 0)
4876 cb->args[2] = ip_idx;
4881 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4883 enum addr_type_t type = UNICAST_ADDR;
4885 return inet6_dump_addr(skb, cb, type);
4888 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4890 enum addr_type_t type = MULTICAST_ADDR;
4892 return inet6_dump_addr(skb, cb, type);
4896 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4898 enum addr_type_t type = ANYCAST_ADDR;
4900 return inet6_dump_addr(skb, cb, type);
4903 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4904 struct netlink_ext_ack *extack)
4906 struct net *net = sock_net(in_skb->sk);
4907 struct ifaddrmsg *ifm;
4908 struct nlattr *tb[IFA_MAX+1];
4909 struct in6_addr *addr = NULL, *peer;
4910 struct net_device *dev = NULL;
4911 struct inet6_ifaddr *ifa;
4912 struct sk_buff *skb;
4915 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4920 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4926 ifm = nlmsg_data(nlh);
4928 dev = __dev_get_by_index(net, ifm->ifa_index);
4930 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4932 err = -EADDRNOTAVAIL;
4936 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4942 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4943 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4945 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4946 WARN_ON(err == -EMSGSIZE);
4950 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4957 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4959 struct sk_buff *skb;
4960 struct net *net = dev_net(ifa->idev->dev);
4963 /* Don't send DELADDR notification for TENTATIVE address,
4964 * since NEWADDR notification is sent only after removing
4967 if (ifa->flags & IFA_F_TENTATIVE && event == RTM_DELADDR)
4970 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4974 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4976 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4977 WARN_ON(err == -EMSGSIZE);
4981 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4985 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4988 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4989 __s32 *array, int bytes)
4991 BUG_ON(bytes < (DEVCONF_MAX * 4));
4993 memset(array, 0, bytes);
4994 array[DEVCONF_FORWARDING] = cnf->forwarding;
4995 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4996 array[DEVCONF_MTU6] = cnf->mtu6;
4997 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4998 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4999 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5000 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5001 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5002 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5003 jiffies_to_msecs(cnf->rtr_solicit_interval);
5004 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5005 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5006 array[DEVCONF_RTR_SOLICIT_DELAY] =
5007 jiffies_to_msecs(cnf->rtr_solicit_delay);
5008 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5009 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5010 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5011 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5012 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5013 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5014 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5015 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5016 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5017 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5018 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5019 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5020 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5021 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5022 #ifdef CONFIG_IPV6_ROUTER_PREF
5023 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5024 array[DEVCONF_RTR_PROBE_INTERVAL] =
5025 jiffies_to_msecs(cnf->rtr_probe_interval);
5026 #ifdef CONFIG_IPV6_ROUTE_INFO
5027 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5028 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5031 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5032 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5033 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5034 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5035 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5037 #ifdef CONFIG_IPV6_MROUTE
5038 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5040 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5041 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5042 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5043 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5044 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5045 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5046 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5047 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5048 /* we omit DEVCONF_STABLE_SECRET for now */
5049 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5050 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5051 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5052 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5053 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5054 #ifdef CONFIG_IPV6_SEG6_HMAC
5055 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5057 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5058 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5059 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5062 static inline size_t inet6_ifla6_size(void)
5064 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5065 + nla_total_size(sizeof(struct ifla_cacheinfo))
