2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
62 #include <linux/sysctl.h>
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
70 #include <net/net_namespace.h>
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
97 /* Set to 3 to get tracing... */
101 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
103 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
106 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108 #define IPV6_MAX_STRLEN \
109 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
111 static inline u32 cstamp_delta(unsigned long cstamp)
113 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
116 static inline s32 rfc3315_s14_backoff_init(s32 irt)
118 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
119 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
120 do_div(tmp, 1000000);
124 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
126 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
127 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
128 do_div(tmp, 1000000);
129 if ((s32)tmp > mrt) {
130 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
131 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
132 do_div(tmp, 1000000);
138 static int addrconf_sysctl_register(struct inet6_dev *idev);
139 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
141 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
146 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
151 static void ipv6_regen_rndid(struct inet6_dev *idev);
152 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
154 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
155 static int ipv6_count_addresses(struct inet6_dev *idev);
156 static int ipv6_generate_stable_address(struct in6_addr *addr,
158 const struct inet6_dev *idev);
161 * Configured unicast address hash table
163 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
164 static DEFINE_SPINLOCK(addrconf_hash_lock);
166 static void addrconf_verify(void);
167 static void addrconf_verify_rtnl(void);
168 static void addrconf_verify_work(struct work_struct *);
170 static struct workqueue_struct *addrconf_wq;
171 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
173 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
174 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
176 static void addrconf_type_change(struct net_device *dev,
177 unsigned long event);
178 static int addrconf_ifdown(struct net_device *dev, int how);
180 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
182 const struct net_device *dev,
183 u32 flags, u32 noflags);
185 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
186 static void addrconf_dad_work(struct work_struct *w);
187 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
189 static void addrconf_dad_run(struct inet6_dev *idev);
190 static void addrconf_rs_timer(unsigned long data);
191 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
192 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
194 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
195 struct prefix_info *pinfo);
196 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
197 struct net_device *dev);
199 static struct ipv6_devconf ipv6_devconf __read_mostly = {
201 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
202 .mtu6 = IPV6_MIN_MTU,
204 .accept_redirects = 1,
206 .force_mld_version = 0,
207 .mldv1_unsolicited_report_interval = 10 * HZ,
208 .mldv2_unsolicited_report_interval = HZ,
210 .rtr_solicits = MAX_RTR_SOLICITATIONS,
211 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
212 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
213 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
215 .temp_valid_lft = TEMP_VALID_LIFETIME,
216 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
217 .regen_max_retry = REGEN_MAX_RETRY,
218 .max_desync_factor = MAX_DESYNC_FACTOR,
219 .max_addresses = IPV6_MAX_ADDRESSES,
220 .accept_ra_defrtr = 1,
221 .accept_ra_from_local = 0,
222 .accept_ra_min_hop_limit= 1,
223 .accept_ra_pinfo = 1,
224 #ifdef CONFIG_IPV6_ROUTER_PREF
225 .accept_ra_rtr_pref = 1,
226 .rtr_probe_interval = 60 * HZ,
227 #ifdef CONFIG_IPV6_ROUTE_INFO
228 .accept_ra_rt_info_min_plen = 0,
229 .accept_ra_rt_info_max_plen = 0,
233 .accept_source_route = 0, /* we do not accept RH0 by default. */
236 .suppress_frag_ndisc = 1,
239 .initialized = false,
241 .use_oif_addrs_only = 0,
242 .ignore_routes_with_linkdown = 0,
243 .keep_addr_on_down = 0,
245 #ifdef CONFIG_IPV6_SEG6_HMAC
246 .seg6_require_hmac = 0,
249 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
253 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
255 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
256 .mtu6 = IPV6_MIN_MTU,
258 .accept_redirects = 1,
260 .force_mld_version = 0,
261 .mldv1_unsolicited_report_interval = 10 * HZ,
262 .mldv2_unsolicited_report_interval = HZ,
264 .rtr_solicits = MAX_RTR_SOLICITATIONS,
265 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
266 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
267 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
269 .temp_valid_lft = TEMP_VALID_LIFETIME,
270 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
271 .regen_max_retry = REGEN_MAX_RETRY,
272 .max_desync_factor = MAX_DESYNC_FACTOR,
273 .max_addresses = IPV6_MAX_ADDRESSES,
274 .accept_ra_defrtr = 1,
275 .accept_ra_from_local = 0,
276 .accept_ra_min_hop_limit= 1,
277 .accept_ra_pinfo = 1,
278 #ifdef CONFIG_IPV6_ROUTER_PREF
279 .accept_ra_rtr_pref = 1,
280 .rtr_probe_interval = 60 * HZ,
281 #ifdef CONFIG_IPV6_ROUTE_INFO
282 .accept_ra_rt_info_min_plen = 0,
283 .accept_ra_rt_info_max_plen = 0,
287 .accept_source_route = 0, /* we do not accept RH0 by default. */
290 .suppress_frag_ndisc = 1,
293 .initialized = false,
295 .use_oif_addrs_only = 0,
296 .ignore_routes_with_linkdown = 0,
297 .keep_addr_on_down = 0,
299 #ifdef CONFIG_IPV6_SEG6_HMAC
300 .seg6_require_hmac = 0,
303 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
307 /* Check if link is ready: is it up and is a valid qdisc available */
308 static inline bool addrconf_link_ready(const struct net_device *dev)
310 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
315 if (del_timer(&idev->rs_timer))
319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
321 if (cancel_delayed_work(&ifp->dad_work))
325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
328 if (!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 netconfmsg *ncm;
621 struct ipv6_devconf *devconf;
622 struct inet6_dev *in6_dev;
623 struct net_device *dev;
627 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
628 devconf_ipv6_policy, extack);
633 if (!tb[NETCONFA_IFINDEX])
636 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
638 case NETCONFA_IFINDEX_ALL:
639 devconf = net->ipv6.devconf_all;
641 case NETCONFA_IFINDEX_DEFAULT:
642 devconf = net->ipv6.devconf_dflt;
645 dev = __dev_get_by_index(net, ifindex);
648 in6_dev = __in6_dev_get(dev);
651 devconf = &in6_dev->cnf;
656 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
660 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
661 NETLINK_CB(in_skb).portid,
662 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
665 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
666 WARN_ON(err == -EMSGSIZE);
670 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
675 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
676 struct netlink_callback *cb)
678 struct net *net = sock_net(skb->sk);
681 struct net_device *dev;
682 struct inet6_dev *idev;
683 struct hlist_head *head;
686 s_idx = idx = cb->args[1];
688 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
690 head = &net->dev_index_head[h];
692 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
694 hlist_for_each_entry_rcu(dev, head, index_hlist) {
697 idev = __in6_dev_get(dev);
701 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
703 NETLINK_CB(cb->skb).portid,
711 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
717 if (h == NETDEV_HASHENTRIES) {
718 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
719 net->ipv6.devconf_all,
720 NETLINK_CB(cb->skb).portid,
722 RTM_NEWNETCONF, NLM_F_MULTI,
728 if (h == NETDEV_HASHENTRIES + 1) {
729 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
730 net->ipv6.devconf_dflt,
731 NETLINK_CB(cb->skb).portid,
733 RTM_NEWNETCONF, NLM_F_MULTI,
747 static void dev_forward_change(struct inet6_dev *idev)
749 struct net_device *dev;
750 struct inet6_ifaddr *ifa;
755 if (idev->cnf.forwarding)
756 dev_disable_lro(dev);
757 if (dev->flags & IFF_MULTICAST) {
758 if (idev->cnf.forwarding) {
759 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
760 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
761 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
763 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
764 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
765 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
769 list_for_each_entry(ifa, &idev->addr_list, if_list) {
770 if (ifa->flags&IFA_F_TENTATIVE)
772 if (idev->cnf.forwarding)
773 addrconf_join_anycast(ifa);
775 addrconf_leave_anycast(ifa);
777 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
779 dev->ifindex, &idev->cnf);
783 static void addrconf_forward_change(struct net *net, __s32 newf)
785 struct net_device *dev;
786 struct inet6_dev *idev;
788 for_each_netdev(net, dev) {
789 idev = __in6_dev_get(dev);
791 int changed = (!idev->cnf.forwarding) ^ (!newf);
792 idev->cnf.forwarding = newf;
794 dev_forward_change(idev);
799 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
805 return restart_syscall();
807 net = (struct net *)table->extra2;
811 if (p == &net->ipv6.devconf_dflt->forwarding) {
812 if ((!newf) ^ (!old))
813 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
815 NETCONFA_IFINDEX_DEFAULT,
816 net->ipv6.devconf_dflt);
821 if (p == &net->ipv6.devconf_all->forwarding) {
822 int old_dflt = net->ipv6.devconf_dflt->forwarding;
824 net->ipv6.devconf_dflt->forwarding = newf;
825 if ((!newf) ^ (!old_dflt))
826 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
828 NETCONFA_IFINDEX_DEFAULT,
829 net->ipv6.devconf_dflt);
831 addrconf_forward_change(net, newf);
832 if ((!newf) ^ (!old))
833 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
835 NETCONFA_IFINDEX_ALL,
836 net->ipv6.devconf_all);
837 } else if ((!newf) ^ (!old))
838 dev_forward_change((struct inet6_dev *)table->extra1);
842 rt6_purge_dflt_routers(net);
846 static void addrconf_linkdown_change(struct net *net, __s32 newf)
848 struct net_device *dev;
849 struct inet6_dev *idev;
851 for_each_netdev(net, dev) {
852 idev = __in6_dev_get(dev);
854 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
856 idev->cnf.ignore_routes_with_linkdown = newf;
858 inet6_netconf_notify_devconf(dev_net(dev),
860 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
867 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
873 return restart_syscall();
875 net = (struct net *)table->extra2;
879 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
880 if ((!newf) ^ (!old))
881 inet6_netconf_notify_devconf(net,
883 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
884 NETCONFA_IFINDEX_DEFAULT,
885 net->ipv6.devconf_dflt);
890 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
891 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
892 addrconf_linkdown_change(net, newf);
893 if ((!newf) ^ (!old))
894 inet6_netconf_notify_devconf(net,
896 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
897 NETCONFA_IFINDEX_ALL,
898 net->ipv6.devconf_all);
907 /* Nobody refers to this ifaddr, destroy it */
908 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
910 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
912 #ifdef NET_REFCNT_DEBUG
913 pr_debug("%s\n", __func__);
916 in6_dev_put(ifp->idev);
918 if (cancel_delayed_work(&ifp->dad_work))
919 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
922 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
923 pr_warn("Freeing alive inet6 address %p\n", ifp);
932 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
935 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
938 * Each device address list is sorted in order of scope -
939 * global before linklocal.
941 list_for_each(p, &idev->addr_list) {
942 struct inet6_ifaddr *ifa
943 = list_entry(p, struct inet6_ifaddr, if_list);
944 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
948 list_add_tail(&ifp->if_list, p);
951 static u32 inet6_addr_hash(const struct in6_addr *addr)
953 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
956 /* On success it returns ifp with increased reference count */
958 static struct inet6_ifaddr *
959 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
960 const struct in6_addr *peer_addr, int pfxlen,
961 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
963 struct net *net = dev_net(idev->dev);
964 struct inet6_ifaddr *ifa = NULL;
966 struct in6_validator_info i6vi;
969 int addr_type = ipv6_addr_type(addr);
971 if (addr_type == IPV6_ADDR_ANY ||
972 addr_type & IPV6_ADDR_MULTICAST ||
973 (!(idev->dev->flags & IFF_LOOPBACK) &&
974 addr_type & IPV6_ADDR_LOOPBACK))
975 return ERR_PTR(-EADDRNOTAVAIL);
982 err = -ENODEV; /*XXX*/
986 if (idev->cnf.disable_ipv6) {
991 i6vi.i6vi_addr = *addr;
992 i6vi.i6vi_dev = idev;
993 rcu_read_unlock_bh();
995 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
998 err = notifier_to_errno(err);
1002 spin_lock(&addrconf_hash_lock);
1004 /* Ignore adding duplicate addresses on an interface */
1005 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
1006 ADBG("ipv6_add_addr: already assigned\n");
1011 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
1014 ADBG("ipv6_add_addr: malloc failed\n");
1019 rt = addrconf_dst_alloc(idev, addr, false);
1025 if (net->ipv6.devconf_all->disable_policy ||
1026 idev->cnf.disable_policy)
1027 rt->dst.flags |= DST_NOPOLICY;
1029 neigh_parms_data_state_setall(idev->nd_parms);
1033 ifa->peer_addr = *peer_addr;
1035 spin_lock_init(&ifa->lock);
1036 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1037 INIT_HLIST_NODE(&ifa->addr_lst);
1039 ifa->prefix_len = pfxlen;
1041 /* No need to add the TENTATIVE flag for addresses with NODAD */
1042 if (!(flags & IFA_F_NODAD))
1043 ifa->flags |= IFA_F_TENTATIVE;
1044 ifa->valid_lft = valid_lft;
1045 ifa->prefered_lft = prefered_lft;
1046 ifa->cstamp = ifa->tstamp = jiffies;
1047 ifa->tokenized = false;
1053 refcount_set(&ifa->refcnt, 1);
1055 /* Add to big hash table */
1056 hash = inet6_addr_hash(addr);
1058 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1059 spin_unlock(&addrconf_hash_lock);
1061 write_lock(&idev->lock);
1062 /* Add to inet6_dev unicast addr list. */
1063 ipv6_link_dev_addr(idev, ifa);
1065 if (ifa->flags&IFA_F_TEMPORARY) {
1066 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1071 write_unlock(&idev->lock);
1073 rcu_read_unlock_bh();
1075 if (likely(err == 0))
1076 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1085 spin_unlock(&addrconf_hash_lock);
1089 enum cleanup_prefix_rt_t {
1090 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1091 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1092 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1096 * Check, whether the prefix for ifp would still need a prefix route
1097 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1100 * 1) we don't purge prefix if address was not permanent.
1101 * prefix is managed by its own lifetime.
1102 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1103 * 3) if there are no addresses, delete prefix.
1104 * 4) if there are still other permanent address(es),
1105 * corresponding prefix is still permanent.
1106 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1107 * don't purge the prefix, assume user space is managing it.
1108 * 6) otherwise, update prefix lifetime to the
1109 * longest valid lifetime among the corresponding
1110 * addresses on the device.
1111 * Note: subsequent RA will update lifetime.
1113 static enum cleanup_prefix_rt_t
1114 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1116 struct inet6_ifaddr *ifa;
1117 struct inet6_dev *idev = ifp->idev;
1118 unsigned long lifetime;
1119 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1123 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1126 if (ifa->prefix_len != ifp->prefix_len ||
1127 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1130 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1131 return CLEANUP_PREFIX_RT_NOP;
1133 action = CLEANUP_PREFIX_RT_EXPIRE;
1135 spin_lock(&ifa->lock);
1137 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1139 * Note: Because this address is
1140 * not permanent, lifetime <
1141 * LONG_MAX / HZ here.
