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 (!timer_pending(&idev->rs_timer))
330 mod_timer(&idev->rs_timer, jiffies + when);
333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
337 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
341 static int snmp6_alloc_dev(struct inet6_dev *idev)
345 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
346 if (!idev->stats.ipv6)
349 for_each_possible_cpu(i) {
350 struct ipstats_mib *addrconf_stats;
351 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
352 u64_stats_init(&addrconf_stats->syncp);
356 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
358 if (!idev->stats.icmpv6dev)
360 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
362 if (!idev->stats.icmpv6msgdev)
368 kfree(idev->stats.icmpv6dev);
370 free_percpu(idev->stats.ipv6);
375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
377 struct inet6_dev *ndev;
382 if (dev->mtu < IPV6_MIN_MTU)
383 return ERR_PTR(-EINVAL);
385 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
389 rwlock_init(&ndev->lock);
391 INIT_LIST_HEAD(&ndev->addr_list);
392 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
393 (unsigned long)ndev);
394 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
396 if (ndev->cnf.stable_secret.initialized)
397 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
399 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
401 ndev->cnf.mtu6 = dev->mtu;
402 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
403 if (!ndev->nd_parms) {
407 if (ndev->cnf.forwarding)
408 dev_disable_lro(dev);
409 /* We refer to the device */
412 if (snmp6_alloc_dev(ndev) < 0) {
414 "%s: cannot allocate memory for statistics; dev=%s.\n",
415 __func__, dev->name);
416 neigh_parms_release(&nd_tbl, ndev->nd_parms);
422 if (snmp6_register_dev(ndev) < 0) {
424 "%s: cannot create /proc/net/dev_snmp6/%s\n",
425 __func__, dev->name);
429 /* One reference from device. */
430 refcount_set(&ndev->refcnt, 1);
432 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
433 ndev->cnf.accept_dad = -1;
435 #if IS_ENABLED(CONFIG_IPV6_SIT)
436 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
437 pr_info("%s: Disabled Multicast RS\n", dev->name);
438 ndev->cnf.rtr_solicits = 0;
442 INIT_LIST_HEAD(&ndev->tempaddr_list);
443 ndev->desync_factor = U32_MAX;
444 if ((dev->flags&IFF_LOOPBACK) ||
445 dev->type == ARPHRD_TUNNEL ||
446 dev->type == ARPHRD_TUNNEL6 ||
447 dev->type == ARPHRD_SIT ||
448 dev->type == ARPHRD_NONE) {
449 ndev->cnf.use_tempaddr = -1;
451 ipv6_regen_rndid(ndev);
453 ndev->token = in6addr_any;
455 if (netif_running(dev) && addrconf_link_ready(dev))
456 ndev->if_flags |= IF_READY;
458 ipv6_mc_init_dev(ndev);
459 ndev->tstamp = jiffies;
460 err = addrconf_sysctl_register(ndev);
462 ipv6_mc_destroy_dev(ndev);
463 snmp6_unregister_dev(ndev);
466 /* protected by rtnl_lock */
467 rcu_assign_pointer(dev->ip6_ptr, ndev);
469 /* Join interface-local all-node multicast group */
470 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
472 /* Join all-node multicast group */
473 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
475 /* Join all-router multicast group if forwarding is set */
476 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
477 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
482 neigh_parms_release(&nd_tbl, ndev->nd_parms);
484 in6_dev_finish_destroy(ndev);
488 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
490 struct inet6_dev *idev;
494 idev = __in6_dev_get(dev);
496 idev = ipv6_add_dev(dev);
501 if (dev->flags&IFF_UP)
506 static int inet6_netconf_msgsize_devconf(int type)
508 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
509 + nla_total_size(4); /* NETCONFA_IFINDEX */
512 if (type == NETCONFA_ALL)
515 if (all || type == NETCONFA_FORWARDING)
516 size += nla_total_size(4);
517 #ifdef CONFIG_IPV6_MROUTE
518 if (all || type == NETCONFA_MC_FORWARDING)
519 size += nla_total_size(4);
521 if (all || type == NETCONFA_PROXY_NEIGH)
522 size += nla_total_size(4);
524 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
525 size += nla_total_size(4);
530 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
531 struct ipv6_devconf *devconf, u32 portid,
532 u32 seq, int event, unsigned int flags,
535 struct nlmsghdr *nlh;
536 struct netconfmsg *ncm;
539 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
544 if (type == NETCONFA_ALL)
547 ncm = nlmsg_data(nlh);
548 ncm->ncm_family = AF_INET6;
550 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
551 goto nla_put_failure;
556 if ((all || type == NETCONFA_FORWARDING) &&
557 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
558 goto nla_put_failure;
559 #ifdef CONFIG_IPV6_MROUTE
560 if ((all || type == NETCONFA_MC_FORWARDING) &&
561 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
562 devconf->mc_forwarding) < 0)
563 goto nla_put_failure;
565 if ((all || type == NETCONFA_PROXY_NEIGH) &&
566 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
567 goto nla_put_failure;
569 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
570 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
571 devconf->ignore_routes_with_linkdown) < 0)
572 goto nla_put_failure;
579 nlmsg_cancel(skb, nlh);
583 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
584 int ifindex, struct ipv6_devconf *devconf)
589 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
593 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
596 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
597 WARN_ON(err == -EMSGSIZE);
601 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
604 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
607 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
608 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
609 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
610 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
611 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
614 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
615 struct nlmsghdr *nlh,
616 struct netlink_ext_ack *extack)
618 struct net *net = sock_net(in_skb->sk);
619 struct nlattr *tb[NETCONFA_MAX+1];
620 struct netconfmsg *ncm;
622 struct ipv6_devconf *devconf;
623 struct inet6_dev *in6_dev;
624 struct net_device *dev;
628 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
629 devconf_ipv6_policy, extack);
634 if (!tb[NETCONFA_IFINDEX])
637 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
639 case NETCONFA_IFINDEX_ALL:
640 devconf = net->ipv6.devconf_all;
642 case NETCONFA_IFINDEX_DEFAULT:
643 devconf = net->ipv6.devconf_dflt;
646 dev = __dev_get_by_index(net, ifindex);
649 in6_dev = __in6_dev_get(dev);
652 devconf = &in6_dev->cnf;
657 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
661 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
662 NETLINK_CB(in_skb).portid,
663 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
666 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
667 WARN_ON(err == -EMSGSIZE);
671 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
676 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
677 struct netlink_callback *cb)
679 struct net *net = sock_net(skb->sk);
682 struct net_device *dev;
683 struct inet6_dev *idev;
684 struct hlist_head *head;
687 s_idx = idx = cb->args[1];
689 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
691 head = &net->dev_index_head[h];
693 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
695 hlist_for_each_entry_rcu(dev, head, index_hlist) {
698 idev = __in6_dev_get(dev);
702 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
704 NETLINK_CB(cb->skb).portid,
712 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
718 if (h == NETDEV_HASHENTRIES) {
719 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
720 net->ipv6.devconf_all,
721 NETLINK_CB(cb->skb).portid,
723 RTM_NEWNETCONF, NLM_F_MULTI,
729 if (h == NETDEV_HASHENTRIES + 1) {
730 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
731 net->ipv6.devconf_dflt,
732 NETLINK_CB(cb->skb).portid,
734 RTM_NEWNETCONF, NLM_F_MULTI,
748 static void dev_forward_change(struct inet6_dev *idev)
750 struct net_device *dev;
751 struct inet6_ifaddr *ifa;
756 if (idev->cnf.forwarding)
757 dev_disable_lro(dev);
758 if (dev->flags & IFF_MULTICAST) {
759 if (idev->cnf.forwarding) {
760 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
761 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
762 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
764 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
765 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
766 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
770 list_for_each_entry(ifa, &idev->addr_list, if_list) {
771 if (ifa->flags&IFA_F_TENTATIVE)
773 if (idev->cnf.forwarding)
774 addrconf_join_anycast(ifa);
776 addrconf_leave_anycast(ifa);
778 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
780 dev->ifindex, &idev->cnf);
784 static void addrconf_forward_change(struct net *net, __s32 newf)
786 struct net_device *dev;
787 struct inet6_dev *idev;
789 for_each_netdev(net, dev) {
790 idev = __in6_dev_get(dev);
792 int changed = (!idev->cnf.forwarding) ^ (!newf);
793 idev->cnf.forwarding = newf;
795 dev_forward_change(idev);
800 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
806 return restart_syscall();
808 net = (struct net *)table->extra2;
812 if (p == &net->ipv6.devconf_dflt->forwarding) {
813 if ((!newf) ^ (!old))
814 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
816 NETCONFA_IFINDEX_DEFAULT,
817 net->ipv6.devconf_dflt);
822 if (p == &net->ipv6.devconf_all->forwarding) {
823 int old_dflt = net->ipv6.devconf_dflt->forwarding;
825 net->ipv6.devconf_dflt->forwarding = newf;
826 if ((!newf) ^ (!old_dflt))
827 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
829 NETCONFA_IFINDEX_DEFAULT,
830 net->ipv6.devconf_dflt);
832 addrconf_forward_change(net, newf);
833 if ((!newf) ^ (!old))
834 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
836 NETCONFA_IFINDEX_ALL,
837 net->ipv6.devconf_all);
838 } else if ((!newf) ^ (!old))
839 dev_forward_change((struct inet6_dev *)table->extra1);
843 rt6_purge_dflt_routers(net);
847 static void addrconf_linkdown_change(struct net *net, __s32 newf)
849 struct net_device *dev;
850 struct inet6_dev *idev;
852 for_each_netdev(net, dev) {
853 idev = __in6_dev_get(dev);
855 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
857 idev->cnf.ignore_routes_with_linkdown = newf;
859 inet6_netconf_notify_devconf(dev_net(dev),
861 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
868 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
874 return restart_syscall();
876 net = (struct net *)table->extra2;
880 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
881 if ((!newf) ^ (!old))
882 inet6_netconf_notify_devconf(net,
884 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
885 NETCONFA_IFINDEX_DEFAULT,
886 net->ipv6.devconf_dflt);
891 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
892 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
893 addrconf_linkdown_change(net, newf);
894 if ((!newf) ^ (!old))
895 inet6_netconf_notify_devconf(net,
897 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
898 NETCONFA_IFINDEX_ALL,
899 net->ipv6.devconf_all);
908 /* Nobody refers to this ifaddr, destroy it */
909 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
911 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
913 #ifdef NET_REFCNT_DEBUG
914 pr_debug("%s\n", __func__);
917 in6_dev_put(ifp->idev);
919 if (cancel_delayed_work(&ifp->dad_work))
920 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
923 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
924 pr_warn("Freeing alive inet6 address %p\n", ifp);
933 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
936 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
939 * Each device address list is sorted in order of scope -
940 * global before linklocal.
942 list_for_each(p, &idev->addr_list) {
943 struct inet6_ifaddr *ifa
944 = list_entry(p, struct inet6_ifaddr, if_list);
945 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
949 list_add_tail(&ifp->if_list, p);
952 static u32 inet6_addr_hash(const struct in6_addr *addr)
954 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
957 /* On success it returns ifp with increased reference count */
959 static struct inet6_ifaddr *
960 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
961 const struct in6_addr *peer_addr, int pfxlen,
962 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
964 struct net *net = dev_net(idev->dev);
965 struct inet6_ifaddr *ifa = NULL;
967 struct in6_validator_info i6vi;
970 int addr_type = ipv6_addr_type(addr);
972 if (addr_type == IPV6_ADDR_ANY ||
973 addr_type & IPV6_ADDR_MULTICAST ||
974 (!(idev->dev->flags & IFF_LOOPBACK) &&
975 addr_type & IPV6_ADDR_LOOPBACK))
976 return ERR_PTR(-EADDRNOTAVAIL);
983 err = -ENODEV; /*XXX*/
987 if (idev->cnf.disable_ipv6) {
992 i6vi.i6vi_addr = *addr;
993 i6vi.i6vi_dev = idev;
994 rcu_read_unlock_bh();
996 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
999 err = notifier_to_errno(err);
1003 spin_lock(&addrconf_hash_lock);
1005 /* Ignore adding duplicate addresses on an interface */
1006 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
1007 ADBG("ipv6_add_addr: already assigned\n");
1012 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
1015 ADBG("ipv6_add_addr: malloc failed\n");
1020 rt = addrconf_dst_alloc(idev, addr, false);
1026 if (net->ipv6.devconf_all->disable_policy ||
1027 idev->cnf.disable_policy)
1028 rt->dst.flags |= DST_NOPOLICY;
1030 neigh_parms_data_state_setall(idev->nd_parms);
1034 ifa->peer_addr = *peer_addr;
1036 spin_lock_init(&ifa->lock);
1037 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1038 INIT_HLIST_NODE(&ifa->addr_lst);
1040 ifa->prefix_len = pfxlen;
1042 /* No need to add the TENTATIVE flag for addresses with NODAD */
1043 if (!(flags & IFA_F_NODAD))
1044 ifa->flags |= IFA_F_TENTATIVE;
1045 ifa->valid_lft = valid_lft;
1046 ifa->prefered_lft = prefered_lft;
1047 ifa->cstamp = ifa->tstamp = jiffies;
1048 ifa->tokenized = false;
1054 refcount_set(&ifa->refcnt, 1);
1056 /* Add to big hash table */
1057 hash = inet6_addr_hash(addr);
1059 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1060 spin_unlock(&addrconf_hash_lock);
1062 write_lock(&idev->lock);
1063 /* Add to inet6_dev unicast addr list. */
1064 ipv6_link_dev_addr(idev, ifa);
1066 if (ifa->flags&IFA_F_TEMPORARY) {
1067 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1072 write_unlock(&idev->lock);
1074 rcu_read_unlock_bh();
1076 if (likely(err == 0))
1077 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1086 spin_unlock(&addrconf_hash_lock);
1090 enum cleanup_prefix_rt_t {
1091 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1092 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1093 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1097 * Check, whether the prefix for ifp would still need a prefix route
1098 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1101 * 1) we don't purge prefix if address was not permanent.
1102 * prefix is managed by its own lifetime.
1103 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1104 * 3) if there are no addresses, delete prefix.
1105 * 4) if there are still other permanent address(es),
1106 * corresponding prefix is still permanent.
1107 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1108 * don't purge the prefix, assume user space is managing it.
1109 * 6) otherwise, update prefix lifetime to the
1110 * longest valid lifetime among the corresponding
1111 * addresses on the device.
1112 * Note: subsequent RA will update lifetime.
1114 static enum cleanup_prefix_rt_t
1115 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1117 struct inet6_ifaddr *ifa;
1118 struct inet6_dev *idev = ifp->idev;
1119 unsigned long lifetime;
1120 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1124 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1127 if (ifa->prefix_len != ifp->prefix_len ||
1128 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1131 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1132 return CLEANUP_PREFIX_RT_NOP;
1134 action = CLEANUP_PREFIX_RT_EXPIRE;
1136 spin_lock(&ifa->lock);
1138 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1140 * Note: Because this address is
1141 * not permanent, lifetime <
1142 * LONG_MAX / HZ here.
