2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
62 #include <linux/sysctl.h>
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
70 #include <net/net_namespace.h>
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
97 #define INFINITY_LIFE_TIME 0xFFFFFFFF
99 #define IPV6_MAX_STRLEN \
100 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
102 static inline u32 cstamp_delta(unsigned long cstamp)
104 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
109 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
110 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
111 do_div(tmp, 1000000);
115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
117 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
118 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
119 do_div(tmp, 1000000);
120 if ((s32)tmp > mrt) {
121 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
122 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
123 do_div(tmp, 1000000);
129 static int addrconf_sysctl_register(struct inet6_dev *idev);
130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
142 static void ipv6_regen_rndid(struct inet6_dev *idev);
143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
146 static int ipv6_count_addresses(const struct inet6_dev *idev);
147 static int ipv6_generate_stable_address(struct in6_addr *addr,
149 const struct inet6_dev *idev);
151 #define IN6_ADDR_HSIZE_SHIFT 8
152 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
154 * Configured unicast address hash table
156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
157 static DEFINE_SPINLOCK(addrconf_hash_lock);
159 static void addrconf_verify(void);
160 static void addrconf_verify_rtnl(void);
161 static void addrconf_verify_work(struct work_struct *);
163 static struct workqueue_struct *addrconf_wq;
164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
169 static void addrconf_type_change(struct net_device *dev,
170 unsigned long event);
171 static int addrconf_ifdown(struct net_device *dev, int how);
173 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
175 const struct net_device *dev,
176 u32 flags, u32 noflags);
178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
179 static void addrconf_dad_work(struct work_struct *w);
180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
182 static void addrconf_dad_run(struct inet6_dev *idev);
183 static void addrconf_rs_timer(struct timer_list *t);
184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
188 struct prefix_info *pinfo);
190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
192 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
193 .mtu6 = IPV6_MIN_MTU,
195 .accept_redirects = 1,
197 .force_mld_version = 0,
198 .mldv1_unsolicited_report_interval = 10 * HZ,
199 .mldv2_unsolicited_report_interval = HZ,
201 .rtr_solicits = MAX_RTR_SOLICITATIONS,
202 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
203 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
204 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
206 .temp_valid_lft = TEMP_VALID_LIFETIME,
207 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
208 .regen_max_retry = REGEN_MAX_RETRY,
209 .max_desync_factor = MAX_DESYNC_FACTOR,
210 .max_addresses = IPV6_MAX_ADDRESSES,
211 .accept_ra_defrtr = 1,
212 .accept_ra_from_local = 0,
213 .accept_ra_min_hop_limit= 1,
214 .accept_ra_pinfo = 1,
215 #ifdef CONFIG_IPV6_ROUTER_PREF
216 .accept_ra_rtr_pref = 1,
217 .rtr_probe_interval = 60 * HZ,
218 #ifdef CONFIG_IPV6_ROUTE_INFO
219 .accept_ra_rt_info_min_plen = 0,
220 .accept_ra_rt_info_max_plen = 0,
224 .accept_source_route = 0, /* we do not accept RH0 by default. */
227 .suppress_frag_ndisc = 1,
230 .initialized = false,
232 .use_oif_addrs_only = 0,
233 .ignore_routes_with_linkdown = 0,
234 .keep_addr_on_down = 0,
236 #ifdef CONFIG_IPV6_SEG6_HMAC
237 .seg6_require_hmac = 0,
240 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
246 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
247 .mtu6 = IPV6_MIN_MTU,
249 .accept_redirects = 1,
251 .force_mld_version = 0,
252 .mldv1_unsolicited_report_interval = 10 * HZ,
253 .mldv2_unsolicited_report_interval = HZ,
255 .rtr_solicits = MAX_RTR_SOLICITATIONS,
256 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
257 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
258 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
260 .temp_valid_lft = TEMP_VALID_LIFETIME,
261 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
262 .regen_max_retry = REGEN_MAX_RETRY,
263 .max_desync_factor = MAX_DESYNC_FACTOR,
264 .max_addresses = IPV6_MAX_ADDRESSES,
265 .accept_ra_defrtr = 1,
266 .accept_ra_from_local = 0,
267 .accept_ra_min_hop_limit= 1,
268 .accept_ra_pinfo = 1,
269 #ifdef CONFIG_IPV6_ROUTER_PREF
270 .accept_ra_rtr_pref = 1,
271 .rtr_probe_interval = 60 * HZ,
272 #ifdef CONFIG_IPV6_ROUTE_INFO
273 .accept_ra_rt_info_min_plen = 0,
274 .accept_ra_rt_info_max_plen = 0,
278 .accept_source_route = 0, /* we do not accept RH0 by default. */
281 .suppress_frag_ndisc = 1,
284 .initialized = false,
286 .use_oif_addrs_only = 0,
287 .ignore_routes_with_linkdown = 0,
288 .keep_addr_on_down = 0,
290 #ifdef CONFIG_IPV6_SEG6_HMAC
291 .seg6_require_hmac = 0,
294 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
298 /* Check if link is ready: is it up and is a valid qdisc available */
299 static inline bool addrconf_link_ready(const struct net_device *dev)
301 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
306 if (del_timer(&idev->rs_timer))
310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
312 if (cancel_delayed_work(&ifp->dad_work))
316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
319 if (!timer_pending(&idev->rs_timer))
321 mod_timer(&idev->rs_timer, jiffies + when);
324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
328 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
332 static int snmp6_alloc_dev(struct inet6_dev *idev)
336 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
337 if (!idev->stats.ipv6)
340 for_each_possible_cpu(i) {
341 struct ipstats_mib *addrconf_stats;
342 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
343 u64_stats_init(&addrconf_stats->syncp);
347 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
349 if (!idev->stats.icmpv6dev)
351 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
353 if (!idev->stats.icmpv6msgdev)
359 kfree(idev->stats.icmpv6dev);
361 free_percpu(idev->stats.ipv6);
366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
368 struct inet6_dev *ndev;
373 if (dev->mtu < IPV6_MIN_MTU)
374 return ERR_PTR(-EINVAL);
376 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
380 rwlock_init(&ndev->lock);
382 INIT_LIST_HEAD(&ndev->addr_list);
383 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
384 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
386 if (ndev->cnf.stable_secret.initialized)
387 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
389 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
391 ndev->cnf.mtu6 = dev->mtu;
392 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
393 if (!ndev->nd_parms) {
397 if (ndev->cnf.forwarding)
398 dev_disable_lro(dev);
399 /* We refer to the device */
402 if (snmp6_alloc_dev(ndev) < 0) {
403 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
405 neigh_parms_release(&nd_tbl, ndev->nd_parms);
411 if (snmp6_register_dev(ndev) < 0) {
412 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
413 __func__, dev->name);
417 /* One reference from device. */
418 refcount_set(&ndev->refcnt, 1);
420 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
421 ndev->cnf.accept_dad = -1;
423 #if IS_ENABLED(CONFIG_IPV6_SIT)
424 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
425 pr_info("%s: Disabled Multicast RS\n", dev->name);
426 ndev->cnf.rtr_solicits = 0;
430 INIT_LIST_HEAD(&ndev->tempaddr_list);
431 ndev->desync_factor = U32_MAX;
432 if ((dev->flags&IFF_LOOPBACK) ||
433 dev->type == ARPHRD_TUNNEL ||
434 dev->type == ARPHRD_TUNNEL6 ||
435 dev->type == ARPHRD_SIT ||
436 dev->type == ARPHRD_NONE) {
437 ndev->cnf.use_tempaddr = -1;
439 ipv6_regen_rndid(ndev);
441 ndev->token = in6addr_any;
443 if (netif_running(dev) && addrconf_link_ready(dev))
444 ndev->if_flags |= IF_READY;
446 ipv6_mc_init_dev(ndev);
447 ndev->tstamp = jiffies;
448 err = addrconf_sysctl_register(ndev);
450 ipv6_mc_destroy_dev(ndev);
451 snmp6_unregister_dev(ndev);
454 /* protected by rtnl_lock */
455 rcu_assign_pointer(dev->ip6_ptr, ndev);
457 /* Join interface-local all-node multicast group */
458 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
460 /* Join all-node multicast group */
461 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
463 /* Join all-router multicast group if forwarding is set */
464 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
465 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
470 neigh_parms_release(&nd_tbl, ndev->nd_parms);
472 in6_dev_finish_destroy(ndev);
476 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
478 struct inet6_dev *idev;
482 idev = __in6_dev_get(dev);
484 idev = ipv6_add_dev(dev);
489 if (dev->flags&IFF_UP)
494 static int inet6_netconf_msgsize_devconf(int type)
496 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
497 + nla_total_size(4); /* NETCONFA_IFINDEX */
500 if (type == NETCONFA_ALL)
503 if (all || type == NETCONFA_FORWARDING)
504 size += nla_total_size(4);
505 #ifdef CONFIG_IPV6_MROUTE
506 if (all || type == NETCONFA_MC_FORWARDING)
507 size += nla_total_size(4);
509 if (all || type == NETCONFA_PROXY_NEIGH)
510 size += nla_total_size(4);
512 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
513 size += nla_total_size(4);
518 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
519 struct ipv6_devconf *devconf, u32 portid,
520 u32 seq, int event, unsigned int flags,
523 struct nlmsghdr *nlh;
524 struct netconfmsg *ncm;
527 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
532 if (type == NETCONFA_ALL)
535 ncm = nlmsg_data(nlh);
536 ncm->ncm_family = AF_INET6;
538 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
539 goto nla_put_failure;
544 if ((all || type == NETCONFA_FORWARDING) &&
545 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
546 goto nla_put_failure;
547 #ifdef CONFIG_IPV6_MROUTE
548 if ((all || type == NETCONFA_MC_FORWARDING) &&
549 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
550 devconf->mc_forwarding) < 0)
551 goto nla_put_failure;
553 if ((all || type == NETCONFA_PROXY_NEIGH) &&
554 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
555 goto nla_put_failure;
557 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
558 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
559 devconf->ignore_routes_with_linkdown) < 0)
560 goto nla_put_failure;
567 nlmsg_cancel(skb, nlh);
571 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
572 int ifindex, struct ipv6_devconf *devconf)
577 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
581 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
584 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
585 WARN_ON(err == -EMSGSIZE);
589 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
592 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
595 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
596 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
597 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
598 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
599 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
602 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
603 struct nlmsghdr *nlh,
604 struct netlink_ext_ack *extack)
606 struct net *net = sock_net(in_skb->sk);
607 struct nlattr *tb[NETCONFA_MAX+1];
608 struct inet6_dev *in6_dev = NULL;
609 struct net_device *dev = NULL;
610 struct netconfmsg *ncm;
612 struct ipv6_devconf *devconf;
616 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
617 devconf_ipv6_policy, extack);
621 if (!tb[NETCONFA_IFINDEX])
625 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
627 case NETCONFA_IFINDEX_ALL:
628 devconf = net->ipv6.devconf_all;
630 case NETCONFA_IFINDEX_DEFAULT:
631 devconf = net->ipv6.devconf_dflt;
634 dev = dev_get_by_index(net, ifindex);
637 in6_dev = in6_dev_get(dev);
640 devconf = &in6_dev->cnf;
645 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
649 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
650 NETLINK_CB(in_skb).portid,
651 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
654 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
655 WARN_ON(err == -EMSGSIZE);
659 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
662 in6_dev_put(in6_dev);
668 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
669 struct netlink_callback *cb)
671 struct net *net = sock_net(skb->sk);
674 struct net_device *dev;
675 struct inet6_dev *idev;
676 struct hlist_head *head;
679 s_idx = idx = cb->args[1];
681 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
683 head = &net->dev_index_head[h];
685 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
687 hlist_for_each_entry_rcu(dev, head, index_hlist) {
690 idev = __in6_dev_get(dev);
694 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
696 NETLINK_CB(cb->skb).portid,
704 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
710 if (h == NETDEV_HASHENTRIES) {
711 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
712 net->ipv6.devconf_all,
713 NETLINK_CB(cb->skb).portid,
715 RTM_NEWNETCONF, NLM_F_MULTI,
721 if (h == NETDEV_HASHENTRIES + 1) {
722 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
723 net->ipv6.devconf_dflt,
724 NETLINK_CB(cb->skb).portid,
726 RTM_NEWNETCONF, NLM_F_MULTI,
740 static void dev_forward_change(struct inet6_dev *idev)
742 struct net_device *dev;
743 struct inet6_ifaddr *ifa;
748 if (idev->cnf.forwarding)
749 dev_disable_lro(dev);
750 if (dev->flags & IFF_MULTICAST) {
751 if (idev->cnf.forwarding) {
752 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
753 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
754 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
756 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
757 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
758 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
762 list_for_each_entry(ifa, &idev->addr_list, if_list) {
763 if (ifa->flags&IFA_F_TENTATIVE)
765 if (idev->cnf.forwarding)
766 addrconf_join_anycast(ifa);
768 addrconf_leave_anycast(ifa);
770 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
772 dev->ifindex, &idev->cnf);
776 static void addrconf_forward_change(struct net *net, __s32 newf)
778 struct net_device *dev;
779 struct inet6_dev *idev;
781 for_each_netdev(net, dev) {
782 idev = __in6_dev_get(dev);
784 int changed = (!idev->cnf.forwarding) ^ (!newf);
785 idev->cnf.forwarding = newf;
787 dev_forward_change(idev);
792 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
798 return restart_syscall();
800 net = (struct net *)table->extra2;
804 if (p == &net->ipv6.devconf_dflt->forwarding) {
805 if ((!newf) ^ (!old))
806 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
808 NETCONFA_IFINDEX_DEFAULT,
809 net->ipv6.devconf_dflt);
814 if (p == &net->ipv6.devconf_all->forwarding) {
815 int old_dflt = net->ipv6.devconf_dflt->forwarding;
817 net->ipv6.devconf_dflt->forwarding = newf;
818 if ((!newf) ^ (!old_dflt))
819 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
821 NETCONFA_IFINDEX_DEFAULT,
822 net->ipv6.devconf_dflt);
824 addrconf_forward_change(net, newf);
825 if ((!newf) ^ (!old))
826 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
828 NETCONFA_IFINDEX_ALL,
829 net->ipv6.devconf_all);
830 } else if ((!newf) ^ (!old))
831 dev_forward_change((struct inet6_dev *)table->extra1);
835 rt6_purge_dflt_routers(net);
839 static void addrconf_linkdown_change(struct net *net, __s32 newf)
841 struct net_device *dev;
842 struct inet6_dev *idev;
844 for_each_netdev(net, dev) {
845 idev = __in6_dev_get(dev);
847 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
849 idev->cnf.ignore_routes_with_linkdown = newf;
851 inet6_netconf_notify_devconf(dev_net(dev),
853 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
860 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
866 return restart_syscall();
868 net = (struct net *)table->extra2;
872 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
873 if ((!newf) ^ (!old))
874 inet6_netconf_notify_devconf(net,
876 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
877 NETCONFA_IFINDEX_DEFAULT,
878 net->ipv6.devconf_dflt);
883 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
884 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
885 addrconf_linkdown_change(net, newf);
886 if ((!newf) ^ (!old))
887 inet6_netconf_notify_devconf(net,
889 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
890 NETCONFA_IFINDEX_ALL,
891 net->ipv6.devconf_all);
900 /* Nobody refers to this ifaddr, destroy it */
901 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
903 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
905 #ifdef NET_REFCNT_DEBUG
906 pr_debug("%s\n", __func__);
909 in6_dev_put(ifp->idev);
911 if (cancel_delayed_work(&ifp->dad_work))
912 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
915 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
916 pr_warn("Freeing alive inet6 address %p\n", ifp);
924 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
927 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
930 * Each device address list is sorted in order of scope -
931 * global before linklocal.
