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(const struct inet6_dev *idev);
156 static int ipv6_generate_stable_address(struct in6_addr *addr,
158 const struct inet6_dev *idev);
160 #define IN6_ADDR_HSIZE_SHIFT 8
161 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
163 * Configured unicast address hash table
165 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
166 static DEFINE_SPINLOCK(addrconf_hash_lock);
168 static void addrconf_verify(void);
169 static void addrconf_verify_rtnl(void);
170 static void addrconf_verify_work(struct work_struct *);
172 static struct workqueue_struct *addrconf_wq;
173 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
175 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
176 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
178 static void addrconf_type_change(struct net_device *dev,
179 unsigned long event);
180 static int addrconf_ifdown(struct net_device *dev, int how);
182 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
184 const struct net_device *dev,
185 u32 flags, u32 noflags);
187 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
188 static void addrconf_dad_work(struct work_struct *w);
189 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id);
190 static void addrconf_dad_run(struct inet6_dev *idev);
191 static void addrconf_rs_timer(unsigned long data);
192 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
193 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
195 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
196 struct prefix_info *pinfo);
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 link is ready: is it up and is a valid qdisc available */
307 static inline bool addrconf_link_ready(const struct net_device *dev)
309 return netif_oper_up(dev) && !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_link_ready(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 inet6_dev *in6_dev = NULL;
620 struct net_device *dev = NULL;
621 struct netconfmsg *ncm;
623 struct ipv6_devconf *devconf;
627 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
628 devconf_ipv6_policy, extack);
632 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_KERNEL);
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);
673 in6_dev_put(in6_dev);
679 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
680 struct netlink_callback *cb)
682 struct net *net = sock_net(skb->sk);
685 struct net_device *dev;
686 struct inet6_dev *idev;
687 struct hlist_head *head;
690 s_idx = idx = cb->args[1];
692 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
694 head = &net->dev_index_head[h];
696 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
698 hlist_for_each_entry_rcu(dev, head, index_hlist) {
701 idev = __in6_dev_get(dev);
705 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
707 NETLINK_CB(cb->skb).portid,
715 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
721 if (h == NETDEV_HASHENTRIES) {
722 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
723 net->ipv6.devconf_all,
724 NETLINK_CB(cb->skb).portid,
726 RTM_NEWNETCONF, NLM_F_MULTI,
732 if (h == NETDEV_HASHENTRIES + 1) {
733 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
734 net->ipv6.devconf_dflt,
735 NETLINK_CB(cb->skb).portid,
737 RTM_NEWNETCONF, NLM_F_MULTI,
751 static void dev_forward_change(struct inet6_dev *idev)
753 struct net_device *dev;
754 struct inet6_ifaddr *ifa;
759 if (idev->cnf.forwarding)
760 dev_disable_lro(dev);
761 if (dev->flags & IFF_MULTICAST) {
762 if (idev->cnf.forwarding) {
763 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
764 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
765 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
767 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
768 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
769 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
773 list_for_each_entry(ifa, &idev->addr_list, if_list) {
774 if (ifa->flags&IFA_F_TENTATIVE)
776 if (idev->cnf.forwarding)
777 addrconf_join_anycast(ifa);
779 addrconf_leave_anycast(ifa);
781 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
783 dev->ifindex, &idev->cnf);
787 static void addrconf_forward_change(struct net *net, __s32 newf)
789 struct net_device *dev;
790 struct inet6_dev *idev;
792 for_each_netdev(net, dev) {
793 idev = __in6_dev_get(dev);
795 int changed = (!idev->cnf.forwarding) ^ (!newf);
796 idev->cnf.forwarding = newf;
798 dev_forward_change(idev);
803 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
809 return restart_syscall();
811 net = (struct net *)table->extra2;
815 if (p == &net->ipv6.devconf_dflt->forwarding) {
816 if ((!newf) ^ (!old))
817 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
819 NETCONFA_IFINDEX_DEFAULT,
820 net->ipv6.devconf_dflt);
825 if (p == &net->ipv6.devconf_all->forwarding) {
826 int old_dflt = net->ipv6.devconf_dflt->forwarding;
828 net->ipv6.devconf_dflt->forwarding = newf;
829 if ((!newf) ^ (!old_dflt))
830 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
832 NETCONFA_IFINDEX_DEFAULT,
833 net->ipv6.devconf_dflt);
835 addrconf_forward_change(net, newf);
836 if ((!newf) ^ (!old))
837 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
839 NETCONFA_IFINDEX_ALL,
840 net->ipv6.devconf_all);
841 } else if ((!newf) ^ (!old))
842 dev_forward_change((struct inet6_dev *)table->extra1);
846 rt6_purge_dflt_routers(net);
850 static void addrconf_linkdown_change(struct net *net, __s32 newf)
852 struct net_device *dev;
853 struct inet6_dev *idev;
855 for_each_netdev(net, dev) {
856 idev = __in6_dev_get(dev);
858 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
860 idev->cnf.ignore_routes_with_linkdown = newf;
862 inet6_netconf_notify_devconf(dev_net(dev),
864 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
871 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
877 return restart_syscall();
879 net = (struct net *)table->extra2;
883 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
884 if ((!newf) ^ (!old))
885 inet6_netconf_notify_devconf(net,
887 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
888 NETCONFA_IFINDEX_DEFAULT,
889 net->ipv6.devconf_dflt);
894 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
895 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
896 addrconf_linkdown_change(net, newf);
897 if ((!newf) ^ (!old))
898 inet6_netconf_notify_devconf(net,
900 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
901 NETCONFA_IFINDEX_ALL,
902 net->ipv6.devconf_all);
911 /* Nobody refers to this ifaddr, destroy it */
912 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
914 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
916 #ifdef NET_REFCNT_DEBUG
917 pr_debug("%s\n", __func__);
920 in6_dev_put(ifp->idev);
922 if (cancel_delayed_work(&ifp->dad_work))
923 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
926 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
927 pr_warn("Freeing alive inet6 address %p\n", ifp);
936 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
939 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
942 * Each device address list is sorted in order of scope -
943 * global before linklocal.
945 list_for_each(p, &idev->addr_list) {
946 struct inet6_ifaddr *ifa
947 = list_entry(p, struct inet6_ifaddr, if_list);
948 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
952 list_add_tail_rcu(&ifp->if_list, p);
955 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
957 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
959 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
962 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
963 struct net_device *dev, unsigned int hash)
965 struct inet6_ifaddr *ifp;
967 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
968 if (!net_eq(dev_net(ifp->idev->dev), net))
970 if (ipv6_addr_equal(&ifp->addr, addr)) {
971 if (!dev || ifp->idev->dev == dev)
978 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
980 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
983 spin_lock(&addrconf_hash_lock);
985 /* Ignore adding duplicate addresses on an interface */
986 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
987 ADBG("ipv6_add_addr: already assigned\n");
990 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
993 spin_unlock(&addrconf_hash_lock);
998 /* On success it returns ifp with increased reference count */
1000 static struct inet6_ifaddr *
1001 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
1002 const struct in6_addr *peer_addr, int pfxlen,
1003 int scope, u32 flags, u32 valid_lft, u32 prefered_lft,
1004 bool can_block, struct netlink_ext_ack *extack)
1006 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1007 struct net *net = dev_net(idev->dev);
1008 struct inet6_ifaddr *ifa = NULL;
1009 struct rt6_info *rt = NULL;
1011 int addr_type = ipv6_addr_type(addr);
1013 if (addr_type == IPV6_ADDR_ANY ||
1014 addr_type & IPV6_ADDR_MULTICAST ||
1015 (!(idev->dev->flags & IFF_LOOPBACK) &&
1016 addr_type & IPV6_ADDR_LOOPBACK))
1017 return ERR_PTR(-EADDRNOTAVAIL);
1020 err = -ENODEV; /*XXX*/
1024 if (idev->cnf.disable_ipv6) {
1029 /* validator notifier needs to be blocking;
1030 * do not call in atomic context
1033 struct in6_validator_info i6vi = {
1039 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1040 err = notifier_to_errno(err);
1045 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1047 ADBG("ipv6_add_addr: malloc failed\n");
1052 rt = addrconf_dst_alloc(idev, addr, false);
1059 if (net->ipv6.devconf_all->disable_policy ||
1060 idev->cnf.disable_policy)
1061 rt->dst.flags |= DST_NOPOLICY;
1063 neigh_parms_data_state_setall(idev->nd_parms);
1067 ifa->peer_addr = *peer_addr;
1069 spin_lock_init(&ifa->lock);
1070 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1071 INIT_HLIST_NODE(&ifa->addr_lst);
1073 ifa->prefix_len = pfxlen;
1075 /* No need to add the TENTATIVE flag for addresses with NODAD */
1076 if (!(flags & IFA_F_NODAD))
1077 ifa->flags |= IFA_F_TENTATIVE;
1078 ifa->valid_lft = valid_lft;
1079 ifa->prefered_lft = prefered_lft;
1080 ifa->cstamp = ifa->tstamp = jiffies;
1081 ifa->tokenized = false;
1089 refcount_set(&ifa->refcnt, 1);
1093 err = ipv6_add_addr_hash(idev->dev, ifa);
1095 rcu_read_unlock_bh();
1099 write_lock(&idev->lock);
1101 /* Add to inet6_dev unicast addr list. */
1102 ipv6_link_dev_addr(idev, ifa);
1104 if (ifa->flags&IFA_F_TEMPORARY) {
1105 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1110 write_unlock(&idev->lock);
1112 rcu_read_unlock_bh();
1114 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1116 if (unlikely(err < 0)) {
1121 in6_dev_put(ifa->idev);
1130 enum cleanup_prefix_rt_t {
1131 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1132 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1133 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1137 * Check, whether the prefix for ifp would still need a prefix route
1138 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1141 * 1) we don't purge prefix if address was not permanent.
1142 * prefix is managed by its own lifetime.
1143 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1144 * 3) if there are no addresses, delete prefix.
1145 * 4) if there are still other permanent address(es),
1146 * corresponding prefix is still permanent.
1147 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1148 * don't purge the prefix, assume user space is managing it.
1149 * 6) otherwise, update prefix lifetime to the
1150 * longest valid lifetime among the corresponding
1151 * addresses on the device.
1152 * Note: subsequent RA will update lifetime.
1154 static enum cleanup_prefix_rt_t
1155 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1157 struct inet6_ifaddr *ifa;
1158 struct inet6_dev *idev = ifp->idev;
1159 unsigned long lifetime;
1160 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1164 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1167 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1170 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1171 return CLEANUP_PREFIX_RT_NOP;
1173 action = CLEANUP_PREFIX_RT_EXPIRE;
1175 spin_lock(&ifa->lock);
1177 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1179 * Note: Because this address is
1180 * not permanent, lifetime <
1181 * LONG_MAX / HZ here.
1183 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1184 *expires = ifa->tstamp + lifetime * HZ;
1185 spin_unlock(&ifa->lock);
1192 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1194 struct rt6_info *rt;
1196 rt = addrconf_get_prefix_route(&ifp->addr,
1199 0, RTF_GATEWAY | RTF_DEFAULT);
1204 if (!(rt->rt6i_flags & RTF_EXPIRES))
1205 rt6_set_expires(rt, expires);
1212 /* This function wants to get referenced ifp and releases it before return */
1214 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1217 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1218 unsigned long expires;
1222 spin_lock_bh(&ifp->lock);
1224 ifp->state = INET6_IFADDR_STATE_DEAD;
1225 spin_unlock_bh(&ifp->lock);
1227 if (state == INET6_IFADDR_STATE_DEAD)
1230 spin_lock_bh(&addrconf_hash_lock);
1231 hlist_del_init_rcu(&ifp->addr_lst);
1232 spin_unlock_bh(&addrconf_hash_lock);
1234 write_lock_bh(&ifp->idev->lock);
1236 if (ifp->flags&IFA_F_TEMPORARY) {
1237 list_del(&ifp->tmp_list);
1239 in6_ifa_put(ifp->ifpub);
1245 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1246 action = check_cleanup_prefix_route(ifp, &expires);
1248 list_del_rcu(&ifp->if_list);
1251 write_unlock_bh(&ifp->idev->lock);
1253 addrconf_del_dad_work(ifp);
1255 ipv6_ifa_notify(RTM_DELADDR, ifp);
1257 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1259 if (action != CLEANUP_PREFIX_RT_NOP) {
1260 cleanup_prefix_route(ifp, expires,
1261 action == CLEANUP_PREFIX_RT_DEL);
1264 /* clean up prefsrc entries */
1265 rt6_remove_prefsrc(ifp);
1270 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1271 struct inet6_ifaddr *ift,
1274 struct inet6_dev *idev = ifp->idev;
1275 struct in6_addr addr, *tmpaddr;
1276 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1277 unsigned long regen_advance;
1281 unsigned long now = jiffies;
1282 long max_desync_factor;
1283 s32 cnf_temp_preferred_lft;
1285 write_lock_bh(&idev->lock);
1287 spin_lock_bh(&ift->lock);
1288 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1289 spin_unlock_bh(&ift->lock);
1296 if (idev->cnf.use_tempaddr <= 0) {
1297 write_unlock_bh(&idev->lock);
1298 pr_info("%s: use_tempaddr is disabled\n", __func__);
1303 spin_lock_bh(&ifp->lock);
1304 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1305 idev->cnf.use_tempaddr = -1; /*XXX*/
1306 spin_unlock_bh(&ifp->lock);
1307 write_unlock_bh(&idev->lock);
1308 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1315 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1316 ipv6_try_regen_rndid(idev, tmpaddr);
1317 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1318 age = (now - ifp->tstamp) / HZ;
1320 regen_advance = idev->cnf.regen_max_retry *
1321 idev->cnf.dad_transmits *
1322 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1324 /* recalculate max_desync_factor each time and update
1325 * idev->desync_factor if it's larger
1327 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1328 max_desync_factor = min_t(__u32,
1329 idev->cnf.max_desync_factor,
1330 cnf_temp_preferred_lft - regen_advance);
1332 if (unlikely(idev->desync_factor > max_desync_factor)) {
1333 if (max_desync_factor > 0) {
1334 get_random_bytes(&idev->desync_factor,
1335 sizeof(idev->desync_factor));
1336 idev->desync_factor %= max_desync_factor;
1338 idev->desync_factor = 0;
1342 tmp_valid_lft = min_t(__u32,
1344 idev->cnf.temp_valid_lft + age);
1345 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1346 idev->desync_factor;
1347 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1348 tmp_plen = ifp->prefix_len;
1349 tmp_tstamp = ifp->tstamp;
1350 spin_unlock_bh(&ifp->lock);
1352 write_unlock_bh(&idev->lock);
1354 /* A temporary address is created only if this calculated Preferred
1355 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1356 * an implementation must not create a temporary address with a zero
1357 * Preferred Lifetime.
1358 * Use age calculation as in addrconf_verify to avoid unnecessary
1359 * temporary addresses being generated.
