1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IPv6 Address [auto]configuration
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
14 * Janos Farkas : delete timer on ifdown
15 * <chexum@bankinf.banki.hu>
16 * Andi Kleen : kill double kfree on module
18 * Maciej W. Rozycki : FDDI support
19 * sekiya@USAGI : Don't send too many RS
21 * yoshfuji@USAGI : Fixed interval between DAD
23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
24 * address validation timer.
25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
28 * address on a same interface.
29 * YOSHIFUJI Hideaki @USAGI : ARCnet support
30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
32 * YOSHIFUJI Hideaki @USAGI : improved source address
33 * selection; consider scope,
37 #define pr_fmt(fmt) "IPv6: " fmt
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
58 #include <linux/sysctl.h>
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
66 #include <net/net_namespace.h>
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
93 #define INFINITY_LIFE_TIME 0xFFFFFFFF
95 #define IPV6_MAX_STRLEN \
96 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98 static inline u32 cstamp_delta(unsigned long cstamp)
100 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
106 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
107 do_div(tmp, 1000000);
111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
114 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
115 do_div(tmp, 1000000);
116 if ((s32)tmp > mrt) {
117 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
118 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
119 do_div(tmp, 1000000);
125 static int addrconf_sysctl_register(struct inet6_dev *idev);
126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
138 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
141 static int ipv6_count_addresses(const struct inet6_dev *idev);
142 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 const struct inet6_dev *idev);
146 #define IN6_ADDR_HSIZE_SHIFT 8
147 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
149 * Configured unicast address hash table
151 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
152 static DEFINE_SPINLOCK(addrconf_hash_lock);
154 static void addrconf_verify(void);
155 static void addrconf_verify_rtnl(void);
156 static void addrconf_verify_work(struct work_struct *);
158 static struct workqueue_struct *addrconf_wq;
159 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
161 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
162 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
164 static void addrconf_type_change(struct net_device *dev,
165 unsigned long event);
166 static int addrconf_ifdown(struct net_device *dev, bool unregister);
168 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
170 const struct net_device *dev,
171 u32 flags, u32 noflags,
174 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
175 static void addrconf_dad_work(struct work_struct *w);
176 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
178 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
179 static void addrconf_rs_timer(struct timer_list *t);
180 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
181 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
183 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
184 struct prefix_info *pinfo);
186 static struct ipv6_devconf ipv6_devconf __read_mostly = {
188 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
189 .mtu6 = IPV6_MIN_MTU,
191 .accept_redirects = 1,
193 .force_mld_version = 0,
194 .mldv1_unsolicited_report_interval = 10 * HZ,
195 .mldv2_unsolicited_report_interval = HZ,
197 .rtr_solicits = MAX_RTR_SOLICITATIONS,
198 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
199 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
200 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
202 .temp_valid_lft = TEMP_VALID_LIFETIME,
203 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
204 .regen_max_retry = REGEN_MAX_RETRY,
205 .max_desync_factor = MAX_DESYNC_FACTOR,
206 .max_addresses = IPV6_MAX_ADDRESSES,
207 .accept_ra_defrtr = 1,
208 .accept_ra_from_local = 0,
209 .accept_ra_min_hop_limit= 1,
210 .accept_ra_pinfo = 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212 .accept_ra_rtr_pref = 1,
213 .rtr_probe_interval = 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215 .accept_ra_rt_info_min_plen = 0,
216 .accept_ra_rt_info_max_plen = 0,
220 .accept_source_route = 0, /* we do not accept RH0 by default. */
223 .suppress_frag_ndisc = 1,
226 .initialized = false,
228 .use_oif_addrs_only = 0,
229 .ignore_routes_with_linkdown = 0,
230 .keep_addr_on_down = 0,
232 #ifdef CONFIG_IPV6_SEG6_HMAC
233 .seg6_require_hmac = 0,
236 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
238 .rpl_seg_enabled = 0,
241 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
243 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
244 .mtu6 = IPV6_MIN_MTU,
246 .accept_redirects = 1,
248 .force_mld_version = 0,
249 .mldv1_unsolicited_report_interval = 10 * HZ,
250 .mldv2_unsolicited_report_interval = HZ,
252 .rtr_solicits = MAX_RTR_SOLICITATIONS,
253 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
254 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
255 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
257 .temp_valid_lft = TEMP_VALID_LIFETIME,
258 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
259 .regen_max_retry = REGEN_MAX_RETRY,
260 .max_desync_factor = MAX_DESYNC_FACTOR,
261 .max_addresses = IPV6_MAX_ADDRESSES,
262 .accept_ra_defrtr = 1,
263 .accept_ra_from_local = 0,
264 .accept_ra_min_hop_limit= 1,
265 .accept_ra_pinfo = 1,
266 #ifdef CONFIG_IPV6_ROUTER_PREF
267 .accept_ra_rtr_pref = 1,
268 .rtr_probe_interval = 60 * HZ,
269 #ifdef CONFIG_IPV6_ROUTE_INFO
270 .accept_ra_rt_info_min_plen = 0,
271 .accept_ra_rt_info_max_plen = 0,
275 .accept_source_route = 0, /* we do not accept RH0 by default. */
278 .suppress_frag_ndisc = 1,
281 .initialized = false,
283 .use_oif_addrs_only = 0,
284 .ignore_routes_with_linkdown = 0,
285 .keep_addr_on_down = 0,
287 #ifdef CONFIG_IPV6_SEG6_HMAC
288 .seg6_require_hmac = 0,
291 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
293 .rpl_seg_enabled = 0,
296 /* Check if link is ready: is it up and is a valid qdisc available */
297 static inline bool addrconf_link_ready(const struct net_device *dev)
299 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
302 static void addrconf_del_rs_timer(struct inet6_dev *idev)
304 if (del_timer(&idev->rs_timer))
308 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
310 if (cancel_delayed_work(&ifp->dad_work))
314 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
317 if (!timer_pending(&idev->rs_timer))
319 mod_timer(&idev->rs_timer, jiffies + when);
322 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
326 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
330 static int snmp6_alloc_dev(struct inet6_dev *idev)
334 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
335 if (!idev->stats.ipv6)
338 for_each_possible_cpu(i) {
339 struct ipstats_mib *addrconf_stats;
340 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
341 u64_stats_init(&addrconf_stats->syncp);
345 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
347 if (!idev->stats.icmpv6dev)
349 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
351 if (!idev->stats.icmpv6msgdev)
357 kfree(idev->stats.icmpv6dev);
359 free_percpu(idev->stats.ipv6);
364 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
366 struct inet6_dev *ndev;
371 if (dev->mtu < IPV6_MIN_MTU)
372 return ERR_PTR(-EINVAL);
374 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
378 rwlock_init(&ndev->lock);
380 INIT_LIST_HEAD(&ndev->addr_list);
381 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
382 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
384 if (ndev->cnf.stable_secret.initialized)
385 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
387 ndev->cnf.mtu6 = dev->mtu;
388 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
389 if (!ndev->nd_parms) {
393 if (ndev->cnf.forwarding)
394 dev_disable_lro(dev);
395 /* We refer to the device */
398 if (snmp6_alloc_dev(ndev) < 0) {
399 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
401 neigh_parms_release(&nd_tbl, ndev->nd_parms);
407 if (snmp6_register_dev(ndev) < 0) {
408 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
409 __func__, dev->name);
413 /* One reference from device. */
414 refcount_set(&ndev->refcnt, 1);
416 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
417 ndev->cnf.accept_dad = -1;
419 #if IS_ENABLED(CONFIG_IPV6_SIT)
420 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
421 pr_info("%s: Disabled Multicast RS\n", dev->name);
422 ndev->cnf.rtr_solicits = 0;
426 INIT_LIST_HEAD(&ndev->tempaddr_list);
427 ndev->desync_factor = U32_MAX;
428 if ((dev->flags&IFF_LOOPBACK) ||
429 dev->type == ARPHRD_TUNNEL ||
430 dev->type == ARPHRD_TUNNEL6 ||
431 dev->type == ARPHRD_SIT ||
432 dev->type == ARPHRD_NONE) {
433 ndev->cnf.use_tempaddr = -1;
436 ndev->token = in6addr_any;
438 if (netif_running(dev) && addrconf_link_ready(dev))
439 ndev->if_flags |= IF_READY;
441 ipv6_mc_init_dev(ndev);
442 ndev->tstamp = jiffies;
443 err = addrconf_sysctl_register(ndev);
445 ipv6_mc_destroy_dev(ndev);
446 snmp6_unregister_dev(ndev);
449 /* protected by rtnl_lock */
450 rcu_assign_pointer(dev->ip6_ptr, ndev);
452 /* Join interface-local all-node multicast group */
453 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
455 /* Join all-node multicast group */
456 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
458 /* Join all-router multicast group if forwarding is set */
459 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
460 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
465 neigh_parms_release(&nd_tbl, ndev->nd_parms);
467 in6_dev_finish_destroy(ndev);
471 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
473 struct inet6_dev *idev;
477 idev = __in6_dev_get(dev);
479 idev = ipv6_add_dev(dev);
484 if (dev->flags&IFF_UP)
489 static int inet6_netconf_msgsize_devconf(int type)
491 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
492 + nla_total_size(4); /* NETCONFA_IFINDEX */
495 if (type == NETCONFA_ALL)
498 if (all || type == NETCONFA_FORWARDING)
499 size += nla_total_size(4);
500 #ifdef CONFIG_IPV6_MROUTE
501 if (all || type == NETCONFA_MC_FORWARDING)
502 size += nla_total_size(4);
504 if (all || type == NETCONFA_PROXY_NEIGH)
505 size += nla_total_size(4);
507 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
508 size += nla_total_size(4);
513 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
514 struct ipv6_devconf *devconf, u32 portid,
515 u32 seq, int event, unsigned int flags,
518 struct nlmsghdr *nlh;
519 struct netconfmsg *ncm;
522 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
527 if (type == NETCONFA_ALL)
530 ncm = nlmsg_data(nlh);
531 ncm->ncm_family = AF_INET6;
533 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
534 goto nla_put_failure;
539 if ((all || type == NETCONFA_FORWARDING) &&
540 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
541 goto nla_put_failure;
542 #ifdef CONFIG_IPV6_MROUTE
543 if ((all || type == NETCONFA_MC_FORWARDING) &&
544 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
545 devconf->mc_forwarding) < 0)
546 goto nla_put_failure;
548 if ((all || type == NETCONFA_PROXY_NEIGH) &&
549 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
550 goto nla_put_failure;
552 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
553 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
554 devconf->ignore_routes_with_linkdown) < 0)
555 goto nla_put_failure;
562 nlmsg_cancel(skb, nlh);
566 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
567 int ifindex, struct ipv6_devconf *devconf)
572 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
576 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
579 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
580 WARN_ON(err == -EMSGSIZE);
584 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
587 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
590 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
591 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
592 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
593 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
594 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
597 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
598 const struct nlmsghdr *nlh,
600 struct netlink_ext_ack *extack)
604 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
605 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
609 if (!netlink_strict_get_check(skb))
610 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
612 devconf_ipv6_policy, extack);
614 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
616 devconf_ipv6_policy, extack);
620 for (i = 0; i <= NETCONFA_MAX; i++) {
625 case NETCONFA_IFINDEX:
628 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
636 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
637 struct nlmsghdr *nlh,
638 struct netlink_ext_ack *extack)
640 struct net *net = sock_net(in_skb->sk);
641 struct nlattr *tb[NETCONFA_MAX+1];
642 struct inet6_dev *in6_dev = NULL;
643 struct net_device *dev = NULL;
645 struct ipv6_devconf *devconf;
649 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
653 if (!tb[NETCONFA_IFINDEX])
657 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
659 case NETCONFA_IFINDEX_ALL:
660 devconf = net->ipv6.devconf_all;
662 case NETCONFA_IFINDEX_DEFAULT:
663 devconf = net->ipv6.devconf_dflt;
666 dev = dev_get_by_index(net, ifindex);
669 in6_dev = in6_dev_get(dev);
672 devconf = &in6_dev->cnf;
677 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
681 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
682 NETLINK_CB(in_skb).portid,
683 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
686 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
687 WARN_ON(err == -EMSGSIZE);
691 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
694 in6_dev_put(in6_dev);
700 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
701 struct netlink_callback *cb)
703 const struct nlmsghdr *nlh = cb->nlh;
704 struct net *net = sock_net(skb->sk);
707 struct net_device *dev;
708 struct inet6_dev *idev;
709 struct hlist_head *head;
711 if (cb->strict_check) {
712 struct netlink_ext_ack *extack = cb->extack;
713 struct netconfmsg *ncm;
715 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
716 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
720 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
721 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
727 s_idx = idx = cb->args[1];
729 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
731 head = &net->dev_index_head[h];
733 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
735 hlist_for_each_entry_rcu(dev, head, index_hlist) {
738 idev = __in6_dev_get(dev);
742 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
744 NETLINK_CB(cb->skb).portid,
752 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
758 if (h == NETDEV_HASHENTRIES) {
759 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
760 net->ipv6.devconf_all,
761 NETLINK_CB(cb->skb).portid,
763 RTM_NEWNETCONF, NLM_F_MULTI,
769 if (h == NETDEV_HASHENTRIES + 1) {
770 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
771 net->ipv6.devconf_dflt,
772 NETLINK_CB(cb->skb).portid,
774 RTM_NEWNETCONF, NLM_F_MULTI,
788 static void dev_forward_change(struct inet6_dev *idev)
790 struct net_device *dev;
791 struct inet6_ifaddr *ifa;
796 if (idev->cnf.forwarding)
797 dev_disable_lro(dev);
798 if (dev->flags & IFF_MULTICAST) {
799 if (idev->cnf.forwarding) {
800 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
801 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
802 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
804 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
805 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
806 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
810 list_for_each_entry(ifa, &idev->addr_list, if_list) {
811 if (ifa->flags&IFA_F_TENTATIVE)
813 if (idev->cnf.forwarding)
814 addrconf_join_anycast(ifa);
816 addrconf_leave_anycast(ifa);
818 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
820 dev->ifindex, &idev->cnf);
824 static void addrconf_forward_change(struct net *net, __s32 newf)
826 struct net_device *dev;
827 struct inet6_dev *idev;
829 for_each_netdev(net, dev) {
830 idev = __in6_dev_get(dev);
832 int changed = (!idev->cnf.forwarding) ^ (!newf);
833 idev->cnf.forwarding = newf;
835 dev_forward_change(idev);
840 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
846 return restart_syscall();
848 net = (struct net *)table->extra2;
852 if (p == &net->ipv6.devconf_dflt->forwarding) {
853 if ((!newf) ^ (!old))
854 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
856 NETCONFA_IFINDEX_DEFAULT,
857 net->ipv6.devconf_dflt);
862 if (p == &net->ipv6.devconf_all->forwarding) {
863 int old_dflt = net->ipv6.devconf_dflt->forwarding;
865 net->ipv6.devconf_dflt->forwarding = newf;
866 if ((!newf) ^ (!old_dflt))
867 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
869 NETCONFA_IFINDEX_DEFAULT,
870 net->ipv6.devconf_dflt);
872 addrconf_forward_change(net, newf);
873 if ((!newf) ^ (!old))
874 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
876 NETCONFA_IFINDEX_ALL,
877 net->ipv6.devconf_all);
878 } else if ((!newf) ^ (!old))
879 dev_forward_change((struct inet6_dev *)table->extra1);
883 rt6_purge_dflt_routers(net);
887 static void addrconf_linkdown_change(struct net *net, __s32 newf)
889 struct net_device *dev;
890 struct inet6_dev *idev;
892 for_each_netdev(net, dev) {
893 idev = __in6_dev_get(dev);
895 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
897 idev->cnf.ignore_routes_with_linkdown = newf;
899 inet6_netconf_notify_devconf(dev_net(dev),
901 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
908 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
914 return restart_syscall();
916 net = (struct net *)table->extra2;
920 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
921 if ((!newf) ^ (!old))
922 inet6_netconf_notify_devconf(net,
924 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
925 NETCONFA_IFINDEX_DEFAULT,
926 net->ipv6.devconf_dflt);
931 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
932 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
933 addrconf_linkdown_change(net, newf);
934 if ((!newf) ^ (!old))
935 inet6_netconf_notify_devconf(net,
937 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
938 NETCONFA_IFINDEX_ALL,
939 net->ipv6.devconf_all);
948 /* Nobody refers to this ifaddr, destroy it */
949 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
951 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
953 #ifdef NET_REFCNT_DEBUG
954 pr_debug("%s\n", __func__);
957 in6_dev_put(ifp->idev);
959 if (cancel_delayed_work(&ifp->dad_work))
960 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
963 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
964 pr_warn("Freeing alive inet6 address %p\n", ifp);
972 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
975 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
978 * Each device address list is sorted in order of scope -
979 * global before linklocal.