1144 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1145 *expires = ifa->tstamp + lifetime * HZ;
1146 spin_unlock(&ifa->lock);
1153 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1155 struct rt6_info *rt;
1157 rt = addrconf_get_prefix_route(&ifp->addr,
1160 0, RTF_GATEWAY | RTF_DEFAULT);
1165 if (!(rt->rt6i_flags & RTF_EXPIRES))
1166 rt6_set_expires(rt, expires);
1173 /* This function wants to get referenced ifp and releases it before return */
1175 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1178 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1179 unsigned long expires;
1183 spin_lock_bh(&ifp->lock);
1185 ifp->state = INET6_IFADDR_STATE_DEAD;
1186 spin_unlock_bh(&ifp->lock);
1188 if (state == INET6_IFADDR_STATE_DEAD)
1191 spin_lock_bh(&addrconf_hash_lock);
1192 hlist_del_init_rcu(&ifp->addr_lst);
1193 spin_unlock_bh(&addrconf_hash_lock);
1195 write_lock_bh(&ifp->idev->lock);
1197 if (ifp->flags&IFA_F_TEMPORARY) {
1198 list_del(&ifp->tmp_list);
1200 in6_ifa_put(ifp->ifpub);
1206 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1207 action = check_cleanup_prefix_route(ifp, &expires);
1209 list_del_init(&ifp->if_list);
1212 write_unlock_bh(&ifp->idev->lock);
1214 addrconf_del_dad_work(ifp);
1216 ipv6_ifa_notify(RTM_DELADDR, ifp);
1218 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1220 if (action != CLEANUP_PREFIX_RT_NOP) {
1221 cleanup_prefix_route(ifp, expires,
1222 action == CLEANUP_PREFIX_RT_DEL);
1225 /* clean up prefsrc entries */
1226 rt6_remove_prefsrc(ifp);
1231 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1233 struct inet6_dev *idev = ifp->idev;
1234 struct in6_addr addr, *tmpaddr;
1235 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1236 unsigned long regen_advance;
1240 unsigned long now = jiffies;
1241 long max_desync_factor;
1242 s32 cnf_temp_preferred_lft;
1244 write_lock_bh(&idev->lock);
1246 spin_lock_bh(&ift->lock);
1247 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1248 spin_unlock_bh(&ift->lock);
1255 if (idev->cnf.use_tempaddr <= 0) {
1256 write_unlock_bh(&idev->lock);
1257 pr_info("%s: use_tempaddr is disabled\n", __func__);
1262 spin_lock_bh(&ifp->lock);
1263 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1264 idev->cnf.use_tempaddr = -1; /*XXX*/
1265 spin_unlock_bh(&ifp->lock);
1266 write_unlock_bh(&idev->lock);
1267 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1274 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1275 ipv6_try_regen_rndid(idev, tmpaddr);
1276 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1277 age = (now - ifp->tstamp) / HZ;
1279 regen_advance = idev->cnf.regen_max_retry *
1280 idev->cnf.dad_transmits *
1281 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1283 /* recalculate max_desync_factor each time and update
1284 * idev->desync_factor if it's larger
1286 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1287 max_desync_factor = min_t(__u32,
1288 idev->cnf.max_desync_factor,
1289 cnf_temp_preferred_lft - regen_advance);
1291 if (unlikely(idev->desync_factor > max_desync_factor)) {
1292 if (max_desync_factor > 0) {
1293 get_random_bytes(&idev->desync_factor,
1294 sizeof(idev->desync_factor));
1295 idev->desync_factor %= max_desync_factor;
1297 idev->desync_factor = 0;
1301 tmp_valid_lft = min_t(__u32,
1303 idev->cnf.temp_valid_lft + age);
1304 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1305 idev->desync_factor;
1306 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1307 tmp_plen = ifp->prefix_len;
1308 tmp_tstamp = ifp->tstamp;
1309 spin_unlock_bh(&ifp->lock);
1311 write_unlock_bh(&idev->lock);
1313 /* A temporary address is created only if this calculated Preferred
1314 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1315 * an implementation must not create a temporary address with a zero
1316 * Preferred Lifetime.
1317 * Use age calculation as in addrconf_verify to avoid unnecessary
1318 * temporary addresses being generated.
1320 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1321 if (tmp_prefered_lft <= regen_advance + age) {
1328 addr_flags = IFA_F_TEMPORARY;
1329 /* set in addrconf_prefix_rcv() */
1330 if (ifp->flags & IFA_F_OPTIMISTIC)
1331 addr_flags |= IFA_F_OPTIMISTIC;
1333 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1334 ipv6_addr_scope(&addr), addr_flags,
1335 tmp_valid_lft, tmp_prefered_lft);
1339 pr_info("%s: retry temporary address regeneration\n", __func__);
1341 write_lock_bh(&idev->lock);
1345 spin_lock_bh(&ift->lock);
1348 ift->tstamp = tmp_tstamp;
1349 spin_unlock_bh(&ift->lock);
1351 addrconf_dad_start(ift);
1359 * Choose an appropriate source address (RFC3484)
1362 IPV6_SADDR_RULE_INIT = 0,
1363 IPV6_SADDR_RULE_LOCAL,
1364 IPV6_SADDR_RULE_SCOPE,
1365 IPV6_SADDR_RULE_PREFERRED,
1366 #ifdef CONFIG_IPV6_MIP6
1367 IPV6_SADDR_RULE_HOA,
1369 IPV6_SADDR_RULE_OIF,
1370 IPV6_SADDR_RULE_LABEL,
1371 IPV6_SADDR_RULE_PRIVACY,
1372 IPV6_SADDR_RULE_ORCHID,
1373 IPV6_SADDR_RULE_PREFIX,
1374 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1375 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1380 struct ipv6_saddr_score {
1383 struct inet6_ifaddr *ifa;
1384 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1389 struct ipv6_saddr_dst {
1390 const struct in6_addr *addr;
1397 static inline int ipv6_saddr_preferred(int type)
1399 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1404 static bool ipv6_use_optimistic_addr(struct net *net,
1405 struct inet6_dev *idev)
1407 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1410 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1412 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1421 static int ipv6_get_saddr_eval(struct net *net,
1422 struct ipv6_saddr_score *score,
1423 struct ipv6_saddr_dst *dst,
1428 if (i <= score->rule) {
1430 case IPV6_SADDR_RULE_SCOPE:
1431 ret = score->scopedist;
1433 case IPV6_SADDR_RULE_PREFIX:
1434 ret = score->matchlen;
1437 ret = !!test_bit(i, score->scorebits);
1443 case IPV6_SADDR_RULE_INIT:
1444 /* Rule 0: remember if hiscore is not ready yet */
1447 case IPV6_SADDR_RULE_LOCAL:
1448 /* Rule 1: Prefer same address */
1449 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1451 case IPV6_SADDR_RULE_SCOPE:
1452 /* Rule 2: Prefer appropriate scope
1457 * ---+--+-+---> scope
1459 * | d is scope of the destination.
1461 * | \ <- smaller scope is better if
1462 * B-15 | \ if scope is enough for destination.
1463 * | ret = B - scope (-1 <= scope >= d <= 15).
1465 * |/ <- greater is better
1466 * -C / if scope is not enough for destination.
1467 * /| ret = scope - C (-1 <= d < scope <= 15).
1469 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1470 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1471 * Assume B = 0 and we get C > 29.
1473 ret = __ipv6_addr_src_scope(score->addr_type);
1474 if (ret >= dst->scope)
1477 ret -= 128; /* 30 is enough */
1478 score->scopedist = ret;
1480 case IPV6_SADDR_RULE_PREFERRED:
1482 /* Rule 3: Avoid deprecated and optimistic addresses */
1483 u8 avoid = IFA_F_DEPRECATED;
1485 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1486 avoid |= IFA_F_OPTIMISTIC;
1487 ret = ipv6_saddr_preferred(score->addr_type) ||
1488 !(score->ifa->flags & avoid);
1491 #ifdef CONFIG_IPV6_MIP6
1492 case IPV6_SADDR_RULE_HOA:
1494 /* Rule 4: Prefer home address */
1495 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1496 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1500 case IPV6_SADDR_RULE_OIF:
1501 /* Rule 5: Prefer outgoing interface */
1502 ret = (!dst->ifindex ||
1503 dst->ifindex == score->ifa->idev->dev->ifindex);
1505 case IPV6_SADDR_RULE_LABEL:
1506 /* Rule 6: Prefer matching label */
1507 ret = ipv6_addr_label(net,
1508 &score->ifa->addr, score->addr_type,
1509 score->ifa->idev->dev->ifindex) == dst->label;
1511 case IPV6_SADDR_RULE_PRIVACY:
1513 /* Rule 7: Prefer public address
1514 * Note: prefer temporary address if use_tempaddr >= 2
1516 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1517 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1518 score->ifa->idev->cnf.use_tempaddr >= 2;
1519 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1522 case IPV6_SADDR_RULE_ORCHID:
1523 /* Rule 8-: Prefer ORCHID vs ORCHID or
1524 * non-ORCHID vs non-ORCHID
1526 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1527 ipv6_addr_orchid(dst->addr));
1529 case IPV6_SADDR_RULE_PREFIX:
1530 /* Rule 8: Use longest matching prefix */
1531 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1532 if (ret > score->ifa->prefix_len)
1533 ret = score->ifa->prefix_len;
1534 score->matchlen = ret;
1536 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1537 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1538 /* Optimistic addresses still have lower precedence than other
1539 * preferred addresses.
1541 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1549 __set_bit(i, score->scorebits);
1555 static int __ipv6_dev_get_saddr(struct net *net,
1556 struct ipv6_saddr_dst *dst,
1557 struct inet6_dev *idev,
1558 struct ipv6_saddr_score *scores,
1561 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1563 read_lock_bh(&idev->lock);
1564 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1568 * - Tentative Address (RFC2462 section 5.4)
1569 * - A tentative address is not considered
1570 * "assigned to an interface" in the traditional
1571 * sense, unless it is also flagged as optimistic.
1572 * - Candidate Source Address (section 4)
1573 * - In any case, anycast addresses, multicast
1574 * addresses, and the unspecified address MUST
1575 * NOT be included in a candidate set.
1577 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1578 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1581 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1583 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1584 score->addr_type & IPV6_ADDR_MULTICAST)) {
1585 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1591 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1593 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1594 int minihiscore, miniscore;
1596 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1597 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1599 if (minihiscore > miniscore) {
1600 if (i == IPV6_SADDR_RULE_SCOPE &&
1601 score->scopedist > 0) {
1604 * each remaining entry
1605 * has too small (not enough)
1606 * scope, because ifa entries
1607 * are sorted by their scope
1613 } else if (minihiscore < miniscore) {
1615 in6_ifa_put(hiscore->ifa);
1617 in6_ifa_hold(score->ifa);
1619 swap(hiscore, score);
1620 hiscore_idx = 1 - hiscore_idx;
1622 /* restore our iterator */
1623 score->ifa = hiscore->ifa;
1630 read_unlock_bh(&idev->lock);
1634 static int ipv6_get_saddr_master(struct net *net,
1635 const struct net_device *dst_dev,
1636 const struct net_device *master,
1637 struct ipv6_saddr_dst *dst,
1638 struct ipv6_saddr_score *scores,
1641 struct inet6_dev *idev;
1643 idev = __in6_dev_get(dst_dev);
1645 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1646 scores, hiscore_idx);
1648 idev = __in6_dev_get(master);
1650 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1651 scores, hiscore_idx);
1656 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1657 const struct in6_addr *daddr, unsigned int prefs,
1658 struct in6_addr *saddr)
1660 struct ipv6_saddr_score scores[2], *hiscore;
1661 struct ipv6_saddr_dst dst;
1662 struct inet6_dev *idev;
1663 struct net_device *dev;
1665 bool use_oif_addr = false;
1666 int hiscore_idx = 0;
1668 dst_type = __ipv6_addr_type(daddr);
1670 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1671 dst.scope = __ipv6_addr_src_scope(dst_type);
1672 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1675 scores[hiscore_idx].rule = -1;
1676 scores[hiscore_idx].ifa = NULL;
1680 /* Candidate Source Address (section 4)
1681 * - multicast and link-local destination address,
1682 * the set of candidate source address MUST only
1683 * include addresses assigned to interfaces
1684 * belonging to the same link as the outgoing
1686 * (- For site-local destination addresses, the
1687 * set of candidate source addresses MUST only
1688 * include addresses assigned to interfaces
1689 * belonging to the same site as the outgoing
1691 * - "It is RECOMMENDED that the candidate source addresses
1692 * be the set of unicast addresses assigned to the
1693 * interface that will be used to send to the destination
1694 * (the 'outgoing' interface)." (RFC 6724)
1697 idev = __in6_dev_get(dst_dev);
1698 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1699 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1700 (idev && idev->cnf.use_oif_addrs_only)) {
1701 use_oif_addr = true;
1707 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1709 const struct net_device *master;
1712 /* if dst_dev exists and is enslaved to an L3 device, then
1713 * prefer addresses from dst_dev and then the master over
1714 * any other enslaved devices in the L3 domain.
1716 master = l3mdev_master_dev_rcu(dst_dev);
1718 master_idx = master->ifindex;
1720 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1722 scores, hiscore_idx);
1724 if (scores[hiscore_idx].ifa)
1728 for_each_netdev_rcu(net, dev) {
1729 /* only consider addresses on devices in the
1732 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1734 idev = __in6_dev_get(dev);
1737 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1744 hiscore = &scores[hiscore_idx];
1746 return -EADDRNOTAVAIL;
1748 *saddr = hiscore->ifa->addr;
1749 in6_ifa_put(hiscore->ifa);
1752 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1754 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1757 struct inet6_ifaddr *ifp;
1758 int err = -EADDRNOTAVAIL;
1760 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1761 if (ifp->scope > IFA_LINK)
1763 if (ifp->scope == IFA_LINK &&
1764 !(ifp->flags & banned_flags)) {
1773 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1776 struct inet6_dev *idev;
1777 int err = -EADDRNOTAVAIL;
1780 idev = __in6_dev_get(dev);
1782 read_lock_bh(&idev->lock);
1783 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1784 read_unlock_bh(&idev->lock);
1790 static int ipv6_count_addresses(struct inet6_dev *idev)
1793 struct inet6_ifaddr *ifp;
1795 read_lock_bh(&idev->lock);
1796 list_for_each_entry(ifp, &idev->addr_list, if_list)
1798 read_unlock_bh(&idev->lock);
1802 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1803 const struct net_device *dev, int strict)
1805 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1807 EXPORT_SYMBOL(ipv6_chk_addr);
1809 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1810 const struct net_device *dev, int strict,
1813 struct inet6_ifaddr *ifp;
1814 unsigned int hash = inet6_addr_hash(addr);
1818 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1819 if (!net_eq(dev_net(ifp->idev->dev), net))
1821 /* Decouple optimistic from tentative for evaluation here.
1822 * Ban optimistic addresses explicitly, when required.
1824 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1825 ? (ifp->flags&~IFA_F_TENTATIVE)
1827 if (ipv6_addr_equal(&ifp->addr, addr) &&
1828 !(ifp_flags&banned_flags) &&
1829 (!dev || ifp->idev->dev == dev ||
1830 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1831 rcu_read_unlock_bh();
1836 rcu_read_unlock_bh();
1839 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1841 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1842 struct net_device *dev)
1844 unsigned int hash = inet6_addr_hash(addr);
1845 struct inet6_ifaddr *ifp;
1847 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1848 if (!net_eq(dev_net(ifp->idev->dev), net))
1850 if (ipv6_addr_equal(&ifp->addr, addr)) {
1851 if (!dev || ifp->idev->dev == dev)
1858 /* Compares an address/prefix_len with addresses on device @dev.
1859 * If one is found it returns true.