1143 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1144 *expires = ifa->tstamp + lifetime * HZ;
1145 spin_unlock(&ifa->lock);
1152 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1154 struct rt6_info *rt;
1156 rt = addrconf_get_prefix_route(&ifp->addr,
1159 0, RTF_GATEWAY | RTF_DEFAULT);
1164 if (!(rt->rt6i_flags & RTF_EXPIRES))
1165 rt6_set_expires(rt, expires);
1172 /* This function wants to get referenced ifp and releases it before return */
1174 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1177 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1178 unsigned long expires;
1182 spin_lock_bh(&ifp->lock);
1184 ifp->state = INET6_IFADDR_STATE_DEAD;
1185 spin_unlock_bh(&ifp->lock);
1187 if (state == INET6_IFADDR_STATE_DEAD)
1190 spin_lock_bh(&addrconf_hash_lock);
1191 hlist_del_init_rcu(&ifp->addr_lst);
1192 spin_unlock_bh(&addrconf_hash_lock);
1194 write_lock_bh(&ifp->idev->lock);
1196 if (ifp->flags&IFA_F_TEMPORARY) {
1197 list_del(&ifp->tmp_list);
1199 in6_ifa_put(ifp->ifpub);
1205 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1206 action = check_cleanup_prefix_route(ifp, &expires);
1208 list_del_init(&ifp->if_list);
1211 write_unlock_bh(&ifp->idev->lock);
1213 addrconf_del_dad_work(ifp);
1215 ipv6_ifa_notify(RTM_DELADDR, ifp);
1217 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1219 if (action != CLEANUP_PREFIX_RT_NOP) {
1220 cleanup_prefix_route(ifp, expires,
1221 action == CLEANUP_PREFIX_RT_DEL);
1224 /* clean up prefsrc entries */
1225 rt6_remove_prefsrc(ifp);
1230 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1232 struct inet6_dev *idev = ifp->idev;
1233 struct in6_addr addr, *tmpaddr;
1234 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1235 unsigned long regen_advance;
1239 unsigned long now = jiffies;
1240 long max_desync_factor;
1241 s32 cnf_temp_preferred_lft;
1243 write_lock_bh(&idev->lock);
1245 spin_lock_bh(&ift->lock);
1246 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1247 spin_unlock_bh(&ift->lock);
1254 if (idev->cnf.use_tempaddr <= 0) {
1255 write_unlock_bh(&idev->lock);
1256 pr_info("%s: use_tempaddr is disabled\n", __func__);
1261 spin_lock_bh(&ifp->lock);
1262 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1263 idev->cnf.use_tempaddr = -1; /*XXX*/
1264 spin_unlock_bh(&ifp->lock);
1265 write_unlock_bh(&idev->lock);
1266 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1273 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1274 ipv6_try_regen_rndid(idev, tmpaddr);
1275 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1276 age = (now - ifp->tstamp) / HZ;
1278 regen_advance = idev->cnf.regen_max_retry *
1279 idev->cnf.dad_transmits *
1280 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1282 /* recalculate max_desync_factor each time and update
1283 * idev->desync_factor if it's larger
1285 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1286 max_desync_factor = min_t(long,
1287 idev->cnf.max_desync_factor,
1288 cnf_temp_preferred_lft - regen_advance);
1290 if (unlikely(idev->desync_factor > max_desync_factor)) {
1291 if (max_desync_factor > 0) {
1292 get_random_bytes(&idev->desync_factor,
1293 sizeof(idev->desync_factor));
1294 idev->desync_factor %= max_desync_factor;
1296 idev->desync_factor = 0;
1300 tmp_valid_lft = min_t(__u32,
1302 idev->cnf.temp_valid_lft + age);
1303 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1304 idev->desync_factor;
1305 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1306 tmp_plen = ifp->prefix_len;
1307 tmp_tstamp = ifp->tstamp;
1308 spin_unlock_bh(&ifp->lock);
1310 write_unlock_bh(&idev->lock);
1312 /* A temporary address is created only if this calculated Preferred
1313 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1314 * an implementation must not create a temporary address with a zero
1315 * Preferred Lifetime.
1316 * Use age calculation as in addrconf_verify to avoid unnecessary
1317 * temporary addresses being generated.
1319 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1320 if (tmp_prefered_lft <= regen_advance + age) {
1327 addr_flags = IFA_F_TEMPORARY;
1328 /* set in addrconf_prefix_rcv() */
1329 if (ifp->flags & IFA_F_OPTIMISTIC)
1330 addr_flags |= IFA_F_OPTIMISTIC;
1332 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1333 ipv6_addr_scope(&addr), addr_flags,
1334 tmp_valid_lft, tmp_prefered_lft);
1338 pr_info("%s: retry temporary address regeneration\n", __func__);
1340 write_lock_bh(&idev->lock);
1344 spin_lock_bh(&ift->lock);
1347 ift->tstamp = tmp_tstamp;
1348 spin_unlock_bh(&ift->lock);
1350 addrconf_dad_start(ift);
1358 * Choose an appropriate source address (RFC3484)
1361 IPV6_SADDR_RULE_INIT = 0,
1362 IPV6_SADDR_RULE_LOCAL,
1363 IPV6_SADDR_RULE_SCOPE,
1364 IPV6_SADDR_RULE_PREFERRED,
1365 #ifdef CONFIG_IPV6_MIP6
1366 IPV6_SADDR_RULE_HOA,
1368 IPV6_SADDR_RULE_OIF,
1369 IPV6_SADDR_RULE_LABEL,
1370 IPV6_SADDR_RULE_PRIVACY,
1371 IPV6_SADDR_RULE_ORCHID,
1372 IPV6_SADDR_RULE_PREFIX,
1373 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1374 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1379 struct ipv6_saddr_score {
1382 struct inet6_ifaddr *ifa;
1383 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1388 struct ipv6_saddr_dst {
1389 const struct in6_addr *addr;
1396 static inline int ipv6_saddr_preferred(int type)
1398 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1403 static bool ipv6_use_optimistic_addr(struct net *net,
1404 struct inet6_dev *idev)
1406 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1409 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1411 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1420 static int ipv6_get_saddr_eval(struct net *net,
1421 struct ipv6_saddr_score *score,
1422 struct ipv6_saddr_dst *dst,
1427 if (i <= score->rule) {
1429 case IPV6_SADDR_RULE_SCOPE:
1430 ret = score->scopedist;
1432 case IPV6_SADDR_RULE_PREFIX:
1433 ret = score->matchlen;
1436 ret = !!test_bit(i, score->scorebits);
1442 case IPV6_SADDR_RULE_INIT:
1443 /* Rule 0: remember if hiscore is not ready yet */
1446 case IPV6_SADDR_RULE_LOCAL:
1447 /* Rule 1: Prefer same address */
1448 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1450 case IPV6_SADDR_RULE_SCOPE:
1451 /* Rule 2: Prefer appropriate scope
1456 * ---+--+-+---> scope
1458 * | d is scope of the destination.
1460 * | \ <- smaller scope is better if
1461 * B-15 | \ if scope is enough for destination.
1462 * | ret = B - scope (-1 <= scope >= d <= 15).
1464 * |/ <- greater is better
1465 * -C / if scope is not enough for destination.
1466 * /| ret = scope - C (-1 <= d < scope <= 15).
1468 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1469 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1470 * Assume B = 0 and we get C > 29.
1472 ret = __ipv6_addr_src_scope(score->addr_type);
1473 if (ret >= dst->scope)
1476 ret -= 128; /* 30 is enough */
1477 score->scopedist = ret;
1479 case IPV6_SADDR_RULE_PREFERRED:
1481 /* Rule 3: Avoid deprecated and optimistic addresses */
1482 u8 avoid = IFA_F_DEPRECATED;
1484 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1485 avoid |= IFA_F_OPTIMISTIC;
1486 ret = ipv6_saddr_preferred(score->addr_type) ||
1487 !(score->ifa->flags & avoid);
1490 #ifdef CONFIG_IPV6_MIP6
1491 case IPV6_SADDR_RULE_HOA:
1493 /* Rule 4: Prefer home address */
1494 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1495 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1499 case IPV6_SADDR_RULE_OIF:
1500 /* Rule 5: Prefer outgoing interface */
1501 ret = (!dst->ifindex ||
1502 dst->ifindex == score->ifa->idev->dev->ifindex);
1504 case IPV6_SADDR_RULE_LABEL:
1505 /* Rule 6: Prefer matching label */
1506 ret = ipv6_addr_label(net,
1507 &score->ifa->addr, score->addr_type,
1508 score->ifa->idev->dev->ifindex) == dst->label;
1510 case IPV6_SADDR_RULE_PRIVACY:
1512 /* Rule 7: Prefer public address
1513 * Note: prefer temporary address if use_tempaddr >= 2
1515 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1516 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1517 score->ifa->idev->cnf.use_tempaddr >= 2;
1518 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1521 case IPV6_SADDR_RULE_ORCHID:
1522 /* Rule 8-: Prefer ORCHID vs ORCHID or
1523 * non-ORCHID vs non-ORCHID
1525 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1526 ipv6_addr_orchid(dst->addr));
1528 case IPV6_SADDR_RULE_PREFIX:
1529 /* Rule 8: Use longest matching prefix */
1530 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1531 if (ret > score->ifa->prefix_len)
1532 ret = score->ifa->prefix_len;
1533 score->matchlen = ret;
1535 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1536 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1537 /* Optimistic addresses still have lower precedence than other
1538 * preferred addresses.
1540 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1548 __set_bit(i, score->scorebits);
1554 static int __ipv6_dev_get_saddr(struct net *net,
1555 struct ipv6_saddr_dst *dst,
1556 struct inet6_dev *idev,
1557 struct ipv6_saddr_score *scores,
1560 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1562 read_lock_bh(&idev->lock);
1563 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1567 * - Tentative Address (RFC2462 section 5.4)
1568 * - A tentative address is not considered
1569 * "assigned to an interface" in the traditional
1570 * sense, unless it is also flagged as optimistic.
1571 * - Candidate Source Address (section 4)
1572 * - In any case, anycast addresses, multicast
1573 * addresses, and the unspecified address MUST
1574 * NOT be included in a candidate set.
1576 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1577 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1580 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1582 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1583 score->addr_type & IPV6_ADDR_MULTICAST)) {
1584 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1590 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1592 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1593 int minihiscore, miniscore;
1595 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1596 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1598 if (minihiscore > miniscore) {
1599 if (i == IPV6_SADDR_RULE_SCOPE &&
1600 score->scopedist > 0) {
1603 * each remaining entry
1604 * has too small (not enough)
1605 * scope, because ifa entries
1606 * are sorted by their scope
1612 } else if (minihiscore < miniscore) {
1614 in6_ifa_put(hiscore->ifa);
1616 in6_ifa_hold(score->ifa);
1618 swap(hiscore, score);
1619 hiscore_idx = 1 - hiscore_idx;
1621 /* restore our iterator */
1622 score->ifa = hiscore->ifa;
1629 read_unlock_bh(&idev->lock);
1633 static int ipv6_get_saddr_master(struct net *net,
1634 const struct net_device *dst_dev,
1635 const struct net_device *master,
1636 struct ipv6_saddr_dst *dst,
1637 struct ipv6_saddr_score *scores,
1640 struct inet6_dev *idev;
1642 idev = __in6_dev_get(dst_dev);
1644 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1645 scores, hiscore_idx);
1647 idev = __in6_dev_get(master);
1649 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1650 scores, hiscore_idx);
1655 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1656 const struct in6_addr *daddr, unsigned int prefs,
1657 struct in6_addr *saddr)
1659 struct ipv6_saddr_score scores[2], *hiscore;
1660 struct ipv6_saddr_dst dst;
1661 struct inet6_dev *idev;
1662 struct net_device *dev;
1664 bool use_oif_addr = false;
1665 int hiscore_idx = 0;
1667 dst_type = __ipv6_addr_type(daddr);
1669 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1670 dst.scope = __ipv6_addr_src_scope(dst_type);
1671 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1674 scores[hiscore_idx].rule = -1;
1675 scores[hiscore_idx].ifa = NULL;
1679 /* Candidate Source Address (section 4)
1680 * - multicast and link-local destination address,
1681 * the set of candidate source address MUST only
1682 * include addresses assigned to interfaces
1683 * belonging to the same link as the outgoing
1685 * (- For site-local destination addresses, the
1686 * set of candidate source addresses MUST only
1687 * include addresses assigned to interfaces
1688 * belonging to the same site as the outgoing
1690 * - "It is RECOMMENDED that the candidate source addresses
1691 * be the set of unicast addresses assigned to the
1692 * interface that will be used to send to the destination
1693 * (the 'outgoing' interface)." (RFC 6724)
1696 idev = __in6_dev_get(dst_dev);
1697 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1698 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1699 (idev && idev->cnf.use_oif_addrs_only)) {
1700 use_oif_addr = true;
1706 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1708 const struct net_device *master;
1711 /* if dst_dev exists and is enslaved to an L3 device, then
1712 * prefer addresses from dst_dev and then the master over
1713 * any other enslaved devices in the L3 domain.
1715 master = l3mdev_master_dev_rcu(dst_dev);
1717 master_idx = master->ifindex;
1719 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1721 scores, hiscore_idx);
1723 if (scores[hiscore_idx].ifa)
1727 for_each_netdev_rcu(net, dev) {
1728 /* only consider addresses on devices in the
1731 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1733 idev = __in6_dev_get(dev);
1736 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1743 hiscore = &scores[hiscore_idx];
1745 return -EADDRNOTAVAIL;
1747 *saddr = hiscore->ifa->addr;
1748 in6_ifa_put(hiscore->ifa);
1751 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1753 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1756 struct inet6_ifaddr *ifp;
1757 int err = -EADDRNOTAVAIL;
1759 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1760 if (ifp->scope > IFA_LINK)
1762 if (ifp->scope == IFA_LINK &&
1763 !(ifp->flags & banned_flags)) {
1772 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1775 struct inet6_dev *idev;
1776 int err = -EADDRNOTAVAIL;
1779 idev = __in6_dev_get(dev);
1781 read_lock_bh(&idev->lock);
1782 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1783 read_unlock_bh(&idev->lock);
1789 static int ipv6_count_addresses(struct inet6_dev *idev)
1792 struct inet6_ifaddr *ifp;
1794 read_lock_bh(&idev->lock);
1795 list_for_each_entry(ifp, &idev->addr_list, if_list)
1797 read_unlock_bh(&idev->lock);
1801 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1802 const struct net_device *dev, int strict)
1804 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1806 EXPORT_SYMBOL(ipv6_chk_addr);
1808 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1809 const struct net_device *dev, int strict,
1812 struct inet6_ifaddr *ifp;
1813 unsigned int hash = inet6_addr_hash(addr);
1817 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1818 if (!net_eq(dev_net(ifp->idev->dev), net))
1820 /* Decouple optimistic from tentative for evaluation here.
1821 * Ban optimistic addresses explicitly, when required.
1823 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1824 ? (ifp->flags&~IFA_F_TENTATIVE)
1826 if (ipv6_addr_equal(&ifp->addr, addr) &&
1827 !(ifp_flags&banned_flags) &&
1828 (!dev || ifp->idev->dev == dev ||
1829 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1830 rcu_read_unlock_bh();
1835 rcu_read_unlock_bh();
1838 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1840 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1841 struct net_device *dev)
1843 unsigned int hash = inet6_addr_hash(addr);
1844 struct inet6_ifaddr *ifp;
1846 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1847 if (!net_eq(dev_net(ifp->idev->dev), net))
1849 if (ipv6_addr_equal(&ifp->addr, addr)) {
1850 if (!dev || ifp->idev->dev == dev)
1857 /* Compares an address/prefix_len with addresses on device @dev.
1858 * If one is found it returns true.