981 list_for_each(p, &idev->addr_list) {
982 struct inet6_ifaddr *ifa
983 = list_entry(p, struct inet6_ifaddr, if_list);
984 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
988 list_add_tail_rcu(&ifp->if_list, p);
991 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
993 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
995 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
998 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
999 struct net_device *dev, unsigned int hash)
1001 struct inet6_ifaddr *ifp;
1003 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1004 if (!net_eq(dev_net(ifp->idev->dev), net))
1006 if (ipv6_addr_equal(&ifp->addr, addr)) {
1007 if (!dev || ifp->idev->dev == dev)
1014 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1016 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1019 spin_lock(&addrconf_hash_lock);
1021 /* Ignore adding duplicate addresses on an interface */
1022 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1023 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1026 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1029 spin_unlock(&addrconf_hash_lock);
1034 /* On success it returns ifp with increased reference count */
1036 static struct inet6_ifaddr *
1037 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1038 bool can_block, struct netlink_ext_ack *extack)
1040 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1041 int addr_type = ipv6_addr_type(cfg->pfx);
1042 struct net *net = dev_net(idev->dev);
1043 struct inet6_ifaddr *ifa = NULL;
1044 struct fib6_info *f6i = NULL;
1047 if (addr_type == IPV6_ADDR_ANY ||
1048 (addr_type & IPV6_ADDR_MULTICAST &&
1049 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1050 (!(idev->dev->flags & IFF_LOOPBACK) &&
1051 !netif_is_l3_master(idev->dev) &&
1052 addr_type & IPV6_ADDR_LOOPBACK))
1053 return ERR_PTR(-EADDRNOTAVAIL);
1056 err = -ENODEV; /*XXX*/
1060 if (idev->cnf.disable_ipv6) {
1065 /* validator notifier needs to be blocking;
1066 * do not call in atomic context
1069 struct in6_validator_info i6vi = {
1070 .i6vi_addr = *cfg->pfx,
1075 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1076 err = notifier_to_errno(err);
1081 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1087 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1094 if (net->ipv6.devconf_all->disable_policy ||
1095 idev->cnf.disable_policy)
1096 f6i->dst_nopolicy = true;
1098 neigh_parms_data_state_setall(idev->nd_parms);
1100 ifa->addr = *cfg->pfx;
1102 ifa->peer_addr = *cfg->peer_pfx;
1104 spin_lock_init(&ifa->lock);
1105 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1106 INIT_HLIST_NODE(&ifa->addr_lst);
1107 ifa->scope = cfg->scope;
1108 ifa->prefix_len = cfg->plen;
1109 ifa->rt_priority = cfg->rt_priority;
1110 ifa->flags = cfg->ifa_flags;
1111 /* No need to add the TENTATIVE flag for addresses with NODAD */
1112 if (!(cfg->ifa_flags & IFA_F_NODAD))
1113 ifa->flags |= IFA_F_TENTATIVE;
1114 ifa->valid_lft = cfg->valid_lft;
1115 ifa->prefered_lft = cfg->preferred_lft;
1116 ifa->cstamp = ifa->tstamp = jiffies;
1117 ifa->tokenized = false;
1125 refcount_set(&ifa->refcnt, 1);
1129 err = ipv6_add_addr_hash(idev->dev, ifa);
1131 rcu_read_unlock_bh();
1135 write_lock(&idev->lock);
1137 /* Add to inet6_dev unicast addr list. */
1138 ipv6_link_dev_addr(idev, ifa);
1140 if (ifa->flags&IFA_F_TEMPORARY) {
1141 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1146 write_unlock(&idev->lock);
1148 rcu_read_unlock_bh();
1150 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1152 if (unlikely(err < 0)) {
1153 fib6_info_release(f6i);
1157 in6_dev_put(ifa->idev);
1166 enum cleanup_prefix_rt_t {
1167 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1168 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1169 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1173 * Check, whether the prefix for ifp would still need a prefix route
1174 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1177 * 1) we don't purge prefix if address was not permanent.
1178 * prefix is managed by its own lifetime.
1179 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1180 * 3) if there are no addresses, delete prefix.
1181 * 4) if there are still other permanent address(es),
1182 * corresponding prefix is still permanent.
1183 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1184 * don't purge the prefix, assume user space is managing it.
1185 * 6) otherwise, update prefix lifetime to the
1186 * longest valid lifetime among the corresponding
1187 * addresses on the device.
1188 * Note: subsequent RA will update lifetime.
1190 static enum cleanup_prefix_rt_t
1191 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1193 struct inet6_ifaddr *ifa;
1194 struct inet6_dev *idev = ifp->idev;
1195 unsigned long lifetime;
1196 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1200 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1203 if (ifa->prefix_len != ifp->prefix_len ||
1204 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1207 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1208 return CLEANUP_PREFIX_RT_NOP;
1210 action = CLEANUP_PREFIX_RT_EXPIRE;
1212 spin_lock(&ifa->lock);
1214 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1216 * Note: Because this address is
1217 * not permanent, lifetime <
1218 * LONG_MAX / HZ here.
1220 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1221 *expires = ifa->tstamp + lifetime * HZ;
1222 spin_unlock(&ifa->lock);
1229 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1230 bool del_rt, bool del_peer)
1232 struct fib6_info *f6i;
1234 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1236 ifp->idev->dev, 0, RTF_DEFAULT, true);
1239 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1241 if (!(f6i->fib6_flags & RTF_EXPIRES))
1242 fib6_set_expires(f6i, expires);
1243 fib6_info_release(f6i);
1249 /* This function wants to get referenced ifp and releases it before return */
1251 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1254 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1255 unsigned long expires;
1259 spin_lock_bh(&ifp->lock);
1261 ifp->state = INET6_IFADDR_STATE_DEAD;
1262 spin_unlock_bh(&ifp->lock);
1264 if (state == INET6_IFADDR_STATE_DEAD)
1267 spin_lock_bh(&addrconf_hash_lock);
1268 hlist_del_init_rcu(&ifp->addr_lst);
1269 spin_unlock_bh(&addrconf_hash_lock);
1271 write_lock_bh(&ifp->idev->lock);
1273 if (ifp->flags&IFA_F_TEMPORARY) {
1274 list_del(&ifp->tmp_list);
1276 in6_ifa_put(ifp->ifpub);
1282 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1283 action = check_cleanup_prefix_route(ifp, &expires);
1285 list_del_rcu(&ifp->if_list);
1288 write_unlock_bh(&ifp->idev->lock);
1290 addrconf_del_dad_work(ifp);
1292 ipv6_ifa_notify(RTM_DELADDR, ifp);
1294 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1296 if (action != CLEANUP_PREFIX_RT_NOP) {
1297 cleanup_prefix_route(ifp, expires,
1298 action == CLEANUP_PREFIX_RT_DEL, false);
1301 /* clean up prefsrc entries */
1302 rt6_remove_prefsrc(ifp);
1307 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1309 struct inet6_dev *idev = ifp->idev;
1310 unsigned long tmp_tstamp, age;
1311 unsigned long regen_advance;
1312 unsigned long now = jiffies;
1313 s32 cnf_temp_preferred_lft;
1314 struct inet6_ifaddr *ift;
1315 struct ifa6_config cfg;
1316 long max_desync_factor;
1317 struct in6_addr addr;
1320 write_lock_bh(&idev->lock);
1324 if (idev->cnf.use_tempaddr <= 0) {
1325 write_unlock_bh(&idev->lock);
1326 pr_info("%s: use_tempaddr is disabled\n", __func__);
1331 spin_lock_bh(&ifp->lock);
1332 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1333 idev->cnf.use_tempaddr = -1; /*XXX*/
1334 spin_unlock_bh(&ifp->lock);
1335 write_unlock_bh(&idev->lock);
1336 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1343 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1344 ipv6_gen_rnd_iid(&addr);
1346 age = (now - ifp->tstamp) / HZ;
1348 regen_advance = idev->cnf.regen_max_retry *
1349 idev->cnf.dad_transmits *
1350 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1352 /* recalculate max_desync_factor each time and update
1353 * idev->desync_factor if it's larger
1355 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1356 max_desync_factor = min_t(__u32,
1357 idev->cnf.max_desync_factor,
1358 cnf_temp_preferred_lft - regen_advance);
1360 if (unlikely(idev->desync_factor > max_desync_factor)) {
1361 if (max_desync_factor > 0) {
1362 get_random_bytes(&idev->desync_factor,
1363 sizeof(idev->desync_factor));
1364 idev->desync_factor %= max_desync_factor;
1366 idev->desync_factor = 0;
1370 memset(&cfg, 0, sizeof(cfg));
1371 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1372 idev->cnf.temp_valid_lft + age);
1373 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1374 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1376 cfg.plen = ifp->prefix_len;
1377 tmp_tstamp = ifp->tstamp;
1378 spin_unlock_bh(&ifp->lock);
1380 write_unlock_bh(&idev->lock);
1382 /* A temporary address is created only if this calculated Preferred
1383 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1384 * an implementation must not create a temporary address with a zero
1385 * Preferred Lifetime.
1386 * Use age calculation as in addrconf_verify to avoid unnecessary
1387 * temporary addresses being generated.
1389 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1390 if (cfg.preferred_lft <= regen_advance + age) {
1397 cfg.ifa_flags = IFA_F_TEMPORARY;
1398 /* set in addrconf_prefix_rcv() */
1399 if (ifp->flags & IFA_F_OPTIMISTIC)
1400 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1403 cfg.scope = ipv6_addr_scope(cfg.pfx);
1405 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1409 pr_info("%s: retry temporary address regeneration\n", __func__);
1410 write_lock_bh(&idev->lock);
1414 spin_lock_bh(&ift->lock);
1417 ift->tstamp = tmp_tstamp;
1418 spin_unlock_bh(&ift->lock);
1420 addrconf_dad_start(ift);
1428 * Choose an appropriate source address (RFC3484)
1431 IPV6_SADDR_RULE_INIT = 0,
1432 IPV6_SADDR_RULE_LOCAL,
1433 IPV6_SADDR_RULE_SCOPE,
1434 IPV6_SADDR_RULE_PREFERRED,
1435 #ifdef CONFIG_IPV6_MIP6
1436 IPV6_SADDR_RULE_HOA,
1438 IPV6_SADDR_RULE_OIF,
1439 IPV6_SADDR_RULE_LABEL,
1440 IPV6_SADDR_RULE_PRIVACY,
1441 IPV6_SADDR_RULE_ORCHID,
1442 IPV6_SADDR_RULE_PREFIX,
1443 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1444 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1449 struct ipv6_saddr_score {
1452 struct inet6_ifaddr *ifa;
1453 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1458 struct ipv6_saddr_dst {
1459 const struct in6_addr *addr;
1466 static inline int ipv6_saddr_preferred(int type)
1468 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1473 static bool ipv6_use_optimistic_addr(struct net *net,
1474 struct inet6_dev *idev)
1476 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1479 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1481 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1490 static bool ipv6_allow_optimistic_dad(struct net *net,
1491 struct inet6_dev *idev)
1493 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1496 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1505 static int ipv6_get_saddr_eval(struct net *net,
1506 struct ipv6_saddr_score *score,
1507 struct ipv6_saddr_dst *dst,
1512 if (i <= score->rule) {
1514 case IPV6_SADDR_RULE_SCOPE:
1515 ret = score->scopedist;
1517 case IPV6_SADDR_RULE_PREFIX:
1518 ret = score->matchlen;
1521 ret = !!test_bit(i, score->scorebits);
1527 case IPV6_SADDR_RULE_INIT:
1528 /* Rule 0: remember if hiscore is not ready yet */
1531 case IPV6_SADDR_RULE_LOCAL:
1532 /* Rule 1: Prefer same address */
1533 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1535 case IPV6_SADDR_RULE_SCOPE:
1536 /* Rule 2: Prefer appropriate scope
1541 * ---+--+-+---> scope
1543 * | d is scope of the destination.
1545 * | \ <- smaller scope is better if
1546 * B-15 | \ if scope is enough for destination.
1547 * | ret = B - scope (-1 <= scope >= d <= 15).
1549 * |/ <- greater is better
1550 * -C / if scope is not enough for destination.
1551 * /| ret = scope - C (-1 <= d < scope <= 15).
1553 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1554 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1555 * Assume B = 0 and we get C > 29.
1557 ret = __ipv6_addr_src_scope(score->addr_type);
1558 if (ret >= dst->scope)
1561 ret -= 128; /* 30 is enough */
1562 score->scopedist = ret;
1564 case IPV6_SADDR_RULE_PREFERRED:
1566 /* Rule 3: Avoid deprecated and optimistic addresses */
1567 u8 avoid = IFA_F_DEPRECATED;
1569 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1570 avoid |= IFA_F_OPTIMISTIC;
1571 ret = ipv6_saddr_preferred(score->addr_type) ||
1572 !(score->ifa->flags & avoid);
1575 #ifdef CONFIG_IPV6_MIP6
1576 case IPV6_SADDR_RULE_HOA:
1578 /* Rule 4: Prefer home address */
1579 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1580 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1584 case IPV6_SADDR_RULE_OIF:
1585 /* Rule 5: Prefer outgoing interface */
1586 ret = (!dst->ifindex ||
1587 dst->ifindex == score->ifa->idev->dev->ifindex);
1589 case IPV6_SADDR_RULE_LABEL:
1590 /* Rule 6: Prefer matching label */
1591 ret = ipv6_addr_label(net,
1592 &score->ifa->addr, score->addr_type,
1593 score->ifa->idev->dev->ifindex) == dst->label;
1595 case IPV6_SADDR_RULE_PRIVACY:
1597 /* Rule 7: Prefer public address
1598 * Note: prefer temporary address if use_tempaddr >= 2
1600 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1601 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1602 score->ifa->idev->cnf.use_tempaddr >= 2;
1603 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1606 case IPV6_SADDR_RULE_ORCHID:
1607 /* Rule 8-: Prefer ORCHID vs ORCHID or
1608 * non-ORCHID vs non-ORCHID
1610 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1611 ipv6_addr_orchid(dst->addr));
1613 case IPV6_SADDR_RULE_PREFIX:
1614 /* Rule 8: Use longest matching prefix */
1615 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1616 if (ret > score->ifa->prefix_len)
1617 ret = score->ifa->prefix_len;
1618 score->matchlen = ret;
1620 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1621 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1622 /* Optimistic addresses still have lower precedence than other
1623 * preferred addresses.
1625 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1633 __set_bit(i, score->scorebits);
1639 static int __ipv6_dev_get_saddr(struct net *net,
1640 struct ipv6_saddr_dst *dst,
1641 struct inet6_dev *idev,
1642 struct ipv6_saddr_score *scores,
1645 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1647 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1651 * - Tentative Address (RFC2462 section 5.4)
1652 * - A tentative address is not considered
1653 * "assigned to an interface" in the traditional
1654 * sense, unless it is also flagged as optimistic.
1655 * - Candidate Source Address (section 4)
1656 * - In any case, anycast addresses, multicast
1657 * addresses, and the unspecified address MUST
1658 * NOT be included in a candidate set.
1660 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1661 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1664 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1666 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1667 score->addr_type & IPV6_ADDR_MULTICAST)) {
1668 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1674 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1676 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1677 int minihiscore, miniscore;
1679 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1680 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1682 if (minihiscore > miniscore) {
1683 if (i == IPV6_SADDR_RULE_SCOPE &&
1684 score->scopedist > 0) {
1687 * each remaining entry
1688 * has too small (not enough)
1689 * scope, because ifa entries
1690 * are sorted by their scope
1696 } else if (minihiscore < miniscore) {
1697 swap(hiscore, score);
1698 hiscore_idx = 1 - hiscore_idx;
1700 /* restore our iterator */
1701 score->ifa = hiscore->ifa;
1711 static int ipv6_get_saddr_master(struct net *net,
1712 const struct net_device *dst_dev,
1713 const struct net_device *master,
1714 struct ipv6_saddr_dst *dst,
1715 struct ipv6_saddr_score *scores,
1718 struct inet6_dev *idev;
1720 idev = __in6_dev_get(dst_dev);
1722 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1723 scores, hiscore_idx);
1725 idev = __in6_dev_get(master);
1727 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1728 scores, hiscore_idx);
1733 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1734 const struct in6_addr *daddr, unsigned int prefs,
1735 struct in6_addr *saddr)
1737 struct ipv6_saddr_score scores[2], *hiscore;
1738 struct ipv6_saddr_dst dst;
1739 struct inet6_dev *idev;
1740 struct net_device *dev;
1742 bool use_oif_addr = false;
1743 int hiscore_idx = 0;
1746 dst_type = __ipv6_addr_type(daddr);
1748 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1749 dst.scope = __ipv6_addr_src_scope(dst_type);
1750 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1753 scores[hiscore_idx].rule = -1;
1754 scores[hiscore_idx].ifa = NULL;
1758 /* Candidate Source Address (section 4)
1759 * - multicast and link-local destination address,
1760 * the set of candidate source address MUST only
1761 * include addresses assigned to interfaces
1762 * belonging to the same link as the outgoing
1764 * (- For site-local destination addresses, the
1765 * set of candidate source addresses MUST only
1766 * include addresses assigned to interfaces
1767 * belonging to the same site as the outgoing
1769 * - "It is RECOMMENDED that the candidate source addresses
1770 * be the set of unicast addresses assigned to the
1771 * interface that will be used to send to the destination
1772 * (the 'outgoing' interface)." (RFC 6724)
1775 idev = __in6_dev_get(dst_dev);
1776 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1777 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1778 (idev && idev->cnf.use_oif_addrs_only)) {
1779 use_oif_addr = true;
1785 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1787 const struct net_device *master;
1790 /* if dst_dev exists and is enslaved to an L3 device, then
1791 * prefer addresses from dst_dev and then the master over
1792 * any other enslaved devices in the L3 domain.
