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,
336 if (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. */
429 refcount_set(&ndev->refcnt, 1);
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;
966 struct in6_validator_info i6vi;
969 int addr_type = ipv6_addr_type(addr);
971 if (addr_type == IPV6_ADDR_ANY ||
972 addr_type & IPV6_ADDR_MULTICAST ||
973 (!(idev->dev->flags & IFF_LOOPBACK) &&
974 addr_type & IPV6_ADDR_LOOPBACK))
975 return ERR_PTR(-EADDRNOTAVAIL);
982 err = -ENODEV; /*XXX*/
986 if (idev->cnf.disable_ipv6) {
991 i6vi.i6vi_addr = *addr;
992 i6vi.i6vi_dev = idev;
993 rcu_read_unlock_bh();
995 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
998 err = notifier_to_errno(err);
1002 spin_lock(&addrconf_hash_lock);
1004 /* Ignore adding duplicate addresses on an interface */
1005 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
1006 ADBG("ipv6_add_addr: already assigned\n");
1011 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
1014 ADBG("ipv6_add_addr: malloc failed\n");
1019 rt = addrconf_dst_alloc(idev, addr, false);
1025 if (net->ipv6.devconf_all->disable_policy ||
1026 idev->cnf.disable_policy)
1027 rt->dst.flags |= DST_NOPOLICY;
1029 neigh_parms_data_state_setall(idev->nd_parms);
1033 ifa->peer_addr = *peer_addr;
1035 spin_lock_init(&ifa->lock);
1036 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1037 INIT_HLIST_NODE(&ifa->addr_lst);
1039 ifa->prefix_len = pfxlen;
1041 /* No need to add the TENTATIVE flag for addresses with NODAD */
1042 if (!(flags & IFA_F_NODAD))
1043 ifa->flags |= IFA_F_TENTATIVE;
1044 ifa->valid_lft = valid_lft;
1045 ifa->prefered_lft = prefered_lft;
1046 ifa->cstamp = ifa->tstamp = jiffies;
1047 ifa->tokenized = false;
1053 refcount_set(&ifa->refcnt, 1);
1055 /* Add to big hash table */
1056 hash = inet6_addr_hash(addr);
1058 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1059 spin_unlock(&addrconf_hash_lock);
1061 write_lock(&idev->lock);
1062 /* Add to inet6_dev unicast addr list. */
1063 ipv6_link_dev_addr(idev, ifa);
1065 if (ifa->flags&IFA_F_TEMPORARY) {
1066 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1071 write_unlock(&idev->lock);
1073 rcu_read_unlock_bh();
1075 if (likely(err == 0))
1076 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1085 spin_unlock(&addrconf_hash_lock);
1089 enum cleanup_prefix_rt_t {
1090 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1091 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1092 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1096 * Check, whether the prefix for ifp would still need a prefix route
1097 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1100 * 1) we don't purge prefix if address was not permanent.
1101 * prefix is managed by its own lifetime.
1102 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1103 * 3) if there are no addresses, delete prefix.
1104 * 4) if there are still other permanent address(es),
1105 * corresponding prefix is still permanent.
1106 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1107 * don't purge the prefix, assume user space is managing it.
1108 * 6) otherwise, update prefix lifetime to the
1109 * longest valid lifetime among the corresponding
1110 * addresses on the device.
1111 * Note: subsequent RA will update lifetime.
1113 static enum cleanup_prefix_rt_t
1114 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1116 struct inet6_ifaddr *ifa;
1117 struct inet6_dev *idev = ifp->idev;
1118 unsigned long lifetime;
1119 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1123 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1126 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1129 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1130 return CLEANUP_PREFIX_RT_NOP;
1132 action = CLEANUP_PREFIX_RT_EXPIRE;
1134 spin_lock(&ifa->lock);
1136 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1138 * Note: Because this address is
1139 * not permanent, lifetime <
1140 * LONG_MAX / HZ here.
1142 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1143 *expires = ifa->tstamp + lifetime * HZ;
1144 spin_unlock(&ifa->lock);
1151 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1153 struct rt6_info *rt;
1155 rt = addrconf_get_prefix_route(&ifp->addr,
1158 0, RTF_GATEWAY | RTF_DEFAULT);
1163 if (!(rt->rt6i_flags & RTF_EXPIRES))
1164 rt6_set_expires(rt, expires);
1171 /* This function wants to get referenced ifp and releases it before return */
1173 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1176 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1177 unsigned long expires;
1181 spin_lock_bh(&ifp->lock);
1183 ifp->state = INET6_IFADDR_STATE_DEAD;
1184 spin_unlock_bh(&ifp->lock);
1186 if (state == INET6_IFADDR_STATE_DEAD)
1189 spin_lock_bh(&addrconf_hash_lock);
1190 hlist_del_init_rcu(&ifp->addr_lst);
1191 spin_unlock_bh(&addrconf_hash_lock);
1193 write_lock_bh(&ifp->idev->lock);
1195 if (ifp->flags&IFA_F_TEMPORARY) {
1196 list_del(&ifp->tmp_list);
1198 in6_ifa_put(ifp->ifpub);
1204 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1205 action = check_cleanup_prefix_route(ifp, &expires);
1207 list_del_init(&ifp->if_list);
1210 write_unlock_bh(&ifp->idev->lock);
1212 addrconf_del_dad_work(ifp);
1214 ipv6_ifa_notify(RTM_DELADDR, ifp);
1216 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1218 if (action != CLEANUP_PREFIX_RT_NOP) {
1219 cleanup_prefix_route(ifp, expires,
1220 action == CLEANUP_PREFIX_RT_DEL);
1223 /* clean up prefsrc entries */
1224 rt6_remove_prefsrc(ifp);
1229 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1231 struct inet6_dev *idev = ifp->idev;
1232 struct in6_addr addr, *tmpaddr;
1233 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1234 unsigned long regen_advance;
1238 unsigned long now = jiffies;
1239 long max_desync_factor;
1240 s32 cnf_temp_preferred_lft;
1242 write_lock_bh(&idev->lock);
1244 spin_lock_bh(&ift->lock);
1245 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1246 spin_unlock_bh(&ift->lock);
1253 if (idev->cnf.use_tempaddr <= 0) {
1254 write_unlock_bh(&idev->lock);
1255 pr_info("%s: use_tempaddr is disabled\n", __func__);
1260 spin_lock_bh(&ifp->lock);
1261 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1262 idev->cnf.use_tempaddr = -1; /*XXX*/
1263 spin_unlock_bh(&ifp->lock);
1264 write_unlock_bh(&idev->lock);
1265 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1272 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1273 ipv6_try_regen_rndid(idev, tmpaddr);
1274 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1275 age = (now - ifp->tstamp) / HZ;
1277 regen_advance = idev->cnf.regen_max_retry *
1278 idev->cnf.dad_transmits *
1279 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1281 /* recalculate max_desync_factor each time and update
1282 * idev->desync_factor if it's larger
1284 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1285 max_desync_factor = min_t(__u32,
1286 idev->cnf.max_desync_factor,
1287 cnf_temp_preferred_lft - regen_advance);
1289 if (unlikely(idev->desync_factor > max_desync_factor)) {
1290 if (max_desync_factor > 0) {
1291 get_random_bytes(&idev->desync_factor,
1292 sizeof(idev->desync_factor));
1293 idev->desync_factor %= max_desync_factor;
1295 idev->desync_factor = 0;
1299 tmp_valid_lft = min_t(__u32,
1301 idev->cnf.temp_valid_lft + age);
1302 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1303 idev->desync_factor;
1304 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1305 tmp_plen = ifp->prefix_len;
1306 tmp_tstamp = ifp->tstamp;
1307 spin_unlock_bh(&ifp->lock);
1309 write_unlock_bh(&idev->lock);
1311 /* A temporary address is created only if this calculated Preferred
1312 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1313 * an implementation must not create a temporary address with a zero
1314 * Preferred Lifetime.
1315 * Use age calculation as in addrconf_verify to avoid unnecessary
1316 * temporary addresses being generated.
1318 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1319 if (tmp_prefered_lft <= regen_advance + age) {
1326 addr_flags = IFA_F_TEMPORARY;
1327 /* set in addrconf_prefix_rcv() */
1328 if (ifp->flags & IFA_F_OPTIMISTIC)
1329 addr_flags |= IFA_F_OPTIMISTIC;
1331 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1332 ipv6_addr_scope(&addr), addr_flags,
1333 tmp_valid_lft, tmp_prefered_lft);
1337 pr_info("%s: retry temporary address regeneration\n", __func__);
1339 write_lock_bh(&idev->lock);
1343 spin_lock_bh(&ift->lock);
1346 ift->tstamp = tmp_tstamp;
1347 spin_unlock_bh(&ift->lock);
1349 addrconf_dad_start(ift);
1357 * Choose an appropriate source address (RFC3484)
1360 IPV6_SADDR_RULE_INIT = 0,
1361 IPV6_SADDR_RULE_LOCAL,
1362 IPV6_SADDR_RULE_SCOPE,
1363 IPV6_SADDR_RULE_PREFERRED,
1364 #ifdef CONFIG_IPV6_MIP6
1365 IPV6_SADDR_RULE_HOA,
1367 IPV6_SADDR_RULE_OIF,
1368 IPV6_SADDR_RULE_LABEL,
1369 IPV6_SADDR_RULE_PRIVACY,
1370 IPV6_SADDR_RULE_ORCHID,
1371 IPV6_SADDR_RULE_PREFIX,
1372 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1373 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1378 struct ipv6_saddr_score {
1381 struct inet6_ifaddr *ifa;
1382 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1387 struct ipv6_saddr_dst {
1388 const struct in6_addr *addr;
1395 static inline int ipv6_saddr_preferred(int type)
1397 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1402 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1404 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1405 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1411 static int ipv6_get_saddr_eval(struct net *net,
1412 struct ipv6_saddr_score *score,
1413 struct ipv6_saddr_dst *dst,
1418 if (i <= score->rule) {
1420 case IPV6_SADDR_RULE_SCOPE:
1421 ret = score->scopedist;
1423 case IPV6_SADDR_RULE_PREFIX:
1424 ret = score->matchlen;
1427 ret = !!test_bit(i, score->scorebits);
1433 case IPV6_SADDR_RULE_INIT:
1434 /* Rule 0: remember if hiscore is not ready yet */
1437 case IPV6_SADDR_RULE_LOCAL:
1438 /* Rule 1: Prefer same address */
1439 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1441 case IPV6_SADDR_RULE_SCOPE:
1442 /* Rule 2: Prefer appropriate scope
1447 * ---+--+-+---> scope
1449 * | d is scope of the destination.
1451 * | \ <- smaller scope is better if
1452 * B-15 | \ if scope is enough for destination.
1453 * | ret = B - scope (-1 <= scope >= d <= 15).
1455 * |/ <- greater is better
1456 * -C / if scope is not enough for destination.
1457 * /| ret = scope - C (-1 <= d < scope <= 15).
1459 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1460 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1461 * Assume B = 0 and we get C > 29.
1463 ret = __ipv6_addr_src_scope(score->addr_type);
1464 if (ret >= dst->scope)
1467 ret -= 128; /* 30 is enough */
1468 score->scopedist = ret;
1470 case IPV6_SADDR_RULE_PREFERRED:
1472 /* Rule 3: Avoid deprecated and optimistic addresses */
1473 u8 avoid = IFA_F_DEPRECATED;
1475 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1476 avoid |= IFA_F_OPTIMISTIC;
1477 ret = ipv6_saddr_preferred(score->addr_type) ||
1478 !(score->ifa->flags & avoid);
1481 #ifdef CONFIG_IPV6_MIP6
1482 case IPV6_SADDR_RULE_HOA:
1484 /* Rule 4: Prefer home address */
1485 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1486 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1490 case IPV6_SADDR_RULE_OIF:
1491 /* Rule 5: Prefer outgoing interface */
1492 ret = (!dst->ifindex ||
1493 dst->ifindex == score->ifa->idev->dev->ifindex);
1495 case IPV6_SADDR_RULE_LABEL:
1496 /* Rule 6: Prefer matching label */
1497 ret = ipv6_addr_label(net,
1498 &score->ifa->addr, score->addr_type,
1499 score->ifa->idev->dev->ifindex) == dst->label;
1501 case IPV6_SADDR_RULE_PRIVACY:
1503 /* Rule 7: Prefer public address
1504 * Note: prefer temporary address if use_tempaddr >= 2
1506 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1507 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1508 score->ifa->idev->cnf.use_tempaddr >= 2;
1509 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1512 case IPV6_SADDR_RULE_ORCHID:
1513 /* Rule 8-: Prefer ORCHID vs ORCHID or
1514 * non-ORCHID vs non-ORCHID
1516 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1517 ipv6_addr_orchid(dst->addr));
1519 case IPV6_SADDR_RULE_PREFIX:
1520 /* Rule 8: Use longest matching prefix */
1521 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1522 if (ret > score->ifa->prefix_len)
1523 ret = score->ifa->prefix_len;
1524 score->matchlen = ret;
1526 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1527 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1528 /* Optimistic addresses still have lower precedence than other
1529 * preferred addresses.
1531 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1539 __set_bit(i, score->scorebits);
1545 static int __ipv6_dev_get_saddr(struct net *net,
1546 struct ipv6_saddr_dst *dst,
1547 struct inet6_dev *idev,
1548 struct ipv6_saddr_score *scores,
1551 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1553 read_lock_bh(&idev->lock);
1554 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1558 * - Tentative Address (RFC2462 section 5.4)
1559 * - A tentative address is not considered
1560 * "assigned to an interface" in the traditional
1561 * sense, unless it is also flagged as optimistic.
1562 * - Candidate Source Address (section 4)
1563 * - In any case, anycast addresses, multicast
1564 * addresses, and the unspecified address MUST
1565 * NOT be included in a candidate set.