1860 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1861 const unsigned int prefix_len, struct net_device *dev)
1863 struct inet6_dev *idev;
1864 struct inet6_ifaddr *ifa;
1868 idev = __in6_dev_get(dev);
1870 read_lock_bh(&idev->lock);
1871 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1872 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1876 read_unlock_bh(&idev->lock);
1882 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1884 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1886 struct inet6_dev *idev;
1887 struct inet6_ifaddr *ifa;
1892 idev = __in6_dev_get(dev);
1894 read_lock_bh(&idev->lock);
1895 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1896 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1901 read_unlock_bh(&idev->lock);
1906 EXPORT_SYMBOL(ipv6_chk_prefix);
1908 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1909 struct net_device *dev, int strict)
1911 struct inet6_ifaddr *ifp, *result = NULL;
1912 unsigned int hash = inet6_addr_hash(addr);
1915 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1916 if (!net_eq(dev_net(ifp->idev->dev), net))
1918 if (ipv6_addr_equal(&ifp->addr, addr)) {
1919 if (!dev || ifp->idev->dev == dev ||
1920 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1927 rcu_read_unlock_bh();
1932 /* Gets referenced address, destroys ifaddr */
1934 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1937 ifp->flags |= IFA_F_DADFAILED;
1939 if (ifp->flags&IFA_F_TEMPORARY) {
1940 struct inet6_ifaddr *ifpub;
1941 spin_lock_bh(&ifp->lock);
1944 in6_ifa_hold(ifpub);
1945 spin_unlock_bh(&ifp->lock);
1946 ipv6_create_tempaddr(ifpub, ifp);
1949 spin_unlock_bh(&ifp->lock);
1952 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1953 spin_lock_bh(&ifp->lock);
1954 addrconf_del_dad_work(ifp);
1955 ifp->flags |= IFA_F_TENTATIVE;
1956 spin_unlock_bh(&ifp->lock);
1958 ipv6_ifa_notify(0, ifp);
1965 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1969 spin_lock_bh(&ifp->lock);
1970 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1971 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1974 spin_unlock_bh(&ifp->lock);
1979 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1981 struct inet6_dev *idev = ifp->idev;
1982 struct net *net = dev_net(ifp->idev->dev);
1984 if (addrconf_dad_end(ifp)) {
1989 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1990 ifp->idev->dev->name, &ifp->addr);
1992 spin_lock_bh(&ifp->lock);
1994 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1995 int scope = ifp->scope;
1996 u32 flags = ifp->flags;
1997 struct in6_addr new_addr;
1998 struct inet6_ifaddr *ifp2;
1999 u32 valid_lft, preferred_lft;
2000 int pfxlen = ifp->prefix_len;
2001 int retries = ifp->stable_privacy_retry + 1;
2003 if (retries > net->ipv6.sysctl.idgen_retries) {
2004 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2005 ifp->idev->dev->name);
2009 new_addr = ifp->addr;
2010 if (ipv6_generate_stable_address(&new_addr, retries,
2014 valid_lft = ifp->valid_lft;
2015 preferred_lft = ifp->prefered_lft;
2017 spin_unlock_bh(&ifp->lock);
2019 if (idev->cnf.max_addresses &&
2020 ipv6_count_addresses(idev) >=
2021 idev->cnf.max_addresses)
2024 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2025 ifp->idev->dev->name);
2027 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2028 scope, flags, valid_lft,
2033 spin_lock_bh(&ifp2->lock);
2034 ifp2->stable_privacy_retry = retries;
2035 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2036 spin_unlock_bh(&ifp2->lock);
2038 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2041 spin_lock_bh(&ifp->lock);
2045 /* transition from _POSTDAD to _ERRDAD */
2046 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2047 spin_unlock_bh(&ifp->lock);
2049 addrconf_mod_dad_work(ifp, 0);
2053 /* Join to solicited addr multicast group.
2054 * caller must hold RTNL */
2055 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2057 struct in6_addr maddr;
2059 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2062 addrconf_addr_solict_mult(addr, &maddr);
2063 ipv6_dev_mc_inc(dev, &maddr);
2066 /* caller must hold RTNL */
2067 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2069 struct in6_addr maddr;
2071 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2074 addrconf_addr_solict_mult(addr, &maddr);
2075 __ipv6_dev_mc_dec(idev, &maddr);
2078 /* caller must hold RTNL */
2079 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2081 struct in6_addr addr;
2083 if (ifp->prefix_len >= 127) /* RFC 6164 */
2085 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2086 if (ipv6_addr_any(&addr))
2088 __ipv6_dev_ac_inc(ifp->idev, &addr);
2091 /* caller must hold RTNL */
2092 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2094 struct in6_addr addr;
2096 if (ifp->prefix_len >= 127) /* RFC 6164 */
2098 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2099 if (ipv6_addr_any(&addr))
2101 __ipv6_dev_ac_dec(ifp->idev, &addr);
2104 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2106 switch (dev->addr_len) {
2108 memcpy(eui, dev->dev_addr, 3);
2111 memcpy(eui + 5, dev->dev_addr + 3, 3);
2113 case EUI64_ADDR_LEN:
2114 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2124 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2126 union fwnet_hwaddr *ha;
2128 if (dev->addr_len != FWNET_ALEN)
2131 ha = (union fwnet_hwaddr *)dev->dev_addr;
2133 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2138 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2140 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2141 if (dev->addr_len != ARCNET_ALEN)
2144 eui[7] = *(u8 *)dev->dev_addr;
2148 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2150 if (dev->addr_len != INFINIBAND_ALEN)
2152 memcpy(eui, dev->dev_addr + 12, 8);
2157 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2161 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2162 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2163 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2164 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2165 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2166 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2170 memcpy(eui + 4, &addr, 4);
2174 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2176 if (dev->priv_flags & IFF_ISATAP)
2177 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2181 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2183 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2186 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2188 memcpy(eui, dev->perm_addr, 3);
2189 memcpy(eui + 5, dev->perm_addr + 3, 3);
2196 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2198 switch (dev->type) {
2201 return addrconf_ifid_eui48(eui, dev);
2203 return addrconf_ifid_arcnet(eui, dev);
2204 case ARPHRD_INFINIBAND:
2205 return addrconf_ifid_infiniband(eui, dev);
2207 return addrconf_ifid_sit(eui, dev);
2210 return addrconf_ifid_gre(eui, dev);
2211 case ARPHRD_6LOWPAN:
2212 return addrconf_ifid_6lowpan(eui, dev);
2213 case ARPHRD_IEEE1394:
2214 return addrconf_ifid_ieee1394(eui, dev);
2215 case ARPHRD_TUNNEL6:
2217 return addrconf_ifid_ip6tnl(eui, dev);
2222 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2225 struct inet6_ifaddr *ifp;
2227 read_lock_bh(&idev->lock);
2228 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2229 if (ifp->scope > IFA_LINK)
2231 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2232 memcpy(eui, ifp->addr.s6_addr+8, 8);
2237 read_unlock_bh(&idev->lock);
2241 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2242 static void ipv6_regen_rndid(struct inet6_dev *idev)
2245 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2246 idev->rndid[0] &= ~0x02;
2249 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2250 * check if generated address is not inappropriate
2252 * - Reserved subnet anycast (RFC 2526)
2253 * 11111101 11....11 1xxxxxxx
2254 * - ISATAP (RFC4214) 6.1
2255 * 00-00-5E-FE-xx-xx-xx-xx
2257 * - XXX: already assigned to an address on the device
2259 if (idev->rndid[0] == 0xfd &&
2260 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2261 (idev->rndid[7]&0x80))
2263 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2264 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2266 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2271 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2273 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2274 ipv6_regen_rndid(idev);
2282 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2283 unsigned long expires, u32 flags)
2285 struct fib6_config cfg = {
2286 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2287 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2288 .fc_ifindex = dev->ifindex,
2289 .fc_expires = expires,
2291 .fc_flags = RTF_UP | flags,
2292 .fc_nlinfo.nl_net = dev_net(dev),
2293 .fc_protocol = RTPROT_KERNEL,
2298 /* Prevent useless cloning on PtP SIT.
2299 This thing is done here expecting that the whole
2300 class of non-broadcast devices need not cloning.
2302 #if IS_ENABLED(CONFIG_IPV6_SIT)
2303 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2304 cfg.fc_flags |= RTF_NONEXTHOP;
2307 ip6_route_add(&cfg, NULL);
2311 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2313 const struct net_device *dev,
2314 u32 flags, u32 noflags)
2316 struct fib6_node *fn;
2317 struct rt6_info *rt = NULL;
2318 struct fib6_table *table;
2319 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2321 table = fib6_get_table(dev_net(dev), tb_id);
2325 read_lock_bh(&table->tb6_lock);
2326 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2330 noflags |= RTF_CACHE;
2331 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2332 if (rt->dst.dev->ifindex != dev->ifindex)
2334 if ((rt->rt6i_flags & flags) != flags)
2336 if ((rt->rt6i_flags & noflags) != 0)
2342 read_unlock_bh(&table->tb6_lock);
2347 /* Create "default" multicast route to the interface */
2349 static void addrconf_add_mroute(struct net_device *dev)
2351 struct fib6_config cfg = {
2352 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2353 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2354 .fc_ifindex = dev->ifindex,
2357 .fc_nlinfo.nl_net = dev_net(dev),
2358 .fc_protocol = RTPROT_KERNEL,
2361 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2363 ip6_route_add(&cfg, NULL);
2366 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2368 struct inet6_dev *idev;
2372 idev = ipv6_find_idev(dev);
2374 return ERR_PTR(-ENOBUFS);
2376 if (idev->cnf.disable_ipv6)
2377 return ERR_PTR(-EACCES);
2379 /* Add default multicast route */
2380 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2381 addrconf_add_mroute(dev);
2386 static void manage_tempaddrs(struct inet6_dev *idev,
2387 struct inet6_ifaddr *ifp,
2388 __u32 valid_lft, __u32 prefered_lft,
2389 bool create, unsigned long now)
2392 struct inet6_ifaddr *ift;
2394 read_lock_bh(&idev->lock);
2395 /* update all temporary addresses in the list */
2396 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2397 int age, max_valid, max_prefered;
2399 if (ifp != ift->ifpub)
2402 /* RFC 4941 section 3.3:
2403 * If a received option will extend the lifetime of a public
2404 * address, the lifetimes of temporary addresses should
2405 * be extended, subject to the overall constraint that no
2406 * temporary addresses should ever remain "valid" or "preferred"
2407 * for a time longer than (TEMP_VALID_LIFETIME) or
2408 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2410 age = (now - ift->cstamp) / HZ;
2411 max_valid = idev->cnf.temp_valid_lft - age;
2415 max_prefered = idev->cnf.temp_prefered_lft -
2416 idev->desync_factor - age;
2417 if (max_prefered < 0)
2420 if (valid_lft > max_valid)
2421 valid_lft = max_valid;
2423 if (prefered_lft > max_prefered)
2424 prefered_lft = max_prefered;
2426 spin_lock(&ift->lock);
2428 ift->valid_lft = valid_lft;
2429 ift->prefered_lft = prefered_lft;
2431 if (prefered_lft > 0)
2432 ift->flags &= ~IFA_F_DEPRECATED;
2434 spin_unlock(&ift->lock);
2435 if (!(flags&IFA_F_TENTATIVE))
2436 ipv6_ifa_notify(0, ift);
2439 /* Also create a temporary address if it's enabled but no temporary
2440 * address currently exists.
2441 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2442 * as part of cleanup (ie. deleting the mngtmpaddr).
2443 * We don't want that to result in creating a new temporary ip address.
2445 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2448 if (create && idev->cnf.use_tempaddr > 0) {
2449 /* When a new public address is created as described
2450 * in [ADDRCONF], also create a new temporary address.
2452 read_unlock_bh(&idev->lock);
2453 ipv6_create_tempaddr(ifp, NULL);
2455 read_unlock_bh(&idev->lock);
2459 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2461 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2462 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2465 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2466 const struct prefix_info *pinfo,
2467 struct inet6_dev *in6_dev,
2468 const struct in6_addr *addr, int addr_type,
2469 u32 addr_flags, bool sllao, bool tokenized,
2470 __u32 valid_lft, u32 prefered_lft)
2472 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2473 int create = 0, update_lft = 0;
2475 if (!ifp && valid_lft) {
2476 int max_addresses = in6_dev->cnf.max_addresses;
2478 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2479 if ((net->ipv6.devconf_all->optimistic_dad ||
2480 in6_dev->cnf.optimistic_dad) &&
2481 !net->ipv6.devconf_all->forwarding && sllao)
2482 addr_flags |= IFA_F_OPTIMISTIC;
2485 /* Do not allow to create too much of autoconfigured
2486 * addresses; this would be too easy way to crash kernel.
2488 if (!max_addresses ||
2489 ipv6_count_addresses(in6_dev) < max_addresses)
2490 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2492 addr_type&IPV6_ADDR_SCOPE_MASK,
2493 addr_flags, valid_lft,
2496 if (IS_ERR_OR_NULL(ifp))
2501 spin_lock_bh(&ifp->lock);
2502 ifp->flags |= IFA_F_MANAGETEMPADDR;
2503 ifp->cstamp = jiffies;
2504 ifp->tokenized = tokenized;
2505 spin_unlock_bh(&ifp->lock);
2506 addrconf_dad_start(ifp);
2514 /* update lifetime (RFC2462 5.5.3 e) */
2515 spin_lock_bh(&ifp->lock);
2517 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2518 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2521 if (!update_lft && !create && stored_lft) {
2522 const u32 minimum_lft = min_t(u32,
2523 stored_lft, MIN_VALID_LIFETIME);
2524 valid_lft = max(valid_lft, minimum_lft);
2526 /* RFC4862 Section 5.5.3e:
2527 * "Note that the preferred lifetime of the
2528 * corresponding address is always reset to
2529 * the Preferred Lifetime in the received
2530 * Prefix Information option, regardless of
2531 * whether the valid lifetime is also reset or
2534 * So we should always update prefered_lft here.
2540 ifp->valid_lft = valid_lft;
2541 ifp->prefered_lft = prefered_lft;
2544 ifp->flags &= ~IFA_F_DEPRECATED;
2545 spin_unlock_bh(&ifp->lock);
2547 if (!(flags&IFA_F_TENTATIVE))
2548 ipv6_ifa_notify(0, ifp);
2550 spin_unlock_bh(&ifp->lock);
2552 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2561 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2563 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2565 struct prefix_info *pinfo;
2570 struct inet6_dev *in6_dev;
2571 struct net *net = dev_net(dev);
2573 pinfo = (struct prefix_info *) opt;
2575 if (len < sizeof(struct prefix_info)) {
2576 ADBG("addrconf: prefix option too short\n");
2581 * Validation checks ([ADDRCONF], page 19)
2584 addr_type = ipv6_addr_type(&pinfo->prefix);
2586 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2589 valid_lft = ntohl(pinfo->valid);
2590 prefered_lft = ntohl(pinfo->prefered);
2592 if (prefered_lft > valid_lft) {
2593 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2597 in6_dev = in6_dev_get(dev);
2600 net_dbg_ratelimited("addrconf: device %s not configured\n",
2606 * Two things going on here:
2607 * 1) Add routes for on-link prefixes
2608 * 2) Configure prefixes with the auto flag set
2611 if (pinfo->onlink) {
2612 struct rt6_info *rt;
2613 unsigned long rt_expires;
2615 /* Avoid arithmetic overflow. Really, we could
2616 * save rt_expires in seconds, likely valid_lft,
2617 * but it would require division in fib gc, that it
2621 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2623 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2625 if (addrconf_finite_timeout(rt_expires))
2628 rt = addrconf_get_prefix_route(&pinfo->prefix,
2631 RTF_ADDRCONF | RTF_PREFIX_RT,
2632 RTF_GATEWAY | RTF_DEFAULT);
2635 /* Autoconf prefix route */
2636 if (valid_lft == 0) {
2639 } else if (addrconf_finite_timeout(rt_expires)) {
2641 rt6_set_expires(rt, jiffies + rt_expires);
2643 rt6_clean_expires(rt);
2645 } else if (valid_lft) {
2646 clock_t expires = 0;
2647 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2648 if (addrconf_finite_timeout(rt_expires)) {
2650 flags |= RTF_EXPIRES;
2651 expires = jiffies_to_clock_t(rt_expires);
2653 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2654 dev, expires, flags);
2659 /* Try to figure out our local address for this prefix */
2661 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2662 struct in6_addr addr;
2663 bool tokenized = false, dev_addr_generated = false;
2665 if (pinfo->prefix_len == 64) {
2666 memcpy(&addr, &pinfo->prefix, 8);
2668 if (!ipv6_addr_any(&in6_dev->token)) {
2669 read_lock_bh(&in6_dev->lock);
2670 memcpy(addr.s6_addr + 8,
2671 in6_dev->token.s6_addr + 8, 8);
2672 read_unlock_bh(&in6_dev->lock);
2674 } else if (is_addr_mode_generate_stable(in6_dev) &&
2675 !ipv6_generate_stable_address(&addr, 0,
2677 addr_flags |= IFA_F_STABLE_PRIVACY;
2679 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2680 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2683 dev_addr_generated = true;
2687 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2692 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2695 tokenized, valid_lft,
2700 /* Ignore error case here because previous prefix add addr was
2701 * successful which will be notified.
2703 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2704 addr_type, addr_flags, sllao,
2705 tokenized, valid_lft,
2707 dev_addr_generated);
2709 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2711 in6_dev_put(in6_dev);
2715 * Set destination address.