1794 master = l3mdev_master_dev_rcu(dst_dev);
1796 master_idx = master->ifindex;
1798 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1800 scores, hiscore_idx);
1802 if (scores[hiscore_idx].ifa)
1806 for_each_netdev_rcu(net, dev) {
1807 /* only consider addresses on devices in the
1810 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1812 idev = __in6_dev_get(dev);
1815 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1820 hiscore = &scores[hiscore_idx];
1822 ret = -EADDRNOTAVAIL;
1824 *saddr = hiscore->ifa->addr;
1829 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1831 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1834 struct inet6_ifaddr *ifp;
1835 int err = -EADDRNOTAVAIL;
1837 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1838 if (ifp->scope > IFA_LINK)
1840 if (ifp->scope == IFA_LINK &&
1841 !(ifp->flags & banned_flags)) {
1850 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1853 struct inet6_dev *idev;
1854 int err = -EADDRNOTAVAIL;
1857 idev = __in6_dev_get(dev);
1859 read_lock_bh(&idev->lock);
1860 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1861 read_unlock_bh(&idev->lock);
1867 static int ipv6_count_addresses(const struct inet6_dev *idev)
1869 const struct inet6_ifaddr *ifp;
1873 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1879 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1880 const struct net_device *dev, int strict)
1882 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1883 strict, IFA_F_TENTATIVE);
1885 EXPORT_SYMBOL(ipv6_chk_addr);
1887 /* device argument is used to find the L3 domain of interest. If
1888 * skip_dev_check is set, then the ifp device is not checked against
1889 * the passed in dev argument. So the 2 cases for addresses checks are:
1890 * 1. does the address exist in the L3 domain that dev is part of
1891 * (skip_dev_check = true), or
1893 * 2. does the address exist on the specific device
1894 * (skip_dev_check = false)
1896 static struct net_device *
1897 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1898 const struct net_device *dev, bool skip_dev_check,
1899 int strict, u32 banned_flags)
1901 unsigned int hash = inet6_addr_hash(net, addr);
1902 struct net_device *l3mdev, *ndev;
1903 struct inet6_ifaddr *ifp;
1908 l3mdev = l3mdev_master_dev_rcu(dev);
1912 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1913 ndev = ifp->idev->dev;
1914 if (!net_eq(dev_net(ndev), net))
1917 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1920 /* Decouple optimistic from tentative for evaluation here.
1921 * Ban optimistic addresses explicitly, when required.
1923 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1924 ? (ifp->flags&~IFA_F_TENTATIVE)
1926 if (ipv6_addr_equal(&ifp->addr, addr) &&
1927 !(ifp_flags&banned_flags) &&
1928 (!dev || ndev == dev ||
1929 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1939 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1940 const struct net_device *dev, bool skip_dev_check,
1941 int strict, u32 banned_flags)
1943 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1944 strict, banned_flags) ? 1 : 0;
1946 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1949 /* Compares an address/prefix_len with addresses on device @dev.
1950 * If one is found it returns true.
1952 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1953 const unsigned int prefix_len, struct net_device *dev)
1955 const struct inet6_ifaddr *ifa;
1956 const struct inet6_dev *idev;
1960 idev = __in6_dev_get(dev);
1962 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1963 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1972 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1974 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1976 const struct inet6_ifaddr *ifa;
1977 const struct inet6_dev *idev;
1982 idev = __in6_dev_get(dev);
1984 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1985 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1994 EXPORT_SYMBOL(ipv6_chk_prefix);
1997 * ipv6_dev_find - find the first device with a given source address.
1998 * @net: the net namespace
1999 * @addr: the source address
2001 * The caller should be protected by RCU, or RTNL.
2003 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2004 struct net_device *dev)
2006 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2009 EXPORT_SYMBOL(ipv6_dev_find);
2011 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2012 struct net_device *dev, int strict)
2014 unsigned int hash = inet6_addr_hash(net, addr);
2015 struct inet6_ifaddr *ifp, *result = NULL;
2018 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2019 if (!net_eq(dev_net(ifp->idev->dev), net))
2021 if (ipv6_addr_equal(&ifp->addr, addr)) {
2022 if (!dev || ifp->idev->dev == dev ||
2023 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2035 /* Gets referenced address, destroys ifaddr */
2037 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2040 ifp->flags |= IFA_F_DADFAILED;
2042 if (ifp->flags&IFA_F_TEMPORARY) {
2043 struct inet6_ifaddr *ifpub;
2044 spin_lock_bh(&ifp->lock);
2047 in6_ifa_hold(ifpub);
2048 spin_unlock_bh(&ifp->lock);
2049 ipv6_create_tempaddr(ifpub, true);
2052 spin_unlock_bh(&ifp->lock);
2055 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2056 spin_lock_bh(&ifp->lock);
2057 addrconf_del_dad_work(ifp);
2058 ifp->flags |= IFA_F_TENTATIVE;
2060 ifp->flags &= ~IFA_F_OPTIMISTIC;
2061 spin_unlock_bh(&ifp->lock);
2063 ipv6_ifa_notify(0, ifp);
2070 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2074 spin_lock_bh(&ifp->lock);
2075 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2076 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2079 spin_unlock_bh(&ifp->lock);
2084 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2086 struct inet6_dev *idev = ifp->idev;
2087 struct net *net = dev_net(ifp->idev->dev);
2089 if (addrconf_dad_end(ifp)) {
2094 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2095 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2097 spin_lock_bh(&ifp->lock);
2099 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2100 struct in6_addr new_addr;
2101 struct inet6_ifaddr *ifp2;
2102 int retries = ifp->stable_privacy_retry + 1;
2103 struct ifa6_config cfg = {
2105 .plen = ifp->prefix_len,
2106 .ifa_flags = ifp->flags,
2107 .valid_lft = ifp->valid_lft,
2108 .preferred_lft = ifp->prefered_lft,
2109 .scope = ifp->scope,
2112 if (retries > net->ipv6.sysctl.idgen_retries) {
2113 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2114 ifp->idev->dev->name);
2118 new_addr = ifp->addr;
2119 if (ipv6_generate_stable_address(&new_addr, retries,
2123 spin_unlock_bh(&ifp->lock);
2125 if (idev->cnf.max_addresses &&
2126 ipv6_count_addresses(idev) >=
2127 idev->cnf.max_addresses)
2130 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2131 ifp->idev->dev->name);
2133 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2137 spin_lock_bh(&ifp2->lock);
2138 ifp2->stable_privacy_retry = retries;
2139 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2140 spin_unlock_bh(&ifp2->lock);
2142 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2145 spin_lock_bh(&ifp->lock);
2149 /* transition from _POSTDAD to _ERRDAD */
2150 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2151 spin_unlock_bh(&ifp->lock);
2153 addrconf_mod_dad_work(ifp, 0);
2157 /* Join to solicited addr multicast group.
2158 * caller must hold RTNL */
2159 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2161 struct in6_addr maddr;
2163 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2166 addrconf_addr_solict_mult(addr, &maddr);
2167 ipv6_dev_mc_inc(dev, &maddr);
2170 /* caller must hold RTNL */
2171 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2173 struct in6_addr maddr;
2175 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2178 addrconf_addr_solict_mult(addr, &maddr);
2179 __ipv6_dev_mc_dec(idev, &maddr);
2182 /* caller must hold RTNL */
2183 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2185 struct in6_addr addr;
2187 if (ifp->prefix_len >= 127) /* RFC 6164 */
2189 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2190 if (ipv6_addr_any(&addr))
2192 __ipv6_dev_ac_inc(ifp->idev, &addr);
2195 /* caller must hold RTNL */
2196 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2198 struct in6_addr addr;
2200 if (ifp->prefix_len >= 127) /* RFC 6164 */
2202 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2203 if (ipv6_addr_any(&addr))
2205 __ipv6_dev_ac_dec(ifp->idev, &addr);
2208 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2210 switch (dev->addr_len) {
2212 memcpy(eui, dev->dev_addr, 3);
2215 memcpy(eui + 5, dev->dev_addr + 3, 3);
2217 case EUI64_ADDR_LEN:
2218 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2228 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2230 union fwnet_hwaddr *ha;
2232 if (dev->addr_len != FWNET_ALEN)
2235 ha = (union fwnet_hwaddr *)dev->dev_addr;
2237 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2242 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2244 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2245 if (dev->addr_len != ARCNET_ALEN)
2248 eui[7] = *(u8 *)dev->dev_addr;
2252 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2254 if (dev->addr_len != INFINIBAND_ALEN)
2256 memcpy(eui, dev->dev_addr + 12, 8);
2261 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2265 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2266 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2267 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2268 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2269 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2270 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2274 memcpy(eui + 4, &addr, 4);
2278 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2280 if (dev->priv_flags & IFF_ISATAP)
2281 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2285 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2287 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2290 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2292 memcpy(eui, dev->perm_addr, 3);
2293 memcpy(eui + 5, dev->perm_addr + 3, 3);
2300 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2302 switch (dev->type) {
2305 return addrconf_ifid_eui48(eui, dev);
2307 return addrconf_ifid_arcnet(eui, dev);
2308 case ARPHRD_INFINIBAND:
2309 return addrconf_ifid_infiniband(eui, dev);
2311 return addrconf_ifid_sit(eui, dev);
2314 return addrconf_ifid_gre(eui, dev);
2315 case ARPHRD_6LOWPAN:
2316 return addrconf_ifid_6lowpan(eui, dev);
2317 case ARPHRD_IEEE1394:
2318 return addrconf_ifid_ieee1394(eui, dev);
2319 case ARPHRD_TUNNEL6:
2322 return addrconf_ifid_ip6tnl(eui, dev);
2327 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2330 struct inet6_ifaddr *ifp;
2332 read_lock_bh(&idev->lock);
2333 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2334 if (ifp->scope > IFA_LINK)
2336 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2337 memcpy(eui, ifp->addr.s6_addr+8, 8);
2342 read_unlock_bh(&idev->lock);
2346 /* Generation of a randomized Interface Identifier
2347 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2350 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2353 get_random_bytes(&addr->s6_addr[8], 8);
2355 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2356 * check if generated address is not inappropriate:
2358 * - Reserved IPv6 Interface Identifers
2359 * - XXX: already assigned to an address on the device
2362 /* Subnet-router anycast: 0000:0000:0000:0000 */
2363 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2366 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2367 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2368 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2370 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2371 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2374 /* Reserved subnet anycast addresses */
2375 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2376 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2385 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2386 struct net_device *dev, unsigned long expires,
2387 u32 flags, gfp_t gfp_flags)
2389 struct fib6_config cfg = {
2390 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2391 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2392 .fc_ifindex = dev->ifindex,
2393 .fc_expires = expires,
2395 .fc_flags = RTF_UP | flags,
2396 .fc_nlinfo.nl_net = dev_net(dev),
2397 .fc_protocol = RTPROT_KERNEL,
2398 .fc_type = RTN_UNICAST,
2403 /* Prevent useless cloning on PtP SIT.
2404 This thing is done here expecting that the whole
2405 class of non-broadcast devices need not cloning.
2407 #if IS_ENABLED(CONFIG_IPV6_SIT)
2408 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2409 cfg.fc_flags |= RTF_NONEXTHOP;
2412 ip6_route_add(&cfg, gfp_flags, NULL);
2416 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2418 const struct net_device *dev,
2419 u32 flags, u32 noflags,
2422 struct fib6_node *fn;
2423 struct fib6_info *rt = NULL;
2424 struct fib6_table *table;
2425 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2427 table = fib6_get_table(dev_net(dev), tb_id);
2432 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2436 for_each_fib6_node_rt_rcu(fn) {
2437 /* prefix routes only use builtin fib6_nh */
2441 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2443 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2445 if ((rt->fib6_flags & flags) != flags)
2447 if ((rt->fib6_flags & noflags) != 0)
2449 if (!fib6_info_hold_safe(rt))
2459 /* Create "default" multicast route to the interface */
2461 static void addrconf_add_mroute(struct net_device *dev)
2463 struct fib6_config cfg = {
2464 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2465 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2466 .fc_ifindex = dev->ifindex,
2469 .fc_type = RTN_UNICAST,
2470 .fc_nlinfo.nl_net = dev_net(dev),
2473 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2475 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2478 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2480 struct inet6_dev *idev;
2484 idev = ipv6_find_idev(dev);
2488 if (idev->cnf.disable_ipv6)
2489 return ERR_PTR(-EACCES);
2491 /* Add default multicast route */
2492 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2493 addrconf_add_mroute(dev);
2498 static void manage_tempaddrs(struct inet6_dev *idev,
2499 struct inet6_ifaddr *ifp,
2500 __u32 valid_lft, __u32 prefered_lft,
2501 bool create, unsigned long now)
2504 struct inet6_ifaddr *ift;
2506 read_lock_bh(&idev->lock);
2507 /* update all temporary addresses in the list */
2508 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2509 int age, max_valid, max_prefered;
2511 if (ifp != ift->ifpub)
2514 /* RFC 4941 section 3.3:
2515 * If a received option will extend the lifetime of a public
2516 * address, the lifetimes of temporary addresses should
2517 * be extended, subject to the overall constraint that no
2518 * temporary addresses should ever remain "valid" or "preferred"
2519 * for a time longer than (TEMP_VALID_LIFETIME) or
2520 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2522 age = (now - ift->cstamp) / HZ;
2523 max_valid = idev->cnf.temp_valid_lft - age;
2527 max_prefered = idev->cnf.temp_prefered_lft -
2528 idev->desync_factor - age;
2529 if (max_prefered < 0)
2532 if (valid_lft > max_valid)
2533 valid_lft = max_valid;
2535 if (prefered_lft > max_prefered)
2536 prefered_lft = max_prefered;
2538 spin_lock(&ift->lock);
2540 ift->valid_lft = valid_lft;
2541 ift->prefered_lft = prefered_lft;
2543 if (prefered_lft > 0)
2544 ift->flags &= ~IFA_F_DEPRECATED;
2546 spin_unlock(&ift->lock);
2547 if (!(flags&IFA_F_TENTATIVE))
2548 ipv6_ifa_notify(0, ift);
2551 if ((create || list_empty(&idev->tempaddr_list)) &&
2552 idev->cnf.use_tempaddr > 0) {
2553 /* When a new public address is created as described
2554 * in [ADDRCONF], also create a new temporary address.
2555 * Also create a temporary address if it's enabled but
2556 * no temporary address currently exists.
2558 read_unlock_bh(&idev->lock);
2559 ipv6_create_tempaddr(ifp, false);
2561 read_unlock_bh(&idev->lock);
2565 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2567 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2568 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2571 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2572 const struct prefix_info *pinfo,
2573 struct inet6_dev *in6_dev,
2574 const struct in6_addr *addr, int addr_type,
2575 u32 addr_flags, bool sllao, bool tokenized,
2576 __u32 valid_lft, u32 prefered_lft)
2578 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2581 if (!ifp && valid_lft) {
2582 int max_addresses = in6_dev->cnf.max_addresses;
2583 struct ifa6_config cfg = {
2585 .plen = pinfo->prefix_len,
2586 .ifa_flags = addr_flags,
2587 .valid_lft = valid_lft,
2588 .preferred_lft = prefered_lft,
2589 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2592 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2593 if ((net->ipv6.devconf_all->optimistic_dad ||
2594 in6_dev->cnf.optimistic_dad) &&
2595 !net->ipv6.devconf_all->forwarding && sllao)
2596 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2599 /* Do not allow to create too much of autoconfigured
2600 * addresses; this would be too easy way to crash kernel.
2602 if (!max_addresses ||
2603 ipv6_count_addresses(in6_dev) < max_addresses)
2604 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2606 if (IS_ERR_OR_NULL(ifp))
2610 spin_lock_bh(&ifp->lock);
2611 ifp->flags |= IFA_F_MANAGETEMPADDR;
2612 ifp->cstamp = jiffies;
2613 ifp->tokenized = tokenized;
2614 spin_unlock_bh(&ifp->lock);
2615 addrconf_dad_start(ifp);
2623 /* Update lifetime (RFC4862 5.5.3 e)
2624 * We deviate from RFC4862 by honoring all Valid Lifetimes to
2625 * improve the reaction of SLAAC to renumbering events
2626 * (draft-gont-6man-slaac-renum-06, Section 4.2)
2628 spin_lock_bh(&ifp->lock);
2630 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2631 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2635 if (!create && stored_lft) {
2636 ifp->valid_lft = valid_lft;
2637 ifp->prefered_lft = prefered_lft;
2640 ifp->flags &= ~IFA_F_DEPRECATED;
2641 spin_unlock_bh(&ifp->lock);
2643 if (!(flags&IFA_F_TENTATIVE))
2644 ipv6_ifa_notify(0, ifp);
2646 spin_unlock_bh(&ifp->lock);
2648 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2657 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2659 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2661 struct prefix_info *pinfo;
2666 struct inet6_dev *in6_dev;
2667 struct net *net = dev_net(dev);
2669 pinfo = (struct prefix_info *) opt;
2671 if (len < sizeof(struct prefix_info)) {
2672 netdev_dbg(dev, "addrconf: prefix option too short\n");
2677 * Validation checks ([ADDRCONF], page 19)
2680 addr_type = ipv6_addr_type(&pinfo->prefix);
2682 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2685 valid_lft = ntohl(pinfo->valid);
2686 prefered_lft = ntohl(pinfo->prefered);
2688 if (prefered_lft > valid_lft) {
2689 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2693 in6_dev = in6_dev_get(dev);
2696 net_dbg_ratelimited("addrconf: device %s not configured\n",
2702 * Two things going on here:
2703 * 1) Add routes for on-link prefixes
2704 * 2) Configure prefixes with the auto flag set
2707 if (pinfo->onlink) {
2708 struct fib6_info *rt;
2709 unsigned long rt_expires;
2711 /* Avoid arithmetic overflow. Really, we could
2712 * save rt_expires in seconds, likely valid_lft,
2713 * but it would require division in fib gc, that it
2717 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2719 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2721 if (addrconf_finite_timeout(rt_expires))
2724 rt = addrconf_get_prefix_route(&pinfo->prefix,
2727 RTF_ADDRCONF | RTF_PREFIX_RT,
2731 /* Autoconf prefix route */
2732 if (valid_lft == 0) {
2733 ip6_del_rt(net, rt, false);
2735 } else if (addrconf_finite_timeout(rt_expires)) {
2737 fib6_set_expires(rt, jiffies + rt_expires);
2739 fib6_clean_expires(rt);
2741 } else if (valid_lft) {
2742 clock_t expires = 0;
2743 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2744 if (addrconf_finite_timeout(rt_expires)) {
2746 flags |= RTF_EXPIRES;
2747 expires = jiffies_to_clock_t(rt_expires);
2749 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2750 0, dev, expires, flags,
2753 fib6_info_release(rt);
2756 /* Try to figure out our local address for this prefix */
2758 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2759 struct in6_addr addr;
2760 bool tokenized = false, dev_addr_generated = false;
2762 if (pinfo->prefix_len == 64) {
2763 memcpy(&addr, &pinfo->prefix, 8);
2765 if (!ipv6_addr_any(&in6_dev->token)) {
2766 read_lock_bh(&in6_dev->lock);
2767 memcpy(addr.s6_addr + 8,
2768 in6_dev->token.s6_addr + 8, 8);
2769 read_unlock_bh(&in6_dev->lock);
2771 } else if (is_addr_mode_generate_stable(in6_dev) &&
2772 !ipv6_generate_stable_address(&addr, 0,
2774 addr_flags |= IFA_F_STABLE_PRIVACY;
2776 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2777 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2780 dev_addr_generated = true;
2784 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2789 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2792 tokenized, valid_lft,
2797 /* Ignore error case here because previous prefix add addr was
2798 * successful which will be notified.