1861 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1862 const unsigned int prefix_len, struct net_device *dev)
1864 struct inet6_dev *idev;
1865 struct inet6_ifaddr *ifa;
1869 idev = __in6_dev_get(dev);
1871 read_lock_bh(&idev->lock);
1872 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1873 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1877 read_unlock_bh(&idev->lock);
1883 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1885 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1887 struct inet6_dev *idev;
1888 struct inet6_ifaddr *ifa;
1893 idev = __in6_dev_get(dev);
1895 read_lock_bh(&idev->lock);
1896 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1897 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1902 read_unlock_bh(&idev->lock);
1907 EXPORT_SYMBOL(ipv6_chk_prefix);
1909 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1910 struct net_device *dev, int strict)
1912 struct inet6_ifaddr *ifp, *result = NULL;
1913 unsigned int hash = inet6_addr_hash(addr);
1916 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1917 if (!net_eq(dev_net(ifp->idev->dev), net))
1919 if (ipv6_addr_equal(&ifp->addr, addr)) {
1920 if (!dev || ifp->idev->dev == dev ||
1921 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1928 rcu_read_unlock_bh();
1933 /* Gets referenced address, destroys ifaddr */
1935 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1938 ifp->flags |= IFA_F_DADFAILED;
1940 if (ifp->flags&IFA_F_TEMPORARY) {
1941 struct inet6_ifaddr *ifpub;
1942 spin_lock_bh(&ifp->lock);
1945 in6_ifa_hold(ifpub);
1946 spin_unlock_bh(&ifp->lock);
1947 ipv6_create_tempaddr(ifpub, ifp);
1950 spin_unlock_bh(&ifp->lock);
1953 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1954 spin_lock_bh(&ifp->lock);
1955 addrconf_del_dad_work(ifp);
1956 ifp->flags |= IFA_F_TENTATIVE;
1957 spin_unlock_bh(&ifp->lock);
1959 ipv6_ifa_notify(0, ifp);
1966 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1970 spin_lock_bh(&ifp->lock);
1971 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1972 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1975 spin_unlock_bh(&ifp->lock);
1980 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1982 struct inet6_dev *idev = ifp->idev;
1983 struct net *net = dev_net(ifp->idev->dev);
1985 if (addrconf_dad_end(ifp)) {
1990 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1991 ifp->idev->dev->name, &ifp->addr);
1993 spin_lock_bh(&ifp->lock);
1995 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1996 int scope = ifp->scope;
1997 u32 flags = ifp->flags;
1998 struct in6_addr new_addr;
1999 struct inet6_ifaddr *ifp2;
2000 u32 valid_lft, preferred_lft;
2001 int pfxlen = ifp->prefix_len;
2002 int retries = ifp->stable_privacy_retry + 1;
2004 if (retries > net->ipv6.sysctl.idgen_retries) {
2005 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2006 ifp->idev->dev->name);
2010 new_addr = ifp->addr;
2011 if (ipv6_generate_stable_address(&new_addr, retries,
2015 valid_lft = ifp->valid_lft;
2016 preferred_lft = ifp->prefered_lft;
2018 spin_unlock_bh(&ifp->lock);
2020 if (idev->cnf.max_addresses &&
2021 ipv6_count_addresses(idev) >=
2022 idev->cnf.max_addresses)
2025 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2026 ifp->idev->dev->name);
2028 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2029 scope, flags, valid_lft,
2034 spin_lock_bh(&ifp2->lock);
2035 ifp2->stable_privacy_retry = retries;
2036 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2037 spin_unlock_bh(&ifp2->lock);
2039 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2042 spin_lock_bh(&ifp->lock);
2046 /* transition from _POSTDAD to _ERRDAD */
2047 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2048 spin_unlock_bh(&ifp->lock);
2050 addrconf_mod_dad_work(ifp, 0);
2054 /* Join to solicited addr multicast group.
2055 * caller must hold RTNL */
2056 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2058 struct in6_addr maddr;
2060 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2063 addrconf_addr_solict_mult(addr, &maddr);
2064 ipv6_dev_mc_inc(dev, &maddr);
2067 /* caller must hold RTNL */
2068 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2070 struct in6_addr maddr;
2072 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2075 addrconf_addr_solict_mult(addr, &maddr);
2076 __ipv6_dev_mc_dec(idev, &maddr);
2079 /* caller must hold RTNL */
2080 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2082 struct in6_addr addr;
2084 if (ifp->prefix_len >= 127) /* RFC 6164 */
2086 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2087 if (ipv6_addr_any(&addr))
2089 __ipv6_dev_ac_inc(ifp->idev, &addr);
2092 /* caller must hold RTNL */
2093 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2095 struct in6_addr addr;
2097 if (ifp->prefix_len >= 127) /* RFC 6164 */
2099 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2100 if (ipv6_addr_any(&addr))
2102 __ipv6_dev_ac_dec(ifp->idev, &addr);
2105 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2107 switch (dev->addr_len) {
2109 memcpy(eui, dev->dev_addr, 3);
2112 memcpy(eui + 5, dev->dev_addr + 3, 3);
2114 case EUI64_ADDR_LEN:
2115 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2125 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2127 union fwnet_hwaddr *ha;
2129 if (dev->addr_len != FWNET_ALEN)
2132 ha = (union fwnet_hwaddr *)dev->dev_addr;
2134 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2139 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2141 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2142 if (dev->addr_len != ARCNET_ALEN)
2145 eui[7] = *(u8 *)dev->dev_addr;
2149 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2151 if (dev->addr_len != INFINIBAND_ALEN)
2153 memcpy(eui, dev->dev_addr + 12, 8);
2158 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2162 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2163 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2164 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2165 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2166 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2167 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2171 memcpy(eui + 4, &addr, 4);
2175 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2177 if (dev->priv_flags & IFF_ISATAP)
2178 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2182 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2184 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2187 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2189 memcpy(eui, dev->perm_addr, 3);
2190 memcpy(eui + 5, dev->perm_addr + 3, 3);
2197 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2199 switch (dev->type) {
2202 return addrconf_ifid_eui48(eui, dev);
2204 return addrconf_ifid_arcnet(eui, dev);
2205 case ARPHRD_INFINIBAND:
2206 return addrconf_ifid_infiniband(eui, dev);
2208 return addrconf_ifid_sit(eui, dev);
2211 return addrconf_ifid_gre(eui, dev);
2212 case ARPHRD_6LOWPAN:
2213 return addrconf_ifid_6lowpan(eui, dev);
2214 case ARPHRD_IEEE1394:
2215 return addrconf_ifid_ieee1394(eui, dev);
2216 case ARPHRD_TUNNEL6:
2218 return addrconf_ifid_ip6tnl(eui, dev);
2223 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2226 struct inet6_ifaddr *ifp;
2228 read_lock_bh(&idev->lock);
2229 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2230 if (ifp->scope > IFA_LINK)
2232 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2233 memcpy(eui, ifp->addr.s6_addr+8, 8);
2238 read_unlock_bh(&idev->lock);
2242 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2243 static void ipv6_regen_rndid(struct inet6_dev *idev)
2246 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2247 idev->rndid[0] &= ~0x02;
2250 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2251 * check if generated address is not inappropriate
2253 * - Reserved subnet anycast (RFC 2526)
2254 * 11111101 11....11 1xxxxxxx
2255 * - ISATAP (RFC4214) 6.1
2256 * 00-00-5E-FE-xx-xx-xx-xx
2258 * - XXX: already assigned to an address on the device
2260 if (idev->rndid[0] == 0xfd &&
2261 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2262 (idev->rndid[7]&0x80))
2264 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2265 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2267 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2272 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2274 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2275 ipv6_regen_rndid(idev);
2283 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2284 unsigned long expires, u32 flags)
2286 struct fib6_config cfg = {
2287 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2288 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2289 .fc_ifindex = dev->ifindex,
2290 .fc_expires = expires,
2292 .fc_flags = RTF_UP | flags,
2293 .fc_nlinfo.nl_net = dev_net(dev),
2294 .fc_protocol = RTPROT_KERNEL,
2299 /* Prevent useless cloning on PtP SIT.
2300 This thing is done here expecting that the whole
2301 class of non-broadcast devices need not cloning.
2303 #if IS_ENABLED(CONFIG_IPV6_SIT)
2304 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2305 cfg.fc_flags |= RTF_NONEXTHOP;
2308 ip6_route_add(&cfg, NULL);
2312 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2314 const struct net_device *dev,
2315 u32 flags, u32 noflags)
2317 struct fib6_node *fn;
2318 struct rt6_info *rt = NULL;
2319 struct fib6_table *table;
2320 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2322 table = fib6_get_table(dev_net(dev), tb_id);
2326 read_lock_bh(&table->tb6_lock);
2327 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2331 noflags |= RTF_CACHE;
2332 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2333 if (rt->dst.dev->ifindex != dev->ifindex)
2335 if ((rt->rt6i_flags & flags) != flags)
2337 if ((rt->rt6i_flags & noflags) != 0)
2343 read_unlock_bh(&table->tb6_lock);
2348 /* Create "default" multicast route to the interface */
2350 static void addrconf_add_mroute(struct net_device *dev)
2352 struct fib6_config cfg = {
2353 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2354 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2355 .fc_ifindex = dev->ifindex,
2358 .fc_nlinfo.nl_net = dev_net(dev),
2359 .fc_protocol = RTPROT_KERNEL,
2362 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2364 ip6_route_add(&cfg, NULL);
2367 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2369 struct inet6_dev *idev;
2373 idev = ipv6_find_idev(dev);
2375 return ERR_PTR(-ENOBUFS);
2377 if (idev->cnf.disable_ipv6)
2378 return ERR_PTR(-EACCES);
2380 /* Add default multicast route */
2381 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2382 addrconf_add_mroute(dev);
2387 static void manage_tempaddrs(struct inet6_dev *idev,
2388 struct inet6_ifaddr *ifp,
2389 __u32 valid_lft, __u32 prefered_lft,
2390 bool create, unsigned long now)
2393 struct inet6_ifaddr *ift;
2395 read_lock_bh(&idev->lock);
2396 /* update all temporary addresses in the list */
2397 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2398 int age, max_valid, max_prefered;
2400 if (ifp != ift->ifpub)
2403 /* RFC 4941 section 3.3:
2404 * If a received option will extend the lifetime of a public
2405 * address, the lifetimes of temporary addresses should
2406 * be extended, subject to the overall constraint that no
2407 * temporary addresses should ever remain "valid" or "preferred"
2408 * for a time longer than (TEMP_VALID_LIFETIME) or
2409 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2411 age = (now - ift->cstamp) / HZ;
2412 max_valid = idev->cnf.temp_valid_lft - age;
2416 max_prefered = idev->cnf.temp_prefered_lft -
2417 idev->desync_factor - age;
2418 if (max_prefered < 0)
2421 if (valid_lft > max_valid)
2422 valid_lft = max_valid;
2424 if (prefered_lft > max_prefered)
2425 prefered_lft = max_prefered;
2427 spin_lock(&ift->lock);
2429 ift->valid_lft = valid_lft;
2430 ift->prefered_lft = prefered_lft;
2432 if (prefered_lft > 0)
2433 ift->flags &= ~IFA_F_DEPRECATED;
2435 spin_unlock(&ift->lock);
2436 if (!(flags&IFA_F_TENTATIVE))
2437 ipv6_ifa_notify(0, ift);
2440 if ((create || list_empty(&idev->tempaddr_list)) &&
2441 idev->cnf.use_tempaddr > 0) {
2442 /* When a new public address is created as described
2443 * in [ADDRCONF], also create a new temporary address.
2444 * Also create a temporary address if it's enabled but
2445 * no temporary address currently exists.
2447 read_unlock_bh(&idev->lock);
2448 ipv6_create_tempaddr(ifp, NULL);
2450 read_unlock_bh(&idev->lock);
2454 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2456 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2457 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2460 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2461 const struct prefix_info *pinfo,
2462 struct inet6_dev *in6_dev,
2463 const struct in6_addr *addr, int addr_type,
2464 u32 addr_flags, bool sllao, bool tokenized,
2465 __u32 valid_lft, u32 prefered_lft)
2467 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2468 int create = 0, update_lft = 0;
2470 if (!ifp && valid_lft) {
2471 int max_addresses = in6_dev->cnf.max_addresses;
2473 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2474 if ((net->ipv6.devconf_all->optimistic_dad ||
2475 in6_dev->cnf.optimistic_dad) &&
2476 !net->ipv6.devconf_all->forwarding && sllao)
2477 addr_flags |= IFA_F_OPTIMISTIC;
2480 /* Do not allow to create too much of autoconfigured
2481 * addresses; this would be too easy way to crash kernel.
2483 if (!max_addresses ||
2484 ipv6_count_addresses(in6_dev) < max_addresses)
2485 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2487 addr_type&IPV6_ADDR_SCOPE_MASK,
2488 addr_flags, valid_lft,
2491 if (IS_ERR_OR_NULL(ifp))
2496 spin_lock_bh(&ifp->lock);
2497 ifp->flags |= IFA_F_MANAGETEMPADDR;
2498 ifp->cstamp = jiffies;
2499 ifp->tokenized = tokenized;
2500 spin_unlock_bh(&ifp->lock);
2501 addrconf_dad_start(ifp);
2509 /* update lifetime (RFC2462 5.5.3 e) */
2510 spin_lock_bh(&ifp->lock);
2512 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2513 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2516 if (!update_lft && !create && stored_lft) {
2517 const u32 minimum_lft = min_t(u32,
2518 stored_lft, MIN_VALID_LIFETIME);
2519 valid_lft = max(valid_lft, minimum_lft);
2521 /* RFC4862 Section 5.5.3e:
2522 * "Note that the preferred lifetime of the
2523 * corresponding address is always reset to
2524 * the Preferred Lifetime in the received
2525 * Prefix Information option, regardless of
2526 * whether the valid lifetime is also reset or
2529 * So we should always update prefered_lft here.
2535 ifp->valid_lft = valid_lft;
2536 ifp->prefered_lft = prefered_lft;
2539 ifp->flags &= ~IFA_F_DEPRECATED;
2540 spin_unlock_bh(&ifp->lock);
2542 if (!(flags&IFA_F_TENTATIVE))
2543 ipv6_ifa_notify(0, ifp);
2545 spin_unlock_bh(&ifp->lock);
2547 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2556 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2558 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2560 struct prefix_info *pinfo;
2565 struct inet6_dev *in6_dev;
2566 struct net *net = dev_net(dev);
2568 pinfo = (struct prefix_info *) opt;
2570 if (len < sizeof(struct prefix_info)) {
2571 ADBG("addrconf: prefix option too short\n");
2576 * Validation checks ([ADDRCONF], page 19)
2579 addr_type = ipv6_addr_type(&pinfo->prefix);
2581 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2584 valid_lft = ntohl(pinfo->valid);
2585 prefered_lft = ntohl(pinfo->prefered);
2587 if (prefered_lft > valid_lft) {
2588 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2592 in6_dev = in6_dev_get(dev);
2595 net_dbg_ratelimited("addrconf: device %s not configured\n",
2601 * Two things going on here:
2602 * 1) Add routes for on-link prefixes
2603 * 2) Configure prefixes with the auto flag set
2606 if (pinfo->onlink) {
2607 struct rt6_info *rt;
2608 unsigned long rt_expires;
2610 /* Avoid arithmetic overflow. Really, we could
2611 * save rt_expires in seconds, likely valid_lft,
2612 * but it would require division in fib gc, that it
2616 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2618 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2620 if (addrconf_finite_timeout(rt_expires))
2623 rt = addrconf_get_prefix_route(&pinfo->prefix,
2626 RTF_ADDRCONF | RTF_PREFIX_RT,
2627 RTF_GATEWAY | RTF_DEFAULT);
2630 /* Autoconf prefix route */
2631 if (valid_lft == 0) {
2634 } else if (addrconf_finite_timeout(rt_expires)) {
2636 rt6_set_expires(rt, jiffies + rt_expires);
2638 rt6_clean_expires(rt);
2640 } else if (valid_lft) {
2641 clock_t expires = 0;
2642 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2643 if (addrconf_finite_timeout(rt_expires)) {
2645 flags |= RTF_EXPIRES;
2646 expires = jiffies_to_clock_t(rt_expires);
2648 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2649 dev, expires, flags);
2654 /* Try to figure out our local address for this prefix */
2656 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2657 struct in6_addr addr;
2658 bool tokenized = false, dev_addr_generated = false;
2660 if (pinfo->prefix_len == 64) {
2661 memcpy(&addr, &pinfo->prefix, 8);
2663 if (!ipv6_addr_any(&in6_dev->token)) {
2664 read_lock_bh(&in6_dev->lock);
2665 memcpy(addr.s6_addr + 8,
2666 in6_dev->token.s6_addr + 8, 8);
2667 read_unlock_bh(&in6_dev->lock);
2669 } else if (is_addr_mode_generate_stable(in6_dev) &&
2670 !ipv6_generate_stable_address(&addr, 0,
2672 addr_flags |= IFA_F_STABLE_PRIVACY;
2674 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2675 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2678 dev_addr_generated = true;
2682 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2687 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2690 tokenized, valid_lft,
2695 /* Ignore error case here because previous prefix add addr was
2696 * successful which will be notified.