1567 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1568 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1571 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1573 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1574 score->addr_type & IPV6_ADDR_MULTICAST)) {
1575 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1581 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1583 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1584 int minihiscore, miniscore;
1586 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1587 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1589 if (minihiscore > miniscore) {
1590 if (i == IPV6_SADDR_RULE_SCOPE &&
1591 score->scopedist > 0) {
1594 * each remaining entry
1595 * has too small (not enough)
1596 * scope, because ifa entries
1597 * are sorted by their scope
1603 } else if (minihiscore < miniscore) {
1605 in6_ifa_put(hiscore->ifa);
1607 in6_ifa_hold(score->ifa);
1609 swap(hiscore, score);
1610 hiscore_idx = 1 - hiscore_idx;
1612 /* restore our iterator */
1613 score->ifa = hiscore->ifa;
1620 read_unlock_bh(&idev->lock);
1624 static int ipv6_get_saddr_master(struct net *net,
1625 const struct net_device *dst_dev,
1626 const struct net_device *master,
1627 struct ipv6_saddr_dst *dst,
1628 struct ipv6_saddr_score *scores,
1631 struct inet6_dev *idev;
1633 idev = __in6_dev_get(dst_dev);
1635 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1636 scores, hiscore_idx);
1638 idev = __in6_dev_get(master);
1640 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1641 scores, hiscore_idx);
1646 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1647 const struct in6_addr *daddr, unsigned int prefs,
1648 struct in6_addr *saddr)
1650 struct ipv6_saddr_score scores[2], *hiscore;
1651 struct ipv6_saddr_dst dst;
1652 struct inet6_dev *idev;
1653 struct net_device *dev;
1655 bool use_oif_addr = false;
1656 int hiscore_idx = 0;
1658 dst_type = __ipv6_addr_type(daddr);
1660 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1661 dst.scope = __ipv6_addr_src_scope(dst_type);
1662 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1665 scores[hiscore_idx].rule = -1;
1666 scores[hiscore_idx].ifa = NULL;
1670 /* Candidate Source Address (section 4)
1671 * - multicast and link-local destination address,
1672 * the set of candidate source address MUST only
1673 * include addresses assigned to interfaces
1674 * belonging to the same link as the outgoing
1676 * (- For site-local destination addresses, the
1677 * set of candidate source addresses MUST only
1678 * include addresses assigned to interfaces
1679 * belonging to the same site as the outgoing
1681 * - "It is RECOMMENDED that the candidate source addresses
1682 * be the set of unicast addresses assigned to the
1683 * interface that will be used to send to the destination
1684 * (the 'outgoing' interface)." (RFC 6724)
1687 idev = __in6_dev_get(dst_dev);
1688 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1689 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1690 (idev && idev->cnf.use_oif_addrs_only)) {
1691 use_oif_addr = true;
1697 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1699 const struct net_device *master;
1702 /* if dst_dev exists and is enslaved to an L3 device, then
1703 * prefer addresses from dst_dev and then the master over
1704 * any other enslaved devices in the L3 domain.
1706 master = l3mdev_master_dev_rcu(dst_dev);
1708 master_idx = master->ifindex;
1710 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1712 scores, hiscore_idx);
1714 if (scores[hiscore_idx].ifa)
1718 for_each_netdev_rcu(net, dev) {
1719 /* only consider addresses on devices in the
1722 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1724 idev = __in6_dev_get(dev);
1727 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1734 hiscore = &scores[hiscore_idx];
1736 return -EADDRNOTAVAIL;
1738 *saddr = hiscore->ifa->addr;
1739 in6_ifa_put(hiscore->ifa);
1742 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1744 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1747 struct inet6_ifaddr *ifp;
1748 int err = -EADDRNOTAVAIL;
1750 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1751 if (ifp->scope > IFA_LINK)
1753 if (ifp->scope == IFA_LINK &&
1754 !(ifp->flags & banned_flags)) {
1763 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1766 struct inet6_dev *idev;
1767 int err = -EADDRNOTAVAIL;
1770 idev = __in6_dev_get(dev);
1772 read_lock_bh(&idev->lock);
1773 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1774 read_unlock_bh(&idev->lock);
1780 static int ipv6_count_addresses(struct inet6_dev *idev)
1783 struct inet6_ifaddr *ifp;
1785 read_lock_bh(&idev->lock);
1786 list_for_each_entry(ifp, &idev->addr_list, if_list)
1788 read_unlock_bh(&idev->lock);
1792 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1793 const struct net_device *dev, int strict)
1795 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1797 EXPORT_SYMBOL(ipv6_chk_addr);
1799 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1800 const struct net_device *dev, int strict,
1803 struct inet6_ifaddr *ifp;
1804 unsigned int hash = inet6_addr_hash(addr);
1808 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1809 if (!net_eq(dev_net(ifp->idev->dev), net))
1811 /* Decouple optimistic from tentative for evaluation here.
1812 * Ban optimistic addresses explicitly, when required.
1814 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1815 ? (ifp->flags&~IFA_F_TENTATIVE)
1817 if (ipv6_addr_equal(&ifp->addr, addr) &&
1818 !(ifp_flags&banned_flags) &&
1819 (!dev || ifp->idev->dev == dev ||
1820 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1821 rcu_read_unlock_bh();
1826 rcu_read_unlock_bh();
1829 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1831 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1832 struct net_device *dev)
1834 unsigned int hash = inet6_addr_hash(addr);
1835 struct inet6_ifaddr *ifp;
1837 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1838 if (!net_eq(dev_net(ifp->idev->dev), net))
1840 if (ipv6_addr_equal(&ifp->addr, addr)) {
1841 if (!dev || ifp->idev->dev == dev)
1848 /* Compares an address/prefix_len with addresses on device @dev.
1849 * If one is found it returns true.
1851 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1852 const unsigned int prefix_len, struct net_device *dev)
1854 struct inet6_dev *idev;
1855 struct inet6_ifaddr *ifa;
1859 idev = __in6_dev_get(dev);
1861 read_lock_bh(&idev->lock);
1862 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1863 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1867 read_unlock_bh(&idev->lock);
1873 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1875 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1877 struct inet6_dev *idev;
1878 struct inet6_ifaddr *ifa;
1883 idev = __in6_dev_get(dev);
1885 read_lock_bh(&idev->lock);
1886 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1887 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1892 read_unlock_bh(&idev->lock);
1897 EXPORT_SYMBOL(ipv6_chk_prefix);
1899 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1900 struct net_device *dev, int strict)
1902 struct inet6_ifaddr *ifp, *result = NULL;
1903 unsigned int hash = inet6_addr_hash(addr);
1906 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1907 if (!net_eq(dev_net(ifp->idev->dev), net))
1909 if (ipv6_addr_equal(&ifp->addr, addr)) {
1910 if (!dev || ifp->idev->dev == dev ||
1911 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1918 rcu_read_unlock_bh();
1923 /* Gets referenced address, destroys ifaddr */
1925 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1928 ifp->flags |= IFA_F_DADFAILED;
1930 if (ifp->flags&IFA_F_TEMPORARY) {
1931 struct inet6_ifaddr *ifpub;
1932 spin_lock_bh(&ifp->lock);
1935 in6_ifa_hold(ifpub);
1936 spin_unlock_bh(&ifp->lock);
1937 ipv6_create_tempaddr(ifpub, ifp);
1940 spin_unlock_bh(&ifp->lock);
1943 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1944 spin_lock_bh(&ifp->lock);
1945 addrconf_del_dad_work(ifp);
1946 ifp->flags |= IFA_F_TENTATIVE;
1947 spin_unlock_bh(&ifp->lock);
1949 ipv6_ifa_notify(0, ifp);
1956 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1960 spin_lock_bh(&ifp->lock);
1961 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1962 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1965 spin_unlock_bh(&ifp->lock);
1970 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1972 struct inet6_dev *idev = ifp->idev;
1973 struct net *net = dev_net(ifp->idev->dev);
1975 if (addrconf_dad_end(ifp)) {
1980 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1981 ifp->idev->dev->name, &ifp->addr);
1983 spin_lock_bh(&ifp->lock);
1985 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1986 int scope = ifp->scope;
1987 u32 flags = ifp->flags;
1988 struct in6_addr new_addr;
1989 struct inet6_ifaddr *ifp2;
1990 u32 valid_lft, preferred_lft;
1991 int pfxlen = ifp->prefix_len;
1992 int retries = ifp->stable_privacy_retry + 1;
1994 if (retries > net->ipv6.sysctl.idgen_retries) {
1995 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1996 ifp->idev->dev->name);
2000 new_addr = ifp->addr;
2001 if (ipv6_generate_stable_address(&new_addr, retries,
2005 valid_lft = ifp->valid_lft;
2006 preferred_lft = ifp->prefered_lft;
2008 spin_unlock_bh(&ifp->lock);
2010 if (idev->cnf.max_addresses &&
2011 ipv6_count_addresses(idev) >=
2012 idev->cnf.max_addresses)
2015 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2016 ifp->idev->dev->name);
2018 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2019 scope, flags, valid_lft,
2024 spin_lock_bh(&ifp2->lock);
2025 ifp2->stable_privacy_retry = retries;
2026 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2027 spin_unlock_bh(&ifp2->lock);
2029 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2032 spin_lock_bh(&ifp->lock);
2036 /* transition from _POSTDAD to _ERRDAD */
2037 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2038 spin_unlock_bh(&ifp->lock);
2040 addrconf_mod_dad_work(ifp, 0);
2044 /* Join to solicited addr multicast group.
2045 * caller must hold RTNL */
2046 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2048 struct in6_addr maddr;
2050 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2053 addrconf_addr_solict_mult(addr, &maddr);
2054 ipv6_dev_mc_inc(dev, &maddr);
2057 /* caller must hold RTNL */
2058 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2060 struct in6_addr maddr;
2062 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2065 addrconf_addr_solict_mult(addr, &maddr);
2066 __ipv6_dev_mc_dec(idev, &maddr);
2069 /* caller must hold RTNL */
2070 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2072 struct in6_addr addr;
2074 if (ifp->prefix_len >= 127) /* RFC 6164 */
2076 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2077 if (ipv6_addr_any(&addr))
2079 __ipv6_dev_ac_inc(ifp->idev, &addr);
2082 /* caller must hold RTNL */
2083 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2085 struct in6_addr addr;
2087 if (ifp->prefix_len >= 127) /* RFC 6164 */
2089 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2090 if (ipv6_addr_any(&addr))
2092 __ipv6_dev_ac_dec(ifp->idev, &addr);
2095 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2097 switch (dev->addr_len) {
2099 memcpy(eui, dev->dev_addr, 3);
2102 memcpy(eui + 5, dev->dev_addr + 3, 3);
2104 case EUI64_ADDR_LEN:
2105 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2115 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2117 union fwnet_hwaddr *ha;
2119 if (dev->addr_len != FWNET_ALEN)
2122 ha = (union fwnet_hwaddr *)dev->dev_addr;
2124 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2129 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2131 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2132 if (dev->addr_len != ARCNET_ALEN)
2135 eui[7] = *(u8 *)dev->dev_addr;
2139 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2141 if (dev->addr_len != INFINIBAND_ALEN)
2143 memcpy(eui, dev->dev_addr + 12, 8);
2148 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2152 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2153 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2154 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2155 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2156 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2157 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2161 memcpy(eui + 4, &addr, 4);
2165 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2167 if (dev->priv_flags & IFF_ISATAP)
2168 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2172 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2174 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2177 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2179 memcpy(eui, dev->perm_addr, 3);
2180 memcpy(eui + 5, dev->perm_addr + 3, 3);
2187 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2189 switch (dev->type) {
2192 return addrconf_ifid_eui48(eui, dev);
2194 return addrconf_ifid_arcnet(eui, dev);
2195 case ARPHRD_INFINIBAND:
2196 return addrconf_ifid_infiniband(eui, dev);
2198 return addrconf_ifid_sit(eui, dev);
2201 return addrconf_ifid_gre(eui, dev);
2202 case ARPHRD_6LOWPAN:
2203 return addrconf_ifid_6lowpan(eui, dev);
2204 case ARPHRD_IEEE1394:
2205 return addrconf_ifid_ieee1394(eui, dev);
2206 case ARPHRD_TUNNEL6:
2208 return addrconf_ifid_ip6tnl(eui, dev);
2213 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2216 struct inet6_ifaddr *ifp;
2218 read_lock_bh(&idev->lock);
2219 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2220 if (ifp->scope > IFA_LINK)
2222 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2223 memcpy(eui, ifp->addr.s6_addr+8, 8);
2228 read_unlock_bh(&idev->lock);
2232 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2233 static void ipv6_regen_rndid(struct inet6_dev *idev)
2236 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2237 idev->rndid[0] &= ~0x02;
2240 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2241 * check if generated address is not inappropriate
2243 * - Reserved subnet anycast (RFC 2526)
2244 * 11111101 11....11 1xxxxxxx
2245 * - ISATAP (RFC4214) 6.1
2246 * 00-00-5E-FE-xx-xx-xx-xx
2248 * - XXX: already assigned to an address on the device
2250 if (idev->rndid[0] == 0xfd &&
2251 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2252 (idev->rndid[7]&0x80))
2254 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2255 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2257 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2262 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2264 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2265 ipv6_regen_rndid(idev);
2273 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2274 unsigned long expires, u32 flags)
2276 struct fib6_config cfg = {
2277 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2278 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2279 .fc_ifindex = dev->ifindex,
2280 .fc_expires = expires,
2282 .fc_flags = RTF_UP | flags,
2283 .fc_nlinfo.nl_net = dev_net(dev),
2284 .fc_protocol = RTPROT_KERNEL,
2289 /* Prevent useless cloning on PtP SIT.
2290 This thing is done here expecting that the whole
2291 class of non-broadcast devices need not cloning.