933 list_for_each(p, &idev->addr_list) {
934 struct inet6_ifaddr *ifa
935 = list_entry(p, struct inet6_ifaddr, if_list);
936 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
940 list_add_tail_rcu(&ifp->if_list, p);
943 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
945 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
947 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
950 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
951 struct net_device *dev, unsigned int hash)
953 struct inet6_ifaddr *ifp;
955 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
956 if (!net_eq(dev_net(ifp->idev->dev), net))
958 if (ipv6_addr_equal(&ifp->addr, addr)) {
959 if (!dev || ifp->idev->dev == dev)
966 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
968 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
971 spin_lock(&addrconf_hash_lock);
973 /* Ignore adding duplicate addresses on an interface */
974 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
975 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
978 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
981 spin_unlock(&addrconf_hash_lock);
986 /* On success it returns ifp with increased reference count */
988 static struct inet6_ifaddr *
989 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
990 bool can_block, struct netlink_ext_ack *extack)
992 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
993 int addr_type = ipv6_addr_type(cfg->pfx);
994 struct net *net = dev_net(idev->dev);
995 struct inet6_ifaddr *ifa = NULL;
996 struct fib6_info *f6i = NULL;
999 if (addr_type == IPV6_ADDR_ANY ||
1000 addr_type & IPV6_ADDR_MULTICAST ||
1001 (!(idev->dev->flags & IFF_LOOPBACK) &&
1002 addr_type & IPV6_ADDR_LOOPBACK))
1003 return ERR_PTR(-EADDRNOTAVAIL);
1006 err = -ENODEV; /*XXX*/
1010 if (idev->cnf.disable_ipv6) {
1015 /* validator notifier needs to be blocking;
1016 * do not call in atomic context
1019 struct in6_validator_info i6vi = {
1020 .i6vi_addr = *cfg->pfx,
1025 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1026 err = notifier_to_errno(err);
1031 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1037 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1044 if (net->ipv6.devconf_all->disable_policy ||
1045 idev->cnf.disable_policy)
1046 f6i->dst_nopolicy = true;
1048 neigh_parms_data_state_setall(idev->nd_parms);
1050 ifa->addr = *cfg->pfx;
1052 ifa->peer_addr = *cfg->peer_pfx;
1054 spin_lock_init(&ifa->lock);
1055 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1056 INIT_HLIST_NODE(&ifa->addr_lst);
1057 ifa->scope = cfg->scope;
1058 ifa->prefix_len = cfg->plen;
1059 ifa->rt_priority = cfg->rt_priority;
1060 ifa->flags = cfg->ifa_flags;
1061 /* No need to add the TENTATIVE flag for addresses with NODAD */
1062 if (!(cfg->ifa_flags & IFA_F_NODAD))
1063 ifa->flags |= IFA_F_TENTATIVE;
1064 ifa->valid_lft = cfg->valid_lft;
1065 ifa->prefered_lft = cfg->preferred_lft;
1066 ifa->cstamp = ifa->tstamp = jiffies;
1067 ifa->tokenized = false;
1075 refcount_set(&ifa->refcnt, 1);
1079 err = ipv6_add_addr_hash(idev->dev, ifa);
1081 rcu_read_unlock_bh();
1085 write_lock(&idev->lock);
1087 /* Add to inet6_dev unicast addr list. */
1088 ipv6_link_dev_addr(idev, ifa);
1090 if (ifa->flags&IFA_F_TEMPORARY) {
1091 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1096 write_unlock(&idev->lock);
1098 rcu_read_unlock_bh();
1100 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1102 if (unlikely(err < 0)) {
1103 fib6_info_release(f6i);
1107 in6_dev_put(ifa->idev);
1116 enum cleanup_prefix_rt_t {
1117 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1118 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1119 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1123 * Check, whether the prefix for ifp would still need a prefix route
1124 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1127 * 1) we don't purge prefix if address was not permanent.
1128 * prefix is managed by its own lifetime.
1129 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1130 * 3) if there are no addresses, delete prefix.
1131 * 4) if there are still other permanent address(es),
1132 * corresponding prefix is still permanent.
1133 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1134 * don't purge the prefix, assume user space is managing it.
1135 * 6) otherwise, update prefix lifetime to the
1136 * longest valid lifetime among the corresponding
1137 * addresses on the device.
1138 * Note: subsequent RA will update lifetime.
1140 static enum cleanup_prefix_rt_t
1141 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1143 struct inet6_ifaddr *ifa;
1144 struct inet6_dev *idev = ifp->idev;
1145 unsigned long lifetime;
1146 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1150 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1153 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1156 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1157 return CLEANUP_PREFIX_RT_NOP;
1159 action = CLEANUP_PREFIX_RT_EXPIRE;
1161 spin_lock(&ifa->lock);
1163 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1165 * Note: Because this address is
1166 * not permanent, lifetime <
1167 * LONG_MAX / HZ here.
1169 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1170 *expires = ifa->tstamp + lifetime * HZ;
1171 spin_unlock(&ifa->lock);
1178 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1180 struct fib6_info *f6i;
1182 f6i = addrconf_get_prefix_route(&ifp->addr,
1185 0, RTF_GATEWAY | RTF_DEFAULT);
1188 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1190 if (!(f6i->fib6_flags & RTF_EXPIRES))
1191 fib6_set_expires(f6i, expires);
1192 fib6_info_release(f6i);
1198 /* This function wants to get referenced ifp and releases it before return */
1200 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1203 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1204 unsigned long expires;
1208 spin_lock_bh(&ifp->lock);
1210 ifp->state = INET6_IFADDR_STATE_DEAD;
1211 spin_unlock_bh(&ifp->lock);
1213 if (state == INET6_IFADDR_STATE_DEAD)
1216 spin_lock_bh(&addrconf_hash_lock);
1217 hlist_del_init_rcu(&ifp->addr_lst);
1218 spin_unlock_bh(&addrconf_hash_lock);
1220 write_lock_bh(&ifp->idev->lock);
1222 if (ifp->flags&IFA_F_TEMPORARY) {
1223 list_del(&ifp->tmp_list);
1225 in6_ifa_put(ifp->ifpub);
1231 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1232 action = check_cleanup_prefix_route(ifp, &expires);
1234 list_del_rcu(&ifp->if_list);
1237 write_unlock_bh(&ifp->idev->lock);
1239 addrconf_del_dad_work(ifp);
1241 ipv6_ifa_notify(RTM_DELADDR, ifp);
1243 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1245 if (action != CLEANUP_PREFIX_RT_NOP) {
1246 cleanup_prefix_route(ifp, expires,
1247 action == CLEANUP_PREFIX_RT_DEL);
1250 /* clean up prefsrc entries */
1251 rt6_remove_prefsrc(ifp);
1256 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1257 struct inet6_ifaddr *ift,
1260 struct inet6_dev *idev = ifp->idev;
1261 struct in6_addr addr, *tmpaddr;
1262 unsigned long tmp_tstamp, age;
1263 unsigned long regen_advance;
1264 struct ifa6_config cfg;
1266 unsigned long now = jiffies;
1267 long max_desync_factor;
1268 s32 cnf_temp_preferred_lft;
1270 write_lock_bh(&idev->lock);
1272 spin_lock_bh(&ift->lock);
1273 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1274 spin_unlock_bh(&ift->lock);
1281 if (idev->cnf.use_tempaddr <= 0) {
1282 write_unlock_bh(&idev->lock);
1283 pr_info("%s: use_tempaddr is disabled\n", __func__);
1288 spin_lock_bh(&ifp->lock);
1289 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1290 idev->cnf.use_tempaddr = -1; /*XXX*/
1291 spin_unlock_bh(&ifp->lock);
1292 write_unlock_bh(&idev->lock);
1293 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1300 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1301 ipv6_try_regen_rndid(idev, tmpaddr);
1302 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1303 age = (now - ifp->tstamp) / HZ;
1305 regen_advance = idev->cnf.regen_max_retry *
1306 idev->cnf.dad_transmits *
1307 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1309 /* recalculate max_desync_factor each time and update
1310 * idev->desync_factor if it's larger
1312 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1313 max_desync_factor = min_t(__u32,
1314 idev->cnf.max_desync_factor,
1315 cnf_temp_preferred_lft - regen_advance);
1317 if (unlikely(idev->desync_factor > max_desync_factor)) {
1318 if (max_desync_factor > 0) {
1319 get_random_bytes(&idev->desync_factor,
1320 sizeof(idev->desync_factor));
1321 idev->desync_factor %= max_desync_factor;
1323 idev->desync_factor = 0;
1327 memset(&cfg, 0, sizeof(cfg));
1328 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1329 idev->cnf.temp_valid_lft + age);
1330 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1331 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1333 cfg.plen = ifp->prefix_len;
1334 tmp_tstamp = ifp->tstamp;
1335 spin_unlock_bh(&ifp->lock);
1337 write_unlock_bh(&idev->lock);
1339 /* A temporary address is created only if this calculated Preferred
1340 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1341 * an implementation must not create a temporary address with a zero
1342 * Preferred Lifetime.
1343 * Use age calculation as in addrconf_verify to avoid unnecessary
1344 * temporary addresses being generated.
1346 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1347 if (cfg.preferred_lft <= regen_advance + age) {
1354 cfg.ifa_flags = IFA_F_TEMPORARY;
1355 /* set in addrconf_prefix_rcv() */
1356 if (ifp->flags & IFA_F_OPTIMISTIC)
1357 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1360 cfg.scope = ipv6_addr_scope(cfg.pfx);
1362 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1366 pr_info("%s: retry temporary address regeneration\n", __func__);
1368 write_lock_bh(&idev->lock);
1372 spin_lock_bh(&ift->lock);
1375 ift->tstamp = tmp_tstamp;
1376 spin_unlock_bh(&ift->lock);
1378 addrconf_dad_start(ift);
1386 * Choose an appropriate source address (RFC3484)
1389 IPV6_SADDR_RULE_INIT = 0,
1390 IPV6_SADDR_RULE_LOCAL,
1391 IPV6_SADDR_RULE_SCOPE,
1392 IPV6_SADDR_RULE_PREFERRED,
1393 #ifdef CONFIG_IPV6_MIP6
1394 IPV6_SADDR_RULE_HOA,
1396 IPV6_SADDR_RULE_OIF,
1397 IPV6_SADDR_RULE_LABEL,
1398 IPV6_SADDR_RULE_PRIVACY,
1399 IPV6_SADDR_RULE_ORCHID,
1400 IPV6_SADDR_RULE_PREFIX,
1401 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1402 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1407 struct ipv6_saddr_score {
1410 struct inet6_ifaddr *ifa;
1411 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1416 struct ipv6_saddr_dst {
1417 const struct in6_addr *addr;
1424 static inline int ipv6_saddr_preferred(int type)
1426 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1431 static bool ipv6_use_optimistic_addr(struct net *net,
1432 struct inet6_dev *idev)
1434 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1437 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1439 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1448 static bool ipv6_allow_optimistic_dad(struct net *net,
1449 struct inet6_dev *idev)
1451 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1454 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1463 static int ipv6_get_saddr_eval(struct net *net,
1464 struct ipv6_saddr_score *score,
1465 struct ipv6_saddr_dst *dst,
1470 if (i <= score->rule) {
1472 case IPV6_SADDR_RULE_SCOPE:
1473 ret = score->scopedist;
1475 case IPV6_SADDR_RULE_PREFIX:
1476 ret = score->matchlen;
1479 ret = !!test_bit(i, score->scorebits);
1485 case IPV6_SADDR_RULE_INIT:
1486 /* Rule 0: remember if hiscore is not ready yet */
1489 case IPV6_SADDR_RULE_LOCAL:
1490 /* Rule 1: Prefer same address */
1491 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1493 case IPV6_SADDR_RULE_SCOPE:
1494 /* Rule 2: Prefer appropriate scope
1499 * ---+--+-+---> scope
1501 * | d is scope of the destination.
1503 * | \ <- smaller scope is better if
1504 * B-15 | \ if scope is enough for destination.
1505 * | ret = B - scope (-1 <= scope >= d <= 15).
1507 * |/ <- greater is better
1508 * -C / if scope is not enough for destination.
1509 * /| ret = scope - C (-1 <= d < scope <= 15).
1511 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1512 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1513 * Assume B = 0 and we get C > 29.
1515 ret = __ipv6_addr_src_scope(score->addr_type);
1516 if (ret >= dst->scope)
1519 ret -= 128; /* 30 is enough */
1520 score->scopedist = ret;
1522 case IPV6_SADDR_RULE_PREFERRED:
1524 /* Rule 3: Avoid deprecated and optimistic addresses */
1525 u8 avoid = IFA_F_DEPRECATED;
1527 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1528 avoid |= IFA_F_OPTIMISTIC;
1529 ret = ipv6_saddr_preferred(score->addr_type) ||
1530 !(score->ifa->flags & avoid);
1533 #ifdef CONFIG_IPV6_MIP6
1534 case IPV6_SADDR_RULE_HOA:
1536 /* Rule 4: Prefer home address */
1537 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1538 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1542 case IPV6_SADDR_RULE_OIF:
1543 /* Rule 5: Prefer outgoing interface */
1544 ret = (!dst->ifindex ||
1545 dst->ifindex == score->ifa->idev->dev->ifindex);
1547 case IPV6_SADDR_RULE_LABEL:
1548 /* Rule 6: Prefer matching label */
1549 ret = ipv6_addr_label(net,
1550 &score->ifa->addr, score->addr_type,
1551 score->ifa->idev->dev->ifindex) == dst->label;
1553 case IPV6_SADDR_RULE_PRIVACY:
1555 /* Rule 7: Prefer public address
1556 * Note: prefer temporary address if use_tempaddr >= 2
1558 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1559 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1560 score->ifa->idev->cnf.use_tempaddr >= 2;
1561 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1564 case IPV6_SADDR_RULE_ORCHID:
1565 /* Rule 8-: Prefer ORCHID vs ORCHID or
1566 * non-ORCHID vs non-ORCHID
1568 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1569 ipv6_addr_orchid(dst->addr));
1571 case IPV6_SADDR_RULE_PREFIX:
1572 /* Rule 8: Use longest matching prefix */
1573 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1574 if (ret > score->ifa->prefix_len)
1575 ret = score->ifa->prefix_len;
1576 score->matchlen = ret;
1578 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1579 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1580 /* Optimistic addresses still have lower precedence than other
1581 * preferred addresses.
1583 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1591 __set_bit(i, score->scorebits);
1597 static int __ipv6_dev_get_saddr(struct net *net,
1598 struct ipv6_saddr_dst *dst,
1599 struct inet6_dev *idev,
1600 struct ipv6_saddr_score *scores,
1603 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1605 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1609 * - Tentative Address (RFC2462 section 5.4)
1610 * - A tentative address is not considered
1611 * "assigned to an interface" in the traditional
1612 * sense, unless it is also flagged as optimistic.
1613 * - Candidate Source Address (section 4)
1614 * - In any case, anycast addresses, multicast
1615 * addresses, and the unspecified address MUST
1616 * NOT be included in a candidate set.
1618 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1619 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1622 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1624 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1625 score->addr_type & IPV6_ADDR_MULTICAST)) {
1626 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1632 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1634 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1635 int minihiscore, miniscore;
1637 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1638 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1640 if (minihiscore > miniscore) {
1641 if (i == IPV6_SADDR_RULE_SCOPE &&
1642 score->scopedist > 0) {
1645 * each remaining entry
1646 * has too small (not enough)
1647 * scope, because ifa entries
1648 * are sorted by their scope
1654 } else if (minihiscore < miniscore) {
1655 swap(hiscore, score);
1656 hiscore_idx = 1 - hiscore_idx;
1658 /* restore our iterator */
1659 score->ifa = hiscore->ifa;
1669 static int ipv6_get_saddr_master(struct net *net,
1670 const struct net_device *dst_dev,
1671 const struct net_device *master,
1672 struct ipv6_saddr_dst *dst,
1673 struct ipv6_saddr_score *scores,
1676 struct inet6_dev *idev;
1678 idev = __in6_dev_get(dst_dev);
1680 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1681 scores, hiscore_idx);
1683 idev = __in6_dev_get(master);
1685 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1686 scores, hiscore_idx);
1691 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1692 const struct in6_addr *daddr, unsigned int prefs,
1693 struct in6_addr *saddr)
1695 struct ipv6_saddr_score scores[2], *hiscore;
1696 struct ipv6_saddr_dst dst;
1697 struct inet6_dev *idev;
1698 struct net_device *dev;
1700 bool use_oif_addr = false;
1701 int hiscore_idx = 0;
1704 dst_type = __ipv6_addr_type(daddr);
1706 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1707 dst.scope = __ipv6_addr_src_scope(dst_type);
1708 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1711 scores[hiscore_idx].rule = -1;
1712 scores[hiscore_idx].ifa = NULL;
1716 /* Candidate Source Address (section 4)
1717 * - multicast and link-local destination address,
1718 * the set of candidate source address MUST only
1719 * include addresses assigned to interfaces
1720 * belonging to the same link as the outgoing
1722 * (- For site-local destination addresses, the
1723 * set of candidate source addresses MUST only
1724 * include addresses assigned to interfaces
1725 * belonging to the same site as the outgoing
1727 * - "It is RECOMMENDED that the candidate source addresses
1728 * be the set of unicast addresses assigned to the
1729 * interface that will be used to send to the destination
1730 * (the 'outgoing' interface)." (RFC 6724)
1733 idev = __in6_dev_get(dst_dev);
1734 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1735 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1736 (idev && idev->cnf.use_oif_addrs_only)) {
1737 use_oif_addr = true;
1743 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1745 const struct net_device *master;
1748 /* if dst_dev exists and is enslaved to an L3 device, then
1749 * prefer addresses from dst_dev and then the master over
1750 * any other enslaved devices in the L3 domain.