1361 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1362 if (tmp_prefered_lft <= regen_advance + age) {
1369 addr_flags = IFA_F_TEMPORARY;
1370 /* set in addrconf_prefix_rcv() */
1371 if (ifp->flags & IFA_F_OPTIMISTIC)
1372 addr_flags |= IFA_F_OPTIMISTIC;
1374 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1375 ipv6_addr_scope(&addr), addr_flags,
1376 tmp_valid_lft, tmp_prefered_lft, block, NULL);
1380 pr_info("%s: retry temporary address regeneration\n", __func__);
1382 write_lock_bh(&idev->lock);
1386 spin_lock_bh(&ift->lock);
1389 ift->tstamp = tmp_tstamp;
1390 spin_unlock_bh(&ift->lock);
1392 addrconf_dad_start(ift);
1400 * Choose an appropriate source address (RFC3484)
1403 IPV6_SADDR_RULE_INIT = 0,
1404 IPV6_SADDR_RULE_LOCAL,
1405 IPV6_SADDR_RULE_SCOPE,
1406 IPV6_SADDR_RULE_PREFERRED,
1407 #ifdef CONFIG_IPV6_MIP6
1408 IPV6_SADDR_RULE_HOA,
1410 IPV6_SADDR_RULE_OIF,
1411 IPV6_SADDR_RULE_LABEL,
1412 IPV6_SADDR_RULE_PRIVACY,
1413 IPV6_SADDR_RULE_ORCHID,
1414 IPV6_SADDR_RULE_PREFIX,
1415 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1416 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1421 struct ipv6_saddr_score {
1424 struct inet6_ifaddr *ifa;
1425 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1430 struct ipv6_saddr_dst {
1431 const struct in6_addr *addr;
1438 static inline int ipv6_saddr_preferred(int type)
1440 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1445 static bool ipv6_use_optimistic_addr(struct net *net,
1446 struct inet6_dev *idev)
1448 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1451 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1453 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1462 static int ipv6_get_saddr_eval(struct net *net,
1463 struct ipv6_saddr_score *score,
1464 struct ipv6_saddr_dst *dst,
1469 if (i <= score->rule) {
1471 case IPV6_SADDR_RULE_SCOPE:
1472 ret = score->scopedist;
1474 case IPV6_SADDR_RULE_PREFIX:
1475 ret = score->matchlen;
1478 ret = !!test_bit(i, score->scorebits);
1484 case IPV6_SADDR_RULE_INIT:
1485 /* Rule 0: remember if hiscore is not ready yet */
1488 case IPV6_SADDR_RULE_LOCAL:
1489 /* Rule 1: Prefer same address */
1490 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1492 case IPV6_SADDR_RULE_SCOPE:
1493 /* Rule 2: Prefer appropriate scope
1498 * ---+--+-+---> scope
1500 * | d is scope of the destination.
1502 * | \ <- smaller scope is better if
1503 * B-15 | \ if scope is enough for destination.
1504 * | ret = B - scope (-1 <= scope >= d <= 15).
1506 * |/ <- greater is better
1507 * -C / if scope is not enough for destination.
1508 * /| ret = scope - C (-1 <= d < scope <= 15).
1510 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1511 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1512 * Assume B = 0 and we get C > 29.
1514 ret = __ipv6_addr_src_scope(score->addr_type);
1515 if (ret >= dst->scope)
1518 ret -= 128; /* 30 is enough */
1519 score->scopedist = ret;
1521 case IPV6_SADDR_RULE_PREFERRED:
1523 /* Rule 3: Avoid deprecated and optimistic addresses */
1524 u8 avoid = IFA_F_DEPRECATED;
1526 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1527 avoid |= IFA_F_OPTIMISTIC;
1528 ret = ipv6_saddr_preferred(score->addr_type) ||
1529 !(score->ifa->flags & avoid);
1532 #ifdef CONFIG_IPV6_MIP6
1533 case IPV6_SADDR_RULE_HOA:
1535 /* Rule 4: Prefer home address */
1536 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1537 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1541 case IPV6_SADDR_RULE_OIF:
1542 /* Rule 5: Prefer outgoing interface */
1543 ret = (!dst->ifindex ||
1544 dst->ifindex == score->ifa->idev->dev->ifindex);
1546 case IPV6_SADDR_RULE_LABEL:
1547 /* Rule 6: Prefer matching label */
1548 ret = ipv6_addr_label(net,
1549 &score->ifa->addr, score->addr_type,
1550 score->ifa->idev->dev->ifindex) == dst->label;
1552 case IPV6_SADDR_RULE_PRIVACY:
1554 /* Rule 7: Prefer public address
1555 * Note: prefer temporary address if use_tempaddr >= 2
1557 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1558 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1559 score->ifa->idev->cnf.use_tempaddr >= 2;
1560 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1563 case IPV6_SADDR_RULE_ORCHID:
1564 /* Rule 8-: Prefer ORCHID vs ORCHID or
1565 * non-ORCHID vs non-ORCHID
1567 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1568 ipv6_addr_orchid(dst->addr));
1570 case IPV6_SADDR_RULE_PREFIX:
1571 /* Rule 8: Use longest matching prefix */
1572 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1573 if (ret > score->ifa->prefix_len)
1574 ret = score->ifa->prefix_len;
1575 score->matchlen = ret;
1577 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1578 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1579 /* Optimistic addresses still have lower precedence than other
1580 * preferred addresses.
1582 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1590 __set_bit(i, score->scorebits);
1596 static int __ipv6_dev_get_saddr(struct net *net,
1597 struct ipv6_saddr_dst *dst,
1598 struct inet6_dev *idev,
1599 struct ipv6_saddr_score *scores,
1602 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1604 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1608 * - Tentative Address (RFC2462 section 5.4)
1609 * - A tentative address is not considered
1610 * "assigned to an interface" in the traditional
1611 * sense, unless it is also flagged as optimistic.
1612 * - Candidate Source Address (section 4)
1613 * - In any case, anycast addresses, multicast
1614 * addresses, and the unspecified address MUST
1615 * NOT be included in a candidate set.
1617 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1618 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1621 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1623 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1624 score->addr_type & IPV6_ADDR_MULTICAST)) {
1625 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1631 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1633 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1634 int minihiscore, miniscore;
1636 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1637 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1639 if (minihiscore > miniscore) {
1640 if (i == IPV6_SADDR_RULE_SCOPE &&
1641 score->scopedist > 0) {
1644 * each remaining entry
1645 * has too small (not enough)
1646 * scope, because ifa entries
1647 * are sorted by their scope
1653 } else if (minihiscore < miniscore) {
1654 swap(hiscore, score);
1655 hiscore_idx = 1 - hiscore_idx;
1657 /* restore our iterator */
1658 score->ifa = hiscore->ifa;
1668 static int ipv6_get_saddr_master(struct net *net,
1669 const struct net_device *dst_dev,
1670 const struct net_device *master,
1671 struct ipv6_saddr_dst *dst,
1672 struct ipv6_saddr_score *scores,
1675 struct inet6_dev *idev;
1677 idev = __in6_dev_get(dst_dev);
1679 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1680 scores, hiscore_idx);
1682 idev = __in6_dev_get(master);
1684 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1685 scores, hiscore_idx);
1690 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1691 const struct in6_addr *daddr, unsigned int prefs,
1692 struct in6_addr *saddr)
1694 struct ipv6_saddr_score scores[2], *hiscore;
1695 struct ipv6_saddr_dst dst;
1696 struct inet6_dev *idev;
1697 struct net_device *dev;
1699 bool use_oif_addr = false;
1700 int hiscore_idx = 0;
1703 dst_type = __ipv6_addr_type(daddr);
1705 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1706 dst.scope = __ipv6_addr_src_scope(dst_type);
1707 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1710 scores[hiscore_idx].rule = -1;
1711 scores[hiscore_idx].ifa = NULL;
1715 /* Candidate Source Address (section 4)
1716 * - multicast and link-local destination address,
1717 * the set of candidate source address MUST only
1718 * include addresses assigned to interfaces
1719 * belonging to the same link as the outgoing
1721 * (- For site-local destination addresses, the
1722 * set of candidate source addresses MUST only
1723 * include addresses assigned to interfaces
1724 * belonging to the same site as the outgoing
1726 * - "It is RECOMMENDED that the candidate source addresses
1727 * be the set of unicast addresses assigned to the
1728 * interface that will be used to send to the destination
1729 * (the 'outgoing' interface)." (RFC 6724)
1732 idev = __in6_dev_get(dst_dev);
1733 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1734 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1735 (idev && idev->cnf.use_oif_addrs_only)) {
1736 use_oif_addr = true;
1742 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1744 const struct net_device *master;
1747 /* if dst_dev exists and is enslaved to an L3 device, then
1748 * prefer addresses from dst_dev and then the master over
1749 * any other enslaved devices in the L3 domain.
1751 master = l3mdev_master_dev_rcu(dst_dev);
1753 master_idx = master->ifindex;
1755 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1757 scores, hiscore_idx);
1759 if (scores[hiscore_idx].ifa)
1763 for_each_netdev_rcu(net, dev) {
1764 /* only consider addresses on devices in the
1767 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1769 idev = __in6_dev_get(dev);
1772 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1777 hiscore = &scores[hiscore_idx];
1779 ret = -EADDRNOTAVAIL;
1781 *saddr = hiscore->ifa->addr;
1786 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1788 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1791 struct inet6_ifaddr *ifp;
1792 int err = -EADDRNOTAVAIL;
1794 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1795 if (ifp->scope > IFA_LINK)
1797 if (ifp->scope == IFA_LINK &&
1798 !(ifp->flags & banned_flags)) {
1807 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1810 struct inet6_dev *idev;
1811 int err = -EADDRNOTAVAIL;
1814 idev = __in6_dev_get(dev);
1816 read_lock_bh(&idev->lock);
1817 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1818 read_unlock_bh(&idev->lock);
1824 static int ipv6_count_addresses(const struct inet6_dev *idev)
1826 const struct inet6_ifaddr *ifp;
1830 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1836 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1837 const struct net_device *dev, int strict)
1839 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1841 EXPORT_SYMBOL(ipv6_chk_addr);
1843 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1844 const struct net_device *dev, int strict,
1847 unsigned int hash = inet6_addr_hash(net, addr);
1848 struct inet6_ifaddr *ifp;
1852 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1853 if (!net_eq(dev_net(ifp->idev->dev), net))
1855 /* Decouple optimistic from tentative for evaluation here.
1856 * Ban optimistic addresses explicitly, when required.
1858 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1859 ? (ifp->flags&~IFA_F_TENTATIVE)
1861 if (ipv6_addr_equal(&ifp->addr, addr) &&
1862 !(ifp_flags&banned_flags) &&
1863 (!dev || ifp->idev->dev == dev ||
1864 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1873 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1876 /* Compares an address/prefix_len with addresses on device @dev.
1877 * If one is found it returns true.
1879 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1880 const unsigned int prefix_len, struct net_device *dev)
1882 const struct inet6_ifaddr *ifa;
1883 const struct inet6_dev *idev;
1887 idev = __in6_dev_get(dev);
1889 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1890 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1899 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1901 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1903 const struct inet6_ifaddr *ifa;
1904 const struct inet6_dev *idev;
1909 idev = __in6_dev_get(dev);
1911 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1912 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1921 EXPORT_SYMBOL(ipv6_chk_prefix);
1923 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1924 struct net_device *dev, int strict)
1926 unsigned int hash = inet6_addr_hash(net, addr);
1927 struct inet6_ifaddr *ifp, *result = NULL;
1930 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1931 if (!net_eq(dev_net(ifp->idev->dev), net))
1933 if (ipv6_addr_equal(&ifp->addr, addr)) {
1934 if (!dev || ifp->idev->dev == dev ||
1935 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1947 /* Gets referenced address, destroys ifaddr */
1949 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1952 ifp->flags |= IFA_F_DADFAILED;
1954 if (ifp->flags&IFA_F_TEMPORARY) {
1955 struct inet6_ifaddr *ifpub;
1956 spin_lock_bh(&ifp->lock);
1959 in6_ifa_hold(ifpub);
1960 spin_unlock_bh(&ifp->lock);
1961 ipv6_create_tempaddr(ifpub, ifp, true);
1964 spin_unlock_bh(&ifp->lock);
1967 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1968 spin_lock_bh(&ifp->lock);
1969 addrconf_del_dad_work(ifp);
1970 ifp->flags |= IFA_F_TENTATIVE;
1971 spin_unlock_bh(&ifp->lock);
1973 ipv6_ifa_notify(0, ifp);
1980 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1984 spin_lock_bh(&ifp->lock);
1985 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1986 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1989 spin_unlock_bh(&ifp->lock);
1994 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
1996 struct inet6_dev *idev = ifp->idev;
1997 struct net *net = dev_net(ifp->idev->dev);
1999 if (addrconf_dad_end(ifp)) {
2004 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2005 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2007 spin_lock_bh(&ifp->lock);
2009 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2010 int scope = ifp->scope;
2011 u32 flags = ifp->flags;
2012 struct in6_addr new_addr;
2013 struct inet6_ifaddr *ifp2;
2014 u32 valid_lft, preferred_lft;
2015 int pfxlen = ifp->prefix_len;
2016 int retries = ifp->stable_privacy_retry + 1;
2018 if (retries > net->ipv6.sysctl.idgen_retries) {
2019 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2020 ifp->idev->dev->name);
2024 new_addr = ifp->addr;
2025 if (ipv6_generate_stable_address(&new_addr, retries,
2029 valid_lft = ifp->valid_lft;
2030 preferred_lft = ifp->prefered_lft;
2032 spin_unlock_bh(&ifp->lock);
2034 if (idev->cnf.max_addresses &&
2035 ipv6_count_addresses(idev) >=
2036 idev->cnf.max_addresses)
2039 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2040 ifp->idev->dev->name);
2042 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2043 scope, flags, valid_lft,
2044 preferred_lft, false, NULL);
2048 spin_lock_bh(&ifp2->lock);
2049 ifp2->stable_privacy_retry = retries;
2050 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2051 spin_unlock_bh(&ifp2->lock);
2053 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2056 spin_lock_bh(&ifp->lock);
2060 /* transition from _POSTDAD to _ERRDAD */
2061 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2062 spin_unlock_bh(&ifp->lock);
2064 addrconf_mod_dad_work(ifp, 0);
2068 /* Join to solicited addr multicast group.