2698 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2699 addr_type, addr_flags, sllao,
2700 tokenized, valid_lft,
2702 dev_addr_generated);
2704 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2706 in6_dev_put(in6_dev);
2710 * Set destination address.
2711 * Special case for SIT interfaces where we create a new "virtual"
2714 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2716 struct in6_ifreq ireq;
2717 struct net_device *dev;
2723 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2726 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2732 #if IS_ENABLED(CONFIG_IPV6_SIT)
2733 if (dev->type == ARPHRD_SIT) {
2734 const struct net_device_ops *ops = dev->netdev_ops;
2736 struct ip_tunnel_parm p;
2738 err = -EADDRNOTAVAIL;
2739 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2742 memset(&p, 0, sizeof(p));
2743 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2747 p.iph.protocol = IPPROTO_IPV6;
2749 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2751 if (ops->ndo_do_ioctl) {
2752 mm_segment_t oldfs = get_fs();
2755 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2762 dev = __dev_get_by_name(net, p.name);
2765 err = dev_open(dev);
2775 static int ipv6_mc_config(struct sock *sk, bool join,
2776 const struct in6_addr *addr, int ifindex)
2784 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2786 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2793 * Manual configuration of address on an interface
2795 static int inet6_addr_add(struct net *net, int ifindex,
2796 const struct in6_addr *pfx,
2797 const struct in6_addr *peer_pfx,
2798 unsigned int plen, __u32 ifa_flags,
2799 __u32 prefered_lft, __u32 valid_lft)
2801 struct inet6_ifaddr *ifp;
2802 struct inet6_dev *idev;
2803 struct net_device *dev;
2804 unsigned long timeout;
2814 /* check the lifetime */
2815 if (!valid_lft || prefered_lft > valid_lft)
2818 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2821 dev = __dev_get_by_index(net, ifindex);
2825 idev = addrconf_add_dev(dev);
2827 return PTR_ERR(idev);
2829 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2830 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2831 true, pfx, ifindex);
2837 scope = ipv6_addr_scope(pfx);
2839 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2840 if (addrconf_finite_timeout(timeout)) {
2841 expires = jiffies_to_clock_t(timeout * HZ);
2842 valid_lft = timeout;
2843 flags = RTF_EXPIRES;
2847 ifa_flags |= IFA_F_PERMANENT;
2850 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2851 if (addrconf_finite_timeout(timeout)) {
2853 ifa_flags |= IFA_F_DEPRECATED;
2854 prefered_lft = timeout;
2857 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2858 valid_lft, prefered_lft);
2861 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2862 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2867 * Note that section 3.1 of RFC 4429 indicates
2868 * that the Optimistic flag should not be set for
2869 * manually configured addresses
2871 addrconf_dad_start(ifp);
2872 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2873 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2876 addrconf_verify_rtnl();
2878 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2879 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2880 false, pfx, ifindex);
2883 return PTR_ERR(ifp);
2886 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2887 const struct in6_addr *pfx, unsigned int plen)
2889 struct inet6_ifaddr *ifp;
2890 struct inet6_dev *idev;
2891 struct net_device *dev;
2896 dev = __dev_get_by_index(net, ifindex);
2900 idev = __in6_dev_get(dev);
2904 read_lock_bh(&idev->lock);
2905 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2906 if (ifp->prefix_len == plen &&
2907 ipv6_addr_equal(pfx, &ifp->addr)) {
2909 read_unlock_bh(&idev->lock);
2911 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2912 (ifa_flags & IFA_F_MANAGETEMPADDR))
2913 manage_tempaddrs(idev, ifp, 0, 0, false,
2916 addrconf_verify_rtnl();
2917 if (ipv6_addr_is_multicast(pfx)) {
2918 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2919 false, pfx, dev->ifindex);
2924 read_unlock_bh(&idev->lock);
2925 return -EADDRNOTAVAIL;
2929 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2931 struct in6_ifreq ireq;
2934 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2937 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2941 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2942 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2943 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2948 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2950 struct in6_ifreq ireq;
2953 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2956 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2960 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2961 ireq.ifr6_prefixlen);
2966 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2967 int plen, int scope)
2969 struct inet6_ifaddr *ifp;
2971 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2972 scope, IFA_F_PERMANENT,
2973 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2975 spin_lock_bh(&ifp->lock);
2976 ifp->flags &= ~IFA_F_TENTATIVE;
2977 spin_unlock_bh(&ifp->lock);
2978 rt_genid_bump_ipv6(dev_net(idev->dev));
2979 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2984 #if IS_ENABLED(CONFIG_IPV6_SIT)
2985 static void sit_add_v4_addrs(struct inet6_dev *idev)
2987 struct in6_addr addr;
2988 struct net_device *dev;
2989 struct net *net = dev_net(idev->dev);
2995 memset(&addr, 0, sizeof(struct in6_addr));
2996 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2998 if (idev->dev->flags&IFF_POINTOPOINT) {
2999 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3002 addr.s6_addr32[0] = htonl(0xfe800000);
3006 scope = IPV6_ADDR_COMPATv4;
3008 pflags |= RTF_NONEXTHOP;
3011 if (addr.s6_addr32[3]) {
3012 add_addr(idev, &addr, plen, scope);
3013 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3017 for_each_netdev(net, dev) {
3018 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3019 if (in_dev && (dev->flags & IFF_UP)) {
3020 struct in_ifaddr *ifa;
3024 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3026 addr.s6_addr32[3] = ifa->ifa_local;
3028 if (ifa->ifa_scope == RT_SCOPE_LINK)
3030 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3031 if (idev->dev->flags&IFF_POINTOPOINT)
3036 add_addr(idev, &addr, plen, flag);
3037 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3045 static void init_loopback(struct net_device *dev)
3047 struct inet6_dev *idev;
3053 idev = ipv6_find_idev(dev);
3055 pr_debug("%s: add_dev failed\n", __func__);
3059 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3062 void addrconf_add_linklocal(struct inet6_dev *idev,
3063 const struct in6_addr *addr, u32 flags)
3065 struct inet6_ifaddr *ifp;
3066 u32 addr_flags = flags | IFA_F_PERMANENT;
3068 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3069 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3070 idev->cnf.optimistic_dad) &&
3071 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3072 addr_flags |= IFA_F_OPTIMISTIC;
3075 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3076 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3078 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3079 addrconf_dad_start(ifp);
3083 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3085 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3087 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3090 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3091 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3094 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3095 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3101 static int ipv6_generate_stable_address(struct in6_addr *address,
3103 const struct inet6_dev *idev)
3105 static DEFINE_SPINLOCK(lock);
3106 static __u32 digest[SHA_DIGEST_WORDS];
3107 static __u32 workspace[SHA_WORKSPACE_WORDS];
3110 char __data[SHA_MESSAGE_BYTES];
3112 struct in6_addr secret;
3114 unsigned char hwaddr[MAX_ADDR_LEN];
3119 struct in6_addr secret;
3120 struct in6_addr temp;
3121 struct net *net = dev_net(idev->dev);
3123 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3125 if (idev->cnf.stable_secret.initialized)
3126 secret = idev->cnf.stable_secret.secret;
3127 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3128 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3133 spin_lock_bh(&lock);
3136 memset(&data, 0, sizeof(data));
3137 memset(workspace, 0, sizeof(workspace));
3138 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3139 data.prefix[0] = address->s6_addr32[0];
3140 data.prefix[1] = address->s6_addr32[1];
3141 data.secret = secret;
3142 data.dad_count = dad_count;
3144 sha_transform(digest, data.__data, workspace);
3147 temp.s6_addr32[2] = (__force __be32)digest[0];
3148 temp.s6_addr32[3] = (__force __be32)digest[1];
3150 spin_unlock_bh(&lock);
3152 if (ipv6_reserved_interfaceid(temp)) {
3154 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3163 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3165 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3169 s = &idev->cnf.stable_secret;
3170 get_random_bytes(&s->secret, sizeof(s->secret));
3171 s->initialized = true;
3174 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3176 struct in6_addr addr;
3178 /* no link local addresses on L3 master devices */
3179 if (netif_is_l3_master(idev->dev))
3182 /* no link local addresses on devices flagged as slaves */
3183 if (idev->dev->flags & IFF_SLAVE)
3186 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3188 switch (idev->cnf.addr_gen_mode) {
3189 case IN6_ADDR_GEN_MODE_RANDOM:
3190 ipv6_gen_mode_random_init(idev);
3192 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3193 if (!ipv6_generate_stable_address(&addr, 0, idev))
3194 addrconf_add_linklocal(idev, &addr,
3195 IFA_F_STABLE_PRIVACY);
3196 else if (prefix_route)
3197 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3199 case IN6_ADDR_GEN_MODE_EUI64:
3200 /* addrconf_add_linklocal also adds a prefix_route and we
3201 * only need to care about prefix routes if ipv6_generate_eui64
3202 * couldn't generate one.
3204 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3205 addrconf_add_linklocal(idev, &addr, 0);
3206 else if (prefix_route)
3207 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3209 case IN6_ADDR_GEN_MODE_NONE:
3211 /* will not add any link local address */
3216 static void addrconf_dev_config(struct net_device *dev)
3218 struct inet6_dev *idev;
3222 if ((dev->type != ARPHRD_ETHER) &&
3223 (dev->type != ARPHRD_FDDI) &&
3224 (dev->type != ARPHRD_ARCNET) &&
3225 (dev->type != ARPHRD_INFINIBAND) &&
3226 (dev->type != ARPHRD_IEEE1394) &&
3227 (dev->type != ARPHRD_TUNNEL6) &&
3228 (dev->type != ARPHRD_6LOWPAN) &&
3229 (dev->type != ARPHRD_IP6GRE) &&
3230 (dev->type != ARPHRD_IPGRE) &&
3231 (dev->type != ARPHRD_TUNNEL) &&
3232 (dev->type != ARPHRD_NONE)) {
3233 /* Alas, we support only Ethernet autoconfiguration. */
3234 idev = __in6_dev_get(dev);
3235 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3236 dev->flags & IFF_MULTICAST)
3241 idev = addrconf_add_dev(dev);
3245 /* this device type has no EUI support */
3246 if (dev->type == ARPHRD_NONE &&
3247 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3248 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3250 addrconf_addr_gen(idev, false);
3253 #if IS_ENABLED(CONFIG_IPV6_SIT)
3254 static void addrconf_sit_config(struct net_device *dev)
3256 struct inet6_dev *idev;
3261 * Configure the tunnel with one of our IPv4
3262 * addresses... we should configure all of
3263 * our v4 addrs in the tunnel
3266 idev = ipv6_find_idev(dev);
3268 pr_debug("%s: add_dev failed\n", __func__);
3272 if (dev->priv_flags & IFF_ISATAP) {
3273 addrconf_addr_gen(idev, false);
3277 sit_add_v4_addrs(idev);
3279 if (dev->flags&IFF_POINTOPOINT)
3280 addrconf_add_mroute(dev);
3284 #if IS_ENABLED(CONFIG_NET_IPGRE)
3285 static void addrconf_gre_config(struct net_device *dev)
3287 struct inet6_dev *idev;
3291 idev = ipv6_find_idev(dev);
3293 pr_debug("%s: add_dev failed\n", __func__);
3297 addrconf_addr_gen(idev, true);
3298 if (dev->flags & IFF_POINTOPOINT)
3299 addrconf_add_mroute(dev);
3303 static int fixup_permanent_addr(struct inet6_dev *idev,
3304 struct inet6_ifaddr *ifp)
3306 /* !rt6i_node means the host route was removed from the
3307 * FIB, for example, if 'lo' device is taken down. In that
3308 * case regenerate the host route.
3310 if (!ifp->rt || !ifp->rt->rt6i_node) {
3311 struct rt6_info *rt, *prev;
3313 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3314 if (unlikely(IS_ERR(rt)))
3317 /* ifp->rt can be accessed outside of rtnl */
3318 spin_lock(&ifp->lock);
3321 spin_unlock(&ifp->lock);
3326 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3327 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3331 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3332 addrconf_dad_start(ifp);
3337 static void addrconf_permanent_addr(struct net_device *dev)
3339 struct inet6_ifaddr *ifp, *tmp;
3340 struct inet6_dev *idev;
3342 idev = __in6_dev_get(dev);
3346 write_lock_bh(&idev->lock);
3348 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3349 if ((ifp->flags & IFA_F_PERMANENT) &&
3350 fixup_permanent_addr(idev, ifp) < 0) {
3351 write_unlock_bh(&idev->lock);
3354 write_lock_bh(&idev->lock);
3356 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3357 idev->dev->name, &ifp->addr);
3361 write_unlock_bh(&idev->lock);
3364 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3367 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3368 struct netdev_notifier_changeupper_info *info;
3369 struct inet6_dev *idev = __in6_dev_get(dev);
3370 struct net *net = dev_net(dev);
3371 int run_pending = 0;
3375 case NETDEV_REGISTER:
3376 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3377 idev = ipv6_add_dev(dev);
3379 return notifier_from_errno(PTR_ERR(idev));
3383 case NETDEV_CHANGEMTU:
3384 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3385 if (dev->mtu < IPV6_MIN_MTU) {
3386 addrconf_ifdown(dev, dev != net->loopback_dev);
3391 rt6_mtu_change(dev, dev->mtu);
3392 idev->cnf.mtu6 = dev->mtu;
3396 /* allocate new idev */
3397 idev = ipv6_add_dev(dev);
3401 /* device is still not ready */
3402 if (!(idev->if_flags & IF_READY))
3411 if (dev->flags & IFF_SLAVE)
3414 if (idev && idev->cnf.disable_ipv6)
3417 if (event == NETDEV_UP) {
3418 /* restore routes for permanent addresses */
3419 addrconf_permanent_addr(dev);
3421 if (!addrconf_link_ready(dev)) {
3422 /* device is not ready yet. */
3423 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3428 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3429 idev = ipv6_add_dev(dev);
3431 if (!IS_ERR_OR_NULL(idev)) {
3432 idev->if_flags |= IF_READY;
3435 } else if (event == NETDEV_CHANGE) {
3436 if (!addrconf_link_ready(dev)) {
3437 /* device is still not ready. */
3442 if (idev->if_flags & IF_READY) {
3443 /* device is already configured -
3444 * but resend MLD reports, we might
3445 * have roamed and need to update
3446 * multicast snooping switches
3451 idev->if_flags |= IF_READY;
3454 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3460 switch (dev->type) {
3461 #if IS_ENABLED(CONFIG_IPV6_SIT)
3463 addrconf_sit_config(dev);
3466 #if IS_ENABLED(CONFIG_NET_IPGRE)
3468 addrconf_gre_config(dev);
3471 case ARPHRD_LOOPBACK:
3476 addrconf_dev_config(dev);
3480 if (!IS_ERR_OR_NULL(idev)) {
3482 addrconf_dad_run(idev);
3485 * If the MTU changed during the interface down,
3486 * when the interface up, the changed MTU must be
3487 * reflected in the idev as well as routers.
3489 if (idev->cnf.mtu6 != dev->mtu &&
3490 dev->mtu >= IPV6_MIN_MTU) {
3491 rt6_mtu_change(dev, dev->mtu);
3492 idev->cnf.mtu6 = dev->mtu;
3494 idev->tstamp = jiffies;
3495 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3498 * If the changed mtu during down is lower than
3499 * IPV6_MIN_MTU stop IPv6 on this interface.
3501 if (dev->mtu < IPV6_MIN_MTU)
3502 addrconf_ifdown(dev, dev != net->loopback_dev);
3507 case NETDEV_UNREGISTER:
3509 * Remove all addresses from this interface.