2800 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2801 addr_type, addr_flags, sllao,
2802 tokenized, valid_lft,
2804 dev_addr_generated);
2806 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2808 in6_dev_put(in6_dev);
2811 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2812 struct in6_ifreq *ireq)
2814 struct ip_tunnel_parm p = { };
2817 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2818 return -EADDRNOTAVAIL;
2820 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2823 p.iph.protocol = IPPROTO_IPV6;
2826 if (!dev->netdev_ops->ndo_tunnel_ctl)
2828 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2832 dev = __dev_get_by_name(net, p.name);
2835 return dev_open(dev, NULL);
2839 * Set destination address.
2840 * Special case for SIT interfaces where we create a new "virtual"
2843 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2845 struct net_device *dev;
2846 struct in6_ifreq ireq;
2849 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2851 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2855 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2856 if (dev && dev->type == ARPHRD_SIT)
2857 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2862 static int ipv6_mc_config(struct sock *sk, bool join,
2863 const struct in6_addr *addr, int ifindex)
2871 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2873 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2880 * Manual configuration of address on an interface
2882 static int inet6_addr_add(struct net *net, int ifindex,
2883 struct ifa6_config *cfg,
2884 struct netlink_ext_ack *extack)
2886 struct inet6_ifaddr *ifp;
2887 struct inet6_dev *idev;
2888 struct net_device *dev;
2889 unsigned long timeout;
2895 if (cfg->plen > 128)
2898 /* check the lifetime */
2899 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2902 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2905 dev = __dev_get_by_index(net, ifindex);
2909 idev = addrconf_add_dev(dev);
2911 return PTR_ERR(idev);
2913 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2914 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2915 true, cfg->pfx, ifindex);
2921 cfg->scope = ipv6_addr_scope(cfg->pfx);
2923 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2924 if (addrconf_finite_timeout(timeout)) {
2925 expires = jiffies_to_clock_t(timeout * HZ);
2926 cfg->valid_lft = timeout;
2927 flags = RTF_EXPIRES;
2931 cfg->ifa_flags |= IFA_F_PERMANENT;
2934 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2935 if (addrconf_finite_timeout(timeout)) {
2937 cfg->ifa_flags |= IFA_F_DEPRECATED;
2938 cfg->preferred_lft = timeout;
2941 ifp = ipv6_add_addr(idev, cfg, true, extack);
2943 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2944 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2945 ifp->rt_priority, dev, expires,
2949 /* Send a netlink notification if DAD is enabled and
2950 * optimistic flag is not set
2952 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2953 ipv6_ifa_notify(0, ifp);
2955 * Note that section 3.1 of RFC 4429 indicates
2956 * that the Optimistic flag should not be set for
2957 * manually configured addresses
2959 addrconf_dad_start(ifp);
2960 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2961 manage_tempaddrs(idev, ifp, cfg->valid_lft,
2962 cfg->preferred_lft, true, jiffies);
2964 addrconf_verify_rtnl();
2966 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2967 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2971 return PTR_ERR(ifp);
2974 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2975 const struct in6_addr *pfx, unsigned int plen)
2977 struct inet6_ifaddr *ifp;
2978 struct inet6_dev *idev;
2979 struct net_device *dev;
2984 dev = __dev_get_by_index(net, ifindex);
2988 idev = __in6_dev_get(dev);
2992 read_lock_bh(&idev->lock);
2993 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2994 if (ifp->prefix_len == plen &&
2995 ipv6_addr_equal(pfx, &ifp->addr)) {
2997 read_unlock_bh(&idev->lock);
2999 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3000 (ifa_flags & IFA_F_MANAGETEMPADDR))
3001 manage_tempaddrs(idev, ifp, 0, 0, false,
3004 addrconf_verify_rtnl();
3005 if (ipv6_addr_is_multicast(pfx)) {
3006 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3007 false, pfx, dev->ifindex);
3012 read_unlock_bh(&idev->lock);
3013 return -EADDRNOTAVAIL;
3017 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3019 struct ifa6_config cfg = {
3020 .ifa_flags = IFA_F_PERMANENT,
3021 .preferred_lft = INFINITY_LIFE_TIME,
3022 .valid_lft = INFINITY_LIFE_TIME,
3024 struct in6_ifreq ireq;
3027 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3030 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3033 cfg.pfx = &ireq.ifr6_addr;
3034 cfg.plen = ireq.ifr6_prefixlen;
3037 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3042 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3044 struct in6_ifreq ireq;
3047 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3050 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3054 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3055 ireq.ifr6_prefixlen);
3060 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3061 int plen, int scope)
3063 struct inet6_ifaddr *ifp;
3064 struct ifa6_config cfg = {
3067 .ifa_flags = IFA_F_PERMANENT,
3068 .valid_lft = INFINITY_LIFE_TIME,
3069 .preferred_lft = INFINITY_LIFE_TIME,
3073 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3075 spin_lock_bh(&ifp->lock);
3076 ifp->flags &= ~IFA_F_TENTATIVE;
3077 spin_unlock_bh(&ifp->lock);
3078 rt_genid_bump_ipv6(dev_net(idev->dev));
3079 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3084 #if IS_ENABLED(CONFIG_IPV6_SIT)
3085 static void sit_add_v4_addrs(struct inet6_dev *idev)
3087 struct in6_addr addr;
3088 struct net_device *dev;
3089 struct net *net = dev_net(idev->dev);
3095 memset(&addr, 0, sizeof(struct in6_addr));
3096 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3098 if (idev->dev->flags&IFF_POINTOPOINT) {
3099 addr.s6_addr32[0] = htonl(0xfe800000);
3103 scope = IPV6_ADDR_COMPATv4;
3105 pflags |= RTF_NONEXTHOP;
3108 if (addr.s6_addr32[3]) {
3109 add_addr(idev, &addr, plen, scope);
3110 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3115 for_each_netdev(net, dev) {
3116 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3117 if (in_dev && (dev->flags & IFF_UP)) {
3118 struct in_ifaddr *ifa;
3121 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3122 addr.s6_addr32[3] = ifa->ifa_local;
3124 if (ifa->ifa_scope == RT_SCOPE_LINK)
3126 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3127 if (idev->dev->flags&IFF_POINTOPOINT)
3132 add_addr(idev, &addr, plen, flag);
3133 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3134 0, pflags, GFP_KERNEL);
3141 static void init_loopback(struct net_device *dev)
3143 struct inet6_dev *idev;
3149 idev = ipv6_find_idev(dev);
3151 pr_debug("%s: add_dev failed\n", __func__);
3155 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3158 void addrconf_add_linklocal(struct inet6_dev *idev,
3159 const struct in6_addr *addr, u32 flags)
3161 struct ifa6_config cfg = {
3164 .ifa_flags = flags | IFA_F_PERMANENT,
3165 .valid_lft = INFINITY_LIFE_TIME,
3166 .preferred_lft = INFINITY_LIFE_TIME,
3169 struct inet6_ifaddr *ifp;
3171 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3172 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3173 idev->cnf.optimistic_dad) &&
3174 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3175 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3178 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3180 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3182 addrconf_dad_start(ifp);
3186 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3188 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3190 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3193 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3194 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3197 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3198 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3204 static int ipv6_generate_stable_address(struct in6_addr *address,
3206 const struct inet6_dev *idev)
3208 static DEFINE_SPINLOCK(lock);
3209 static __u32 digest[SHA1_DIGEST_WORDS];
3210 static __u32 workspace[SHA1_WORKSPACE_WORDS];
3213 char __data[SHA1_BLOCK_SIZE];
3215 struct in6_addr secret;
3217 unsigned char hwaddr[MAX_ADDR_LEN];
3222 struct in6_addr secret;
3223 struct in6_addr temp;
3224 struct net *net = dev_net(idev->dev);
3226 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3228 if (idev->cnf.stable_secret.initialized)
3229 secret = idev->cnf.stable_secret.secret;
3230 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3231 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3236 spin_lock_bh(&lock);
3239 memset(&data, 0, sizeof(data));
3240 memset(workspace, 0, sizeof(workspace));
3241 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3242 data.prefix[0] = address->s6_addr32[0];
3243 data.prefix[1] = address->s6_addr32[1];
3244 data.secret = secret;
3245 data.dad_count = dad_count;
3247 sha1_transform(digest, data.__data, workspace);
3250 temp.s6_addr32[2] = (__force __be32)digest[0];
3251 temp.s6_addr32[3] = (__force __be32)digest[1];
3253 spin_unlock_bh(&lock);
3255 if (ipv6_reserved_interfaceid(temp)) {
3257 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3266 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3268 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3272 s = &idev->cnf.stable_secret;
3273 get_random_bytes(&s->secret, sizeof(s->secret));
3274 s->initialized = true;
3277 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3279 struct in6_addr addr;
3281 /* no link local addresses on L3 master devices */
3282 if (netif_is_l3_master(idev->dev))
3285 /* no link local addresses on devices flagged as slaves */
3286 if (idev->dev->flags & IFF_SLAVE)
3289 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3291 switch (idev->cnf.addr_gen_mode) {
3292 case IN6_ADDR_GEN_MODE_RANDOM:
3293 ipv6_gen_mode_random_init(idev);
3295 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3296 if (!ipv6_generate_stable_address(&addr, 0, idev))
3297 addrconf_add_linklocal(idev, &addr,
3298 IFA_F_STABLE_PRIVACY);
3299 else if (prefix_route)
3300 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3303 case IN6_ADDR_GEN_MODE_EUI64:
3304 /* addrconf_add_linklocal also adds a prefix_route and we
3305 * only need to care about prefix routes if ipv6_generate_eui64
3306 * couldn't generate one.
3308 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3309 addrconf_add_linklocal(idev, &addr, 0);
3310 else if (prefix_route)
3311 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3314 case IN6_ADDR_GEN_MODE_NONE:
3316 /* will not add any link local address */
3321 static void addrconf_dev_config(struct net_device *dev)
3323 struct inet6_dev *idev;
3327 if ((dev->type != ARPHRD_ETHER) &&
3328 (dev->type != ARPHRD_FDDI) &&
3329 (dev->type != ARPHRD_ARCNET) &&
3330 (dev->type != ARPHRD_INFINIBAND) &&
3331 (dev->type != ARPHRD_IEEE1394) &&
3332 (dev->type != ARPHRD_TUNNEL6) &&
3333 (dev->type != ARPHRD_6LOWPAN) &&
3334 (dev->type != ARPHRD_IP6GRE) &&
3335 (dev->type != ARPHRD_IPGRE) &&
3336 (dev->type != ARPHRD_TUNNEL) &&
3337 (dev->type != ARPHRD_NONE) &&
3338 (dev->type != ARPHRD_RAWIP)) {
3339 /* Alas, we support only Ethernet autoconfiguration. */
3340 idev = __in6_dev_get(dev);
3341 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3342 dev->flags & IFF_MULTICAST)
3347 idev = addrconf_add_dev(dev);
3351 /* this device type has no EUI support */
3352 if (dev->type == ARPHRD_NONE &&
3353 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3354 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3356 addrconf_addr_gen(idev, false);
3359 #if IS_ENABLED(CONFIG_IPV6_SIT)
3360 static void addrconf_sit_config(struct net_device *dev)
3362 struct inet6_dev *idev;
3367 * Configure the tunnel with one of our IPv4
3368 * addresses... we should configure all of
3369 * our v4 addrs in the tunnel
3372 idev = ipv6_find_idev(dev);
3374 pr_debug("%s: add_dev failed\n", __func__);
3378 if (dev->priv_flags & IFF_ISATAP) {
3379 addrconf_addr_gen(idev, false);
3383 sit_add_v4_addrs(idev);
3385 if (dev->flags&IFF_POINTOPOINT)
3386 addrconf_add_mroute(dev);
3390 #if IS_ENABLED(CONFIG_NET_IPGRE)
3391 static void addrconf_gre_config(struct net_device *dev)
3393 struct inet6_dev *idev;
3397 idev = ipv6_find_idev(dev);
3399 pr_debug("%s: add_dev failed\n", __func__);
3403 addrconf_addr_gen(idev, true);
3404 if (dev->flags & IFF_POINTOPOINT)
3405 addrconf_add_mroute(dev);
3409 static int fixup_permanent_addr(struct net *net,
3410 struct inet6_dev *idev,
3411 struct inet6_ifaddr *ifp)
3413 /* !fib6_node means the host route was removed from the
3414 * FIB, for example, if 'lo' device is taken down. In that
3415 * case regenerate the host route.
3417 if (!ifp->rt || !ifp->rt->fib6_node) {
3418 struct fib6_info *f6i, *prev;
3420 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3423 return PTR_ERR(f6i);
3425 /* ifp->rt can be accessed outside of rtnl */
3426 spin_lock(&ifp->lock);
3429 spin_unlock(&ifp->lock);
3431 fib6_info_release(prev);
3434 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3435 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3436 ifp->rt_priority, idev->dev, 0, 0,
3440 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3441 addrconf_dad_start(ifp);
3446 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3448 struct inet6_ifaddr *ifp, *tmp;
3449 struct inet6_dev *idev;
3451 idev = __in6_dev_get(dev);
3455 write_lock_bh(&idev->lock);
3457 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3458 if ((ifp->flags & IFA_F_PERMANENT) &&
3459 fixup_permanent_addr(net, idev, ifp) < 0) {
3460 write_unlock_bh(&idev->lock);
3463 write_lock_bh(&idev->lock);
3465 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3466 idev->dev->name, &ifp->addr);
3470 write_unlock_bh(&idev->lock);
3473 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3476 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3477 struct netdev_notifier_change_info *change_info;
3478 struct netdev_notifier_changeupper_info *info;
3479 struct inet6_dev *idev = __in6_dev_get(dev);
3480 struct net *net = dev_net(dev);
3481 int run_pending = 0;
3485 case NETDEV_REGISTER:
3486 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3487 idev = ipv6_add_dev(dev);
3489 return notifier_from_errno(PTR_ERR(idev));
3493 case NETDEV_CHANGEMTU:
3494 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3495 if (dev->mtu < IPV6_MIN_MTU) {
3496 addrconf_ifdown(dev, dev != net->loopback_dev);
3501 rt6_mtu_change(dev, dev->mtu);
3502 idev->cnf.mtu6 = dev->mtu;
3506 /* allocate new idev */
3507 idev = ipv6_add_dev(dev);
3511 /* device is still not ready */
3512 if (!(idev->if_flags & IF_READY))
3519 if (dev->flags & IFF_SLAVE)
3522 if (idev && idev->cnf.disable_ipv6)
3525 if (event == NETDEV_UP) {
3526 /* restore routes for permanent addresses */
3527 addrconf_permanent_addr(net, dev);
3529 if (!addrconf_link_ready(dev)) {
3530 /* device is not ready yet. */
3531 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3536 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3537 idev = ipv6_add_dev(dev);
3539 if (!IS_ERR_OR_NULL(idev)) {
3540 idev->if_flags |= IF_READY;
3543 } else if (event == NETDEV_CHANGE) {
3544 if (!addrconf_link_ready(dev)) {
3545 /* device is still not ready. */
3546 rt6_sync_down_dev(dev, event);
3550 if (!IS_ERR_OR_NULL(idev)) {
3551 if (idev->if_flags & IF_READY) {
3552 /* device is already configured -
3553 * but resend MLD reports, we might
3554 * have roamed and need to update
3555 * multicast snooping switches
3559 if (change_info->flags_changed & IFF_NOARP)
3560 addrconf_dad_run(idev, true);
3561 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3564 idev->if_flags |= IF_READY;
3567 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3573 switch (dev->type) {
3574 #if IS_ENABLED(CONFIG_IPV6_SIT)
3576 addrconf_sit_config(dev);
3579 #if IS_ENABLED(CONFIG_NET_IPGRE)
3581 addrconf_gre_config(dev);
3584 case ARPHRD_LOOPBACK:
3589 addrconf_dev_config(dev);
3593 if (!IS_ERR_OR_NULL(idev)) {
3595 addrconf_dad_run(idev, false);
3597 /* Device has an address by now */
3598 rt6_sync_up(dev, RTNH_F_DEAD);
3601 * If the MTU changed during the interface down,
3602 * when the interface up, the changed MTU must be
3603 * reflected in the idev as well as routers.
3605 if (idev->cnf.mtu6 != dev->mtu &&
3606 dev->mtu >= IPV6_MIN_MTU) {
3607 rt6_mtu_change(dev, dev->mtu);
3608 idev->cnf.mtu6 = dev->mtu;
3610 idev->tstamp = jiffies;
3611 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3614 * If the changed mtu during down is lower than
3615 * IPV6_MIN_MTU stop IPv6 on this interface.
3617 if (dev->mtu < IPV6_MIN_MTU)
3618 addrconf_ifdown(dev, dev != net->loopback_dev);
3623 case NETDEV_UNREGISTER:
3625 * Remove all addresses from this interface.
