2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
62 #include <linux/sysctl.h>
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
70 #include <net/net_namespace.h>
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
97 /* Set to 3 to get tracing... */
101 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
103 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
106 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108 #define IPV6_MAX_STRLEN \
109 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
111 static inline u32 cstamp_delta(unsigned long cstamp)
113 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
116 static inline s32 rfc3315_s14_backoff_init(s32 irt)
118 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
119 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
120 do_div(tmp, 1000000);
124 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
126 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
127 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
128 do_div(tmp, 1000000);
129 if ((s32)tmp > mrt) {
130 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
131 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
132 do_div(tmp, 1000000);
138 static int addrconf_sysctl_register(struct inet6_dev *idev);
139 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
141 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
146 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
151 static void ipv6_regen_rndid(struct inet6_dev *idev);
152 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
154 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
155 static int ipv6_count_addresses(const struct inet6_dev *idev);
156 static int ipv6_generate_stable_address(struct in6_addr *addr,
158 const struct inet6_dev *idev);
160 #define IN6_ADDR_HSIZE_SHIFT 8
161 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
163 * Configured unicast address hash table
165 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
166 static DEFINE_SPINLOCK(addrconf_hash_lock);
168 static void addrconf_verify(void);
169 static void addrconf_verify_rtnl(void);
170 static void addrconf_verify_work(struct work_struct *);
172 static struct workqueue_struct *addrconf_wq;
173 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
175 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
176 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
178 static void addrconf_type_change(struct net_device *dev,
179 unsigned long event);
180 static int addrconf_ifdown(struct net_device *dev, int how);
182 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
184 const struct net_device *dev,
185 u32 flags, u32 noflags);
187 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
188 static void addrconf_dad_work(struct work_struct *w);
189 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
191 static void addrconf_dad_run(struct inet6_dev *idev);
192 static void addrconf_rs_timer(struct timer_list *t);
193 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
194 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
196 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
197 struct prefix_info *pinfo);
199 static struct ipv6_devconf ipv6_devconf __read_mostly = {
201 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
202 .mtu6 = IPV6_MIN_MTU,
204 .accept_redirects = 1,
206 .force_mld_version = 0,
207 .mldv1_unsolicited_report_interval = 10 * HZ,
208 .mldv2_unsolicited_report_interval = HZ,
210 .rtr_solicits = MAX_RTR_SOLICITATIONS,
211 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
212 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
213 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
215 .temp_valid_lft = TEMP_VALID_LIFETIME,
216 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
217 .regen_max_retry = REGEN_MAX_RETRY,
218 .max_desync_factor = MAX_DESYNC_FACTOR,
219 .max_addresses = IPV6_MAX_ADDRESSES,
220 .accept_ra_defrtr = 1,
221 .accept_ra_from_local = 0,
222 .accept_ra_min_hop_limit= 1,
223 .accept_ra_pinfo = 1,
224 #ifdef CONFIG_IPV6_ROUTER_PREF
225 .accept_ra_rtr_pref = 1,
226 .rtr_probe_interval = 60 * HZ,
227 #ifdef CONFIG_IPV6_ROUTE_INFO
228 .accept_ra_rt_info_min_plen = 0,
229 .accept_ra_rt_info_max_plen = 0,
233 .accept_source_route = 0, /* we do not accept RH0 by default. */
236 .suppress_frag_ndisc = 1,
239 .initialized = false,
241 .use_oif_addrs_only = 0,
242 .ignore_routes_with_linkdown = 0,
243 .keep_addr_on_down = 0,
245 #ifdef CONFIG_IPV6_SEG6_HMAC
246 .seg6_require_hmac = 0,
249 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
253 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
255 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
256 .mtu6 = IPV6_MIN_MTU,
258 .accept_redirects = 1,
260 .force_mld_version = 0,
261 .mldv1_unsolicited_report_interval = 10 * HZ,
262 .mldv2_unsolicited_report_interval = HZ,
264 .rtr_solicits = MAX_RTR_SOLICITATIONS,
265 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
266 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
267 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
269 .temp_valid_lft = TEMP_VALID_LIFETIME,
270 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
271 .regen_max_retry = REGEN_MAX_RETRY,
272 .max_desync_factor = MAX_DESYNC_FACTOR,
273 .max_addresses = IPV6_MAX_ADDRESSES,
274 .accept_ra_defrtr = 1,
275 .accept_ra_from_local = 0,
276 .accept_ra_min_hop_limit= 1,
277 .accept_ra_pinfo = 1,
278 #ifdef CONFIG_IPV6_ROUTER_PREF
279 .accept_ra_rtr_pref = 1,
280 .rtr_probe_interval = 60 * HZ,
281 #ifdef CONFIG_IPV6_ROUTE_INFO
282 .accept_ra_rt_info_min_plen = 0,
283 .accept_ra_rt_info_max_plen = 0,
287 .accept_source_route = 0, /* we do not accept RH0 by default. */
290 .suppress_frag_ndisc = 1,
293 .initialized = false,
295 .use_oif_addrs_only = 0,
296 .ignore_routes_with_linkdown = 0,
297 .keep_addr_on_down = 0,
299 #ifdef CONFIG_IPV6_SEG6_HMAC
300 .seg6_require_hmac = 0,
303 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
307 /* Check if link is ready: is it up and is a valid qdisc available */
308 static inline bool addrconf_link_ready(const struct net_device *dev)
310 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
315 if (del_timer(&idev->rs_timer))
319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