5066 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5067 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5068 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5069 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5072 static inline size_t inet6_if_nlmsg_size(void)
5074 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5075 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5076 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5077 + nla_total_size(4) /* IFLA_MTU */
5078 + nla_total_size(4) /* IFLA_LINK */
5079 + nla_total_size(1) /* IFLA_OPERSTATE */
5080 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5083 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5087 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5090 /* Use put_unaligned() because stats may not be aligned for u64. */
5091 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5092 for (i = 1; i < ICMP6_MIB_MAX; i++)
5093 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5095 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5098 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5099 int bytes, size_t syncpoff)
5102 u64 buff[IPSTATS_MIB_MAX];
5103 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5107 memset(buff, 0, sizeof(buff));
5108 buff[0] = IPSTATS_MIB_MAX;
5110 for_each_possible_cpu(c) {
5111 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5112 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5115 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5116 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5119 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5123 case IFLA_INET6_STATS:
5124 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5125 offsetof(struct ipstats_mib, syncp));
5127 case IFLA_INET6_ICMP6STATS:
5128 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5133 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5134 u32 ext_filter_mask)
5137 struct ifla_cacheinfo ci;
5139 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5140 goto nla_put_failure;
5141 ci.max_reasm_len = IPV6_MAXPLEN;
5142 ci.tstamp = cstamp_delta(idev->tstamp);
5143 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5144 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5145 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5146 goto nla_put_failure;
5147 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5149 goto nla_put_failure;
5150 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5152 /* XXX - MC not implemented */
5154 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5157 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5159 goto nla_put_failure;
5160 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5162 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5164 goto nla_put_failure;
5165 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5167 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5169 goto nla_put_failure;
5171 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5172 goto nla_put_failure;
5174 read_lock_bh(&idev->lock);
5175 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5176 read_unlock_bh(&idev->lock);
5184 static size_t inet6_get_link_af_size(const struct net_device *dev,
5185 u32 ext_filter_mask)
5187 if (!__in6_dev_get(dev))
5190 return inet6_ifla6_size();
5193 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5194 u32 ext_filter_mask)
5196 struct inet6_dev *idev = __in6_dev_get(dev);
5201 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5207 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5209 struct inet6_ifaddr *ifp;
5210 struct net_device *dev = idev->dev;
5211 bool clear_token, update_rs = false;
5212 struct in6_addr ll_addr;
5218 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5220 if (!ipv6_accept_ra(idev))
5222 if (idev->cnf.rtr_solicits == 0)
5225 write_lock_bh(&idev->lock);
5227 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5228 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5230 write_unlock_bh(&idev->lock);
5232 clear_token = ipv6_addr_any(token);
5236 if (!idev->dead && (idev->if_flags & IF_READY) &&
5237 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5238 IFA_F_OPTIMISTIC)) {
5239 /* If we're not ready, then normal ifup will take care
5240 * of this. Otherwise, we need to request our rs here.
5242 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5247 write_lock_bh(&idev->lock);
5250 idev->if_flags |= IF_RS_SENT;
5251 idev->rs_interval = rfc3315_s14_backoff_init(
5252 idev->cnf.rtr_solicit_interval);
5253 idev->rs_probes = 1;
5254 addrconf_mod_rs_timer(idev, idev->rs_interval);
5257 /* Well, that's kinda nasty ... */
5258 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5259 spin_lock(&ifp->lock);
5260 if (ifp->tokenized) {
5262 ifp->prefered_lft = 0;
5264 spin_unlock(&ifp->lock);
5267 write_unlock_bh(&idev->lock);
5268 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5269 addrconf_verify_rtnl();
5273 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5274 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5275 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5278 static int inet6_validate_link_af(const struct net_device *dev,
5279 const struct nlattr *nla)
5281 struct nlattr *tb[IFLA_INET6_MAX + 1];
5283 if (dev && !__in6_dev_get(dev))
5284 return -EAFNOSUPPORT;
5286 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5290 static int check_addr_gen_mode(int mode)
5292 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5293 mode != IN6_ADDR_GEN_MODE_NONE &&
5294 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5295 mode != IN6_ADDR_GEN_MODE_RANDOM)
5300 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5303 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5304 !idev->cnf.stable_secret.initialized &&
5305 !net->ipv6.devconf_dflt->stable_secret.initialized)