3511 addrconf_ifdown(dev, event != NETDEV_DOWN);
3514 case NETDEV_CHANGENAME:
3516 snmp6_unregister_dev(idev);
3517 addrconf_sysctl_unregister(idev);
3518 err = addrconf_sysctl_register(idev);
3520 return notifier_from_errno(err);
3521 err = snmp6_register_dev(idev);
3523 addrconf_sysctl_unregister(idev);
3524 return notifier_from_errno(err);
3529 case NETDEV_PRE_TYPE_CHANGE:
3530 case NETDEV_POST_TYPE_CHANGE:
3532 addrconf_type_change(dev, event);
3535 case NETDEV_CHANGEUPPER:
3538 /* flush all routes if dev is linked to or unlinked from
3539 * an L3 master device (e.g., VRF)
3541 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3542 addrconf_ifdown(dev, 0);
3549 * addrconf module should be notified of a device going up
3551 static struct notifier_block ipv6_dev_notf = {
3552 .notifier_call = addrconf_notify,
3553 .priority = ADDRCONF_NOTIFY_PRIORITY,
3556 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3558 struct inet6_dev *idev;
3561 idev = __in6_dev_get(dev);
3563 if (event == NETDEV_POST_TYPE_CHANGE)
3564 ipv6_mc_remap(idev);
3565 else if (event == NETDEV_PRE_TYPE_CHANGE)
3566 ipv6_mc_unmap(idev);
3569 static bool addr_is_local(const struct in6_addr *addr)
3571 return ipv6_addr_type(addr) &
3572 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3575 static int addrconf_ifdown(struct net_device *dev, int how)
3577 struct net *net = dev_net(dev);
3578 struct inet6_dev *idev;
3579 struct inet6_ifaddr *ifa, *tmp;
3580 struct list_head del_list;
3587 rt6_ifdown(net, dev);
3588 neigh_ifdown(&nd_tbl, dev);
3590 idev = __in6_dev_get(dev);
3595 * Step 1: remove reference to ipv6 device from parent device.
3601 /* protected by rtnl_lock */
3602 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3604 /* Step 1.5: remove snmp6 entry */
3605 snmp6_unregister_dev(idev);
3609 /* aggregate the system setting and interface setting */
3610 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3612 _keep_addr = idev->cnf.keep_addr_on_down;
3614 /* combine the user config with event to determine if permanent
3615 * addresses are to be removed from address hash table
3617 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3619 /* Step 2: clear hash table */
3620 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3621 struct hlist_head *h = &inet6_addr_lst[i];
3623 spin_lock_bh(&addrconf_hash_lock);
3625 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3626 if (ifa->idev == idev) {
3627 addrconf_del_dad_work(ifa);
3628 /* combined flag + permanent flag decide if
3629 * address is retained on a down event
3632 !(ifa->flags & IFA_F_PERMANENT) ||
3633 addr_is_local(&ifa->addr)) {
3634 hlist_del_init_rcu(&ifa->addr_lst);
3639 spin_unlock_bh(&addrconf_hash_lock);
3642 write_lock_bh(&idev->lock);
3644 addrconf_del_rs_timer(idev);
3646 /* Step 2: clear flags for stateless addrconf */
3648 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3650 /* Step 3: clear tempaddr list */
3651 while (!list_empty(&idev->tempaddr_list)) {
3652 ifa = list_first_entry(&idev->tempaddr_list,
3653 struct inet6_ifaddr, tmp_list);
3654 list_del(&ifa->tmp_list);
3655 write_unlock_bh(&idev->lock);
3656 spin_lock_bh(&ifa->lock);
3659 in6_ifa_put(ifa->ifpub);
3662 spin_unlock_bh(&ifa->lock);
3664 write_lock_bh(&idev->lock);
3667 /* re-combine the user config with event to determine if permanent
3668 * addresses are to be removed from the interface list
3670 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3672 INIT_LIST_HEAD(&del_list);
3673 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3674 struct rt6_info *rt = NULL;
3677 addrconf_del_dad_work(ifa);
3679 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3680 !addr_is_local(&ifa->addr);
3682 list_move(&ifa->if_list, &del_list);
3684 write_unlock_bh(&idev->lock);
3685 spin_lock_bh(&ifa->lock);
3688 /* set state to skip the notifier below */
3689 state = INET6_IFADDR_STATE_DEAD;
3690 ifa->state = INET6_IFADDR_STATE_PREDAD;
3691 if (!(ifa->flags & IFA_F_NODAD))
3692 ifa->flags |= IFA_F_TENTATIVE;
3698 ifa->state = INET6_IFADDR_STATE_DEAD;
3701 spin_unlock_bh(&ifa->lock);
3706 if (state != INET6_IFADDR_STATE_DEAD) {
3707 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3708 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3710 if (idev->cnf.forwarding)
3711 addrconf_leave_anycast(ifa);
3712 addrconf_leave_solict(ifa->idev, &ifa->addr);
3715 write_lock_bh(&idev->lock);
3718 write_unlock_bh(&idev->lock);
3720 /* now clean up addresses to be removed */
3721 while (!list_empty(&del_list)) {
3722 ifa = list_first_entry(&del_list,
3723 struct inet6_ifaddr, if_list);
3724 list_del(&ifa->if_list);
3729 /* Step 5: Discard anycast and multicast list */
3731 ipv6_ac_destroy_dev(idev);
3732 ipv6_mc_destroy_dev(idev);
3737 idev->tstamp = jiffies;
3739 /* Last: Shot the device (if unregistered) */
3741 addrconf_sysctl_unregister(idev);
3742 neigh_parms_release(&nd_tbl, idev->nd_parms);
3743 neigh_ifdown(&nd_tbl, dev);
3749 static void addrconf_rs_timer(unsigned long data)
3751 struct inet6_dev *idev = (struct inet6_dev *)data;
3752 struct net_device *dev = idev->dev;
3753 struct in6_addr lladdr;
3755 write_lock(&idev->lock);
3756 if (idev->dead || !(idev->if_flags & IF_READY))
3759 if (!ipv6_accept_ra(idev))
3762 /* Announcement received after solicitation was sent */
3763 if (idev->if_flags & IF_RA_RCVD)
3766 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3767 write_unlock(&idev->lock);
3768 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3769 ndisc_send_rs(dev, &lladdr,
3770 &in6addr_linklocal_allrouters);
3774 write_lock(&idev->lock);
3775 idev->rs_interval = rfc3315_s14_backoff_update(
3776 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3777 /* The wait after the last probe can be shorter */
3778 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3779 idev->cnf.rtr_solicits) ?
3780 idev->cnf.rtr_solicit_delay :
3784 * Note: we do not support deprecated "all on-link"
3785 * assumption any longer.
3787 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3791 write_unlock(&idev->lock);
3797 * Duplicate Address Detection
3799 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3801 unsigned long rand_num;
3802 struct inet6_dev *idev = ifp->idev;
3805 if (ifp->flags & IFA_F_OPTIMISTIC)
3808 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3811 if (idev->cnf.enhanced_dad ||
3812 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3814 get_random_bytes(&nonce, 6);
3817 ifp->dad_nonce = nonce;
3818 ifp->dad_probes = idev->cnf.dad_transmits;
3819 addrconf_mod_dad_work(ifp, rand_num);
3822 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3824 struct inet6_dev *idev = ifp->idev;
3825 struct net_device *dev = idev->dev;
3826 bool bump_id, notify = false;
3828 addrconf_join_solict(dev, &ifp->addr);
3830 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3832 read_lock_bh(&idev->lock);
3833 spin_lock(&ifp->lock);
3834 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3837 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3838 (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3839 idev->cnf.accept_dad < 1) ||
3840 !(ifp->flags&IFA_F_TENTATIVE) ||
3841 ifp->flags & IFA_F_NODAD) {
3842 bool send_na = false;
3844 if (ifp->flags & IFA_F_TENTATIVE &&
3845 !(ifp->flags & IFA_F_OPTIMISTIC))
3847 bump_id = ifp->flags & IFA_F_TENTATIVE;
3848 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3849 spin_unlock(&ifp->lock);
3850 read_unlock_bh(&idev->lock);
3852 addrconf_dad_completed(ifp, bump_id, send_na);
3856 if (!(idev->if_flags & IF_READY)) {
3857 spin_unlock(&ifp->lock);
3858 read_unlock_bh(&idev->lock);
3860 * If the device is not ready:
3861 * - keep it tentative if it is a permanent address.
3862 * - otherwise, kill it.
3865 addrconf_dad_stop(ifp, 0);
3870 * Optimistic nodes can start receiving
3873 if (ifp->flags & IFA_F_OPTIMISTIC) {
3874 ip6_ins_rt(ifp->rt);
3875 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3876 /* Because optimistic nodes can use this address,
3877 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3883 addrconf_dad_kick(ifp);
3885 spin_unlock(&ifp->lock);
3886 read_unlock_bh(&idev->lock);
3888 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3891 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3893 bool begin_dad = false;
3895 spin_lock_bh(&ifp->lock);
3896 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3897 ifp->state = INET6_IFADDR_STATE_PREDAD;
3900 spin_unlock_bh(&ifp->lock);
3903 addrconf_mod_dad_work(ifp, 0);
3906 static void addrconf_dad_work(struct work_struct *w)
3908 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3909 struct inet6_ifaddr,
3911 struct inet6_dev *idev = ifp->idev;
3912 bool bump_id, disable_ipv6 = false;
3913 struct in6_addr mcaddr;
3919 } action = DAD_PROCESS;
3923 spin_lock_bh(&ifp->lock);
3924 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3926 ifp->state = INET6_IFADDR_STATE_DAD;
3927 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3929 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3931 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3932 idev->cnf.accept_dad > 1) &&
3933 !idev->cnf.disable_ipv6 &&
3934 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3935 struct in6_addr addr;
3937 addr.s6_addr32[0] = htonl(0xfe800000);
3938 addr.s6_addr32[1] = 0;
3940 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3941 ipv6_addr_equal(&ifp->addr, &addr)) {
3942 /* DAD failed for link-local based on MAC */
3943 idev->cnf.disable_ipv6 = 1;
3945 pr_info("%s: IPv6 being disabled!\n",
3946 ifp->idev->dev->name);
3947 disable_ipv6 = true;
3951 spin_unlock_bh(&ifp->lock);
3953 if (action == DAD_BEGIN) {
3954 addrconf_dad_begin(ifp);
3956 } else if (action == DAD_ABORT) {
3958 addrconf_dad_stop(ifp, 1);
3960 addrconf_ifdown(idev->dev, 0);
3964 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3967 write_lock_bh(&idev->lock);
3968 if (idev->dead || !(idev->if_flags & IF_READY)) {
3969 write_unlock_bh(&idev->lock);
3973 spin_lock(&ifp->lock);
3974 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3975 spin_unlock(&ifp->lock);
3976 write_unlock_bh(&idev->lock);
3980 if (ifp->dad_probes == 0) {
3981 bool send_na = false;
3984 * DAD was successful
3987 if (ifp->flags & IFA_F_TENTATIVE &&
3988 !(ifp->flags & IFA_F_OPTIMISTIC))
3990 bump_id = ifp->flags & IFA_F_TENTATIVE;
3991 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3992 spin_unlock(&ifp->lock);
3993 write_unlock_bh(&idev->lock);
3995 addrconf_dad_completed(ifp, bump_id, send_na);
4001 addrconf_mod_dad_work(ifp,
4002 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4003 spin_unlock(&ifp->lock);
4004 write_unlock_bh(&idev->lock);
4006 /* send a neighbour solicitation for our addr */
4007 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4008 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4015 /* ifp->idev must be at least read locked */
4016 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4018 struct inet6_ifaddr *ifpiter;
4019 struct inet6_dev *idev = ifp->idev;
4021 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4022 if (ifpiter->scope > IFA_LINK)
4024 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4025 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4026 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4033 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4036 struct net_device *dev = ifp->idev->dev;
4037 struct in6_addr lladdr;
4038 bool send_rs, send_mld;
4040 addrconf_del_dad_work(ifp);
4043 * Configure the address for reception. Now it is valid.
4046 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4048 /* If added prefix is link local and we are prepared to process
4049 router advertisements, start sending router solicitations.
4052 read_lock_bh(&ifp->idev->lock);
4053 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4054 send_rs = send_mld &&
4055 ipv6_accept_ra(ifp->idev) &&
4056 ifp->idev->cnf.rtr_solicits != 0 &&
4057 (dev->flags&IFF_LOOPBACK) == 0;
4058 read_unlock_bh(&ifp->idev->lock);
4060 /* While dad is in progress mld report's source address is in6_addrany.
4061 * Resend with proper ll now.
4064 ipv6_mc_dad_complete(ifp->idev);
4066 /* send unsolicited NA if enabled */
4068 (ifp->idev->cnf.ndisc_notify ||
4069 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4070 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4071 /*router=*/ !!ifp->idev->cnf.forwarding,
4072 /*solicited=*/ false, /*override=*/ true,
4078 * If a host as already performed a random delay
4079 * [...] as part of DAD [...] there is no need
4080 * to delay again before sending the first RS
4082 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4084 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4086 write_lock_bh(&ifp->idev->lock);
4087 spin_lock(&ifp->lock);
4088 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4089 ifp->idev->cnf.rtr_solicit_interval);
4090 ifp->idev->rs_probes = 1;
4091 ifp->idev->if_flags |= IF_RS_SENT;
4092 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4093 spin_unlock(&ifp->lock);
4094 write_unlock_bh(&ifp->idev->lock);
4098 rt_genid_bump_ipv6(dev_net(dev));
4100 /* Make sure that a new temporary address will be created
4101 * before this temporary address becomes deprecated.
4103 if (ifp->flags & IFA_F_TEMPORARY)
4104 addrconf_verify_rtnl();
4107 static void addrconf_dad_run(struct inet6_dev *idev)
4109 struct inet6_ifaddr *ifp;
4111 read_lock_bh(&idev->lock);
4112 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4113 spin_lock(&ifp->lock);
4114 if (ifp->flags & IFA_F_TENTATIVE &&
4115 ifp->state == INET6_IFADDR_STATE_DAD)
4116 addrconf_dad_kick(ifp);
4117 spin_unlock(&ifp->lock);
4119 read_unlock_bh(&idev->lock);
4122 #ifdef CONFIG_PROC_FS
4123 struct if6_iter_state {
4124 struct seq_net_private p;
4129 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4131 struct inet6_ifaddr *ifa = NULL;
4132 struct if6_iter_state *state = seq->private;
4133 struct net *net = seq_file_net(seq);
4136 /* initial bucket if pos is 0 */
4142 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4143 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4145 if (!net_eq(dev_net(ifa->idev->dev), net))
4147 /* sync with offset */
4148 if (p < state->offset) {
4155 /* prepare for next bucket */
4162 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4163 struct inet6_ifaddr *ifa)
4165 struct if6_iter_state *state = seq->private;
4166 struct net *net = seq_file_net(seq);
4168 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4169 if (!net_eq(dev_net(ifa->idev->dev), net))
4176 while (++state->bucket < IN6_ADDR_HSIZE) {
4177 hlist_for_each_entry_rcu_bh(ifa,
4178 &inet6_addr_lst[state->bucket], addr_lst) {
4179 if (!net_eq(dev_net(ifa->idev->dev), net))
4188 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4192 return if6_get_first(seq, *pos);
4195 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4197 struct inet6_ifaddr *ifa;
4199 ifa = if6_get_next(seq, v);
4204 static void if6_seq_stop(struct seq_file *seq, void *v)
4207 rcu_read_unlock_bh();
4210 static int if6_seq_show(struct seq_file *seq, void *v)
4212 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4213 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4215 ifp->idev->dev->ifindex,
4219 ifp->idev->dev->name);
4223 static const struct seq_operations if6_seq_ops = {
4224 .start = if6_seq_start,
4225 .next = if6_seq_next,
4226 .show = if6_seq_show,
4227 .stop = if6_seq_stop,
4230 static int if6_seq_open(struct inode *inode, struct file *file)
4232 return seq_open_net(inode, file, &if6_seq_ops,
4233 sizeof(struct if6_iter_state));
4236 static const struct file_operations if6_fops = {
4237 .owner = THIS_MODULE,
4238 .open = if6_seq_open,
4240 .llseek = seq_lseek,
4241 .release = seq_release_net,
4244 static int __net_init if6_proc_net_init(struct net *net)
4246 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4251 static void __net_exit if6_proc_net_exit(struct net *net)
4253 remove_proc_entry("if_inet6", net->proc_net);
4256 static struct pernet_operations if6_proc_net_ops = {
4257 .init = if6_proc_net_init,
4258 .exit = if6_proc_net_exit,
4261 int __init if6_proc_init(void)
4263 return register_pernet_subsys(&if6_proc_net_ops);
4266 void if6_proc_exit(void)
4268 unregister_pernet_subsys(&if6_proc_net_ops);
4270 #endif /* CONFIG_PROC_FS */
4272 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4273 /* Check if address is a home address configured on any interface. */
4274 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4277 struct inet6_ifaddr *ifp = NULL;
4278 unsigned int hash = inet6_addr_hash(addr);
4281 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4282 if (!net_eq(dev_net(ifp->idev->dev), net))
4284 if (ipv6_addr_equal(&ifp->addr, addr) &&
4285 (ifp->flags & IFA_F_HOMEADDRESS)) {
4290 rcu_read_unlock_bh();
4296 * Periodic address status verification
4299 static void addrconf_verify_rtnl(void)
4301 unsigned long now, next, next_sec, next_sched;
4302 struct inet6_ifaddr *ifp;
4309 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4311 cancel_delayed_work(&addr_chk_work);
4313 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4315 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4318 /* When setting preferred_lft to a value not zero or
4319 * infinity, while valid_lft is infinity
4320 * IFA_F_PERMANENT has a non-infinity life time.