2293 #if IS_ENABLED(CONFIG_IPV6_SIT)
2294 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2295 cfg.fc_flags |= RTF_NONEXTHOP;
2298 ip6_route_add(&cfg, NULL);
2302 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2304 const struct net_device *dev,
2305 u32 flags, u32 noflags)
2307 struct fib6_node *fn;
2308 struct rt6_info *rt = NULL;
2309 struct fib6_table *table;
2310 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2312 table = fib6_get_table(dev_net(dev), tb_id);
2316 read_lock_bh(&table->tb6_lock);
2317 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2321 noflags |= RTF_CACHE;
2322 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2323 if (rt->dst.dev->ifindex != dev->ifindex)
2325 if ((rt->rt6i_flags & flags) != flags)
2327 if ((rt->rt6i_flags & noflags) != 0)
2333 read_unlock_bh(&table->tb6_lock);
2338 /* Create "default" multicast route to the interface */
2340 static void addrconf_add_mroute(struct net_device *dev)
2342 struct fib6_config cfg = {
2343 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2344 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2345 .fc_ifindex = dev->ifindex,
2348 .fc_nlinfo.nl_net = dev_net(dev),
2351 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2353 ip6_route_add(&cfg, NULL);
2356 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2358 struct inet6_dev *idev;
2362 idev = ipv6_find_idev(dev);
2364 return ERR_PTR(-ENOBUFS);
2366 if (idev->cnf.disable_ipv6)
2367 return ERR_PTR(-EACCES);
2369 /* Add default multicast route */
2370 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2371 addrconf_add_mroute(dev);
2376 static void manage_tempaddrs(struct inet6_dev *idev,
2377 struct inet6_ifaddr *ifp,
2378 __u32 valid_lft, __u32 prefered_lft,
2379 bool create, unsigned long now)
2382 struct inet6_ifaddr *ift;
2384 read_lock_bh(&idev->lock);
2385 /* update all temporary addresses in the list */
2386 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2387 int age, max_valid, max_prefered;
2389 if (ifp != ift->ifpub)
2392 /* RFC 4941 section 3.3:
2393 * If a received option will extend the lifetime of a public
2394 * address, the lifetimes of temporary addresses should
2395 * be extended, subject to the overall constraint that no
2396 * temporary addresses should ever remain "valid" or "preferred"
2397 * for a time longer than (TEMP_VALID_LIFETIME) or
2398 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2400 age = (now - ift->cstamp) / HZ;
2401 max_valid = idev->cnf.temp_valid_lft - age;
2405 max_prefered = idev->cnf.temp_prefered_lft -
2406 idev->desync_factor - age;
2407 if (max_prefered < 0)
2410 if (valid_lft > max_valid)
2411 valid_lft = max_valid;
2413 if (prefered_lft > max_prefered)
2414 prefered_lft = max_prefered;
2416 spin_lock(&ift->lock);
2418 ift->valid_lft = valid_lft;
2419 ift->prefered_lft = prefered_lft;
2421 if (prefered_lft > 0)
2422 ift->flags &= ~IFA_F_DEPRECATED;
2424 spin_unlock(&ift->lock);
2425 if (!(flags&IFA_F_TENTATIVE))
2426 ipv6_ifa_notify(0, ift);
2429 if ((create || list_empty(&idev->tempaddr_list)) &&
2430 idev->cnf.use_tempaddr > 0) {
2431 /* When a new public address is created as described
2432 * in [ADDRCONF], also create a new temporary address.
2433 * Also create a temporary address if it's enabled but
2434 * no temporary address currently exists.
2436 read_unlock_bh(&idev->lock);
2437 ipv6_create_tempaddr(ifp, NULL);
2439 read_unlock_bh(&idev->lock);
2443 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2445 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2446 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2449 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2450 const struct prefix_info *pinfo,
2451 struct inet6_dev *in6_dev,
2452 const struct in6_addr *addr, int addr_type,
2453 u32 addr_flags, bool sllao, bool tokenized,
2454 __u32 valid_lft, u32 prefered_lft)
2456 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2457 int create = 0, update_lft = 0;
2459 if (!ifp && valid_lft) {
2460 int max_addresses = in6_dev->cnf.max_addresses;
2462 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2463 if (in6_dev->cnf.optimistic_dad &&
2464 !net->ipv6.devconf_all->forwarding && sllao)
2465 addr_flags |= IFA_F_OPTIMISTIC;
2468 /* Do not allow to create too much of autoconfigured
2469 * addresses; this would be too easy way to crash kernel.
2471 if (!max_addresses ||
2472 ipv6_count_addresses(in6_dev) < max_addresses)
2473 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2475 addr_type&IPV6_ADDR_SCOPE_MASK,
2476 addr_flags, valid_lft,
2479 if (IS_ERR_OR_NULL(ifp))
2484 spin_lock_bh(&ifp->lock);
2485 ifp->flags |= IFA_F_MANAGETEMPADDR;
2486 ifp->cstamp = jiffies;
2487 ifp->tokenized = tokenized;
2488 spin_unlock_bh(&ifp->lock);
2489 addrconf_dad_start(ifp);
2497 /* update lifetime (RFC2462 5.5.3 e) */
2498 spin_lock_bh(&ifp->lock);
2500 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2501 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2504 if (!update_lft && !create && stored_lft) {
2505 const u32 minimum_lft = min_t(u32,
2506 stored_lft, MIN_VALID_LIFETIME);
2507 valid_lft = max(valid_lft, minimum_lft);
2509 /* RFC4862 Section 5.5.3e:
2510 * "Note that the preferred lifetime of the
2511 * corresponding address is always reset to
2512 * the Preferred Lifetime in the received
2513 * Prefix Information option, regardless of
2514 * whether the valid lifetime is also reset or
2517 * So we should always update prefered_lft here.
2523 ifp->valid_lft = valid_lft;
2524 ifp->prefered_lft = prefered_lft;
2527 ifp->flags &= ~IFA_F_DEPRECATED;
2528 spin_unlock_bh(&ifp->lock);
2530 if (!(flags&IFA_F_TENTATIVE))
2531 ipv6_ifa_notify(0, ifp);
2533 spin_unlock_bh(&ifp->lock);
2535 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2544 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2546 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2548 struct prefix_info *pinfo;
2553 struct inet6_dev *in6_dev;
2554 struct net *net = dev_net(dev);
2556 pinfo = (struct prefix_info *) opt;
2558 if (len < sizeof(struct prefix_info)) {
2559 ADBG("addrconf: prefix option too short\n");
2564 * Validation checks ([ADDRCONF], page 19)
2567 addr_type = ipv6_addr_type(&pinfo->prefix);
2569 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2572 valid_lft = ntohl(pinfo->valid);
2573 prefered_lft = ntohl(pinfo->prefered);
2575 if (prefered_lft > valid_lft) {
2576 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2580 in6_dev = in6_dev_get(dev);
2583 net_dbg_ratelimited("addrconf: device %s not configured\n",
2589 * Two things going on here:
2590 * 1) Add routes for on-link prefixes
2591 * 2) Configure prefixes with the auto flag set
2594 if (pinfo->onlink) {
2595 struct rt6_info *rt;
2596 unsigned long rt_expires;
2598 /* Avoid arithmetic overflow. Really, we could
2599 * save rt_expires in seconds, likely valid_lft,
2600 * but it would require division in fib gc, that it
2604 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2606 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2608 if (addrconf_finite_timeout(rt_expires))
2611 rt = addrconf_get_prefix_route(&pinfo->prefix,
2614 RTF_ADDRCONF | RTF_PREFIX_RT,
2615 RTF_GATEWAY | RTF_DEFAULT);
2618 /* Autoconf prefix route */
2619 if (valid_lft == 0) {
2622 } else if (addrconf_finite_timeout(rt_expires)) {
2624 rt6_set_expires(rt, jiffies + rt_expires);
2626 rt6_clean_expires(rt);
2628 } else if (valid_lft) {
2629 clock_t expires = 0;
2630 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2631 if (addrconf_finite_timeout(rt_expires)) {
2633 flags |= RTF_EXPIRES;
2634 expires = jiffies_to_clock_t(rt_expires);
2636 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2637 dev, expires, flags);
2642 /* Try to figure out our local address for this prefix */
2644 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2645 struct in6_addr addr;
2646 bool tokenized = false, dev_addr_generated = false;
2648 if (pinfo->prefix_len == 64) {
2649 memcpy(&addr, &pinfo->prefix, 8);
2651 if (!ipv6_addr_any(&in6_dev->token)) {
2652 read_lock_bh(&in6_dev->lock);
2653 memcpy(addr.s6_addr + 8,
2654 in6_dev->token.s6_addr + 8, 8);
2655 read_unlock_bh(&in6_dev->lock);
2657 } else if (is_addr_mode_generate_stable(in6_dev) &&
2658 !ipv6_generate_stable_address(&addr, 0,
2660 addr_flags |= IFA_F_STABLE_PRIVACY;
2662 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2663 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2666 dev_addr_generated = true;
2670 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2675 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2678 tokenized, valid_lft,
2683 /* Ignore error case here because previous prefix add addr was
2684 * successful which will be notified.
2686 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2687 addr_type, addr_flags, sllao,
2688 tokenized, valid_lft,
2690 dev_addr_generated);
2692 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2694 in6_dev_put(in6_dev);
2698 * Set destination address.
2699 * Special case for SIT interfaces where we create a new "virtual"
2702 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2704 struct in6_ifreq ireq;
2705 struct net_device *dev;
2711 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2714 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2720 #if IS_ENABLED(CONFIG_IPV6_SIT)
2721 if (dev->type == ARPHRD_SIT) {
2722 const struct net_device_ops *ops = dev->netdev_ops;
2724 struct ip_tunnel_parm p;
2726 err = -EADDRNOTAVAIL;
2727 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2730 memset(&p, 0, sizeof(p));
2731 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2735 p.iph.protocol = IPPROTO_IPV6;
2737 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2739 if (ops->ndo_do_ioctl) {
2740 mm_segment_t oldfs = get_fs();
2743 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2750 dev = __dev_get_by_name(net, p.name);
2753 err = dev_open(dev);
2763 static int ipv6_mc_config(struct sock *sk, bool join,
2764 const struct in6_addr *addr, int ifindex)
2772 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2774 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2781 * Manual configuration of address on an interface
2783 static int inet6_addr_add(struct net *net, int ifindex,
2784 const struct in6_addr *pfx,
2785 const struct in6_addr *peer_pfx,
2786 unsigned int plen, __u32 ifa_flags,
2787 __u32 prefered_lft, __u32 valid_lft)
2789 struct inet6_ifaddr *ifp;
2790 struct inet6_dev *idev;
2791 struct net_device *dev;
2792 unsigned long timeout;
2802 /* check the lifetime */
2803 if (!valid_lft || prefered_lft > valid_lft)
2806 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2809 dev = __dev_get_by_index(net, ifindex);
2813 idev = addrconf_add_dev(dev);
2815 return PTR_ERR(idev);
2817 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2818 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2819 true, pfx, ifindex);
2825 scope = ipv6_addr_scope(pfx);
2827 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2828 if (addrconf_finite_timeout(timeout)) {
2829 expires = jiffies_to_clock_t(timeout * HZ);
2830 valid_lft = timeout;
2831 flags = RTF_EXPIRES;
2835 ifa_flags |= IFA_F_PERMANENT;
2838 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2839 if (addrconf_finite_timeout(timeout)) {
2841 ifa_flags |= IFA_F_DEPRECATED;
2842 prefered_lft = timeout;
2845 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2846 valid_lft, prefered_lft);
2849 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2850 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2855 * Note that section 3.1 of RFC 4429 indicates
2856 * that the Optimistic flag should not be set for
2857 * manually configured addresses
2859 addrconf_dad_start(ifp);
2860 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2861 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2864 addrconf_verify_rtnl();
2866 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2867 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2868 false, pfx, ifindex);
2871 return PTR_ERR(ifp);
2874 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2875 const struct in6_addr *pfx, unsigned int plen)
2877 struct inet6_ifaddr *ifp;
2878 struct inet6_dev *idev;
2879 struct net_device *dev;
2884 dev = __dev_get_by_index(net, ifindex);
2888 idev = __in6_dev_get(dev);
2892 read_lock_bh(&idev->lock);
2893 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2894 if (ifp->prefix_len == plen &&
2895 ipv6_addr_equal(pfx, &ifp->addr)) {
2897 read_unlock_bh(&idev->lock);
2899 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2900 (ifa_flags & IFA_F_MANAGETEMPADDR))
2901 manage_tempaddrs(idev, ifp, 0, 0, false,
2904 addrconf_verify_rtnl();
2905 if (ipv6_addr_is_multicast(pfx)) {
2906 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2907 false, pfx, dev->ifindex);
2912 read_unlock_bh(&idev->lock);
2913 return -EADDRNOTAVAIL;
2917 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2919 struct in6_ifreq ireq;
2922 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2925 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2929 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2930 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2931 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2936 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2938 struct in6_ifreq ireq;
2941 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2944 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2948 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2949 ireq.ifr6_prefixlen);
2954 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2955 int plen, int scope)
2957 struct inet6_ifaddr *ifp;
2959 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2960 scope, IFA_F_PERMANENT,
2961 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2963 spin_lock_bh(&ifp->lock);
2964 ifp->flags &= ~IFA_F_TENTATIVE;
2965 spin_unlock_bh(&ifp->lock);
2966 rt_genid_bump_ipv6(dev_net(idev->dev));
2967 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2972 #if IS_ENABLED(CONFIG_IPV6_SIT)
2973 static void sit_add_v4_addrs(struct inet6_dev *idev)
2975 struct in6_addr addr;
2976 struct net_device *dev;
2977 struct net *net = dev_net(idev->dev);
2983 memset(&addr, 0, sizeof(struct in6_addr));
2984 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2986 if (idev->dev->flags&IFF_POINTOPOINT) {
2987 addr.s6_addr32[0] = htonl(0xfe800000);
2991 scope = IPV6_ADDR_COMPATv4;
2993 pflags |= RTF_NONEXTHOP;
2996 if (addr.s6_addr32[3]) {
2997 add_addr(idev, &addr, plen, scope);
2998 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3002 for_each_netdev(net, dev) {
3003 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3004 if (in_dev && (dev->flags & IFF_UP)) {
3005 struct in_ifaddr *ifa;
3009 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3011 addr.s6_addr32[3] = ifa->ifa_local;
3013 if (ifa->ifa_scope == RT_SCOPE_LINK)
3015 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3016 if (idev->dev->flags&IFF_POINTOPOINT)
3021 add_addr(idev, &addr, plen, flag);
3022 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3030 static void init_loopback(struct net_device *dev)
3032 struct inet6_dev *idev;
3038 idev = ipv6_find_idev(dev);
3040 pr_debug("%s: add_dev failed\n", __func__);
3044 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3047 void addrconf_add_linklocal(struct inet6_dev *idev,
3048 const struct in6_addr *addr, u32 flags)
3050 struct inet6_ifaddr *ifp;
3051 u32 addr_flags = flags | IFA_F_PERMANENT;
3053 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3054 if (idev->cnf.optimistic_dad &&
3055 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3056 addr_flags |= IFA_F_OPTIMISTIC;
3059 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3060 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3062 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3063 addrconf_dad_start(ifp);
3067 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3069 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3071 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3074 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3075 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3078 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3079 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3085 static int ipv6_generate_stable_address(struct in6_addr *address,
3087 const struct inet6_dev *idev)
3089 static DEFINE_SPINLOCK(lock);
3090 static __u32 digest[SHA_DIGEST_WORDS];
3091 static __u32 workspace[SHA_WORKSPACE_WORDS];
3094 char __data[SHA_MESSAGE_BYTES];
3096 struct in6_addr secret;
3098 unsigned char hwaddr[MAX_ADDR_LEN];
3103 struct in6_addr secret;
3104 struct in6_addr temp;
3105 struct net *net = dev_net(idev->dev);
3107 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3109 if (idev->cnf.stable_secret.initialized)
3110 secret = idev->cnf.stable_secret.secret;
3111 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3112 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3117 spin_lock_bh(&lock);
3120 memset(&data, 0, sizeof(data));
3121 memset(workspace, 0, sizeof(workspace));
3122 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3123 data.prefix[0] = address->s6_addr32[0];
3124 data.prefix[1] = address->s6_addr32[1];
3125 data.secret = secret;
3126 data.dad_count = dad_count;
3128 sha_transform(digest, data.__data, workspace);
3131 temp.s6_addr32[2] = (__force __be32)digest[0];
3132 temp.s6_addr32[3] = (__force __be32)digest[1];
3134 spin_unlock_bh(&lock);
3136 if (ipv6_reserved_interfaceid(temp)) {
3138 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3147 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3149 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3153 s = &idev->cnf.stable_secret;
3154 get_random_bytes(&s->secret, sizeof(s->secret));
3155 s->initialized = true;
3158 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3160 struct in6_addr addr;
3162 /* no link local addresses on L3 master devices */
3163 if (netif_is_l3_master(idev->dev))
3166 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3168 switch (idev->cnf.addr_gen_mode) {
3169 case IN6_ADDR_GEN_MODE_RANDOM:
3170 ipv6_gen_mode_random_init(idev);
3172 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3173 if (!ipv6_generate_stable_address(&addr, 0, idev))
3174 addrconf_add_linklocal(idev, &addr,
3175 IFA_F_STABLE_PRIVACY);
3176 else if (prefix_route)
3177 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3179 case IN6_ADDR_GEN_MODE_EUI64:
3180 /* addrconf_add_linklocal also adds a prefix_route and we
3181 * only need to care about prefix routes if ipv6_generate_eui64
3182 * couldn't generate one.