5310 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5313 struct inet6_dev *idev = __in6_dev_get(dev);
5314 struct nlattr *tb[IFLA_INET6_MAX + 1];
5317 return -EAFNOSUPPORT;
5319 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5322 if (tb[IFLA_INET6_TOKEN]) {
5323 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5328 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5329 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5331 if (check_addr_gen_mode(mode) < 0 ||
5332 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5335 idev->cnf.addr_gen_mode = mode;
5342 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5343 u32 portid, u32 seq, int event, unsigned int flags)
5345 struct net_device *dev = idev->dev;
5346 struct ifinfomsg *hdr;
5347 struct nlmsghdr *nlh;
5350 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5354 hdr = nlmsg_data(nlh);
5355 hdr->ifi_family = AF_INET6;
5357 hdr->ifi_type = dev->type;
5358 hdr->ifi_index = dev->ifindex;
5359 hdr->ifi_flags = dev_get_flags(dev);
5360 hdr->ifi_change = 0;
5362 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5364 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5365 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5366 (dev->ifindex != dev_get_iflink(dev) &&
5367 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5368 nla_put_u8(skb, IFLA_OPERSTATE,
5369 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5370 goto nla_put_failure;
5371 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5373 goto nla_put_failure;
5375 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5376 goto nla_put_failure;
5378 nla_nest_end(skb, protoinfo);
5379 nlmsg_end(skb, nlh);
5383 nlmsg_cancel(skb, nlh);
5387 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5389 struct net *net = sock_net(skb->sk);
5392 struct net_device *dev;
5393 struct inet6_dev *idev;
5394 struct hlist_head *head;
5397 s_idx = cb->args[1];
5400 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5402 head = &net->dev_index_head[h];
5403 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5406 idev = __in6_dev_get(dev);
5409 if (inet6_fill_ifinfo(skb, idev,
5410 NETLINK_CB(cb->skb).portid,
5412 RTM_NEWLINK, NLM_F_MULTI) < 0)
5426 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5428 struct sk_buff *skb;
5429 struct net *net = dev_net(idev->dev);
5432 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5436 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5438 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5439 WARN_ON(err == -EMSGSIZE);
5443 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5447 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5450 static inline size_t inet6_prefix_nlmsg_size(void)
5452 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5453 + nla_total_size(sizeof(struct in6_addr))
5454 + nla_total_size(sizeof(struct prefix_cacheinfo));
5457 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5458 struct prefix_info *pinfo, u32 portid, u32 seq,
5459 int event, unsigned int flags)
5461 struct prefixmsg *pmsg;
5462 struct nlmsghdr *nlh;
5463 struct prefix_cacheinfo ci;
5465 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5469 pmsg = nlmsg_data(nlh);
5470 pmsg->prefix_family = AF_INET6;
5471 pmsg->prefix_pad1 = 0;
5472 pmsg->prefix_pad2 = 0;
5473 pmsg->prefix_ifindex = idev->dev->ifindex;
5474 pmsg->prefix_len = pinfo->prefix_len;
5475 pmsg->prefix_type = pinfo->type;
5476 pmsg->prefix_pad3 = 0;
5477 pmsg->prefix_flags = 0;
5479 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5480 if (pinfo->autoconf)
5481 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5483 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5484 goto nla_put_failure;
5485 ci.preferred_time = ntohl(pinfo->prefered);
5486 ci.valid_time = ntohl(pinfo->valid);
5487 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5488 goto nla_put_failure;
5489 nlmsg_end(skb, nlh);
5493 nlmsg_cancel(skb, nlh);
5497 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5498 struct prefix_info *pinfo)
5500 struct sk_buff *skb;
5501 struct net *net = dev_net(idev->dev);
5504 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5508 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5510 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5511 WARN_ON(err == -EMSGSIZE);
5515 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5519 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5522 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5524 struct net *net = dev_net(ifp->idev->dev);
5529 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5534 * If the address was optimistic
5535 * we inserted the route at the start of
5536 * our DAD process, so we don't need
5539 if (!(ifp->rt->rt6i_node))
5540 ip6_ins_rt(ifp->rt);
5541 if (ifp->idev->cnf.forwarding)
5542 addrconf_join_anycast(ifp);
5543 if (!ipv6_addr_any(&ifp->peer_addr))
5544 addrconf_prefix_route(&ifp->peer_addr, 128,
5545 ifp->idev->dev, 0, 0);
5548 if (ifp->idev->cnf.forwarding)
5549 addrconf_leave_anycast(ifp);
5550 addrconf_leave_solict(ifp->idev, &ifp->addr);
5551 if (!ipv6_addr_any(&ifp->peer_addr)) {
5552 struct rt6_info *rt;
5554 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5555 ifp->idev->dev, 0, 0);
5560 dst_hold(&ifp->rt->dst);
5561 ip6_del_rt(ifp->rt);
5563 rt_genid_bump_ipv6(net);
5566 atomic_inc(&net->ipv6.dev_addr_genid);
5569 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5572 if (likely(ifp->idev->dead == 0))
5573 __ipv6_ifa_notify(event, ifp);
5574 rcu_read_unlock_bh();
5577 #ifdef CONFIG_SYSCTL
5580 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5581 void __user *buffer, size_t *lenp, loff_t *ppos)
5583 int *valp = ctl->data;
5586 struct ctl_table lctl;
5590 * ctl->data points to idev->cnf.forwarding, we should
5591 * not modify it until we get the rtnl lock.