2069 * caller must hold RTNL */
2070 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2072 struct in6_addr maddr;
2074 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2077 addrconf_addr_solict_mult(addr, &maddr);
2078 ipv6_dev_mc_inc(dev, &maddr);
2081 /* caller must hold RTNL */
2082 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2084 struct in6_addr maddr;
2086 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2089 addrconf_addr_solict_mult(addr, &maddr);
2090 __ipv6_dev_mc_dec(idev, &maddr);
2093 /* caller must hold RTNL */
2094 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2096 struct in6_addr addr;
2098 if (ifp->prefix_len >= 127) /* RFC 6164 */
2100 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2101 if (ipv6_addr_any(&addr))
2103 __ipv6_dev_ac_inc(ifp->idev, &addr);
2106 /* caller must hold RTNL */
2107 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2109 struct in6_addr addr;
2111 if (ifp->prefix_len >= 127) /* RFC 6164 */
2113 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2114 if (ipv6_addr_any(&addr))
2116 __ipv6_dev_ac_dec(ifp->idev, &addr);
2119 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2121 switch (dev->addr_len) {
2123 memcpy(eui, dev->dev_addr, 3);
2126 memcpy(eui + 5, dev->dev_addr + 3, 3);
2128 case EUI64_ADDR_LEN:
2129 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2139 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2141 union fwnet_hwaddr *ha;
2143 if (dev->addr_len != FWNET_ALEN)
2146 ha = (union fwnet_hwaddr *)dev->dev_addr;
2148 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2153 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2155 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2156 if (dev->addr_len != ARCNET_ALEN)
2159 eui[7] = *(u8 *)dev->dev_addr;
2163 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2165 if (dev->addr_len != INFINIBAND_ALEN)
2167 memcpy(eui, dev->dev_addr + 12, 8);
2172 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2176 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2177 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2178 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2179 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2180 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2181 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2185 memcpy(eui + 4, &addr, 4);
2189 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2191 if (dev->priv_flags & IFF_ISATAP)
2192 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2196 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2198 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2201 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2203 memcpy(eui, dev->perm_addr, 3);
2204 memcpy(eui + 5, dev->perm_addr + 3, 3);
2211 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2213 switch (dev->type) {
2216 return addrconf_ifid_eui48(eui, dev);
2218 return addrconf_ifid_arcnet(eui, dev);
2219 case ARPHRD_INFINIBAND:
2220 return addrconf_ifid_infiniband(eui, dev);
2222 return addrconf_ifid_sit(eui, dev);
2225 return addrconf_ifid_gre(eui, dev);
2226 case ARPHRD_6LOWPAN:
2227 return addrconf_ifid_6lowpan(eui, dev);
2228 case ARPHRD_IEEE1394:
2229 return addrconf_ifid_ieee1394(eui, dev);
2230 case ARPHRD_TUNNEL6:
2232 return addrconf_ifid_ip6tnl(eui, dev);
2237 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2240 struct inet6_ifaddr *ifp;
2242 read_lock_bh(&idev->lock);
2243 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2244 if (ifp->scope > IFA_LINK)
2246 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2247 memcpy(eui, ifp->addr.s6_addr+8, 8);
2252 read_unlock_bh(&idev->lock);
2256 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2257 static void ipv6_regen_rndid(struct inet6_dev *idev)
2260 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2261 idev->rndid[0] &= ~0x02;
2264 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2265 * check if generated address is not inappropriate
2267 * - Reserved subnet anycast (RFC 2526)
2268 * 11111101 11....11 1xxxxxxx
2269 * - ISATAP (RFC4214) 6.1
2270 * 00-00-5E-FE-xx-xx-xx-xx
2272 * - XXX: already assigned to an address on the device
2274 if (idev->rndid[0] == 0xfd &&
2275 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2276 (idev->rndid[7]&0x80))
2278 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2279 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2281 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2286 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2288 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2289 ipv6_regen_rndid(idev);
2297 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2298 unsigned long expires, u32 flags)
2300 struct fib6_config cfg = {
2301 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2302 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2303 .fc_ifindex = dev->ifindex,
2304 .fc_expires = expires,
2306 .fc_flags = RTF_UP | flags,
2307 .fc_nlinfo.nl_net = dev_net(dev),
2308 .fc_protocol = RTPROT_KERNEL,
2313 /* Prevent useless cloning on PtP SIT.
2314 This thing is done here expecting that the whole
2315 class of non-broadcast devices need not cloning.
2317 #if IS_ENABLED(CONFIG_IPV6_SIT)
2318 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2319 cfg.fc_flags |= RTF_NONEXTHOP;
2322 ip6_route_add(&cfg, NULL);
2326 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2328 const struct net_device *dev,
2329 u32 flags, u32 noflags)
2331 struct fib6_node *fn;
2332 struct rt6_info *rt = NULL;
2333 struct fib6_table *table;
2334 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2336 table = fib6_get_table(dev_net(dev), tb_id);
2341 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2345 for_each_fib6_node_rt_rcu(fn) {
2346 if (rt->dst.dev->ifindex != dev->ifindex)
2348 if ((rt->rt6i_flags & flags) != flags)
2350 if ((rt->rt6i_flags & noflags) != 0)
2352 if (!dst_hold_safe(&rt->dst))
2362 /* Create "default" multicast route to the interface */
2364 static void addrconf_add_mroute(struct net_device *dev)
2366 struct fib6_config cfg = {
2367 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2368 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2369 .fc_ifindex = dev->ifindex,
2372 .fc_nlinfo.nl_net = dev_net(dev),
2375 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2377 ip6_route_add(&cfg, NULL);
2380 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2382 struct inet6_dev *idev;
2386 idev = ipv6_find_idev(dev);
2388 return ERR_PTR(-ENOBUFS);
2390 if (idev->cnf.disable_ipv6)
2391 return ERR_PTR(-EACCES);
2393 /* Add default multicast route */
2394 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2395 addrconf_add_mroute(dev);
2400 static void manage_tempaddrs(struct inet6_dev *idev,
2401 struct inet6_ifaddr *ifp,
2402 __u32 valid_lft, __u32 prefered_lft,
2403 bool create, unsigned long now)
2406 struct inet6_ifaddr *ift;
2408 read_lock_bh(&idev->lock);
2409 /* update all temporary addresses in the list */
2410 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2411 int age, max_valid, max_prefered;
2413 if (ifp != ift->ifpub)
2416 /* RFC 4941 section 3.3:
2417 * If a received option will extend the lifetime of a public
2418 * address, the lifetimes of temporary addresses should
2419 * be extended, subject to the overall constraint that no
2420 * temporary addresses should ever remain "valid" or "preferred"
2421 * for a time longer than (TEMP_VALID_LIFETIME) or
2422 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2424 age = (now - ift->cstamp) / HZ;
2425 max_valid = idev->cnf.temp_valid_lft - age;
2429 max_prefered = idev->cnf.temp_prefered_lft -
2430 idev->desync_factor - age;
2431 if (max_prefered < 0)
2434 if (valid_lft > max_valid)
2435 valid_lft = max_valid;
2437 if (prefered_lft > max_prefered)
2438 prefered_lft = max_prefered;
2440 spin_lock(&ift->lock);
2442 ift->valid_lft = valid_lft;
2443 ift->prefered_lft = prefered_lft;
2445 if (prefered_lft > 0)
2446 ift->flags &= ~IFA_F_DEPRECATED;
2448 spin_unlock(&ift->lock);
2449 if (!(flags&IFA_F_TENTATIVE))
2450 ipv6_ifa_notify(0, ift);
2453 if ((create || list_empty(&idev->tempaddr_list)) &&
2454 idev->cnf.use_tempaddr > 0) {
2455 /* When a new public address is created as described
2456 * in [ADDRCONF], also create a new temporary address.
2457 * Also create a temporary address if it's enabled but
2458 * no temporary address currently exists.
2460 read_unlock_bh(&idev->lock);
2461 ipv6_create_tempaddr(ifp, NULL, false);
2463 read_unlock_bh(&idev->lock);
2467 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2469 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2470 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2473 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2474 const struct prefix_info *pinfo,
2475 struct inet6_dev *in6_dev,
2476 const struct in6_addr *addr, int addr_type,
2477 u32 addr_flags, bool sllao, bool tokenized,
2478 __u32 valid_lft, u32 prefered_lft)
2480 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2481 int create = 0, update_lft = 0;
2483 if (!ifp && valid_lft) {
2484 int max_addresses = in6_dev->cnf.max_addresses;
2486 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2487 if ((net->ipv6.devconf_all->optimistic_dad ||
2488 in6_dev->cnf.optimistic_dad) &&
2489 !net->ipv6.devconf_all->forwarding && sllao)
2490 addr_flags |= IFA_F_OPTIMISTIC;
2493 /* Do not allow to create too much of autoconfigured
2494 * addresses; this would be too easy way to crash kernel.
2496 if (!max_addresses ||
2497 ipv6_count_addresses(in6_dev) < max_addresses)
2498 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2500 addr_type&IPV6_ADDR_SCOPE_MASK,
2501 addr_flags, valid_lft,
2502 prefered_lft, false, NULL);
2504 if (IS_ERR_OR_NULL(ifp))
2509 spin_lock_bh(&ifp->lock);
2510 ifp->flags |= IFA_F_MANAGETEMPADDR;
2511 ifp->cstamp = jiffies;
2512 ifp->tokenized = tokenized;
2513 spin_unlock_bh(&ifp->lock);
2514 addrconf_dad_start(ifp);
2522 /* update lifetime (RFC2462 5.5.3 e) */
2523 spin_lock_bh(&ifp->lock);
2525 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2526 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2529 if (!update_lft && !create && stored_lft) {
2530 const u32 minimum_lft = min_t(u32,
2531 stored_lft, MIN_VALID_LIFETIME);
2532 valid_lft = max(valid_lft, minimum_lft);
2534 /* RFC4862 Section 5.5.3e:
2535 * "Note that the preferred lifetime of the
2536 * corresponding address is always reset to
2537 * the Preferred Lifetime in the received
2538 * Prefix Information option, regardless of
2539 * whether the valid lifetime is also reset or
2542 * So we should always update prefered_lft here.
2548 ifp->valid_lft = valid_lft;
2549 ifp->prefered_lft = prefered_lft;
2552 ifp->flags &= ~IFA_F_DEPRECATED;
2553 spin_unlock_bh(&ifp->lock);
2555 if (!(flags&IFA_F_TENTATIVE))
2556 ipv6_ifa_notify(0, ifp);
2558 spin_unlock_bh(&ifp->lock);
2560 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2569 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2571 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2573 struct prefix_info *pinfo;
2578 struct inet6_dev *in6_dev;
2579 struct net *net = dev_net(dev);
2581 pinfo = (struct prefix_info *) opt;
2583 if (len < sizeof(struct prefix_info)) {
2584 ADBG("addrconf: prefix option too short\n");
2589 * Validation checks ([ADDRCONF], page 19)
2592 addr_type = ipv6_addr_type(&pinfo->prefix);
2594 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2597 valid_lft = ntohl(pinfo->valid);
2598 prefered_lft = ntohl(pinfo->prefered);
2600 if (prefered_lft > valid_lft) {
2601 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2605 in6_dev = in6_dev_get(dev);
2608 net_dbg_ratelimited("addrconf: device %s not configured\n",
2614 * Two things going on here:
2615 * 1) Add routes for on-link prefixes
2616 * 2) Configure prefixes with the auto flag set
2619 if (pinfo->onlink) {
2620 struct rt6_info *rt;
2621 unsigned long rt_expires;
2623 /* Avoid arithmetic overflow. Really, we could
2624 * save rt_expires in seconds, likely valid_lft,
2625 * but it would require division in fib gc, that it
2629 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2631 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2633 if (addrconf_finite_timeout(rt_expires))
2636 rt = addrconf_get_prefix_route(&pinfo->prefix,
2639 RTF_ADDRCONF | RTF_PREFIX_RT,
2640 RTF_GATEWAY | RTF_DEFAULT);
2643 /* Autoconf prefix route */
2644 if (valid_lft == 0) {
2647 } else if (addrconf_finite_timeout(rt_expires)) {
2649 rt6_set_expires(rt, jiffies + rt_expires);
2651 rt6_clean_expires(rt);
2653 } else if (valid_lft) {
2654 clock_t expires = 0;
2655 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2656 if (addrconf_finite_timeout(rt_expires)) {
2658 flags |= RTF_EXPIRES;
2659 expires = jiffies_to_clock_t(rt_expires);
2661 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2662 dev, expires, flags);
2667 /* Try to figure out our local address for this prefix */
2669 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2670 struct in6_addr addr;
2671 bool tokenized = false, dev_addr_generated = false;
2673 if (pinfo->prefix_len == 64) {
2674 memcpy(&addr, &pinfo->prefix, 8);
2676 if (!ipv6_addr_any(&in6_dev->token)) {
2677 read_lock_bh(&in6_dev->lock);
2678 memcpy(addr.s6_addr + 8,
2679 in6_dev->token.s6_addr + 8, 8);
2680 read_unlock_bh(&in6_dev->lock);
2682 } else if (is_addr_mode_generate_stable(in6_dev) &&
2683 !ipv6_generate_stable_address(&addr, 0,
2685 addr_flags |= IFA_F_STABLE_PRIVACY;
2687 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2688 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2691 dev_addr_generated = true;
2695 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2700 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2703 tokenized, valid_lft,
2708 /* Ignore error case here because previous prefix add addr was
2709 * successful which will be notified.
2711 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2712 addr_type, addr_flags, sllao,
2713 tokenized, valid_lft,
2715 dev_addr_generated);
2717 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2719 in6_dev_put(in6_dev);
2723 * Set destination address.