3627 addrconf_ifdown(dev, event != NETDEV_DOWN);
3630 case NETDEV_CHANGENAME:
3632 snmp6_unregister_dev(idev);
3633 addrconf_sysctl_unregister(idev);
3634 err = addrconf_sysctl_register(idev);
3636 return notifier_from_errno(err);
3637 err = snmp6_register_dev(idev);
3639 addrconf_sysctl_unregister(idev);
3640 return notifier_from_errno(err);
3645 case NETDEV_PRE_TYPE_CHANGE:
3646 case NETDEV_POST_TYPE_CHANGE:
3648 addrconf_type_change(dev, event);
3651 case NETDEV_CHANGEUPPER:
3654 /* flush all routes if dev is linked to or unlinked from
3655 * an L3 master device (e.g., VRF)
3657 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3658 addrconf_ifdown(dev, false);
3665 * addrconf module should be notified of a device going up
3667 static struct notifier_block ipv6_dev_notf = {
3668 .notifier_call = addrconf_notify,
3669 .priority = ADDRCONF_NOTIFY_PRIORITY,
3672 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3674 struct inet6_dev *idev;
3677 idev = __in6_dev_get(dev);
3679 if (event == NETDEV_POST_TYPE_CHANGE)
3680 ipv6_mc_remap(idev);
3681 else if (event == NETDEV_PRE_TYPE_CHANGE)
3682 ipv6_mc_unmap(idev);
3685 static bool addr_is_local(const struct in6_addr *addr)
3687 return ipv6_addr_type(addr) &
3688 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3691 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3693 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3694 struct net *net = dev_net(dev);
3695 struct inet6_dev *idev;
3696 struct inet6_ifaddr *ifa, *tmp;
3697 bool keep_addr = false;
3702 rt6_disable_ip(dev, event);
3704 idev = __in6_dev_get(dev);
3709 * Step 1: remove reference to ipv6 device from parent device.
3715 /* protected by rtnl_lock */
3716 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3718 /* Step 1.5: remove snmp6 entry */
3719 snmp6_unregister_dev(idev);
3723 /* combine the user config with event to determine if permanent
3724 * addresses are to be removed from address hash table
3726 if (!unregister && !idev->cnf.disable_ipv6) {
3727 /* aggregate the system setting and interface setting */
3728 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3731 _keep_addr = idev->cnf.keep_addr_on_down;
3733 keep_addr = (_keep_addr > 0);
3736 /* Step 2: clear hash table */
3737 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3738 struct hlist_head *h = &inet6_addr_lst[i];
3740 spin_lock_bh(&addrconf_hash_lock);
3742 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3743 if (ifa->idev == idev) {
3744 addrconf_del_dad_work(ifa);
3745 /* combined flag + permanent flag decide if
3746 * address is retained on a down event
3749 !(ifa->flags & IFA_F_PERMANENT) ||
3750 addr_is_local(&ifa->addr)) {
3751 hlist_del_init_rcu(&ifa->addr_lst);
3756 spin_unlock_bh(&addrconf_hash_lock);
3759 write_lock_bh(&idev->lock);
3761 addrconf_del_rs_timer(idev);
3763 /* Step 2: clear flags for stateless addrconf */
3765 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3767 /* Step 3: clear tempaddr list */
3768 while (!list_empty(&idev->tempaddr_list)) {
3769 ifa = list_first_entry(&idev->tempaddr_list,
3770 struct inet6_ifaddr, tmp_list);
3771 list_del(&ifa->tmp_list);
3772 write_unlock_bh(&idev->lock);
3773 spin_lock_bh(&ifa->lock);
3776 in6_ifa_put(ifa->ifpub);
3779 spin_unlock_bh(&ifa->lock);
3781 write_lock_bh(&idev->lock);
3784 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3785 struct fib6_info *rt = NULL;
3788 addrconf_del_dad_work(ifa);
3790 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3791 !addr_is_local(&ifa->addr);
3793 write_unlock_bh(&idev->lock);
3794 spin_lock_bh(&ifa->lock);
3797 /* set state to skip the notifier below */
3798 state = INET6_IFADDR_STATE_DEAD;
3799 ifa->state = INET6_IFADDR_STATE_PREDAD;
3800 if (!(ifa->flags & IFA_F_NODAD))
3801 ifa->flags |= IFA_F_TENTATIVE;
3807 ifa->state = INET6_IFADDR_STATE_DEAD;
3810 spin_unlock_bh(&ifa->lock);
3813 ip6_del_rt(net, rt, false);
3815 if (state != INET6_IFADDR_STATE_DEAD) {
3816 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3817 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3819 if (idev->cnf.forwarding)
3820 addrconf_leave_anycast(ifa);
3821 addrconf_leave_solict(ifa->idev, &ifa->addr);
3824 write_lock_bh(&idev->lock);
3826 list_del_rcu(&ifa->if_list);
3831 write_unlock_bh(&idev->lock);
3833 /* Step 5: Discard anycast and multicast list */
3835 ipv6_ac_destroy_dev(idev);
3836 ipv6_mc_destroy_dev(idev);
3841 idev->tstamp = jiffies;
3843 /* Last: Shot the device (if unregistered) */
3845 addrconf_sysctl_unregister(idev);
3846 neigh_parms_release(&nd_tbl, idev->nd_parms);
3847 neigh_ifdown(&nd_tbl, dev);
3853 static void addrconf_rs_timer(struct timer_list *t)
3855 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3856 struct net_device *dev = idev->dev;
3857 struct in6_addr lladdr;
3859 write_lock(&idev->lock);
3860 if (idev->dead || !(idev->if_flags & IF_READY))
3863 if (!ipv6_accept_ra(idev))
3866 /* Announcement received after solicitation was sent */
3867 if (idev->if_flags & IF_RA_RCVD)
3870 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3871 write_unlock(&idev->lock);
3872 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3873 ndisc_send_rs(dev, &lladdr,
3874 &in6addr_linklocal_allrouters);
3878 write_lock(&idev->lock);
3879 idev->rs_interval = rfc3315_s14_backoff_update(
3880 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3881 /* The wait after the last probe can be shorter */
3882 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3883 idev->cnf.rtr_solicits) ?
3884 idev->cnf.rtr_solicit_delay :
3888 * Note: we do not support deprecated "all on-link"
3889 * assumption any longer.
3891 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3895 write_unlock(&idev->lock);
3901 * Duplicate Address Detection
3903 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3905 unsigned long rand_num;
3906 struct inet6_dev *idev = ifp->idev;
3909 if (ifp->flags & IFA_F_OPTIMISTIC)
3912 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3915 if (idev->cnf.enhanced_dad ||
3916 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3918 get_random_bytes(&nonce, 6);
3921 ifp->dad_nonce = nonce;
3922 ifp->dad_probes = idev->cnf.dad_transmits;
3923 addrconf_mod_dad_work(ifp, rand_num);
3926 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3928 struct inet6_dev *idev = ifp->idev;
3929 struct net_device *dev = idev->dev;
3930 bool bump_id, notify = false;
3933 addrconf_join_solict(dev, &ifp->addr);
3935 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3937 read_lock_bh(&idev->lock);
3938 spin_lock(&ifp->lock);
3939 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3943 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3944 (net->ipv6.devconf_all->accept_dad < 1 &&
3945 idev->cnf.accept_dad < 1) ||
3946 !(ifp->flags&IFA_F_TENTATIVE) ||
3947 ifp->flags & IFA_F_NODAD) {
3948 bool send_na = false;
3950 if (ifp->flags & IFA_F_TENTATIVE &&
3951 !(ifp->flags & IFA_F_OPTIMISTIC))
3953 bump_id = ifp->flags & IFA_F_TENTATIVE;
3954 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3955 spin_unlock(&ifp->lock);
3956 read_unlock_bh(&idev->lock);
3958 addrconf_dad_completed(ifp, bump_id, send_na);
3962 if (!(idev->if_flags & IF_READY)) {
3963 spin_unlock(&ifp->lock);
3964 read_unlock_bh(&idev->lock);
3966 * If the device is not ready:
3967 * - keep it tentative if it is a permanent address.
3968 * - otherwise, kill it.
3971 addrconf_dad_stop(ifp, 0);
3976 * Optimistic nodes can start receiving
3979 if (ifp->flags & IFA_F_OPTIMISTIC) {
3980 ip6_ins_rt(net, ifp->rt);
3981 if (ipv6_use_optimistic_addr(net, idev)) {
3982 /* Because optimistic nodes can use this address,
3983 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3989 addrconf_dad_kick(ifp);
3991 spin_unlock(&ifp->lock);
3992 read_unlock_bh(&idev->lock);
3994 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3997 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3999 bool begin_dad = false;
4001 spin_lock_bh(&ifp->lock);
4002 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4003 ifp->state = INET6_IFADDR_STATE_PREDAD;
4006 spin_unlock_bh(&ifp->lock);
4009 addrconf_mod_dad_work(ifp, 0);
4012 static void addrconf_dad_work(struct work_struct *w)
4014 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4015 struct inet6_ifaddr,
4017 struct inet6_dev *idev = ifp->idev;
4018 bool bump_id, disable_ipv6 = false;
4019 struct in6_addr mcaddr;
4025 } action = DAD_PROCESS;
4029 spin_lock_bh(&ifp->lock);
4030 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4032 ifp->state = INET6_IFADDR_STATE_DAD;
4033 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4035 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4037 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4038 idev->cnf.accept_dad > 1) &&
4039 !idev->cnf.disable_ipv6 &&
4040 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4041 struct in6_addr addr;
4043 addr.s6_addr32[0] = htonl(0xfe800000);
4044 addr.s6_addr32[1] = 0;
4046 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4047 ipv6_addr_equal(&ifp->addr, &addr)) {
4048 /* DAD failed for link-local based on MAC */
4049 idev->cnf.disable_ipv6 = 1;
4051 pr_info("%s: IPv6 being disabled!\n",
4052 ifp->idev->dev->name);
4053 disable_ipv6 = true;
4057 spin_unlock_bh(&ifp->lock);
4059 if (action == DAD_BEGIN) {
4060 addrconf_dad_begin(ifp);
4062 } else if (action == DAD_ABORT) {
4064 addrconf_dad_stop(ifp, 1);
4066 addrconf_ifdown(idev->dev, false);
4070 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4073 write_lock_bh(&idev->lock);
4074 if (idev->dead || !(idev->if_flags & IF_READY)) {
4075 write_unlock_bh(&idev->lock);
4079 spin_lock(&ifp->lock);
4080 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4081 spin_unlock(&ifp->lock);
4082 write_unlock_bh(&idev->lock);
4086 if (ifp->dad_probes == 0) {
4087 bool send_na = false;
4090 * DAD was successful
4093 if (ifp->flags & IFA_F_TENTATIVE &&
4094 !(ifp->flags & IFA_F_OPTIMISTIC))
4096 bump_id = ifp->flags & IFA_F_TENTATIVE;
4097 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4098 spin_unlock(&ifp->lock);
4099 write_unlock_bh(&idev->lock);
4101 addrconf_dad_completed(ifp, bump_id, send_na);
4107 addrconf_mod_dad_work(ifp,
4108 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4110 spin_unlock(&ifp->lock);
4111 write_unlock_bh(&idev->lock);
4113 /* send a neighbour solicitation for our addr */
4114 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4115 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4122 /* ifp->idev must be at least read locked */
4123 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4125 struct inet6_ifaddr *ifpiter;
4126 struct inet6_dev *idev = ifp->idev;
4128 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4129 if (ifpiter->scope > IFA_LINK)
4131 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4132 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4133 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4140 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4143 struct net_device *dev = ifp->idev->dev;
4144 struct in6_addr lladdr;
4145 bool send_rs, send_mld;
4147 addrconf_del_dad_work(ifp);
4150 * Configure the address for reception. Now it is valid.
4153 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4155 /* If added prefix is link local and we are prepared to process
4156 router advertisements, start sending router solicitations.
4159 read_lock_bh(&ifp->idev->lock);
4160 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4161 send_rs = send_mld &&
4162 ipv6_accept_ra(ifp->idev) &&
4163 ifp->idev->cnf.rtr_solicits != 0 &&
4164 (dev->flags&IFF_LOOPBACK) == 0;
4165 read_unlock_bh(&ifp->idev->lock);
4167 /* While dad is in progress mld report's source address is in6_addrany.
4168 * Resend with proper ll now.
4171 ipv6_mc_dad_complete(ifp->idev);
4173 /* send unsolicited NA if enabled */
4175 (ifp->idev->cnf.ndisc_notify ||
4176 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4177 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4178 /*router=*/ !!ifp->idev->cnf.forwarding,
4179 /*solicited=*/ false, /*override=*/ true,
4185 * If a host as already performed a random delay
4186 * [...] as part of DAD [...] there is no need
4187 * to delay again before sending the first RS
4189 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4191 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4193 write_lock_bh(&ifp->idev->lock);
4194 spin_lock(&ifp->lock);
4195 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4196 ifp->idev->cnf.rtr_solicit_interval);
4197 ifp->idev->rs_probes = 1;
4198 ifp->idev->if_flags |= IF_RS_SENT;
4199 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4200 spin_unlock(&ifp->lock);
4201 write_unlock_bh(&ifp->idev->lock);
4205 rt_genid_bump_ipv6(dev_net(dev));
4207 /* Make sure that a new temporary address will be created
4208 * before this temporary address becomes deprecated.
4210 if (ifp->flags & IFA_F_TEMPORARY)
4211 addrconf_verify_rtnl();
4214 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4216 struct inet6_ifaddr *ifp;
4218 read_lock_bh(&idev->lock);
4219 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4220 spin_lock(&ifp->lock);
4221 if ((ifp->flags & IFA_F_TENTATIVE &&
4222 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4224 ifp->state = INET6_IFADDR_STATE_PREDAD;
4225 addrconf_dad_kick(ifp);
4227 spin_unlock(&ifp->lock);
4229 read_unlock_bh(&idev->lock);
4232 #ifdef CONFIG_PROC_FS
4233 struct if6_iter_state {
4234 struct seq_net_private p;
4239 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4241 struct if6_iter_state *state = seq->private;
4242 struct net *net = seq_file_net(seq);
4243 struct inet6_ifaddr *ifa = NULL;
4246 /* initial bucket if pos is 0 */
4252 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4253 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4255 if (!net_eq(dev_net(ifa->idev->dev), net))
4257 /* sync with offset */
4258 if (p < state->offset) {
4265 /* prepare for next bucket */
4272 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4273 struct inet6_ifaddr *ifa)
4275 struct if6_iter_state *state = seq->private;
4276 struct net *net = seq_file_net(seq);
4278 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4279 if (!net_eq(dev_net(ifa->idev->dev), net))
4286 while (++state->bucket < IN6_ADDR_HSIZE) {
4287 hlist_for_each_entry_rcu(ifa,
4288 &inet6_addr_lst[state->bucket], addr_lst) {
4289 if (!net_eq(dev_net(ifa->idev->dev), net))
4298 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4302 return if6_get_first(seq, *pos);
4305 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4307 struct inet6_ifaddr *ifa;
4309 ifa = if6_get_next(seq, v);
4314 static void if6_seq_stop(struct seq_file *seq, void *v)
4320 static int if6_seq_show(struct seq_file *seq, void *v)
4322 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4323 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4325 ifp->idev->dev->ifindex,
4329 ifp->idev->dev->name);
4333 static const struct seq_operations if6_seq_ops = {
4334 .start = if6_seq_start,
4335 .next = if6_seq_next,
4336 .show = if6_seq_show,
4337 .stop = if6_seq_stop,
4340 static int __net_init if6_proc_net_init(struct net *net)
4342 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4343 sizeof(struct if6_iter_state)))
4348 static void __net_exit if6_proc_net_exit(struct net *net)
4350 remove_proc_entry("if_inet6", net->proc_net);
4353 static struct pernet_operations if6_proc_net_ops = {
4354 .init = if6_proc_net_init,
4355 .exit = if6_proc_net_exit,
4358 int __init if6_proc_init(void)
4360 return register_pernet_subsys(&if6_proc_net_ops);
4363 void if6_proc_exit(void)
4365 unregister_pernet_subsys(&if6_proc_net_ops);
4367 #endif /* CONFIG_PROC_FS */
4369 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4370 /* Check if address is a home address configured on any interface. */
4371 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4373 unsigned int hash = inet6_addr_hash(net, addr);
4374 struct inet6_ifaddr *ifp = NULL;
4378 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4379 if (!net_eq(dev_net(ifp->idev->dev), net))
4381 if (ipv6_addr_equal(&ifp->addr, addr) &&
4382 (ifp->flags & IFA_F_HOMEADDRESS)) {
4392 /* RFC6554 has some algorithm to avoid loops in segment routing by
4393 * checking if the segments contains any of a local interface address.
4397 * To detect loops in the SRH, a router MUST determine if the SRH
4398 * includes multiple addresses assigned to any interface on that router.
4399 * If such addresses appear more than once and are separated by at least
4400 * one address not assigned to that router.
4402 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4403 unsigned char nsegs)
4405 const struct in6_addr *addr;
4406 int i, ret = 0, found = 0;
4407 struct inet6_ifaddr *ifp;
4408 bool separated = false;
4413 for (i = 0; i < nsegs; i++) {
4415 hash = inet6_addr_hash(net, addr);
4418 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4419 if (!net_eq(dev_net(ifp->idev->dev), net))
4422 if (ipv6_addr_equal(&ifp->addr, addr)) {
4429 if (found > 1 && separated) {
4446 * Periodic address status verification
4449 static void addrconf_verify_rtnl(void)
4451 unsigned long now, next, next_sec, next_sched;
4452 struct inet6_ifaddr *ifp;
4459 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4461 cancel_delayed_work(&addr_chk_work);
4463 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4465 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4468 /* When setting preferred_lft to a value not zero or
4469 * infinity, while valid_lft is infinity
4470 * IFA_F_PERMANENT has a non-infinity life time.