5596 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5599 ret = addrconf_fixup_forwarding(ctl, valp, val);
5606 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5607 void __user *buffer, size_t *lenp, loff_t *ppos)
5609 struct inet6_dev *idev = ctl->extra1;
5610 int min_mtu = IPV6_MIN_MTU;
5611 struct ctl_table lctl;
5614 lctl.extra1 = &min_mtu;
5615 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5617 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5620 static void dev_disable_change(struct inet6_dev *idev)
5622 struct netdev_notifier_info info;
5624 if (!idev || !idev->dev)
5627 netdev_notifier_info_init(&info, idev->dev);
5628 if (idev->cnf.disable_ipv6)
5629 addrconf_notify(NULL, NETDEV_DOWN, &info);
5631 addrconf_notify(NULL, NETDEV_UP, &info);
5634 static void addrconf_disable_change(struct net *net, __s32 newf)
5636 struct net_device *dev;
5637 struct inet6_dev *idev;
5639 for_each_netdev(net, dev) {
5640 idev = __in6_dev_get(dev);
5642 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5643 idev->cnf.disable_ipv6 = newf;
5645 dev_disable_change(idev);
5650 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5655 if (!rtnl_trylock())
5656 return restart_syscall();
5658 net = (struct net *)table->extra2;
5662 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5667 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5668 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5669 addrconf_disable_change(net, newf);
5670 } else if ((!newf) ^ (!old))
5671 dev_disable_change((struct inet6_dev *)table->extra1);
5678 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5679 void __user *buffer, size_t *lenp, loff_t *ppos)
5681 int *valp = ctl->data;
5684 struct ctl_table lctl;
5688 * ctl->data points to idev->cnf.disable_ipv6, we should
5689 * not modify it until we get the rtnl lock.
5694 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5697 ret = addrconf_disable_ipv6(ctl, valp, val);
5704 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5705 void __user *buffer, size_t *lenp, loff_t *ppos)
5707 int *valp = ctl->data;
5712 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5715 if (write && old != new) {
5716 struct net *net = ctl->extra2;
5718 if (!rtnl_trylock())
5719 return restart_syscall();
5721 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5722 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5723 NETCONFA_PROXY_NEIGH,
5724 NETCONFA_IFINDEX_DEFAULT,
5725 net->ipv6.devconf_dflt);
5726 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5727 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5728 NETCONFA_PROXY_NEIGH,
5729 NETCONFA_IFINDEX_ALL,
5730 net->ipv6.devconf_all);
5732 struct inet6_dev *idev = ctl->extra1;
5734 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5735 NETCONFA_PROXY_NEIGH,
5745 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5746 void __user *buffer, size_t *lenp,
5751 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5752 struct net *net = (struct net *)ctl->extra2;
5754 if (!rtnl_trylock())
5755 return restart_syscall();
5757 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5760 new_val = *((int *)ctl->data);
5762 if (check_addr_gen_mode(new_val) < 0) {
5767 /* request for default */
5768 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5769 ipv6_devconf_dflt.addr_gen_mode = new_val;
5771 /* request for individual net device */
5776 if (check_stable_privacy(idev, net, new_val) < 0) {
5781 if (idev->cnf.addr_gen_mode != new_val) {
5782 idev->cnf.addr_gen_mode = new_val;
5783 addrconf_dev_config(idev->dev);
5794 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5795 void __user *buffer, size_t *lenp,
5799 struct in6_addr addr;
5800 char str[IPV6_MAX_STRLEN];
5801 struct ctl_table lctl = *ctl;
5802 struct net *net = ctl->extra2;
5803 struct ipv6_stable_secret *secret = ctl->data;
5805 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5808 lctl.maxlen = IPV6_MAX_STRLEN;
5811 if (!rtnl_trylock())
5812 return restart_syscall();
5814 if (!write && !secret->initialized) {
5819 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5820 if (err >= sizeof(str)) {
5825 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5829 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5834 secret->initialized = true;
5835 secret->secret = addr;
5837 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5838 struct net_device *dev;
5840 for_each_netdev(net, dev) {
5841 struct inet6_dev *idev = __in6_dev_get(dev);
5844 idev->cnf.addr_gen_mode =
5845 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5849 struct inet6_dev *idev = ctl->extra1;
5851 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5861 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5863 void __user *buffer,
5867 int *valp = ctl->data;
5870 struct ctl_table lctl;
5873 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5874 * we should not modify it until we get the rtnl lock.