1752 master = l3mdev_master_dev_rcu(dst_dev);
1754 master_idx = master->ifindex;
1756 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1758 scores, hiscore_idx);
1760 if (scores[hiscore_idx].ifa)
1764 for_each_netdev_rcu(net, dev) {
1765 /* only consider addresses on devices in the
1768 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1770 idev = __in6_dev_get(dev);
1773 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1778 hiscore = &scores[hiscore_idx];
1780 ret = -EADDRNOTAVAIL;
1782 *saddr = hiscore->ifa->addr;
1787 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1789 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1792 struct inet6_ifaddr *ifp;
1793 int err = -EADDRNOTAVAIL;
1795 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1796 if (ifp->scope > IFA_LINK)
1798 if (ifp->scope == IFA_LINK &&
1799 !(ifp->flags & banned_flags)) {
1808 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1811 struct inet6_dev *idev;
1812 int err = -EADDRNOTAVAIL;
1815 idev = __in6_dev_get(dev);
1817 read_lock_bh(&idev->lock);
1818 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1819 read_unlock_bh(&idev->lock);
1825 static int ipv6_count_addresses(const struct inet6_dev *idev)
1827 const struct inet6_ifaddr *ifp;
1831 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1837 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1838 const struct net_device *dev, int strict)
1840 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1841 strict, IFA_F_TENTATIVE);
1843 EXPORT_SYMBOL(ipv6_chk_addr);
1845 /* device argument is used to find the L3 domain of interest. If
1846 * skip_dev_check is set, then the ifp device is not checked against
1847 * the passed in dev argument. So the 2 cases for addresses checks are:
1848 * 1. does the address exist in the L3 domain that dev is part of
1849 * (skip_dev_check = true), or
1851 * 2. does the address exist on the specific device
1852 * (skip_dev_check = false)
1854 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1855 const struct net_device *dev, bool skip_dev_check,
1856 int strict, u32 banned_flags)
1858 unsigned int hash = inet6_addr_hash(net, addr);
1859 const struct net_device *l3mdev;
1860 struct inet6_ifaddr *ifp;
1865 l3mdev = l3mdev_master_dev_rcu(dev);
1869 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1870 if (!net_eq(dev_net(ifp->idev->dev), net))
1873 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1876 /* Decouple optimistic from tentative for evaluation here.
1877 * Ban optimistic addresses explicitly, when required.
1879 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1880 ? (ifp->flags&~IFA_F_TENTATIVE)
1882 if (ipv6_addr_equal(&ifp->addr, addr) &&
1883 !(ifp_flags&banned_flags) &&
1884 (!dev || ifp->idev->dev == dev ||
1885 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1894 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1897 /* Compares an address/prefix_len with addresses on device @dev.
1898 * If one is found it returns true.
1900 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1901 const unsigned int prefix_len, struct net_device *dev)
1903 const struct inet6_ifaddr *ifa;
1904 const struct inet6_dev *idev;
1908 idev = __in6_dev_get(dev);
1910 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1911 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1920 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1922 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1924 const struct inet6_ifaddr *ifa;
1925 const struct inet6_dev *idev;
1930 idev = __in6_dev_get(dev);
1932 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1933 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1942 EXPORT_SYMBOL(ipv6_chk_prefix);
1944 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1945 struct net_device *dev, int strict)
1947 unsigned int hash = inet6_addr_hash(net, addr);
1948 struct inet6_ifaddr *ifp, *result = NULL;
1951 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1952 if (!net_eq(dev_net(ifp->idev->dev), net))
1954 if (ipv6_addr_equal(&ifp->addr, addr)) {
1955 if (!dev || ifp->idev->dev == dev ||
1956 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1968 /* Gets referenced address, destroys ifaddr */
1970 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1973 ifp->flags |= IFA_F_DADFAILED;
1975 if (ifp->flags&IFA_F_TEMPORARY) {
1976 struct inet6_ifaddr *ifpub;
1977 spin_lock_bh(&ifp->lock);
1980 in6_ifa_hold(ifpub);
1981 spin_unlock_bh(&ifp->lock);
1982 ipv6_create_tempaddr(ifpub, ifp, true);
1985 spin_unlock_bh(&ifp->lock);
1988 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1989 spin_lock_bh(&ifp->lock);
1990 addrconf_del_dad_work(ifp);
1991 ifp->flags |= IFA_F_TENTATIVE;
1993 ifp->flags &= ~IFA_F_OPTIMISTIC;
1994 spin_unlock_bh(&ifp->lock);
1996 ipv6_ifa_notify(0, ifp);
2003 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2007 spin_lock_bh(&ifp->lock);
2008 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2009 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2012 spin_unlock_bh(&ifp->lock);
2017 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2019 struct inet6_dev *idev = ifp->idev;
2020 struct net *net = dev_net(ifp->idev->dev);
2022 if (addrconf_dad_end(ifp)) {
2027 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2028 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2030 spin_lock_bh(&ifp->lock);
2032 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2033 struct in6_addr new_addr;
2034 struct inet6_ifaddr *ifp2;
2035 int retries = ifp->stable_privacy_retry + 1;
2036 struct ifa6_config cfg = {
2038 .plen = ifp->prefix_len,
2039 .ifa_flags = ifp->flags,
2040 .valid_lft = ifp->valid_lft,
2041 .preferred_lft = ifp->prefered_lft,
2042 .scope = ifp->scope,
2045 if (retries > net->ipv6.sysctl.idgen_retries) {
2046 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2047 ifp->idev->dev->name);
2051 new_addr = ifp->addr;
2052 if (ipv6_generate_stable_address(&new_addr, retries,
2056 spin_unlock_bh(&ifp->lock);
2058 if (idev->cnf.max_addresses &&
2059 ipv6_count_addresses(idev) >=
2060 idev->cnf.max_addresses)
2063 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2064 ifp->idev->dev->name);
2066 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2070 spin_lock_bh(&ifp2->lock);
2071 ifp2->stable_privacy_retry = retries;
2072 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2073 spin_unlock_bh(&ifp2->lock);
2075 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2078 spin_lock_bh(&ifp->lock);
2082 /* transition from _POSTDAD to _ERRDAD */
2083 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2084 spin_unlock_bh(&ifp->lock);
2086 addrconf_mod_dad_work(ifp, 0);
2090 /* Join to solicited addr multicast group.
2091 * caller must hold RTNL */
2092 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2094 struct in6_addr maddr;
2096 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2099 addrconf_addr_solict_mult(addr, &maddr);
2100 ipv6_dev_mc_inc(dev, &maddr);
2103 /* caller must hold RTNL */
2104 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2106 struct in6_addr maddr;
2108 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2111 addrconf_addr_solict_mult(addr, &maddr);
2112 __ipv6_dev_mc_dec(idev, &maddr);
2115 /* caller must hold RTNL */
2116 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2118 struct in6_addr addr;
2120 if (ifp->prefix_len >= 127) /* RFC 6164 */
2122 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2123 if (ipv6_addr_any(&addr))
2125 __ipv6_dev_ac_inc(ifp->idev, &addr);
2128 /* caller must hold RTNL */
2129 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2131 struct in6_addr addr;
2133 if (ifp->prefix_len >= 127) /* RFC 6164 */
2135 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2136 if (ipv6_addr_any(&addr))
2138 __ipv6_dev_ac_dec(ifp->idev, &addr);
2141 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2143 switch (dev->addr_len) {
2145 memcpy(eui, dev->dev_addr, 3);
2148 memcpy(eui + 5, dev->dev_addr + 3, 3);
2150 case EUI64_ADDR_LEN:
2151 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2161 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2163 union fwnet_hwaddr *ha;
2165 if (dev->addr_len != FWNET_ALEN)
2168 ha = (union fwnet_hwaddr *)dev->dev_addr;
2170 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2175 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2177 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2178 if (dev->addr_len != ARCNET_ALEN)
2181 eui[7] = *(u8 *)dev->dev_addr;
2185 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2187 if (dev->addr_len != INFINIBAND_ALEN)
2189 memcpy(eui, dev->dev_addr + 12, 8);
2194 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2198 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2199 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2200 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2201 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2202 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2203 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2207 memcpy(eui + 4, &addr, 4);
2211 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2213 if (dev->priv_flags & IFF_ISATAP)
2214 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2218 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2220 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2223 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2225 memcpy(eui, dev->perm_addr, 3);
2226 memcpy(eui + 5, dev->perm_addr + 3, 3);
2233 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2235 switch (dev->type) {
2238 return addrconf_ifid_eui48(eui, dev);
2240 return addrconf_ifid_arcnet(eui, dev);
2241 case ARPHRD_INFINIBAND:
2242 return addrconf_ifid_infiniband(eui, dev);
2244 return addrconf_ifid_sit(eui, dev);
2247 return addrconf_ifid_gre(eui, dev);
2248 case ARPHRD_6LOWPAN:
2249 return addrconf_ifid_6lowpan(eui, dev);
2250 case ARPHRD_IEEE1394:
2251 return addrconf_ifid_ieee1394(eui, dev);
2252 case ARPHRD_TUNNEL6:
2255 return addrconf_ifid_ip6tnl(eui, dev);
2260 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2263 struct inet6_ifaddr *ifp;
2265 read_lock_bh(&idev->lock);
2266 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2267 if (ifp->scope > IFA_LINK)
2269 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2270 memcpy(eui, ifp->addr.s6_addr+8, 8);
2275 read_unlock_bh(&idev->lock);
2279 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2280 static void ipv6_regen_rndid(struct inet6_dev *idev)
2283 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2284 idev->rndid[0] &= ~0x02;
2287 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2288 * check if generated address is not inappropriate
2290 * - Reserved subnet anycast (RFC 2526)
2291 * 11111101 11....11 1xxxxxxx
2292 * - ISATAP (RFC4214) 6.1
2293 * 00-00-5E-FE-xx-xx-xx-xx
2295 * - XXX: already assigned to an address on the device
2297 if (idev->rndid[0] == 0xfd &&
2298 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2299 (idev->rndid[7]&0x80))
2301 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2302 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2304 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2309 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2311 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2312 ipv6_regen_rndid(idev);
2320 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2321 struct net_device *dev, unsigned long expires,
2322 u32 flags, gfp_t gfp_flags)
2324 struct fib6_config cfg = {
2325 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2326 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2327 .fc_ifindex = dev->ifindex,
2328 .fc_expires = expires,
2330 .fc_flags = RTF_UP | flags,
2331 .fc_nlinfo.nl_net = dev_net(dev),
2332 .fc_protocol = RTPROT_KERNEL,
2333 .fc_type = RTN_UNICAST,
2338 /* Prevent useless cloning on PtP SIT.
2339 This thing is done here expecting that the whole
2340 class of non-broadcast devices need not cloning.
2342 #if IS_ENABLED(CONFIG_IPV6_SIT)
2343 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2344 cfg.fc_flags |= RTF_NONEXTHOP;
2347 ip6_route_add(&cfg, gfp_flags, NULL);
2351 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2353 const struct net_device *dev,
2354 u32 flags, u32 noflags)
2356 struct fib6_node *fn;
2357 struct fib6_info *rt = NULL;
2358 struct fib6_table *table;
2359 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2361 table = fib6_get_table(dev_net(dev), tb_id);
2366 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2370 for_each_fib6_node_rt_rcu(fn) {
2371 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2373 if ((rt->fib6_flags & flags) != flags)
2375 if ((rt->fib6_flags & noflags) != 0)
2386 /* Create "default" multicast route to the interface */
2388 static void addrconf_add_mroute(struct net_device *dev)
2390 struct fib6_config cfg = {
2391 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2392 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2393 .fc_ifindex = dev->ifindex,
2396 .fc_type = RTN_UNICAST,
2397 .fc_nlinfo.nl_net = dev_net(dev),
2400 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2402 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
2405 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2407 struct inet6_dev *idev;
2411 idev = ipv6_find_idev(dev);
2413 return ERR_PTR(-ENOBUFS);
2415 if (idev->cnf.disable_ipv6)
2416 return ERR_PTR(-EACCES);
2418 /* Add default multicast route */
2419 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2420 addrconf_add_mroute(dev);
2425 static void manage_tempaddrs(struct inet6_dev *idev,
2426 struct inet6_ifaddr *ifp,
2427 __u32 valid_lft, __u32 prefered_lft,
2428 bool create, unsigned long now)
2431 struct inet6_ifaddr *ift;
2433 read_lock_bh(&idev->lock);
2434 /* update all temporary addresses in the list */
2435 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2436 int age, max_valid, max_prefered;
2438 if (ifp != ift->ifpub)
2441 /* RFC 4941 section 3.3:
2442 * If a received option will extend the lifetime of a public
2443 * address, the lifetimes of temporary addresses should
2444 * be extended, subject to the overall constraint that no
2445 * temporary addresses should ever remain "valid" or "preferred"
2446 * for a time longer than (TEMP_VALID_LIFETIME) or
2447 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2449 age = (now - ift->cstamp) / HZ;
2450 max_valid = idev->cnf.temp_valid_lft - age;
2454 max_prefered = idev->cnf.temp_prefered_lft -
2455 idev->desync_factor - age;
2456 if (max_prefered < 0)
2459 if (valid_lft > max_valid)
2460 valid_lft = max_valid;
2462 if (prefered_lft > max_prefered)
2463 prefered_lft = max_prefered;
2465 spin_lock(&ift->lock);
2467 ift->valid_lft = valid_lft;
2468 ift->prefered_lft = prefered_lft;
2470 if (prefered_lft > 0)
2471 ift->flags &= ~IFA_F_DEPRECATED;
2473 spin_unlock(&ift->lock);
2474 if (!(flags&IFA_F_TENTATIVE))
2475 ipv6_ifa_notify(0, ift);
2478 if ((create || list_empty(&idev->tempaddr_list)) &&
2479 idev->cnf.use_tempaddr > 0) {
2480 /* When a new public address is created as described
2481 * in [ADDRCONF], also create a new temporary address.
2482 * Also create a temporary address if it's enabled but
2483 * no temporary address currently exists.
2485 read_unlock_bh(&idev->lock);
2486 ipv6_create_tempaddr(ifp, NULL, false);
2488 read_unlock_bh(&idev->lock);
2492 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2494 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2495 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2498 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2499 const struct prefix_info *pinfo,
2500 struct inet6_dev *in6_dev,
2501 const struct in6_addr *addr, int addr_type,
2502 u32 addr_flags, bool sllao, bool tokenized,
2503 __u32 valid_lft, u32 prefered_lft)
2505 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2506 int create = 0, update_lft = 0;
2508 if (!ifp && valid_lft) {
2509 int max_addresses = in6_dev->cnf.max_addresses;
2510 struct ifa6_config cfg = {
2512 .plen = pinfo->prefix_len,
2513 .ifa_flags = addr_flags,
2514 .valid_lft = valid_lft,
2515 .preferred_lft = prefered_lft,
2516 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2519 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2520 if ((net->ipv6.devconf_all->optimistic_dad ||
2521 in6_dev->cnf.optimistic_dad) &&
2522 !net->ipv6.devconf_all->forwarding && sllao)
2523 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2526 /* Do not allow to create too much of autoconfigured
2527 * addresses; this would be too easy way to crash kernel.
2529 if (!max_addresses ||
2530 ipv6_count_addresses(in6_dev) < max_addresses)
2531 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2533 if (IS_ERR_OR_NULL(ifp))
2537 spin_lock_bh(&ifp->lock);
2538 ifp->flags |= IFA_F_MANAGETEMPADDR;
2539 ifp->cstamp = jiffies;
2540 ifp->tokenized = tokenized;
2541 spin_unlock_bh(&ifp->lock);
2542 addrconf_dad_start(ifp);
2550 /* update lifetime (RFC2462 5.5.3 e) */
2551 spin_lock_bh(&ifp->lock);
2553 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2554 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2557 if (!create && stored_lft) {
2558 const u32 minimum_lft = min_t(u32,
2559 stored_lft, MIN_VALID_LIFETIME);
2560 valid_lft = max(valid_lft, minimum_lft);
2562 /* RFC4862 Section 5.5.3e:
2563 * "Note that the preferred lifetime of the
2564 * corresponding address is always reset to
2565 * the Preferred Lifetime in the received
2566 * Prefix Information option, regardless of
2567 * whether the valid lifetime is also reset or
2570 * So we should always update prefered_lft here.