4322 if ((ifp->flags & IFA_F_PERMANENT) &&
4323 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4326 spin_lock(&ifp->lock);
4327 /* We try to batch several events at once. */
4328 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4330 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4331 age >= ifp->valid_lft) {
4332 spin_unlock(&ifp->lock);
4336 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4337 spin_unlock(&ifp->lock);
4339 } else if (age >= ifp->prefered_lft) {
4340 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4343 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4345 ifp->flags |= IFA_F_DEPRECATED;
4348 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4349 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4350 next = ifp->tstamp + ifp->valid_lft * HZ;
4352 spin_unlock(&ifp->lock);
4357 ipv6_ifa_notify(0, ifp);
4361 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4362 !(ifp->flags&IFA_F_TENTATIVE)) {
4363 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4364 ifp->idev->cnf.dad_transmits *
4365 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4367 if (age >= ifp->prefered_lft - regen_advance) {
4368 struct inet6_ifaddr *ifpub = ifp->ifpub;
4369 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4370 next = ifp->tstamp + ifp->prefered_lft * HZ;
4371 if (!ifp->regen_count && ifpub) {
4374 in6_ifa_hold(ifpub);
4375 spin_unlock(&ifp->lock);
4377 spin_lock(&ifpub->lock);
4378 ifpub->regen_count = 0;
4379 spin_unlock(&ifpub->lock);
4380 ipv6_create_tempaddr(ifpub, ifp);
4385 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4386 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4387 spin_unlock(&ifp->lock);
4389 /* ifp->prefered_lft <= ifp->valid_lft */
4390 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4391 next = ifp->tstamp + ifp->prefered_lft * HZ;
4392 spin_unlock(&ifp->lock);
4397 next_sec = round_jiffies_up(next);
4400 /* If rounded timeout is accurate enough, accept it. */
4401 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4402 next_sched = next_sec;
4404 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4405 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4406 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4408 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4409 now, next, next_sec, next_sched);
4410 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4411 rcu_read_unlock_bh();
4414 static void addrconf_verify_work(struct work_struct *w)
4417 addrconf_verify_rtnl();
4421 static void addrconf_verify(void)
4423 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4426 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4427 struct in6_addr **peer_pfx)
4429 struct in6_addr *pfx = NULL;
4434 pfx = nla_data(addr);
4437 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4439 pfx = nla_data(local);
4445 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4446 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4447 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4448 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4449 [IFA_FLAGS] = { .len = sizeof(u32) },
4453 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4454 struct netlink_ext_ack *extack)
4456 struct net *net = sock_net(skb->sk);
4457 struct ifaddrmsg *ifm;
4458 struct nlattr *tb[IFA_MAX+1];
4459 struct in6_addr *pfx, *peer_pfx;
4463 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4468 ifm = nlmsg_data(nlh);
4469 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4473 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4475 /* We ignore other flags so far. */
4476 ifa_flags &= IFA_F_MANAGETEMPADDR;
4478 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4479 ifm->ifa_prefixlen);
4482 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4483 u32 prefered_lft, u32 valid_lft)
4487 unsigned long timeout;
4488 bool was_managetempaddr;
4489 bool had_prefixroute;
4493 if (!valid_lft || (prefered_lft > valid_lft))
4496 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4497 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4500 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4501 if (addrconf_finite_timeout(timeout)) {
4502 expires = jiffies_to_clock_t(timeout * HZ);
4503 valid_lft = timeout;
4504 flags = RTF_EXPIRES;
4508 ifa_flags |= IFA_F_PERMANENT;
4511 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4512 if (addrconf_finite_timeout(timeout)) {
4514 ifa_flags |= IFA_F_DEPRECATED;
4515 prefered_lft = timeout;
4518 spin_lock_bh(&ifp->lock);
4519 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4520 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4521 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4522 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4523 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4524 IFA_F_NOPREFIXROUTE);
4525 ifp->flags |= ifa_flags;
4526 ifp->tstamp = jiffies;
4527 ifp->valid_lft = valid_lft;
4528 ifp->prefered_lft = prefered_lft;
4530 spin_unlock_bh(&ifp->lock);
4531 if (!(ifp->flags&IFA_F_TENTATIVE))
4532 ipv6_ifa_notify(0, ifp);
4534 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4535 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4537 } else if (had_prefixroute) {
4538 enum cleanup_prefix_rt_t action;
4539 unsigned long rt_expires;
4541 write_lock_bh(&ifp->idev->lock);
4542 action = check_cleanup_prefix_route(ifp, &rt_expires);
4543 write_unlock_bh(&ifp->idev->lock);
4545 if (action != CLEANUP_PREFIX_RT_NOP) {
4546 cleanup_prefix_route(ifp, rt_expires,
4547 action == CLEANUP_PREFIX_RT_DEL);
4551 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4552 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4553 valid_lft = prefered_lft = 0;
4554 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4555 !was_managetempaddr, jiffies);
4558 addrconf_verify_rtnl();
4564 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4565 struct netlink_ext_ack *extack)
4567 struct net *net = sock_net(skb->sk);
4568 struct ifaddrmsg *ifm;
4569 struct nlattr *tb[IFA_MAX+1];
4570 struct in6_addr *pfx, *peer_pfx;
4571 struct inet6_ifaddr *ifa;
4572 struct net_device *dev;
4573 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4577 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4582 ifm = nlmsg_data(nlh);
4583 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4587 if (tb[IFA_CACHEINFO]) {
4588 struct ifa_cacheinfo *ci;
4590 ci = nla_data(tb[IFA_CACHEINFO]);
4591 valid_lft = ci->ifa_valid;
4592 preferred_lft = ci->ifa_prefered;
4594 preferred_lft = INFINITY_LIFE_TIME;
4595 valid_lft = INFINITY_LIFE_TIME;
4598 dev = __dev_get_by_index(net, ifm->ifa_index);
4602 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4604 /* We ignore other flags so far. */
4605 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4606 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4608 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4611 * It would be best to check for !NLM_F_CREATE here but
4612 * userspace already relies on not having to provide this.
4614 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4615 ifm->ifa_prefixlen, ifa_flags,
4616 preferred_lft, valid_lft);
4619 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4620 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4623 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4630 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4631 u8 scope, int ifindex)
4633 struct ifaddrmsg *ifm;
4635 ifm = nlmsg_data(nlh);
4636 ifm->ifa_family = AF_INET6;
4637 ifm->ifa_prefixlen = prefixlen;
4638 ifm->ifa_flags = flags;
4639 ifm->ifa_scope = scope;
4640 ifm->ifa_index = ifindex;
4643 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4644 unsigned long tstamp, u32 preferred, u32 valid)
4646 struct ifa_cacheinfo ci;
4648 ci.cstamp = cstamp_delta(cstamp);
4649 ci.tstamp = cstamp_delta(tstamp);
4650 ci.ifa_prefered = preferred;
4651 ci.ifa_valid = valid;
4653 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4656 static inline int rt_scope(int ifa_scope)
4658 if (ifa_scope & IFA_HOST)
4659 return RT_SCOPE_HOST;
4660 else if (ifa_scope & IFA_LINK)
4661 return RT_SCOPE_LINK;
4662 else if (ifa_scope & IFA_SITE)
4663 return RT_SCOPE_SITE;
4665 return RT_SCOPE_UNIVERSE;
4668 static inline int inet6_ifaddr_msgsize(void)
4670 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4671 + nla_total_size(16) /* IFA_LOCAL */
4672 + nla_total_size(16) /* IFA_ADDRESS */
4673 + nla_total_size(sizeof(struct ifa_cacheinfo))
4674 + nla_total_size(4) /* IFA_FLAGS */;
4677 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4678 u32 portid, u32 seq, int event, unsigned int flags)
4680 struct nlmsghdr *nlh;
4681 u32 preferred, valid;
4683 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4687 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4688 ifa->idev->dev->ifindex);
4690 if (!((ifa->flags&IFA_F_PERMANENT) &&
4691 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4692 preferred = ifa->prefered_lft;
4693 valid = ifa->valid_lft;
4694 if (preferred != INFINITY_LIFE_TIME) {
4695 long tval = (jiffies - ifa->tstamp)/HZ;
4696 if (preferred > tval)
4700 if (valid != INFINITY_LIFE_TIME) {
4708 preferred = INFINITY_LIFE_TIME;
4709 valid = INFINITY_LIFE_TIME;
4712 if (!ipv6_addr_any(&ifa->peer_addr)) {
4713 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4714 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4717 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4720 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4723 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4726 nlmsg_end(skb, nlh);
4730 nlmsg_cancel(skb, nlh);
4734 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4735 u32 portid, u32 seq, int event, u16 flags)
4737 struct nlmsghdr *nlh;
4738 u8 scope = RT_SCOPE_UNIVERSE;
4739 int ifindex = ifmca->idev->dev->ifindex;
4741 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4742 scope = RT_SCOPE_SITE;
4744 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4748 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4749 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4750 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4751 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4752 nlmsg_cancel(skb, nlh);
4756 nlmsg_end(skb, nlh);
4760 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4761 u32 portid, u32 seq, int event, unsigned int flags)
4763 struct nlmsghdr *nlh;
4764 u8 scope = RT_SCOPE_UNIVERSE;
4765 int ifindex = ifaca->aca_idev->dev->ifindex;
4767 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4768 scope = RT_SCOPE_SITE;
4770 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4774 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4775 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4776 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4777 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4778 nlmsg_cancel(skb, nlh);
4782 nlmsg_end(skb, nlh);
4792 /* called with rcu_read_lock() */
4793 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4794 struct netlink_callback *cb, enum addr_type_t type,
4795 int s_ip_idx, int *p_ip_idx)
4797 struct ifmcaddr6 *ifmca;
4798 struct ifacaddr6 *ifaca;
4800 int ip_idx = *p_ip_idx;
4802 read_lock_bh(&idev->lock);
4804 case UNICAST_ADDR: {
4805 struct inet6_ifaddr *ifa;
4807 /* unicast address incl. temp addr */
4808 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4809 if (ip_idx < s_ip_idx)
4811 err = inet6_fill_ifaddr(skb, ifa,
4812 NETLINK_CB(cb->skb).portid,
4818 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4824 case MULTICAST_ADDR:
4825 /* multicast address */
4826 for (ifmca = idev->mc_list; ifmca;
4827 ifmca = ifmca->next, ip_idx++) {
4828 if (ip_idx < s_ip_idx)
4830 err = inet6_fill_ifmcaddr(skb, ifmca,
4831 NETLINK_CB(cb->skb).portid,
4840 /* anycast address */
4841 for (ifaca = idev->ac_list; ifaca;
4842 ifaca = ifaca->aca_next, ip_idx++) {
4843 if (ip_idx < s_ip_idx)
4845 err = inet6_fill_ifacaddr(skb, ifaca,
4846 NETLINK_CB(cb->skb).portid,
4857 read_unlock_bh(&idev->lock);
4862 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4863 enum addr_type_t type)
4865 struct net *net = sock_net(skb->sk);
4868 int s_idx, s_ip_idx;
4869 struct net_device *dev;
4870 struct inet6_dev *idev;
4871 struct hlist_head *head;
4874 s_idx = idx = cb->args[1];
4875 s_ip_idx = ip_idx = cb->args[2];
4878 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4879 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4881 head = &net->dev_index_head[h];
4882 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4885 if (h > s_h || idx > s_idx)
4888 idev = __in6_dev_get(dev);
4892 if (in6_dump_addrs(idev, skb, cb, type,
4893 s_ip_idx, &ip_idx) < 0)
4903 cb->args[2] = ip_idx;
4908 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4910 enum addr_type_t type = UNICAST_ADDR;
4912 return inet6_dump_addr(skb, cb, type);
4915 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4917 enum addr_type_t type = MULTICAST_ADDR;
4919 return inet6_dump_addr(skb, cb, type);
4923 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4925 enum addr_type_t type = ANYCAST_ADDR;
4927 return inet6_dump_addr(skb, cb, type);
4930 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4931 struct netlink_ext_ack *extack)
4933 struct net *net = sock_net(in_skb->sk);
4934 struct ifaddrmsg *ifm;
4935 struct nlattr *tb[IFA_MAX+1];
4936 struct in6_addr *addr = NULL, *peer;
4937 struct net_device *dev = NULL;
4938 struct inet6_ifaddr *ifa;
4939 struct sk_buff *skb;
4942 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4947 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4953 ifm = nlmsg_data(nlh);
4955 dev = __dev_get_by_index(net, ifm->ifa_index);
4957 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4959 err = -EADDRNOTAVAIL;
4963 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4969 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4970 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4972 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4973 WARN_ON(err == -EMSGSIZE);
4977 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4984 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4986 struct sk_buff *skb;
4987 struct net *net = dev_net(ifa->idev->dev);
4990 /* Don't send DELADDR notification for TENTATIVE address,
4991 * since NEWADDR notification is sent only after removing
4992 * TENTATIVE flag, if DAD has not failed.