5879 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5882 ret = addrconf_fixup_linkdown(ctl, valp, val);
5889 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5893 rt->dst.flags |= DST_NOPOLICY;
5895 rt->dst.flags &= ~DST_NOPOLICY;
5900 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5902 struct inet6_ifaddr *ifa;
5904 read_lock_bh(&idev->lock);
5905 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5906 spin_lock(&ifa->lock);
5908 struct rt6_info *rt = ifa->rt;
5909 struct fib6_table *table = rt->rt6i_table;
5912 read_lock(&table->tb6_lock);
5913 addrconf_set_nopolicy(ifa->rt, val);
5914 if (rt->rt6i_pcpu) {
5915 for_each_possible_cpu(cpu) {
5916 struct rt6_info **rtp;
5918 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5919 addrconf_set_nopolicy(*rtp, val);
5922 read_unlock(&table->tb6_lock);
5924 spin_unlock(&ifa->lock);
5926 read_unlock_bh(&idev->lock);
5930 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5932 struct inet6_dev *idev;
5935 if (!rtnl_trylock())
5936 return restart_syscall();
5940 net = (struct net *)ctl->extra2;
5941 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5946 if (valp == &net->ipv6.devconf_all->disable_policy) {
5947 struct net_device *dev;
5949 for_each_netdev(net, dev) {
5950 idev = __in6_dev_get(dev);
5952 addrconf_disable_policy_idev(idev, val);
5955 idev = (struct inet6_dev *)ctl->extra1;
5956 addrconf_disable_policy_idev(idev, val);
5964 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
5965 void __user *buffer, size_t *lenp,
5968 int *valp = ctl->data;
5971 struct ctl_table lctl;
5976 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5978 if (write && (*valp != val))
5979 ret = addrconf_disable_policy(ctl, valp, val);
5987 static int minus_one = -1;
5988 static const int one = 1;
5989 static const int two_five_five = 255;
5991 static const struct ctl_table addrconf_sysctl[] = {
5993 .procname = "forwarding",
5994 .data = &ipv6_devconf.forwarding,
5995 .maxlen = sizeof(int),
5997 .proc_handler = addrconf_sysctl_forward,
6000 .procname = "hop_limit",
6001 .data = &ipv6_devconf.hop_limit,
6002 .maxlen = sizeof(int),
6004 .proc_handler = proc_dointvec_minmax,
6005 .extra1 = (void *)&one,
6006 .extra2 = (void *)&two_five_five,
6010 .data = &ipv6_devconf.mtu6,
6011 .maxlen = sizeof(int),
6013 .proc_handler = addrconf_sysctl_mtu,
6016 .procname = "accept_ra",
6017 .data = &ipv6_devconf.accept_ra,
6018 .maxlen = sizeof(int),
6020 .proc_handler = proc_dointvec,
6023 .procname = "accept_redirects",
6024 .data = &ipv6_devconf.accept_redirects,
6025 .maxlen = sizeof(int),
6027 .proc_handler = proc_dointvec,
6030 .procname = "autoconf",
6031 .data = &ipv6_devconf.autoconf,
6032 .maxlen = sizeof(int),
6034 .proc_handler = proc_dointvec,
6037 .procname = "dad_transmits",
6038 .data = &ipv6_devconf.dad_transmits,
6039 .maxlen = sizeof(int),
6041 .proc_handler = proc_dointvec,
6044 .procname = "router_solicitations",
6045 .data = &ipv6_devconf.rtr_solicits,
6046 .maxlen = sizeof(int),
6048 .proc_handler = proc_dointvec_minmax,
6049 .extra1 = &minus_one,
6052 .procname = "router_solicitation_interval",
6053 .data = &ipv6_devconf.rtr_solicit_interval,
6054 .maxlen = sizeof(int),
6056 .proc_handler = proc_dointvec_jiffies,
6059 .procname = "router_solicitation_max_interval",
6060 .data = &ipv6_devconf.rtr_solicit_max_interval,
6061 .maxlen = sizeof(int),
6063 .proc_handler = proc_dointvec_jiffies,
6066 .procname = "router_solicitation_delay",
6067 .data = &ipv6_devconf.rtr_solicit_delay,
6068 .maxlen = sizeof(int),
6070 .proc_handler = proc_dointvec_jiffies,
6073 .procname = "force_mld_version",
6074 .data = &ipv6_devconf.force_mld_version,
6075 .maxlen = sizeof(int),
6077 .proc_handler = proc_dointvec,
6080 .procname = "mldv1_unsolicited_report_interval",
6082 &ipv6_devconf.mldv1_unsolicited_report_interval,
6083 .maxlen = sizeof(int),
6085 .proc_handler = proc_dointvec_ms_jiffies,
6088 .procname = "mldv2_unsolicited_report_interval",
6090 &ipv6_devconf.mldv2_unsolicited_report_interval,