2576 ifp->valid_lft = valid_lft;
2577 ifp->prefered_lft = prefered_lft;
2580 ifp->flags &= ~IFA_F_DEPRECATED;
2581 spin_unlock_bh(&ifp->lock);
2583 if (!(flags&IFA_F_TENTATIVE))
2584 ipv6_ifa_notify(0, ifp);
2586 spin_unlock_bh(&ifp->lock);
2588 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2597 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2599 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2601 struct prefix_info *pinfo;
2606 struct inet6_dev *in6_dev;
2607 struct net *net = dev_net(dev);
2609 pinfo = (struct prefix_info *) opt;
2611 if (len < sizeof(struct prefix_info)) {
2612 netdev_dbg(dev, "addrconf: prefix option too short\n");
2617 * Validation checks ([ADDRCONF], page 19)
2620 addr_type = ipv6_addr_type(&pinfo->prefix);
2622 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2625 valid_lft = ntohl(pinfo->valid);
2626 prefered_lft = ntohl(pinfo->prefered);
2628 if (prefered_lft > valid_lft) {
2629 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2633 in6_dev = in6_dev_get(dev);
2636 net_dbg_ratelimited("addrconf: device %s not configured\n",
2642 * Two things going on here:
2643 * 1) Add routes for on-link prefixes
2644 * 2) Configure prefixes with the auto flag set
2647 if (pinfo->onlink) {
2648 struct fib6_info *rt;
2649 unsigned long rt_expires;
2651 /* Avoid arithmetic overflow. Really, we could
2652 * save rt_expires in seconds, likely valid_lft,
2653 * but it would require division in fib gc, that it
2657 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2659 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2661 if (addrconf_finite_timeout(rt_expires))
2664 rt = addrconf_get_prefix_route(&pinfo->prefix,
2667 RTF_ADDRCONF | RTF_PREFIX_RT,
2668 RTF_GATEWAY | RTF_DEFAULT);
2671 /* Autoconf prefix route */
2672 if (valid_lft == 0) {
2673 ip6_del_rt(net, rt);
2675 } else if (addrconf_finite_timeout(rt_expires)) {
2677 fib6_set_expires(rt, jiffies + rt_expires);
2679 fib6_clean_expires(rt);
2681 } else if (valid_lft) {
2682 clock_t expires = 0;
2683 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2684 if (addrconf_finite_timeout(rt_expires)) {
2686 flags |= RTF_EXPIRES;
2687 expires = jiffies_to_clock_t(rt_expires);
2689 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2690 0, dev, expires, flags,
2693 fib6_info_release(rt);
2696 /* Try to figure out our local address for this prefix */
2698 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2699 struct in6_addr addr;
2700 bool tokenized = false, dev_addr_generated = false;
2702 if (pinfo->prefix_len == 64) {
2703 memcpy(&addr, &pinfo->prefix, 8);
2705 if (!ipv6_addr_any(&in6_dev->token)) {
2706 read_lock_bh(&in6_dev->lock);
2707 memcpy(addr.s6_addr + 8,
2708 in6_dev->token.s6_addr + 8, 8);
2709 read_unlock_bh(&in6_dev->lock);
2711 } else if (is_addr_mode_generate_stable(in6_dev) &&
2712 !ipv6_generate_stable_address(&addr, 0,
2714 addr_flags |= IFA_F_STABLE_PRIVACY;
2716 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2717 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2720 dev_addr_generated = true;
2724 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2729 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2732 tokenized, valid_lft,
2737 /* Ignore error case here because previous prefix add addr was
2738 * successful which will be notified.
2740 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2741 addr_type, addr_flags, sllao,
2742 tokenized, valid_lft,
2744 dev_addr_generated);
2746 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2748 in6_dev_put(in6_dev);
2752 * Set destination address.
2753 * Special case for SIT interfaces where we create a new "virtual"
2756 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2758 struct in6_ifreq ireq;
2759 struct net_device *dev;
2765 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2768 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2774 #if IS_ENABLED(CONFIG_IPV6_SIT)
2775 if (dev->type == ARPHRD_SIT) {
2776 const struct net_device_ops *ops = dev->netdev_ops;
2778 struct ip_tunnel_parm p;
2780 err = -EADDRNOTAVAIL;
2781 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2784 memset(&p, 0, sizeof(p));
2785 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2789 p.iph.protocol = IPPROTO_IPV6;
2791 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2793 if (ops->ndo_do_ioctl) {
2794 mm_segment_t oldfs = get_fs();
2797 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2804 dev = __dev_get_by_name(net, p.name);
2807 err = dev_open(dev);
2817 static int ipv6_mc_config(struct sock *sk, bool join,
2818 const struct in6_addr *addr, int ifindex)
2826 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2828 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2835 * Manual configuration of address on an interface
2837 static int inet6_addr_add(struct net *net, int ifindex,
2838 struct ifa6_config *cfg,
2839 struct netlink_ext_ack *extack)
2841 struct inet6_ifaddr *ifp;
2842 struct inet6_dev *idev;
2843 struct net_device *dev;
2844 unsigned long timeout;
2850 if (cfg->plen > 128)
2853 /* check the lifetime */
2854 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2857 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2860 dev = __dev_get_by_index(net, ifindex);
2864 idev = addrconf_add_dev(dev);
2866 return PTR_ERR(idev);
2868 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2869 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2870 true, cfg->pfx, ifindex);
2876 cfg->scope = ipv6_addr_scope(cfg->pfx);
2878 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2879 if (addrconf_finite_timeout(timeout)) {
2880 expires = jiffies_to_clock_t(timeout * HZ);
2881 cfg->valid_lft = timeout;
2882 flags = RTF_EXPIRES;
2886 cfg->ifa_flags |= IFA_F_PERMANENT;
2889 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2890 if (addrconf_finite_timeout(timeout)) {
2892 cfg->ifa_flags |= IFA_F_DEPRECATED;
2893 cfg->preferred_lft = timeout;
2896 ifp = ipv6_add_addr(idev, cfg, true, extack);
2898 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2899 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2900 ifp->rt_priority, dev, expires,
2904 /* Send a netlink notification if DAD is enabled and
2905 * optimistic flag is not set
2907 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2908 ipv6_ifa_notify(0, ifp);
2910 * Note that section 3.1 of RFC 4429 indicates
2911 * that the Optimistic flag should not be set for
2912 * manually configured addresses
2914 addrconf_dad_start(ifp);
2915 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2916 manage_tempaddrs(idev, ifp, cfg->valid_lft,
2917 cfg->preferred_lft, true, jiffies);
2919 addrconf_verify_rtnl();
2921 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2922 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2926 return PTR_ERR(ifp);
2929 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2930 const struct in6_addr *pfx, unsigned int plen)
2932 struct inet6_ifaddr *ifp;
2933 struct inet6_dev *idev;
2934 struct net_device *dev;
2939 dev = __dev_get_by_index(net, ifindex);
2943 idev = __in6_dev_get(dev);
2947 read_lock_bh(&idev->lock);
2948 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2949 if (ifp->prefix_len == plen &&
2950 ipv6_addr_equal(pfx, &ifp->addr)) {
2952 read_unlock_bh(&idev->lock);
2954 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2955 (ifa_flags & IFA_F_MANAGETEMPADDR))
2956 manage_tempaddrs(idev, ifp, 0, 0, false,
2959 addrconf_verify_rtnl();
2960 if (ipv6_addr_is_multicast(pfx)) {
2961 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2962 false, pfx, dev->ifindex);
2967 read_unlock_bh(&idev->lock);
2968 return -EADDRNOTAVAIL;
2972 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2974 struct ifa6_config cfg = {
2975 .ifa_flags = IFA_F_PERMANENT,
2976 .preferred_lft = INFINITY_LIFE_TIME,
2977 .valid_lft = INFINITY_LIFE_TIME,
2979 struct in6_ifreq ireq;
2982 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2985 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2988 cfg.pfx = &ireq.ifr6_addr;
2989 cfg.plen = ireq.ifr6_prefixlen;
2992 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
2997 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2999 struct in6_ifreq ireq;
3002 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3005 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3009 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3010 ireq.ifr6_prefixlen);
3015 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3016 int plen, int scope)
3018 struct inet6_ifaddr *ifp;
3019 struct ifa6_config cfg = {
3022 .ifa_flags = IFA_F_PERMANENT,
3023 .valid_lft = INFINITY_LIFE_TIME,
3024 .preferred_lft = INFINITY_LIFE_TIME,
3028 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3030 spin_lock_bh(&ifp->lock);
3031 ifp->flags &= ~IFA_F_TENTATIVE;
3032 spin_unlock_bh(&ifp->lock);
3033 rt_genid_bump_ipv6(dev_net(idev->dev));
3034 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3039 #if IS_ENABLED(CONFIG_IPV6_SIT)
3040 static void sit_add_v4_addrs(struct inet6_dev *idev)
3042 struct in6_addr addr;
3043 struct net_device *dev;
3044 struct net *net = dev_net(idev->dev);
3050 memset(&addr, 0, sizeof(struct in6_addr));
3051 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3053 if (idev->dev->flags&IFF_POINTOPOINT) {
3054 addr.s6_addr32[0] = htonl(0xfe800000);
3058 scope = IPV6_ADDR_COMPATv4;
3060 pflags |= RTF_NONEXTHOP;
3063 if (addr.s6_addr32[3]) {
3064 add_addr(idev, &addr, plen, scope);
3065 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3070 for_each_netdev(net, dev) {
3071 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3072 if (in_dev && (dev->flags & IFF_UP)) {
3073 struct in_ifaddr *ifa;
3077 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3079 addr.s6_addr32[3] = ifa->ifa_local;
3081 if (ifa->ifa_scope == RT_SCOPE_LINK)
3083 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3084 if (idev->dev->flags&IFF_POINTOPOINT)
3089 add_addr(idev, &addr, plen, flag);
3090 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3091 0, pflags, GFP_ATOMIC);
3098 static void init_loopback(struct net_device *dev)
3100 struct inet6_dev *idev;
3106 idev = ipv6_find_idev(dev);
3108 pr_debug("%s: add_dev failed\n", __func__);
3112 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3115 void addrconf_add_linklocal(struct inet6_dev *idev,
3116 const struct in6_addr *addr, u32 flags)
3118 struct ifa6_config cfg = {
3121 .ifa_flags = flags | IFA_F_PERMANENT,
3122 .valid_lft = INFINITY_LIFE_TIME,
3123 .preferred_lft = INFINITY_LIFE_TIME,
3126 struct inet6_ifaddr *ifp;
3128 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3129 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3130 idev->cnf.optimistic_dad) &&
3131 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3132 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3135 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3137 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3139 addrconf_dad_start(ifp);
3143 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3145 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3147 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3150 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3151 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3154 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3155 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3161 static int ipv6_generate_stable_address(struct in6_addr *address,
3163 const struct inet6_dev *idev)
3165 static DEFINE_SPINLOCK(lock);
3166 static __u32 digest[SHA_DIGEST_WORDS];
3167 static __u32 workspace[SHA_WORKSPACE_WORDS];
3170 char __data[SHA_MESSAGE_BYTES];
3172 struct in6_addr secret;
3174 unsigned char hwaddr[MAX_ADDR_LEN];
3179 struct in6_addr secret;
3180 struct in6_addr temp;
3181 struct net *net = dev_net(idev->dev);
3183 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3185 if (idev->cnf.stable_secret.initialized)
3186 secret = idev->cnf.stable_secret.secret;
3187 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3188 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3193 spin_lock_bh(&lock);
3196 memset(&data, 0, sizeof(data));
3197 memset(workspace, 0, sizeof(workspace));
3198 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3199 data.prefix[0] = address->s6_addr32[0];
3200 data.prefix[1] = address->s6_addr32[1];
3201 data.secret = secret;
3202 data.dad_count = dad_count;
3204 sha_transform(digest, data.__data, workspace);
3207 temp.s6_addr32[2] = (__force __be32)digest[0];
3208 temp.s6_addr32[3] = (__force __be32)digest[1];
3210 spin_unlock_bh(&lock);
3212 if (ipv6_reserved_interfaceid(temp)) {
3214 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3223 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3225 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3229 s = &idev->cnf.stable_secret;
3230 get_random_bytes(&s->secret, sizeof(s->secret));
3231 s->initialized = true;
3234 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3236 struct in6_addr addr;
3238 /* no link local addresses on L3 master devices */
3239 if (netif_is_l3_master(idev->dev))
3242 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3244 switch (idev->cnf.addr_gen_mode) {
3245 case IN6_ADDR_GEN_MODE_RANDOM:
3246 ipv6_gen_mode_random_init(idev);
3248 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3249 if (!ipv6_generate_stable_address(&addr, 0, idev))
3250 addrconf_add_linklocal(idev, &addr,
3251 IFA_F_STABLE_PRIVACY);
3252 else if (prefix_route)
3253 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3256 case IN6_ADDR_GEN_MODE_EUI64:
3257 /* addrconf_add_linklocal also adds a prefix_route and we
3258 * only need to care about prefix routes if ipv6_generate_eui64
3259 * couldn't generate one.
3261 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3262 addrconf_add_linklocal(idev, &addr, 0);
3263 else if (prefix_route)
3264 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3267 case IN6_ADDR_GEN_MODE_NONE:
3269 /* will not add any link local address */
3274 static void addrconf_dev_config(struct net_device *dev)
3276 struct inet6_dev *idev;
3280 if ((dev->type != ARPHRD_ETHER) &&
3281 (dev->type != ARPHRD_FDDI) &&
3282 (dev->type != ARPHRD_ARCNET) &&
3283 (dev->type != ARPHRD_INFINIBAND) &&
3284 (dev->type != ARPHRD_IEEE1394) &&
3285 (dev->type != ARPHRD_TUNNEL6) &&
3286 (dev->type != ARPHRD_6LOWPAN) &&
3287 (dev->type != ARPHRD_IP6GRE) &&
3288 (dev->type != ARPHRD_IPGRE) &&
3289 (dev->type != ARPHRD_TUNNEL) &&
3290 (dev->type != ARPHRD_NONE) &&
3291 (dev->type != ARPHRD_RAWIP)) {
3292 /* Alas, we support only Ethernet autoconfiguration. */
3296 idev = addrconf_add_dev(dev);
3300 /* this device type has no EUI support */
3301 if (dev->type == ARPHRD_NONE &&
3302 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3303 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3305 addrconf_addr_gen(idev, false);
3308 #if IS_ENABLED(CONFIG_IPV6_SIT)
3309 static void addrconf_sit_config(struct net_device *dev)
3311 struct inet6_dev *idev;
3316 * Configure the tunnel with one of our IPv4
3317 * addresses... we should configure all of
3318 * our v4 addrs in the tunnel
3321 idev = ipv6_find_idev(dev);
3323 pr_debug("%s: add_dev failed\n", __func__);
3327 if (dev->priv_flags & IFF_ISATAP) {
3328 addrconf_addr_gen(idev, false);
3332 sit_add_v4_addrs(idev);
3334 if (dev->flags&IFF_POINTOPOINT)
3335 addrconf_add_mroute(dev);
3339 #if IS_ENABLED(CONFIG_NET_IPGRE)
3340 static void addrconf_gre_config(struct net_device *dev)
3342 struct inet6_dev *idev;
3346 idev = ipv6_find_idev(dev);
3348 pr_debug("%s: add_dev failed\n", __func__);
3352 addrconf_addr_gen(idev, true);
3353 if (dev->flags & IFF_POINTOPOINT)
3354 addrconf_add_mroute(dev);
3358 static int fixup_permanent_addr(struct net *net,
3359 struct inet6_dev *idev,
3360 struct inet6_ifaddr *ifp)
3362 /* !fib6_node means the host route was removed from the
3363 * FIB, for example, if 'lo' device is taken down. In that
3364 * case regenerate the host route.
3366 if (!ifp->rt || !ifp->rt->fib6_node) {
3367 struct fib6_info *f6i, *prev;
3369 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3372 return PTR_ERR(f6i);
3374 /* ifp->rt can be accessed outside of rtnl */
3375 spin_lock(&ifp->lock);
3378 spin_unlock(&ifp->lock);
3380 fib6_info_release(prev);
3383 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3384 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3385 ifp->rt_priority, idev->dev, 0, 0,
3389 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3390 addrconf_dad_start(ifp);
3395 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3397 struct inet6_ifaddr *ifp, *tmp;
3398 struct inet6_dev *idev;
3400 idev = __in6_dev_get(dev);
3404 write_lock_bh(&idev->lock);
3406 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3407 if ((ifp->flags & IFA_F_PERMANENT) &&
3408 fixup_permanent_addr(net, idev, ifp) < 0) {
3409 write_unlock_bh(&idev->lock);
3412 write_lock_bh(&idev->lock);
3414 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3415 idev->dev->name, &ifp->addr);
3419 write_unlock_bh(&idev->lock);
3422 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3425 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3426 struct netdev_notifier_changeupper_info *info;
3427 struct inet6_dev *idev = __in6_dev_get(dev);
3428 struct net *net = dev_net(dev);
3429 int run_pending = 0;
3433 case NETDEV_REGISTER:
3434 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3435 idev = ipv6_add_dev(dev);
3437 return notifier_from_errno(PTR_ERR(idev));
3441 case NETDEV_CHANGEMTU:
3442 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3443 if (dev->mtu < IPV6_MIN_MTU) {
3444 addrconf_ifdown(dev, dev != net->loopback_dev);
3449 rt6_mtu_change(dev, dev->mtu);
3450 idev->cnf.mtu6 = dev->mtu;
3454 /* allocate new idev */
3455 idev = ipv6_add_dev(dev);
3459 /* device is still not ready */
3460 if (!(idev->if_flags & IF_READY))
3469 if (dev->flags & IFF_SLAVE)
3472 if (idev && idev->cnf.disable_ipv6)
3475 if (event == NETDEV_UP) {
3476 /* restore routes for permanent addresses */
3477 addrconf_permanent_addr(net, dev);
3479 if (!addrconf_link_ready(dev)) {
3480 /* device is not ready yet. */
3481 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3486 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3487 idev = ipv6_add_dev(dev);
3489 if (!IS_ERR_OR_NULL(idev)) {
3490 idev->if_flags |= IF_READY;
3493 } else if (event == NETDEV_CHANGE) {
3494 if (!addrconf_link_ready(dev)) {
3495 /* device is still not ready. */
3496 rt6_sync_down_dev(dev, event);
3501 if (idev->if_flags & IF_READY) {
3502 /* device is already configured -
3503 * but resend MLD reports, we might
3504 * have roamed and need to update
3505 * multicast snooping switches
3508 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3511 idev->if_flags |= IF_READY;
3514 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3520 switch (dev->type) {
3521 #if IS_ENABLED(CONFIG_IPV6_SIT)
3523 addrconf_sit_config(dev);
3526 #if IS_ENABLED(CONFIG_NET_IPGRE)
3528 addrconf_gre_config(dev);
3531 case ARPHRD_LOOPBACK:
3536 addrconf_dev_config(dev);
3540 if (!IS_ERR_OR_NULL(idev)) {
3542 addrconf_dad_run(idev);
3544 /* Device has an address by now */
3545 rt6_sync_up(dev, RTNH_F_DEAD);
3548 * If the MTU changed during the interface down,
3549 * when the interface up, the changed MTU must be
3550 * reflected in the idev as well as routers.