4994 if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
4995 event == RTM_DELADDR)
4998 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5002 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5004 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5005 WARN_ON(err == -EMSGSIZE);
5009 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5013 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5016 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5017 __s32 *array, int bytes)
5019 BUG_ON(bytes < (DEVCONF_MAX * 4));
5021 memset(array, 0, bytes);
5022 array[DEVCONF_FORWARDING] = cnf->forwarding;
5023 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5024 array[DEVCONF_MTU6] = cnf->mtu6;
5025 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5026 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5027 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5028 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5029 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5030 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5031 jiffies_to_msecs(cnf->rtr_solicit_interval);
5032 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5033 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5034 array[DEVCONF_RTR_SOLICIT_DELAY] =
5035 jiffies_to_msecs(cnf->rtr_solicit_delay);
5036 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5037 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5038 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5039 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5040 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5041 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5042 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5043 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5044 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5045 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5046 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5047 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5048 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5049 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5050 #ifdef CONFIG_IPV6_ROUTER_PREF
5051 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5052 array[DEVCONF_RTR_PROBE_INTERVAL] =
5053 jiffies_to_msecs(cnf->rtr_probe_interval);
5054 #ifdef CONFIG_IPV6_ROUTE_INFO
5055 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5056 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5059 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5060 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5061 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5062 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5063 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5065 #ifdef CONFIG_IPV6_MROUTE
5066 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5068 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5069 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5070 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5071 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5072 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5073 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5074 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5075 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5076 /* we omit DEVCONF_STABLE_SECRET for now */
5077 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5078 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5079 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5080 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5081 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5082 #ifdef CONFIG_IPV6_SEG6_HMAC
5083 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5085 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5086 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5087 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5090 static inline size_t inet6_ifla6_size(void)
5092 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5093 + nla_total_size(sizeof(struct ifla_cacheinfo))
5094 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5095 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5096 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5097 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5100 static inline size_t inet6_if_nlmsg_size(void)
5102 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5103 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5104 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5105 + nla_total_size(4) /* IFLA_MTU */
5106 + nla_total_size(4) /* IFLA_LINK */
5107 + nla_total_size(1) /* IFLA_OPERSTATE */
5108 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5111 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5115 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5118 /* Use put_unaligned() because stats may not be aligned for u64. */
5119 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5120 for (i = 1; i < ICMP6_MIB_MAX; i++)
5121 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5123 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5126 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5127 int bytes, size_t syncpoff)
5130 u64 buff[IPSTATS_MIB_MAX];
5131 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5135 memset(buff, 0, sizeof(buff));
5136 buff[0] = IPSTATS_MIB_MAX;
5138 for_each_possible_cpu(c) {
5139 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5140 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5143 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5144 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5147 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5151 case IFLA_INET6_STATS:
5152 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5153 offsetof(struct ipstats_mib, syncp));
5155 case IFLA_INET6_ICMP6STATS:
5156 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5161 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5162 u32 ext_filter_mask)
5165 struct ifla_cacheinfo ci;
5167 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5168 goto nla_put_failure;
5169 ci.max_reasm_len = IPV6_MAXPLEN;
5170 ci.tstamp = cstamp_delta(idev->tstamp);
5171 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5172 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5173 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5174 goto nla_put_failure;
5175 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5177 goto nla_put_failure;
5178 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5180 /* XXX - MC not implemented */
5182 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5185 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5187 goto nla_put_failure;
5188 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5190 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5192 goto nla_put_failure;
5193 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5195 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5197 goto nla_put_failure;
5199 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5200 goto nla_put_failure;
5202 read_lock_bh(&idev->lock);
5203 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5204 read_unlock_bh(&idev->lock);
5212 static size_t inet6_get_link_af_size(const struct net_device *dev,
5213 u32 ext_filter_mask)
5215 if (!__in6_dev_get(dev))
5218 return inet6_ifla6_size();
5221 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5222 u32 ext_filter_mask)
5224 struct inet6_dev *idev = __in6_dev_get(dev);
5229 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5235 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5237 struct inet6_ifaddr *ifp;
5238 struct net_device *dev = idev->dev;
5239 bool clear_token, update_rs = false;
5240 struct in6_addr ll_addr;
5246 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5248 if (!ipv6_accept_ra(idev))
5250 if (idev->cnf.rtr_solicits == 0)
5253 write_lock_bh(&idev->lock);
5255 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5256 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5258 write_unlock_bh(&idev->lock);
5260 clear_token = ipv6_addr_any(token);
5264 if (!idev->dead && (idev->if_flags & IF_READY) &&
5265 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5266 IFA_F_OPTIMISTIC)) {
5267 /* If we're not ready, then normal ifup will take care
5268 * of this. Otherwise, we need to request our rs here.
5270 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5275 write_lock_bh(&idev->lock);
5278 idev->if_flags |= IF_RS_SENT;
5279 idev->rs_interval = rfc3315_s14_backoff_init(
5280 idev->cnf.rtr_solicit_interval);
5281 idev->rs_probes = 1;
5282 addrconf_mod_rs_timer(idev, idev->rs_interval);
5285 /* Well, that's kinda nasty ... */
5286 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5287 spin_lock(&ifp->lock);
5288 if (ifp->tokenized) {
5290 ifp->prefered_lft = 0;
5292 spin_unlock(&ifp->lock);
5295 write_unlock_bh(&idev->lock);
5296 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5297 addrconf_verify_rtnl();
5301 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5302 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5303 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5306 static int inet6_validate_link_af(const struct net_device *dev,
5307 const struct nlattr *nla)
5309 struct nlattr *tb[IFLA_INET6_MAX + 1];
5311 if (dev && !__in6_dev_get(dev))
5312 return -EAFNOSUPPORT;
5314 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5318 static int check_addr_gen_mode(int mode)
5320 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5321 mode != IN6_ADDR_GEN_MODE_NONE &&
5322 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5323 mode != IN6_ADDR_GEN_MODE_RANDOM)
5328 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5331 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5332 !idev->cnf.stable_secret.initialized &&
5333 !net->ipv6.devconf_dflt->stable_secret.initialized)
5338 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5341 struct inet6_dev *idev = __in6_dev_get(dev);
5342 struct nlattr *tb[IFLA_INET6_MAX + 1];
5345 return -EAFNOSUPPORT;
5347 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5350 if (tb[IFLA_INET6_TOKEN]) {
5351 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5356 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5357 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5359 if (check_addr_gen_mode(mode) < 0 ||
5360 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5363 idev->cnf.addr_gen_mode = mode;
5370 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5371 u32 portid, u32 seq, int event, unsigned int flags)
5373 struct net_device *dev = idev->dev;
5374 struct ifinfomsg *hdr;
5375 struct nlmsghdr *nlh;
5378 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5382 hdr = nlmsg_data(nlh);
5383 hdr->ifi_family = AF_INET6;
5385 hdr->ifi_type = dev->type;
5386 hdr->ifi_index = dev->ifindex;
5387 hdr->ifi_flags = dev_get_flags(dev);
5388 hdr->ifi_change = 0;
5390 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5392 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5393 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5394 (dev->ifindex != dev_get_iflink(dev) &&
5395 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5396 nla_put_u8(skb, IFLA_OPERSTATE,
5397 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5398 goto nla_put_failure;
5399 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5401 goto nla_put_failure;
5403 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5404 goto nla_put_failure;
5406 nla_nest_end(skb, protoinfo);
5407 nlmsg_end(skb, nlh);
5411 nlmsg_cancel(skb, nlh);
5415 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5417 struct net *net = sock_net(skb->sk);
5420 struct net_device *dev;
5421 struct inet6_dev *idev;
5422 struct hlist_head *head;
5425 s_idx = cb->args[1];
5428 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5430 head = &net->dev_index_head[h];
5431 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5434 idev = __in6_dev_get(dev);
5437 if (inet6_fill_ifinfo(skb, idev,
5438 NETLINK_CB(cb->skb).portid,
5440 RTM_NEWLINK, NLM_F_MULTI) < 0)
5454 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5456 struct sk_buff *skb;
5457 struct net *net = dev_net(idev->dev);
5460 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5464 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5466 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5467 WARN_ON(err == -EMSGSIZE);
5471 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5475 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5478 static inline size_t inet6_prefix_nlmsg_size(void)
5480 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5481 + nla_total_size(sizeof(struct in6_addr))
5482 + nla_total_size(sizeof(struct prefix_cacheinfo));
5485 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5486 struct prefix_info *pinfo, u32 portid, u32 seq,
5487 int event, unsigned int flags)
5489 struct prefixmsg *pmsg;
5490 struct nlmsghdr *nlh;
5491 struct prefix_cacheinfo ci;
5493 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5497 pmsg = nlmsg_data(nlh);
5498 pmsg->prefix_family = AF_INET6;
5499 pmsg->prefix_pad1 = 0;
5500 pmsg->prefix_pad2 = 0;
5501 pmsg->prefix_ifindex = idev->dev->ifindex;
5502 pmsg->prefix_len = pinfo->prefix_len;
5503 pmsg->prefix_type = pinfo->type;
5504 pmsg->prefix_pad3 = 0;
5505 pmsg->prefix_flags = 0;
5507 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5508 if (pinfo->autoconf)
5509 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5511 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5512 goto nla_put_failure;
5513 ci.preferred_time = ntohl(pinfo->prefered);
5514 ci.valid_time = ntohl(pinfo->valid);
5515 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5516 goto nla_put_failure;
5517 nlmsg_end(skb, nlh);
5521 nlmsg_cancel(skb, nlh);
5525 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5526 struct prefix_info *pinfo)
5528 struct sk_buff *skb;
5529 struct net *net = dev_net(idev->dev);
5532 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5536 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5538 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5539 WARN_ON(err == -EMSGSIZE);
5543 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5547 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5550 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5552 struct net *net = dev_net(ifp->idev->dev);
5557 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5562 * If the address was optimistic we inserted the route at the
5563 * start of our DAD process, so we don't need to do it again.
5564 * If the device was taken down in the middle of the DAD
5565 * cycle there is a race where we could get here without a
5566 * host route, so nothing to insert. That will be fixed when
5567 * the device is brought up.
5569 if (ifp->rt && !rcu_access_pointer(ifp->rt->rt6i_node)) {
5570 ip6_ins_rt(ifp->rt);
5571 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
5572 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
5573 &ifp->addr, ifp->idev->dev->name);
5576 if (ifp->idev->cnf.forwarding)
5577 addrconf_join_anycast(ifp);
5578 if (!ipv6_addr_any(&ifp->peer_addr))
5579 addrconf_prefix_route(&ifp->peer_addr, 128,
5580 ifp->idev->dev, 0, 0);
5583 if (ifp->idev->cnf.forwarding)
5584 addrconf_leave_anycast(ifp);
5585 addrconf_leave_solict(ifp->idev, &ifp->addr);
5586 if (!ipv6_addr_any(&ifp->peer_addr)) {
5587 struct rt6_info *rt;
5589 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5590 ifp->idev->dev, 0, 0);
5595 if (dst_hold_safe(&ifp->rt->dst))
5596 ip6_del_rt(ifp->rt);
5598 rt_genid_bump_ipv6(net);
5601 atomic_inc(&net->ipv6.dev_addr_genid);
5604 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5607 if (likely(ifp->idev->dead == 0))
5608 __ipv6_ifa_notify(event, ifp);
5609 rcu_read_unlock_bh();
5612 #ifdef CONFIG_SYSCTL
5615 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5616 void __user *buffer, size_t *lenp, loff_t *ppos)
5618 int *valp = ctl->data;
5621 struct ctl_table lctl;
5625 * ctl->data points to idev->cnf.forwarding, we should
5626 * not modify it until we get the rtnl lock.
5631 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5634 ret = addrconf_fixup_forwarding(ctl, valp, val);
5641 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5642 void __user *buffer, size_t *lenp, loff_t *ppos)
5644 struct inet6_dev *idev = ctl->extra1;
5645 int min_mtu = IPV6_MIN_MTU;
5646 struct ctl_table lctl;
5649 lctl.extra1 = &min_mtu;
5650 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5652 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5655 static void dev_disable_change(struct inet6_dev *idev)
5657 struct netdev_notifier_info info;
5659 if (!idev || !idev->dev)
5662 netdev_notifier_info_init(&info, idev->dev);
5663 if (idev->cnf.disable_ipv6)
5664 addrconf_notify(NULL, NETDEV_DOWN, &info);
5666 addrconf_notify(NULL, NETDEV_UP, &info);
5669 static void addrconf_disable_change(struct net *net, __s32 newf)
5671 struct net_device *dev;
5672 struct inet6_dev *idev;
5674 for_each_netdev(net, dev) {
5675 idev = __in6_dev_get(dev);
5677 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5678 idev->cnf.disable_ipv6 = newf;
5680 dev_disable_change(idev);
5685 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5690 if (!rtnl_trylock())
5691 return restart_syscall();
5693 net = (struct net *)table->extra2;
5697 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5702 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5703 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5704 addrconf_disable_change(net, newf);
5705 } else if ((!newf) ^ (!old))
5706 dev_disable_change((struct inet6_dev *)table->extra1);
5713 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5714 void __user *buffer, size_t *lenp, loff_t *ppos)
5716 int *valp = ctl->data;
5719 struct ctl_table lctl;
5723 * ctl->data points to idev->cnf.disable_ipv6, we should
5724 * not modify it until we get the rtnl lock.
5729 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5732 ret = addrconf_disable_ipv6(ctl, valp, val);
5739 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5740 void __user *buffer, size_t *lenp, loff_t *ppos)
5742 int *valp = ctl->data;
5747 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5750 if (write && old != new) {
5751 struct net *net = ctl->extra2;
5753 if (!rtnl_trylock())
5754 return restart_syscall();
5756 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5757 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5758 NETCONFA_PROXY_NEIGH,
5759 NETCONFA_IFINDEX_DEFAULT,
5760 net->ipv6.devconf_dflt);
5761 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5762 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5763 NETCONFA_PROXY_NEIGH,
5764 NETCONFA_IFINDEX_ALL,
5765 net->ipv6.devconf_all);
5767 struct inet6_dev *idev = ctl->extra1;
5769 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5770 NETCONFA_PROXY_NEIGH,
5780 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5781 void __user *buffer, size_t *lenp,
5786 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5787 struct net *net = (struct net *)ctl->extra2;
5789 if (!rtnl_trylock())
5790 return restart_syscall();
5792 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5795 new_val = *((int *)ctl->data);
5797 if (check_addr_gen_mode(new_val) < 0) {
5802 /* request for default */
5803 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5804 ipv6_devconf_dflt.addr_gen_mode = new_val;
5806 /* request for individual net device */
5811 if (check_stable_privacy(idev, net, new_val) < 0) {
5816 if (idev->cnf.addr_gen_mode != new_val) {
5817 idev->cnf.addr_gen_mode = new_val;
5818 addrconf_dev_config(idev->dev);
5829 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5830 void __user *buffer, size_t *lenp,
5834 struct in6_addr addr;
5835 char str[IPV6_MAX_STRLEN];
5836 struct ctl_table lctl = *ctl;
5837 struct net *net = ctl->extra2;
5838 struct ipv6_stable_secret *secret = ctl->data;
5840 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5843 lctl.maxlen = IPV6_MAX_STRLEN;
5846 if (!rtnl_trylock())
5847 return restart_syscall();
5849 if (!write && !secret->initialized) {
5854 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5855 if (err >= sizeof(str)) {
5860 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5864 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5869 secret->initialized = true;
5870 secret->secret = addr;
5872 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5873 struct net_device *dev;
5875 for_each_netdev(net, dev) {
5876 struct inet6_dev *idev = __in6_dev_get(dev);
5879 idev->cnf.addr_gen_mode =
5880 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5884 struct inet6_dev *idev = ctl->extra1;
5886 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5896 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5898 void __user *buffer,
5902 int *valp = ctl->data;
5905 struct ctl_table lctl;
5908 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5909 * we should not modify it until we get the rtnl lock.