4472 if ((ifp->flags & IFA_F_PERMANENT) &&
4473 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4476 spin_lock(&ifp->lock);
4477 /* We try to batch several events at once. */
4478 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4480 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4481 age >= ifp->valid_lft) {
4482 spin_unlock(&ifp->lock);
4486 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4487 spin_unlock(&ifp->lock);
4489 } else if (age >= ifp->prefered_lft) {
4490 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4493 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4495 ifp->flags |= IFA_F_DEPRECATED;
4498 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4499 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4500 next = ifp->tstamp + ifp->valid_lft * HZ;
4502 spin_unlock(&ifp->lock);
4507 ipv6_ifa_notify(0, ifp);
4511 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4512 !(ifp->flags&IFA_F_TENTATIVE)) {
4513 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4514 ifp->idev->cnf.dad_transmits *
4515 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4517 if (age >= ifp->prefered_lft - regen_advance) {
4518 struct inet6_ifaddr *ifpub = ifp->ifpub;
4519 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4520 next = ifp->tstamp + ifp->prefered_lft * HZ;
4521 if (!ifp->regen_count && ifpub) {
4524 in6_ifa_hold(ifpub);
4525 spin_unlock(&ifp->lock);
4527 spin_lock(&ifpub->lock);
4528 ifpub->regen_count = 0;
4529 spin_unlock(&ifpub->lock);
4530 rcu_read_unlock_bh();
4531 ipv6_create_tempaddr(ifpub, true);
4537 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4538 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4539 spin_unlock(&ifp->lock);
4541 /* ifp->prefered_lft <= ifp->valid_lft */
4542 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4543 next = ifp->tstamp + ifp->prefered_lft * HZ;
4544 spin_unlock(&ifp->lock);
4549 next_sec = round_jiffies_up(next);
4552 /* If rounded timeout is accurate enough, accept it. */
4553 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4554 next_sched = next_sec;
4556 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4557 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4558 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4560 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4561 now, next, next_sec, next_sched);
4562 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4563 rcu_read_unlock_bh();
4566 static void addrconf_verify_work(struct work_struct *w)
4569 addrconf_verify_rtnl();
4573 static void addrconf_verify(void)
4575 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4578 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4579 struct in6_addr **peer_pfx)
4581 struct in6_addr *pfx = NULL;
4586 pfx = nla_data(addr);
4589 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4591 pfx = nla_data(local);
4597 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4598 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4599 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4600 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4601 [IFA_FLAGS] = { .len = sizeof(u32) },
4602 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4603 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4607 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4608 struct netlink_ext_ack *extack)
4610 struct net *net = sock_net(skb->sk);
4611 struct ifaddrmsg *ifm;
4612 struct nlattr *tb[IFA_MAX+1];
4613 struct in6_addr *pfx, *peer_pfx;
4617 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4618 ifa_ipv6_policy, extack);
4622 ifm = nlmsg_data(nlh);
4623 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4627 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4629 /* We ignore other flags so far. */
4630 ifa_flags &= IFA_F_MANAGETEMPADDR;
4632 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4633 ifm->ifa_prefixlen);
4636 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4637 unsigned long expires, u32 flags,
4640 struct fib6_info *f6i;
4643 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4645 ifp->idev->dev, 0, RTF_DEFAULT, true);
4649 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4650 if (f6i->fib6_metric != prio) {
4651 /* delete old one */
4652 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4655 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4657 ifp->rt_priority, ifp->idev->dev,
4658 expires, flags, GFP_KERNEL);
4661 fib6_clean_expires(f6i);
4663 fib6_set_expires(f6i, expires);
4665 fib6_info_release(f6i);
4671 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4675 unsigned long timeout;
4676 bool was_managetempaddr;
4677 bool had_prefixroute;
4678 bool new_peer = false;
4682 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4685 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4686 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4689 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4690 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4692 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4693 if (addrconf_finite_timeout(timeout)) {
4694 expires = jiffies_to_clock_t(timeout * HZ);
4695 cfg->valid_lft = timeout;
4696 flags = RTF_EXPIRES;
4700 cfg->ifa_flags |= IFA_F_PERMANENT;
4703 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4704 if (addrconf_finite_timeout(timeout)) {
4706 cfg->ifa_flags |= IFA_F_DEPRECATED;
4707 cfg->preferred_lft = timeout;
4710 if (cfg->peer_pfx &&
4711 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4712 if (!ipv6_addr_any(&ifp->peer_addr))
4713 cleanup_prefix_route(ifp, expires, true, true);
4717 spin_lock_bh(&ifp->lock);
4718 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4719 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4720 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4721 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4722 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4723 IFA_F_NOPREFIXROUTE);
4724 ifp->flags |= cfg->ifa_flags;
4725 ifp->tstamp = jiffies;
4726 ifp->valid_lft = cfg->valid_lft;
4727 ifp->prefered_lft = cfg->preferred_lft;
4729 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4730 ifp->rt_priority = cfg->rt_priority;
4733 ifp->peer_addr = *cfg->peer_pfx;
4735 spin_unlock_bh(&ifp->lock);
4736 if (!(ifp->flags&IFA_F_TENTATIVE))
4737 ipv6_ifa_notify(0, ifp);
4739 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4742 if (had_prefixroute)
4743 rc = modify_prefix_route(ifp, expires, flags, false);
4745 /* prefix route could have been deleted; if so restore it */
4746 if (rc == -ENOENT) {
4747 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4748 ifp->rt_priority, ifp->idev->dev,
4749 expires, flags, GFP_KERNEL);
4752 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4753 rc = modify_prefix_route(ifp, expires, flags, true);
4755 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4756 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4757 ifp->rt_priority, ifp->idev->dev,
4758 expires, flags, GFP_KERNEL);
4760 } else if (had_prefixroute) {
4761 enum cleanup_prefix_rt_t action;
4762 unsigned long rt_expires;
4764 write_lock_bh(&ifp->idev->lock);
4765 action = check_cleanup_prefix_route(ifp, &rt_expires);
4766 write_unlock_bh(&ifp->idev->lock);
4768 if (action != CLEANUP_PREFIX_RT_NOP) {
4769 cleanup_prefix_route(ifp, rt_expires,
4770 action == CLEANUP_PREFIX_RT_DEL, false);
4774 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4775 if (was_managetempaddr &&
4776 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4778 cfg->preferred_lft = 0;
4780 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4781 cfg->preferred_lft, !was_managetempaddr,
4785 addrconf_verify_rtnl();
4791 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4792 struct netlink_ext_ack *extack)
4794 struct net *net = sock_net(skb->sk);
4795 struct ifaddrmsg *ifm;
4796 struct nlattr *tb[IFA_MAX+1];
4797 struct in6_addr *peer_pfx;
4798 struct inet6_ifaddr *ifa;
4799 struct net_device *dev;
4800 struct inet6_dev *idev;
4801 struct ifa6_config cfg;
4804 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4805 ifa_ipv6_policy, extack);
4809 memset(&cfg, 0, sizeof(cfg));
4811 ifm = nlmsg_data(nlh);
4812 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4816 cfg.peer_pfx = peer_pfx;
4817 cfg.plen = ifm->ifa_prefixlen;
4818 if (tb[IFA_RT_PRIORITY])
4819 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4821 cfg.valid_lft = INFINITY_LIFE_TIME;
4822 cfg.preferred_lft = INFINITY_LIFE_TIME;
4824 if (tb[IFA_CACHEINFO]) {
4825 struct ifa_cacheinfo *ci;
4827 ci = nla_data(tb[IFA_CACHEINFO]);
4828 cfg.valid_lft = ci->ifa_valid;
4829 cfg.preferred_lft = ci->ifa_prefered;
4832 dev = __dev_get_by_index(net, ifm->ifa_index);
4837 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4839 cfg.ifa_flags = ifm->ifa_flags;
4841 /* We ignore other flags so far. */
4842 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4843 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4844 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4846 idev = ipv6_find_idev(dev);
4848 return PTR_ERR(idev);
4850 if (!ipv6_allow_optimistic_dad(net, idev))
4851 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4853 if (cfg.ifa_flags & IFA_F_NODAD &&
4854 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4855 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4859 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4862 * It would be best to check for !NLM_F_CREATE here but
4863 * userspace already relies on not having to provide this.
4865 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4868 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4869 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4872 err = inet6_addr_modify(ifa, &cfg);
4879 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4880 u8 scope, int ifindex)
4882 struct ifaddrmsg *ifm;
4884 ifm = nlmsg_data(nlh);
4885 ifm->ifa_family = AF_INET6;
4886 ifm->ifa_prefixlen = prefixlen;
4887 ifm->ifa_flags = flags;
4888 ifm->ifa_scope = scope;
4889 ifm->ifa_index = ifindex;
4892 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4893 unsigned long tstamp, u32 preferred, u32 valid)
4895 struct ifa_cacheinfo ci;
4897 ci.cstamp = cstamp_delta(cstamp);
4898 ci.tstamp = cstamp_delta(tstamp);
4899 ci.ifa_prefered = preferred;
4900 ci.ifa_valid = valid;
4902 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4905 static inline int rt_scope(int ifa_scope)
4907 if (ifa_scope & IFA_HOST)
4908 return RT_SCOPE_HOST;
4909 else if (ifa_scope & IFA_LINK)
4910 return RT_SCOPE_LINK;
4911 else if (ifa_scope & IFA_SITE)
4912 return RT_SCOPE_SITE;
4914 return RT_SCOPE_UNIVERSE;
4917 static inline int inet6_ifaddr_msgsize(void)
4919 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4920 + nla_total_size(16) /* IFA_LOCAL */
4921 + nla_total_size(16) /* IFA_ADDRESS */
4922 + nla_total_size(sizeof(struct ifa_cacheinfo))
4923 + nla_total_size(4) /* IFA_FLAGS */
4924 + nla_total_size(4) /* IFA_RT_PRIORITY */;
4933 struct inet6_fill_args {
4940 enum addr_type_t type;
4943 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4944 struct inet6_fill_args *args)
4946 struct nlmsghdr *nlh;
4947 u32 preferred, valid;
4949 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4950 sizeof(struct ifaddrmsg), args->flags);
4954 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4955 ifa->idev->dev->ifindex);
4957 if (args->netnsid >= 0 &&
4958 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4961 if (!((ifa->flags&IFA_F_PERMANENT) &&
4962 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4963 preferred = ifa->prefered_lft;
4964 valid = ifa->valid_lft;
4965 if (preferred != INFINITY_LIFE_TIME) {
4966 long tval = (jiffies - ifa->tstamp)/HZ;
4967 if (preferred > tval)
4971 if (valid != INFINITY_LIFE_TIME) {
4979 preferred = INFINITY_LIFE_TIME;
4980 valid = INFINITY_LIFE_TIME;
4983 if (!ipv6_addr_any(&ifa->peer_addr)) {
4984 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4985 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4988 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4991 if (ifa->rt_priority &&
4992 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4995 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4998 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5001 nlmsg_end(skb, nlh);
5005 nlmsg_cancel(skb, nlh);
5009 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5010 struct inet6_fill_args *args)
5012 struct nlmsghdr *nlh;
5013 u8 scope = RT_SCOPE_UNIVERSE;
5014 int ifindex = ifmca->idev->dev->ifindex;
5016 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5017 scope = RT_SCOPE_SITE;
5019 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5020 sizeof(struct ifaddrmsg), args->flags);
5024 if (args->netnsid >= 0 &&
5025 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5028 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5029 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5030 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5031 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5032 nlmsg_cancel(skb, nlh);
5036 nlmsg_end(skb, nlh);
5040 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5041 struct inet6_fill_args *args)
5043 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5044 int ifindex = dev ? dev->ifindex : 1;
5045 struct nlmsghdr *nlh;
5046 u8 scope = RT_SCOPE_UNIVERSE;
5048 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5049 scope = RT_SCOPE_SITE;
5051 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5052 sizeof(struct ifaddrmsg), args->flags);
5056 if (args->netnsid >= 0 &&
5057 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5060 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5061 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5062 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5063 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5064 nlmsg_cancel(skb, nlh);
5068 nlmsg_end(skb, nlh);
5072 /* called with rcu_read_lock() */
5073 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5074 struct netlink_callback *cb, int s_ip_idx,
5075 struct inet6_fill_args *fillargs)
5077 struct ifmcaddr6 *ifmca;
5078 struct ifacaddr6 *ifaca;
5082 read_lock_bh(&idev->lock);
5083 switch (fillargs->type) {
5084 case UNICAST_ADDR: {
5085 struct inet6_ifaddr *ifa;
5086 fillargs->event = RTM_NEWADDR;
5088 /* unicast address incl. temp addr */
5089 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5090 if (ip_idx < s_ip_idx)
5092 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5095 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5101 case MULTICAST_ADDR:
5102 fillargs->event = RTM_GETMULTICAST;
5104 /* multicast address */
5105 for (ifmca = idev->mc_list; ifmca;
5106 ifmca = ifmca->next, ip_idx++) {
5107 if (ip_idx < s_ip_idx)
5109 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5115 fillargs->event = RTM_GETANYCAST;
5116 /* anycast address */
5117 for (ifaca = idev->ac_list; ifaca;
5118 ifaca = ifaca->aca_next, ip_idx++) {
5119 if (ip_idx < s_ip_idx)
5121 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5129 read_unlock_bh(&idev->lock);
5130 cb->args[2] = ip_idx;
5134 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5135 struct inet6_fill_args *fillargs,
5136 struct net **tgt_net, struct sock *sk,
5137 struct netlink_callback *cb)
5139 struct netlink_ext_ack *extack = cb->extack;
5140 struct nlattr *tb[IFA_MAX+1];
5141 struct ifaddrmsg *ifm;
5144 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5145 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5149 ifm = nlmsg_data(nlh);
5150 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5151 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5155 fillargs->ifindex = ifm->ifa_index;
5156 if (fillargs->ifindex) {
5157 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5158 fillargs->flags |= NLM_F_DUMP_FILTERED;
5161 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5162 ifa_ipv6_policy, extack);
5166 for (i = 0; i <= IFA_MAX; ++i) {
5170 if (i == IFA_TARGET_NETNSID) {
5173 fillargs->netnsid = nla_get_s32(tb[i]);
5174 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5176 fillargs->netnsid = -1;
5177 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5178 return PTR_ERR(net);
5182 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5190 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5191 enum addr_type_t type)
5193 const struct nlmsghdr *nlh = cb->nlh;
5194 struct inet6_fill_args fillargs = {
5195 .portid = NETLINK_CB(cb->skb).portid,
5196 .seq = cb->nlh->nlmsg_seq,
5197 .flags = NLM_F_MULTI,
5201 struct net *net = sock_net(skb->sk);
5202 struct net *tgt_net = net;
5203 int idx, s_idx, s_ip_idx;
5205 struct net_device *dev;
5206 struct inet6_dev *idev;
5207 struct hlist_head *head;
5211 s_idx = idx = cb->args[1];
5212 s_ip_idx = cb->args[2];
5214 if (cb->strict_check) {
5215 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5221 if (fillargs.ifindex) {
5222 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5227 idev = __in6_dev_get(dev);
5229 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5239 cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5240 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5242 head = &tgt_net->dev_index_head[h];
5243 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5246 if (h > s_h || idx > s_idx)
5248 idev = __in6_dev_get(dev);
5252 if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5264 if (fillargs.netnsid >= 0)
5267 return skb->len ? : err;
5270 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5272 enum addr_type_t type = UNICAST_ADDR;
5274 return inet6_dump_addr(skb, cb, type);
5277 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5279 enum addr_type_t type = MULTICAST_ADDR;
5281 return inet6_dump_addr(skb, cb, type);
5285 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5287 enum addr_type_t type = ANYCAST_ADDR;
5289 return inet6_dump_addr(skb, cb, type);
5292 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5293 const struct nlmsghdr *nlh,
5295 struct netlink_ext_ack *extack)
5297 struct ifaddrmsg *ifm;
5300 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5301 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5305 if (!netlink_strict_get_check(skb))
5306 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5307 ifa_ipv6_policy, extack);
5309 ifm = nlmsg_data(nlh);
5310 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5311 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5315 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5316 ifa_ipv6_policy, extack);
5320 for (i = 0; i <= IFA_MAX; i++) {
5325 case IFA_TARGET_NETNSID:
5330 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5338 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5339 struct netlink_ext_ack *extack)
5341 struct net *net = sock_net(in_skb->sk);
5342 struct inet6_fill_args fillargs = {
5343 .portid = NETLINK_CB(in_skb).portid,
5344 .seq = nlh->nlmsg_seq,
5345 .event = RTM_NEWADDR,
5349 struct net *tgt_net = net;
5350 struct ifaddrmsg *ifm;
5351 struct nlattr *tb[IFA_MAX+1];
5352 struct in6_addr *addr = NULL, *peer;
5353 struct net_device *dev = NULL;
5354 struct inet6_ifaddr *ifa;
5355 struct sk_buff *skb;
5358 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5362 if (tb[IFA_TARGET_NETNSID]) {
5363 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5365 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5367 if (IS_ERR(tgt_net))
5368 return PTR_ERR(tgt_net);
5371 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5375 ifm = nlmsg_data(nlh);
5377 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5379 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5381 err = -EADDRNOTAVAIL;
5385 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5391 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5393 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5394 WARN_ON(err == -EMSGSIZE);
5398 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5404 if (fillargs.netnsid >= 0)
5410 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5412 struct sk_buff *skb;
5413 struct net *net = dev_net(ifa->idev->dev);
5414 struct inet6_fill_args fillargs = {
5423 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5427 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5429 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5430 WARN_ON(err == -EMSGSIZE);
5434 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5438 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5441 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5442 __s32 *array, int bytes)
5444 BUG_ON(bytes < (DEVCONF_MAX * 4));
5446 memset(array, 0, bytes);
5447 array[DEVCONF_FORWARDING] = cnf->forwarding;
5448 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5449 array[DEVCONF_MTU6] = cnf->mtu6;
5450 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5451 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5452 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5453 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5454 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5455 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5456 jiffies_to_msecs(cnf->rtr_solicit_interval);
5457 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5458 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5459 array[DEVCONF_RTR_SOLICIT_DELAY] =
5460 jiffies_to_msecs(cnf->rtr_solicit_delay);
5461 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5462 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5463 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5464 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5465 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5466 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5467 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5468 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5469 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5470 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5471 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5472 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5473 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5474 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5475 #ifdef CONFIG_IPV6_ROUTER_PREF
5476 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5477 array[DEVCONF_RTR_PROBE_INTERVAL] =
5478 jiffies_to_msecs(cnf->rtr_probe_interval);
5479 #ifdef CONFIG_IPV6_ROUTE_INFO
5480 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5481 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5484 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5485 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5486 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5487 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5488 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5490 #ifdef CONFIG_IPV6_MROUTE
5491 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5493 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5494 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5495 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5496 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5497 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5498 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5499 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5500 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5501 /* we omit DEVCONF_STABLE_SECRET for now */
5502 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5503 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5504 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5505 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5506 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5507 #ifdef CONFIG_IPV6_SEG6_HMAC
5508 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5510 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5511 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5512 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5513 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5514 array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5517 static inline size_t inet6_ifla6_size(void)
5519 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5520 + nla_total_size(sizeof(struct ifla_cacheinfo))
5521 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5522 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5523 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5524 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5525 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5529 static inline size_t inet6_if_nlmsg_size(void)
5531 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5532 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5533 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5534 + nla_total_size(4) /* IFLA_MTU */
5535 + nla_total_size(4) /* IFLA_LINK */
5536 + nla_total_size(1) /* IFLA_OPERSTATE */
5537 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5540 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5544 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5547 /* Use put_unaligned() because stats may not be aligned for u64. */
5548 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5549 for (i = 1; i < ICMP6_MIB_MAX; i++)
5550 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5552 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5555 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5556 int bytes, size_t syncpoff)
5559 u64 buff[IPSTATS_MIB_MAX];
5560 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5564 memset(buff, 0, sizeof(buff));
5565 buff[0] = IPSTATS_MIB_MAX;
5567 for_each_possible_cpu(c) {
5568 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5569 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5572 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5573 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5576 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5580 case IFLA_INET6_STATS:
5581 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5582 offsetof(struct ipstats_mib, syncp));
5584 case IFLA_INET6_ICMP6STATS:
5585 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5590 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5591 u32 ext_filter_mask)
5594 struct ifla_cacheinfo ci;
5596 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5597 goto nla_put_failure;
5598 ci.max_reasm_len = IPV6_MAXPLEN;
5599 ci.tstamp = cstamp_delta(idev->tstamp);
5600 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5601 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5602 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5603 goto nla_put_failure;
5604 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5606 goto nla_put_failure;
5607 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5609 /* XXX - MC not implemented */
5611 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5614 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5616 goto nla_put_failure;
5617 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5619 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5621 goto nla_put_failure;
5622 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5624 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5626 goto nla_put_failure;
5627 read_lock_bh(&idev->lock);
5628 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5629 read_unlock_bh(&idev->lock);
5631 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5632 goto nla_put_failure;
5640 static size_t inet6_get_link_af_size(const struct net_device *dev,
5641 u32 ext_filter_mask)
5643 if (!__in6_dev_get(dev))
5646 return inet6_ifla6_size();
5649 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5650 u32 ext_filter_mask)
5652 struct inet6_dev *idev = __in6_dev_get(dev);
5657 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5663 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5665 struct inet6_ifaddr *ifp;
5666 struct net_device *dev = idev->dev;
5667 bool clear_token, update_rs = false;
5668 struct in6_addr ll_addr;
5674 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5676 if (!ipv6_accept_ra(idev))
5678 if (idev->cnf.rtr_solicits == 0)
5681 write_lock_bh(&idev->lock);
5683 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5684 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5686 write_unlock_bh(&idev->lock);
5688 clear_token = ipv6_addr_any(token);
5692 if (!idev->dead && (idev->if_flags & IF_READY) &&
5693 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5694 IFA_F_OPTIMISTIC)) {
5695 /* If we're not ready, then normal ifup will take care
5696 * of this. Otherwise, we need to request our rs here.