2716 * Special case for SIT interfaces where we create a new "virtual"
2719 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2721 struct in6_ifreq ireq;
2722 struct net_device *dev;
2728 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2731 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2737 #if IS_ENABLED(CONFIG_IPV6_SIT)
2738 if (dev->type == ARPHRD_SIT) {
2739 const struct net_device_ops *ops = dev->netdev_ops;
2741 struct ip_tunnel_parm p;
2743 err = -EADDRNOTAVAIL;
2744 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2747 memset(&p, 0, sizeof(p));
2748 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2752 p.iph.protocol = IPPROTO_IPV6;
2754 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2756 if (ops->ndo_do_ioctl) {
2757 mm_segment_t oldfs = get_fs();
2760 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2767 dev = __dev_get_by_name(net, p.name);
2770 err = dev_open(dev);
2780 static int ipv6_mc_config(struct sock *sk, bool join,
2781 const struct in6_addr *addr, int ifindex)
2789 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2791 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2798 * Manual configuration of address on an interface
2800 static int inet6_addr_add(struct net *net, int ifindex,
2801 const struct in6_addr *pfx,
2802 const struct in6_addr *peer_pfx,
2803 unsigned int plen, __u32 ifa_flags,
2804 __u32 prefered_lft, __u32 valid_lft)
2806 struct inet6_ifaddr *ifp;
2807 struct inet6_dev *idev;
2808 struct net_device *dev;
2809 unsigned long timeout;
2819 /* check the lifetime */
2820 if (!valid_lft || prefered_lft > valid_lft)
2823 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2826 dev = __dev_get_by_index(net, ifindex);
2830 idev = addrconf_add_dev(dev);
2832 return PTR_ERR(idev);
2834 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2835 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2836 true, pfx, ifindex);
2842 scope = ipv6_addr_scope(pfx);
2844 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2845 if (addrconf_finite_timeout(timeout)) {
2846 expires = jiffies_to_clock_t(timeout * HZ);
2847 valid_lft = timeout;
2848 flags = RTF_EXPIRES;
2852 ifa_flags |= IFA_F_PERMANENT;
2855 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2856 if (addrconf_finite_timeout(timeout)) {
2858 ifa_flags |= IFA_F_DEPRECATED;
2859 prefered_lft = timeout;
2862 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2863 valid_lft, prefered_lft);
2866 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2867 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2872 * Note that section 3.1 of RFC 4429 indicates
2873 * that the Optimistic flag should not be set for
2874 * manually configured addresses
2876 addrconf_dad_start(ifp);
2877 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2878 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2881 addrconf_verify_rtnl();
2883 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2884 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2885 false, pfx, ifindex);
2888 return PTR_ERR(ifp);
2891 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2892 const struct in6_addr *pfx, unsigned int plen)
2894 struct inet6_ifaddr *ifp;
2895 struct inet6_dev *idev;
2896 struct net_device *dev;
2901 dev = __dev_get_by_index(net, ifindex);
2905 idev = __in6_dev_get(dev);
2909 read_lock_bh(&idev->lock);
2910 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2911 if (ifp->prefix_len == plen &&
2912 ipv6_addr_equal(pfx, &ifp->addr)) {
2914 read_unlock_bh(&idev->lock);
2916 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2917 (ifa_flags & IFA_F_MANAGETEMPADDR))
2918 manage_tempaddrs(idev, ifp, 0, 0, false,
2921 addrconf_verify_rtnl();
2922 if (ipv6_addr_is_multicast(pfx)) {
2923 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2924 false, pfx, dev->ifindex);
2929 read_unlock_bh(&idev->lock);
2930 return -EADDRNOTAVAIL;
2934 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2936 struct in6_ifreq ireq;
2939 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2942 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2946 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2947 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2948 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2953 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2955 struct in6_ifreq ireq;
2958 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2961 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2965 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2966 ireq.ifr6_prefixlen);
2971 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2972 int plen, int scope)
2974 struct inet6_ifaddr *ifp;
2976 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2977 scope, IFA_F_PERMANENT,
2978 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2980 spin_lock_bh(&ifp->lock);
2981 ifp->flags &= ~IFA_F_TENTATIVE;
2982 spin_unlock_bh(&ifp->lock);
2983 rt_genid_bump_ipv6(dev_net(idev->dev));
2984 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2989 #if IS_ENABLED(CONFIG_IPV6_SIT)
2990 static void sit_add_v4_addrs(struct inet6_dev *idev)
2992 struct in6_addr addr;
2993 struct net_device *dev;
2994 struct net *net = dev_net(idev->dev);
3000 memset(&addr, 0, sizeof(struct in6_addr));
3001 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3003 if (idev->dev->flags&IFF_POINTOPOINT) {
3004 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3007 addr.s6_addr32[0] = htonl(0xfe800000);
3011 scope = IPV6_ADDR_COMPATv4;
3013 pflags |= RTF_NONEXTHOP;
3016 if (addr.s6_addr32[3]) {
3017 add_addr(idev, &addr, plen, scope);
3018 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3022 for_each_netdev(net, dev) {
3023 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3024 if (in_dev && (dev->flags & IFF_UP)) {
3025 struct in_ifaddr *ifa;
3029 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3031 addr.s6_addr32[3] = ifa->ifa_local;
3033 if (ifa->ifa_scope == RT_SCOPE_LINK)
3035 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3036 if (idev->dev->flags&IFF_POINTOPOINT)
3041 add_addr(idev, &addr, plen, flag);
3042 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3050 static void init_loopback(struct net_device *dev)
3052 struct inet6_dev *idev;
3058 idev = ipv6_find_idev(dev);
3060 pr_debug("%s: add_dev failed\n", __func__);
3064 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3067 void addrconf_add_linklocal(struct inet6_dev *idev,
3068 const struct in6_addr *addr, u32 flags)
3070 struct inet6_ifaddr *ifp;
3071 u32 addr_flags = flags | IFA_F_PERMANENT;
3073 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3074 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3075 idev->cnf.optimistic_dad) &&
3076 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3077 addr_flags |= IFA_F_OPTIMISTIC;
3080 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3081 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3083 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3084 addrconf_dad_start(ifp);
3088 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3090 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3092 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3095 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3096 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3099 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3100 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3106 static int ipv6_generate_stable_address(struct in6_addr *address,
3108 const struct inet6_dev *idev)
3110 static DEFINE_SPINLOCK(lock);
3111 static __u32 digest[SHA_DIGEST_WORDS];
3112 static __u32 workspace[SHA_WORKSPACE_WORDS];
3115 char __data[SHA_MESSAGE_BYTES];
3117 struct in6_addr secret;
3119 unsigned char hwaddr[MAX_ADDR_LEN];
3124 struct in6_addr secret;
3125 struct in6_addr temp;
3126 struct net *net = dev_net(idev->dev);
3128 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3130 if (idev->cnf.stable_secret.initialized)
3131 secret = idev->cnf.stable_secret.secret;
3132 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3133 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3138 spin_lock_bh(&lock);
3141 memset(&data, 0, sizeof(data));
3142 memset(workspace, 0, sizeof(workspace));
3143 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3144 data.prefix[0] = address->s6_addr32[0];
3145 data.prefix[1] = address->s6_addr32[1];
3146 data.secret = secret;
3147 data.dad_count = dad_count;
3149 sha_transform(digest, data.__data, workspace);
3152 temp.s6_addr32[2] = (__force __be32)digest[0];
3153 temp.s6_addr32[3] = (__force __be32)digest[1];
3155 spin_unlock_bh(&lock);
3157 if (ipv6_reserved_interfaceid(temp)) {
3159 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3168 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3170 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3174 s = &idev->cnf.stable_secret;
3175 get_random_bytes(&s->secret, sizeof(s->secret));
3176 s->initialized = true;
3179 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3181 struct in6_addr addr;
3183 /* no link local addresses on L3 master devices */
3184 if (netif_is_l3_master(idev->dev))
3187 /* no link local addresses on devices flagged as slaves */
3188 if (idev->dev->flags & IFF_SLAVE)
3191 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3193 switch (idev->cnf.addr_gen_mode) {
3194 case IN6_ADDR_GEN_MODE_RANDOM:
3195 ipv6_gen_mode_random_init(idev);
3197 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3198 if (!ipv6_generate_stable_address(&addr, 0, idev))
3199 addrconf_add_linklocal(idev, &addr,
3200 IFA_F_STABLE_PRIVACY);
3201 else if (prefix_route)
3202 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3204 case IN6_ADDR_GEN_MODE_EUI64:
3205 /* addrconf_add_linklocal also adds a prefix_route and we
3206 * only need to care about prefix routes if ipv6_generate_eui64
3207 * couldn't generate one.
3209 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3210 addrconf_add_linklocal(idev, &addr, 0);
3211 else if (prefix_route)
3212 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3214 case IN6_ADDR_GEN_MODE_NONE:
3216 /* will not add any link local address */
3221 static void addrconf_dev_config(struct net_device *dev)
3223 struct inet6_dev *idev;
3227 if ((dev->type != ARPHRD_ETHER) &&
3228 (dev->type != ARPHRD_FDDI) &&
3229 (dev->type != ARPHRD_ARCNET) &&
3230 (dev->type != ARPHRD_INFINIBAND) &&
3231 (dev->type != ARPHRD_IEEE1394) &&
3232 (dev->type != ARPHRD_TUNNEL6) &&
3233 (dev->type != ARPHRD_6LOWPAN) &&
3234 (dev->type != ARPHRD_IP6GRE) &&
3235 (dev->type != ARPHRD_IPGRE) &&
3236 (dev->type != ARPHRD_TUNNEL) &&
3237 (dev->type != ARPHRD_NONE)) {
3238 /* Alas, we support only Ethernet autoconfiguration. */
3239 idev = __in6_dev_get(dev);
3240 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3241 dev->flags & IFF_MULTICAST)
3246 idev = addrconf_add_dev(dev);
3250 /* this device type has no EUI support */
3251 if (dev->type == ARPHRD_NONE &&
3252 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3253 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3255 addrconf_addr_gen(idev, false);
3258 #if IS_ENABLED(CONFIG_IPV6_SIT)
3259 static void addrconf_sit_config(struct net_device *dev)
3261 struct inet6_dev *idev;
3266 * Configure the tunnel with one of our IPv4
3267 * addresses... we should configure all of
3268 * our v4 addrs in the tunnel
3271 idev = ipv6_find_idev(dev);
3273 pr_debug("%s: add_dev failed\n", __func__);
3277 if (dev->priv_flags & IFF_ISATAP) {
3278 addrconf_addr_gen(idev, false);
3282 sit_add_v4_addrs(idev);
3284 if (dev->flags&IFF_POINTOPOINT)
3285 addrconf_add_mroute(dev);
3289 #if IS_ENABLED(CONFIG_NET_IPGRE)
3290 static void addrconf_gre_config(struct net_device *dev)
3292 struct inet6_dev *idev;
3296 idev = ipv6_find_idev(dev);
3298 pr_debug("%s: add_dev failed\n", __func__);
3302 addrconf_addr_gen(idev, true);
3303 if (dev->flags & IFF_POINTOPOINT)
3304 addrconf_add_mroute(dev);
3308 static int fixup_permanent_addr(struct inet6_dev *idev,
3309 struct inet6_ifaddr *ifp)
3311 /* !rt6i_node means the host route was removed from the
3312 * FIB, for example, if 'lo' device is taken down. In that
3313 * case regenerate the host route.
3315 if (!ifp->rt || !ifp->rt->rt6i_node) {
3316 struct rt6_info *rt, *prev;
3318 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3319 if (unlikely(IS_ERR(rt)))
3322 /* ifp->rt can be accessed outside of rtnl */
3323 spin_lock(&ifp->lock);
3326 spin_unlock(&ifp->lock);
3331 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3332 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3336 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3337 addrconf_dad_start(ifp);
3342 static void addrconf_permanent_addr(struct net_device *dev)
3344 struct inet6_ifaddr *ifp, *tmp;
3345 struct inet6_dev *idev;
3347 idev = __in6_dev_get(dev);
3351 write_lock_bh(&idev->lock);
3353 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3354 if ((ifp->flags & IFA_F_PERMANENT) &&
3355 fixup_permanent_addr(idev, ifp) < 0) {
3356 write_unlock_bh(&idev->lock);
3359 write_lock_bh(&idev->lock);
3361 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3362 idev->dev->name, &ifp->addr);
3366 write_unlock_bh(&idev->lock);
3369 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3372 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3373 struct netdev_notifier_changeupper_info *info;
3374 struct inet6_dev *idev = __in6_dev_get(dev);
3375 struct net *net = dev_net(dev);
3376 int run_pending = 0;
3380 case NETDEV_REGISTER:
3381 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3382 idev = ipv6_add_dev(dev);
3384 return notifier_from_errno(PTR_ERR(idev));
3388 case NETDEV_CHANGEMTU:
3389 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3390 if (dev->mtu < IPV6_MIN_MTU) {
3391 addrconf_ifdown(dev, dev != net->loopback_dev);
3396 rt6_mtu_change(dev, dev->mtu);
3397 idev->cnf.mtu6 = dev->mtu;
3401 /* allocate new idev */
3402 idev = ipv6_add_dev(dev);
3406 /* device is still not ready */
3407 if (!(idev->if_flags & IF_READY))
3416 if (dev->flags & IFF_SLAVE)
3419 if (idev && idev->cnf.disable_ipv6)
3422 if (event == NETDEV_UP) {
3423 /* restore routes for permanent addresses */
3424 addrconf_permanent_addr(dev);
3426 if (!addrconf_link_ready(dev)) {
3427 /* device is not ready yet. */
3428 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3433 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3434 idev = ipv6_add_dev(dev);
3436 if (!IS_ERR_OR_NULL(idev)) {
3437 idev->if_flags |= IF_READY;
3440 } else if (event == NETDEV_CHANGE) {
3441 if (!addrconf_link_ready(dev)) {
3442 /* device is still not ready. */
3447 if (idev->if_flags & IF_READY) {
3448 /* device is already configured -
3449 * but resend MLD reports, we might
3450 * have roamed and need to update
3451 * multicast snooping switches
3456 idev->if_flags |= IF_READY;
3459 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3465 switch (dev->type) {
3466 #if IS_ENABLED(CONFIG_IPV6_SIT)
3468 addrconf_sit_config(dev);
3471 #if IS_ENABLED(CONFIG_NET_IPGRE)
3473 addrconf_gre_config(dev);
3476 case ARPHRD_LOOPBACK:
3481 addrconf_dev_config(dev);
3485 if (!IS_ERR_OR_NULL(idev)) {
3487 addrconf_dad_run(idev);
3490 * If the MTU changed during the interface down,
3491 * when the interface up, the changed MTU must be
3492 * reflected in the idev as well as routers.
3494 if (idev->cnf.mtu6 != dev->mtu &&
3495 dev->mtu >= IPV6_MIN_MTU) {
3496 rt6_mtu_change(dev, dev->mtu);
3497 idev->cnf.mtu6 = dev->mtu;
3499 idev->tstamp = jiffies;
3500 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3503 * If the changed mtu during down is lower than
3504 * IPV6_MIN_MTU stop IPv6 on this interface.
3506 if (dev->mtu < IPV6_MIN_MTU)
3507 addrconf_ifdown(dev, dev != net->loopback_dev);
3512 case NETDEV_UNREGISTER:
3514 * Remove all addresses from this interface.