3184 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3185 addrconf_add_linklocal(idev, &addr, 0);
3186 else if (prefix_route)
3187 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3189 case IN6_ADDR_GEN_MODE_NONE:
3191 /* will not add any link local address */
3196 static void addrconf_dev_config(struct net_device *dev)
3198 struct inet6_dev *idev;
3202 if ((dev->type != ARPHRD_ETHER) &&
3203 (dev->type != ARPHRD_FDDI) &&
3204 (dev->type != ARPHRD_ARCNET) &&
3205 (dev->type != ARPHRD_INFINIBAND) &&
3206 (dev->type != ARPHRD_IEEE1394) &&
3207 (dev->type != ARPHRD_TUNNEL6) &&
3208 (dev->type != ARPHRD_6LOWPAN) &&
3209 (dev->type != ARPHRD_IP6GRE) &&
3210 (dev->type != ARPHRD_IPGRE) &&
3211 (dev->type != ARPHRD_TUNNEL) &&
3212 (dev->type != ARPHRD_NONE)) {
3213 /* Alas, we support only Ethernet autoconfiguration. */
3217 idev = addrconf_add_dev(dev);
3221 /* this device type has no EUI support */
3222 if (dev->type == ARPHRD_NONE &&
3223 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3224 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3226 addrconf_addr_gen(idev, false);
3229 #if IS_ENABLED(CONFIG_IPV6_SIT)
3230 static void addrconf_sit_config(struct net_device *dev)
3232 struct inet6_dev *idev;
3237 * Configure the tunnel with one of our IPv4
3238 * addresses... we should configure all of
3239 * our v4 addrs in the tunnel
3242 idev = ipv6_find_idev(dev);
3244 pr_debug("%s: add_dev failed\n", __func__);
3248 if (dev->priv_flags & IFF_ISATAP) {
3249 addrconf_addr_gen(idev, false);
3253 sit_add_v4_addrs(idev);
3255 if (dev->flags&IFF_POINTOPOINT)
3256 addrconf_add_mroute(dev);
3260 #if IS_ENABLED(CONFIG_NET_IPGRE)
3261 static void addrconf_gre_config(struct net_device *dev)
3263 struct inet6_dev *idev;
3267 idev = ipv6_find_idev(dev);
3269 pr_debug("%s: add_dev failed\n", __func__);
3273 addrconf_addr_gen(idev, true);
3274 if (dev->flags & IFF_POINTOPOINT)
3275 addrconf_add_mroute(dev);
3279 static int fixup_permanent_addr(struct inet6_dev *idev,
3280 struct inet6_ifaddr *ifp)
3282 /* !rt6i_node means the host route was removed from the
3283 * FIB, for example, if 'lo' device is taken down. In that
3284 * case regenerate the host route.
3286 if (!ifp->rt || !ifp->rt->rt6i_node) {
3287 struct rt6_info *rt, *prev;
3289 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3290 if (unlikely(IS_ERR(rt)))
3293 /* ifp->rt can be accessed outside of rtnl */
3294 spin_lock(&ifp->lock);
3297 spin_unlock(&ifp->lock);
3302 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3303 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3307 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3308 addrconf_dad_start(ifp);
3313 static void addrconf_permanent_addr(struct net_device *dev)
3315 struct inet6_ifaddr *ifp, *tmp;
3316 struct inet6_dev *idev;
3318 idev = __in6_dev_get(dev);
3322 write_lock_bh(&idev->lock);
3324 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3325 if ((ifp->flags & IFA_F_PERMANENT) &&
3326 fixup_permanent_addr(idev, ifp) < 0) {
3327 write_unlock_bh(&idev->lock);
3329 write_lock_bh(&idev->lock);
3331 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3332 idev->dev->name, &ifp->addr);
3336 write_unlock_bh(&idev->lock);
3339 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3342 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3343 struct netdev_notifier_changeupper_info *info;
3344 struct inet6_dev *idev = __in6_dev_get(dev);
3345 struct net *net = dev_net(dev);
3346 int run_pending = 0;
3350 case NETDEV_REGISTER:
3351 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3352 idev = ipv6_add_dev(dev);
3354 return notifier_from_errno(PTR_ERR(idev));
3358 case NETDEV_CHANGEMTU:
3359 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3360 if (dev->mtu < IPV6_MIN_MTU) {
3361 addrconf_ifdown(dev, dev != net->loopback_dev);
3366 rt6_mtu_change(dev, dev->mtu);
3367 idev->cnf.mtu6 = dev->mtu;
3371 /* allocate new idev */
3372 idev = ipv6_add_dev(dev);
3376 /* device is still not ready */
3377 if (!(idev->if_flags & IF_READY))
3386 if (dev->flags & IFF_SLAVE)
3389 if (idev && idev->cnf.disable_ipv6)
3392 if (event == NETDEV_UP) {
3393 /* restore routes for permanent addresses */
3394 addrconf_permanent_addr(dev);
3396 if (!addrconf_qdisc_ok(dev)) {
3397 /* device is not ready yet. */
3398 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3403 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3404 idev = ipv6_add_dev(dev);
3406 if (!IS_ERR_OR_NULL(idev)) {
3407 idev->if_flags |= IF_READY;
3410 } else if (event == NETDEV_CHANGE) {
3411 if (!addrconf_qdisc_ok(dev)) {
3412 /* device is still not ready. */
3417 if (idev->if_flags & IF_READY) {
3418 /* device is already configured -
3419 * but resend MLD reports, we might
3420 * have roamed and need to update
3421 * multicast snooping switches
3426 idev->if_flags |= IF_READY;
3429 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3435 switch (dev->type) {
3436 #if IS_ENABLED(CONFIG_IPV6_SIT)
3438 addrconf_sit_config(dev);
3441 #if IS_ENABLED(CONFIG_NET_IPGRE)
3443 addrconf_gre_config(dev);
3446 case ARPHRD_LOOPBACK:
3451 addrconf_dev_config(dev);
3455 if (!IS_ERR_OR_NULL(idev)) {
3457 addrconf_dad_run(idev);
3460 * If the MTU changed during the interface down,
3461 * when the interface up, the changed MTU must be
3462 * reflected in the idev as well as routers.
3464 if (idev->cnf.mtu6 != dev->mtu &&
3465 dev->mtu >= IPV6_MIN_MTU) {
3466 rt6_mtu_change(dev, dev->mtu);
3467 idev->cnf.mtu6 = dev->mtu;
3469 idev->tstamp = jiffies;
3470 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3473 * If the changed mtu during down is lower than
3474 * IPV6_MIN_MTU stop IPv6 on this interface.
3476 if (dev->mtu < IPV6_MIN_MTU)
3477 addrconf_ifdown(dev, dev != net->loopback_dev);
3482 case NETDEV_UNREGISTER:
3484 * Remove all addresses from this interface.
3486 addrconf_ifdown(dev, event != NETDEV_DOWN);
3489 case NETDEV_CHANGENAME:
3491 snmp6_unregister_dev(idev);
3492 addrconf_sysctl_unregister(idev);
3493 err = addrconf_sysctl_register(idev);
3495 return notifier_from_errno(err);
3496 err = snmp6_register_dev(idev);
3498 addrconf_sysctl_unregister(idev);
3499 return notifier_from_errno(err);
3504 case NETDEV_PRE_TYPE_CHANGE:
3505 case NETDEV_POST_TYPE_CHANGE:
3507 addrconf_type_change(dev, event);
3510 case NETDEV_CHANGEUPPER:
3513 /* flush all routes if dev is linked to or unlinked from
3514 * an L3 master device (e.g., VRF)
3516 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3517 addrconf_ifdown(dev, 0);
3524 * addrconf module should be notified of a device going up
3526 static struct notifier_block ipv6_dev_notf = {
3527 .notifier_call = addrconf_notify,
3528 .priority = ADDRCONF_NOTIFY_PRIORITY,
3531 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3533 struct inet6_dev *idev;
3536 idev = __in6_dev_get(dev);
3538 if (event == NETDEV_POST_TYPE_CHANGE)
3539 ipv6_mc_remap(idev);
3540 else if (event == NETDEV_PRE_TYPE_CHANGE)
3541 ipv6_mc_unmap(idev);
3544 static bool addr_is_local(const struct in6_addr *addr)
3546 return ipv6_addr_type(addr) &
3547 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3550 static int addrconf_ifdown(struct net_device *dev, int how)
3552 struct net *net = dev_net(dev);
3553 struct inet6_dev *idev;
3554 struct inet6_ifaddr *ifa, *tmp;
3555 struct list_head del_list;
3562 rt6_ifdown(net, dev);
3563 neigh_ifdown(&nd_tbl, dev);
3565 idev = __in6_dev_get(dev);
3570 * Step 1: remove reference to ipv6 device from parent device.
3576 /* protected by rtnl_lock */
3577 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3579 /* Step 1.5: remove snmp6 entry */
3580 snmp6_unregister_dev(idev);
3584 /* aggregate the system setting and interface setting */
3585 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3587 _keep_addr = idev->cnf.keep_addr_on_down;
3589 /* combine the user config with event to determine if permanent
3590 * addresses are to be removed from address hash table
3592 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3594 /* Step 2: clear hash table */
3595 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3596 struct hlist_head *h = &inet6_addr_lst[i];
3598 spin_lock_bh(&addrconf_hash_lock);
3600 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3601 if (ifa->idev == idev) {
3602 addrconf_del_dad_work(ifa);
3603 /* combined flag + permanent flag decide if
3604 * address is retained on a down event
3607 !(ifa->flags & IFA_F_PERMANENT) ||
3608 addr_is_local(&ifa->addr)) {
3609 hlist_del_init_rcu(&ifa->addr_lst);
3614 spin_unlock_bh(&addrconf_hash_lock);
3617 write_lock_bh(&idev->lock);
3619 addrconf_del_rs_timer(idev);
3621 /* Step 2: clear flags for stateless addrconf */
3623 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3625 /* Step 3: clear tempaddr list */
3626 while (!list_empty(&idev->tempaddr_list)) {
3627 ifa = list_first_entry(&idev->tempaddr_list,
3628 struct inet6_ifaddr, tmp_list);
3629 list_del(&ifa->tmp_list);
3630 write_unlock_bh(&idev->lock);
3631 spin_lock_bh(&ifa->lock);
3634 in6_ifa_put(ifa->ifpub);
3637 spin_unlock_bh(&ifa->lock);
3639 write_lock_bh(&idev->lock);
3642 /* re-combine the user config with event to determine if permanent
3643 * addresses are to be removed from the interface list
3645 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3647 INIT_LIST_HEAD(&del_list);
3648 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3649 struct rt6_info *rt = NULL;
3652 addrconf_del_dad_work(ifa);
3654 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3655 !addr_is_local(&ifa->addr);
3657 list_move(&ifa->if_list, &del_list);
3659 write_unlock_bh(&idev->lock);
3660 spin_lock_bh(&ifa->lock);
3663 /* set state to skip the notifier below */
3664 state = INET6_IFADDR_STATE_DEAD;
3665 ifa->state = INET6_IFADDR_STATE_PREDAD;
3666 if (!(ifa->flags & IFA_F_NODAD))
3667 ifa->flags |= IFA_F_TENTATIVE;
3673 ifa->state = INET6_IFADDR_STATE_DEAD;
3676 spin_unlock_bh(&ifa->lock);
3681 if (state != INET6_IFADDR_STATE_DEAD) {
3682 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3683 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3685 if (idev->cnf.forwarding)
3686 addrconf_leave_anycast(ifa);
3687 addrconf_leave_solict(ifa->idev, &ifa->addr);
3690 write_lock_bh(&idev->lock);
3693 write_unlock_bh(&idev->lock);
3695 /* now clean up addresses to be removed */
3696 while (!list_empty(&del_list)) {
3697 ifa = list_first_entry(&del_list,
3698 struct inet6_ifaddr, if_list);
3699 list_del(&ifa->if_list);
3704 /* Step 5: Discard anycast and multicast list */
3706 ipv6_ac_destroy_dev(idev);
3707 ipv6_mc_destroy_dev(idev);
3712 idev->tstamp = jiffies;
3714 /* Last: Shot the device (if unregistered) */
3716 addrconf_sysctl_unregister(idev);
3717 neigh_parms_release(&nd_tbl, idev->nd_parms);
3718 neigh_ifdown(&nd_tbl, dev);
3724 static void addrconf_rs_timer(unsigned long data)
3726 struct inet6_dev *idev = (struct inet6_dev *)data;
3727 struct net_device *dev = idev->dev;
3728 struct in6_addr lladdr;
3730 write_lock(&idev->lock);
3731 if (idev->dead || !(idev->if_flags & IF_READY))
3734 if (!ipv6_accept_ra(idev))
3737 /* Announcement received after solicitation was sent */
3738 if (idev->if_flags & IF_RA_RCVD)
3741 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3742 write_unlock(&idev->lock);
3743 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3744 ndisc_send_rs(dev, &lladdr,
3745 &in6addr_linklocal_allrouters);
3749 write_lock(&idev->lock);
3750 idev->rs_interval = rfc3315_s14_backoff_update(
3751 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3752 /* The wait after the last probe can be shorter */
3753 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3754 idev->cnf.rtr_solicits) ?