3552 if (idev->cnf.mtu6 != dev->mtu &&
3553 dev->mtu >= IPV6_MIN_MTU) {
3554 rt6_mtu_change(dev, dev->mtu);
3555 idev->cnf.mtu6 = dev->mtu;
3557 idev->tstamp = jiffies;
3558 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3561 * If the changed mtu during down is lower than
3562 * IPV6_MIN_MTU stop IPv6 on this interface.
3564 if (dev->mtu < IPV6_MIN_MTU)
3565 addrconf_ifdown(dev, dev != net->loopback_dev);
3570 case NETDEV_UNREGISTER:
3572 * Remove all addresses from this interface.
3574 addrconf_ifdown(dev, event != NETDEV_DOWN);
3577 case NETDEV_CHANGENAME:
3579 snmp6_unregister_dev(idev);
3580 addrconf_sysctl_unregister(idev);
3581 err = addrconf_sysctl_register(idev);
3583 return notifier_from_errno(err);
3584 err = snmp6_register_dev(idev);
3586 addrconf_sysctl_unregister(idev);
3587 return notifier_from_errno(err);
3592 case NETDEV_PRE_TYPE_CHANGE:
3593 case NETDEV_POST_TYPE_CHANGE:
3595 addrconf_type_change(dev, event);
3598 case NETDEV_CHANGEUPPER:
3601 /* flush all routes if dev is linked to or unlinked from
3602 * an L3 master device (e.g., VRF)
3604 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3605 addrconf_ifdown(dev, 0);
3612 * addrconf module should be notified of a device going up
3614 static struct notifier_block ipv6_dev_notf = {
3615 .notifier_call = addrconf_notify,
3616 .priority = ADDRCONF_NOTIFY_PRIORITY,
3619 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3621 struct inet6_dev *idev;
3624 idev = __in6_dev_get(dev);
3626 if (event == NETDEV_POST_TYPE_CHANGE)
3627 ipv6_mc_remap(idev);
3628 else if (event == NETDEV_PRE_TYPE_CHANGE)
3629 ipv6_mc_unmap(idev);
3632 static bool addr_is_local(const struct in6_addr *addr)
3634 return ipv6_addr_type(addr) &
3635 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3638 static int addrconf_ifdown(struct net_device *dev, int how)
3640 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3641 struct net *net = dev_net(dev);
3642 struct inet6_dev *idev;
3643 struct inet6_ifaddr *ifa, *tmp;
3644 bool keep_addr = false;
3649 rt6_disable_ip(dev, event);
3651 idev = __in6_dev_get(dev);
3656 * Step 1: remove reference to ipv6 device from parent device.
3662 /* protected by rtnl_lock */
3663 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3665 /* Step 1.5: remove snmp6 entry */
3666 snmp6_unregister_dev(idev);
3670 /* combine the user config with event to determine if permanent
3671 * addresses are to be removed from address hash table
3673 if (!how && !idev->cnf.disable_ipv6) {
3674 /* aggregate the system setting and interface setting */
3675 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3678 _keep_addr = idev->cnf.keep_addr_on_down;
3680 keep_addr = (_keep_addr > 0);
3683 /* Step 2: clear hash table */
3684 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3685 struct hlist_head *h = &inet6_addr_lst[i];
3687 spin_lock_bh(&addrconf_hash_lock);
3689 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3690 if (ifa->idev == idev) {
3691 addrconf_del_dad_work(ifa);
3692 /* combined flag + permanent flag decide if
3693 * address is retained on a down event
3696 !(ifa->flags & IFA_F_PERMANENT) ||
3697 addr_is_local(&ifa->addr)) {
3698 hlist_del_init_rcu(&ifa->addr_lst);
3703 spin_unlock_bh(&addrconf_hash_lock);
3706 write_lock_bh(&idev->lock);
3708 addrconf_del_rs_timer(idev);
3710 /* Step 2: clear flags for stateless addrconf */
3712 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3714 /* Step 3: clear tempaddr list */
3715 while (!list_empty(&idev->tempaddr_list)) {
3716 ifa = list_first_entry(&idev->tempaddr_list,
3717 struct inet6_ifaddr, tmp_list);
3718 list_del(&ifa->tmp_list);
3719 write_unlock_bh(&idev->lock);
3720 spin_lock_bh(&ifa->lock);
3723 in6_ifa_put(ifa->ifpub);
3726 spin_unlock_bh(&ifa->lock);
3728 write_lock_bh(&idev->lock);
3731 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3732 struct fib6_info *rt = NULL;
3735 addrconf_del_dad_work(ifa);
3737 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3738 !addr_is_local(&ifa->addr);
3740 write_unlock_bh(&idev->lock);
3741 spin_lock_bh(&ifa->lock);
3744 /* set state to skip the notifier below */
3745 state = INET6_IFADDR_STATE_DEAD;
3746 ifa->state = INET6_IFADDR_STATE_PREDAD;
3747 if (!(ifa->flags & IFA_F_NODAD))
3748 ifa->flags |= IFA_F_TENTATIVE;
3754 ifa->state = INET6_IFADDR_STATE_DEAD;
3757 spin_unlock_bh(&ifa->lock);
3760 ip6_del_rt(net, rt);
3762 if (state != INET6_IFADDR_STATE_DEAD) {
3763 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3764 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3766 if (idev->cnf.forwarding)
3767 addrconf_leave_anycast(ifa);
3768 addrconf_leave_solict(ifa->idev, &ifa->addr);
3771 write_lock_bh(&idev->lock);
3773 list_del_rcu(&ifa->if_list);
3778 write_unlock_bh(&idev->lock);
3780 /* Step 5: Discard anycast and multicast list */
3782 ipv6_ac_destroy_dev(idev);
3783 ipv6_mc_destroy_dev(idev);
3788 idev->tstamp = jiffies;
3790 /* Last: Shot the device (if unregistered) */
3792 addrconf_sysctl_unregister(idev);
3793 neigh_parms_release(&nd_tbl, idev->nd_parms);
3794 neigh_ifdown(&nd_tbl, dev);
3800 static void addrconf_rs_timer(struct timer_list *t)
3802 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3803 struct net_device *dev = idev->dev;
3804 struct in6_addr lladdr;
3806 write_lock(&idev->lock);
3807 if (idev->dead || !(idev->if_flags & IF_READY))
3810 if (!ipv6_accept_ra(idev))
3813 /* Announcement received after solicitation was sent */
3814 if (idev->if_flags & IF_RA_RCVD)
3817 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3818 write_unlock(&idev->lock);
3819 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3820 ndisc_send_rs(dev, &lladdr,
3821 &in6addr_linklocal_allrouters);
3825 write_lock(&idev->lock);
3826 idev->rs_interval = rfc3315_s14_backoff_update(
3827 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3828 /* The wait after the last probe can be shorter */
3829 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3830 idev->cnf.rtr_solicits) ?
3831 idev->cnf.rtr_solicit_delay :
3835 * Note: we do not support deprecated "all on-link"
3836 * assumption any longer.
3838 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3842 write_unlock(&idev->lock);
3848 * Duplicate Address Detection
3850 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3852 unsigned long rand_num;
3853 struct inet6_dev *idev = ifp->idev;
3856 if (ifp->flags & IFA_F_OPTIMISTIC)
3859 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3862 if (idev->cnf.enhanced_dad ||
3863 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3865 get_random_bytes(&nonce, 6);
3868 ifp->dad_nonce = nonce;
3869 ifp->dad_probes = idev->cnf.dad_transmits;
3870 addrconf_mod_dad_work(ifp, rand_num);
3873 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3875 struct inet6_dev *idev = ifp->idev;
3876 struct net_device *dev = idev->dev;
3877 bool bump_id, notify = false;
3880 addrconf_join_solict(dev, &ifp->addr);
3882 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3884 read_lock_bh(&idev->lock);
3885 spin_lock(&ifp->lock);
3886 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3890 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3891 (net->ipv6.devconf_all->accept_dad < 1 &&
3892 idev->cnf.accept_dad < 1) ||
3893 !(ifp->flags&IFA_F_TENTATIVE) ||
3894 ifp->flags & IFA_F_NODAD) {
3895 bool send_na = false;
3897 if (ifp->flags & IFA_F_TENTATIVE &&
3898 !(ifp->flags & IFA_F_OPTIMISTIC))
3900 bump_id = ifp->flags & IFA_F_TENTATIVE;
3901 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3902 spin_unlock(&ifp->lock);
3903 read_unlock_bh(&idev->lock);
3905 addrconf_dad_completed(ifp, bump_id, send_na);
3909 if (!(idev->if_flags & IF_READY)) {
3910 spin_unlock(&ifp->lock);
3911 read_unlock_bh(&idev->lock);
3913 * If the device is not ready:
3914 * - keep it tentative if it is a permanent address.
3915 * - otherwise, kill it.
3918 addrconf_dad_stop(ifp, 0);
3923 * Optimistic nodes can start receiving
3926 if (ifp->flags & IFA_F_OPTIMISTIC) {
3927 ip6_ins_rt(net, ifp->rt);
3928 if (ipv6_use_optimistic_addr(net, idev)) {
3929 /* Because optimistic nodes can use this address,
3930 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3936 addrconf_dad_kick(ifp);
3938 spin_unlock(&ifp->lock);
3939 read_unlock_bh(&idev->lock);
3941 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3944 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3946 bool begin_dad = false;
3948 spin_lock_bh(&ifp->lock);
3949 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3950 ifp->state = INET6_IFADDR_STATE_PREDAD;
3953 spin_unlock_bh(&ifp->lock);
3956 addrconf_mod_dad_work(ifp, 0);
3959 static void addrconf_dad_work(struct work_struct *w)
3961 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3962 struct inet6_ifaddr,
3964 struct inet6_dev *idev = ifp->idev;
3965 bool bump_id, disable_ipv6 = false;
3966 struct in6_addr mcaddr;
3972 } action = DAD_PROCESS;
3976 spin_lock_bh(&ifp->lock);
3977 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3979 ifp->state = INET6_IFADDR_STATE_DAD;
3980 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3982 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3984 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3985 idev->cnf.accept_dad > 1) &&
3986 !idev->cnf.disable_ipv6 &&
3987 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3988 struct in6_addr addr;
3990 addr.s6_addr32[0] = htonl(0xfe800000);
3991 addr.s6_addr32[1] = 0;
3993 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3994 ipv6_addr_equal(&ifp->addr, &addr)) {
3995 /* DAD failed for link-local based on MAC */
3996 idev->cnf.disable_ipv6 = 1;
3998 pr_info("%s: IPv6 being disabled!\n",
3999 ifp->idev->dev->name);
4000 disable_ipv6 = true;
4004 spin_unlock_bh(&ifp->lock);
4006 if (action == DAD_BEGIN) {
4007 addrconf_dad_begin(ifp);
4009 } else if (action == DAD_ABORT) {
4011 addrconf_dad_stop(ifp, 1);
4013 addrconf_ifdown(idev->dev, 0);
4017 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4020 write_lock_bh(&idev->lock);
4021 if (idev->dead || !(idev->if_flags & IF_READY)) {
4022 write_unlock_bh(&idev->lock);
4026 spin_lock(&ifp->lock);
4027 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4028 spin_unlock(&ifp->lock);
4029 write_unlock_bh(&idev->lock);
4033 if (ifp->dad_probes == 0) {
4034 bool send_na = false;
4037 * DAD was successful
4040 if (ifp->flags & IFA_F_TENTATIVE &&
4041 !(ifp->flags & IFA_F_OPTIMISTIC))
4043 bump_id = ifp->flags & IFA_F_TENTATIVE;
4044 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4045 spin_unlock(&ifp->lock);
4046 write_unlock_bh(&idev->lock);
4048 addrconf_dad_completed(ifp, bump_id, send_na);
4054 addrconf_mod_dad_work(ifp,
4055 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4056 spin_unlock(&ifp->lock);
4057 write_unlock_bh(&idev->lock);
4059 /* send a neighbour solicitation for our addr */
4060 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4061 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4068 /* ifp->idev must be at least read locked */
4069 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4071 struct inet6_ifaddr *ifpiter;
4072 struct inet6_dev *idev = ifp->idev;
4074 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4075 if (ifpiter->scope > IFA_LINK)
4077 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4078 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4079 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4086 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4089 struct net_device *dev = ifp->idev->dev;
4090 struct in6_addr lladdr;
4091 bool send_rs, send_mld;
4093 addrconf_del_dad_work(ifp);
4096 * Configure the address for reception. Now it is valid.
4099 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4101 /* If added prefix is link local and we are prepared to process
4102 router advertisements, start sending router solicitations.
4105 read_lock_bh(&ifp->idev->lock);
4106 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4107 send_rs = send_mld &&
4108 ipv6_accept_ra(ifp->idev) &&
4109 ifp->idev->cnf.rtr_solicits != 0 &&
4110 (dev->flags&IFF_LOOPBACK) == 0;
4111 read_unlock_bh(&ifp->idev->lock);
4113 /* While dad is in progress mld report's source address is in6_addrany.
4114 * Resend with proper ll now.
4117 ipv6_mc_dad_complete(ifp->idev);
4119 /* send unsolicited NA if enabled */
4121 (ifp->idev->cnf.ndisc_notify ||
4122 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4123 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4124 /*router=*/ !!ifp->idev->cnf.forwarding,
4125 /*solicited=*/ false, /*override=*/ true,
4131 * If a host as already performed a random delay
4132 * [...] as part of DAD [...] there is no need
4133 * to delay again before sending the first RS
4135 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4137 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4139 write_lock_bh(&ifp->idev->lock);
4140 spin_lock(&ifp->lock);
4141 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4142 ifp->idev->cnf.rtr_solicit_interval);
4143 ifp->idev->rs_probes = 1;
4144 ifp->idev->if_flags |= IF_RS_SENT;
4145 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4146 spin_unlock(&ifp->lock);
4147 write_unlock_bh(&ifp->idev->lock);
4151 rt_genid_bump_ipv6(dev_net(dev));
4153 /* Make sure that a new temporary address will be created
4154 * before this temporary address becomes deprecated.