6091 .maxlen = sizeof(int),
6093 .proc_handler = proc_dointvec_ms_jiffies,
6096 .procname = "use_tempaddr",
6097 .data = &ipv6_devconf.use_tempaddr,
6098 .maxlen = sizeof(int),
6100 .proc_handler = proc_dointvec,
6103 .procname = "temp_valid_lft",
6104 .data = &ipv6_devconf.temp_valid_lft,
6105 .maxlen = sizeof(int),
6107 .proc_handler = proc_dointvec,
6110 .procname = "temp_prefered_lft",
6111 .data = &ipv6_devconf.temp_prefered_lft,
6112 .maxlen = sizeof(int),
6114 .proc_handler = proc_dointvec,
6117 .procname = "regen_max_retry",
6118 .data = &ipv6_devconf.regen_max_retry,
6119 .maxlen = sizeof(int),
6121 .proc_handler = proc_dointvec,
6124 .procname = "max_desync_factor",
6125 .data = &ipv6_devconf.max_desync_factor,
6126 .maxlen = sizeof(int),
6128 .proc_handler = proc_dointvec,
6131 .procname = "max_addresses",
6132 .data = &ipv6_devconf.max_addresses,
6133 .maxlen = sizeof(int),
6135 .proc_handler = proc_dointvec,
6138 .procname = "accept_ra_defrtr",
6139 .data = &ipv6_devconf.accept_ra_defrtr,
6140 .maxlen = sizeof(int),
6142 .proc_handler = proc_dointvec,
6145 .procname = "accept_ra_min_hop_limit",
6146 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6147 .maxlen = sizeof(int),
6149 .proc_handler = proc_dointvec,
6152 .procname = "accept_ra_pinfo",
6153 .data = &ipv6_devconf.accept_ra_pinfo,
6154 .maxlen = sizeof(int),
6156 .proc_handler = proc_dointvec,
6158 #ifdef CONFIG_IPV6_ROUTER_PREF
6160 .procname = "accept_ra_rtr_pref",
6161 .data = &ipv6_devconf.accept_ra_rtr_pref,
6162 .maxlen = sizeof(int),
6164 .proc_handler = proc_dointvec,
6167 .procname = "router_probe_interval",
6168 .data = &ipv6_devconf.rtr_probe_interval,
6169 .maxlen = sizeof(int),
6171 .proc_handler = proc_dointvec_jiffies,
6173 #ifdef CONFIG_IPV6_ROUTE_INFO
6175 .procname = "accept_ra_rt_info_min_plen",
6176 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6177 .maxlen = sizeof(int),
6179 .proc_handler = proc_dointvec,
6182 .procname = "accept_ra_rt_info_max_plen",
6183 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6184 .maxlen = sizeof(int),
6186 .proc_handler = proc_dointvec,
6191 .procname = "proxy_ndp",
6192 .data = &ipv6_devconf.proxy_ndp,
6193 .maxlen = sizeof(int),
6195 .proc_handler = addrconf_sysctl_proxy_ndp,
6198 .procname = "accept_source_route",
6199 .data = &ipv6_devconf.accept_source_route,
6200 .maxlen = sizeof(int),
6202 .proc_handler = proc_dointvec,
6204 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6206 .procname = "optimistic_dad",
6207 .data = &ipv6_devconf.optimistic_dad,
6208 .maxlen = sizeof(int),
6210 .proc_handler = proc_dointvec,
6213 .procname = "use_optimistic",
6214 .data = &ipv6_devconf.use_optimistic,
6215 .maxlen = sizeof(int),
6217 .proc_handler = proc_dointvec,
6220 #ifdef CONFIG_IPV6_MROUTE
6222 .procname = "mc_forwarding",
6223 .data = &ipv6_devconf.mc_forwarding,
6224 .maxlen = sizeof(int),
6226 .proc_handler = proc_dointvec,
6230 .procname = "disable_ipv6",
6231 .data = &ipv6_devconf.disable_ipv6,
6232 .maxlen = sizeof(int),
6234 .proc_handler = addrconf_sysctl_disable,
6237 .procname = "accept_dad",
6238 .data = &ipv6_devconf.accept_dad,
6239 .maxlen = sizeof(int),
6241 .proc_handler = proc_dointvec,
6244 .procname = "force_tllao",
6245 .data = &ipv6_devconf.force_tllao,
6246 .maxlen = sizeof(int),
6248 .proc_handler = proc_dointvec
6251 .procname = "ndisc_notify",
6252 .data = &ipv6_devconf.ndisc_notify,
6253 .maxlen = sizeof(int),
6255 .proc_handler = proc_dointvec
6258 .procname = "suppress_frag_ndisc",
6259 .data = &ipv6_devconf.suppress_frag_ndisc,
6260 .maxlen = sizeof(int),
6262 .proc_handler = proc_dointvec
6265 .procname = "accept_ra_from_local",
6266 .data = &ipv6_devconf.accept_ra_from_local,
6267 .maxlen = sizeof(int),
6269 .proc_handler = proc_dointvec,
6272 .procname = "accept_ra_mtu",
6273 .data = &ipv6_devconf.accept_ra_mtu,
6274 .maxlen = sizeof(int),
6276 .proc_handler = proc_dointvec,
6279 .procname = "stable_secret",
6280 .data = &ipv6_devconf.stable_secret,