3755 idev->cnf.rtr_solicit_delay :
3759 * Note: we do not support deprecated "all on-link"
3760 * assumption any longer.
3762 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3766 write_unlock(&idev->lock);
3772 * Duplicate Address Detection
3774 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3776 unsigned long rand_num;
3777 struct inet6_dev *idev = ifp->idev;
3780 if (ifp->flags & IFA_F_OPTIMISTIC)
3783 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3786 if (idev->cnf.enhanced_dad ||
3787 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3789 get_random_bytes(&nonce, 6);
3792 ifp->dad_nonce = nonce;
3793 ifp->dad_probes = idev->cnf.dad_transmits;
3794 addrconf_mod_dad_work(ifp, rand_num);
3797 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3799 struct inet6_dev *idev = ifp->idev;
3800 struct net_device *dev = idev->dev;
3801 bool bump_id, notify = false;
3803 addrconf_join_solict(dev, &ifp->addr);
3805 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3807 read_lock_bh(&idev->lock);
3808 spin_lock(&ifp->lock);
3809 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3812 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3813 idev->cnf.accept_dad < 1 ||
3814 !(ifp->flags&IFA_F_TENTATIVE) ||
3815 ifp->flags & IFA_F_NODAD) {
3816 bump_id = ifp->flags & IFA_F_TENTATIVE;
3817 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3818 spin_unlock(&ifp->lock);
3819 read_unlock_bh(&idev->lock);
3821 addrconf_dad_completed(ifp, bump_id);
3825 if (!(idev->if_flags & IF_READY)) {
3826 spin_unlock(&ifp->lock);
3827 read_unlock_bh(&idev->lock);
3829 * If the device is not ready:
3830 * - keep it tentative if it is a permanent address.
3831 * - otherwise, kill it.
3834 addrconf_dad_stop(ifp, 0);
3839 * Optimistic nodes can start receiving
3842 if (ifp->flags & IFA_F_OPTIMISTIC) {
3843 ip6_ins_rt(ifp->rt);
3844 if (ipv6_use_optimistic_addr(idev)) {
3845 /* Because optimistic nodes can use this address,
3846 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3852 addrconf_dad_kick(ifp);
3854 spin_unlock(&ifp->lock);
3855 read_unlock_bh(&idev->lock);
3857 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3860 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3862 bool begin_dad = false;
3864 spin_lock_bh(&ifp->lock);
3865 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3866 ifp->state = INET6_IFADDR_STATE_PREDAD;
3869 spin_unlock_bh(&ifp->lock);
3872 addrconf_mod_dad_work(ifp, 0);
3875 static void addrconf_dad_work(struct work_struct *w)
3877 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3878 struct inet6_ifaddr,
3880 struct inet6_dev *idev = ifp->idev;
3881 bool bump_id, disable_ipv6 = false;
3882 struct in6_addr mcaddr;
3888 } action = DAD_PROCESS;
3892 spin_lock_bh(&ifp->lock);
3893 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3895 ifp->state = INET6_IFADDR_STATE_DAD;
3896 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3898 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3900 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3901 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3902 struct in6_addr addr;
3904 addr.s6_addr32[0] = htonl(0xfe800000);
3905 addr.s6_addr32[1] = 0;
3907 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3908 ipv6_addr_equal(&ifp->addr, &addr)) {
3909 /* DAD failed for link-local based on MAC */
3910 idev->cnf.disable_ipv6 = 1;
3912 pr_info("%s: IPv6 being disabled!\n",
3913 ifp->idev->dev->name);
3914 disable_ipv6 = true;
3918 spin_unlock_bh(&ifp->lock);
3920 if (action == DAD_BEGIN) {
3921 addrconf_dad_begin(ifp);
3923 } else if (action == DAD_ABORT) {
3925 addrconf_dad_stop(ifp, 1);
3927 addrconf_ifdown(idev->dev, 0);
3931 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3934 write_lock_bh(&idev->lock);
3935 if (idev->dead || !(idev->if_flags & IF_READY)) {
3936 write_unlock_bh(&idev->lock);
3940 spin_lock(&ifp->lock);
3941 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3942 spin_unlock(&ifp->lock);
3943 write_unlock_bh(&idev->lock);
3947 if (ifp->dad_probes == 0) {
3949 * DAD was successful
3952 bump_id = ifp->flags & IFA_F_TENTATIVE;
3953 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3954 spin_unlock(&ifp->lock);
3955 write_unlock_bh(&idev->lock);
3957 addrconf_dad_completed(ifp, bump_id);
3963 addrconf_mod_dad_work(ifp,
3964 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3965 spin_unlock(&ifp->lock);
3966 write_unlock_bh(&idev->lock);
3968 /* send a neighbour solicitation for our addr */
3969 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3970 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
3977 /* ifp->idev must be at least read locked */
3978 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3980 struct inet6_ifaddr *ifpiter;
3981 struct inet6_dev *idev = ifp->idev;
3983 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3984 if (ifpiter->scope > IFA_LINK)
3986 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3987 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3988 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3995 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
3997 struct net_device *dev = ifp->idev->dev;
3998 struct in6_addr lladdr;
3999 bool send_rs, send_mld;
4001 addrconf_del_dad_work(ifp);
4004 * Configure the address for reception. Now it is valid.
4007 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4009 /* If added prefix is link local and we are prepared to process
4010 router advertisements, start sending router solicitations.
4013 read_lock_bh(&ifp->idev->lock);
4014 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4015 send_rs = send_mld &&
4016 ipv6_accept_ra(ifp->idev) &&
4017 ifp->idev->cnf.rtr_solicits != 0 &&
4018 (dev->flags&IFF_LOOPBACK) == 0;
4019 read_unlock_bh(&ifp->idev->lock);
4021 /* While dad is in progress mld report's source address is in6_addrany.
4022 * Resend with proper ll now.
4025 ipv6_mc_dad_complete(ifp->idev);
4029 * If a host as already performed a random delay
4030 * [...] as part of DAD [...] there is no need
4031 * to delay again before sending the first RS
4033 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4035 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4037 write_lock_bh(&ifp->idev->lock);
4038 spin_lock(&ifp->lock);
4039 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4040 ifp->idev->cnf.rtr_solicit_interval);
4041 ifp->idev->rs_probes = 1;
4042 ifp->idev->if_flags |= IF_RS_SENT;
4043 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4044 spin_unlock(&ifp->lock);
4045 write_unlock_bh(&ifp->idev->lock);
4049 rt_genid_bump_ipv6(dev_net(dev));
4051 /* Make sure that a new temporary address will be created
4052 * before this temporary address becomes deprecated.
4054 if (ifp->flags & IFA_F_TEMPORARY)
4055 addrconf_verify_rtnl();
4058 static void addrconf_dad_run(struct inet6_dev *idev)
4060 struct inet6_ifaddr *ifp;
4062 read_lock_bh(&idev->lock);
4063 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4064 spin_lock(&ifp->lock);
4065 if (ifp->flags & IFA_F_TENTATIVE &&
4066 ifp->state == INET6_IFADDR_STATE_DAD)
4067 addrconf_dad_kick(ifp);
4068 spin_unlock(&ifp->lock);
4070 read_unlock_bh(&idev->lock);
4073 #ifdef CONFIG_PROC_FS
4074 struct if6_iter_state {
4075 struct seq_net_private p;
4080 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4082 struct inet6_ifaddr *ifa = NULL;
4083 struct if6_iter_state *state = seq->private;
4084 struct net *net = seq_file_net(seq);
4087 /* initial bucket if pos is 0 */
4093 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4094 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4096 if (!net_eq(dev_net(ifa->idev->dev), net))
4098 /* sync with offset */
4099 if (p < state->offset) {
4107 /* prepare for next bucket */
4114 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4115 struct inet6_ifaddr *ifa)
4117 struct if6_iter_state *state = seq->private;
4118 struct net *net = seq_file_net(seq);
4120 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4121 if (!net_eq(dev_net(ifa->idev->dev), net))
4127 while (++state->bucket < IN6_ADDR_HSIZE) {
4129 hlist_for_each_entry_rcu_bh(ifa,
4130 &inet6_addr_lst[state->bucket], addr_lst) {
4131 if (!net_eq(dev_net(ifa->idev->dev), net))
4141 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4145 return if6_get_first(seq, *pos);
4148 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4150 struct inet6_ifaddr *ifa;
4152 ifa = if6_get_next(seq, v);
4157 static void if6_seq_stop(struct seq_file *seq, void *v)
4160 rcu_read_unlock_bh();
4163 static int if6_seq_show(struct seq_file *seq, void *v)
4165 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4166 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4168 ifp->idev->dev->ifindex,
4172 ifp->idev->dev->name);
4176 static const struct seq_operations if6_seq_ops = {
4177 .start = if6_seq_start,
4178 .next = if6_seq_next,
4179 .show = if6_seq_show,
4180 .stop = if6_seq_stop,
4183 static int if6_seq_open(struct inode *inode, struct file *file)
4185 return seq_open_net(inode, file, &if6_seq_ops,
4186 sizeof(struct if6_iter_state));
4189 static const struct file_operations if6_fops = {
4190 .owner = THIS_MODULE,
4191 .open = if6_seq_open,
4193 .llseek = seq_lseek,
4194 .release = seq_release_net,
4197 static int __net_init if6_proc_net_init(struct net *net)
4199 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4204 static void __net_exit if6_proc_net_exit(struct net *net)
4206 remove_proc_entry("if_inet6", net->proc_net);
4209 static struct pernet_operations if6_proc_net_ops = {
4210 .init = if6_proc_net_init,
4211 .exit = if6_proc_net_exit,
4214 int __init if6_proc_init(void)
4216 return register_pernet_subsys(&if6_proc_net_ops);
4219 void if6_proc_exit(void)
4221 unregister_pernet_subsys(&if6_proc_net_ops);
4223 #endif /* CONFIG_PROC_FS */
4225 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4226 /* Check if address is a home address configured on any interface. */
4227 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4230 struct inet6_ifaddr *ifp = NULL;
4231 unsigned int hash = inet6_addr_hash(addr);
4234 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4235 if (!net_eq(dev_net(ifp->idev->dev), net))
4237 if (ipv6_addr_equal(&ifp->addr, addr) &&
4238 (ifp->flags & IFA_F_HOMEADDRESS)) {
4243 rcu_read_unlock_bh();
4249 * Periodic address status verification
4252 static void addrconf_verify_rtnl(void)
4254 unsigned long now, next, next_sec, next_sched;
4255 struct inet6_ifaddr *ifp;
4262 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4264 cancel_delayed_work(&addr_chk_work);
4266 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4268 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4271 /* When setting preferred_lft to a value not zero or
4272 * infinity, while valid_lft is infinity
4273 * IFA_F_PERMANENT has a non-infinity life time.
4275 if ((ifp->flags & IFA_F_PERMANENT) &&
4276 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4279 spin_lock(&ifp->lock);
4280 /* We try to batch several events at once. */
4281 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4283 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4284 age >= ifp->valid_lft) {
4285 spin_unlock(&ifp->lock);
4289 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4290 spin_unlock(&ifp->lock);
4292 } else if (age >= ifp->prefered_lft) {
4293 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4296 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4298 ifp->flags |= IFA_F_DEPRECATED;
4301 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4302 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4303 next = ifp->tstamp + ifp->valid_lft * HZ;
4305 spin_unlock(&ifp->lock);
4310 ipv6_ifa_notify(0, ifp);
4314 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4315 !(ifp->flags&IFA_F_TENTATIVE)) {
4316 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4317 ifp->idev->cnf.dad_transmits *
4318 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4320 if (age >= ifp->prefered_lft - regen_advance) {
4321 struct inet6_ifaddr *ifpub = ifp->ifpub;
4322 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4323 next = ifp->tstamp + ifp->prefered_lft * HZ;
4324 if (!ifp->regen_count && ifpub) {
4327 in6_ifa_hold(ifpub);
4328 spin_unlock(&ifp->lock);
4330 spin_lock(&ifpub->lock);
4331 ifpub->regen_count = 0;
4332 spin_unlock(&ifpub->lock);
4333 ipv6_create_tempaddr(ifpub, ifp);
4338 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4339 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4340 spin_unlock(&ifp->lock);
4342 /* ifp->prefered_lft <= ifp->valid_lft */
4343 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4344 next = ifp->tstamp + ifp->prefered_lft * HZ;
4345 spin_unlock(&ifp->lock);
4350 next_sec = round_jiffies_up(next);
4353 /* If rounded timeout is accurate enough, accept it. */
4354 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4355 next_sched = next_sec;
4357 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4358 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4359 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4361 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4362 now, next, next_sec, next_sched);
4363 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4364 rcu_read_unlock_bh();
4367 static void addrconf_verify_work(struct work_struct *w)
4370 addrconf_verify_rtnl();
4374 static void addrconf_verify(void)
4376 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4379 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4380 struct in6_addr **peer_pfx)
4382 struct in6_addr *pfx = NULL;
4387 pfx = nla_data(addr);
4390 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4392 pfx = nla_data(local);
4398 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4399 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4400 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4401 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4402 [IFA_FLAGS] = { .len = sizeof(u32) },
4406 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4407 struct netlink_ext_ack *extack)
4409 struct net *net = sock_net(skb->sk);
4410 struct ifaddrmsg *ifm;
4411 struct nlattr *tb[IFA_MAX+1];
4412 struct in6_addr *pfx, *peer_pfx;
4416 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4421 ifm = nlmsg_data(nlh);
4422 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4426 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4428 /* We ignore other flags so far. */
4429 ifa_flags &= IFA_F_MANAGETEMPADDR;
4431 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4432 ifm->ifa_prefixlen);
4435 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4436 u32 prefered_lft, u32 valid_lft)
4440 unsigned long timeout;
4441 bool was_managetempaddr;
4442 bool had_prefixroute;
4446 if (!valid_lft || (prefered_lft > valid_lft))
4449 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4450 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4453 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4454 if (addrconf_finite_timeout(timeout)) {
4455 expires = jiffies_to_clock_t(timeout * HZ);
4456 valid_lft = timeout;
4457 flags = RTF_EXPIRES;
4461 ifa_flags |= IFA_F_PERMANENT;
4464 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4465 if (addrconf_finite_timeout(timeout)) {
4467 ifa_flags |= IFA_F_DEPRECATED;
4468 prefered_lft = timeout;
4471 spin_lock_bh(&ifp->lock);
4472 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4473 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4474 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4475 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4476 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4477 IFA_F_NOPREFIXROUTE);
4478 ifp->flags |= ifa_flags;
4479 ifp->tstamp = jiffies;
4480 ifp->valid_lft = valid_lft;
4481 ifp->prefered_lft = prefered_lft;
4483 spin_unlock_bh(&ifp->lock);
4484 if (!(ifp->flags&IFA_F_TENTATIVE))
4485 ipv6_ifa_notify(0, ifp);
4487 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4488 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4490 } else if (had_prefixroute) {
4491 enum cleanup_prefix_rt_t action;
4492 unsigned long rt_expires;
4494 write_lock_bh(&ifp->idev->lock);
4495 action = check_cleanup_prefix_route(ifp, &rt_expires);
4496 write_unlock_bh(&ifp->idev->lock);
4498 if (action != CLEANUP_PREFIX_RT_NOP) {
4499 cleanup_prefix_route(ifp, rt_expires,
4500 action == CLEANUP_PREFIX_RT_DEL);
4504 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4505 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4506 valid_lft = prefered_lft = 0;
4507 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4508 !was_managetempaddr, jiffies);
4511 addrconf_verify_rtnl();
4517 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4518 struct netlink_ext_ack *extack)
4520 struct net *net = sock_net(skb->sk);
4521 struct ifaddrmsg *ifm;
4522 struct nlattr *tb[IFA_MAX+1];
4523 struct in6_addr *pfx, *peer_pfx;
4524 struct inet6_ifaddr *ifa;
4525 struct net_device *dev;
4526 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4530 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4535 ifm = nlmsg_data(nlh);
4536 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4540 if (tb[IFA_CACHEINFO]) {
4541 struct ifa_cacheinfo *ci;
4543 ci = nla_data(tb[IFA_CACHEINFO]);
4544 valid_lft = ci->ifa_valid;
4545 preferred_lft = ci->ifa_prefered;
4547 preferred_lft = INFINITY_LIFE_TIME;
4548 valid_lft = INFINITY_LIFE_TIME;
4551 dev = __dev_get_by_index(net, ifm->ifa_index);
4555 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4557 /* We ignore other flags so far. */
4558 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4559 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4561 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4564 * It would be best to check for !NLM_F_CREATE here but
4565 * userspace already relies on not having to provide this.