3516 addrconf_ifdown(dev, event != NETDEV_DOWN);
3519 case NETDEV_CHANGENAME:
3521 snmp6_unregister_dev(idev);
3522 addrconf_sysctl_unregister(idev);
3523 err = addrconf_sysctl_register(idev);
3525 return notifier_from_errno(err);
3526 err = snmp6_register_dev(idev);
3528 addrconf_sysctl_unregister(idev);
3529 return notifier_from_errno(err);
3534 case NETDEV_PRE_TYPE_CHANGE:
3535 case NETDEV_POST_TYPE_CHANGE:
3537 addrconf_type_change(dev, event);
3540 case NETDEV_CHANGEUPPER:
3543 /* flush all routes if dev is linked to or unlinked from
3544 * an L3 master device (e.g., VRF)
3546 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3547 addrconf_ifdown(dev, 0);
3554 * addrconf module should be notified of a device going up
3556 static struct notifier_block ipv6_dev_notf = {
3557 .notifier_call = addrconf_notify,
3558 .priority = ADDRCONF_NOTIFY_PRIORITY,
3561 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3563 struct inet6_dev *idev;
3566 idev = __in6_dev_get(dev);
3568 if (event == NETDEV_POST_TYPE_CHANGE)
3569 ipv6_mc_remap(idev);
3570 else if (event == NETDEV_PRE_TYPE_CHANGE)
3571 ipv6_mc_unmap(idev);
3574 static bool addr_is_local(const struct in6_addr *addr)
3576 return ipv6_addr_type(addr) &
3577 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3580 static int addrconf_ifdown(struct net_device *dev, int how)
3582 struct net *net = dev_net(dev);
3583 struct inet6_dev *idev;
3584 struct inet6_ifaddr *ifa, *tmp;
3585 struct list_head del_list;
3593 rt6_ifdown(net, dev);
3594 neigh_ifdown(&nd_tbl, dev);
3596 idev = __in6_dev_get(dev);
3601 * Step 1: remove reference to ipv6 device from parent device.
3607 /* protected by rtnl_lock */
3608 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3610 /* Step 1.5: remove snmp6 entry */
3611 snmp6_unregister_dev(idev);
3615 /* aggregate the system setting and interface setting */
3616 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3618 _keep_addr = idev->cnf.keep_addr_on_down;
3620 /* combine the user config with event to determine if permanent
3621 * addresses are to be removed from address hash table
3623 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3625 /* Step 2: clear hash table */
3626 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3627 struct hlist_head *h = &inet6_addr_lst[i];
3629 spin_lock_bh(&addrconf_hash_lock);
3631 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3632 if (ifa->idev == idev) {
3633 addrconf_del_dad_work(ifa);
3634 /* combined flag + permanent flag decide if
3635 * address is retained on a down event
3638 !(ifa->flags & IFA_F_PERMANENT) ||
3639 addr_is_local(&ifa->addr)) {
3640 hlist_del_init_rcu(&ifa->addr_lst);
3645 spin_unlock_bh(&addrconf_hash_lock);
3648 write_lock_bh(&idev->lock);
3650 addrconf_del_rs_timer(idev);
3652 /* Step 2: clear flags for stateless addrconf, repeated down
3655 was_ready = idev->if_flags & IF_READY;
3657 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3659 /* Step 3: clear tempaddr list */
3660 while (!list_empty(&idev->tempaddr_list)) {
3661 ifa = list_first_entry(&idev->tempaddr_list,
3662 struct inet6_ifaddr, tmp_list);
3663 list_del(&ifa->tmp_list);
3664 write_unlock_bh(&idev->lock);
3665 spin_lock_bh(&ifa->lock);
3668 in6_ifa_put(ifa->ifpub);
3671 spin_unlock_bh(&ifa->lock);
3673 write_lock_bh(&idev->lock);
3676 /* re-combine the user config with event to determine if permanent
3677 * addresses are to be removed from the interface list
3679 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3681 INIT_LIST_HEAD(&del_list);
3682 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3683 struct rt6_info *rt = NULL;
3686 addrconf_del_dad_work(ifa);
3688 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3689 !addr_is_local(&ifa->addr);
3691 list_move(&ifa->if_list, &del_list);
3693 write_unlock_bh(&idev->lock);
3694 spin_lock_bh(&ifa->lock);
3697 /* set state to skip the notifier below */
3698 state = INET6_IFADDR_STATE_DEAD;
3699 ifa->state = INET6_IFADDR_STATE_PREDAD;
3700 if (!(ifa->flags & IFA_F_NODAD))
3701 ifa->flags |= IFA_F_TENTATIVE;
3707 ifa->state = INET6_IFADDR_STATE_DEAD;
3710 spin_unlock_bh(&ifa->lock);
3715 if (state != INET6_IFADDR_STATE_DEAD) {
3716 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3717 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3719 if (idev->cnf.forwarding)
3720 addrconf_leave_anycast(ifa);
3721 addrconf_leave_solict(ifa->idev, &ifa->addr);
3724 write_lock_bh(&idev->lock);
3727 write_unlock_bh(&idev->lock);
3729 /* now clean up addresses to be removed */
3730 while (!list_empty(&del_list)) {
3731 ifa = list_first_entry(&del_list,
3732 struct inet6_ifaddr, if_list);
3733 list_del(&ifa->if_list);
3738 /* Step 5: Discard anycast and multicast list */
3740 ipv6_ac_destroy_dev(idev);
3741 ipv6_mc_destroy_dev(idev);
3742 } else if (was_ready) {
3746 idev->tstamp = jiffies;
3748 /* Last: Shot the device (if unregistered) */
3750 addrconf_sysctl_unregister(idev);
3751 neigh_parms_release(&nd_tbl, idev->nd_parms);
3752 neigh_ifdown(&nd_tbl, dev);
3758 static void addrconf_rs_timer(unsigned long data)
3760 struct inet6_dev *idev = (struct inet6_dev *)data;
3761 struct net_device *dev = idev->dev;
3762 struct in6_addr lladdr;
3764 write_lock(&idev->lock);
3765 if (idev->dead || !(idev->if_flags & IF_READY))
3768 if (!ipv6_accept_ra(idev))
3771 /* Announcement received after solicitation was sent */
3772 if (idev->if_flags & IF_RA_RCVD)
3775 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3776 write_unlock(&idev->lock);
3777 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3778 ndisc_send_rs(dev, &lladdr,
3779 &in6addr_linklocal_allrouters);
3783 write_lock(&idev->lock);
3784 idev->rs_interval = rfc3315_s14_backoff_update(
3785 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3786 /* The wait after the last probe can be shorter */
3787 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3788 idev->cnf.rtr_solicits) ?
3789 idev->cnf.rtr_solicit_delay :
3793 * Note: we do not support deprecated "all on-link"
3794 * assumption any longer.
3796 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3800 write_unlock(&idev->lock);
3806 * Duplicate Address Detection
3808 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3810 unsigned long rand_num;
3811 struct inet6_dev *idev = ifp->idev;
3814 if (ifp->flags & IFA_F_OPTIMISTIC)
3817 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3820 if (idev->cnf.enhanced_dad ||
3821 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3823 get_random_bytes(&nonce, 6);
3826 ifp->dad_nonce = nonce;
3827 ifp->dad_probes = idev->cnf.dad_transmits;
3828 addrconf_mod_dad_work(ifp, rand_num);
3831 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3833 struct inet6_dev *idev = ifp->idev;
3834 struct net_device *dev = idev->dev;
3835 bool bump_id, notify = false;
3837 addrconf_join_solict(dev, &ifp->addr);
3839 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3841 read_lock_bh(&idev->lock);
3842 spin_lock(&ifp->lock);
3843 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3846 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3847 (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3848 idev->cnf.accept_dad < 1) ||
3849 !(ifp->flags&IFA_F_TENTATIVE) ||
3850 ifp->flags & IFA_F_NODAD) {
3851 bool send_na = false;
3853 if (ifp->flags & IFA_F_TENTATIVE &&
3854 !(ifp->flags & IFA_F_OPTIMISTIC))
3856 bump_id = ifp->flags & IFA_F_TENTATIVE;
3857 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3858 spin_unlock(&ifp->lock);
3859 read_unlock_bh(&idev->lock);
3861 addrconf_dad_completed(ifp, bump_id, send_na);
3865 if (!(idev->if_flags & IF_READY)) {
3866 spin_unlock(&ifp->lock);
3867 read_unlock_bh(&idev->lock);
3869 * If the device is not ready:
3870 * - keep it tentative if it is a permanent address.
3871 * - otherwise, kill it.
3874 addrconf_dad_stop(ifp, 0);
3879 * Optimistic nodes can start receiving
3882 if (ifp->flags & IFA_F_OPTIMISTIC) {
3883 ip6_ins_rt(ifp->rt);
3884 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3885 /* Because optimistic nodes can use this address,
3886 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3892 addrconf_dad_kick(ifp);
3894 spin_unlock(&ifp->lock);
3895 read_unlock_bh(&idev->lock);
3897 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3900 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3902 bool begin_dad = false;
3904 spin_lock_bh(&ifp->lock);
3905 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3906 ifp->state = INET6_IFADDR_STATE_PREDAD;
3909 spin_unlock_bh(&ifp->lock);
3912 addrconf_mod_dad_work(ifp, 0);
3915 static void addrconf_dad_work(struct work_struct *w)
3917 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3918 struct inet6_ifaddr,
3920 struct inet6_dev *idev = ifp->idev;
3921 bool bump_id, disable_ipv6 = false;
3922 struct in6_addr mcaddr;
3928 } action = DAD_PROCESS;
3932 spin_lock_bh(&ifp->lock);
3933 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3935 ifp->state = INET6_IFADDR_STATE_DAD;
3936 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3938 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3940 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3941 idev->cnf.accept_dad > 1) &&
3942 !idev->cnf.disable_ipv6 &&
3943 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3944 struct in6_addr addr;
3946 addr.s6_addr32[0] = htonl(0xfe800000);
3947 addr.s6_addr32[1] = 0;
3949 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3950 ipv6_addr_equal(&ifp->addr, &addr)) {
3951 /* DAD failed for link-local based on MAC */
3952 idev->cnf.disable_ipv6 = 1;
3954 pr_info("%s: IPv6 being disabled!\n",
3955 ifp->idev->dev->name);
3956 disable_ipv6 = true;
3960 spin_unlock_bh(&ifp->lock);
3962 if (action == DAD_BEGIN) {
3963 addrconf_dad_begin(ifp);
3965 } else if (action == DAD_ABORT) {
3967 addrconf_dad_stop(ifp, 1);
3969 addrconf_ifdown(idev->dev, 0);
3973 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3976 write_lock_bh(&idev->lock);
3977 if (idev->dead || !(idev->if_flags & IF_READY)) {
3978 write_unlock_bh(&idev->lock);
3982 spin_lock(&ifp->lock);
3983 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3984 spin_unlock(&ifp->lock);
3985 write_unlock_bh(&idev->lock);
3989 if (ifp->dad_probes == 0) {
3990 bool send_na = false;
3993 * DAD was successful
3996 if (ifp->flags & IFA_F_TENTATIVE &&
3997 !(ifp->flags & IFA_F_OPTIMISTIC))
3999 bump_id = ifp->flags & IFA_F_TENTATIVE;
4000 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4001 spin_unlock(&ifp->lock);
4002 write_unlock_bh(&idev->lock);
4004 addrconf_dad_completed(ifp, bump_id, send_na);
4010 addrconf_mod_dad_work(ifp,
4011 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4012 spin_unlock(&ifp->lock);
4013 write_unlock_bh(&idev->lock);
4015 /* send a neighbour solicitation for our addr */
4016 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4017 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4024 /* ifp->idev must be at least read locked */
4025 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4027 struct inet6_ifaddr *ifpiter;
4028 struct inet6_dev *idev = ifp->idev;
4030 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4031 if (ifpiter->scope > IFA_LINK)
4033 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4034 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4035 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4042 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4045 struct net_device *dev = ifp->idev->dev;
4046 struct in6_addr lladdr;
4047 bool send_rs, send_mld;
4049 addrconf_del_dad_work(ifp);
4052 * Configure the address for reception. Now it is valid.
4055 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4057 /* If added prefix is link local and we are prepared to process
4058 router advertisements, start sending router solicitations.
4061 read_lock_bh(&ifp->idev->lock);
4062 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4063 send_rs = send_mld &&
4064 ipv6_accept_ra(ifp->idev) &&
4065 ifp->idev->cnf.rtr_solicits != 0 &&
4066 (dev->flags & IFF_LOOPBACK) == 0 &&
4067 (dev->type != ARPHRD_TUNNEL);
4068 read_unlock_bh(&ifp->idev->lock);
4070 /* While dad is in progress mld report's source address is in6_addrany.
4071 * Resend with proper ll now.
4074 ipv6_mc_dad_complete(ifp->idev);
4076 /* send unsolicited NA if enabled */
4078 (ifp->idev->cnf.ndisc_notify ||
4079 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4080 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4081 /*router=*/ !!ifp->idev->cnf.forwarding,
4082 /*solicited=*/ false, /*override=*/ true,
4088 * If a host as already performed a random delay
4089 * [...] as part of DAD [...] there is no need
4090 * to delay again before sending the first RS
4092 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4094 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4096 write_lock_bh(&ifp->idev->lock);
4097 spin_lock(&ifp->lock);
4098 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4099 ifp->idev->cnf.rtr_solicit_interval);
4100 ifp->idev->rs_probes = 1;
4101 ifp->idev->if_flags |= IF_RS_SENT;
4102 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4103 spin_unlock(&ifp->lock);
4104 write_unlock_bh(&ifp->idev->lock);
4108 rt_genid_bump_ipv6(dev_net(dev));
4110 /* Make sure that a new temporary address will be created
4111 * before this temporary address becomes deprecated.
4113 if (ifp->flags & IFA_F_TEMPORARY)
4114 addrconf_verify_rtnl();
4117 static void addrconf_dad_run(struct inet6_dev *idev)
4119 struct inet6_ifaddr *ifp;
4121 read_lock_bh(&idev->lock);
4122 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4123 spin_lock(&ifp->lock);
4124 if (ifp->flags & IFA_F_TENTATIVE &&
4125 ifp->state == INET6_IFADDR_STATE_DAD)
4126 addrconf_dad_kick(ifp);
4127 spin_unlock(&ifp->lock);
4129 read_unlock_bh(&idev->lock);
4132 #ifdef CONFIG_PROC_FS
4133 struct if6_iter_state {
4134 struct seq_net_private p;
4139 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4141 struct inet6_ifaddr *ifa = NULL;
4142 struct if6_iter_state *state = seq->private;
4143 struct net *net = seq_file_net(seq);
4146 /* initial bucket if pos is 0 */
4152 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4153 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4155 if (!net_eq(dev_net(ifa->idev->dev), net))
4157 /* sync with offset */
4158 if (p < state->offset) {
4165 /* prepare for next bucket */
4172 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4173 struct inet6_ifaddr *ifa)
4175 struct if6_iter_state *state = seq->private;
4176 struct net *net = seq_file_net(seq);
4178 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4179 if (!net_eq(dev_net(ifa->idev->dev), net))
4186 while (++state->bucket < IN6_ADDR_HSIZE) {
4187 hlist_for_each_entry_rcu_bh(ifa,
4188 &inet6_addr_lst[state->bucket], addr_lst) {
4189 if (!net_eq(dev_net(ifa->idev->dev), net))
4198 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4202 return if6_get_first(seq, *pos);
4205 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4207 struct inet6_ifaddr *ifa;
4209 ifa = if6_get_next(seq, v);
4214 static void if6_seq_stop(struct seq_file *seq, void *v)
4217 rcu_read_unlock_bh();
4220 static int if6_seq_show(struct seq_file *seq, void *v)
4222 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4223 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4225 ifp->idev->dev->ifindex,
4229 ifp->idev->dev->name);
4233 static const struct seq_operations if6_seq_ops = {
4234 .start = if6_seq_start,
4235 .next = if6_seq_next,
4236 .show = if6_seq_show,
4237 .stop = if6_seq_stop,
4240 static int if6_seq_open(struct inode *inode, struct file *file)
4242 return seq_open_net(inode, file, &if6_seq_ops,
4243 sizeof(struct if6_iter_state));
4246 static const struct file_operations if6_fops = {
4247 .owner = THIS_MODULE,
4248 .open = if6_seq_open,
4250 .llseek = seq_lseek,
4251 .release = seq_release_net,
4254 static int __net_init if6_proc_net_init(struct net *net)
4256 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4261 static void __net_exit if6_proc_net_exit(struct net *net)
4263 remove_proc_entry("if_inet6", net->proc_net);
4266 static struct pernet_operations if6_proc_net_ops = {
4267 .init = if6_proc_net_init,
4268 .exit = if6_proc_net_exit,
4271 int __init if6_proc_init(void)
4273 return register_pernet_subsys(&if6_proc_net_ops);
4276 void if6_proc_exit(void)
4278 unregister_pernet_subsys(&if6_proc_net_ops);
4280 #endif /* CONFIG_PROC_FS */
4282 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4283 /* Check if address is a home address configured on any interface. */
4284 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4287 struct inet6_ifaddr *ifp = NULL;
4288 unsigned int hash = inet6_addr_hash(addr);
4291 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4292 if (!net_eq(dev_net(ifp->idev->dev), net))
4294 if (ipv6_addr_equal(&ifp->addr, addr) &&
4295 (ifp->flags & IFA_F_HOMEADDRESS)) {
4300 rcu_read_unlock_bh();
4306 * Periodic address status verification
4309 static void addrconf_verify_rtnl(void)
4311 unsigned long now, next, next_sec, next_sched;
4312 struct inet6_ifaddr *ifp;
4319 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4321 cancel_delayed_work(&addr_chk_work);
4323 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4325 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4328 /* When setting preferred_lft to a value not zero or
4329 * infinity, while valid_lft is infinity
4330 * IFA_F_PERMANENT has a non-infinity life time.