4156 if (ifp->flags & IFA_F_TEMPORARY)
4157 addrconf_verify_rtnl();
4160 static void addrconf_dad_run(struct inet6_dev *idev)
4162 struct inet6_ifaddr *ifp;
4164 read_lock_bh(&idev->lock);
4165 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4166 spin_lock(&ifp->lock);
4167 if (ifp->flags & IFA_F_TENTATIVE &&
4168 ifp->state == INET6_IFADDR_STATE_DAD)
4169 addrconf_dad_kick(ifp);
4170 spin_unlock(&ifp->lock);
4172 read_unlock_bh(&idev->lock);
4175 #ifdef CONFIG_PROC_FS
4176 struct if6_iter_state {
4177 struct seq_net_private p;
4182 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4184 struct if6_iter_state *state = seq->private;
4185 struct net *net = seq_file_net(seq);
4186 struct inet6_ifaddr *ifa = NULL;
4189 /* initial bucket if pos is 0 */
4195 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4196 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4198 if (!net_eq(dev_net(ifa->idev->dev), net))
4200 /* sync with offset */
4201 if (p < state->offset) {
4209 /* prepare for next bucket */
4216 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4217 struct inet6_ifaddr *ifa)
4219 struct if6_iter_state *state = seq->private;
4220 struct net *net = seq_file_net(seq);
4222 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4223 if (!net_eq(dev_net(ifa->idev->dev), net))
4229 while (++state->bucket < IN6_ADDR_HSIZE) {
4231 hlist_for_each_entry_rcu(ifa,
4232 &inet6_addr_lst[state->bucket], addr_lst) {
4233 if (!net_eq(dev_net(ifa->idev->dev), net))
4243 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4247 return if6_get_first(seq, *pos);
4250 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4252 struct inet6_ifaddr *ifa;
4254 ifa = if6_get_next(seq, v);
4259 static void if6_seq_stop(struct seq_file *seq, void *v)
4265 static int if6_seq_show(struct seq_file *seq, void *v)
4267 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4268 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4270 ifp->idev->dev->ifindex,
4274 ifp->idev->dev->name);
4278 static const struct seq_operations if6_seq_ops = {
4279 .start = if6_seq_start,
4280 .next = if6_seq_next,
4281 .show = if6_seq_show,
4282 .stop = if6_seq_stop,
4285 static int __net_init if6_proc_net_init(struct net *net)
4287 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4288 sizeof(struct if6_iter_state)))
4293 static void __net_exit if6_proc_net_exit(struct net *net)
4295 remove_proc_entry("if_inet6", net->proc_net);
4298 static struct pernet_operations if6_proc_net_ops = {
4299 .init = if6_proc_net_init,
4300 .exit = if6_proc_net_exit,
4303 int __init if6_proc_init(void)
4305 return register_pernet_subsys(&if6_proc_net_ops);
4308 void if6_proc_exit(void)
4310 unregister_pernet_subsys(&if6_proc_net_ops);
4312 #endif /* CONFIG_PROC_FS */
4314 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4315 /* Check if address is a home address configured on any interface. */
4316 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4318 unsigned int hash = inet6_addr_hash(net, addr);
4319 struct inet6_ifaddr *ifp = NULL;
4323 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4324 if (!net_eq(dev_net(ifp->idev->dev), net))
4326 if (ipv6_addr_equal(&ifp->addr, addr) &&
4327 (ifp->flags & IFA_F_HOMEADDRESS)) {
4338 * Periodic address status verification
4341 static void addrconf_verify_rtnl(void)
4343 unsigned long now, next, next_sec, next_sched;
4344 struct inet6_ifaddr *ifp;
4351 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4353 cancel_delayed_work(&addr_chk_work);
4355 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4357 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4360 /* When setting preferred_lft to a value not zero or
4361 * infinity, while valid_lft is infinity
4362 * IFA_F_PERMANENT has a non-infinity life time.
4364 if ((ifp->flags & IFA_F_PERMANENT) &&
4365 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4368 spin_lock(&ifp->lock);
4369 /* We try to batch several events at once. */
4370 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4372 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4373 age >= ifp->valid_lft) {
4374 spin_unlock(&ifp->lock);
4378 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4379 spin_unlock(&ifp->lock);
4381 } else if (age >= ifp->prefered_lft) {
4382 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4385 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4387 ifp->flags |= IFA_F_DEPRECATED;
4390 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4391 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4392 next = ifp->tstamp + ifp->valid_lft * HZ;
4394 spin_unlock(&ifp->lock);
4399 ipv6_ifa_notify(0, ifp);
4403 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4404 !(ifp->flags&IFA_F_TENTATIVE)) {
4405 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4406 ifp->idev->cnf.dad_transmits *
4407 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4409 if (age >= ifp->prefered_lft - regen_advance) {
4410 struct inet6_ifaddr *ifpub = ifp->ifpub;
4411 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4412 next = ifp->tstamp + ifp->prefered_lft * HZ;
4413 if (!ifp->regen_count && ifpub) {
4416 in6_ifa_hold(ifpub);
4417 spin_unlock(&ifp->lock);
4419 spin_lock(&ifpub->lock);
4420 ifpub->regen_count = 0;
4421 spin_unlock(&ifpub->lock);
4422 rcu_read_unlock_bh();
4423 ipv6_create_tempaddr(ifpub, ifp, true);
4429 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4430 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4431 spin_unlock(&ifp->lock);
4433 /* ifp->prefered_lft <= ifp->valid_lft */
4434 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4435 next = ifp->tstamp + ifp->prefered_lft * HZ;
4436 spin_unlock(&ifp->lock);
4441 next_sec = round_jiffies_up(next);
4444 /* If rounded timeout is accurate enough, accept it. */
4445 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4446 next_sched = next_sec;
4448 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4449 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4450 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4452 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4453 now, next, next_sec, next_sched);
4454 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4455 rcu_read_unlock_bh();
4458 static void addrconf_verify_work(struct work_struct *w)
4461 addrconf_verify_rtnl();
4465 static void addrconf_verify(void)
4467 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4470 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4471 struct in6_addr **peer_pfx)
4473 struct in6_addr *pfx = NULL;
4478 pfx = nla_data(addr);
4481 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4483 pfx = nla_data(local);
4489 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4490 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4491 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4492 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4493 [IFA_FLAGS] = { .len = sizeof(u32) },
4494 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4498 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4499 struct netlink_ext_ack *extack)
4501 struct net *net = sock_net(skb->sk);
4502 struct ifaddrmsg *ifm;
4503 struct nlattr *tb[IFA_MAX+1];
4504 struct in6_addr *pfx, *peer_pfx;
4508 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4513 ifm = nlmsg_data(nlh);
4514 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4518 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4520 /* We ignore other flags so far. */
4521 ifa_flags &= IFA_F_MANAGETEMPADDR;
4523 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4524 ifm->ifa_prefixlen);
4527 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4528 unsigned long expires, u32 flags)
4530 struct fib6_info *f6i;
4533 f6i = addrconf_get_prefix_route(&ifp->addr,
4536 0, RTF_GATEWAY | RTF_DEFAULT);
4540 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4541 if (f6i->fib6_metric != prio) {
4542 /* delete old one */
4543 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4546 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4547 ifp->rt_priority, ifp->idev->dev,
4548 expires, flags, GFP_KERNEL);
4551 fib6_clean_expires(f6i);
4553 fib6_set_expires(f6i, expires);
4555 fib6_info_release(f6i);
4561 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4565 unsigned long timeout;
4566 bool was_managetempaddr;
4567 bool had_prefixroute;
4571 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4574 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4575 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4578 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4579 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4581 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4582 if (addrconf_finite_timeout(timeout)) {
4583 expires = jiffies_to_clock_t(timeout * HZ);
4584 cfg->valid_lft = timeout;
4585 flags = RTF_EXPIRES;
4589 cfg->ifa_flags |= IFA_F_PERMANENT;
4592 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4593 if (addrconf_finite_timeout(timeout)) {
4595 cfg->ifa_flags |= IFA_F_DEPRECATED;
4596 cfg->preferred_lft = timeout;
4599 spin_lock_bh(&ifp->lock);
4600 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4601 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4602 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4603 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4604 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4605 IFA_F_NOPREFIXROUTE);
4606 ifp->flags |= cfg->ifa_flags;
4607 ifp->tstamp = jiffies;
4608 ifp->valid_lft = cfg->valid_lft;
4609 ifp->prefered_lft = cfg->preferred_lft;
4611 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4612 ifp->rt_priority = cfg->rt_priority;
4614 spin_unlock_bh(&ifp->lock);
4615 if (!(ifp->flags&IFA_F_TENTATIVE))
4616 ipv6_ifa_notify(0, ifp);
4618 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4621 if (had_prefixroute)
4622 rc = modify_prefix_route(ifp, expires, flags);
4624 /* prefix route could have been deleted; if so restore it */
4625 if (rc == -ENOENT) {
4626 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4627 ifp->rt_priority, ifp->idev->dev,
4628 expires, flags, GFP_KERNEL);
4630 } else if (had_prefixroute) {
4631 enum cleanup_prefix_rt_t action;
4632 unsigned long rt_expires;
4634 write_lock_bh(&ifp->idev->lock);
4635 action = check_cleanup_prefix_route(ifp, &rt_expires);
4636 write_unlock_bh(&ifp->idev->lock);
4638 if (action != CLEANUP_PREFIX_RT_NOP) {
4639 cleanup_prefix_route(ifp, rt_expires,
4640 action == CLEANUP_PREFIX_RT_DEL);
4644 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4645 if (was_managetempaddr &&
4646 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4648 cfg->preferred_lft = 0;
4650 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4651 cfg->preferred_lft, !was_managetempaddr,
4655 addrconf_verify_rtnl();
4661 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4662 struct netlink_ext_ack *extack)
4664 struct net *net = sock_net(skb->sk);
4665 struct ifaddrmsg *ifm;
4666 struct nlattr *tb[IFA_MAX+1];
4667 struct in6_addr *peer_pfx;
4668 struct inet6_ifaddr *ifa;
4669 struct net_device *dev;
4670 struct inet6_dev *idev;
4671 struct ifa6_config cfg;
4674 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4679 memset(&cfg, 0, sizeof(cfg));
4681 ifm = nlmsg_data(nlh);
4682 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4686 cfg.peer_pfx = peer_pfx;
4687 cfg.plen = ifm->ifa_prefixlen;
4688 if (tb[IFA_RT_PRIORITY])
4689 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4691 cfg.valid_lft = INFINITY_LIFE_TIME;
4692 cfg.preferred_lft = INFINITY_LIFE_TIME;
4694 if (tb[IFA_CACHEINFO]) {
4695 struct ifa_cacheinfo *ci;
4697 ci = nla_data(tb[IFA_CACHEINFO]);
4698 cfg.valid_lft = ci->ifa_valid;
4699 cfg.preferred_lft = ci->ifa_prefered;
4702 dev = __dev_get_by_index(net, ifm->ifa_index);
4707 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4709 cfg.ifa_flags = ifm->ifa_flags;
4711 /* We ignore other flags so far. */
4712 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4713 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4714 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4716 idev = ipv6_find_idev(dev);
4718 return PTR_ERR(idev);
4720 if (!ipv6_allow_optimistic_dad(net, idev))
4721 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4723 if (cfg.ifa_flags & IFA_F_NODAD &&
4724 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4725 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4729 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4732 * It would be best to check for !NLM_F_CREATE here but
4733 * userspace already relies on not having to provide this.
4735 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4738 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4739 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4742 err = inet6_addr_modify(ifa, &cfg);
4749 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4750 u8 scope, int ifindex)
4752 struct ifaddrmsg *ifm;
4754 ifm = nlmsg_data(nlh);
4755 ifm->ifa_family = AF_INET6;
4756 ifm->ifa_prefixlen = prefixlen;
4757 ifm->ifa_flags = flags;
4758 ifm->ifa_scope = scope;
4759 ifm->ifa_index = ifindex;
4762 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4763 unsigned long tstamp, u32 preferred, u32 valid)
4765 struct ifa_cacheinfo ci;
4767 ci.cstamp = cstamp_delta(cstamp);
4768 ci.tstamp = cstamp_delta(tstamp);
4769 ci.ifa_prefered = preferred;
4770 ci.ifa_valid = valid;
4772 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4775 static inline int rt_scope(int ifa_scope)
4777 if (ifa_scope & IFA_HOST)
4778 return RT_SCOPE_HOST;
4779 else if (ifa_scope & IFA_LINK)
4780 return RT_SCOPE_LINK;
4781 else if (ifa_scope & IFA_SITE)
4782 return RT_SCOPE_SITE;
4784 return RT_SCOPE_UNIVERSE;
4787 static inline int inet6_ifaddr_msgsize(void)
4789 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4790 + nla_total_size(16) /* IFA_LOCAL */
4791 + nla_total_size(16) /* IFA_ADDRESS */
4792 + nla_total_size(sizeof(struct ifa_cacheinfo))
4793 + nla_total_size(4) /* IFA_FLAGS */
4794 + nla_total_size(4) /* IFA_RT_PRIORITY */;
4797 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4798 u32 portid, u32 seq, int event, unsigned int flags)
4800 struct nlmsghdr *nlh;
4801 u32 preferred, valid;
4803 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4807 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4808 ifa->idev->dev->ifindex);
4810 if (!((ifa->flags&IFA_F_PERMANENT) &&
4811 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4812 preferred = ifa->prefered_lft;
4813 valid = ifa->valid_lft;
4814 if (preferred != INFINITY_LIFE_TIME) {
4815 long tval = (jiffies - ifa->tstamp)/HZ;
4816 if (preferred > tval)
4820 if (valid != INFINITY_LIFE_TIME) {
4828 preferred = INFINITY_LIFE_TIME;
4829 valid = INFINITY_LIFE_TIME;
4832 if (!ipv6_addr_any(&ifa->peer_addr)) {
4833 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4834 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4837 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4840 if (ifa->rt_priority &&
4841 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4844 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4847 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4850 nlmsg_end(skb, nlh);
4854 nlmsg_cancel(skb, nlh);
4858 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4859 u32 portid, u32 seq, int event, u16 flags)
4861 struct nlmsghdr *nlh;
4862 u8 scope = RT_SCOPE_UNIVERSE;
4863 int ifindex = ifmca->idev->dev->ifindex;
4865 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4866 scope = RT_SCOPE_SITE;
4868 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4872 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4873 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4874 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4875 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4876 nlmsg_cancel(skb, nlh);
4880 nlmsg_end(skb, nlh);
4884 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4885 u32 portid, u32 seq, int event, unsigned int flags)
4887 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4888 int ifindex = dev ? dev->ifindex : 1;
4889 struct nlmsghdr *nlh;
4890 u8 scope = RT_SCOPE_UNIVERSE;
4892 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4893 scope = RT_SCOPE_SITE;
4895 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4899 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4900 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4901 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4902 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4903 nlmsg_cancel(skb, nlh);
4907 nlmsg_end(skb, nlh);
4917 /* called with rcu_read_lock() */
4918 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4919 struct netlink_callback *cb, enum addr_type_t type,
4920 int s_ip_idx, int *p_ip_idx)
4922 struct ifmcaddr6 *ifmca;
4923 struct ifacaddr6 *ifaca;
4925 int ip_idx = *p_ip_idx;
4927 read_lock_bh(&idev->lock);
4929 case UNICAST_ADDR: {
4930 struct inet6_ifaddr *ifa;
4932 /* unicast address incl. temp addr */
4933 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4934 if (++ip_idx < s_ip_idx)
4936 err = inet6_fill_ifaddr(skb, ifa,
4937 NETLINK_CB(cb->skb).portid,
4943 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4947 case MULTICAST_ADDR:
4948 /* multicast address */
4949 for (ifmca = idev->mc_list; ifmca;
4950 ifmca = ifmca->next, ip_idx++) {
4951 if (ip_idx < s_ip_idx)
4953 err = inet6_fill_ifmcaddr(skb, ifmca,
4954 NETLINK_CB(cb->skb).portid,
4963 /* anycast address */
4964 for (ifaca = idev->ac_list; ifaca;
4965 ifaca = ifaca->aca_next, ip_idx++) {
4966 if (ip_idx < s_ip_idx)
4968 err = inet6_fill_ifacaddr(skb, ifaca,
4969 NETLINK_CB(cb->skb).portid,
4980 read_unlock_bh(&idev->lock);
4985 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4986 enum addr_type_t type)
4988 struct net *net = sock_net(skb->sk);
4991 int s_idx, s_ip_idx;
4992 struct net_device *dev;
4993 struct inet6_dev *idev;
4994 struct hlist_head *head;
4997 s_idx = idx = cb->args[1];
4998 s_ip_idx = ip_idx = cb->args[2];
5001 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
5002 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5004 head = &net->dev_index_head[h];
5005 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5008 if (h > s_h || idx > s_idx)
5011 idev = __in6_dev_get(dev);
5015 if (in6_dump_addrs(idev, skb, cb, type,
5016 s_ip_idx, &ip_idx) < 0)
5026 cb->args[2] = ip_idx;
5031 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5033 enum addr_type_t type = UNICAST_ADDR;
5035 return inet6_dump_addr(skb, cb, type);
5038 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5040 enum addr_type_t type = MULTICAST_ADDR;
5042 return inet6_dump_addr(skb, cb, type);
5046 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5048 enum addr_type_t type = ANYCAST_ADDR;
5050 return inet6_dump_addr(skb, cb, type);
5053 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5054 struct netlink_ext_ack *extack)
5056 struct net *net = sock_net(in_skb->sk);
5057 struct ifaddrmsg *ifm;
5058 struct nlattr *tb[IFA_MAX+1];
5059 struct in6_addr *addr = NULL, *peer;
5060 struct net_device *dev = NULL;
5061 struct inet6_ifaddr *ifa;
5062 struct sk_buff *skb;
5065 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5070 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5074 ifm = nlmsg_data(nlh);
5076 dev = dev_get_by_index(net, ifm->ifa_index);
5078 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5080 err = -EADDRNOTAVAIL;
5084 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5090 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5091 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5093 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5094 WARN_ON(err == -EMSGSIZE);
5098 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5107 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5109 struct sk_buff *skb;
5110 struct net *net = dev_net(ifa->idev->dev);
5113 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5117 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5119 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5120 WARN_ON(err == -EMSGSIZE);
5124 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5128 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5131 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5132 __s32 *array, int bytes)
5134 BUG_ON(bytes < (DEVCONF_MAX * 4));
5136 memset(array, 0, bytes);
5137 array[DEVCONF_FORWARDING] = cnf->forwarding;
5138 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5139 array[DEVCONF_MTU6] = cnf->mtu6;
5140 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5141 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5142 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5143 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5144 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5145 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5146 jiffies_to_msecs(cnf->rtr_solicit_interval);
5147 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5148 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5149 array[DEVCONF_RTR_SOLICIT_DELAY] =
5150 jiffies_to_msecs(cnf->rtr_solicit_delay);
5151 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5152 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5153 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5154 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5155 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5156 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5157 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5158 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5159 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5160 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5161 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5162 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5163 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5164 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5165 #ifdef CONFIG_IPV6_ROUTER_PREF
5166 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5167 array[DEVCONF_RTR_PROBE_INTERVAL] =
5168 jiffies_to_msecs(cnf->rtr_probe_interval);
5169 #ifdef CONFIG_IPV6_ROUTE_INFO
5170 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5171 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5174 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5175 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5176 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5177 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5178 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5180 #ifdef CONFIG_IPV6_MROUTE
5181 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5183 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5184 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5185 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5186 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5187 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5188 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5189 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5190 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5191 /* we omit DEVCONF_STABLE_SECRET for now */
5192 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5193 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5194 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5195 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5196 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5197 #ifdef CONFIG_IPV6_SEG6_HMAC
5198 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5200 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5201 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5202 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5203 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5206 static inline size_t inet6_ifla6_size(void)
5208 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5209 + nla_total_size(sizeof(struct ifla_cacheinfo))
5210 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5211 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5212 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5213 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5216 static inline size_t inet6_if_nlmsg_size(void)
5218 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5219 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5220 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5221 + nla_total_size(4) /* IFLA_MTU */
5222 + nla_total_size(4) /* IFLA_LINK */
5223 + nla_total_size(1) /* IFLA_OPERSTATE */
5224 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5227 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5231 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5234 /* Use put_unaligned() because stats may not be aligned for u64. */
5235 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5236 for (i = 1; i < ICMP6_MIB_MAX; i++)
5237 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5239 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5242 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5243 int bytes, size_t syncpoff)
5246 u64 buff[IPSTATS_MIB_MAX];
5247 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5251 memset(buff, 0, sizeof(buff));
5252 buff[0] = IPSTATS_MIB_MAX;
5254 for_each_possible_cpu(c) {
5255 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5256 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5259 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5260 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5263 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5267 case IFLA_INET6_STATS:
5268 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5269 offsetof(struct ipstats_mib, syncp));
5271 case IFLA_INET6_ICMP6STATS:
5272 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5277 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5278 u32 ext_filter_mask)
5281 struct ifla_cacheinfo ci;
5283 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5284 goto nla_put_failure;
5285 ci.max_reasm_len = IPV6_MAXPLEN;
5286 ci.tstamp = cstamp_delta(idev->tstamp);
5287 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5288 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5289 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5290 goto nla_put_failure;
5291 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5293 goto nla_put_failure;
5294 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5296 /* XXX - MC not implemented */
5298 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5301 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5303 goto nla_put_failure;
5304 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5306 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5308 goto nla_put_failure;
5309 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5311 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5313 goto nla_put_failure;
5315 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5316 goto nla_put_failure;
5318 read_lock_bh(&idev->lock);
5319 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5320 read_unlock_bh(&idev->lock);
5328 static size_t inet6_get_link_af_size(const struct net_device *dev,
5329 u32 ext_filter_mask)
5331 if (!__in6_dev_get(dev))
5334 return inet6_ifla6_size();
5337 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5338 u32 ext_filter_mask)
5340 struct inet6_dev *idev = __in6_dev_get(dev);
5345 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5351 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5353 struct inet6_ifaddr *ifp;
5354 struct net_device *dev = idev->dev;
5355 bool clear_token, update_rs = false;
5356 struct in6_addr ll_addr;
5362 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5364 if (!ipv6_accept_ra(idev))
5366 if (idev->cnf.rtr_solicits == 0)
5369 write_lock_bh(&idev->lock);
5371 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5372 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5374 write_unlock_bh(&idev->lock);
5376 clear_token = ipv6_addr_any(token);
5380 if (!idev->dead && (idev->if_flags & IF_READY) &&
5381 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5382 IFA_F_OPTIMISTIC)) {
5383 /* If we're not ready, then normal ifup will take care
5384 * of this. Otherwise, we need to request our rs here.