2724 * Special case for SIT interfaces where we create a new "virtual"
2727 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2729 struct in6_ifreq ireq;
2730 struct net_device *dev;
2736 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2739 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2745 #if IS_ENABLED(CONFIG_IPV6_SIT)
2746 if (dev->type == ARPHRD_SIT) {
2747 const struct net_device_ops *ops = dev->netdev_ops;
2749 struct ip_tunnel_parm p;
2751 err = -EADDRNOTAVAIL;
2752 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2755 memset(&p, 0, sizeof(p));
2756 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2760 p.iph.protocol = IPPROTO_IPV6;
2762 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2764 if (ops->ndo_do_ioctl) {
2765 mm_segment_t oldfs = get_fs();
2768 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2775 dev = __dev_get_by_name(net, p.name);
2778 err = dev_open(dev);
2788 static int ipv6_mc_config(struct sock *sk, bool join,
2789 const struct in6_addr *addr, int ifindex)
2797 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2799 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2806 * Manual configuration of address on an interface
2808 static int inet6_addr_add(struct net *net, int ifindex,
2809 const struct in6_addr *pfx,
2810 const struct in6_addr *peer_pfx,
2811 unsigned int plen, __u32 ifa_flags,
2812 __u32 prefered_lft, __u32 valid_lft,
2813 struct netlink_ext_ack *extack)
2815 struct inet6_ifaddr *ifp;
2816 struct inet6_dev *idev;
2817 struct net_device *dev;
2818 unsigned long timeout;
2828 /* check the lifetime */
2829 if (!valid_lft || prefered_lft > valid_lft)
2832 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2835 dev = __dev_get_by_index(net, ifindex);
2839 idev = addrconf_add_dev(dev);
2841 return PTR_ERR(idev);
2843 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2844 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2845 true, pfx, ifindex);
2851 scope = ipv6_addr_scope(pfx);
2853 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2854 if (addrconf_finite_timeout(timeout)) {
2855 expires = jiffies_to_clock_t(timeout * HZ);
2856 valid_lft = timeout;
2857 flags = RTF_EXPIRES;
2861 ifa_flags |= IFA_F_PERMANENT;
2864 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2865 if (addrconf_finite_timeout(timeout)) {
2867 ifa_flags |= IFA_F_DEPRECATED;
2868 prefered_lft = timeout;
2871 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2872 valid_lft, prefered_lft, true, extack);
2875 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2876 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2881 * Note that section 3.1 of RFC 4429 indicates
2882 * that the Optimistic flag should not be set for
2883 * manually configured addresses
2885 addrconf_dad_start(ifp);
2886 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2887 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2890 addrconf_verify_rtnl();
2892 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2893 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2894 false, pfx, ifindex);
2897 return PTR_ERR(ifp);
2900 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2901 const struct in6_addr *pfx, unsigned int plen)
2903 struct inet6_ifaddr *ifp;
2904 struct inet6_dev *idev;
2905 struct net_device *dev;
2910 dev = __dev_get_by_index(net, ifindex);
2914 idev = __in6_dev_get(dev);
2918 read_lock_bh(&idev->lock);
2919 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2920 if (ifp->prefix_len == plen &&
2921 ipv6_addr_equal(pfx, &ifp->addr)) {
2923 read_unlock_bh(&idev->lock);
2925 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2926 (ifa_flags & IFA_F_MANAGETEMPADDR))
2927 manage_tempaddrs(idev, ifp, 0, 0, false,
2930 addrconf_verify_rtnl();
2931 if (ipv6_addr_is_multicast(pfx)) {
2932 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2933 false, pfx, dev->ifindex);
2938 read_unlock_bh(&idev->lock);
2939 return -EADDRNOTAVAIL;
2943 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2945 struct in6_ifreq ireq;
2948 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2951 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2955 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2956 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2957 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2962 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2964 struct in6_ifreq ireq;
2967 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2970 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2974 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2975 ireq.ifr6_prefixlen);
2980 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2981 int plen, int scope)
2983 struct inet6_ifaddr *ifp;
2985 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2986 scope, IFA_F_PERMANENT,
2987 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
2990 spin_lock_bh(&ifp->lock);
2991 ifp->flags &= ~IFA_F_TENTATIVE;
2992 spin_unlock_bh(&ifp->lock);
2993 rt_genid_bump_ipv6(dev_net(idev->dev));
2994 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2999 #if IS_ENABLED(CONFIG_IPV6_SIT)
3000 static void sit_add_v4_addrs(struct inet6_dev *idev)
3002 struct in6_addr addr;
3003 struct net_device *dev;
3004 struct net *net = dev_net(idev->dev);
3010 memset(&addr, 0, sizeof(struct in6_addr));
3011 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3013 if (idev->dev->flags&IFF_POINTOPOINT) {
3014 addr.s6_addr32[0] = htonl(0xfe800000);
3018 scope = IPV6_ADDR_COMPATv4;
3020 pflags |= RTF_NONEXTHOP;
3023 if (addr.s6_addr32[3]) {
3024 add_addr(idev, &addr, plen, scope);
3025 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3029 for_each_netdev(net, dev) {
3030 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3031 if (in_dev && (dev->flags & IFF_UP)) {
3032 struct in_ifaddr *ifa;
3036 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3038 addr.s6_addr32[3] = ifa->ifa_local;
3040 if (ifa->ifa_scope == RT_SCOPE_LINK)
3042 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3043 if (idev->dev->flags&IFF_POINTOPOINT)
3048 add_addr(idev, &addr, plen, flag);
3049 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3057 static void init_loopback(struct net_device *dev)
3059 struct inet6_dev *idev;
3065 idev = ipv6_find_idev(dev);
3067 pr_debug("%s: add_dev failed\n", __func__);
3071 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3074 void addrconf_add_linklocal(struct inet6_dev *idev,
3075 const struct in6_addr *addr, u32 flags)
3077 struct inet6_ifaddr *ifp;
3078 u32 addr_flags = flags | IFA_F_PERMANENT;
3080 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3081 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3082 idev->cnf.optimistic_dad) &&
3083 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3084 addr_flags |= IFA_F_OPTIMISTIC;
3087 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3088 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3090 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3091 addrconf_dad_start(ifp);
3095 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3097 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3099 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3102 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3103 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3106 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3107 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3113 static int ipv6_generate_stable_address(struct in6_addr *address,
3115 const struct inet6_dev *idev)
3117 static DEFINE_SPINLOCK(lock);
3118 static __u32 digest[SHA_DIGEST_WORDS];
3119 static __u32 workspace[SHA_WORKSPACE_WORDS];
3122 char __data[SHA_MESSAGE_BYTES];
3124 struct in6_addr secret;
3126 unsigned char hwaddr[MAX_ADDR_LEN];
3131 struct in6_addr secret;
3132 struct in6_addr temp;
3133 struct net *net = dev_net(idev->dev);
3135 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3137 if (idev->cnf.stable_secret.initialized)
3138 secret = idev->cnf.stable_secret.secret;
3139 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3140 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3145 spin_lock_bh(&lock);
3148 memset(&data, 0, sizeof(data));
3149 memset(workspace, 0, sizeof(workspace));
3150 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3151 data.prefix[0] = address->s6_addr32[0];
3152 data.prefix[1] = address->s6_addr32[1];
3153 data.secret = secret;
3154 data.dad_count = dad_count;
3156 sha_transform(digest, data.__data, workspace);
3159 temp.s6_addr32[2] = (__force __be32)digest[0];
3160 temp.s6_addr32[3] = (__force __be32)digest[1];
3162 spin_unlock_bh(&lock);
3164 if (ipv6_reserved_interfaceid(temp)) {
3166 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3175 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3177 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3181 s = &idev->cnf.stable_secret;
3182 get_random_bytes(&s->secret, sizeof(s->secret));
3183 s->initialized = true;
3186 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3188 struct in6_addr addr;
3190 /* no link local addresses on L3 master devices */
3191 if (netif_is_l3_master(idev->dev))
3194 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3196 switch (idev->cnf.addr_gen_mode) {
3197 case IN6_ADDR_GEN_MODE_RANDOM:
3198 ipv6_gen_mode_random_init(idev);
3200 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3201 if (!ipv6_generate_stable_address(&addr, 0, idev))
3202 addrconf_add_linklocal(idev, &addr,
3203 IFA_F_STABLE_PRIVACY);
3204 else if (prefix_route)
3205 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3207 case IN6_ADDR_GEN_MODE_EUI64:
3208 /* addrconf_add_linklocal also adds a prefix_route and we
3209 * only need to care about prefix routes if ipv6_generate_eui64
3210 * couldn't generate one.
3212 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3213 addrconf_add_linklocal(idev, &addr, 0);
3214 else if (prefix_route)
3215 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3217 case IN6_ADDR_GEN_MODE_NONE:
3219 /* will not add any link local address */
3224 static void addrconf_dev_config(struct net_device *dev)
3226 struct inet6_dev *idev;
3230 if ((dev->type != ARPHRD_ETHER) &&
3231 (dev->type != ARPHRD_FDDI) &&
3232 (dev->type != ARPHRD_ARCNET) &&
3233 (dev->type != ARPHRD_INFINIBAND) &&
3234 (dev->type != ARPHRD_IEEE1394) &&
3235 (dev->type != ARPHRD_TUNNEL6) &&
3236 (dev->type != ARPHRD_6LOWPAN) &&
3237 (dev->type != ARPHRD_IP6GRE) &&
3238 (dev->type != ARPHRD_IPGRE) &&
3239 (dev->type != ARPHRD_TUNNEL) &&
3240 (dev->type != ARPHRD_NONE)) {
3241 /* Alas, we support only Ethernet autoconfiguration. */
3245 idev = addrconf_add_dev(dev);
3249 /* this device type has no EUI support */
3250 if (dev->type == ARPHRD_NONE &&
3251 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3252 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3254 addrconf_addr_gen(idev, false);
3257 #if IS_ENABLED(CONFIG_IPV6_SIT)
3258 static void addrconf_sit_config(struct net_device *dev)
3260 struct inet6_dev *idev;
3265 * Configure the tunnel with one of our IPv4
3266 * addresses... we should configure all of
3267 * our v4 addrs in the tunnel
3270 idev = ipv6_find_idev(dev);
3272 pr_debug("%s: add_dev failed\n", __func__);
3276 if (dev->priv_flags & IFF_ISATAP) {
3277 addrconf_addr_gen(idev, false);
3281 sit_add_v4_addrs(idev);
3283 if (dev->flags&IFF_POINTOPOINT)
3284 addrconf_add_mroute(dev);
3288 #if IS_ENABLED(CONFIG_NET_IPGRE)
3289 static void addrconf_gre_config(struct net_device *dev)
3291 struct inet6_dev *idev;
3295 idev = ipv6_find_idev(dev);
3297 pr_debug("%s: add_dev failed\n", __func__);
3301 addrconf_addr_gen(idev, true);
3302 if (dev->flags & IFF_POINTOPOINT)
3303 addrconf_add_mroute(dev);
3307 static int fixup_permanent_addr(struct inet6_dev *idev,
3308 struct inet6_ifaddr *ifp)
3310 /* !rt6i_node means the host route was removed from the
3311 * FIB, for example, if 'lo' device is taken down. In that
3312 * case regenerate the host route.
3314 if (!ifp->rt || !ifp->rt->rt6i_node) {
3315 struct rt6_info *rt, *prev;
3317 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3321 /* ifp->rt can be accessed outside of rtnl */
3322 spin_lock(&ifp->lock);
3325 spin_unlock(&ifp->lock);
3330 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3331 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3335 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3336 addrconf_dad_start(ifp);
3341 static void addrconf_permanent_addr(struct net_device *dev)
3343 struct inet6_ifaddr *ifp, *tmp;
3344 struct inet6_dev *idev;
3346 idev = __in6_dev_get(dev);
3350 write_lock_bh(&idev->lock);
3352 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3353 if ((ifp->flags & IFA_F_PERMANENT) &&
3354 fixup_permanent_addr(idev, ifp) < 0) {
3355 write_unlock_bh(&idev->lock);
3358 write_lock_bh(&idev->lock);
3360 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3361 idev->dev->name, &ifp->addr);
3365 write_unlock_bh(&idev->lock);
3368 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3371 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3372 struct netdev_notifier_changeupper_info *info;
3373 struct inet6_dev *idev = __in6_dev_get(dev);
3374 struct net *net = dev_net(dev);
3375 int run_pending = 0;
3379 case NETDEV_REGISTER:
3380 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3381 idev = ipv6_add_dev(dev);
3383 return notifier_from_errno(PTR_ERR(idev));
3387 case NETDEV_CHANGEMTU:
3388 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3389 if (dev->mtu < IPV6_MIN_MTU) {
3390 addrconf_ifdown(dev, dev != net->loopback_dev);
3395 rt6_mtu_change(dev, dev->mtu);
3396 idev->cnf.mtu6 = dev->mtu;
3400 /* allocate new idev */
3401 idev = ipv6_add_dev(dev);
3405 /* device is still not ready */
3406 if (!(idev->if_flags & IF_READY))
3415 if (dev->flags & IFF_SLAVE)
3418 if (idev && idev->cnf.disable_ipv6)
3421 if (event == NETDEV_UP) {
3422 /* restore routes for permanent addresses */
3423 addrconf_permanent_addr(dev);
3425 if (!addrconf_link_ready(dev)) {
3426 /* device is not ready yet. */
3427 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3432 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3433 idev = ipv6_add_dev(dev);
3435 if (!IS_ERR_OR_NULL(idev)) {
3436 idev->if_flags |= IF_READY;
3439 } else if (event == NETDEV_CHANGE) {
3440 if (!addrconf_link_ready(dev)) {
3441 /* device is still not ready. */
3446 if (idev->if_flags & IF_READY) {
3447 /* device is already configured -
3448 * but resend MLD reports, we might
3449 * have roamed and need to update
3450 * multicast snooping switches
3455 idev->if_flags |= IF_READY;
3458 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3464 switch (dev->type) {
3465 #if IS_ENABLED(CONFIG_IPV6_SIT)
3467 addrconf_sit_config(dev);
3470 #if IS_ENABLED(CONFIG_NET_IPGRE)
3472 addrconf_gre_config(dev);
3475 case ARPHRD_LOOPBACK:
3480 addrconf_dev_config(dev);
3484 if (!IS_ERR_OR_NULL(idev)) {
3486 addrconf_dad_run(idev);
3489 * If the MTU changed during the interface down,
3490 * when the interface up, the changed MTU must be
3491 * reflected in the idev as well as routers.
3493 if (idev->cnf.mtu6 != dev->mtu &&
3494 dev->mtu >= IPV6_MIN_MTU) {
3495 rt6_mtu_change(dev, dev->mtu);
3496 idev->cnf.mtu6 = dev->mtu;
3498 idev->tstamp = jiffies;
3499 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3502 * If the changed mtu during down is lower than
3503 * IPV6_MIN_MTU stop IPv6 on this interface.
3505 if (dev->mtu < IPV6_MIN_MTU)
3506 addrconf_ifdown(dev, dev != net->loopback_dev);
3511 case NETDEV_UNREGISTER:
3513 * Remove all addresses from this interface.
3515 addrconf_ifdown(dev, event != NETDEV_DOWN);
3518 case NETDEV_CHANGENAME:
3520 snmp6_unregister_dev(idev);
3521 addrconf_sysctl_unregister(idev);
3522 err = addrconf_sysctl_register(idev);
3524 return notifier_from_errno(err);
3525 err = snmp6_register_dev(idev);
3527 addrconf_sysctl_unregister(idev);
3528 return notifier_from_errno(err);
3533 case NETDEV_PRE_TYPE_CHANGE:
3534 case NETDEV_POST_TYPE_CHANGE:
3536 addrconf_type_change(dev, event);
3539 case NETDEV_CHANGEUPPER:
3542 /* flush all routes if dev is linked to or unlinked from
3543 * an L3 master device (e.g., VRF)
3545 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3546 addrconf_ifdown(dev, 0);
3553 * addrconf module should be notified of a device going up
3555 static struct notifier_block ipv6_dev_notf = {
3556 .notifier_call = addrconf_notify,
3557 .priority = ADDRCONF_NOTIFY_PRIORITY,
3560 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3562 struct inet6_dev *idev;
3565 idev = __in6_dev_get(dev);
3567 if (event == NETDEV_POST_TYPE_CHANGE)
3568 ipv6_mc_remap(idev);
3569 else if (event == NETDEV_PRE_TYPE_CHANGE)
3570 ipv6_mc_unmap(idev);
3573 static bool addr_is_local(const struct in6_addr *addr)
3575 return ipv6_addr_type(addr) &
3576 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3579 static int addrconf_ifdown(struct net_device *dev, int how)
3581 struct net *net = dev_net(dev);
3582 struct inet6_dev *idev;
3583 struct inet6_ifaddr *ifa, *tmp;
3590 rt6_ifdown(net, dev);
3591 neigh_ifdown(&nd_tbl, dev);
3593 idev = __in6_dev_get(dev);
3598 * Step 1: remove reference to ipv6 device from parent device.