5698 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5703 write_lock_bh(&idev->lock);
5706 idev->if_flags |= IF_RS_SENT;
5707 idev->rs_interval = rfc3315_s14_backoff_init(
5708 idev->cnf.rtr_solicit_interval);
5709 idev->rs_probes = 1;
5710 addrconf_mod_rs_timer(idev, idev->rs_interval);
5713 /* Well, that's kinda nasty ... */
5714 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5715 spin_lock(&ifp->lock);
5716 if (ifp->tokenized) {
5718 ifp->prefered_lft = 0;
5720 spin_unlock(&ifp->lock);
5723 write_unlock_bh(&idev->lock);
5724 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5725 addrconf_verify_rtnl();
5729 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5730 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5731 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5734 static int check_addr_gen_mode(int mode)
5736 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5737 mode != IN6_ADDR_GEN_MODE_NONE &&
5738 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5739 mode != IN6_ADDR_GEN_MODE_RANDOM)
5744 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5747 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5748 !idev->cnf.stable_secret.initialized &&
5749 !net->ipv6.devconf_dflt->stable_secret.initialized)
5754 static int inet6_validate_link_af(const struct net_device *dev,
5755 const struct nlattr *nla)
5757 struct nlattr *tb[IFLA_INET6_MAX + 1];
5758 struct inet6_dev *idev = NULL;
5762 idev = __in6_dev_get(dev);
5764 return -EAFNOSUPPORT;
5767 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5768 inet6_af_policy, NULL);
5772 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5775 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5776 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5778 if (check_addr_gen_mode(mode) < 0)
5780 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5787 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5789 struct inet6_dev *idev = __in6_dev_get(dev);
5790 struct nlattr *tb[IFLA_INET6_MAX + 1];
5794 return -EAFNOSUPPORT;
5796 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5799 if (tb[IFLA_INET6_TOKEN]) {
5800 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5805 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5806 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5808 idev->cnf.addr_gen_mode = mode;
5814 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5815 u32 portid, u32 seq, int event, unsigned int flags)
5817 struct net_device *dev = idev->dev;
5818 struct ifinfomsg *hdr;
5819 struct nlmsghdr *nlh;
5822 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5826 hdr = nlmsg_data(nlh);
5827 hdr->ifi_family = AF_INET6;
5829 hdr->ifi_type = dev->type;
5830 hdr->ifi_index = dev->ifindex;
5831 hdr->ifi_flags = dev_get_flags(dev);
5832 hdr->ifi_change = 0;
5834 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5836 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5837 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5838 (dev->ifindex != dev_get_iflink(dev) &&
5839 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5840 nla_put_u8(skb, IFLA_OPERSTATE,
5841 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5842 goto nla_put_failure;
5843 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5845 goto nla_put_failure;
5847 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5848 goto nla_put_failure;
5850 nla_nest_end(skb, protoinfo);
5851 nlmsg_end(skb, nlh);
5855 nlmsg_cancel(skb, nlh);
5859 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5860 struct netlink_ext_ack *extack)
5862 struct ifinfomsg *ifm;
5864 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5865 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5869 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5870 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5874 ifm = nlmsg_data(nlh);
5875 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5876 ifm->ifi_change || ifm->ifi_index) {
5877 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5884 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5886 struct net *net = sock_net(skb->sk);
5889 struct net_device *dev;
5890 struct inet6_dev *idev;
5891 struct hlist_head *head;
5893 /* only requests using strict checking can pass data to
5894 * influence the dump
5896 if (cb->strict_check) {
5897 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5904 s_idx = cb->args[1];
5907 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5909 head = &net->dev_index_head[h];
5910 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5913 idev = __in6_dev_get(dev);
5916 if (inet6_fill_ifinfo(skb, idev,
5917 NETLINK_CB(cb->skb).portid,
5919 RTM_NEWLINK, NLM_F_MULTI) < 0)
5933 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5935 struct sk_buff *skb;
5936 struct net *net = dev_net(idev->dev);
5939 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5943 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5945 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5946 WARN_ON(err == -EMSGSIZE);
5950 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5954 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5957 static inline size_t inet6_prefix_nlmsg_size(void)
5959 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5960 + nla_total_size(sizeof(struct in6_addr))
5961 + nla_total_size(sizeof(struct prefix_cacheinfo));
5964 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5965 struct prefix_info *pinfo, u32 portid, u32 seq,
5966 int event, unsigned int flags)
5968 struct prefixmsg *pmsg;
5969 struct nlmsghdr *nlh;
5970 struct prefix_cacheinfo ci;
5972 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5976 pmsg = nlmsg_data(nlh);
5977 pmsg->prefix_family = AF_INET6;
5978 pmsg->prefix_pad1 = 0;
5979 pmsg->prefix_pad2 = 0;
5980 pmsg->prefix_ifindex = idev->dev->ifindex;
5981 pmsg->prefix_len = pinfo->prefix_len;
5982 pmsg->prefix_type = pinfo->type;
5983 pmsg->prefix_pad3 = 0;
5984 pmsg->prefix_flags = 0;
5986 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5987 if (pinfo->autoconf)
5988 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5990 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5991 goto nla_put_failure;
5992 ci.preferred_time = ntohl(pinfo->prefered);
5993 ci.valid_time = ntohl(pinfo->valid);
5994 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5995 goto nla_put_failure;
5996 nlmsg_end(skb, nlh);
6000 nlmsg_cancel(skb, nlh);
6004 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6005 struct prefix_info *pinfo)
6007 struct sk_buff *skb;
6008 struct net *net = dev_net(idev->dev);
6011 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6015 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6017 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6018 WARN_ON(err == -EMSGSIZE);
6022 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6026 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6029 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6031 struct net *net = dev_net(ifp->idev->dev);
6036 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6041 * If the address was optimistic we inserted the route at the
6042 * start of our DAD process, so we don't need to do it again.
6043 * If the device was taken down in the middle of the DAD
6044 * cycle there is a race where we could get here without a
6045 * host route, so nothing to insert. That will be fixed when
6046 * the device is brought up.
6048 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6049 ip6_ins_rt(net, ifp->rt);
6050 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6051 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6052 &ifp->addr, ifp->idev->dev->name);
6055 if (ifp->idev->cnf.forwarding)
6056 addrconf_join_anycast(ifp);
6057 if (!ipv6_addr_any(&ifp->peer_addr))
6058 addrconf_prefix_route(&ifp->peer_addr, 128,
6059 ifp->rt_priority, ifp->idev->dev,
6063 if (ifp->idev->cnf.forwarding)
6064 addrconf_leave_anycast(ifp);
6065 addrconf_leave_solict(ifp->idev, &ifp->addr);
6066 if (!ipv6_addr_any(&ifp->peer_addr)) {
6067 struct fib6_info *rt;
6069 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6070 ifp->idev->dev, 0, 0,
6073 ip6_del_rt(net, rt, false);
6076 ip6_del_rt(net, ifp->rt, false);
6079 rt_genid_bump_ipv6(net);
6082 atomic_inc(&net->ipv6.dev_addr_genid);
6085 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6088 if (likely(ifp->idev->dead == 0))
6089 __ipv6_ifa_notify(event, ifp);
6090 rcu_read_unlock_bh();
6093 #ifdef CONFIG_SYSCTL
6095 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6096 void *buffer, size_t *lenp, loff_t *ppos)
6098 int *valp = ctl->data;
6101 struct ctl_table lctl;
6105 * ctl->data points to idev->cnf.forwarding, we should
6106 * not modify it until we get the rtnl lock.
6111 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6114 ret = addrconf_fixup_forwarding(ctl, valp, val);
6120 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6121 void *buffer, size_t *lenp, loff_t *ppos)
6123 struct inet6_dev *idev = ctl->extra1;
6124 int min_mtu = IPV6_MIN_MTU;
6125 struct ctl_table lctl;
6128 lctl.extra1 = &min_mtu;
6129 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6131 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6134 static void dev_disable_change(struct inet6_dev *idev)
6136 struct netdev_notifier_info info;
6138 if (!idev || !idev->dev)
6141 netdev_notifier_info_init(&info, idev->dev);
6142 if (idev->cnf.disable_ipv6)
6143 addrconf_notify(NULL, NETDEV_DOWN, &info);
6145 addrconf_notify(NULL, NETDEV_UP, &info);
6148 static void addrconf_disable_change(struct net *net, __s32 newf)
6150 struct net_device *dev;
6151 struct inet6_dev *idev;
6153 for_each_netdev(net, dev) {
6154 idev = __in6_dev_get(dev);
6156 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6157 idev->cnf.disable_ipv6 = newf;
6159 dev_disable_change(idev);
6164 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6169 if (!rtnl_trylock())
6170 return restart_syscall();
6172 net = (struct net *)table->extra2;
6176 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6181 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6182 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6183 addrconf_disable_change(net, newf);
6184 } else if ((!newf) ^ (!old))
6185 dev_disable_change((struct inet6_dev *)table->extra1);
6191 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6192 void *buffer, size_t *lenp, loff_t *ppos)
6194 int *valp = ctl->data;
6197 struct ctl_table lctl;
6201 * ctl->data points to idev->cnf.disable_ipv6, we should
6202 * not modify it until we get the rtnl lock.
6207 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6210 ret = addrconf_disable_ipv6(ctl, valp, val);
6216 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6217 void *buffer, size_t *lenp, loff_t *ppos)
6219 int *valp = ctl->data;
6224 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6227 if (write && old != new) {
6228 struct net *net = ctl->extra2;
6230 if (!rtnl_trylock())
6231 return restart_syscall();
6233 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6234 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6235 NETCONFA_PROXY_NEIGH,
6236 NETCONFA_IFINDEX_DEFAULT,
6237 net->ipv6.devconf_dflt);
6238 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6239 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6240 NETCONFA_PROXY_NEIGH,
6241 NETCONFA_IFINDEX_ALL,
6242 net->ipv6.devconf_all);
6244 struct inet6_dev *idev = ctl->extra1;
6246 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6247 NETCONFA_PROXY_NEIGH,
6257 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6258 void *buffer, size_t *lenp,
6263 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6264 struct net *net = (struct net *)ctl->extra2;
6265 struct ctl_table tmp = {
6267 .maxlen = sizeof(new_val),
6271 if (!rtnl_trylock())
6272 return restart_syscall();
6274 new_val = *((u32 *)ctl->data);
6276 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6281 if (check_addr_gen_mode(new_val) < 0) {
6287 if (check_stable_privacy(idev, net, new_val) < 0) {
6292 if (idev->cnf.addr_gen_mode != new_val) {
6293 idev->cnf.addr_gen_mode = new_val;
6294 addrconf_dev_config(idev->dev);
6296 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6297 struct net_device *dev;
6299 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6300 for_each_netdev(net, dev) {
6301 idev = __in6_dev_get(dev);
6303 idev->cnf.addr_gen_mode != new_val) {
6304 idev->cnf.addr_gen_mode = new_val;
6305 addrconf_dev_config(idev->dev);
6310 *((u32 *)ctl->data) = new_val;
6319 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6320 void *buffer, size_t *lenp,
6324 struct in6_addr addr;
6325 char str[IPV6_MAX_STRLEN];
6326 struct ctl_table lctl = *ctl;
6327 struct net *net = ctl->extra2;
6328 struct ipv6_stable_secret *secret = ctl->data;
6330 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6333 lctl.maxlen = IPV6_MAX_STRLEN;
6336 if (!rtnl_trylock())
6337 return restart_syscall();
6339 if (!write && !secret->initialized) {
6344 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6345 if (err >= sizeof(str)) {
6350 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6354 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6359 secret->initialized = true;
6360 secret->secret = addr;
6362 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6363 struct net_device *dev;
6365 for_each_netdev(net, dev) {
6366 struct inet6_dev *idev = __in6_dev_get(dev);
6369 idev->cnf.addr_gen_mode =
6370 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6374 struct inet6_dev *idev = ctl->extra1;
6376 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6386 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6387 int write, void *buffer,
6391 int *valp = ctl->data;
6394 struct ctl_table lctl;
6397 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6398 * we should not modify it until we get the rtnl lock.