4567 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4568 ifm->ifa_prefixlen, ifa_flags,
4569 preferred_lft, valid_lft);
4572 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4573 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4576 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4583 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4584 u8 scope, int ifindex)
4586 struct ifaddrmsg *ifm;
4588 ifm = nlmsg_data(nlh);
4589 ifm->ifa_family = AF_INET6;
4590 ifm->ifa_prefixlen = prefixlen;
4591 ifm->ifa_flags = flags;
4592 ifm->ifa_scope = scope;
4593 ifm->ifa_index = ifindex;
4596 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4597 unsigned long tstamp, u32 preferred, u32 valid)
4599 struct ifa_cacheinfo ci;
4601 ci.cstamp = cstamp_delta(cstamp);
4602 ci.tstamp = cstamp_delta(tstamp);
4603 ci.ifa_prefered = preferred;
4604 ci.ifa_valid = valid;
4606 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4609 static inline int rt_scope(int ifa_scope)
4611 if (ifa_scope & IFA_HOST)
4612 return RT_SCOPE_HOST;
4613 else if (ifa_scope & IFA_LINK)
4614 return RT_SCOPE_LINK;
4615 else if (ifa_scope & IFA_SITE)
4616 return RT_SCOPE_SITE;
4618 return RT_SCOPE_UNIVERSE;
4621 static inline int inet6_ifaddr_msgsize(void)
4623 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4624 + nla_total_size(16) /* IFA_LOCAL */
4625 + nla_total_size(16) /* IFA_ADDRESS */
4626 + nla_total_size(sizeof(struct ifa_cacheinfo))
4627 + nla_total_size(4) /* IFA_FLAGS */;
4630 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4631 u32 portid, u32 seq, int event, unsigned int flags)
4633 struct nlmsghdr *nlh;
4634 u32 preferred, valid;
4636 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4640 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4641 ifa->idev->dev->ifindex);
4643 if (!((ifa->flags&IFA_F_PERMANENT) &&
4644 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4645 preferred = ifa->prefered_lft;
4646 valid = ifa->valid_lft;
4647 if (preferred != INFINITY_LIFE_TIME) {
4648 long tval = (jiffies - ifa->tstamp)/HZ;
4649 if (preferred > tval)
4653 if (valid != INFINITY_LIFE_TIME) {
4661 preferred = INFINITY_LIFE_TIME;
4662 valid = INFINITY_LIFE_TIME;
4665 if (!ipv6_addr_any(&ifa->peer_addr)) {
4666 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4667 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4670 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4673 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4676 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4679 nlmsg_end(skb, nlh);
4683 nlmsg_cancel(skb, nlh);
4687 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4688 u32 portid, u32 seq, int event, u16 flags)
4690 struct nlmsghdr *nlh;
4691 u8 scope = RT_SCOPE_UNIVERSE;
4692 int ifindex = ifmca->idev->dev->ifindex;
4694 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4695 scope = RT_SCOPE_SITE;
4697 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4701 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4702 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4703 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4704 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4705 nlmsg_cancel(skb, nlh);
4709 nlmsg_end(skb, nlh);
4713 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4714 u32 portid, u32 seq, int event, unsigned int flags)
4716 struct nlmsghdr *nlh;
4717 u8 scope = RT_SCOPE_UNIVERSE;
4718 int ifindex = ifaca->aca_idev->dev->ifindex;
4720 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4721 scope = RT_SCOPE_SITE;
4723 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4727 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4728 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4729 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4730 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4731 nlmsg_cancel(skb, nlh);
4735 nlmsg_end(skb, nlh);
4745 /* called with rcu_read_lock() */
4746 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4747 struct netlink_callback *cb, enum addr_type_t type,
4748 int s_ip_idx, int *p_ip_idx)
4750 struct ifmcaddr6 *ifmca;
4751 struct ifacaddr6 *ifaca;
4753 int ip_idx = *p_ip_idx;
4755 read_lock_bh(&idev->lock);
4757 case UNICAST_ADDR: {
4758 struct inet6_ifaddr *ifa;
4760 /* unicast address incl. temp addr */
4761 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4762 if (++ip_idx < s_ip_idx)
4764 err = inet6_fill_ifaddr(skb, ifa,
4765 NETLINK_CB(cb->skb).portid,
4771 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4775 case MULTICAST_ADDR:
4776 /* multicast address */
4777 for (ifmca = idev->mc_list; ifmca;
4778 ifmca = ifmca->next, ip_idx++) {
4779 if (ip_idx < s_ip_idx)
4781 err = inet6_fill_ifmcaddr(skb, ifmca,
4782 NETLINK_CB(cb->skb).portid,
4791 /* anycast address */
4792 for (ifaca = idev->ac_list; ifaca;
4793 ifaca = ifaca->aca_next, ip_idx++) {
4794 if (ip_idx < s_ip_idx)
4796 err = inet6_fill_ifacaddr(skb, ifaca,
4797 NETLINK_CB(cb->skb).portid,
4808 read_unlock_bh(&idev->lock);
4813 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4814 enum addr_type_t type)
4816 struct net *net = sock_net(skb->sk);
4819 int s_idx, s_ip_idx;
4820 struct net_device *dev;
4821 struct inet6_dev *idev;
4822 struct hlist_head *head;
4825 s_idx = idx = cb->args[1];
4826 s_ip_idx = ip_idx = cb->args[2];
4829 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4830 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4832 head = &net->dev_index_head[h];
4833 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4836 if (h > s_h || idx > s_idx)
4839 idev = __in6_dev_get(dev);
4843 if (in6_dump_addrs(idev, skb, cb, type,
4844 s_ip_idx, &ip_idx) < 0)
4854 cb->args[2] = ip_idx;
4859 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4861 enum addr_type_t type = UNICAST_ADDR;
4863 return inet6_dump_addr(skb, cb, type);
4866 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4868 enum addr_type_t type = MULTICAST_ADDR;
4870 return inet6_dump_addr(skb, cb, type);
4874 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4876 enum addr_type_t type = ANYCAST_ADDR;
4878 return inet6_dump_addr(skb, cb, type);
4881 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4882 struct netlink_ext_ack *extack)
4884 struct net *net = sock_net(in_skb->sk);
4885 struct ifaddrmsg *ifm;
4886 struct nlattr *tb[IFA_MAX+1];
4887 struct in6_addr *addr = NULL, *peer;
4888 struct net_device *dev = NULL;
4889 struct inet6_ifaddr *ifa;
4890 struct sk_buff *skb;
4893 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4898 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4904 ifm = nlmsg_data(nlh);
4906 dev = __dev_get_by_index(net, ifm->ifa_index);
4908 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4910 err = -EADDRNOTAVAIL;
4914 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4920 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4921 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4923 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4924 WARN_ON(err == -EMSGSIZE);
4928 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4935 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4937 struct sk_buff *skb;
4938 struct net *net = dev_net(ifa->idev->dev);
4941 /* Don't send DELADDR notification for TENTATIVE address,
4942 * since NEWADDR notification is sent only after removing
4943 * TENTATIVE flag, if DAD has not failed.
4945 if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
4946 event == RTM_DELADDR)
4949 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4953 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4955 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4956 WARN_ON(err == -EMSGSIZE);
4960 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4964 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4967 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4968 __s32 *array, int bytes)
4970 BUG_ON(bytes < (DEVCONF_MAX * 4));
4972 memset(array, 0, bytes);
4973 array[DEVCONF_FORWARDING] = cnf->forwarding;
4974 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4975 array[DEVCONF_MTU6] = cnf->mtu6;
4976 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4977 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4978 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4979 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4980 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4981 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4982 jiffies_to_msecs(cnf->rtr_solicit_interval);
4983 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4984 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4985 array[DEVCONF_RTR_SOLICIT_DELAY] =
4986 jiffies_to_msecs(cnf->rtr_solicit_delay);
4987 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4988 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4989 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4990 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4991 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4992 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4993 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4994 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4995 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4996 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4997 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4998 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4999 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5000 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5001 #ifdef CONFIG_IPV6_ROUTER_PREF
5002 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5003 array[DEVCONF_RTR_PROBE_INTERVAL] =
5004 jiffies_to_msecs(cnf->rtr_probe_interval);
5005 #ifdef CONFIG_IPV6_ROUTE_INFO
5006 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5007 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5010 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5011 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5012 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5013 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5014 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5016 #ifdef CONFIG_IPV6_MROUTE
5017 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5019 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5020 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5021 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5022 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5023 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5024 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5025 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5026 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5027 /* we omit DEVCONF_STABLE_SECRET for now */
5028 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5029 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5030 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5031 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5032 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5033 #ifdef CONFIG_IPV6_SEG6_HMAC
5034 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5036 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5037 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5038 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5041 static inline size_t inet6_ifla6_size(void)
5043 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5044 + nla_total_size(sizeof(struct ifla_cacheinfo))
5045 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5046 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5047 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5048 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5051 static inline size_t inet6_if_nlmsg_size(void)
5053 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5054 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5055 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5056 + nla_total_size(4) /* IFLA_MTU */
5057 + nla_total_size(4) /* IFLA_LINK */
5058 + nla_total_size(1) /* IFLA_OPERSTATE */
5059 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5062 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5066 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5069 /* Use put_unaligned() because stats may not be aligned for u64. */
5070 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5071 for (i = 1; i < ICMP6_MIB_MAX; i++)
5072 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5074 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5077 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5078 int bytes, size_t syncpoff)
5081 u64 buff[IPSTATS_MIB_MAX];
5082 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5086 memset(buff, 0, sizeof(buff));
5087 buff[0] = IPSTATS_MIB_MAX;
5089 for_each_possible_cpu(c) {
5090 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5091 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5094 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5095 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5098 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5102 case IFLA_INET6_STATS:
5103 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5104 offsetof(struct ipstats_mib, syncp));
5106 case IFLA_INET6_ICMP6STATS:
5107 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5112 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5113 u32 ext_filter_mask)
5116 struct ifla_cacheinfo ci;
5118 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5119 goto nla_put_failure;
5120 ci.max_reasm_len = IPV6_MAXPLEN;
5121 ci.tstamp = cstamp_delta(idev->tstamp);
5122 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5123 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5124 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5125 goto nla_put_failure;
5126 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5128 goto nla_put_failure;
5129 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5131 /* XXX - MC not implemented */
5133 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5136 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5138 goto nla_put_failure;
5139 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5141 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5143 goto nla_put_failure;
5144 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5146 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5148 goto nla_put_failure;
5150 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5151 goto nla_put_failure;
5153 read_lock_bh(&idev->lock);
5154 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5155 read_unlock_bh(&idev->lock);
5163 static size_t inet6_get_link_af_size(const struct net_device *dev,
5164 u32 ext_filter_mask)
5166 if (!__in6_dev_get(dev))
5169 return inet6_ifla6_size();
5172 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5173 u32 ext_filter_mask)
5175 struct inet6_dev *idev = __in6_dev_get(dev);
5180 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5186 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5188 struct inet6_ifaddr *ifp;
5189 struct net_device *dev = idev->dev;
5190 bool clear_token, update_rs = false;
5191 struct in6_addr ll_addr;
5197 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5199 if (!ipv6_accept_ra(idev))
5201 if (idev->cnf.rtr_solicits == 0)
5204 write_lock_bh(&idev->lock);
5206 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5207 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5209 write_unlock_bh(&idev->lock);
5211 clear_token = ipv6_addr_any(token);
5215 if (!idev->dead && (idev->if_flags & IF_READY) &&
5216 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5217 IFA_F_OPTIMISTIC)) {
5218 /* If we're not ready, then normal ifup will take care
5219 * of this. Otherwise, we need to request our rs here.