4332 if ((ifp->flags & IFA_F_PERMANENT) &&
4333 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4336 spin_lock(&ifp->lock);
4337 /* We try to batch several events at once. */
4338 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4340 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4341 age >= ifp->valid_lft) {
4342 spin_unlock(&ifp->lock);
4346 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4347 spin_unlock(&ifp->lock);
4349 } else if (age >= ifp->prefered_lft) {
4350 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4353 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4355 ifp->flags |= IFA_F_DEPRECATED;
4358 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4359 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4360 next = ifp->tstamp + ifp->valid_lft * HZ;
4362 spin_unlock(&ifp->lock);
4367 ipv6_ifa_notify(0, ifp);
4371 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4372 !(ifp->flags&IFA_F_TENTATIVE)) {
4373 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4374 ifp->idev->cnf.dad_transmits *
4375 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4377 if (age >= ifp->prefered_lft - regen_advance) {
4378 struct inet6_ifaddr *ifpub = ifp->ifpub;
4379 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4380 next = ifp->tstamp + ifp->prefered_lft * HZ;
4381 if (!ifp->regen_count && ifpub) {
4384 in6_ifa_hold(ifpub);
4385 spin_unlock(&ifp->lock);
4387 spin_lock(&ifpub->lock);
4388 ifpub->regen_count = 0;
4389 spin_unlock(&ifpub->lock);
4390 ipv6_create_tempaddr(ifpub, ifp);
4395 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4396 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4397 spin_unlock(&ifp->lock);
4399 /* ifp->prefered_lft <= ifp->valid_lft */
4400 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4401 next = ifp->tstamp + ifp->prefered_lft * HZ;
4402 spin_unlock(&ifp->lock);
4407 next_sec = round_jiffies_up(next);
4410 /* If rounded timeout is accurate enough, accept it. */
4411 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4412 next_sched = next_sec;
4414 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4415 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4416 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4418 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4419 now, next, next_sec, next_sched);
4420 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4421 rcu_read_unlock_bh();
4424 static void addrconf_verify_work(struct work_struct *w)
4427 addrconf_verify_rtnl();
4431 static void addrconf_verify(void)
4433 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4436 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4437 struct in6_addr **peer_pfx)
4439 struct in6_addr *pfx = NULL;
4444 pfx = nla_data(addr);
4447 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4449 pfx = nla_data(local);
4455 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4456 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4457 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4458 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4459 [IFA_FLAGS] = { .len = sizeof(u32) },
4463 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4464 struct netlink_ext_ack *extack)
4466 struct net *net = sock_net(skb->sk);
4467 struct ifaddrmsg *ifm;
4468 struct nlattr *tb[IFA_MAX+1];
4469 struct in6_addr *pfx, *peer_pfx;
4473 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4478 ifm = nlmsg_data(nlh);
4479 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4483 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4485 /* We ignore other flags so far. */
4486 ifa_flags &= IFA_F_MANAGETEMPADDR;
4488 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4489 ifm->ifa_prefixlen);
4492 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4493 u32 prefered_lft, u32 valid_lft)
4497 unsigned long timeout;
4498 bool was_managetempaddr;
4499 bool had_prefixroute;
4503 if (!valid_lft || (prefered_lft > valid_lft))
4506 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4507 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4510 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4511 if (addrconf_finite_timeout(timeout)) {
4512 expires = jiffies_to_clock_t(timeout * HZ);
4513 valid_lft = timeout;
4514 flags = RTF_EXPIRES;
4518 ifa_flags |= IFA_F_PERMANENT;
4521 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4522 if (addrconf_finite_timeout(timeout)) {
4524 ifa_flags |= IFA_F_DEPRECATED;
4525 prefered_lft = timeout;
4528 spin_lock_bh(&ifp->lock);
4529 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4530 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4531 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4532 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4533 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4534 IFA_F_NOPREFIXROUTE);
4535 ifp->flags |= ifa_flags;
4536 ifp->tstamp = jiffies;
4537 ifp->valid_lft = valid_lft;
4538 ifp->prefered_lft = prefered_lft;
4540 spin_unlock_bh(&ifp->lock);
4541 if (!(ifp->flags&IFA_F_TENTATIVE))
4542 ipv6_ifa_notify(0, ifp);
4544 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4545 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4547 } else if (had_prefixroute) {
4548 enum cleanup_prefix_rt_t action;
4549 unsigned long rt_expires;
4551 write_lock_bh(&ifp->idev->lock);
4552 action = check_cleanup_prefix_route(ifp, &rt_expires);
4553 write_unlock_bh(&ifp->idev->lock);
4555 if (action != CLEANUP_PREFIX_RT_NOP) {
4556 cleanup_prefix_route(ifp, rt_expires,
4557 action == CLEANUP_PREFIX_RT_DEL);
4561 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4562 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4563 valid_lft = prefered_lft = 0;
4564 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4565 !was_managetempaddr, jiffies);
4568 addrconf_verify_rtnl();
4574 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4575 struct netlink_ext_ack *extack)
4577 struct net *net = sock_net(skb->sk);
4578 struct ifaddrmsg *ifm;
4579 struct nlattr *tb[IFA_MAX+1];
4580 struct in6_addr *pfx, *peer_pfx;
4581 struct inet6_ifaddr *ifa;
4582 struct net_device *dev;
4583 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4587 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4592 ifm = nlmsg_data(nlh);
4593 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4597 if (tb[IFA_CACHEINFO]) {
4598 struct ifa_cacheinfo *ci;
4600 ci = nla_data(tb[IFA_CACHEINFO]);
4601 valid_lft = ci->ifa_valid;
4602 preferred_lft = ci->ifa_prefered;
4604 preferred_lft = INFINITY_LIFE_TIME;
4605 valid_lft = INFINITY_LIFE_TIME;
4608 dev = __dev_get_by_index(net, ifm->ifa_index);
4612 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4614 /* We ignore other flags so far. */
4615 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4616 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4618 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4621 * It would be best to check for !NLM_F_CREATE here but
4622 * userspace already relies on not having to provide this.
4624 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4625 ifm->ifa_prefixlen, ifa_flags,
4626 preferred_lft, valid_lft);
4629 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4630 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4633 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4640 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4641 u8 scope, int ifindex)
4643 struct ifaddrmsg *ifm;
4645 ifm = nlmsg_data(nlh);
4646 ifm->ifa_family = AF_INET6;
4647 ifm->ifa_prefixlen = prefixlen;
4648 ifm->ifa_flags = flags;
4649 ifm->ifa_scope = scope;
4650 ifm->ifa_index = ifindex;
4653 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4654 unsigned long tstamp, u32 preferred, u32 valid)
4656 struct ifa_cacheinfo ci;
4658 ci.cstamp = cstamp_delta(cstamp);
4659 ci.tstamp = cstamp_delta(tstamp);
4660 ci.ifa_prefered = preferred;
4661 ci.ifa_valid = valid;
4663 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4666 static inline int rt_scope(int ifa_scope)
4668 if (ifa_scope & IFA_HOST)
4669 return RT_SCOPE_HOST;
4670 else if (ifa_scope & IFA_LINK)
4671 return RT_SCOPE_LINK;
4672 else if (ifa_scope & IFA_SITE)
4673 return RT_SCOPE_SITE;
4675 return RT_SCOPE_UNIVERSE;
4678 static inline int inet6_ifaddr_msgsize(void)
4680 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4681 + nla_total_size(16) /* IFA_LOCAL */
4682 + nla_total_size(16) /* IFA_ADDRESS */
4683 + nla_total_size(sizeof(struct ifa_cacheinfo))
4684 + nla_total_size(4) /* IFA_FLAGS */;
4687 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4688 u32 portid, u32 seq, int event, unsigned int flags)
4690 struct nlmsghdr *nlh;
4691 u32 preferred, valid;
4693 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4697 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4698 ifa->idev->dev->ifindex);
4700 if (!((ifa->flags&IFA_F_PERMANENT) &&
4701 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4702 preferred = ifa->prefered_lft;
4703 valid = ifa->valid_lft;
4704 if (preferred != INFINITY_LIFE_TIME) {
4705 long tval = (jiffies - ifa->tstamp)/HZ;
4706 if (preferred > tval)
4710 if (valid != INFINITY_LIFE_TIME) {
4718 preferred = INFINITY_LIFE_TIME;
4719 valid = INFINITY_LIFE_TIME;
4722 if (!ipv6_addr_any(&ifa->peer_addr)) {
4723 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4724 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4727 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4730 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4733 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4736 nlmsg_end(skb, nlh);
4740 nlmsg_cancel(skb, nlh);
4744 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4745 u32 portid, u32 seq, int event, u16 flags)
4747 struct nlmsghdr *nlh;
4748 u8 scope = RT_SCOPE_UNIVERSE;
4749 int ifindex = ifmca->idev->dev->ifindex;
4751 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4752 scope = RT_SCOPE_SITE;
4754 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4758 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4759 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4760 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4761 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4762 nlmsg_cancel(skb, nlh);
4766 nlmsg_end(skb, nlh);
4770 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4771 u32 portid, u32 seq, int event, unsigned int flags)
4773 struct nlmsghdr *nlh;
4774 u8 scope = RT_SCOPE_UNIVERSE;
4775 int ifindex = ifaca->aca_idev->dev->ifindex;
4777 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4778 scope = RT_SCOPE_SITE;
4780 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4784 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4785 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4786 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4787 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4788 nlmsg_cancel(skb, nlh);
4792 nlmsg_end(skb, nlh);
4802 /* called with rcu_read_lock() */
4803 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4804 struct netlink_callback *cb, enum addr_type_t type,
4805 int s_ip_idx, int *p_ip_idx)
4807 struct ifmcaddr6 *ifmca;
4808 struct ifacaddr6 *ifaca;
4810 int ip_idx = *p_ip_idx;
4812 read_lock_bh(&idev->lock);
4814 case UNICAST_ADDR: {
4815 struct inet6_ifaddr *ifa;
4817 /* unicast address incl. temp addr */
4818 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4819 if (ip_idx < s_ip_idx)
4821 err = inet6_fill_ifaddr(skb, ifa,
4822 NETLINK_CB(cb->skb).portid,
4828 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4834 case MULTICAST_ADDR:
4835 /* multicast address */
4836 for (ifmca = idev->mc_list; ifmca;
4837 ifmca = ifmca->next, ip_idx++) {
4838 if (ip_idx < s_ip_idx)
4840 err = inet6_fill_ifmcaddr(skb, ifmca,
4841 NETLINK_CB(cb->skb).portid,
4850 /* anycast address */
4851 for (ifaca = idev->ac_list; ifaca;
4852 ifaca = ifaca->aca_next, ip_idx++) {
4853 if (ip_idx < s_ip_idx)
4855 err = inet6_fill_ifacaddr(skb, ifaca,
4856 NETLINK_CB(cb->skb).portid,
4867 read_unlock_bh(&idev->lock);
4872 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4873 enum addr_type_t type)
4875 struct net *net = sock_net(skb->sk);
4878 int s_idx, s_ip_idx;
4879 struct net_device *dev;
4880 struct inet6_dev *idev;
4881 struct hlist_head *head;
4884 s_idx = idx = cb->args[1];
4885 s_ip_idx = ip_idx = cb->args[2];
4888 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4889 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4891 head = &net->dev_index_head[h];
4892 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4895 if (h > s_h || idx > s_idx)
4898 idev = __in6_dev_get(dev);
4902 if (in6_dump_addrs(idev, skb, cb, type,
4903 s_ip_idx, &ip_idx) < 0)
4913 cb->args[2] = ip_idx;
4918 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4920 enum addr_type_t type = UNICAST_ADDR;
4922 return inet6_dump_addr(skb, cb, type);
4925 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4927 enum addr_type_t type = MULTICAST_ADDR;
4929 return inet6_dump_addr(skb, cb, type);
4933 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4935 enum addr_type_t type = ANYCAST_ADDR;
4937 return inet6_dump_addr(skb, cb, type);
4940 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4941 struct netlink_ext_ack *extack)
4943 struct net *net = sock_net(in_skb->sk);
4944 struct ifaddrmsg *ifm;
4945 struct nlattr *tb[IFA_MAX+1];
4946 struct in6_addr *addr = NULL, *peer;
4947 struct net_device *dev = NULL;
4948 struct inet6_ifaddr *ifa;
4949 struct sk_buff *skb;
4952 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4957 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4963 ifm = nlmsg_data(nlh);
4965 dev = __dev_get_by_index(net, ifm->ifa_index);
4967 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4969 err = -EADDRNOTAVAIL;
4973 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4979 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4980 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4982 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4983 WARN_ON(err == -EMSGSIZE);
4987 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4994 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4996 struct sk_buff *skb;
4997 struct net *net = dev_net(ifa->idev->dev);
5000 /* Don't send DELADDR notification for TENTATIVE address,
5001 * since NEWADDR notification is sent only after removing
5002 * TENTATIVE flag, if DAD has not failed.