5386 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5391 write_lock_bh(&idev->lock);
5394 idev->if_flags |= IF_RS_SENT;
5395 idev->rs_interval = rfc3315_s14_backoff_init(
5396 idev->cnf.rtr_solicit_interval);
5397 idev->rs_probes = 1;
5398 addrconf_mod_rs_timer(idev, idev->rs_interval);
5401 /* Well, that's kinda nasty ... */
5402 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5403 spin_lock(&ifp->lock);
5404 if (ifp->tokenized) {
5406 ifp->prefered_lft = 0;
5408 spin_unlock(&ifp->lock);
5411 write_unlock_bh(&idev->lock);
5412 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5413 addrconf_verify_rtnl();
5417 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5418 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5419 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5422 static int inet6_validate_link_af(const struct net_device *dev,
5423 const struct nlattr *nla)
5425 struct nlattr *tb[IFLA_INET6_MAX + 1];
5427 if (dev && !__in6_dev_get(dev))
5428 return -EAFNOSUPPORT;
5430 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5434 static int check_addr_gen_mode(int mode)
5436 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5437 mode != IN6_ADDR_GEN_MODE_NONE &&
5438 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5439 mode != IN6_ADDR_GEN_MODE_RANDOM)
5444 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5447 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5448 !idev->cnf.stable_secret.initialized &&
5449 !net->ipv6.devconf_dflt->stable_secret.initialized)
5454 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5457 struct inet6_dev *idev = __in6_dev_get(dev);
5458 struct nlattr *tb[IFLA_INET6_MAX + 1];
5461 return -EAFNOSUPPORT;
5463 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5466 if (tb[IFLA_INET6_TOKEN]) {
5467 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5472 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5473 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5475 if (check_addr_gen_mode(mode) < 0 ||
5476 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5479 idev->cnf.addr_gen_mode = mode;
5486 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5487 u32 portid, u32 seq, int event, unsigned int flags)
5489 struct net_device *dev = idev->dev;
5490 struct ifinfomsg *hdr;
5491 struct nlmsghdr *nlh;
5494 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5498 hdr = nlmsg_data(nlh);
5499 hdr->ifi_family = AF_INET6;
5501 hdr->ifi_type = dev->type;
5502 hdr->ifi_index = dev->ifindex;
5503 hdr->ifi_flags = dev_get_flags(dev);
5504 hdr->ifi_change = 0;
5506 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5508 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5509 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5510 (dev->ifindex != dev_get_iflink(dev) &&
5511 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5512 nla_put_u8(skb, IFLA_OPERSTATE,
5513 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5514 goto nla_put_failure;
5515 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5517 goto nla_put_failure;
5519 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5520 goto nla_put_failure;
5522 nla_nest_end(skb, protoinfo);
5523 nlmsg_end(skb, nlh);
5527 nlmsg_cancel(skb, nlh);
5531 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5533 struct net *net = sock_net(skb->sk);
5536 struct net_device *dev;
5537 struct inet6_dev *idev;
5538 struct hlist_head *head;
5541 s_idx = cb->args[1];
5544 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5546 head = &net->dev_index_head[h];
5547 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5550 idev = __in6_dev_get(dev);
5553 if (inet6_fill_ifinfo(skb, idev,
5554 NETLINK_CB(cb->skb).portid,
5556 RTM_NEWLINK, NLM_F_MULTI) < 0)
5570 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5572 struct sk_buff *skb;
5573 struct net *net = dev_net(idev->dev);
5576 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5580 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5582 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5583 WARN_ON(err == -EMSGSIZE);
5587 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5591 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5594 static inline size_t inet6_prefix_nlmsg_size(void)
5596 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5597 + nla_total_size(sizeof(struct in6_addr))
5598 + nla_total_size(sizeof(struct prefix_cacheinfo));
5601 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5602 struct prefix_info *pinfo, u32 portid, u32 seq,
5603 int event, unsigned int flags)
5605 struct prefixmsg *pmsg;
5606 struct nlmsghdr *nlh;
5607 struct prefix_cacheinfo ci;
5609 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5613 pmsg = nlmsg_data(nlh);
5614 pmsg->prefix_family = AF_INET6;
5615 pmsg->prefix_pad1 = 0;
5616 pmsg->prefix_pad2 = 0;
5617 pmsg->prefix_ifindex = idev->dev->ifindex;
5618 pmsg->prefix_len = pinfo->prefix_len;
5619 pmsg->prefix_type = pinfo->type;
5620 pmsg->prefix_pad3 = 0;
5621 pmsg->prefix_flags = 0;
5623 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5624 if (pinfo->autoconf)
5625 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5627 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5628 goto nla_put_failure;
5629 ci.preferred_time = ntohl(pinfo->prefered);
5630 ci.valid_time = ntohl(pinfo->valid);
5631 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5632 goto nla_put_failure;
5633 nlmsg_end(skb, nlh);
5637 nlmsg_cancel(skb, nlh);
5641 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5642 struct prefix_info *pinfo)
5644 struct sk_buff *skb;
5645 struct net *net = dev_net(idev->dev);
5648 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5652 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5654 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5655 WARN_ON(err == -EMSGSIZE);
5659 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5663 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5666 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5668 struct net *net = dev_net(ifp->idev->dev);
5673 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5678 * If the address was optimistic
5679 * we inserted the route at the start of
5680 * our DAD process, so we don't need
5683 if (!rcu_access_pointer(ifp->rt->fib6_node))
5684 ip6_ins_rt(net, ifp->rt);
5685 if (ifp->idev->cnf.forwarding)
5686 addrconf_join_anycast(ifp);
5687 if (!ipv6_addr_any(&ifp->peer_addr))
5688 addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5689 ifp->idev->dev, 0, 0,
5693 if (ifp->idev->cnf.forwarding)
5694 addrconf_leave_anycast(ifp);
5695 addrconf_leave_solict(ifp->idev, &ifp->addr);
5696 if (!ipv6_addr_any(&ifp->peer_addr)) {
5697 struct fib6_info *rt;
5699 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5700 ifp->idev->dev, 0, 0);
5702 ip6_del_rt(net, rt);
5705 ip6_del_rt(net, ifp->rt);
5708 rt_genid_bump_ipv6(net);
5711 atomic_inc(&net->ipv6.dev_addr_genid);
5714 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5717 if (likely(ifp->idev->dead == 0))
5718 __ipv6_ifa_notify(event, ifp);
5719 rcu_read_unlock_bh();
5722 #ifdef CONFIG_SYSCTL
5725 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5726 void __user *buffer, size_t *lenp, loff_t *ppos)
5728 int *valp = ctl->data;
5731 struct ctl_table lctl;
5735 * ctl->data points to idev->cnf.forwarding, we should
5736 * not modify it until we get the rtnl lock.
5741 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5744 ret = addrconf_fixup_forwarding(ctl, valp, val);
5751 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5752 void __user *buffer, size_t *lenp, loff_t *ppos)
5754 struct inet6_dev *idev = ctl->extra1;
5755 int min_mtu = IPV6_MIN_MTU;
5756 struct ctl_table lctl;
5759 lctl.extra1 = &min_mtu;
5760 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5762 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5765 static void dev_disable_change(struct inet6_dev *idev)
5767 struct netdev_notifier_info info;
5769 if (!idev || !idev->dev)
5772 netdev_notifier_info_init(&info, idev->dev);
5773 if (idev->cnf.disable_ipv6)
5774 addrconf_notify(NULL, NETDEV_DOWN, &info);
5776 addrconf_notify(NULL, NETDEV_UP, &info);
5779 static void addrconf_disable_change(struct net *net, __s32 newf)
5781 struct net_device *dev;
5782 struct inet6_dev *idev;
5784 for_each_netdev(net, dev) {
5785 idev = __in6_dev_get(dev);
5787 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5788 idev->cnf.disable_ipv6 = newf;
5790 dev_disable_change(idev);
5795 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5800 if (!rtnl_trylock())
5801 return restart_syscall();
5803 net = (struct net *)table->extra2;
5807 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5812 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5813 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5814 addrconf_disable_change(net, newf);
5815 } else if ((!newf) ^ (!old))
5816 dev_disable_change((struct inet6_dev *)table->extra1);
5823 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5824 void __user *buffer, size_t *lenp, loff_t *ppos)
5826 int *valp = ctl->data;
5829 struct ctl_table lctl;
5833 * ctl->data points to idev->cnf.disable_ipv6, we should
5834 * not modify it until we get the rtnl lock.
5839 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5842 ret = addrconf_disable_ipv6(ctl, valp, val);
5849 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5850 void __user *buffer, size_t *lenp, loff_t *ppos)
5852 int *valp = ctl->data;
5857 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5860 if (write && old != new) {
5861 struct net *net = ctl->extra2;
5863 if (!rtnl_trylock())
5864 return restart_syscall();
5866 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5867 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5868 NETCONFA_PROXY_NEIGH,
5869 NETCONFA_IFINDEX_DEFAULT,
5870 net->ipv6.devconf_dflt);
5871 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5872 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5873 NETCONFA_PROXY_NEIGH,
5874 NETCONFA_IFINDEX_ALL,
5875 net->ipv6.devconf_all);
5877 struct inet6_dev *idev = ctl->extra1;
5879 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5880 NETCONFA_PROXY_NEIGH,
5890 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5891 void __user *buffer, size_t *lenp,
5896 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5897 struct net *net = (struct net *)ctl->extra2;
5899 if (!rtnl_trylock())
5900 return restart_syscall();
5902 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5905 new_val = *((int *)ctl->data);
5907 if (check_addr_gen_mode(new_val) < 0) {
5912 /* request for default */
5913 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5914 ipv6_devconf_dflt.addr_gen_mode = new_val;
5916 /* request for individual net device */
5921 if (check_stable_privacy(idev, net, new_val) < 0) {
5926 if (idev->cnf.addr_gen_mode != new_val) {
5927 idev->cnf.addr_gen_mode = new_val;
5928 addrconf_dev_config(idev->dev);
5939 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5940 void __user *buffer, size_t *lenp,
5944 struct in6_addr addr;
5945 char str[IPV6_MAX_STRLEN];
5946 struct ctl_table lctl = *ctl;
5947 struct net *net = ctl->extra2;
5948 struct ipv6_stable_secret *secret = ctl->data;
5950 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5953 lctl.maxlen = IPV6_MAX_STRLEN;
5956 if (!rtnl_trylock())
5957 return restart_syscall();
5959 if (!write && !secret->initialized) {
5964 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5965 if (err >= sizeof(str)) {
5970 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5974 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5979 secret->initialized = true;
5980 secret->secret = addr;
5982 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5983 struct net_device *dev;
5985 for_each_netdev(net, dev) {
5986 struct inet6_dev *idev = __in6_dev_get(dev);
5989 idev->cnf.addr_gen_mode =
5990 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5994 struct inet6_dev *idev = ctl->extra1;
5996 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6006 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6008 void __user *buffer,
6012 int *valp = ctl->data;
6015 struct ctl_table lctl;
6018 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6019 * we should not modify it until we get the rtnl lock.