5221 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5226 write_lock_bh(&idev->lock);
5229 idev->if_flags |= IF_RS_SENT;
5230 idev->rs_interval = rfc3315_s14_backoff_init(
5231 idev->cnf.rtr_solicit_interval);
5232 idev->rs_probes = 1;
5233 addrconf_mod_rs_timer(idev, idev->rs_interval);
5236 /* Well, that's kinda nasty ... */
5237 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5238 spin_lock(&ifp->lock);
5239 if (ifp->tokenized) {
5241 ifp->prefered_lft = 0;
5243 spin_unlock(&ifp->lock);
5246 write_unlock_bh(&idev->lock);
5247 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5248 addrconf_verify_rtnl();
5252 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5253 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5254 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5257 static int inet6_validate_link_af(const struct net_device *dev,
5258 const struct nlattr *nla)
5260 struct nlattr *tb[IFLA_INET6_MAX + 1];
5262 if (dev && !__in6_dev_get(dev))
5263 return -EAFNOSUPPORT;
5265 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5269 static int check_addr_gen_mode(int mode)
5271 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5272 mode != IN6_ADDR_GEN_MODE_NONE &&
5273 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5274 mode != IN6_ADDR_GEN_MODE_RANDOM)
5279 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5282 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5283 !idev->cnf.stable_secret.initialized &&
5284 !net->ipv6.devconf_dflt->stable_secret.initialized)
5289 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5292 struct inet6_dev *idev = __in6_dev_get(dev);
5293 struct nlattr *tb[IFLA_INET6_MAX + 1];
5296 return -EAFNOSUPPORT;
5298 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5301 if (tb[IFLA_INET6_TOKEN]) {
5302 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5307 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5308 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5310 if (check_addr_gen_mode(mode) < 0 ||
5311 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5314 idev->cnf.addr_gen_mode = mode;
5321 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5322 u32 portid, u32 seq, int event, unsigned int flags)
5324 struct net_device *dev = idev->dev;
5325 struct ifinfomsg *hdr;
5326 struct nlmsghdr *nlh;
5329 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5333 hdr = nlmsg_data(nlh);
5334 hdr->ifi_family = AF_INET6;
5336 hdr->ifi_type = dev->type;
5337 hdr->ifi_index = dev->ifindex;
5338 hdr->ifi_flags = dev_get_flags(dev);
5339 hdr->ifi_change = 0;
5341 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5343 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5344 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5345 (dev->ifindex != dev_get_iflink(dev) &&
5346 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5347 nla_put_u8(skb, IFLA_OPERSTATE,
5348 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5349 goto nla_put_failure;
5350 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5352 goto nla_put_failure;
5354 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5355 goto nla_put_failure;
5357 nla_nest_end(skb, protoinfo);
5358 nlmsg_end(skb, nlh);
5362 nlmsg_cancel(skb, nlh);
5366 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5368 struct net *net = sock_net(skb->sk);
5371 struct net_device *dev;
5372 struct inet6_dev *idev;
5373 struct hlist_head *head;
5376 s_idx = cb->args[1];
5379 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5381 head = &net->dev_index_head[h];
5382 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5385 idev = __in6_dev_get(dev);
5388 if (inet6_fill_ifinfo(skb, idev,
5389 NETLINK_CB(cb->skb).portid,
5391 RTM_NEWLINK, NLM_F_MULTI) < 0)
5405 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5407 struct sk_buff *skb;
5408 struct net *net = dev_net(idev->dev);
5411 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5415 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5417 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5418 WARN_ON(err == -EMSGSIZE);
5422 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5426 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5429 static inline size_t inet6_prefix_nlmsg_size(void)
5431 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5432 + nla_total_size(sizeof(struct in6_addr))
5433 + nla_total_size(sizeof(struct prefix_cacheinfo));
5436 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5437 struct prefix_info *pinfo, u32 portid, u32 seq,
5438 int event, unsigned int flags)
5440 struct prefixmsg *pmsg;
5441 struct nlmsghdr *nlh;
5442 struct prefix_cacheinfo ci;
5444 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5448 pmsg = nlmsg_data(nlh);
5449 pmsg->prefix_family = AF_INET6;
5450 pmsg->prefix_pad1 = 0;
5451 pmsg->prefix_pad2 = 0;
5452 pmsg->prefix_ifindex = idev->dev->ifindex;
5453 pmsg->prefix_len = pinfo->prefix_len;
5454 pmsg->prefix_type = pinfo->type;
5455 pmsg->prefix_pad3 = 0;
5456 pmsg->prefix_flags = 0;
5458 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5459 if (pinfo->autoconf)
5460 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5462 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5463 goto nla_put_failure;
5464 ci.preferred_time = ntohl(pinfo->prefered);
5465 ci.valid_time = ntohl(pinfo->valid);
5466 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5467 goto nla_put_failure;
5468 nlmsg_end(skb, nlh);
5472 nlmsg_cancel(skb, nlh);
5476 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5477 struct prefix_info *pinfo)
5479 struct sk_buff *skb;
5480 struct net *net = dev_net(idev->dev);
5483 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5487 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5489 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5490 WARN_ON(err == -EMSGSIZE);
5494 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5498 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5501 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5503 struct net *net = dev_net(ifp->idev->dev);
5508 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5513 * If the address was optimistic
5514 * we inserted the route at the start of
5515 * our DAD process, so we don't need
5518 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5519 ip6_ins_rt(ifp->rt);
5520 if (ifp->idev->cnf.forwarding)
5521 addrconf_join_anycast(ifp);
5522 if (!ipv6_addr_any(&ifp->peer_addr))
5523 addrconf_prefix_route(&ifp->peer_addr, 128,
5524 ifp->idev->dev, 0, 0);
5527 if (ifp->idev->cnf.forwarding)
5528 addrconf_leave_anycast(ifp);
5529 addrconf_leave_solict(ifp->idev, &ifp->addr);
5530 if (!ipv6_addr_any(&ifp->peer_addr)) {
5531 struct rt6_info *rt;
5533 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5534 ifp->idev->dev, 0, 0);
5539 if (dst_hold_safe(&ifp->rt->dst))
5540 ip6_del_rt(ifp->rt);
5542 rt_genid_bump_ipv6(net);
5545 atomic_inc(&net->ipv6.dev_addr_genid);
5548 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5551 if (likely(ifp->idev->dead == 0))
5552 __ipv6_ifa_notify(event, ifp);
5553 rcu_read_unlock_bh();
5556 #ifdef CONFIG_SYSCTL
5559 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5560 void __user *buffer, size_t *lenp, loff_t *ppos)
5562 int *valp = ctl->data;
5565 struct ctl_table lctl;
5569 * ctl->data points to idev->cnf.forwarding, we should
5570 * not modify it until we get the rtnl lock.
5575 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5578 ret = addrconf_fixup_forwarding(ctl, valp, val);
5585 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5586 void __user *buffer, size_t *lenp, loff_t *ppos)
5588 struct inet6_dev *idev = ctl->extra1;
5589 int min_mtu = IPV6_MIN_MTU;
5590 struct ctl_table lctl;
5593 lctl.extra1 = &min_mtu;
5594 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5596 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5599 static void dev_disable_change(struct inet6_dev *idev)
5601 struct netdev_notifier_info info;
5603 if (!idev || !idev->dev)
5606 netdev_notifier_info_init(&info, idev->dev);
5607 if (idev->cnf.disable_ipv6)
5608 addrconf_notify(NULL, NETDEV_DOWN, &info);
5610 addrconf_notify(NULL, NETDEV_UP, &info);
5613 static void addrconf_disable_change(struct net *net, __s32 newf)
5615 struct net_device *dev;
5616 struct inet6_dev *idev;
5618 for_each_netdev(net, dev) {
5619 idev = __in6_dev_get(dev);
5621 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5622 idev->cnf.disable_ipv6 = newf;
5624 dev_disable_change(idev);
5629 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5634 if (!rtnl_trylock())
5635 return restart_syscall();
5637 net = (struct net *)table->extra2;
5641 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5646 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5647 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5648 addrconf_disable_change(net, newf);
5649 } else if ((!newf) ^ (!old))
5650 dev_disable_change((struct inet6_dev *)table->extra1);
5657 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5658 void __user *buffer, size_t *lenp, loff_t *ppos)
5660 int *valp = ctl->data;
5663 struct ctl_table lctl;
5667 * ctl->data points to idev->cnf.disable_ipv6, we should
5668 * not modify it until we get the rtnl lock.
5673 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5676 ret = addrconf_disable_ipv6(ctl, valp, val);
5683 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5684 void __user *buffer, size_t *lenp, loff_t *ppos)
5686 int *valp = ctl->data;
5691 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5694 if (write && old != new) {
5695 struct net *net = ctl->extra2;
5697 if (!rtnl_trylock())
5698 return restart_syscall();
5700 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5701 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5702 NETCONFA_PROXY_NEIGH,
5703 NETCONFA_IFINDEX_DEFAULT,
5704 net->ipv6.devconf_dflt);
5705 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5706 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5707 NETCONFA_PROXY_NEIGH,
5708 NETCONFA_IFINDEX_ALL,
5709 net->ipv6.devconf_all);
5711 struct inet6_dev *idev = ctl->extra1;
5713 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5714 NETCONFA_PROXY_NEIGH,
5724 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5725 void __user *buffer, size_t *lenp,
5730 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5731 struct net *net = (struct net *)ctl->extra2;
5733 if (!rtnl_trylock())
5734 return restart_syscall();
5736 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5739 new_val = *((int *)ctl->data);
5741 if (check_addr_gen_mode(new_val) < 0) {
5746 /* request for default */
5747 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5748 ipv6_devconf_dflt.addr_gen_mode = new_val;
5750 /* request for individual net device */
5755 if (check_stable_privacy(idev, net, new_val) < 0) {
5760 if (idev->cnf.addr_gen_mode != new_val) {
5761 idev->cnf.addr_gen_mode = new_val;
5762 addrconf_dev_config(idev->dev);
5773 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5774 void __user *buffer, size_t *lenp,
5778 struct in6_addr addr;
5779 char str[IPV6_MAX_STRLEN];
5780 struct ctl_table lctl = *ctl;
5781 struct net *net = ctl->extra2;
5782 struct ipv6_stable_secret *secret = ctl->data;
5784 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5787 lctl.maxlen = IPV6_MAX_STRLEN;
5790 if (!rtnl_trylock())
5791 return restart_syscall();
5793 if (!write && !secret->initialized) {
5798 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5799 if (err >= sizeof(str)) {
5804 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5808 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5813 secret->initialized = true;
5814 secret->secret = addr;
5816 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5817 struct net_device *dev;
5819 for_each_netdev(net, dev) {
5820 struct inet6_dev *idev = __in6_dev_get(dev);
5823 idev->cnf.addr_gen_mode =
5824 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5828 struct inet6_dev *idev = ctl->extra1;
5830 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5840 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5842 void __user *buffer,
5846 int *valp = ctl->data;
5849 struct ctl_table lctl;
5852 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5853 * we should not modify it until we get the rtnl lock.