5004 if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
5005 event == RTM_DELADDR)
5008 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5012 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5014 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5015 WARN_ON(err == -EMSGSIZE);
5019 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5023 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5026 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5027 __s32 *array, int bytes)
5029 BUG_ON(bytes < (DEVCONF_MAX * 4));
5031 memset(array, 0, bytes);
5032 array[DEVCONF_FORWARDING] = cnf->forwarding;
5033 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5034 array[DEVCONF_MTU6] = cnf->mtu6;
5035 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5036 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5037 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5038 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5039 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5040 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5041 jiffies_to_msecs(cnf->rtr_solicit_interval);
5042 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5043 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5044 array[DEVCONF_RTR_SOLICIT_DELAY] =
5045 jiffies_to_msecs(cnf->rtr_solicit_delay);
5046 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5047 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5048 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5049 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5050 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5051 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5052 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5053 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5054 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5055 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5056 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5057 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5058 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5059 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5060 #ifdef CONFIG_IPV6_ROUTER_PREF
5061 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5062 array[DEVCONF_RTR_PROBE_INTERVAL] =
5063 jiffies_to_msecs(cnf->rtr_probe_interval);
5064 #ifdef CONFIG_IPV6_ROUTE_INFO
5065 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5066 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5069 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5070 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5071 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5072 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5073 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5075 #ifdef CONFIG_IPV6_MROUTE
5076 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5078 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5079 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5080 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5081 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5082 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5083 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5084 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5085 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5086 /* we omit DEVCONF_STABLE_SECRET for now */
5087 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5088 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5089 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5090 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5091 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5092 #ifdef CONFIG_IPV6_SEG6_HMAC
5093 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5095 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5096 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5097 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5100 static inline size_t inet6_ifla6_size(void)
5102 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5103 + nla_total_size(sizeof(struct ifla_cacheinfo))
5104 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5105 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5106 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5107 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5110 static inline size_t inet6_if_nlmsg_size(void)
5112 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5113 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5114 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5115 + nla_total_size(4) /* IFLA_MTU */
5116 + nla_total_size(4) /* IFLA_LINK */
5117 + nla_total_size(1) /* IFLA_OPERSTATE */
5118 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5121 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5125 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5128 /* Use put_unaligned() because stats may not be aligned for u64. */
5129 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5130 for (i = 1; i < ICMP6_MIB_MAX; i++)
5131 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5133 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5136 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5137 int bytes, size_t syncpoff)
5140 u64 buff[IPSTATS_MIB_MAX];
5141 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5145 memset(buff, 0, sizeof(buff));
5146 buff[0] = IPSTATS_MIB_MAX;
5148 for_each_possible_cpu(c) {
5149 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5150 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5153 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5154 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5157 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5161 case IFLA_INET6_STATS:
5162 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5163 offsetof(struct ipstats_mib, syncp));
5165 case IFLA_INET6_ICMP6STATS:
5166 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5171 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5172 u32 ext_filter_mask)
5175 struct ifla_cacheinfo ci;
5177 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5178 goto nla_put_failure;
5179 ci.max_reasm_len = IPV6_MAXPLEN;
5180 ci.tstamp = cstamp_delta(idev->tstamp);
5181 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5182 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5183 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5184 goto nla_put_failure;
5185 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5187 goto nla_put_failure;
5188 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5190 /* XXX - MC not implemented */
5192 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5195 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5197 goto nla_put_failure;
5198 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5200 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5202 goto nla_put_failure;
5203 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5205 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5207 goto nla_put_failure;
5209 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5210 goto nla_put_failure;
5212 read_lock_bh(&idev->lock);
5213 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5214 read_unlock_bh(&idev->lock);
5222 static size_t inet6_get_link_af_size(const struct net_device *dev,
5223 u32 ext_filter_mask)
5225 if (!__in6_dev_get(dev))
5228 return inet6_ifla6_size();
5231 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5232 u32 ext_filter_mask)
5234 struct inet6_dev *idev = __in6_dev_get(dev);
5239 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5245 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5247 struct inet6_ifaddr *ifp;
5248 struct net_device *dev = idev->dev;
5249 bool clear_token, update_rs = false;
5250 struct in6_addr ll_addr;
5256 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5258 if (!ipv6_accept_ra(idev))
5260 if (idev->cnf.rtr_solicits == 0)
5263 write_lock_bh(&idev->lock);
5265 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5266 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5268 write_unlock_bh(&idev->lock);
5270 clear_token = ipv6_addr_any(token);
5274 if (!idev->dead && (idev->if_flags & IF_READY) &&
5275 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5276 IFA_F_OPTIMISTIC)) {
5277 /* If we're not ready, then normal ifup will take care
5278 * of this. Otherwise, we need to request our rs here.
5280 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5285 write_lock_bh(&idev->lock);
5288 idev->if_flags |= IF_RS_SENT;
5289 idev->rs_interval = rfc3315_s14_backoff_init(
5290 idev->cnf.rtr_solicit_interval);
5291 idev->rs_probes = 1;
5292 addrconf_mod_rs_timer(idev, idev->rs_interval);
5295 /* Well, that's kinda nasty ... */
5296 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5297 spin_lock(&ifp->lock);
5298 if (ifp->tokenized) {
5300 ifp->prefered_lft = 0;
5302 spin_unlock(&ifp->lock);
5305 write_unlock_bh(&idev->lock);
5306 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5307 addrconf_verify_rtnl();
5311 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5312 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5313 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5316 static int inet6_validate_link_af(const struct net_device *dev,
5317 const struct nlattr *nla)
5319 struct nlattr *tb[IFLA_INET6_MAX + 1];
5321 if (dev && !__in6_dev_get(dev))
5322 return -EAFNOSUPPORT;
5324 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5328 static int check_addr_gen_mode(int mode)
5330 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5331 mode != IN6_ADDR_GEN_MODE_NONE &&
5332 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5333 mode != IN6_ADDR_GEN_MODE_RANDOM)
5338 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5341 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5342 !idev->cnf.stable_secret.initialized &&
5343 !net->ipv6.devconf_dflt->stable_secret.initialized)
5348 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5351 struct inet6_dev *idev = __in6_dev_get(dev);
5352 struct nlattr *tb[IFLA_INET6_MAX + 1];
5355 return -EAFNOSUPPORT;
5357 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5360 if (tb[IFLA_INET6_TOKEN]) {
5361 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5366 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5367 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5369 if (check_addr_gen_mode(mode) < 0 ||
5370 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5373 idev->cnf.addr_gen_mode = mode;
5380 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5381 u32 portid, u32 seq, int event, unsigned int flags)
5383 struct net_device *dev = idev->dev;
5384 struct ifinfomsg *hdr;
5385 struct nlmsghdr *nlh;
5388 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5392 hdr = nlmsg_data(nlh);
5393 hdr->ifi_family = AF_INET6;
5395 hdr->ifi_type = dev->type;
5396 hdr->ifi_index = dev->ifindex;
5397 hdr->ifi_flags = dev_get_flags(dev);
5398 hdr->ifi_change = 0;
5400 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5402 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5403 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5404 (dev->ifindex != dev_get_iflink(dev) &&
5405 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5406 nla_put_u8(skb, IFLA_OPERSTATE,
5407 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5408 goto nla_put_failure;
5409 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5411 goto nla_put_failure;
5413 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5414 goto nla_put_failure;
5416 nla_nest_end(skb, protoinfo);
5417 nlmsg_end(skb, nlh);
5421 nlmsg_cancel(skb, nlh);
5425 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5427 struct net *net = sock_net(skb->sk);
5430 struct net_device *dev;
5431 struct inet6_dev *idev;
5432 struct hlist_head *head;
5435 s_idx = cb->args[1];
5438 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5440 head = &net->dev_index_head[h];
5441 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5444 idev = __in6_dev_get(dev);
5447 if (inet6_fill_ifinfo(skb, idev,
5448 NETLINK_CB(cb->skb).portid,
5450 RTM_NEWLINK, NLM_F_MULTI) < 0)
5464 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5466 struct sk_buff *skb;
5467 struct net *net = dev_net(idev->dev);
5470 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5474 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5476 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5477 WARN_ON(err == -EMSGSIZE);
5481 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5485 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5488 static inline size_t inet6_prefix_nlmsg_size(void)
5490 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5491 + nla_total_size(sizeof(struct in6_addr))
5492 + nla_total_size(sizeof(struct prefix_cacheinfo));
5495 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5496 struct prefix_info *pinfo, u32 portid, u32 seq,
5497 int event, unsigned int flags)
5499 struct prefixmsg *pmsg;
5500 struct nlmsghdr *nlh;
5501 struct prefix_cacheinfo ci;
5503 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5507 pmsg = nlmsg_data(nlh);
5508 pmsg->prefix_family = AF_INET6;
5509 pmsg->prefix_pad1 = 0;
5510 pmsg->prefix_pad2 = 0;
5511 pmsg->prefix_ifindex = idev->dev->ifindex;
5512 pmsg->prefix_len = pinfo->prefix_len;
5513 pmsg->prefix_type = pinfo->type;
5514 pmsg->prefix_pad3 = 0;
5515 pmsg->prefix_flags = 0;
5517 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5518 if (pinfo->autoconf)
5519 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5521 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5522 goto nla_put_failure;
5523 ci.preferred_time = ntohl(pinfo->prefered);
5524 ci.valid_time = ntohl(pinfo->valid);
5525 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5526 goto nla_put_failure;
5527 nlmsg_end(skb, nlh);
5531 nlmsg_cancel(skb, nlh);
5535 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5536 struct prefix_info *pinfo)
5538 struct sk_buff *skb;
5539 struct net *net = dev_net(idev->dev);
5542 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5546 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5548 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5549 WARN_ON(err == -EMSGSIZE);
5553 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5557 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5560 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5562 struct net *net = dev_net(ifp->idev->dev);
5567 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5572 * If the address was optimistic we inserted the route at the
5573 * start of our DAD process, so we don't need to do it again.
5574 * If the device was taken down in the middle of the DAD
5575 * cycle there is a race where we could get here without a
5576 * host route, so nothing to insert. That will be fixed when
5577 * the device is brought up.
5579 if (ifp->rt && !rcu_access_pointer(ifp->rt->rt6i_node)) {
5580 ip6_ins_rt(ifp->rt);
5581 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
5582 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
5583 &ifp->addr, ifp->idev->dev->name);
5586 if (ifp->idev->cnf.forwarding)
5587 addrconf_join_anycast(ifp);
5588 if (!ipv6_addr_any(&ifp->peer_addr))
5589 addrconf_prefix_route(&ifp->peer_addr, 128,
5590 ifp->idev->dev, 0, 0);
5593 if (ifp->idev->cnf.forwarding)
5594 addrconf_leave_anycast(ifp);
5595 addrconf_leave_solict(ifp->idev, &ifp->addr);
5596 if (!ipv6_addr_any(&ifp->peer_addr)) {
5597 struct rt6_info *rt;
5599 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5600 ifp->idev->dev, 0, 0);
5605 if (dst_hold_safe(&ifp->rt->dst))
5606 ip6_del_rt(ifp->rt);
5608 rt_genid_bump_ipv6(net);
5611 atomic_inc(&net->ipv6.dev_addr_genid);
5614 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5617 if (likely(ifp->idev->dead == 0))
5618 __ipv6_ifa_notify(event, ifp);
5619 rcu_read_unlock_bh();
5622 #ifdef CONFIG_SYSCTL
5625 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5626 void __user *buffer, size_t *lenp, loff_t *ppos)
5628 int *valp = ctl->data;
5631 struct ctl_table lctl;
5635 * ctl->data points to idev->cnf.forwarding, we should
5636 * not modify it until we get the rtnl lock.
5641 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5644 ret = addrconf_fixup_forwarding(ctl, valp, val);
5651 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5652 void __user *buffer, size_t *lenp, loff_t *ppos)
5654 struct inet6_dev *idev = ctl->extra1;
5655 int min_mtu = IPV6_MIN_MTU;
5656 struct ctl_table lctl;
5659 lctl.extra1 = &min_mtu;
5660 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5662 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5665 static void dev_disable_change(struct inet6_dev *idev)
5667 struct netdev_notifier_info info;
5669 if (!idev || !idev->dev)
5672 netdev_notifier_info_init(&info, idev->dev);
5673 if (idev->cnf.disable_ipv6)
5674 addrconf_notify(NULL, NETDEV_DOWN, &info);
5676 addrconf_notify(NULL, NETDEV_UP, &info);
5679 static void addrconf_disable_change(struct net *net, __s32 newf)
5681 struct net_device *dev;
5682 struct inet6_dev *idev;
5684 for_each_netdev(net, dev) {
5685 idev = __in6_dev_get(dev);
5687 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5688 idev->cnf.disable_ipv6 = newf;
5690 dev_disable_change(idev);
5695 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5700 if (!rtnl_trylock())
5701 return restart_syscall();
5703 net = (struct net *)table->extra2;
5707 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5712 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5713 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5714 addrconf_disable_change(net, newf);
5715 } else if ((!newf) ^ (!old))
5716 dev_disable_change((struct inet6_dev *)table->extra1);
5723 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5724 void __user *buffer, size_t *lenp, loff_t *ppos)
5726 int *valp = ctl->data;
5729 struct ctl_table lctl;
5733 * ctl->data points to idev->cnf.disable_ipv6, we should
5734 * not modify it until we get the rtnl lock.
5739 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5742 ret = addrconf_disable_ipv6(ctl, valp, val);
5749 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5750 void __user *buffer, size_t *lenp, loff_t *ppos)
5752 int *valp = ctl->data;
5757 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5760 if (write && old != new) {
5761 struct net *net = ctl->extra2;
5763 if (!rtnl_trylock())
5764 return restart_syscall();
5766 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5767 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5768 NETCONFA_PROXY_NEIGH,
5769 NETCONFA_IFINDEX_DEFAULT,
5770 net->ipv6.devconf_dflt);
5771 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5772 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5773 NETCONFA_PROXY_NEIGH,
5774 NETCONFA_IFINDEX_ALL,
5775 net->ipv6.devconf_all);
5777 struct inet6_dev *idev = ctl->extra1;
5779 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5780 NETCONFA_PROXY_NEIGH,
5790 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5791 void __user *buffer, size_t *lenp,
5796 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5797 struct net *net = (struct net *)ctl->extra2;
5799 if (!rtnl_trylock())
5800 return restart_syscall();
5802 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5805 new_val = *((int *)ctl->data);
5807 if (check_addr_gen_mode(new_val) < 0) {
5812 /* request for default */
5813 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5814 ipv6_devconf_dflt.addr_gen_mode = new_val;
5816 /* request for individual net device */
5821 if (check_stable_privacy(idev, net, new_val) < 0) {
5826 if (idev->cnf.addr_gen_mode != new_val) {
5827 idev->cnf.addr_gen_mode = new_val;
5828 addrconf_dev_config(idev->dev);
5839 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5840 void __user *buffer, size_t *lenp,
5844 struct in6_addr addr;
5845 char str[IPV6_MAX_STRLEN];
5846 struct ctl_table lctl = *ctl;
5847 struct net *net = ctl->extra2;
5848 struct ipv6_stable_secret *secret = ctl->data;
5850 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5853 lctl.maxlen = IPV6_MAX_STRLEN;
5856 if (!rtnl_trylock())
5857 return restart_syscall();
5859 if (!write && !secret->initialized) {
5864 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5865 if (err >= sizeof(str)) {
5870 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5874 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5879 secret->initialized = true;
5880 secret->secret = addr;
5882 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5883 struct net_device *dev;
5885 for_each_netdev(net, dev) {
5886 struct inet6_dev *idev = __in6_dev_get(dev);
5889 idev->cnf.addr_gen_mode =
5890 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5894 struct inet6_dev *idev = ctl->extra1;
5896 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5906 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5908 void __user *buffer,
5912 int *valp = ctl->data;
5915 struct ctl_table lctl;
5918 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5919 * we should not modify it until we get the rtnl lock.