3604 /* protected by rtnl_lock */
3605 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3607 /* Step 1.5: remove snmp6 entry */
3608 snmp6_unregister_dev(idev);
3612 /* aggregate the system setting and interface setting */
3613 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3615 _keep_addr = idev->cnf.keep_addr_on_down;
3617 /* combine the user config with event to determine if permanent
3618 * addresses are to be removed from address hash table
3620 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3622 /* Step 2: clear hash table */
3623 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3624 struct hlist_head *h = &inet6_addr_lst[i];
3626 spin_lock_bh(&addrconf_hash_lock);
3628 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3629 if (ifa->idev == idev) {
3630 addrconf_del_dad_work(ifa);
3631 /* combined flag + permanent flag decide if
3632 * address is retained on a down event
3635 !(ifa->flags & IFA_F_PERMANENT) ||
3636 addr_is_local(&ifa->addr)) {
3637 hlist_del_init_rcu(&ifa->addr_lst);
3642 spin_unlock_bh(&addrconf_hash_lock);
3645 write_lock_bh(&idev->lock);
3647 addrconf_del_rs_timer(idev);
3649 /* Step 2: clear flags for stateless addrconf */
3651 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3653 /* Step 3: clear tempaddr list */
3654 while (!list_empty(&idev->tempaddr_list)) {
3655 ifa = list_first_entry(&idev->tempaddr_list,
3656 struct inet6_ifaddr, tmp_list);
3657 list_del(&ifa->tmp_list);
3658 write_unlock_bh(&idev->lock);
3659 spin_lock_bh(&ifa->lock);
3662 in6_ifa_put(ifa->ifpub);
3665 spin_unlock_bh(&ifa->lock);
3667 write_lock_bh(&idev->lock);
3670 /* re-combine the user config with event to determine if permanent
3671 * addresses are to be removed from the interface list
3673 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3675 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3676 struct rt6_info *rt = NULL;
3679 addrconf_del_dad_work(ifa);
3681 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3682 !addr_is_local(&ifa->addr);
3684 write_unlock_bh(&idev->lock);
3685 spin_lock_bh(&ifa->lock);
3688 /* set state to skip the notifier below */
3689 state = INET6_IFADDR_STATE_DEAD;
3690 ifa->state = INET6_IFADDR_STATE_PREDAD;
3691 if (!(ifa->flags & IFA_F_NODAD))
3692 ifa->flags |= IFA_F_TENTATIVE;
3698 ifa->state = INET6_IFADDR_STATE_DEAD;
3701 spin_unlock_bh(&ifa->lock);
3706 if (state != INET6_IFADDR_STATE_DEAD) {
3707 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3708 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3710 if (idev->cnf.forwarding)
3711 addrconf_leave_anycast(ifa);
3712 addrconf_leave_solict(ifa->idev, &ifa->addr);
3715 write_lock_bh(&idev->lock);
3717 list_del_rcu(&ifa->if_list);
3722 write_unlock_bh(&idev->lock);
3724 /* Step 5: Discard anycast and multicast list */
3726 ipv6_ac_destroy_dev(idev);
3727 ipv6_mc_destroy_dev(idev);
3732 idev->tstamp = jiffies;
3734 /* Last: Shot the device (if unregistered) */
3736 addrconf_sysctl_unregister(idev);
3737 neigh_parms_release(&nd_tbl, idev->nd_parms);
3738 neigh_ifdown(&nd_tbl, dev);
3744 static void addrconf_rs_timer(unsigned long data)
3746 struct inet6_dev *idev = (struct inet6_dev *)data;
3747 struct net_device *dev = idev->dev;
3748 struct in6_addr lladdr;
3750 write_lock(&idev->lock);
3751 if (idev->dead || !(idev->if_flags & IF_READY))
3754 if (!ipv6_accept_ra(idev))
3757 /* Announcement received after solicitation was sent */
3758 if (idev->if_flags & IF_RA_RCVD)
3761 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3762 write_unlock(&idev->lock);
3763 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3764 ndisc_send_rs(dev, &lladdr,
3765 &in6addr_linklocal_allrouters);
3769 write_lock(&idev->lock);
3770 idev->rs_interval = rfc3315_s14_backoff_update(
3771 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3772 /* The wait after the last probe can be shorter */
3773 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3774 idev->cnf.rtr_solicits) ?
3775 idev->cnf.rtr_solicit_delay :
3779 * Note: we do not support deprecated "all on-link"
3780 * assumption any longer.
3782 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3786 write_unlock(&idev->lock);
3792 * Duplicate Address Detection
3794 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3796 unsigned long rand_num;
3797 struct inet6_dev *idev = ifp->idev;
3800 if (ifp->flags & IFA_F_OPTIMISTIC)
3803 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3806 if (idev->cnf.enhanced_dad ||
3807 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3809 get_random_bytes(&nonce, 6);
3812 ifp->dad_nonce = nonce;
3813 ifp->dad_probes = idev->cnf.dad_transmits;
3814 addrconf_mod_dad_work(ifp, rand_num);
3817 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3819 struct inet6_dev *idev = ifp->idev;
3820 struct net_device *dev = idev->dev;
3821 bool bump_id, notify = false;
3823 addrconf_join_solict(dev, &ifp->addr);
3825 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3827 read_lock_bh(&idev->lock);
3828 spin_lock(&ifp->lock);
3829 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3832 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3833 (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3834 idev->cnf.accept_dad < 1) ||
3835 !(ifp->flags&IFA_F_TENTATIVE) ||
3836 ifp->flags & IFA_F_NODAD) {
3837 bump_id = ifp->flags & IFA_F_TENTATIVE;
3838 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3839 spin_unlock(&ifp->lock);
3840 read_unlock_bh(&idev->lock);
3842 addrconf_dad_completed(ifp, bump_id);
3846 if (!(idev->if_flags & IF_READY)) {
3847 spin_unlock(&ifp->lock);
3848 read_unlock_bh(&idev->lock);
3850 * If the device is not ready:
3851 * - keep it tentative if it is a permanent address.
3852 * - otherwise, kill it.
3855 addrconf_dad_stop(ifp, 0);
3860 * Optimistic nodes can start receiving
3863 if (ifp->flags & IFA_F_OPTIMISTIC) {
3864 ip6_ins_rt(ifp->rt);
3865 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3866 /* Because optimistic nodes can use this address,
3867 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3873 addrconf_dad_kick(ifp);
3875 spin_unlock(&ifp->lock);
3876 read_unlock_bh(&idev->lock);
3878 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3881 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3883 bool begin_dad = false;
3885 spin_lock_bh(&ifp->lock);
3886 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3887 ifp->state = INET6_IFADDR_STATE_PREDAD;
3890 spin_unlock_bh(&ifp->lock);
3893 addrconf_mod_dad_work(ifp, 0);
3896 static void addrconf_dad_work(struct work_struct *w)
3898 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3899 struct inet6_ifaddr,
3901 struct inet6_dev *idev = ifp->idev;
3902 bool bump_id, disable_ipv6 = false;
3903 struct in6_addr mcaddr;
3909 } action = DAD_PROCESS;
3913 spin_lock_bh(&ifp->lock);
3914 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3916 ifp->state = INET6_IFADDR_STATE_DAD;
3917 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3919 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3921 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3922 idev->cnf.accept_dad > 1) &&
3923 !idev->cnf.disable_ipv6 &&
3924 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3925 struct in6_addr addr;
3927 addr.s6_addr32[0] = htonl(0xfe800000);
3928 addr.s6_addr32[1] = 0;
3930 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3931 ipv6_addr_equal(&ifp->addr, &addr)) {
3932 /* DAD failed for link-local based on MAC */
3933 idev->cnf.disable_ipv6 = 1;
3935 pr_info("%s: IPv6 being disabled!\n",
3936 ifp->idev->dev->name);
3937 disable_ipv6 = true;
3941 spin_unlock_bh(&ifp->lock);
3943 if (action == DAD_BEGIN) {
3944 addrconf_dad_begin(ifp);
3946 } else if (action == DAD_ABORT) {
3948 addrconf_dad_stop(ifp, 1);
3950 addrconf_ifdown(idev->dev, 0);
3954 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3957 write_lock_bh(&idev->lock);
3958 if (idev->dead || !(idev->if_flags & IF_READY)) {
3959 write_unlock_bh(&idev->lock);
3963 spin_lock(&ifp->lock);
3964 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3965 spin_unlock(&ifp->lock);
3966 write_unlock_bh(&idev->lock);
3970 if (ifp->dad_probes == 0) {
3972 * DAD was successful
3975 bump_id = ifp->flags & IFA_F_TENTATIVE;
3976 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3977 spin_unlock(&ifp->lock);
3978 write_unlock_bh(&idev->lock);
3980 addrconf_dad_completed(ifp, bump_id);
3986 addrconf_mod_dad_work(ifp,
3987 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3988 spin_unlock(&ifp->lock);
3989 write_unlock_bh(&idev->lock);
3991 /* send a neighbour solicitation for our addr */
3992 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3993 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4000 /* ifp->idev must be at least read locked */
4001 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4003 struct inet6_ifaddr *ifpiter;
4004 struct inet6_dev *idev = ifp->idev;
4006 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4007 if (ifpiter->scope > IFA_LINK)
4009 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4010 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4011 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4018 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
4020 struct net_device *dev = ifp->idev->dev;
4021 struct in6_addr lladdr;
4022 bool send_rs, send_mld;
4024 addrconf_del_dad_work(ifp);
4027 * Configure the address for reception. Now it is valid.
4030 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4032 /* If added prefix is link local and we are prepared to process
4033 router advertisements, start sending router solicitations.
4036 read_lock_bh(&ifp->idev->lock);
4037 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4038 send_rs = send_mld &&
4039 ipv6_accept_ra(ifp->idev) &&
4040 ifp->idev->cnf.rtr_solicits != 0 &&
4041 (dev->flags&IFF_LOOPBACK) == 0;
4042 read_unlock_bh(&ifp->idev->lock);
4044 /* While dad is in progress mld report's source address is in6_addrany.
4045 * Resend with proper ll now.
4048 ipv6_mc_dad_complete(ifp->idev);
4052 * If a host as already performed a random delay
4053 * [...] as part of DAD [...] there is no need
4054 * to delay again before sending the first RS
4056 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4058 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4060 write_lock_bh(&ifp->idev->lock);
4061 spin_lock(&ifp->lock);
4062 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4063 ifp->idev->cnf.rtr_solicit_interval);
4064 ifp->idev->rs_probes = 1;
4065 ifp->idev->if_flags |= IF_RS_SENT;
4066 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4067 spin_unlock(&ifp->lock);
4068 write_unlock_bh(&ifp->idev->lock);
4072 rt_genid_bump_ipv6(dev_net(dev));
4074 /* Make sure that a new temporary address will be created
4075 * before this temporary address becomes deprecated.
4077 if (ifp->flags & IFA_F_TEMPORARY)
4078 addrconf_verify_rtnl();
4081 static void addrconf_dad_run(struct inet6_dev *idev)
4083 struct inet6_ifaddr *ifp;
4085 read_lock_bh(&idev->lock);
4086 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4087 spin_lock(&ifp->lock);
4088 if (ifp->flags & IFA_F_TENTATIVE &&
4089 ifp->state == INET6_IFADDR_STATE_DAD)
4090 addrconf_dad_kick(ifp);
4091 spin_unlock(&ifp->lock);
4093 read_unlock_bh(&idev->lock);
4096 #ifdef CONFIG_PROC_FS
4097 struct if6_iter_state {
4098 struct seq_net_private p;
4103 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4105 struct if6_iter_state *state = seq->private;
4106 struct net *net = seq_file_net(seq);
4107 struct inet6_ifaddr *ifa = NULL;
4110 /* initial bucket if pos is 0 */
4116 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4117 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4119 if (!net_eq(dev_net(ifa->idev->dev), net))
4121 /* sync with offset */
4122 if (p < state->offset) {
4130 /* prepare for next bucket */
4137 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4138 struct inet6_ifaddr *ifa)
4140 struct if6_iter_state *state = seq->private;
4141 struct net *net = seq_file_net(seq);
4143 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4144 if (!net_eq(dev_net(ifa->idev->dev), net))
4150 while (++state->bucket < IN6_ADDR_HSIZE) {
4152 hlist_for_each_entry_rcu(ifa,
4153 &inet6_addr_lst[state->bucket], addr_lst) {
4154 if (!net_eq(dev_net(ifa->idev->dev), net))
4164 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4168 return if6_get_first(seq, *pos);
4171 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4173 struct inet6_ifaddr *ifa;
4175 ifa = if6_get_next(seq, v);
4180 static void if6_seq_stop(struct seq_file *seq, void *v)
4186 static int if6_seq_show(struct seq_file *seq, void *v)
4188 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4189 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4191 ifp->idev->dev->ifindex,
4195 ifp->idev->dev->name);
4199 static const struct seq_operations if6_seq_ops = {
4200 .start = if6_seq_start,
4201 .next = if6_seq_next,
4202 .show = if6_seq_show,
4203 .stop = if6_seq_stop,
4206 static int if6_seq_open(struct inode *inode, struct file *file)
4208 return seq_open_net(inode, file, &if6_seq_ops,
4209 sizeof(struct if6_iter_state));
4212 static const struct file_operations if6_fops = {
4213 .owner = THIS_MODULE,
4214 .open = if6_seq_open,
4216 .llseek = seq_lseek,
4217 .release = seq_release_net,
4220 static int __net_init if6_proc_net_init(struct net *net)
4222 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4227 static void __net_exit if6_proc_net_exit(struct net *net)
4229 remove_proc_entry("if_inet6", net->proc_net);
4232 static struct pernet_operations if6_proc_net_ops = {
4233 .init = if6_proc_net_init,
4234 .exit = if6_proc_net_exit,
4237 int __init if6_proc_init(void)
4239 return register_pernet_subsys(&if6_proc_net_ops);
4242 void if6_proc_exit(void)
4244 unregister_pernet_subsys(&if6_proc_net_ops);
4246 #endif /* CONFIG_PROC_FS */
4248 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4249 /* Check if address is a home address configured on any interface. */
4250 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4252 unsigned int hash = inet6_addr_hash(net, addr);
4253 struct inet6_ifaddr *ifp = NULL;
4257 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4258 if (!net_eq(dev_net(ifp->idev->dev), net))
4260 if (ipv6_addr_equal(&ifp->addr, addr) &&
4261 (ifp->flags & IFA_F_HOMEADDRESS)) {
4272 * Periodic address status verification
4275 static void addrconf_verify_rtnl(void)
4277 unsigned long now, next, next_sec, next_sched;
4278 struct inet6_ifaddr *ifp;
4285 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4287 cancel_delayed_work(&addr_chk_work);
4289 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4291 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4294 /* When setting preferred_lft to a value not zero or
4295 * infinity, while valid_lft is infinity
4296 * IFA_F_PERMANENT has a non-infinity life time.