6024 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6027 ret = addrconf_fixup_linkdown(ctl, valp, val);
6034 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6038 rt->dst.flags |= DST_NOPOLICY;
6040 rt->dst.flags &= ~DST_NOPOLICY;
6045 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6047 struct inet6_ifaddr *ifa;
6049 read_lock_bh(&idev->lock);
6050 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6051 spin_lock(&ifa->lock);
6053 struct fib6_info *rt = ifa->rt;
6057 ifa->rt->dst_nopolicy = val ? true : false;
6058 if (rt->rt6i_pcpu) {
6059 for_each_possible_cpu(cpu) {
6060 struct rt6_info **rtp;
6062 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6063 addrconf_set_nopolicy(*rtp, val);
6068 spin_unlock(&ifa->lock);
6070 read_unlock_bh(&idev->lock);
6074 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6076 struct inet6_dev *idev;
6079 if (!rtnl_trylock())
6080 return restart_syscall();
6084 net = (struct net *)ctl->extra2;
6085 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6090 if (valp == &net->ipv6.devconf_all->disable_policy) {
6091 struct net_device *dev;
6093 for_each_netdev(net, dev) {
6094 idev = __in6_dev_get(dev);
6096 addrconf_disable_policy_idev(idev, val);
6099 idev = (struct inet6_dev *)ctl->extra1;
6100 addrconf_disable_policy_idev(idev, val);
6108 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6109 void __user *buffer, size_t *lenp,
6112 int *valp = ctl->data;
6115 struct ctl_table lctl;
6120 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6122 if (write && (*valp != val))
6123 ret = addrconf_disable_policy(ctl, valp, val);
6131 static int minus_one = -1;
6132 static const int zero = 0;
6133 static const int one = 1;
6134 static const int two_five_five = 255;
6136 static const struct ctl_table addrconf_sysctl[] = {
6138 .procname = "forwarding",
6139 .data = &ipv6_devconf.forwarding,
6140 .maxlen = sizeof(int),
6142 .proc_handler = addrconf_sysctl_forward,
6145 .procname = "hop_limit",
6146 .data = &ipv6_devconf.hop_limit,
6147 .maxlen = sizeof(int),
6149 .proc_handler = proc_dointvec_minmax,
6150 .extra1 = (void *)&one,
6151 .extra2 = (void *)&two_five_five,
6155 .data = &ipv6_devconf.mtu6,
6156 .maxlen = sizeof(int),
6158 .proc_handler = addrconf_sysctl_mtu,
6161 .procname = "accept_ra",
6162 .data = &ipv6_devconf.accept_ra,
6163 .maxlen = sizeof(int),
6165 .proc_handler = proc_dointvec,
6168 .procname = "accept_redirects",
6169 .data = &ipv6_devconf.accept_redirects,
6170 .maxlen = sizeof(int),
6172 .proc_handler = proc_dointvec,
6175 .procname = "autoconf",
6176 .data = &ipv6_devconf.autoconf,
6177 .maxlen = sizeof(int),
6179 .proc_handler = proc_dointvec,
6182 .procname = "dad_transmits",
6183 .data = &ipv6_devconf.dad_transmits,
6184 .maxlen = sizeof(int),
6186 .proc_handler = proc_dointvec,
6189 .procname = "router_solicitations",
6190 .data = &ipv6_devconf.rtr_solicits,
6191 .maxlen = sizeof(int),
6193 .proc_handler = proc_dointvec_minmax,
6194 .extra1 = &minus_one,
6197 .procname = "router_solicitation_interval",
6198 .data = &ipv6_devconf.rtr_solicit_interval,
6199 .maxlen = sizeof(int),
6201 .proc_handler = proc_dointvec_jiffies,
6204 .procname = "router_solicitation_max_interval",
6205 .data = &ipv6_devconf.rtr_solicit_max_interval,
6206 .maxlen = sizeof(int),
6208 .proc_handler = proc_dointvec_jiffies,
6211 .procname = "router_solicitation_delay",
6212 .data = &ipv6_devconf.rtr_solicit_delay,
6213 .maxlen = sizeof(int),
6215 .proc_handler = proc_dointvec_jiffies,
6218 .procname = "force_mld_version",
6219 .data = &ipv6_devconf.force_mld_version,
6220 .maxlen = sizeof(int),
6222 .proc_handler = proc_dointvec,
6225 .procname = "mldv1_unsolicited_report_interval",
6227 &ipv6_devconf.mldv1_unsolicited_report_interval,
6228 .maxlen = sizeof(int),
6230 .proc_handler = proc_dointvec_ms_jiffies,
6233 .procname = "mldv2_unsolicited_report_interval",
6235 &ipv6_devconf.mldv2_unsolicited_report_interval,
6236 .maxlen = sizeof(int),
6238 .proc_handler = proc_dointvec_ms_jiffies,
6241 .procname = "use_tempaddr",
6242 .data = &ipv6_devconf.use_tempaddr,
6243 .maxlen = sizeof(int),
6245 .proc_handler = proc_dointvec,
6248 .procname = "temp_valid_lft",
6249 .data = &ipv6_devconf.temp_valid_lft,
6250 .maxlen = sizeof(int),
6252 .proc_handler = proc_dointvec,
6255 .procname = "temp_prefered_lft",
6256 .data = &ipv6_devconf.temp_prefered_lft,
6257 .maxlen = sizeof(int),
6259 .proc_handler = proc_dointvec,
6262 .procname = "regen_max_retry",
6263 .data = &ipv6_devconf.regen_max_retry,
6264 .maxlen = sizeof(int),
6266 .proc_handler = proc_dointvec,
6269 .procname = "max_desync_factor",
6270 .data = &ipv6_devconf.max_desync_factor,
6271 .maxlen = sizeof(int),
6273 .proc_handler = proc_dointvec,
6276 .procname = "max_addresses",
6277 .data = &ipv6_devconf.max_addresses,
6278 .maxlen = sizeof(int),
6280 .proc_handler = proc_dointvec,
6283 .procname = "accept_ra_defrtr",
6284 .data = &ipv6_devconf.accept_ra_defrtr,
6285 .maxlen = sizeof(int),
6287 .proc_handler = proc_dointvec,
6290 .procname = "accept_ra_min_hop_limit",
6291 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6292 .maxlen = sizeof(int),
6294 .proc_handler = proc_dointvec,
6297 .procname = "accept_ra_pinfo",
6298 .data = &ipv6_devconf.accept_ra_pinfo,
6299 .maxlen = sizeof(int),
6301 .proc_handler = proc_dointvec,
6303 #ifdef CONFIG_IPV6_ROUTER_PREF
6305 .procname = "accept_ra_rtr_pref",
6306 .data = &ipv6_devconf.accept_ra_rtr_pref,
6307 .maxlen = sizeof(int),
6309 .proc_handler = proc_dointvec,
6312 .procname = "router_probe_interval",
6313 .data = &ipv6_devconf.rtr_probe_interval,
6314 .maxlen = sizeof(int),
6316 .proc_handler = proc_dointvec_jiffies,
6318 #ifdef CONFIG_IPV6_ROUTE_INFO
6320 .procname = "accept_ra_rt_info_min_plen",
6321 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6322 .maxlen = sizeof(int),
6324 .proc_handler = proc_dointvec,
6327 .procname = "accept_ra_rt_info_max_plen",
6328 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6329 .maxlen = sizeof(int),
6331 .proc_handler = proc_dointvec,
6336 .procname = "proxy_ndp",
6337 .data = &ipv6_devconf.proxy_ndp,
6338 .maxlen = sizeof(int),
6340 .proc_handler = addrconf_sysctl_proxy_ndp,
6343 .procname = "accept_source_route",
6344 .data = &ipv6_devconf.accept_source_route,
6345 .maxlen = sizeof(int),
6347 .proc_handler = proc_dointvec,
6349 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6351 .procname = "optimistic_dad",
6352 .data = &ipv6_devconf.optimistic_dad,
6353 .maxlen = sizeof(int),
6355 .proc_handler = proc_dointvec,
6358 .procname = "use_optimistic",
6359 .data = &ipv6_devconf.use_optimistic,
6360 .maxlen = sizeof(int),
6362 .proc_handler = proc_dointvec,
6365 #ifdef CONFIG_IPV6_MROUTE
6367 .procname = "mc_forwarding",
6368 .data = &ipv6_devconf.mc_forwarding,
6369 .maxlen = sizeof(int),
6371 .proc_handler = proc_dointvec,
6375 .procname = "disable_ipv6",
6376 .data = &ipv6_devconf.disable_ipv6,
6377 .maxlen = sizeof(int),
6379 .proc_handler = addrconf_sysctl_disable,
6382 .procname = "accept_dad",
6383 .data = &ipv6_devconf.accept_dad,
6384 .maxlen = sizeof(int),
6386 .proc_handler = proc_dointvec,
6389 .procname = "force_tllao",
6390 .data = &ipv6_devconf.force_tllao,
6391 .maxlen = sizeof(int),
6393 .proc_handler = proc_dointvec
6396 .procname = "ndisc_notify",
6397 .data = &ipv6_devconf.ndisc_notify,
6398 .maxlen = sizeof(int),
6400 .proc_handler = proc_dointvec
6403 .procname = "suppress_frag_ndisc",
6404 .data = &ipv6_devconf.suppress_frag_ndisc,
6405 .maxlen = sizeof(int),
6407 .proc_handler = proc_dointvec
6410 .procname = "accept_ra_from_local",
6411 .data = &ipv6_devconf.accept_ra_from_local,
6412 .maxlen = sizeof(int),
6414 .proc_handler = proc_dointvec,
6417 .procname = "accept_ra_mtu",
6418 .data = &ipv6_devconf.accept_ra_mtu,
6419 .maxlen = sizeof(int),
6421 .proc_handler = proc_dointvec,
6424 .procname = "stable_secret",
6425 .data = &ipv6_devconf.stable_secret,
6426 .maxlen = IPV6_MAX_STRLEN,
6428 .proc_handler = addrconf_sysctl_stable_secret,
6431 .procname = "use_oif_addrs_only",
6432 .data = &ipv6_devconf.use_oif_addrs_only,
6433 .maxlen = sizeof(int),
6435 .proc_handler = proc_dointvec,
6438 .procname = "ignore_routes_with_linkdown",
6439 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6440 .maxlen = sizeof(int),
6442 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6445 .procname = "drop_unicast_in_l2_multicast",
6446 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6447 .maxlen = sizeof(int),
6449 .proc_handler = proc_dointvec,
6452 .procname = "drop_unsolicited_na",
6453 .data = &ipv6_devconf.drop_unsolicited_na,
6454 .maxlen = sizeof(int),
6456 .proc_handler = proc_dointvec,
6459 .procname = "keep_addr_on_down",
6460 .data = &ipv6_devconf.keep_addr_on_down,
6461 .maxlen = sizeof(int),
6463 .proc_handler = proc_dointvec,
6467 .procname = "seg6_enabled",
6468 .data = &ipv6_devconf.seg6_enabled,
6469 .maxlen = sizeof(int),
6471 .proc_handler = proc_dointvec,
6473 #ifdef CONFIG_IPV6_SEG6_HMAC
6475 .procname = "seg6_require_hmac",
6476 .data = &ipv6_devconf.seg6_require_hmac,
6477 .maxlen = sizeof(int),
6479 .proc_handler = proc_dointvec,
6483 .procname = "enhanced_dad",
6484 .data = &ipv6_devconf.enhanced_dad,
6485 .maxlen = sizeof(int),
6487 .proc_handler = proc_dointvec,
6490 .procname = "addr_gen_mode",
6491 .data = &ipv6_devconf.addr_gen_mode,
6492 .maxlen = sizeof(int),
6494 .proc_handler = addrconf_sysctl_addr_gen_mode,
6497 .procname = "disable_policy",
6498 .data = &ipv6_devconf.disable_policy,
6499 .maxlen = sizeof(int),
6501 .proc_handler = addrconf_sysctl_disable_policy,
6504 .procname = "ndisc_tclass",
6505 .data = &ipv6_devconf.ndisc_tclass,
6506 .maxlen = sizeof(int),
6508 .proc_handler = proc_dointvec_minmax,
6509 .extra1 = (void *)&zero,
6510 .extra2 = (void *)&two_five_five,
6517 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6518 struct inet6_dev *idev, struct ipv6_devconf *p)
6521 struct ctl_table *table;
6522 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6524 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6528 for (i = 0; table[i].data; i++) {
6529 table[i].data += (char *)p - (char *)&ipv6_devconf;
6530 /* If one of these is already set, then it is not safe to
6531 * overwrite either of them: this makes proc_dointvec_minmax
6534 if (!table[i].extra1 && !table[i].extra2) {
6535 table[i].extra1 = idev; /* embedded; no ref */
6536 table[i].extra2 = net;
6540 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6542 p->sysctl_header = register_net_sysctl(net, path, table);
6543 if (!p->sysctl_header)
6546 if (!strcmp(dev_name, "all"))
6547 ifindex = NETCONFA_IFINDEX_ALL;
6548 else if (!strcmp(dev_name, "default"))
6549 ifindex = NETCONFA_IFINDEX_DEFAULT;
6551 ifindex = idev->dev->ifindex;
6552 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6562 static void __addrconf_sysctl_unregister(struct net *net,
6563 struct ipv6_devconf *p, int ifindex)
6565 struct ctl_table *table;
6567 if (!p->sysctl_header)
6570 table = p->sysctl_header->ctl_table_arg;
6571 unregister_net_sysctl_table(p->sysctl_header);
6572 p->sysctl_header = NULL;
6575 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6578 static int addrconf_sysctl_register(struct inet6_dev *idev)
6582 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6585 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6586 &ndisc_ifinfo_sysctl_change);
6589 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6592 neigh_sysctl_unregister(idev->nd_parms);
6597 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6599 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6600 idev->dev->ifindex);
6601 neigh_sysctl_unregister(idev->nd_parms);
6607 static int __net_init addrconf_init_net(struct net *net)
6610 struct ipv6_devconf *all, *dflt;
6612 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6616 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6618 goto err_alloc_dflt;
6620 /* these will be inherited by all namespaces */
6621 dflt->autoconf = ipv6_defaults.autoconf;
6622 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6624 dflt->stable_secret.initialized = false;
6625 all->stable_secret.initialized = false;
6627 net->ipv6.devconf_all = all;
6628 net->ipv6.devconf_dflt = dflt;
6630 #ifdef CONFIG_SYSCTL
6631 err = __addrconf_sysctl_register(net, "all", NULL, all);
6635 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6641 #ifdef CONFIG_SYSCTL
6643 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6653 static void __net_exit addrconf_exit_net(struct net *net)
6655 #ifdef CONFIG_SYSCTL
6656 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6657 NETCONFA_IFINDEX_DEFAULT);
6658 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6659 NETCONFA_IFINDEX_ALL);
6661 kfree(net->ipv6.devconf_dflt);
6662 kfree(net->ipv6.devconf_all);
6665 static struct pernet_operations addrconf_ops = {
6666 .init = addrconf_init_net,
6667 .exit = addrconf_exit_net,
6670 static struct rtnl_af_ops inet6_ops __read_mostly = {
6672 .fill_link_af = inet6_fill_link_af,
6673 .get_link_af_size = inet6_get_link_af_size,
6674 .validate_link_af = inet6_validate_link_af,
6675 .set_link_af = inet6_set_link_af,
6679 * Init / cleanup code
6682 int __init addrconf_init(void)
6684 struct inet6_dev *idev;
6687 err = ipv6_addr_label_init();
6689 pr_crit("%s: cannot initialize default policy table: %d\n",
6694 err = register_pernet_subsys(&addrconf_ops);
6698 addrconf_wq = create_workqueue("ipv6_addrconf");
6704 /* The addrconf netdev notifier requires that loopback_dev
6705 * has it's ipv6 private information allocated and setup
6706 * before it can bring up and give link-local addresses
6707 * to other devices which are up.
6709 * Unfortunately, loopback_dev is not necessarily the first
6710 * entry in the global dev_base list of net devices. In fact,
6711 * it is likely to be the very last entry on that list.
6712 * So this causes the notifier registry below to try and
6713 * give link-local addresses to all devices besides loopback_dev
6714 * first, then loopback_dev, which cases all the non-loopback_dev
6715 * devices to fail to get a link-local address.
6717 * So, as a temporary fix, allocate the ipv6 structure for
6718 * loopback_dev first by hand.
6719 * Longer term, all of the dependencies ipv6 has upon the loopback
6720 * device and it being up should be removed.
6723 idev = ipv6_add_dev(init_net.loopback_dev);
6726 err = PTR_ERR(idev);
6730 ip6_route_init_special_entries();
6732 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6733 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6735 register_netdevice_notifier(&ipv6_dev_notf);
6739 rtnl_af_register(&inet6_ops);
6741 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6742 NULL, inet6_dump_ifinfo, 0);
6746 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6747 inet6_rtm_newaddr, NULL, 0);
6750 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6751 inet6_rtm_deladdr, NULL, 0);
6754 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6755 inet6_rtm_getaddr, inet6_dump_ifaddr,
6756 RTNL_FLAG_DOIT_UNLOCKED);
6759 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6760 NULL, inet6_dump_ifmcaddr, 0);
6763 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6764 NULL, inet6_dump_ifacaddr, 0);
6767 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6768 inet6_netconf_get_devconf,
6769 inet6_netconf_dump_devconf,
6770 RTNL_FLAG_DOIT_UNLOCKED);
6773 err = ipv6_addr_label_rtnl_register();
6779 rtnl_unregister_all(PF_INET6);
6780 rtnl_af_unregister(&inet6_ops);
6781 unregister_netdevice_notifier(&ipv6_dev_notf);
6783 destroy_workqueue(addrconf_wq);
6785 unregister_pernet_subsys(&addrconf_ops);
6787 ipv6_addr_label_cleanup();
6792 void addrconf_cleanup(void)
6794 struct net_device *dev;
6797 unregister_netdevice_notifier(&ipv6_dev_notf);
6798 unregister_pernet_subsys(&addrconf_ops);
6799 ipv6_addr_label_cleanup();
6801 rtnl_af_unregister(&inet6_ops);
6805 /* clean dev list */
6806 for_each_netdev(&init_net, dev) {
6807 if (__in6_dev_get(dev) == NULL)
6809 addrconf_ifdown(dev, 1);
6811 addrconf_ifdown(init_net.loopback_dev, 2);
6816 spin_lock_bh(&addrconf_hash_lock);
6817 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6818 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6819 spin_unlock_bh(&addrconf_hash_lock);
6820 cancel_delayed_work(&addr_chk_work);
6823 destroy_workqueue(addrconf_wq);