5924 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5927 ret = addrconf_fixup_linkdown(ctl, valp, val);
5934 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5938 rt->dst.flags |= DST_NOPOLICY;
5940 rt->dst.flags &= ~DST_NOPOLICY;
5945 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5947 struct inet6_ifaddr *ifa;
5949 read_lock_bh(&idev->lock);
5950 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5951 spin_lock(&ifa->lock);
5953 struct rt6_info *rt = ifa->rt;
5954 struct fib6_table *table = rt->rt6i_table;
5957 read_lock(&table->tb6_lock);
5958 addrconf_set_nopolicy(ifa->rt, val);
5959 if (rt->rt6i_pcpu) {
5960 for_each_possible_cpu(cpu) {
5961 struct rt6_info **rtp;
5963 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5964 addrconf_set_nopolicy(*rtp, val);
5967 read_unlock(&table->tb6_lock);
5969 spin_unlock(&ifa->lock);
5971 read_unlock_bh(&idev->lock);
5975 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5977 struct inet6_dev *idev;
5980 if (!rtnl_trylock())
5981 return restart_syscall();
5985 net = (struct net *)ctl->extra2;
5986 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5991 if (valp == &net->ipv6.devconf_all->disable_policy) {
5992 struct net_device *dev;
5994 for_each_netdev(net, dev) {
5995 idev = __in6_dev_get(dev);
5997 addrconf_disable_policy_idev(idev, val);
6000 idev = (struct inet6_dev *)ctl->extra1;
6001 addrconf_disable_policy_idev(idev, val);
6009 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6010 void __user *buffer, size_t *lenp,
6013 int *valp = ctl->data;
6016 struct ctl_table lctl;
6021 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6023 if (write && (*valp != val))
6024 ret = addrconf_disable_policy(ctl, valp, val);
6032 static int minus_one = -1;
6033 static const int one = 1;
6034 static const int two_five_five = 255;
6036 static const struct ctl_table addrconf_sysctl[] = {
6038 .procname = "forwarding",
6039 .data = &ipv6_devconf.forwarding,
6040 .maxlen = sizeof(int),
6042 .proc_handler = addrconf_sysctl_forward,
6045 .procname = "hop_limit",
6046 .data = &ipv6_devconf.hop_limit,
6047 .maxlen = sizeof(int),
6049 .proc_handler = proc_dointvec_minmax,
6050 .extra1 = (void *)&one,
6051 .extra2 = (void *)&two_five_five,
6055 .data = &ipv6_devconf.mtu6,
6056 .maxlen = sizeof(int),
6058 .proc_handler = addrconf_sysctl_mtu,
6061 .procname = "accept_ra",
6062 .data = &ipv6_devconf.accept_ra,
6063 .maxlen = sizeof(int),
6065 .proc_handler = proc_dointvec,
6068 .procname = "accept_redirects",
6069 .data = &ipv6_devconf.accept_redirects,
6070 .maxlen = sizeof(int),
6072 .proc_handler = proc_dointvec,
6075 .procname = "autoconf",
6076 .data = &ipv6_devconf.autoconf,
6077 .maxlen = sizeof(int),
6079 .proc_handler = proc_dointvec,
6082 .procname = "dad_transmits",
6083 .data = &ipv6_devconf.dad_transmits,
6084 .maxlen = sizeof(int),
6086 .proc_handler = proc_dointvec,
6089 .procname = "router_solicitations",
6090 .data = &ipv6_devconf.rtr_solicits,
6091 .maxlen = sizeof(int),
6093 .proc_handler = proc_dointvec_minmax,
6094 .extra1 = &minus_one,
6097 .procname = "router_solicitation_interval",
6098 .data = &ipv6_devconf.rtr_solicit_interval,
6099 .maxlen = sizeof(int),
6101 .proc_handler = proc_dointvec_jiffies,
6104 .procname = "router_solicitation_max_interval",
6105 .data = &ipv6_devconf.rtr_solicit_max_interval,
6106 .maxlen = sizeof(int),
6108 .proc_handler = proc_dointvec_jiffies,
6111 .procname = "router_solicitation_delay",
6112 .data = &ipv6_devconf.rtr_solicit_delay,
6113 .maxlen = sizeof(int),
6115 .proc_handler = proc_dointvec_jiffies,
6118 .procname = "force_mld_version",
6119 .data = &ipv6_devconf.force_mld_version,
6120 .maxlen = sizeof(int),
6122 .proc_handler = proc_dointvec,
6125 .procname = "mldv1_unsolicited_report_interval",
6127 &ipv6_devconf.mldv1_unsolicited_report_interval,
6128 .maxlen = sizeof(int),
6130 .proc_handler = proc_dointvec_ms_jiffies,
6133 .procname = "mldv2_unsolicited_report_interval",
6135 &ipv6_devconf.mldv2_unsolicited_report_interval,
6136 .maxlen = sizeof(int),
6138 .proc_handler = proc_dointvec_ms_jiffies,
6141 .procname = "use_tempaddr",
6142 .data = &ipv6_devconf.use_tempaddr,
6143 .maxlen = sizeof(int),
6145 .proc_handler = proc_dointvec,
6148 .procname = "temp_valid_lft",
6149 .data = &ipv6_devconf.temp_valid_lft,
6150 .maxlen = sizeof(int),
6152 .proc_handler = proc_dointvec,
6155 .procname = "temp_prefered_lft",
6156 .data = &ipv6_devconf.temp_prefered_lft,
6157 .maxlen = sizeof(int),
6159 .proc_handler = proc_dointvec,
6162 .procname = "regen_max_retry",
6163 .data = &ipv6_devconf.regen_max_retry,
6164 .maxlen = sizeof(int),
6166 .proc_handler = proc_dointvec,
6169 .procname = "max_desync_factor",
6170 .data = &ipv6_devconf.max_desync_factor,
6171 .maxlen = sizeof(int),
6173 .proc_handler = proc_dointvec,
6176 .procname = "max_addresses",
6177 .data = &ipv6_devconf.max_addresses,
6178 .maxlen = sizeof(int),
6180 .proc_handler = proc_dointvec,
6183 .procname = "accept_ra_defrtr",
6184 .data = &ipv6_devconf.accept_ra_defrtr,
6185 .maxlen = sizeof(int),
6187 .proc_handler = proc_dointvec,
6190 .procname = "accept_ra_min_hop_limit",
6191 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6192 .maxlen = sizeof(int),
6194 .proc_handler = proc_dointvec,
6197 .procname = "accept_ra_pinfo",
6198 .data = &ipv6_devconf.accept_ra_pinfo,
6199 .maxlen = sizeof(int),
6201 .proc_handler = proc_dointvec,
6203 #ifdef CONFIG_IPV6_ROUTER_PREF
6205 .procname = "accept_ra_rtr_pref",
6206 .data = &ipv6_devconf.accept_ra_rtr_pref,
6207 .maxlen = sizeof(int),
6209 .proc_handler = proc_dointvec,
6212 .procname = "router_probe_interval",
6213 .data = &ipv6_devconf.rtr_probe_interval,
6214 .maxlen = sizeof(int),
6216 .proc_handler = proc_dointvec_jiffies,
6218 #ifdef CONFIG_IPV6_ROUTE_INFO
6220 .procname = "accept_ra_rt_info_min_plen",
6221 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6222 .maxlen = sizeof(int),
6224 .proc_handler = proc_dointvec,
6227 .procname = "accept_ra_rt_info_max_plen",
6228 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6229 .maxlen = sizeof(int),
6231 .proc_handler = proc_dointvec,
6236 .procname = "proxy_ndp",
6237 .data = &ipv6_devconf.proxy_ndp,
6238 .maxlen = sizeof(int),
6240 .proc_handler = addrconf_sysctl_proxy_ndp,
6243 .procname = "accept_source_route",
6244 .data = &ipv6_devconf.accept_source_route,
6245 .maxlen = sizeof(int),
6247 .proc_handler = proc_dointvec,
6249 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6251 .procname = "optimistic_dad",
6252 .data = &ipv6_devconf.optimistic_dad,
6253 .maxlen = sizeof(int),
6255 .proc_handler = proc_dointvec,
6258 .procname = "use_optimistic",
6259 .data = &ipv6_devconf.use_optimistic,
6260 .maxlen = sizeof(int),
6262 .proc_handler = proc_dointvec,
6265 #ifdef CONFIG_IPV6_MROUTE
6267 .procname = "mc_forwarding",
6268 .data = &ipv6_devconf.mc_forwarding,
6269 .maxlen = sizeof(int),
6271 .proc_handler = proc_dointvec,
6275 .procname = "disable_ipv6",
6276 .data = &ipv6_devconf.disable_ipv6,
6277 .maxlen = sizeof(int),
6279 .proc_handler = addrconf_sysctl_disable,
6282 .procname = "accept_dad",
6283 .data = &ipv6_devconf.accept_dad,
6284 .maxlen = sizeof(int),
6286 .proc_handler = proc_dointvec,
6289 .procname = "force_tllao",
6290 .data = &ipv6_devconf.force_tllao,
6291 .maxlen = sizeof(int),
6293 .proc_handler = proc_dointvec
6296 .procname = "ndisc_notify",
6297 .data = &ipv6_devconf.ndisc_notify,
6298 .maxlen = sizeof(int),
6300 .proc_handler = proc_dointvec
6303 .procname = "suppress_frag_ndisc",
6304 .data = &ipv6_devconf.suppress_frag_ndisc,
6305 .maxlen = sizeof(int),
6307 .proc_handler = proc_dointvec
6310 .procname = "accept_ra_from_local",
6311 .data = &ipv6_devconf.accept_ra_from_local,
6312 .maxlen = sizeof(int),
6314 .proc_handler = proc_dointvec,
6317 .procname = "accept_ra_mtu",
6318 .data = &ipv6_devconf.accept_ra_mtu,
6319 .maxlen = sizeof(int),
6321 .proc_handler = proc_dointvec,
6324 .procname = "stable_secret",
6325 .data = &ipv6_devconf.stable_secret,
6326 .maxlen = IPV6_MAX_STRLEN,
6328 .proc_handler = addrconf_sysctl_stable_secret,
6331 .procname = "use_oif_addrs_only",
6332 .data = &ipv6_devconf.use_oif_addrs_only,
6333 .maxlen = sizeof(int),
6335 .proc_handler = proc_dointvec,
6338 .procname = "ignore_routes_with_linkdown",
6339 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6340 .maxlen = sizeof(int),
6342 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6345 .procname = "drop_unicast_in_l2_multicast",
6346 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6347 .maxlen = sizeof(int),
6349 .proc_handler = proc_dointvec,
6352 .procname = "drop_unsolicited_na",
6353 .data = &ipv6_devconf.drop_unsolicited_na,
6354 .maxlen = sizeof(int),
6356 .proc_handler = proc_dointvec,
6359 .procname = "keep_addr_on_down",
6360 .data = &ipv6_devconf.keep_addr_on_down,
6361 .maxlen = sizeof(int),
6363 .proc_handler = proc_dointvec,
6367 .procname = "seg6_enabled",
6368 .data = &ipv6_devconf.seg6_enabled,
6369 .maxlen = sizeof(int),
6371 .proc_handler = proc_dointvec,
6373 #ifdef CONFIG_IPV6_SEG6_HMAC
6375 .procname = "seg6_require_hmac",
6376 .data = &ipv6_devconf.seg6_require_hmac,
6377 .maxlen = sizeof(int),
6379 .proc_handler = proc_dointvec,
6383 .procname = "enhanced_dad",
6384 .data = &ipv6_devconf.enhanced_dad,
6385 .maxlen = sizeof(int),
6387 .proc_handler = proc_dointvec,
6390 .procname = "addr_gen_mode",
6391 .data = &ipv6_devconf.addr_gen_mode,
6392 .maxlen = sizeof(int),
6394 .proc_handler = addrconf_sysctl_addr_gen_mode,
6397 .procname = "disable_policy",
6398 .data = &ipv6_devconf.disable_policy,
6399 .maxlen = sizeof(int),
6401 .proc_handler = addrconf_sysctl_disable_policy,
6408 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6409 struct inet6_dev *idev, struct ipv6_devconf *p)
6412 struct ctl_table *table;
6413 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6415 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6419 for (i = 0; table[i].data; i++) {
6420 table[i].data += (char *)p - (char *)&ipv6_devconf;
6421 /* If one of these is already set, then it is not safe to
6422 * overwrite either of them: this makes proc_dointvec_minmax
6425 if (!table[i].extra1 && !table[i].extra2) {
6426 table[i].extra1 = idev; /* embedded; no ref */
6427 table[i].extra2 = net;
6431 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6433 p->sysctl_header = register_net_sysctl(net, path, table);
6434 if (!p->sysctl_header)
6437 if (!strcmp(dev_name, "all"))
6438 ifindex = NETCONFA_IFINDEX_ALL;
6439 else if (!strcmp(dev_name, "default"))
6440 ifindex = NETCONFA_IFINDEX_DEFAULT;
6442 ifindex = idev->dev->ifindex;
6443 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6453 static void __addrconf_sysctl_unregister(struct net *net,
6454 struct ipv6_devconf *p, int ifindex)
6456 struct ctl_table *table;
6458 if (!p->sysctl_header)
6461 table = p->sysctl_header->ctl_table_arg;
6462 unregister_net_sysctl_table(p->sysctl_header);
6463 p->sysctl_header = NULL;
6466 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6469 static int addrconf_sysctl_register(struct inet6_dev *idev)
6473 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6476 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6477 &ndisc_ifinfo_sysctl_change);
6480 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6483 neigh_sysctl_unregister(idev->nd_parms);
6488 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6490 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6491 idev->dev->ifindex);
6492 neigh_sysctl_unregister(idev->nd_parms);
6498 static int __net_init addrconf_init_net(struct net *net)
6501 struct ipv6_devconf *all, *dflt;
6503 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6507 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6509 goto err_alloc_dflt;
6511 /* these will be inherited by all namespaces */
6512 dflt->autoconf = ipv6_defaults.autoconf;
6513 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6515 dflt->stable_secret.initialized = false;
6516 all->stable_secret.initialized = false;
6518 net->ipv6.devconf_all = all;
6519 net->ipv6.devconf_dflt = dflt;
6521 #ifdef CONFIG_SYSCTL
6522 err = __addrconf_sysctl_register(net, "all", NULL, all);
6526 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6532 #ifdef CONFIG_SYSCTL
6534 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6544 static void __net_exit addrconf_exit_net(struct net *net)
6546 #ifdef CONFIG_SYSCTL
6547 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6548 NETCONFA_IFINDEX_DEFAULT);
6549 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6550 NETCONFA_IFINDEX_ALL);
6552 kfree(net->ipv6.devconf_dflt);
6553 kfree(net->ipv6.devconf_all);
6556 static struct pernet_operations addrconf_ops = {
6557 .init = addrconf_init_net,
6558 .exit = addrconf_exit_net,
6561 static struct rtnl_af_ops inet6_ops __read_mostly = {
6563 .fill_link_af = inet6_fill_link_af,
6564 .get_link_af_size = inet6_get_link_af_size,
6565 .validate_link_af = inet6_validate_link_af,
6566 .set_link_af = inet6_set_link_af,
6570 * Init / cleanup code
6573 int __init addrconf_init(void)
6575 struct inet6_dev *idev;
6578 err = ipv6_addr_label_init();
6580 pr_crit("%s: cannot initialize default policy table: %d\n",
6585 err = register_pernet_subsys(&addrconf_ops);
6589 addrconf_wq = create_workqueue("ipv6_addrconf");
6595 /* The addrconf netdev notifier requires that loopback_dev
6596 * has it's ipv6 private information allocated and setup
6597 * before it can bring up and give link-local addresses
6598 * to other devices which are up.
6600 * Unfortunately, loopback_dev is not necessarily the first
6601 * entry in the global dev_base list of net devices. In fact,
6602 * it is likely to be the very last entry on that list.
6603 * So this causes the notifier registry below to try and
6604 * give link-local addresses to all devices besides loopback_dev
6605 * first, then loopback_dev, which cases all the non-loopback_dev
6606 * devices to fail to get a link-local address.
6608 * So, as a temporary fix, allocate the ipv6 structure for
6609 * loopback_dev first by hand.
6610 * Longer term, all of the dependencies ipv6 has upon the loopback
6611 * device and it being up should be removed.
6614 idev = ipv6_add_dev(init_net.loopback_dev);
6617 err = PTR_ERR(idev);
6621 ip6_route_init_special_entries();
6623 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6624 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6626 register_netdevice_notifier(&ipv6_dev_notf);
6630 rtnl_af_register(&inet6_ops);
6632 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6637 /* Only the first call to __rtnl_register can fail */
6638 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, 0);
6639 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, 0);
6640 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6641 inet6_dump_ifaddr, 0);
6642 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6643 inet6_dump_ifmcaddr, 0);
6644 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6645 inet6_dump_ifacaddr, 0);
6646 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6647 inet6_netconf_dump_devconf, 0);
6649 ipv6_addr_label_rtnl_register();
6653 rtnl_af_unregister(&inet6_ops);
6654 unregister_netdevice_notifier(&ipv6_dev_notf);
6656 destroy_workqueue(addrconf_wq);
6658 unregister_pernet_subsys(&addrconf_ops);
6660 ipv6_addr_label_cleanup();
6665 void addrconf_cleanup(void)
6667 struct net_device *dev;
6670 unregister_netdevice_notifier(&ipv6_dev_notf);
6671 unregister_pernet_subsys(&addrconf_ops);
6672 ipv6_addr_label_cleanup();
6676 __rtnl_af_unregister(&inet6_ops);
6678 /* clean dev list */
6679 for_each_netdev(&init_net, dev) {
6680 if (__in6_dev_get(dev) == NULL)
6682 addrconf_ifdown(dev, 1);
6684 addrconf_ifdown(init_net.loopback_dev, 2);
6689 spin_lock_bh(&addrconf_hash_lock);
6690 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6691 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6692 spin_unlock_bh(&addrconf_hash_lock);
6693 cancel_delayed_work(&addr_chk_work);
6696 destroy_workqueue(addrconf_wq);