4298 if ((ifp->flags & IFA_F_PERMANENT) &&
4299 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4302 spin_lock(&ifp->lock);
4303 /* We try to batch several events at once. */
4304 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4306 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4307 age >= ifp->valid_lft) {
4308 spin_unlock(&ifp->lock);
4312 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4313 spin_unlock(&ifp->lock);
4315 } else if (age >= ifp->prefered_lft) {
4316 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4319 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4321 ifp->flags |= IFA_F_DEPRECATED;
4324 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4325 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4326 next = ifp->tstamp + ifp->valid_lft * HZ;
4328 spin_unlock(&ifp->lock);
4333 ipv6_ifa_notify(0, ifp);
4337 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4338 !(ifp->flags&IFA_F_TENTATIVE)) {
4339 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4340 ifp->idev->cnf.dad_transmits *
4341 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4343 if (age >= ifp->prefered_lft - regen_advance) {
4344 struct inet6_ifaddr *ifpub = ifp->ifpub;
4345 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4346 next = ifp->tstamp + ifp->prefered_lft * HZ;
4347 if (!ifp->regen_count && ifpub) {
4350 in6_ifa_hold(ifpub);
4351 spin_unlock(&ifp->lock);
4353 spin_lock(&ifpub->lock);
4354 ifpub->regen_count = 0;
4355 spin_unlock(&ifpub->lock);
4356 ipv6_create_tempaddr(ifpub, ifp, true);
4361 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4362 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4363 spin_unlock(&ifp->lock);
4365 /* ifp->prefered_lft <= ifp->valid_lft */
4366 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4367 next = ifp->tstamp + ifp->prefered_lft * HZ;
4368 spin_unlock(&ifp->lock);
4373 next_sec = round_jiffies_up(next);
4376 /* If rounded timeout is accurate enough, accept it. */
4377 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4378 next_sched = next_sec;
4380 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4381 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4382 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4384 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4385 now, next, next_sec, next_sched);
4386 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4387 rcu_read_unlock_bh();
4390 static void addrconf_verify_work(struct work_struct *w)
4393 addrconf_verify_rtnl();
4397 static void addrconf_verify(void)
4399 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4402 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4403 struct in6_addr **peer_pfx)
4405 struct in6_addr *pfx = NULL;
4410 pfx = nla_data(addr);
4413 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4415 pfx = nla_data(local);
4421 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4422 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4423 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4424 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4425 [IFA_FLAGS] = { .len = sizeof(u32) },
4429 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4430 struct netlink_ext_ack *extack)
4432 struct net *net = sock_net(skb->sk);
4433 struct ifaddrmsg *ifm;
4434 struct nlattr *tb[IFA_MAX+1];
4435 struct in6_addr *pfx, *peer_pfx;
4439 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4444 ifm = nlmsg_data(nlh);
4445 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4449 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4451 /* We ignore other flags so far. */
4452 ifa_flags &= IFA_F_MANAGETEMPADDR;
4454 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4455 ifm->ifa_prefixlen);
4458 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4459 u32 prefered_lft, u32 valid_lft)
4463 unsigned long timeout;
4464 bool was_managetempaddr;
4465 bool had_prefixroute;
4469 if (!valid_lft || (prefered_lft > valid_lft))
4472 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4473 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4476 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4477 if (addrconf_finite_timeout(timeout)) {
4478 expires = jiffies_to_clock_t(timeout * HZ);
4479 valid_lft = timeout;
4480 flags = RTF_EXPIRES;
4484 ifa_flags |= IFA_F_PERMANENT;
4487 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4488 if (addrconf_finite_timeout(timeout)) {
4490 ifa_flags |= IFA_F_DEPRECATED;
4491 prefered_lft = timeout;
4494 spin_lock_bh(&ifp->lock);
4495 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4496 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4497 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4498 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4499 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4500 IFA_F_NOPREFIXROUTE);
4501 ifp->flags |= ifa_flags;
4502 ifp->tstamp = jiffies;
4503 ifp->valid_lft = valid_lft;
4504 ifp->prefered_lft = prefered_lft;
4506 spin_unlock_bh(&ifp->lock);
4507 if (!(ifp->flags&IFA_F_TENTATIVE))
4508 ipv6_ifa_notify(0, ifp);
4510 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4511 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4513 } else if (had_prefixroute) {
4514 enum cleanup_prefix_rt_t action;
4515 unsigned long rt_expires;
4517 write_lock_bh(&ifp->idev->lock);
4518 action = check_cleanup_prefix_route(ifp, &rt_expires);
4519 write_unlock_bh(&ifp->idev->lock);
4521 if (action != CLEANUP_PREFIX_RT_NOP) {
4522 cleanup_prefix_route(ifp, rt_expires,
4523 action == CLEANUP_PREFIX_RT_DEL);
4527 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4528 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4529 valid_lft = prefered_lft = 0;
4530 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4531 !was_managetempaddr, jiffies);
4534 addrconf_verify_rtnl();
4540 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4541 struct netlink_ext_ack *extack)
4543 struct net *net = sock_net(skb->sk);
4544 struct ifaddrmsg *ifm;
4545 struct nlattr *tb[IFA_MAX+1];
4546 struct in6_addr *pfx, *peer_pfx;
4547 struct inet6_ifaddr *ifa;
4548 struct net_device *dev;
4549 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4553 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4558 ifm = nlmsg_data(nlh);
4559 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4563 if (tb[IFA_CACHEINFO]) {
4564 struct ifa_cacheinfo *ci;
4566 ci = nla_data(tb[IFA_CACHEINFO]);
4567 valid_lft = ci->ifa_valid;
4568 preferred_lft = ci->ifa_prefered;
4570 preferred_lft = INFINITY_LIFE_TIME;
4571 valid_lft = INFINITY_LIFE_TIME;
4574 dev = __dev_get_by_index(net, ifm->ifa_index);
4578 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4580 /* We ignore other flags so far. */
4581 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4582 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4584 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4587 * It would be best to check for !NLM_F_CREATE here but
4588 * userspace already relies on not having to provide this.
4590 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4591 ifm->ifa_prefixlen, ifa_flags,
4592 preferred_lft, valid_lft, extack);
4595 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4596 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4599 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4606 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4607 u8 scope, int ifindex)
4609 struct ifaddrmsg *ifm;
4611 ifm = nlmsg_data(nlh);
4612 ifm->ifa_family = AF_INET6;
4613 ifm->ifa_prefixlen = prefixlen;
4614 ifm->ifa_flags = flags;
4615 ifm->ifa_scope = scope;
4616 ifm->ifa_index = ifindex;
4619 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4620 unsigned long tstamp, u32 preferred, u32 valid)
4622 struct ifa_cacheinfo ci;
4624 ci.cstamp = cstamp_delta(cstamp);
4625 ci.tstamp = cstamp_delta(tstamp);
4626 ci.ifa_prefered = preferred;
4627 ci.ifa_valid = valid;
4629 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4632 static inline int rt_scope(int ifa_scope)
4634 if (ifa_scope & IFA_HOST)
4635 return RT_SCOPE_HOST;
4636 else if (ifa_scope & IFA_LINK)
4637 return RT_SCOPE_LINK;
4638 else if (ifa_scope & IFA_SITE)
4639 return RT_SCOPE_SITE;
4641 return RT_SCOPE_UNIVERSE;
4644 static inline int inet6_ifaddr_msgsize(void)
4646 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4647 + nla_total_size(16) /* IFA_LOCAL */
4648 + nla_total_size(16) /* IFA_ADDRESS */
4649 + nla_total_size(sizeof(struct ifa_cacheinfo))
4650 + nla_total_size(4) /* IFA_FLAGS */;
4653 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4654 u32 portid, u32 seq, int event, unsigned int flags)
4656 struct nlmsghdr *nlh;
4657 u32 preferred, valid;
4659 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4663 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4664 ifa->idev->dev->ifindex);
4666 if (!((ifa->flags&IFA_F_PERMANENT) &&
4667 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4668 preferred = ifa->prefered_lft;
4669 valid = ifa->valid_lft;
4670 if (preferred != INFINITY_LIFE_TIME) {
4671 long tval = (jiffies - ifa->tstamp)/HZ;
4672 if (preferred > tval)
4676 if (valid != INFINITY_LIFE_TIME) {
4684 preferred = INFINITY_LIFE_TIME;
4685 valid = INFINITY_LIFE_TIME;
4688 if (!ipv6_addr_any(&ifa->peer_addr)) {
4689 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4690 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4693 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4696 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4699 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4702 nlmsg_end(skb, nlh);
4706 nlmsg_cancel(skb, nlh);
4710 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4711 u32 portid, u32 seq, int event, u16 flags)
4713 struct nlmsghdr *nlh;
4714 u8 scope = RT_SCOPE_UNIVERSE;
4715 int ifindex = ifmca->idev->dev->ifindex;
4717 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4718 scope = RT_SCOPE_SITE;
4720 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4724 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4725 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4726 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4727 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4728 nlmsg_cancel(skb, nlh);
4732 nlmsg_end(skb, nlh);
4736 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4737 u32 portid, u32 seq, int event, unsigned int flags)
4739 struct nlmsghdr *nlh;
4740 u8 scope = RT_SCOPE_UNIVERSE;
4741 int ifindex = ifaca->aca_idev->dev->ifindex;
4743 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4744 scope = RT_SCOPE_SITE;
4746 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4750 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4751 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4752 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4753 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4754 nlmsg_cancel(skb, nlh);
4758 nlmsg_end(skb, nlh);
4768 /* called with rcu_read_lock() */
4769 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4770 struct netlink_callback *cb, enum addr_type_t type,
4771 int s_ip_idx, int *p_ip_idx)
4773 struct ifmcaddr6 *ifmca;
4774 struct ifacaddr6 *ifaca;
4776 int ip_idx = *p_ip_idx;
4778 read_lock_bh(&idev->lock);
4780 case UNICAST_ADDR: {
4781 struct inet6_ifaddr *ifa;
4783 /* unicast address incl. temp addr */
4784 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4785 if (++ip_idx < s_ip_idx)
4787 err = inet6_fill_ifaddr(skb, ifa,
4788 NETLINK_CB(cb->skb).portid,
4794 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4798 case MULTICAST_ADDR:
4799 /* multicast address */
4800 for (ifmca = idev->mc_list; ifmca;
4801 ifmca = ifmca->next, ip_idx++) {
4802 if (ip_idx < s_ip_idx)
4804 err = inet6_fill_ifmcaddr(skb, ifmca,
4805 NETLINK_CB(cb->skb).portid,
4814 /* anycast address */
4815 for (ifaca = idev->ac_list; ifaca;
4816 ifaca = ifaca->aca_next, ip_idx++) {
4817 if (ip_idx < s_ip_idx)
4819 err = inet6_fill_ifacaddr(skb, ifaca,
4820 NETLINK_CB(cb->skb).portid,
4831 read_unlock_bh(&idev->lock);
4836 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4837 enum addr_type_t type)
4839 struct net *net = sock_net(skb->sk);
4842 int s_idx, s_ip_idx;
4843 struct net_device *dev;
4844 struct inet6_dev *idev;
4845 struct hlist_head *head;
4848 s_idx = idx = cb->args[1];
4849 s_ip_idx = ip_idx = cb->args[2];
4852 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4853 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4855 head = &net->dev_index_head[h];
4856 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4859 if (h > s_h || idx > s_idx)
4862 idev = __in6_dev_get(dev);
4866 if (in6_dump_addrs(idev, skb, cb, type,
4867 s_ip_idx, &ip_idx) < 0)
4877 cb->args[2] = ip_idx;
4882 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4884 enum addr_type_t type = UNICAST_ADDR;
4886 return inet6_dump_addr(skb, cb, type);
4889 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4891 enum addr_type_t type = MULTICAST_ADDR;
4893 return inet6_dump_addr(skb, cb, type);
4897 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4899 enum addr_type_t type = ANYCAST_ADDR;
4901 return inet6_dump_addr(skb, cb, type);
4904 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4905 struct netlink_ext_ack *extack)
4907 struct net *net = sock_net(in_skb->sk);
4908 struct ifaddrmsg *ifm;
4909 struct nlattr *tb[IFA_MAX+1];
4910 struct in6_addr *addr = NULL, *peer;
4911 struct net_device *dev = NULL;
4912 struct inet6_ifaddr *ifa;
4913 struct sk_buff *skb;
4916 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4921 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4925 ifm = nlmsg_data(nlh);
4927 dev = dev_get_by_index(net, ifm->ifa_index);
4929 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4931 err = -EADDRNOTAVAIL;
4935 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4941 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4942 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4944 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4945 WARN_ON(err == -EMSGSIZE);
4949 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4958 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4960 struct sk_buff *skb;
4961 struct net *net = dev_net(ifa->idev->dev);
4964 /* Don't send DELADDR notification for TENTATIVE address,
4965 * since NEWADDR notification is sent only after removing
4966 * TENTATIVE flag, if DAD has not failed.
4968 if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
4969 event == RTM_DELADDR)
4972 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4976 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4978 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4979 WARN_ON(err == -EMSGSIZE);
4983 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4987 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4990 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4991 __s32 *array, int bytes)
4993 BUG_ON(bytes < (DEVCONF_MAX * 4));
4995 memset(array, 0, bytes);
4996 array[DEVCONF_FORWARDING] = cnf->forwarding;
4997 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4998 array[DEVCONF_MTU6] = cnf->mtu6;
4999 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5000 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5001 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5002 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5003 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5004 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5005 jiffies_to_msecs(cnf->rtr_solicit_interval);
5006 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5007 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5008 array[DEVCONF_RTR_SOLICIT_DELAY] =
5009 jiffies_to_msecs(cnf->rtr_solicit_delay);
5010 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5011 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5012 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5013 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5014 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5015 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5016 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5017 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5018 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5019 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5020 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5021 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5022 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5023 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5024 #ifdef CONFIG_IPV6_ROUTER_PREF
5025 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5026 array[DEVCONF_RTR_PROBE_INTERVAL] =
5027 jiffies_to_msecs(cnf->rtr_probe_interval);
5028 #ifdef CONFIG_IPV6_ROUTE_INFO
5029 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5030 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5033 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5034 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5035 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5036 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5037 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5039 #ifdef CONFIG_IPV6_MROUTE
5040 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5042 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5043 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5044 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5045 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5046 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5047 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5048 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5049 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5050 /* we omit DEVCONF_STABLE_SECRET for now */
5051 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5052 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5053 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5054 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5055 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5056 #ifdef CONFIG_IPV6_SEG6_HMAC
5057 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5059 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5060 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5061 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5062 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5065 static inline size_t inet6_ifla6_size(void)
5067 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5068 + nla_total_size(sizeof(struct ifla_cacheinfo))
5069 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5070 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5071 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5072 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5075 static inline size_t inet6_if_nlmsg_size(void)
5077 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5078 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5079 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5080 + nla_total_size(4) /* IFLA_MTU */
5081 + nla_total_size(4) /* IFLA_LINK */
5082 + nla_total_size(1) /* IFLA_OPERSTATE */
5083 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5086 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5090 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5093 /* Use put_unaligned() because stats may not be aligned for u64. */
5094 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5095 for (i = 1; i < ICMP6_MIB_MAX; i++)
5096 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5098 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5101 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5102 int bytes, size_t syncpoff)
5105 u64 buff[IPSTATS_MIB_MAX];
5106 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5110 memset(buff, 0, sizeof(buff));
5111 buff[0] = IPSTATS_MIB_MAX;
5113 for_each_possible_cpu(c) {
5114 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5115 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5118 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5119 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5122 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5126 case IFLA_INET6_STATS:
5127 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5128 offsetof(struct ipstats_mib, syncp));
5130 case IFLA_INET6_ICMP6STATS:
5131 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5136 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5137 u32 ext_filter_mask)
5140 struct ifla_cacheinfo ci;
5142 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5143 goto nla_put_failure;
5144 ci.max_reasm_len = IPV6_MAXPLEN;
5145 ci.tstamp = cstamp_delta(idev->tstamp);
5146 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5147 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5148 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5149 goto nla_put_failure;
5150 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5152 goto nla_put_failure;
5153 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5155 /* XXX - MC not implemented */
5157 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5160 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5162 goto nla_put_failure;
5163 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5165 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5167 goto nla_put_failure;
5168 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5170 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5172 goto nla_put_failure;
5174 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5175 goto nla_put_failure;
5177 read_lock_bh(&idev->lock);
5178 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5179 read_unlock_bh(&idev->lock);
5187 static size_t inet6_get_link_af_size(const struct net_device *dev,
5188 u32 ext_filter_mask)
5190 if (!__in6_dev_get(dev))
5193 return inet6_ifla6_size();
5196 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5197 u32 ext_filter_mask)
5199 struct inet6_dev *idev = __in6_dev_get(dev);
5204 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5210 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5212 struct inet6_ifaddr *ifp;
5213 struct net_device *dev = idev->dev;
5214 bool clear_token, update_rs = false;
5215 struct in6_addr ll_addr;
5221 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5223 if (!ipv6_accept_ra(idev))
5225 if (idev->cnf.rtr_solicits == 0)
5228 write_lock_bh(&idev->lock);
5230 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5231 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5233 write_unlock_bh(&idev->lock);
5235 clear_token = ipv6_addr_any(token);
5239 if (!idev->dead && (idev->if_flags & IF_READY) &&
5240 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5241 IFA_F_OPTIMISTIC)) {
5242 /* If we're not ready, then normal ifup will take care
5243 * of this. Otherwise, we need to request our rs here.