6281 .maxlen = IPV6_MAX_STRLEN,
6283 .proc_handler = addrconf_sysctl_stable_secret,
6286 .procname = "use_oif_addrs_only",
6287 .data = &ipv6_devconf.use_oif_addrs_only,
6288 .maxlen = sizeof(int),
6290 .proc_handler = proc_dointvec,
6293 .procname = "ignore_routes_with_linkdown",
6294 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6295 .maxlen = sizeof(int),
6297 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6300 .procname = "drop_unicast_in_l2_multicast",
6301 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6302 .maxlen = sizeof(int),
6304 .proc_handler = proc_dointvec,
6307 .procname = "drop_unsolicited_na",
6308 .data = &ipv6_devconf.drop_unsolicited_na,
6309 .maxlen = sizeof(int),
6311 .proc_handler = proc_dointvec,
6314 .procname = "keep_addr_on_down",
6315 .data = &ipv6_devconf.keep_addr_on_down,
6316 .maxlen = sizeof(int),
6318 .proc_handler = proc_dointvec,
6322 .procname = "seg6_enabled",
6323 .data = &ipv6_devconf.seg6_enabled,
6324 .maxlen = sizeof(int),
6326 .proc_handler = proc_dointvec,
6328 #ifdef CONFIG_IPV6_SEG6_HMAC
6330 .procname = "seg6_require_hmac",
6331 .data = &ipv6_devconf.seg6_require_hmac,
6332 .maxlen = sizeof(int),
6334 .proc_handler = proc_dointvec,
6338 .procname = "enhanced_dad",
6339 .data = &ipv6_devconf.enhanced_dad,
6340 .maxlen = sizeof(int),
6342 .proc_handler = proc_dointvec,
6345 .procname = "addr_gen_mode",
6346 .data = &ipv6_devconf.addr_gen_mode,
6347 .maxlen = sizeof(int),
6349 .proc_handler = addrconf_sysctl_addr_gen_mode,
6352 .procname = "disable_policy",
6353 .data = &ipv6_devconf.disable_policy,
6354 .maxlen = sizeof(int),
6356 .proc_handler = addrconf_sysctl_disable_policy,
6363 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6364 struct inet6_dev *idev, struct ipv6_devconf *p)
6367 struct ctl_table *table;
6368 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6370 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6374 for (i = 0; table[i].data; i++) {
6375 table[i].data += (char *)p - (char *)&ipv6_devconf;
6376 /* If one of these is already set, then it is not safe to
6377 * overwrite either of them: this makes proc_dointvec_minmax
6380 if (!table[i].extra1 && !table[i].extra2) {
6381 table[i].extra1 = idev; /* embedded; no ref */
6382 table[i].extra2 = net;
6386 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6388 p->sysctl_header = register_net_sysctl(net, path, table);
6389 if (!p->sysctl_header)
6392 if (!strcmp(dev_name, "all"))
6393 ifindex = NETCONFA_IFINDEX_ALL;
6394 else if (!strcmp(dev_name, "default"))
6395 ifindex = NETCONFA_IFINDEX_DEFAULT;
6397 ifindex = idev->dev->ifindex;
6398 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6408 static void __addrconf_sysctl_unregister(struct net *net,
6409 struct ipv6_devconf *p, int ifindex)
6411 struct ctl_table *table;
6413 if (!p->sysctl_header)
6416 table = p->sysctl_header->ctl_table_arg;
6417 unregister_net_sysctl_table(p->sysctl_header);
6418 p->sysctl_header = NULL;
6421 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6424 static int addrconf_sysctl_register(struct inet6_dev *idev)
6428 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6431 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6432 &ndisc_ifinfo_sysctl_change);
6435 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6438 neigh_sysctl_unregister(idev->nd_parms);
6443 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6445 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6446 idev->dev->ifindex);
6447 neigh_sysctl_unregister(idev->nd_parms);
6453 static int __net_init addrconf_init_net(struct net *net)
6456 struct ipv6_devconf *all, *dflt;
6458 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6462 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6464 goto err_alloc_dflt;
6466 /* these will be inherited by all namespaces */
6467 dflt->autoconf = ipv6_defaults.autoconf;
6468 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6470 dflt->stable_secret.initialized = false;
6471 all->stable_secret.initialized = false;