6403 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6406 ret = addrconf_fixup_linkdown(ctl, valp, val);
6413 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6417 rt->dst.flags |= DST_NOPOLICY;
6419 rt->dst.flags &= ~DST_NOPOLICY;
6424 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6426 struct inet6_ifaddr *ifa;
6428 read_lock_bh(&idev->lock);
6429 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6430 spin_lock(&ifa->lock);
6432 /* host routes only use builtin fib6_nh */
6433 struct fib6_nh *nh = ifa->rt->fib6_nh;
6437 ifa->rt->dst_nopolicy = val ? true : false;
6438 if (nh->rt6i_pcpu) {
6439 for_each_possible_cpu(cpu) {
6440 struct rt6_info **rtp;
6442 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6443 addrconf_set_nopolicy(*rtp, val);
6448 spin_unlock(&ifa->lock);
6450 read_unlock_bh(&idev->lock);
6454 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6456 struct inet6_dev *idev;
6459 if (!rtnl_trylock())
6460 return restart_syscall();
6464 net = (struct net *)ctl->extra2;
6465 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6470 if (valp == &net->ipv6.devconf_all->disable_policy) {
6471 struct net_device *dev;
6473 for_each_netdev(net, dev) {
6474 idev = __in6_dev_get(dev);
6476 addrconf_disable_policy_idev(idev, val);
6479 idev = (struct inet6_dev *)ctl->extra1;
6480 addrconf_disable_policy_idev(idev, val);
6487 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6488 void *buffer, size_t *lenp, loff_t *ppos)
6490 int *valp = ctl->data;
6493 struct ctl_table lctl;
6498 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6500 if (write && (*valp != val))
6501 ret = addrconf_disable_policy(ctl, valp, val);
6509 static int minus_one = -1;
6510 static const int two_five_five = 255;
6512 static const struct ctl_table addrconf_sysctl[] = {
6514 .procname = "forwarding",
6515 .data = &ipv6_devconf.forwarding,
6516 .maxlen = sizeof(int),
6518 .proc_handler = addrconf_sysctl_forward,
6521 .procname = "hop_limit",
6522 .data = &ipv6_devconf.hop_limit,
6523 .maxlen = sizeof(int),
6525 .proc_handler = proc_dointvec_minmax,
6526 .extra1 = (void *)SYSCTL_ONE,
6527 .extra2 = (void *)&two_five_five,
6531 .data = &ipv6_devconf.mtu6,
6532 .maxlen = sizeof(int),
6534 .proc_handler = addrconf_sysctl_mtu,
6537 .procname = "accept_ra",
6538 .data = &ipv6_devconf.accept_ra,
6539 .maxlen = sizeof(int),
6541 .proc_handler = proc_dointvec,
6544 .procname = "accept_redirects",
6545 .data = &ipv6_devconf.accept_redirects,
6546 .maxlen = sizeof(int),
6548 .proc_handler = proc_dointvec,
6551 .procname = "autoconf",
6552 .data = &ipv6_devconf.autoconf,
6553 .maxlen = sizeof(int),
6555 .proc_handler = proc_dointvec,
6558 .procname = "dad_transmits",
6559 .data = &ipv6_devconf.dad_transmits,
6560 .maxlen = sizeof(int),
6562 .proc_handler = proc_dointvec,
6565 .procname = "router_solicitations",
6566 .data = &ipv6_devconf.rtr_solicits,
6567 .maxlen = sizeof(int),
6569 .proc_handler = proc_dointvec_minmax,
6570 .extra1 = &minus_one,
6573 .procname = "router_solicitation_interval",
6574 .data = &ipv6_devconf.rtr_solicit_interval,
6575 .maxlen = sizeof(int),
6577 .proc_handler = proc_dointvec_jiffies,
6580 .procname = "router_solicitation_max_interval",
6581 .data = &ipv6_devconf.rtr_solicit_max_interval,
6582 .maxlen = sizeof(int),
6584 .proc_handler = proc_dointvec_jiffies,
6587 .procname = "router_solicitation_delay",
6588 .data = &ipv6_devconf.rtr_solicit_delay,
6589 .maxlen = sizeof(int),
6591 .proc_handler = proc_dointvec_jiffies,
6594 .procname = "force_mld_version",
6595 .data = &ipv6_devconf.force_mld_version,
6596 .maxlen = sizeof(int),
6598 .proc_handler = proc_dointvec,
6601 .procname = "mldv1_unsolicited_report_interval",
6603 &ipv6_devconf.mldv1_unsolicited_report_interval,
6604 .maxlen = sizeof(int),
6606 .proc_handler = proc_dointvec_ms_jiffies,
6609 .procname = "mldv2_unsolicited_report_interval",
6611 &ipv6_devconf.mldv2_unsolicited_report_interval,
6612 .maxlen = sizeof(int),
6614 .proc_handler = proc_dointvec_ms_jiffies,
6617 .procname = "use_tempaddr",
6618 .data = &ipv6_devconf.use_tempaddr,
6619 .maxlen = sizeof(int),
6621 .proc_handler = proc_dointvec,
6624 .procname = "temp_valid_lft",
6625 .data = &ipv6_devconf.temp_valid_lft,
6626 .maxlen = sizeof(int),
6628 .proc_handler = proc_dointvec,
6631 .procname = "temp_prefered_lft",
6632 .data = &ipv6_devconf.temp_prefered_lft,
6633 .maxlen = sizeof(int),
6635 .proc_handler = proc_dointvec,
6638 .procname = "regen_max_retry",
6639 .data = &ipv6_devconf.regen_max_retry,
6640 .maxlen = sizeof(int),
6642 .proc_handler = proc_dointvec,
6645 .procname = "max_desync_factor",
6646 .data = &ipv6_devconf.max_desync_factor,
6647 .maxlen = sizeof(int),
6649 .proc_handler = proc_dointvec,
6652 .procname = "max_addresses",
6653 .data = &ipv6_devconf.max_addresses,
6654 .maxlen = sizeof(int),
6656 .proc_handler = proc_dointvec,
6659 .procname = "accept_ra_defrtr",
6660 .data = &ipv6_devconf.accept_ra_defrtr,
6661 .maxlen = sizeof(int),
6663 .proc_handler = proc_dointvec,
6666 .procname = "accept_ra_min_hop_limit",
6667 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6668 .maxlen = sizeof(int),
6670 .proc_handler = proc_dointvec,
6673 .procname = "accept_ra_pinfo",
6674 .data = &ipv6_devconf.accept_ra_pinfo,
6675 .maxlen = sizeof(int),
6677 .proc_handler = proc_dointvec,
6679 #ifdef CONFIG_IPV6_ROUTER_PREF
6681 .procname = "accept_ra_rtr_pref",
6682 .data = &ipv6_devconf.accept_ra_rtr_pref,
6683 .maxlen = sizeof(int),
6685 .proc_handler = proc_dointvec,
6688 .procname = "router_probe_interval",
6689 .data = &ipv6_devconf.rtr_probe_interval,
6690 .maxlen = sizeof(int),
6692 .proc_handler = proc_dointvec_jiffies,
6694 #ifdef CONFIG_IPV6_ROUTE_INFO
6696 .procname = "accept_ra_rt_info_min_plen",
6697 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6698 .maxlen = sizeof(int),
6700 .proc_handler = proc_dointvec,
6703 .procname = "accept_ra_rt_info_max_plen",
6704 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6705 .maxlen = sizeof(int),
6707 .proc_handler = proc_dointvec,
6712 .procname = "proxy_ndp",
6713 .data = &ipv6_devconf.proxy_ndp,
6714 .maxlen = sizeof(int),
6716 .proc_handler = addrconf_sysctl_proxy_ndp,
6719 .procname = "accept_source_route",
6720 .data = &ipv6_devconf.accept_source_route,
6721 .maxlen = sizeof(int),
6723 .proc_handler = proc_dointvec,
6725 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6727 .procname = "optimistic_dad",
6728 .data = &ipv6_devconf.optimistic_dad,
6729 .maxlen = sizeof(int),
6731 .proc_handler = proc_dointvec,
6734 .procname = "use_optimistic",
6735 .data = &ipv6_devconf.use_optimistic,
6736 .maxlen = sizeof(int),
6738 .proc_handler = proc_dointvec,
6741 #ifdef CONFIG_IPV6_MROUTE
6743 .procname = "mc_forwarding",
6744 .data = &ipv6_devconf.mc_forwarding,
6745 .maxlen = sizeof(int),
6747 .proc_handler = proc_dointvec,
6751 .procname = "disable_ipv6",
6752 .data = &ipv6_devconf.disable_ipv6,
6753 .maxlen = sizeof(int),
6755 .proc_handler = addrconf_sysctl_disable,
6758 .procname = "accept_dad",
6759 .data = &ipv6_devconf.accept_dad,
6760 .maxlen = sizeof(int),
6762 .proc_handler = proc_dointvec,
6765 .procname = "force_tllao",
6766 .data = &ipv6_devconf.force_tllao,
6767 .maxlen = sizeof(int),
6769 .proc_handler = proc_dointvec
6772 .procname = "ndisc_notify",
6773 .data = &ipv6_devconf.ndisc_notify,
6774 .maxlen = sizeof(int),
6776 .proc_handler = proc_dointvec
6779 .procname = "suppress_frag_ndisc",
6780 .data = &ipv6_devconf.suppress_frag_ndisc,
6781 .maxlen = sizeof(int),
6783 .proc_handler = proc_dointvec
6786 .procname = "accept_ra_from_local",
6787 .data = &ipv6_devconf.accept_ra_from_local,
6788 .maxlen = sizeof(int),
6790 .proc_handler = proc_dointvec,
6793 .procname = "accept_ra_mtu",
6794 .data = &ipv6_devconf.accept_ra_mtu,
6795 .maxlen = sizeof(int),
6797 .proc_handler = proc_dointvec,
6800 .procname = "stable_secret",
6801 .data = &ipv6_devconf.stable_secret,
6802 .maxlen = IPV6_MAX_STRLEN,
6804 .proc_handler = addrconf_sysctl_stable_secret,
6807 .procname = "use_oif_addrs_only",
6808 .data = &ipv6_devconf.use_oif_addrs_only,
6809 .maxlen = sizeof(int),
6811 .proc_handler = proc_dointvec,
6814 .procname = "ignore_routes_with_linkdown",
6815 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6816 .maxlen = sizeof(int),
6818 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6821 .procname = "drop_unicast_in_l2_multicast",
6822 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6823 .maxlen = sizeof(int),
6825 .proc_handler = proc_dointvec,
6828 .procname = "drop_unsolicited_na",
6829 .data = &ipv6_devconf.drop_unsolicited_na,
6830 .maxlen = sizeof(int),
6832 .proc_handler = proc_dointvec,
6835 .procname = "keep_addr_on_down",
6836 .data = &ipv6_devconf.keep_addr_on_down,
6837 .maxlen = sizeof(int),
6839 .proc_handler = proc_dointvec,
6843 .procname = "seg6_enabled",
6844 .data = &ipv6_devconf.seg6_enabled,
6845 .maxlen = sizeof(int),
6847 .proc_handler = proc_dointvec,
6849 #ifdef CONFIG_IPV6_SEG6_HMAC
6851 .procname = "seg6_require_hmac",
6852 .data = &ipv6_devconf.seg6_require_hmac,
6853 .maxlen = sizeof(int),
6855 .proc_handler = proc_dointvec,
6859 .procname = "enhanced_dad",
6860 .data = &ipv6_devconf.enhanced_dad,
6861 .maxlen = sizeof(int),
6863 .proc_handler = proc_dointvec,
6866 .procname = "addr_gen_mode",
6867 .data = &ipv6_devconf.addr_gen_mode,
6868 .maxlen = sizeof(int),
6870 .proc_handler = addrconf_sysctl_addr_gen_mode,
6873 .procname = "disable_policy",
6874 .data = &ipv6_devconf.disable_policy,
6875 .maxlen = sizeof(int),
6877 .proc_handler = addrconf_sysctl_disable_policy,
6880 .procname = "ndisc_tclass",
6881 .data = &ipv6_devconf.ndisc_tclass,
6882 .maxlen = sizeof(int),
6884 .proc_handler = proc_dointvec_minmax,
6885 .extra1 = (void *)SYSCTL_ZERO,
6886 .extra2 = (void *)&two_five_five,
6889 .procname = "rpl_seg_enabled",
6890 .data = &ipv6_devconf.rpl_seg_enabled,
6891 .maxlen = sizeof(int),
6893 .proc_handler = proc_dointvec,
6900 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6901 struct inet6_dev *idev, struct ipv6_devconf *p)
6904 struct ctl_table *table;
6905 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6907 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6911 for (i = 0; table[i].data; i++) {
6912 table[i].data += (char *)p - (char *)&ipv6_devconf;
6913 /* If one of these is already set, then it is not safe to
6914 * overwrite either of them: this makes proc_dointvec_minmax
6917 if (!table[i].extra1 && !table[i].extra2) {
6918 table[i].extra1 = idev; /* embedded; no ref */
6919 table[i].extra2 = net;
6923 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6925 p->sysctl_header = register_net_sysctl(net, path, table);
6926 if (!p->sysctl_header)
6929 if (!strcmp(dev_name, "all"))
6930 ifindex = NETCONFA_IFINDEX_ALL;
6931 else if (!strcmp(dev_name, "default"))
6932 ifindex = NETCONFA_IFINDEX_DEFAULT;
6934 ifindex = idev->dev->ifindex;
6935 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6945 static void __addrconf_sysctl_unregister(struct net *net,
6946 struct ipv6_devconf *p, int ifindex)
6948 struct ctl_table *table;
6950 if (!p->sysctl_header)
6953 table = p->sysctl_header->ctl_table_arg;
6954 unregister_net_sysctl_table(p->sysctl_header);
6955 p->sysctl_header = NULL;
6958 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6961 static int addrconf_sysctl_register(struct inet6_dev *idev)
6965 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6968 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6969 &ndisc_ifinfo_sysctl_change);
6972 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6975 neigh_sysctl_unregister(idev->nd_parms);
6980 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6982 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6983 idev->dev->ifindex);
6984 neigh_sysctl_unregister(idev->nd_parms);
6990 static int __net_init addrconf_init_net(struct net *net)
6993 struct ipv6_devconf *all, *dflt;
6995 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6999 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7001 goto err_alloc_dflt;
7003 if (IS_ENABLED(CONFIG_SYSCTL) &&
7004 !net_eq(net, &init_net)) {
7005 switch (sysctl_devconf_inherit_init_net) {
7006 case 1: /* copy from init_net */
7007 memcpy(all, init_net.ipv6.devconf_all,
7008 sizeof(ipv6_devconf));
7009 memcpy(dflt, init_net.ipv6.devconf_dflt,
7010 sizeof(ipv6_devconf_dflt));
7012 case 3: /* copy from the current netns */
7013 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7014 sizeof(ipv6_devconf));
7016 current->nsproxy->net_ns->ipv6.devconf_dflt,
7017 sizeof(ipv6_devconf_dflt));
7021 /* use compiled values */
7026 /* these will be inherited by all namespaces */
7027 dflt->autoconf = ipv6_defaults.autoconf;
7028 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7030 dflt->stable_secret.initialized = false;
7031 all->stable_secret.initialized = false;
7033 net->ipv6.devconf_all = all;
7034 net->ipv6.devconf_dflt = dflt;
7036 #ifdef CONFIG_SYSCTL
7037 err = __addrconf_sysctl_register(net, "all", NULL, all);
7041 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7047 #ifdef CONFIG_SYSCTL
7049 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7059 static void __net_exit addrconf_exit_net(struct net *net)
7061 #ifdef CONFIG_SYSCTL
7062 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7063 NETCONFA_IFINDEX_DEFAULT);
7064 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7065 NETCONFA_IFINDEX_ALL);
7067 kfree(net->ipv6.devconf_dflt);
7068 kfree(net->ipv6.devconf_all);
7071 static struct pernet_operations addrconf_ops = {
7072 .init = addrconf_init_net,
7073 .exit = addrconf_exit_net,
7076 static struct rtnl_af_ops inet6_ops __read_mostly = {
7078 .fill_link_af = inet6_fill_link_af,
7079 .get_link_af_size = inet6_get_link_af_size,
7080 .validate_link_af = inet6_validate_link_af,
7081 .set_link_af = inet6_set_link_af,
7085 * Init / cleanup code
7088 int __init addrconf_init(void)
7090 struct inet6_dev *idev;
7093 err = ipv6_addr_label_init();
7095 pr_crit("%s: cannot initialize default policy table: %d\n",
7100 err = register_pernet_subsys(&addrconf_ops);
7104 addrconf_wq = create_workqueue("ipv6_addrconf");
7110 /* The addrconf netdev notifier requires that loopback_dev
7111 * has it's ipv6 private information allocated and setup
7112 * before it can bring up and give link-local addresses
7113 * to other devices which are up.
7115 * Unfortunately, loopback_dev is not necessarily the first
7116 * entry in the global dev_base list of net devices. In fact,
7117 * it is likely to be the very last entry on that list.
7118 * So this causes the notifier registry below to try and
7119 * give link-local addresses to all devices besides loopback_dev
7120 * first, then loopback_dev, which cases all the non-loopback_dev
7121 * devices to fail to get a link-local address.
7123 * So, as a temporary fix, allocate the ipv6 structure for
7124 * loopback_dev first by hand.
7125 * Longer term, all of the dependencies ipv6 has upon the loopback
7126 * device and it being up should be removed.
7129 idev = ipv6_add_dev(init_net.loopback_dev);
7132 err = PTR_ERR(idev);
7136 ip6_route_init_special_entries();
7138 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7139 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7141 register_netdevice_notifier(&ipv6_dev_notf);
7145 rtnl_af_register(&inet6_ops);
7147 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7148 NULL, inet6_dump_ifinfo, 0);
7152 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7153 inet6_rtm_newaddr, NULL, 0);
7156 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7157 inet6_rtm_deladdr, NULL, 0);
7160 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7161 inet6_rtm_getaddr, inet6_dump_ifaddr,
7162 RTNL_FLAG_DOIT_UNLOCKED);
7165 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7166 NULL, inet6_dump_ifmcaddr, 0);
7169 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7170 NULL, inet6_dump_ifacaddr, 0);
7173 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7174 inet6_netconf_get_devconf,
7175 inet6_netconf_dump_devconf,
7176 RTNL_FLAG_DOIT_UNLOCKED);
7179 err = ipv6_addr_label_rtnl_register();
7185 rtnl_unregister_all(PF_INET6);
7186 rtnl_af_unregister(&inet6_ops);
7187 unregister_netdevice_notifier(&ipv6_dev_notf);
7189 destroy_workqueue(addrconf_wq);
7191 unregister_pernet_subsys(&addrconf_ops);
7193 ipv6_addr_label_cleanup();
7198 void addrconf_cleanup(void)
7200 struct net_device *dev;
7203 unregister_netdevice_notifier(&ipv6_dev_notf);
7204 unregister_pernet_subsys(&addrconf_ops);
7205 ipv6_addr_label_cleanup();
7207 rtnl_af_unregister(&inet6_ops);
7211 /* clean dev list */
7212 for_each_netdev(&init_net, dev) {
7213 if (__in6_dev_get(dev) == NULL)
7215 addrconf_ifdown(dev, true);
7217 addrconf_ifdown(init_net.loopback_dev, true);
7222 spin_lock_bh(&addrconf_hash_lock);
7223 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7224 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7225 spin_unlock_bh(&addrconf_hash_lock);
7226 cancel_delayed_work(&addr_chk_work);
7229 destroy_workqueue(addrconf_wq);