5858 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5861 ret = addrconf_fixup_linkdown(ctl, valp, val);
5868 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5872 rt->dst.flags |= DST_NOPOLICY;
5874 rt->dst.flags &= ~DST_NOPOLICY;
5879 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5881 struct inet6_ifaddr *ifa;
5883 read_lock_bh(&idev->lock);
5884 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5885 spin_lock(&ifa->lock);
5887 struct rt6_info *rt = ifa->rt;
5888 struct fib6_table *table = rt->rt6i_table;
5891 read_lock(&table->tb6_lock);
5892 addrconf_set_nopolicy(ifa->rt, val);
5893 if (rt->rt6i_pcpu) {
5894 for_each_possible_cpu(cpu) {
5895 struct rt6_info **rtp;
5897 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5898 addrconf_set_nopolicy(*rtp, val);
5901 read_unlock(&table->tb6_lock);
5903 spin_unlock(&ifa->lock);
5905 read_unlock_bh(&idev->lock);
5909 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5911 struct inet6_dev *idev;
5914 if (!rtnl_trylock())
5915 return restart_syscall();
5919 net = (struct net *)ctl->extra2;
5920 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5925 if (valp == &net->ipv6.devconf_all->disable_policy) {
5926 struct net_device *dev;
5928 for_each_netdev(net, dev) {
5929 idev = __in6_dev_get(dev);
5931 addrconf_disable_policy_idev(idev, val);
5934 idev = (struct inet6_dev *)ctl->extra1;
5935 addrconf_disable_policy_idev(idev, val);
5943 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
5944 void __user *buffer, size_t *lenp,
5947 int *valp = ctl->data;
5950 struct ctl_table lctl;
5955 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5957 if (write && (*valp != val))
5958 ret = addrconf_disable_policy(ctl, valp, val);
5966 static int minus_one = -1;
5967 static const int one = 1;
5968 static const int two_five_five = 255;
5970 static const struct ctl_table addrconf_sysctl[] = {
5972 .procname = "forwarding",
5973 .data = &ipv6_devconf.forwarding,
5974 .maxlen = sizeof(int),
5976 .proc_handler = addrconf_sysctl_forward,
5979 .procname = "hop_limit",
5980 .data = &ipv6_devconf.hop_limit,
5981 .maxlen = sizeof(int),
5983 .proc_handler = proc_dointvec_minmax,
5984 .extra1 = (void *)&one,
5985 .extra2 = (void *)&two_five_five,
5989 .data = &ipv6_devconf.mtu6,
5990 .maxlen = sizeof(int),
5992 .proc_handler = addrconf_sysctl_mtu,
5995 .procname = "accept_ra",
5996 .data = &ipv6_devconf.accept_ra,
5997 .maxlen = sizeof(int),
5999 .proc_handler = proc_dointvec,
6002 .procname = "accept_redirects",
6003 .data = &ipv6_devconf.accept_redirects,
6004 .maxlen = sizeof(int),
6006 .proc_handler = proc_dointvec,
6009 .procname = "autoconf",
6010 .data = &ipv6_devconf.autoconf,
6011 .maxlen = sizeof(int),
6013 .proc_handler = proc_dointvec,
6016 .procname = "dad_transmits",
6017 .data = &ipv6_devconf.dad_transmits,
6018 .maxlen = sizeof(int),
6020 .proc_handler = proc_dointvec,
6023 .procname = "router_solicitations",
6024 .data = &ipv6_devconf.rtr_solicits,
6025 .maxlen = sizeof(int),
6027 .proc_handler = proc_dointvec_minmax,
6028 .extra1 = &minus_one,
6031 .procname = "router_solicitation_interval",
6032 .data = &ipv6_devconf.rtr_solicit_interval,
6033 .maxlen = sizeof(int),
6035 .proc_handler = proc_dointvec_jiffies,
6038 .procname = "router_solicitation_max_interval",
6039 .data = &ipv6_devconf.rtr_solicit_max_interval,
6040 .maxlen = sizeof(int),
6042 .proc_handler = proc_dointvec_jiffies,
6045 .procname = "router_solicitation_delay",
6046 .data = &ipv6_devconf.rtr_solicit_delay,
6047 .maxlen = sizeof(int),
6049 .proc_handler = proc_dointvec_jiffies,
6052 .procname = "force_mld_version",
6053 .data = &ipv6_devconf.force_mld_version,
6054 .maxlen = sizeof(int),
6056 .proc_handler = proc_dointvec,
6059 .procname = "mldv1_unsolicited_report_interval",
6061 &ipv6_devconf.mldv1_unsolicited_report_interval,
6062 .maxlen = sizeof(int),
6064 .proc_handler = proc_dointvec_ms_jiffies,
6067 .procname = "mldv2_unsolicited_report_interval",
6069 &ipv6_devconf.mldv2_unsolicited_report_interval,
6070 .maxlen = sizeof(int),
6072 .proc_handler = proc_dointvec_ms_jiffies,
6075 .procname = "use_tempaddr",
6076 .data = &ipv6_devconf.use_tempaddr,
6077 .maxlen = sizeof(int),
6079 .proc_handler = proc_dointvec,
6082 .procname = "temp_valid_lft",
6083 .data = &ipv6_devconf.temp_valid_lft,
6084 .maxlen = sizeof(int),
6086 .proc_handler = proc_dointvec,
6089 .procname = "temp_prefered_lft",
6090 .data = &ipv6_devconf.temp_prefered_lft,
6091 .maxlen = sizeof(int),
6093 .proc_handler = proc_dointvec,
6096 .procname = "regen_max_retry",
6097 .data = &ipv6_devconf.regen_max_retry,
6098 .maxlen = sizeof(int),
6100 .proc_handler = proc_dointvec,
6103 .procname = "max_desync_factor",
6104 .data = &ipv6_devconf.max_desync_factor,
6105 .maxlen = sizeof(int),
6107 .proc_handler = proc_dointvec,
6110 .procname = "max_addresses",
6111 .data = &ipv6_devconf.max_addresses,
6112 .maxlen = sizeof(int),
6114 .proc_handler = proc_dointvec,
6117 .procname = "accept_ra_defrtr",
6118 .data = &ipv6_devconf.accept_ra_defrtr,
6119 .maxlen = sizeof(int),
6121 .proc_handler = proc_dointvec,
6124 .procname = "accept_ra_min_hop_limit",
6125 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6126 .maxlen = sizeof(int),
6128 .proc_handler = proc_dointvec,
6131 .procname = "accept_ra_pinfo",
6132 .data = &ipv6_devconf.accept_ra_pinfo,
6133 .maxlen = sizeof(int),
6135 .proc_handler = proc_dointvec,
6137 #ifdef CONFIG_IPV6_ROUTER_PREF
6139 .procname = "accept_ra_rtr_pref",
6140 .data = &ipv6_devconf.accept_ra_rtr_pref,
6141 .maxlen = sizeof(int),
6143 .proc_handler = proc_dointvec,
6146 .procname = "router_probe_interval",
6147 .data = &ipv6_devconf.rtr_probe_interval,
6148 .maxlen = sizeof(int),
6150 .proc_handler = proc_dointvec_jiffies,
6152 #ifdef CONFIG_IPV6_ROUTE_INFO
6154 .procname = "accept_ra_rt_info_min_plen",
6155 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6156 .maxlen = sizeof(int),
6158 .proc_handler = proc_dointvec,
6161 .procname = "accept_ra_rt_info_max_plen",
6162 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6163 .maxlen = sizeof(int),
6165 .proc_handler = proc_dointvec,
6170 .procname = "proxy_ndp",
6171 .data = &ipv6_devconf.proxy_ndp,
6172 .maxlen = sizeof(int),
6174 .proc_handler = addrconf_sysctl_proxy_ndp,
6177 .procname = "accept_source_route",
6178 .data = &ipv6_devconf.accept_source_route,
6179 .maxlen = sizeof(int),
6181 .proc_handler = proc_dointvec,
6183 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6185 .procname = "optimistic_dad",
6186 .data = &ipv6_devconf.optimistic_dad,
6187 .maxlen = sizeof(int),
6189 .proc_handler = proc_dointvec,
6192 .procname = "use_optimistic",
6193 .data = &ipv6_devconf.use_optimistic,
6194 .maxlen = sizeof(int),
6196 .proc_handler = proc_dointvec,
6199 #ifdef CONFIG_IPV6_MROUTE
6201 .procname = "mc_forwarding",
6202 .data = &ipv6_devconf.mc_forwarding,
6203 .maxlen = sizeof(int),
6205 .proc_handler = proc_dointvec,
6209 .procname = "disable_ipv6",
6210 .data = &ipv6_devconf.disable_ipv6,
6211 .maxlen = sizeof(int),
6213 .proc_handler = addrconf_sysctl_disable,
6216 .procname = "accept_dad",
6217 .data = &ipv6_devconf.accept_dad,
6218 .maxlen = sizeof(int),
6220 .proc_handler = proc_dointvec,
6223 .procname = "force_tllao",
6224 .data = &ipv6_devconf.force_tllao,
6225 .maxlen = sizeof(int),
6227 .proc_handler = proc_dointvec
6230 .procname = "ndisc_notify",
6231 .data = &ipv6_devconf.ndisc_notify,
6232 .maxlen = sizeof(int),
6234 .proc_handler = proc_dointvec
6237 .procname = "suppress_frag_ndisc",
6238 .data = &ipv6_devconf.suppress_frag_ndisc,
6239 .maxlen = sizeof(int),
6241 .proc_handler = proc_dointvec
6244 .procname = "accept_ra_from_local",
6245 .data = &ipv6_devconf.accept_ra_from_local,
6246 .maxlen = sizeof(int),
6248 .proc_handler = proc_dointvec,
6251 .procname = "accept_ra_mtu",
6252 .data = &ipv6_devconf.accept_ra_mtu,
6253 .maxlen = sizeof(int),
6255 .proc_handler = proc_dointvec,
6258 .procname = "stable_secret",
6259 .data = &ipv6_devconf.stable_secret,
6260 .maxlen = IPV6_MAX_STRLEN,
6262 .proc_handler = addrconf_sysctl_stable_secret,
6265 .procname = "use_oif_addrs_only",
6266 .data = &ipv6_devconf.use_oif_addrs_only,
6267 .maxlen = sizeof(int),
6269 .proc_handler = proc_dointvec,
6272 .procname = "ignore_routes_with_linkdown",
6273 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6274 .maxlen = sizeof(int),
6276 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6279 .procname = "drop_unicast_in_l2_multicast",
6280 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6281 .maxlen = sizeof(int),
6283 .proc_handler = proc_dointvec,
6286 .procname = "drop_unsolicited_na",
6287 .data = &ipv6_devconf.drop_unsolicited_na,
6288 .maxlen = sizeof(int),
6290 .proc_handler = proc_dointvec,
6293 .procname = "keep_addr_on_down",
6294 .data = &ipv6_devconf.keep_addr_on_down,
6295 .maxlen = sizeof(int),
6297 .proc_handler = proc_dointvec,
6301 .procname = "seg6_enabled",
6302 .data = &ipv6_devconf.seg6_enabled,
6303 .maxlen = sizeof(int),
6305 .proc_handler = proc_dointvec,
6307 #ifdef CONFIG_IPV6_SEG6_HMAC
6309 .procname = "seg6_require_hmac",
6310 .data = &ipv6_devconf.seg6_require_hmac,
6311 .maxlen = sizeof(int),
6313 .proc_handler = proc_dointvec,
6317 .procname = "enhanced_dad",
6318 .data = &ipv6_devconf.enhanced_dad,
6319 .maxlen = sizeof(int),
6321 .proc_handler = proc_dointvec,
6324 .procname = "addr_gen_mode",
6325 .data = &ipv6_devconf.addr_gen_mode,
6326 .maxlen = sizeof(int),
6328 .proc_handler = addrconf_sysctl_addr_gen_mode,
6331 .procname = "disable_policy",
6332 .data = &ipv6_devconf.disable_policy,
6333 .maxlen = sizeof(int),
6335 .proc_handler = addrconf_sysctl_disable_policy,
6342 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6343 struct inet6_dev *idev, struct ipv6_devconf *p)
6346 struct ctl_table *table;
6347 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6349 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6353 for (i = 0; table[i].data; i++) {
6354 table[i].data += (char *)p - (char *)&ipv6_devconf;
6355 /* If one of these is already set, then it is not safe to
6356 * overwrite either of them: this makes proc_dointvec_minmax
6359 if (!table[i].extra1 && !table[i].extra2) {
6360 table[i].extra1 = idev; /* embedded; no ref */
6361 table[i].extra2 = net;
6365 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6367 p->sysctl_header = register_net_sysctl(net, path, table);
6368 if (!p->sysctl_header)
6371 if (!strcmp(dev_name, "all"))
6372 ifindex = NETCONFA_IFINDEX_ALL;
6373 else if (!strcmp(dev_name, "default"))
6374 ifindex = NETCONFA_IFINDEX_DEFAULT;
6376 ifindex = idev->dev->ifindex;
6377 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6387 static void __addrconf_sysctl_unregister(struct net *net,
6388 struct ipv6_devconf *p, int ifindex)
6390 struct ctl_table *table;
6392 if (!p->sysctl_header)
6395 table = p->sysctl_header->ctl_table_arg;
6396 unregister_net_sysctl_table(p->sysctl_header);
6397 p->sysctl_header = NULL;
6400 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6403 static int addrconf_sysctl_register(struct inet6_dev *idev)
6407 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6410 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6411 &ndisc_ifinfo_sysctl_change);
6414 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6417 neigh_sysctl_unregister(idev->nd_parms);
6422 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6424 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6425 idev->dev->ifindex);
6426 neigh_sysctl_unregister(idev->nd_parms);
6432 static int __net_init addrconf_init_net(struct net *net)
6435 struct ipv6_devconf *all, *dflt;
6437 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6441 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6443 goto err_alloc_dflt;
6445 /* these will be inherited by all namespaces */
6446 dflt->autoconf = ipv6_defaults.autoconf;
6447 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6449 dflt->stable_secret.initialized = false;
6450 all->stable_secret.initialized = false;
6452 net->ipv6.devconf_all = all;
6453 net->ipv6.devconf_dflt = dflt;
6455 #ifdef CONFIG_SYSCTL
6456 err = __addrconf_sysctl_register(net, "all", NULL, all);
6460 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6466 #ifdef CONFIG_SYSCTL
6468 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6478 static void __net_exit addrconf_exit_net(struct net *net)
6480 #ifdef CONFIG_SYSCTL
6481 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6482 NETCONFA_IFINDEX_DEFAULT);
6483 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6484 NETCONFA_IFINDEX_ALL);
6486 kfree(net->ipv6.devconf_dflt);
6487 kfree(net->ipv6.devconf_all);
6490 static struct pernet_operations addrconf_ops = {
6491 .init = addrconf_init_net,
6492 .exit = addrconf_exit_net,
6495 static struct rtnl_af_ops inet6_ops __read_mostly = {
6497 .fill_link_af = inet6_fill_link_af,
6498 .get_link_af_size = inet6_get_link_af_size,
6499 .validate_link_af = inet6_validate_link_af,
6500 .set_link_af = inet6_set_link_af,
6504 * Init / cleanup code
6507 int __init addrconf_init(void)
6509 struct inet6_dev *idev;
6512 err = ipv6_addr_label_init();
6514 pr_crit("%s: cannot initialize default policy table: %d\n",
6519 err = register_pernet_subsys(&addrconf_ops);
6523 addrconf_wq = create_workqueue("ipv6_addrconf");
6529 /* The addrconf netdev notifier requires that loopback_dev
6530 * has it's ipv6 private information allocated and setup
6531 * before it can bring up and give link-local addresses
6532 * to other devices which are up.
6534 * Unfortunately, loopback_dev is not necessarily the first
6535 * entry in the global dev_base list of net devices. In fact,
6536 * it is likely to be the very last entry on that list.
6537 * So this causes the notifier registry below to try and
6538 * give link-local addresses to all devices besides loopback_dev
6539 * first, then loopback_dev, which cases all the non-loopback_dev
6540 * devices to fail to get a link-local address.
6542 * So, as a temporary fix, allocate the ipv6 structure for
6543 * loopback_dev first by hand.
6544 * Longer term, all of the dependencies ipv6 has upon the loopback
6545 * device and it being up should be removed.
6548 idev = ipv6_add_dev(init_net.loopback_dev);
6551 err = PTR_ERR(idev);
6555 ip6_route_init_special_entries();
6557 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6558 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6560 register_netdevice_notifier(&ipv6_dev_notf);
6564 rtnl_af_register(&inet6_ops);
6566 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6571 /* Only the first call to __rtnl_register can fail */
6572 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, 0);
6573 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, 0);
6574 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6575 inet6_dump_ifaddr, 0);
6576 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6577 inet6_dump_ifmcaddr, 0);
6578 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6579 inet6_dump_ifacaddr, 0);
6580 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6581 inet6_netconf_dump_devconf, 0);
6583 ipv6_addr_label_rtnl_register();
6587 rtnl_af_unregister(&inet6_ops);
6588 unregister_netdevice_notifier(&ipv6_dev_notf);
6590 destroy_workqueue(addrconf_wq);
6592 unregister_pernet_subsys(&addrconf_ops);
6594 ipv6_addr_label_cleanup();
6599 void addrconf_cleanup(void)
6601 struct net_device *dev;
6604 unregister_netdevice_notifier(&ipv6_dev_notf);
6605 unregister_pernet_subsys(&addrconf_ops);
6606 ipv6_addr_label_cleanup();
6610 __rtnl_af_unregister(&inet6_ops);
6612 /* clean dev list */
6613 for_each_netdev(&init_net, dev) {
6614 if (__in6_dev_get(dev) == NULL)
6616 addrconf_ifdown(dev, 1);
6618 addrconf_ifdown(init_net.loopback_dev, 2);
6623 spin_lock_bh(&addrconf_hash_lock);
6624 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6625 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6626 spin_unlock_bh(&addrconf_hash_lock);
6627 cancel_delayed_work(&addr_chk_work);
6630 destroy_workqueue(addrconf_wq);