5914 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5917 ret = addrconf_fixup_linkdown(ctl, valp, val);
5924 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5928 rt->dst.flags |= DST_NOPOLICY;
5930 rt->dst.flags &= ~DST_NOPOLICY;
5935 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5937 struct inet6_ifaddr *ifa;
5939 read_lock_bh(&idev->lock);
5940 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5941 spin_lock(&ifa->lock);
5943 struct rt6_info *rt = ifa->rt;
5944 struct fib6_table *table = rt->rt6i_table;
5947 read_lock(&table->tb6_lock);
5948 addrconf_set_nopolicy(ifa->rt, val);
5949 if (rt->rt6i_pcpu) {
5950 for_each_possible_cpu(cpu) {
5951 struct rt6_info **rtp;
5953 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5954 addrconf_set_nopolicy(*rtp, val);
5957 read_unlock(&table->tb6_lock);
5959 spin_unlock(&ifa->lock);
5961 read_unlock_bh(&idev->lock);
5965 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5967 struct inet6_dev *idev;
5970 if (!rtnl_trylock())
5971 return restart_syscall();
5975 net = (struct net *)ctl->extra2;
5976 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5981 if (valp == &net->ipv6.devconf_all->disable_policy) {
5982 struct net_device *dev;
5984 for_each_netdev(net, dev) {
5985 idev = __in6_dev_get(dev);
5987 addrconf_disable_policy_idev(idev, val);
5990 idev = (struct inet6_dev *)ctl->extra1;
5991 addrconf_disable_policy_idev(idev, val);
5999 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6000 void __user *buffer, size_t *lenp,
6003 int *valp = ctl->data;
6006 struct ctl_table lctl;
6011 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6013 if (write && (*valp != val))
6014 ret = addrconf_disable_policy(ctl, valp, val);
6022 static int minus_one = -1;
6023 static const int one = 1;
6024 static const int two_five_five = 255;
6026 static const struct ctl_table addrconf_sysctl[] = {
6028 .procname = "forwarding",
6029 .data = &ipv6_devconf.forwarding,
6030 .maxlen = sizeof(int),
6032 .proc_handler = addrconf_sysctl_forward,
6035 .procname = "hop_limit",
6036 .data = &ipv6_devconf.hop_limit,
6037 .maxlen = sizeof(int),
6039 .proc_handler = proc_dointvec_minmax,
6040 .extra1 = (void *)&one,
6041 .extra2 = (void *)&two_five_five,
6045 .data = &ipv6_devconf.mtu6,
6046 .maxlen = sizeof(int),
6048 .proc_handler = addrconf_sysctl_mtu,
6051 .procname = "accept_ra",
6052 .data = &ipv6_devconf.accept_ra,
6053 .maxlen = sizeof(int),
6055 .proc_handler = proc_dointvec,
6058 .procname = "accept_redirects",
6059 .data = &ipv6_devconf.accept_redirects,
6060 .maxlen = sizeof(int),
6062 .proc_handler = proc_dointvec,
6065 .procname = "autoconf",
6066 .data = &ipv6_devconf.autoconf,
6067 .maxlen = sizeof(int),
6069 .proc_handler = proc_dointvec,
6072 .procname = "dad_transmits",
6073 .data = &ipv6_devconf.dad_transmits,
6074 .maxlen = sizeof(int),
6076 .proc_handler = proc_dointvec,
6079 .procname = "router_solicitations",
6080 .data = &ipv6_devconf.rtr_solicits,
6081 .maxlen = sizeof(int),
6083 .proc_handler = proc_dointvec_minmax,
6084 .extra1 = &minus_one,
6087 .procname = "router_solicitation_interval",
6088 .data = &ipv6_devconf.rtr_solicit_interval,
6089 .maxlen = sizeof(int),
6091 .proc_handler = proc_dointvec_jiffies,
6094 .procname = "router_solicitation_max_interval",
6095 .data = &ipv6_devconf.rtr_solicit_max_interval,
6096 .maxlen = sizeof(int),
6098 .proc_handler = proc_dointvec_jiffies,
6101 .procname = "router_solicitation_delay",
6102 .data = &ipv6_devconf.rtr_solicit_delay,
6103 .maxlen = sizeof(int),
6105 .proc_handler = proc_dointvec_jiffies,
6108 .procname = "force_mld_version",
6109 .data = &ipv6_devconf.force_mld_version,
6110 .maxlen = sizeof(int),
6112 .proc_handler = proc_dointvec,
6115 .procname = "mldv1_unsolicited_report_interval",
6117 &ipv6_devconf.mldv1_unsolicited_report_interval,
6118 .maxlen = sizeof(int),
6120 .proc_handler = proc_dointvec_ms_jiffies,
6123 .procname = "mldv2_unsolicited_report_interval",
6125 &ipv6_devconf.mldv2_unsolicited_report_interval,
6126 .maxlen = sizeof(int),
6128 .proc_handler = proc_dointvec_ms_jiffies,
6131 .procname = "use_tempaddr",
6132 .data = &ipv6_devconf.use_tempaddr,
6133 .maxlen = sizeof(int),
6135 .proc_handler = proc_dointvec,
6138 .procname = "temp_valid_lft",
6139 .data = &ipv6_devconf.temp_valid_lft,
6140 .maxlen = sizeof(int),
6142 .proc_handler = proc_dointvec,
6145 .procname = "temp_prefered_lft",
6146 .data = &ipv6_devconf.temp_prefered_lft,
6147 .maxlen = sizeof(int),
6149 .proc_handler = proc_dointvec,
6152 .procname = "regen_max_retry",
6153 .data = &ipv6_devconf.regen_max_retry,
6154 .maxlen = sizeof(int),
6156 .proc_handler = proc_dointvec,
6159 .procname = "max_desync_factor",
6160 .data = &ipv6_devconf.max_desync_factor,
6161 .maxlen = sizeof(int),
6163 .proc_handler = proc_dointvec,
6166 .procname = "max_addresses",
6167 .data = &ipv6_devconf.max_addresses,
6168 .maxlen = sizeof(int),
6170 .proc_handler = proc_dointvec,
6173 .procname = "accept_ra_defrtr",
6174 .data = &ipv6_devconf.accept_ra_defrtr,
6175 .maxlen = sizeof(int),
6177 .proc_handler = proc_dointvec,
6180 .procname = "accept_ra_min_hop_limit",
6181 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6182 .maxlen = sizeof(int),
6184 .proc_handler = proc_dointvec,
6187 .procname = "accept_ra_pinfo",
6188 .data = &ipv6_devconf.accept_ra_pinfo,
6189 .maxlen = sizeof(int),
6191 .proc_handler = proc_dointvec,
6193 #ifdef CONFIG_IPV6_ROUTER_PREF
6195 .procname = "accept_ra_rtr_pref",
6196 .data = &ipv6_devconf.accept_ra_rtr_pref,
6197 .maxlen = sizeof(int),
6199 .proc_handler = proc_dointvec,
6202 .procname = "router_probe_interval",
6203 .data = &ipv6_devconf.rtr_probe_interval,
6204 .maxlen = sizeof(int),
6206 .proc_handler = proc_dointvec_jiffies,
6208 #ifdef CONFIG_IPV6_ROUTE_INFO
6210 .procname = "accept_ra_rt_info_min_plen",
6211 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6212 .maxlen = sizeof(int),
6214 .proc_handler = proc_dointvec,
6217 .procname = "accept_ra_rt_info_max_plen",
6218 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6219 .maxlen = sizeof(int),
6221 .proc_handler = proc_dointvec,
6226 .procname = "proxy_ndp",
6227 .data = &ipv6_devconf.proxy_ndp,
6228 .maxlen = sizeof(int),
6230 .proc_handler = addrconf_sysctl_proxy_ndp,
6233 .procname = "accept_source_route",
6234 .data = &ipv6_devconf.accept_source_route,
6235 .maxlen = sizeof(int),
6237 .proc_handler = proc_dointvec,
6239 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6241 .procname = "optimistic_dad",
6242 .data = &ipv6_devconf.optimistic_dad,
6243 .maxlen = sizeof(int),
6245 .proc_handler = proc_dointvec,
6248 .procname = "use_optimistic",
6249 .data = &ipv6_devconf.use_optimistic,
6250 .maxlen = sizeof(int),
6252 .proc_handler = proc_dointvec,
6255 #ifdef CONFIG_IPV6_MROUTE
6257 .procname = "mc_forwarding",
6258 .data = &ipv6_devconf.mc_forwarding,
6259 .maxlen = sizeof(int),
6261 .proc_handler = proc_dointvec,
6265 .procname = "disable_ipv6",
6266 .data = &ipv6_devconf.disable_ipv6,
6267 .maxlen = sizeof(int),
6269 .proc_handler = addrconf_sysctl_disable,
6272 .procname = "accept_dad",
6273 .data = &ipv6_devconf.accept_dad,
6274 .maxlen = sizeof(int),
6276 .proc_handler = proc_dointvec,
6279 .procname = "force_tllao",
6280 .data = &ipv6_devconf.force_tllao,
6281 .maxlen = sizeof(int),
6283 .proc_handler = proc_dointvec
6286 .procname = "ndisc_notify",
6287 .data = &ipv6_devconf.ndisc_notify,
6288 .maxlen = sizeof(int),
6290 .proc_handler = proc_dointvec
6293 .procname = "suppress_frag_ndisc",
6294 .data = &ipv6_devconf.suppress_frag_ndisc,
6295 .maxlen = sizeof(int),
6297 .proc_handler = proc_dointvec
6300 .procname = "accept_ra_from_local",
6301 .data = &ipv6_devconf.accept_ra_from_local,
6302 .maxlen = sizeof(int),
6304 .proc_handler = proc_dointvec,
6307 .procname = "accept_ra_mtu",
6308 .data = &ipv6_devconf.accept_ra_mtu,
6309 .maxlen = sizeof(int),
6311 .proc_handler = proc_dointvec,
6314 .procname = "stable_secret",
6315 .data = &ipv6_devconf.stable_secret,
6316 .maxlen = IPV6_MAX_STRLEN,
6318 .proc_handler = addrconf_sysctl_stable_secret,
6321 .procname = "use_oif_addrs_only",
6322 .data = &ipv6_devconf.use_oif_addrs_only,
6323 .maxlen = sizeof(int),
6325 .proc_handler = proc_dointvec,
6328 .procname = "ignore_routes_with_linkdown",
6329 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6330 .maxlen = sizeof(int),
6332 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6335 .procname = "drop_unicast_in_l2_multicast",
6336 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6337 .maxlen = sizeof(int),
6339 .proc_handler = proc_dointvec,
6342 .procname = "drop_unsolicited_na",
6343 .data = &ipv6_devconf.drop_unsolicited_na,
6344 .maxlen = sizeof(int),
6346 .proc_handler = proc_dointvec,
6349 .procname = "keep_addr_on_down",
6350 .data = &ipv6_devconf.keep_addr_on_down,
6351 .maxlen = sizeof(int),
6353 .proc_handler = proc_dointvec,
6357 .procname = "seg6_enabled",
6358 .data = &ipv6_devconf.seg6_enabled,
6359 .maxlen = sizeof(int),
6361 .proc_handler = proc_dointvec,
6363 #ifdef CONFIG_IPV6_SEG6_HMAC
6365 .procname = "seg6_require_hmac",
6366 .data = &ipv6_devconf.seg6_require_hmac,
6367 .maxlen = sizeof(int),
6369 .proc_handler = proc_dointvec,
6373 .procname = "enhanced_dad",
6374 .data = &ipv6_devconf.enhanced_dad,
6375 .maxlen = sizeof(int),
6377 .proc_handler = proc_dointvec,
6380 .procname = "addr_gen_mode",
6381 .data = &ipv6_devconf.addr_gen_mode,
6382 .maxlen = sizeof(int),
6384 .proc_handler = addrconf_sysctl_addr_gen_mode,
6387 .procname = "disable_policy",
6388 .data = &ipv6_devconf.disable_policy,
6389 .maxlen = sizeof(int),
6391 .proc_handler = addrconf_sysctl_disable_policy,
6398 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6399 struct inet6_dev *idev, struct ipv6_devconf *p)
6402 struct ctl_table *table;
6403 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6405 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6409 for (i = 0; table[i].data; i++) {
6410 table[i].data += (char *)p - (char *)&ipv6_devconf;
6411 /* If one of these is already set, then it is not safe to
6412 * overwrite either of them: this makes proc_dointvec_minmax
6415 if (!table[i].extra1 && !table[i].extra2) {
6416 table[i].extra1 = idev; /* embedded; no ref */
6417 table[i].extra2 = net;
6421 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6423 p->sysctl_header = register_net_sysctl(net, path, table);
6424 if (!p->sysctl_header)
6427 if (!strcmp(dev_name, "all"))
6428 ifindex = NETCONFA_IFINDEX_ALL;
6429 else if (!strcmp(dev_name, "default"))
6430 ifindex = NETCONFA_IFINDEX_DEFAULT;
6432 ifindex = idev->dev->ifindex;
6433 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6443 static void __addrconf_sysctl_unregister(struct net *net,
6444 struct ipv6_devconf *p, int ifindex)
6446 struct ctl_table *table;
6448 if (!p->sysctl_header)
6451 table = p->sysctl_header->ctl_table_arg;
6452 unregister_net_sysctl_table(p->sysctl_header);
6453 p->sysctl_header = NULL;
6456 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6459 static int addrconf_sysctl_register(struct inet6_dev *idev)
6463 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6466 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6467 &ndisc_ifinfo_sysctl_change);
6470 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6473 neigh_sysctl_unregister(idev->nd_parms);
6478 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6480 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6481 idev->dev->ifindex);
6482 neigh_sysctl_unregister(idev->nd_parms);
6488 static int __net_init addrconf_init_net(struct net *net)
6491 struct ipv6_devconf *all, *dflt;
6493 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6497 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6499 goto err_alloc_dflt;
6501 /* these will be inherited by all namespaces */
6502 dflt->autoconf = ipv6_defaults.autoconf;
6503 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6505 dflt->stable_secret.initialized = false;
6506 all->stable_secret.initialized = false;
6508 net->ipv6.devconf_all = all;
6509 net->ipv6.devconf_dflt = dflt;
6511 #ifdef CONFIG_SYSCTL
6512 err = __addrconf_sysctl_register(net, "all", NULL, all);
6516 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6522 #ifdef CONFIG_SYSCTL
6524 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6534 static void __net_exit addrconf_exit_net(struct net *net)
6536 #ifdef CONFIG_SYSCTL
6537 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6538 NETCONFA_IFINDEX_DEFAULT);
6539 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6540 NETCONFA_IFINDEX_ALL);
6542 kfree(net->ipv6.devconf_dflt);
6543 kfree(net->ipv6.devconf_all);
6546 static struct pernet_operations addrconf_ops = {
6547 .init = addrconf_init_net,
6548 .exit = addrconf_exit_net,
6551 static struct rtnl_af_ops inet6_ops __read_mostly = {
6553 .fill_link_af = inet6_fill_link_af,
6554 .get_link_af_size = inet6_get_link_af_size,
6555 .validate_link_af = inet6_validate_link_af,
6556 .set_link_af = inet6_set_link_af,
6560 * Init / cleanup code
6563 int __init addrconf_init(void)
6565 struct inet6_dev *idev;
6568 err = ipv6_addr_label_init();
6570 pr_crit("%s: cannot initialize default policy table: %d\n",
6575 err = register_pernet_subsys(&addrconf_ops);
6579 addrconf_wq = create_workqueue("ipv6_addrconf");
6585 /* The addrconf netdev notifier requires that loopback_dev
6586 * has it's ipv6 private information allocated and setup
6587 * before it can bring up and give link-local addresses
6588 * to other devices which are up.
6590 * Unfortunately, loopback_dev is not necessarily the first
6591 * entry in the global dev_base list of net devices. In fact,
6592 * it is likely to be the very last entry on that list.
6593 * So this causes the notifier registry below to try and
6594 * give link-local addresses to all devices besides loopback_dev
6595 * first, then loopback_dev, which cases all the non-loopback_dev
6596 * devices to fail to get a link-local address.
6598 * So, as a temporary fix, allocate the ipv6 structure for
6599 * loopback_dev first by hand.
6600 * Longer term, all of the dependencies ipv6 has upon the loopback
6601 * device and it being up should be removed.
6604 idev = ipv6_add_dev(init_net.loopback_dev);
6607 err = PTR_ERR(idev);
6611 ip6_route_init_special_entries();
6613 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6614 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6616 register_netdevice_notifier(&ipv6_dev_notf);
6620 rtnl_af_register(&inet6_ops);
6622 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6627 /* Only the first call to __rtnl_register can fail */
6628 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, 0);
6629 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, 0);
6630 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6631 inet6_dump_ifaddr, 0);
6632 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6633 inet6_dump_ifmcaddr, 0);
6634 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6635 inet6_dump_ifacaddr, 0);
6636 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6637 inet6_netconf_dump_devconf, 0);
6639 ipv6_addr_label_rtnl_register();
6643 rtnl_af_unregister(&inet6_ops);
6644 unregister_netdevice_notifier(&ipv6_dev_notf);
6646 destroy_workqueue(addrconf_wq);
6648 unregister_pernet_subsys(&addrconf_ops);
6650 ipv6_addr_label_cleanup();
6655 void addrconf_cleanup(void)
6657 struct net_device *dev;
6660 unregister_netdevice_notifier(&ipv6_dev_notf);
6661 unregister_pernet_subsys(&addrconf_ops);
6662 ipv6_addr_label_cleanup();
6666 __rtnl_af_unregister(&inet6_ops);
6668 /* clean dev list */
6669 for_each_netdev(&init_net, dev) {
6670 if (__in6_dev_get(dev) == NULL)
6672 addrconf_ifdown(dev, 1);
6674 addrconf_ifdown(init_net.loopback_dev, 2);
6679 spin_lock_bh(&addrconf_hash_lock);
6680 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6681 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6682 spin_unlock_bh(&addrconf_hash_lock);
6683 cancel_delayed_work(&addr_chk_work);
6686 destroy_workqueue(addrconf_wq);