5245 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5250 write_lock_bh(&idev->lock);
5253 idev->if_flags |= IF_RS_SENT;
5254 idev->rs_interval = rfc3315_s14_backoff_init(
5255 idev->cnf.rtr_solicit_interval);
5256 idev->rs_probes = 1;
5257 addrconf_mod_rs_timer(idev, idev->rs_interval);
5260 /* Well, that's kinda nasty ... */
5261 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5262 spin_lock(&ifp->lock);
5263 if (ifp->tokenized) {
5265 ifp->prefered_lft = 0;
5267 spin_unlock(&ifp->lock);
5270 write_unlock_bh(&idev->lock);
5271 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5272 addrconf_verify_rtnl();
5276 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5277 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5278 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5281 static int inet6_validate_link_af(const struct net_device *dev,
5282 const struct nlattr *nla)
5284 struct nlattr *tb[IFLA_INET6_MAX + 1];
5286 if (dev && !__in6_dev_get(dev))
5287 return -EAFNOSUPPORT;
5289 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5293 static int check_addr_gen_mode(int mode)
5295 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5296 mode != IN6_ADDR_GEN_MODE_NONE &&
5297 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5298 mode != IN6_ADDR_GEN_MODE_RANDOM)
5303 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5306 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5307 !idev->cnf.stable_secret.initialized &&
5308 !net->ipv6.devconf_dflt->stable_secret.initialized)
5313 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5316 struct inet6_dev *idev = __in6_dev_get(dev);
5317 struct nlattr *tb[IFLA_INET6_MAX + 1];
5320 return -EAFNOSUPPORT;
5322 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5325 if (tb[IFLA_INET6_TOKEN]) {
5326 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5331 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5332 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5334 if (check_addr_gen_mode(mode) < 0 ||
5335 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5338 idev->cnf.addr_gen_mode = mode;
5345 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5346 u32 portid, u32 seq, int event, unsigned int flags)
5348 struct net_device *dev = idev->dev;
5349 struct ifinfomsg *hdr;
5350 struct nlmsghdr *nlh;
5353 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5357 hdr = nlmsg_data(nlh);
5358 hdr->ifi_family = AF_INET6;
5360 hdr->ifi_type = dev->type;
5361 hdr->ifi_index = dev->ifindex;
5362 hdr->ifi_flags = dev_get_flags(dev);
5363 hdr->ifi_change = 0;
5365 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5367 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5368 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5369 (dev->ifindex != dev_get_iflink(dev) &&
5370 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5371 nla_put_u8(skb, IFLA_OPERSTATE,
5372 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5373 goto nla_put_failure;
5374 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5376 goto nla_put_failure;
5378 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5379 goto nla_put_failure;
5381 nla_nest_end(skb, protoinfo);
5382 nlmsg_end(skb, nlh);
5386 nlmsg_cancel(skb, nlh);
5390 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5392 struct net *net = sock_net(skb->sk);
5395 struct net_device *dev;
5396 struct inet6_dev *idev;
5397 struct hlist_head *head;
5400 s_idx = cb->args[1];
5403 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5405 head = &net->dev_index_head[h];
5406 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5409 idev = __in6_dev_get(dev);
5412 if (inet6_fill_ifinfo(skb, idev,
5413 NETLINK_CB(cb->skb).portid,
5415 RTM_NEWLINK, NLM_F_MULTI) < 0)
5429 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5431 struct sk_buff *skb;
5432 struct net *net = dev_net(idev->dev);
5435 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5439 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5441 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5442 WARN_ON(err == -EMSGSIZE);
5446 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5450 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5453 static inline size_t inet6_prefix_nlmsg_size(void)
5455 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5456 + nla_total_size(sizeof(struct in6_addr))
5457 + nla_total_size(sizeof(struct prefix_cacheinfo));
5460 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5461 struct prefix_info *pinfo, u32 portid, u32 seq,
5462 int event, unsigned int flags)
5464 struct prefixmsg *pmsg;
5465 struct nlmsghdr *nlh;
5466 struct prefix_cacheinfo ci;
5468 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5472 pmsg = nlmsg_data(nlh);
5473 pmsg->prefix_family = AF_INET6;
5474 pmsg->prefix_pad1 = 0;
5475 pmsg->prefix_pad2 = 0;
5476 pmsg->prefix_ifindex = idev->dev->ifindex;
5477 pmsg->prefix_len = pinfo->prefix_len;
5478 pmsg->prefix_type = pinfo->type;
5479 pmsg->prefix_pad3 = 0;
5480 pmsg->prefix_flags = 0;
5482 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5483 if (pinfo->autoconf)
5484 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5486 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5487 goto nla_put_failure;
5488 ci.preferred_time = ntohl(pinfo->prefered);
5489 ci.valid_time = ntohl(pinfo->valid);
5490 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5491 goto nla_put_failure;
5492 nlmsg_end(skb, nlh);
5496 nlmsg_cancel(skb, nlh);
5500 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5501 struct prefix_info *pinfo)
5503 struct sk_buff *skb;
5504 struct net *net = dev_net(idev->dev);
5507 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5511 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5513 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5514 WARN_ON(err == -EMSGSIZE);
5518 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5522 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5525 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5527 struct net *net = dev_net(ifp->idev->dev);
5532 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5537 * If the address was optimistic
5538 * we inserted the route at the start of
5539 * our DAD process, so we don't need
5542 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5543 ip6_ins_rt(ifp->rt);
5544 if (ifp->idev->cnf.forwarding)
5545 addrconf_join_anycast(ifp);
5546 if (!ipv6_addr_any(&ifp->peer_addr))
5547 addrconf_prefix_route(&ifp->peer_addr, 128,
5548 ifp->idev->dev, 0, 0);
5551 if (ifp->idev->cnf.forwarding)
5552 addrconf_leave_anycast(ifp);
5553 addrconf_leave_solict(ifp->idev, &ifp->addr);
5554 if (!ipv6_addr_any(&ifp->peer_addr)) {
5555 struct rt6_info *rt;
5557 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5558 ifp->idev->dev, 0, 0);
5563 if (dst_hold_safe(&ifp->rt->dst))
5564 ip6_del_rt(ifp->rt);
5566 rt_genid_bump_ipv6(net);
5569 atomic_inc(&net->ipv6.dev_addr_genid);
5572 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5575 if (likely(ifp->idev->dead == 0))
5576 __ipv6_ifa_notify(event, ifp);
5577 rcu_read_unlock_bh();
5580 #ifdef CONFIG_SYSCTL
5583 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5584 void __user *buffer, size_t *lenp, loff_t *ppos)
5586 int *valp = ctl->data;
5589 struct ctl_table lctl;
5593 * ctl->data points to idev->cnf.forwarding, we should
5594 * not modify it until we get the rtnl lock.
5599 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5602 ret = addrconf_fixup_forwarding(ctl, valp, val);
5609 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5610 void __user *buffer, size_t *lenp, loff_t *ppos)
5612 struct inet6_dev *idev = ctl->extra1;
5613 int min_mtu = IPV6_MIN_MTU;
5614 struct ctl_table lctl;
5617 lctl.extra1 = &min_mtu;
5618 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5620 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5623 static void dev_disable_change(struct inet6_dev *idev)
5625 struct netdev_notifier_info info;
5627 if (!idev || !idev->dev)
5630 netdev_notifier_info_init(&info, idev->dev);
5631 if (idev->cnf.disable_ipv6)
5632 addrconf_notify(NULL, NETDEV_DOWN, &info);
5634 addrconf_notify(NULL, NETDEV_UP, &info);
5637 static void addrconf_disable_change(struct net *net, __s32 newf)
5639 struct net_device *dev;
5640 struct inet6_dev *idev;
5642 for_each_netdev(net, dev) {
5643 idev = __in6_dev_get(dev);
5645 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5646 idev->cnf.disable_ipv6 = newf;
5648 dev_disable_change(idev);
5653 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5658 if (!rtnl_trylock())
5659 return restart_syscall();
5661 net = (struct net *)table->extra2;
5665 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5670 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5671 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5672 addrconf_disable_change(net, newf);
5673 } else if ((!newf) ^ (!old))
5674 dev_disable_change((struct inet6_dev *)table->extra1);
5681 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5682 void __user *buffer, size_t *lenp, loff_t *ppos)
5684 int *valp = ctl->data;
5687 struct ctl_table lctl;
5691 * ctl->data points to idev->cnf.disable_ipv6, we should
5692 * not modify it until we get the rtnl lock.
5697 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5700 ret = addrconf_disable_ipv6(ctl, valp, val);
5707 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5708 void __user *buffer, size_t *lenp, loff_t *ppos)
5710 int *valp = ctl->data;
5715 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5718 if (write && old != new) {
5719 struct net *net = ctl->extra2;
5721 if (!rtnl_trylock())
5722 return restart_syscall();
5724 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5725 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5726 NETCONFA_PROXY_NEIGH,
5727 NETCONFA_IFINDEX_DEFAULT,
5728 net->ipv6.devconf_dflt);
5729 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5730 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5731 NETCONFA_PROXY_NEIGH,
5732 NETCONFA_IFINDEX_ALL,
5733 net->ipv6.devconf_all);
5735 struct inet6_dev *idev = ctl->extra1;
5737 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5738 NETCONFA_PROXY_NEIGH,
5748 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5749 void __user *buffer, size_t *lenp,
5754 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5755 struct net *net = (struct net *)ctl->extra2;
5757 if (!rtnl_trylock())
5758 return restart_syscall();
5760 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5763 new_val = *((int *)ctl->data);
5765 if (check_addr_gen_mode(new_val) < 0) {
5770 /* request for default */
5771 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5772 ipv6_devconf_dflt.addr_gen_mode = new_val;
5774 /* request for individual net device */
5779 if (check_stable_privacy(idev, net, new_val) < 0) {
5784 if (idev->cnf.addr_gen_mode != new_val) {
5785 idev->cnf.addr_gen_mode = new_val;
5786 addrconf_dev_config(idev->dev);
5797 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5798 void __user *buffer, size_t *lenp,
5802 struct in6_addr addr;
5803 char str[IPV6_MAX_STRLEN];
5804 struct ctl_table lctl = *ctl;
5805 struct net *net = ctl->extra2;
5806 struct ipv6_stable_secret *secret = ctl->data;
5808 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5811 lctl.maxlen = IPV6_MAX_STRLEN;
5814 if (!rtnl_trylock())
5815 return restart_syscall();
5817 if (!write && !secret->initialized) {
5822 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5823 if (err >= sizeof(str)) {
5828 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5832 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5837 secret->initialized = true;
5838 secret->secret = addr;
5840 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5841 struct net_device *dev;
5843 for_each_netdev(net, dev) {
5844 struct inet6_dev *idev = __in6_dev_get(dev);
5847 idev->cnf.addr_gen_mode =
5848 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5852 struct inet6_dev *idev = ctl->extra1;
5854 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5864 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5866 void __user *buffer,
5870 int *valp = ctl->data;
5873 struct ctl_table lctl;
5876 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5877 * we should not modify it until we get the rtnl lock.