6473 net->ipv6.devconf_all = all;
6474 net->ipv6.devconf_dflt = dflt;
6476 #ifdef CONFIG_SYSCTL
6477 err = __addrconf_sysctl_register(net, "all", NULL, all);
6481 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6487 #ifdef CONFIG_SYSCTL
6489 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6499 static void __net_exit addrconf_exit_net(struct net *net)
6501 #ifdef CONFIG_SYSCTL
6502 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6503 NETCONFA_IFINDEX_DEFAULT);
6504 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6505 NETCONFA_IFINDEX_ALL);
6507 kfree(net->ipv6.devconf_dflt);
6508 kfree(net->ipv6.devconf_all);
6511 static struct pernet_operations addrconf_ops = {
6512 .init = addrconf_init_net,
6513 .exit = addrconf_exit_net,
6516 static struct rtnl_af_ops inet6_ops __read_mostly = {
6518 .fill_link_af = inet6_fill_link_af,
6519 .get_link_af_size = inet6_get_link_af_size,
6520 .validate_link_af = inet6_validate_link_af,
6521 .set_link_af = inet6_set_link_af,
6525 * Init / cleanup code
6528 int __init addrconf_init(void)
6530 struct inet6_dev *idev;
6533 err = ipv6_addr_label_init();
6535 pr_crit("%s: cannot initialize default policy table: %d\n",
6540 err = register_pernet_subsys(&addrconf_ops);
6544 addrconf_wq = create_workqueue("ipv6_addrconf");
6550 /* The addrconf netdev notifier requires that loopback_dev
6551 * has it's ipv6 private information allocated and setup
6552 * before it can bring up and give link-local addresses
6553 * to other devices which are up.
6555 * Unfortunately, loopback_dev is not necessarily the first
6556 * entry in the global dev_base list of net devices. In fact,
6557 * it is likely to be the very last entry on that list.
6558 * So this causes the notifier registry below to try and
6559 * give link-local addresses to all devices besides loopback_dev
6560 * first, then loopback_dev, which cases all the non-loopback_dev
6561 * devices to fail to get a link-local address.
6563 * So, as a temporary fix, allocate the ipv6 structure for
6564 * loopback_dev first by hand.
6565 * Longer term, all of the dependencies ipv6 has upon the loopback
6566 * device and it being up should be removed.
6569 idev = ipv6_add_dev(init_net.loopback_dev);
6572 err = PTR_ERR(idev);
6576 ip6_route_init_special_entries();
6578 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6579 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6581 register_netdevice_notifier(&ipv6_dev_notf);
6585 rtnl_af_register(&inet6_ops);
6587 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6592 /* Only the first call to __rtnl_register can fail */
6593 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6594 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6595 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6596 inet6_dump_ifaddr, NULL);
6597 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6598 inet6_dump_ifmcaddr, NULL);
6599 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6600 inet6_dump_ifacaddr, NULL);
6601 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6602 inet6_netconf_dump_devconf, NULL);
6604 ipv6_addr_label_rtnl_register();
6608 rtnl_af_unregister(&inet6_ops);
6609 unregister_netdevice_notifier(&ipv6_dev_notf);
6611 destroy_workqueue(addrconf_wq);
6613 unregister_pernet_subsys(&addrconf_ops);
6615 ipv6_addr_label_cleanup();
6620 void addrconf_cleanup(void)
6622 struct net_device *dev;
6625 unregister_netdevice_notifier(&ipv6_dev_notf);
6626 unregister_pernet_subsys(&addrconf_ops);
6627 ipv6_addr_label_cleanup();
6631 __rtnl_af_unregister(&inet6_ops);
6633 /* clean dev list */
6634 for_each_netdev(&init_net, dev) {
6635 if (__in6_dev_get(dev) == NULL)
6637 addrconf_ifdown(dev, 1);
6639 addrconf_ifdown(init_net.loopback_dev, 2);
6644 spin_lock_bh(&addrconf_hash_lock);
6645 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6646 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6647 spin_unlock_bh(&addrconf_hash_lock);
6648 cancel_delayed_work(&addr_chk_work);
6651 destroy_workqueue(addrconf_wq);