5882 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5885 ret = addrconf_fixup_linkdown(ctl, valp, val);
5892 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5896 rt->dst.flags |= DST_NOPOLICY;
5898 rt->dst.flags &= ~DST_NOPOLICY;
5903 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5905 struct inet6_ifaddr *ifa;
5907 read_lock_bh(&idev->lock);
5908 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5909 spin_lock(&ifa->lock);
5911 struct rt6_info *rt = ifa->rt;
5915 addrconf_set_nopolicy(ifa->rt, val);
5916 if (rt->rt6i_pcpu) {
5917 for_each_possible_cpu(cpu) {
5918 struct rt6_info **rtp;
5920 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5921 addrconf_set_nopolicy(*rtp, val);
5926 spin_unlock(&ifa->lock);
5928 read_unlock_bh(&idev->lock);
5932 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5934 struct inet6_dev *idev;
5937 if (!rtnl_trylock())
5938 return restart_syscall();
5942 net = (struct net *)ctl->extra2;
5943 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5948 if (valp == &net->ipv6.devconf_all->disable_policy) {
5949 struct net_device *dev;
5951 for_each_netdev(net, dev) {
5952 idev = __in6_dev_get(dev);
5954 addrconf_disable_policy_idev(idev, val);
5957 idev = (struct inet6_dev *)ctl->extra1;
5958 addrconf_disable_policy_idev(idev, val);
5966 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
5967 void __user *buffer, size_t *lenp,
5970 int *valp = ctl->data;
5973 struct ctl_table lctl;
5978 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5980 if (write && (*valp != val))
5981 ret = addrconf_disable_policy(ctl, valp, val);
5989 static int minus_one = -1;
5990 static const int zero = 0;
5991 static const int one = 1;
5992 static const int two_five_five = 255;
5994 static const struct ctl_table addrconf_sysctl[] = {
5996 .procname = "forwarding",
5997 .data = &ipv6_devconf.forwarding,
5998 .maxlen = sizeof(int),
6000 .proc_handler = addrconf_sysctl_forward,
6003 .procname = "hop_limit",
6004 .data = &ipv6_devconf.hop_limit,
6005 .maxlen = sizeof(int),
6007 .proc_handler = proc_dointvec_minmax,
6008 .extra1 = (void *)&one,
6009 .extra2 = (void *)&two_five_five,
6013 .data = &ipv6_devconf.mtu6,
6014 .maxlen = sizeof(int),
6016 .proc_handler = addrconf_sysctl_mtu,
6019 .procname = "accept_ra",
6020 .data = &ipv6_devconf.accept_ra,
6021 .maxlen = sizeof(int),
6023 .proc_handler = proc_dointvec,
6026 .procname = "accept_redirects",
6027 .data = &ipv6_devconf.accept_redirects,
6028 .maxlen = sizeof(int),
6030 .proc_handler = proc_dointvec,
6033 .procname = "autoconf",
6034 .data = &ipv6_devconf.autoconf,
6035 .maxlen = sizeof(int),
6037 .proc_handler = proc_dointvec,
6040 .procname = "dad_transmits",
6041 .data = &ipv6_devconf.dad_transmits,
6042 .maxlen = sizeof(int),
6044 .proc_handler = proc_dointvec,
6047 .procname = "router_solicitations",
6048 .data = &ipv6_devconf.rtr_solicits,
6049 .maxlen = sizeof(int),
6051 .proc_handler = proc_dointvec_minmax,
6052 .extra1 = &minus_one,
6055 .procname = "router_solicitation_interval",
6056 .data = &ipv6_devconf.rtr_solicit_interval,
6057 .maxlen = sizeof(int),
6059 .proc_handler = proc_dointvec_jiffies,
6062 .procname = "router_solicitation_max_interval",
6063 .data = &ipv6_devconf.rtr_solicit_max_interval,
6064 .maxlen = sizeof(int),
6066 .proc_handler = proc_dointvec_jiffies,
6069 .procname = "router_solicitation_delay",
6070 .data = &ipv6_devconf.rtr_solicit_delay,
6071 .maxlen = sizeof(int),
6073 .proc_handler = proc_dointvec_jiffies,
6076 .procname = "force_mld_version",
6077 .data = &ipv6_devconf.force_mld_version,
6078 .maxlen = sizeof(int),
6080 .proc_handler = proc_dointvec,
6083 .procname = "mldv1_unsolicited_report_interval",
6085 &ipv6_devconf.mldv1_unsolicited_report_interval,
6086 .maxlen = sizeof(int),
6088 .proc_handler = proc_dointvec_ms_jiffies,
6091 .procname = "mldv2_unsolicited_report_interval",
6093 &ipv6_devconf.mldv2_unsolicited_report_interval,
6094 .maxlen = sizeof(int),
6096 .proc_handler = proc_dointvec_ms_jiffies,
6099 .procname = "use_tempaddr",
6100 .data = &ipv6_devconf.use_tempaddr,
6101 .maxlen = sizeof(int),
6103 .proc_handler = proc_dointvec,
6106 .procname = "temp_valid_lft",
6107 .data = &ipv6_devconf.temp_valid_lft,
6108 .maxlen = sizeof(int),
6110 .proc_handler = proc_dointvec,
6113 .procname = "temp_prefered_lft",
6114 .data = &ipv6_devconf.temp_prefered_lft,
6115 .maxlen = sizeof(int),
6117 .proc_handler = proc_dointvec,
6120 .procname = "regen_max_retry",
6121 .data = &ipv6_devconf.regen_max_retry,
6122 .maxlen = sizeof(int),
6124 .proc_handler = proc_dointvec,
6127 .procname = "max_desync_factor",
6128 .data = &ipv6_devconf.max_desync_factor,
6129 .maxlen = sizeof(int),
6131 .proc_handler = proc_dointvec,
6134 .procname = "max_addresses",
6135 .data = &ipv6_devconf.max_addresses,
6136 .maxlen = sizeof(int),
6138 .proc_handler = proc_dointvec,
6141 .procname = "accept_ra_defrtr",
6142 .data = &ipv6_devconf.accept_ra_defrtr,
6143 .maxlen = sizeof(int),
6145 .proc_handler = proc_dointvec,
6148 .procname = "accept_ra_min_hop_limit",
6149 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6150 .maxlen = sizeof(int),
6152 .proc_handler = proc_dointvec,
6155 .procname = "accept_ra_pinfo",
6156 .data = &ipv6_devconf.accept_ra_pinfo,
6157 .maxlen = sizeof(int),
6159 .proc_handler = proc_dointvec,
6161 #ifdef CONFIG_IPV6_ROUTER_PREF
6163 .procname = "accept_ra_rtr_pref",
6164 .data = &ipv6_devconf.accept_ra_rtr_pref,
6165 .maxlen = sizeof(int),
6167 .proc_handler = proc_dointvec,
6170 .procname = "router_probe_interval",
6171 .data = &ipv6_devconf.rtr_probe_interval,
6172 .maxlen = sizeof(int),
6174 .proc_handler = proc_dointvec_jiffies,
6176 #ifdef CONFIG_IPV6_ROUTE_INFO
6178 .procname = "accept_ra_rt_info_min_plen",
6179 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6180 .maxlen = sizeof(int),
6182 .proc_handler = proc_dointvec,
6185 .procname = "accept_ra_rt_info_max_plen",
6186 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6187 .maxlen = sizeof(int),
6189 .proc_handler = proc_dointvec,
6194 .procname = "proxy_ndp",
6195 .data = &ipv6_devconf.proxy_ndp,
6196 .maxlen = sizeof(int),
6198 .proc_handler = addrconf_sysctl_proxy_ndp,
6201 .procname = "accept_source_route",
6202 .data = &ipv6_devconf.accept_source_route,
6203 .maxlen = sizeof(int),
6205 .proc_handler = proc_dointvec,
6207 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6209 .procname = "optimistic_dad",
6210 .data = &ipv6_devconf.optimistic_dad,
6211 .maxlen = sizeof(int),
6213 .proc_handler = proc_dointvec,
6216 .procname = "use_optimistic",
6217 .data = &ipv6_devconf.use_optimistic,
6218 .maxlen = sizeof(int),
6220 .proc_handler = proc_dointvec,
6223 #ifdef CONFIG_IPV6_MROUTE
6225 .procname = "mc_forwarding",
6226 .data = &ipv6_devconf.mc_forwarding,
6227 .maxlen = sizeof(int),
6229 .proc_handler = proc_dointvec,
6233 .procname = "disable_ipv6",
6234 .data = &ipv6_devconf.disable_ipv6,
6235 .maxlen = sizeof(int),
6237 .proc_handler = addrconf_sysctl_disable,
6240 .procname = "accept_dad",
6241 .data = &ipv6_devconf.accept_dad,
6242 .maxlen = sizeof(int),
6244 .proc_handler = proc_dointvec,
6247 .procname = "force_tllao",
6248 .data = &ipv6_devconf.force_tllao,
6249 .maxlen = sizeof(int),
6251 .proc_handler = proc_dointvec
6254 .procname = "ndisc_notify",
6255 .data = &ipv6_devconf.ndisc_notify,
6256 .maxlen = sizeof(int),
6258 .proc_handler = proc_dointvec
6261 .procname = "suppress_frag_ndisc",
6262 .data = &ipv6_devconf.suppress_frag_ndisc,
6263 .maxlen = sizeof(int),
6265 .proc_handler = proc_dointvec
6268 .procname = "accept_ra_from_local",
6269 .data = &ipv6_devconf.accept_ra_from_local,
6270 .maxlen = sizeof(int),
6272 .proc_handler = proc_dointvec,
6275 .procname = "accept_ra_mtu",
6276 .data = &ipv6_devconf.accept_ra_mtu,
6277 .maxlen = sizeof(int),
6279 .proc_handler = proc_dointvec,
6282 .procname = "stable_secret",
6283 .data = &ipv6_devconf.stable_secret,
6284 .maxlen = IPV6_MAX_STRLEN,
6286 .proc_handler = addrconf_sysctl_stable_secret,
6289 .procname = "use_oif_addrs_only",
6290 .data = &ipv6_devconf.use_oif_addrs_only,
6291 .maxlen = sizeof(int),
6293 .proc_handler = proc_dointvec,
6296 .procname = "ignore_routes_with_linkdown",
6297 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6298 .maxlen = sizeof(int),
6300 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6303 .procname = "drop_unicast_in_l2_multicast",
6304 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6305 .maxlen = sizeof(int),
6307 .proc_handler = proc_dointvec,
6310 .procname = "drop_unsolicited_na",
6311 .data = &ipv6_devconf.drop_unsolicited_na,
6312 .maxlen = sizeof(int),
6314 .proc_handler = proc_dointvec,
6317 .procname = "keep_addr_on_down",
6318 .data = &ipv6_devconf.keep_addr_on_down,
6319 .maxlen = sizeof(int),
6321 .proc_handler = proc_dointvec,
6325 .procname = "seg6_enabled",
6326 .data = &ipv6_devconf.seg6_enabled,
6327 .maxlen = sizeof(int),
6329 .proc_handler = proc_dointvec,
6331 #ifdef CONFIG_IPV6_SEG6_HMAC
6333 .procname = "seg6_require_hmac",
6334 .data = &ipv6_devconf.seg6_require_hmac,
6335 .maxlen = sizeof(int),
6337 .proc_handler = proc_dointvec,
6341 .procname = "enhanced_dad",
6342 .data = &ipv6_devconf.enhanced_dad,
6343 .maxlen = sizeof(int),
6345 .proc_handler = proc_dointvec,
6348 .procname = "addr_gen_mode",
6349 .data = &ipv6_devconf.addr_gen_mode,
6350 .maxlen = sizeof(int),
6352 .proc_handler = addrconf_sysctl_addr_gen_mode,
6355 .procname = "disable_policy",
6356 .data = &ipv6_devconf.disable_policy,
6357 .maxlen = sizeof(int),
6359 .proc_handler = addrconf_sysctl_disable_policy,
6362 .procname = "ndisc_tclass",
6363 .data = &ipv6_devconf.ndisc_tclass,
6364 .maxlen = sizeof(int),
6366 .proc_handler = proc_dointvec_minmax,
6367 .extra1 = (void *)&zero,
6368 .extra2 = (void *)&two_five_five,
6375 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6376 struct inet6_dev *idev, struct ipv6_devconf *p)
6379 struct ctl_table *table;
6380 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6382 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6386 for (i = 0; table[i].data; i++) {
6387 table[i].data += (char *)p - (char *)&ipv6_devconf;
6388 /* If one of these is already set, then it is not safe to
6389 * overwrite either of them: this makes proc_dointvec_minmax
6392 if (!table[i].extra1 && !table[i].extra2) {
6393 table[i].extra1 = idev; /* embedded; no ref */
6394 table[i].extra2 = net;
6398 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6400 p->sysctl_header = register_net_sysctl(net, path, table);
6401 if (!p->sysctl_header)
6404 if (!strcmp(dev_name, "all"))
6405 ifindex = NETCONFA_IFINDEX_ALL;
6406 else if (!strcmp(dev_name, "default"))
6407 ifindex = NETCONFA_IFINDEX_DEFAULT;
6409 ifindex = idev->dev->ifindex;
6410 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6420 static void __addrconf_sysctl_unregister(struct net *net,
6421 struct ipv6_devconf *p, int ifindex)
6423 struct ctl_table *table;
6425 if (!p->sysctl_header)
6428 table = p->sysctl_header->ctl_table_arg;
6429 unregister_net_sysctl_table(p->sysctl_header);
6430 p->sysctl_header = NULL;
6433 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6436 static int addrconf_sysctl_register(struct inet6_dev *idev)
6440 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6443 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6444 &ndisc_ifinfo_sysctl_change);
6447 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6450 neigh_sysctl_unregister(idev->nd_parms);
6455 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6457 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6458 idev->dev->ifindex);
6459 neigh_sysctl_unregister(idev->nd_parms);
6465 static int __net_init addrconf_init_net(struct net *net)
6468 struct ipv6_devconf *all, *dflt;
6470 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6474 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6476 goto err_alloc_dflt;
6478 /* these will be inherited by all namespaces */
6479 dflt->autoconf = ipv6_defaults.autoconf;
6480 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6482 dflt->stable_secret.initialized = false;
6483 all->stable_secret.initialized = false;
6485 net->ipv6.devconf_all = all;
6486 net->ipv6.devconf_dflt = dflt;
6488 #ifdef CONFIG_SYSCTL
6489 err = __addrconf_sysctl_register(net, "all", NULL, all);
6493 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6499 #ifdef CONFIG_SYSCTL
6501 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6511 static void __net_exit addrconf_exit_net(struct net *net)
6513 #ifdef CONFIG_SYSCTL
6514 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6515 NETCONFA_IFINDEX_DEFAULT);
6516 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6517 NETCONFA_IFINDEX_ALL);
6519 kfree(net->ipv6.devconf_dflt);
6520 kfree(net->ipv6.devconf_all);
6523 static struct pernet_operations addrconf_ops = {
6524 .init = addrconf_init_net,
6525 .exit = addrconf_exit_net,
6528 static struct rtnl_af_ops inet6_ops __read_mostly = {
6530 .fill_link_af = inet6_fill_link_af,
6531 .get_link_af_size = inet6_get_link_af_size,
6532 .validate_link_af = inet6_validate_link_af,
6533 .set_link_af = inet6_set_link_af,
6537 * Init / cleanup code
6540 int __init addrconf_init(void)
6542 struct inet6_dev *idev;
6545 err = ipv6_addr_label_init();
6547 pr_crit("%s: cannot initialize default policy table: %d\n",
6552 err = register_pernet_subsys(&addrconf_ops);
6556 addrconf_wq = create_workqueue("ipv6_addrconf");
6562 /* The addrconf netdev notifier requires that loopback_dev
6563 * has it's ipv6 private information allocated and setup
6564 * before it can bring up and give link-local addresses
6565 * to other devices which are up.
6567 * Unfortunately, loopback_dev is not necessarily the first
6568 * entry in the global dev_base list of net devices. In fact,
6569 * it is likely to be the very last entry on that list.
6570 * So this causes the notifier registry below to try and
6571 * give link-local addresses to all devices besides loopback_dev
6572 * first, then loopback_dev, which cases all the non-loopback_dev
6573 * devices to fail to get a link-local address.
6575 * So, as a temporary fix, allocate the ipv6 structure for
6576 * loopback_dev first by hand.
6577 * Longer term, all of the dependencies ipv6 has upon the loopback
6578 * device and it being up should be removed.
6581 idev = ipv6_add_dev(init_net.loopback_dev);
6584 err = PTR_ERR(idev);
6588 ip6_route_init_special_entries();
6590 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6591 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6593 register_netdevice_notifier(&ipv6_dev_notf);
6597 rtnl_af_register(&inet6_ops);
6599 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6604 /* Only the first call to __rtnl_register can fail */
6605 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, 0);
6606 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, 0);
6607 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6608 inet6_dump_ifaddr, RTNL_FLAG_DOIT_UNLOCKED);
6609 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6610 inet6_dump_ifmcaddr, 0);
6611 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6612 inet6_dump_ifacaddr, 0);
6613 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6614 inet6_netconf_dump_devconf, RTNL_FLAG_DOIT_UNLOCKED);
6616 ipv6_addr_label_rtnl_register();
6620 rtnl_af_unregister(&inet6_ops);
6621 unregister_netdevice_notifier(&ipv6_dev_notf);
6623 destroy_workqueue(addrconf_wq);
6625 unregister_pernet_subsys(&addrconf_ops);
6627 ipv6_addr_label_cleanup();
6632 void addrconf_cleanup(void)
6634 struct net_device *dev;
6637 unregister_netdevice_notifier(&ipv6_dev_notf);
6638 unregister_pernet_subsys(&addrconf_ops);
6639 ipv6_addr_label_cleanup();
6641 rtnl_af_unregister(&inet6_ops);
6645 /* clean dev list */
6646 for_each_netdev(&init_net, dev) {
6647 if (__in6_dev_get(dev) == NULL)
6649 addrconf_ifdown(dev, 1);
6651 addrconf_ifdown(init_net.loopback_dev, 2);
6656 spin_lock_bh(&addrconf_hash_lock);
6657 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6658 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6659 spin_unlock_bh(&addrconf_hash_lock);
6660 cancel_delayed_work(&addr_chk_work);
6663 destroy_workqueue(addrconf_wq);