321 if (cancel_delayed_work(&ifp->dad_work))
325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
328 if (!timer_pending(&idev->rs_timer))
330 mod_timer(&idev->rs_timer, jiffies + when);
333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
337 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
341 static int snmp6_alloc_dev(struct inet6_dev *idev)
345 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
346 if (!idev->stats.ipv6)
349 for_each_possible_cpu(i) {
350 struct ipstats_mib *addrconf_stats;
351 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
352 u64_stats_init(&addrconf_stats->syncp);
356 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
358 if (!idev->stats.icmpv6dev)
360 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
362 if (!idev->stats.icmpv6msgdev)
368 kfree(idev->stats.icmpv6dev);
370 free_percpu(idev->stats.ipv6);
375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
377 struct inet6_dev *ndev;
382 if (dev->mtu < IPV6_MIN_MTU)
383 return ERR_PTR(-EINVAL);
385 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
389 rwlock_init(&ndev->lock);
391 INIT_LIST_HEAD(&ndev->addr_list);
392 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
393 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
395 if (ndev->cnf.stable_secret.initialized)
396 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
398 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
400 ndev->cnf.mtu6 = dev->mtu;
401 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
402 if (!ndev->nd_parms) {
406 if (ndev->cnf.forwarding)
407 dev_disable_lro(dev);
408 /* We refer to the device */
411 if (snmp6_alloc_dev(ndev) < 0) {
413 "%s: cannot allocate memory for statistics; dev=%s.\n",
414 __func__, dev->name);
415 neigh_parms_release(&nd_tbl, ndev->nd_parms);
421 if (snmp6_register_dev(ndev) < 0) {
423 "%s: cannot create /proc/net/dev_snmp6/%s\n",
424 __func__, dev->name);
428 /* One reference from device. */
429 refcount_set(&ndev->refcnt, 1);
431 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
432 ndev->cnf.accept_dad = -1;
434 #if IS_ENABLED(CONFIG_IPV6_SIT)
435 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
436 pr_info("%s: Disabled Multicast RS\n", dev->name);
437 ndev->cnf.rtr_solicits = 0;
441 INIT_LIST_HEAD(&ndev->tempaddr_list);
442 ndev->desync_factor = U32_MAX;
443 if ((dev->flags&IFF_LOOPBACK) ||
444 dev->type == ARPHRD_TUNNEL ||
445 dev->type == ARPHRD_TUNNEL6 ||
446 dev->type == ARPHRD_SIT ||
447 dev->type == ARPHRD_NONE) {
448 ndev->cnf.use_tempaddr = -1;
450 ipv6_regen_rndid(ndev);
452 ndev->token = in6addr_any;
454 if (netif_running(dev) && addrconf_link_ready(dev))
455 ndev->if_flags |= IF_READY;
457 ipv6_mc_init_dev(ndev);
458 ndev->tstamp = jiffies;
459 err = addrconf_sysctl_register(ndev);
461 ipv6_mc_destroy_dev(ndev);
462 snmp6_unregister_dev(ndev);
465 /* protected by rtnl_lock */
466 rcu_assign_pointer(dev->ip6_ptr, ndev);
468 /* Join interface-local all-node multicast group */
469 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
471 /* Join all-node multicast group */
472 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
474 /* Join all-router multicast group if forwarding is set */
475 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
476 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
481 neigh_parms_release(&nd_tbl, ndev->nd_parms);
483 in6_dev_finish_destroy(ndev);
487 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
489 struct inet6_dev *idev;
493 idev = __in6_dev_get(dev);
495 idev = ipv6_add_dev(dev);
500 if (dev->flags&IFF_UP)
505 static int inet6_netconf_msgsize_devconf(int type)
507 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
508 + nla_total_size(4); /* NETCONFA_IFINDEX */
511 if (type == NETCONFA_ALL)
514 if (all || type == NETCONFA_FORWARDING)
515 size += nla_total_size(4);
516 #ifdef CONFIG_IPV6_MROUTE
517 if (all || type == NETCONFA_MC_FORWARDING)
518 size += nla_total_size(4);
520 if (all || type == NETCONFA_PROXY_NEIGH)
521 size += nla_total_size(4);
523 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
524 size += nla_total_size(4);
529 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
530 struct ipv6_devconf *devconf, u32 portid,
531 u32 seq, int event, unsigned int flags,
534 struct nlmsghdr *nlh;
535 struct netconfmsg *ncm;
538 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
543 if (type == NETCONFA_ALL)
546 ncm = nlmsg_data(nlh);
547 ncm->ncm_family = AF_INET6;
549 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
550 goto nla_put_failure;
555 if ((all || type == NETCONFA_FORWARDING) &&
556 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
557 goto nla_put_failure;
558 #ifdef CONFIG_IPV6_MROUTE
559 if ((all || type == NETCONFA_MC_FORWARDING) &&
560 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
561 devconf->mc_forwarding) < 0)
562 goto nla_put_failure;
564 if ((all || type == NETCONFA_PROXY_NEIGH) &&
565 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
566 goto nla_put_failure;
568 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
569 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
570 devconf->ignore_routes_with_linkdown) < 0)
571 goto nla_put_failure;
578 nlmsg_cancel(skb, nlh);
582 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
583 int ifindex, struct ipv6_devconf *devconf)
588 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
592 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
595 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
596 WARN_ON(err == -EMSGSIZE);
600 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
603 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
606 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
607 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
608 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
609 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
610 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
613 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
614 struct nlmsghdr *nlh,
615 struct netlink_ext_ack *extack)
617 struct net *net = sock_net(in_skb->sk);
618 struct nlattr *tb[NETCONFA_MAX+1];
619 struct inet6_dev *in6_dev = NULL;
620 struct net_device *dev = NULL;
621 struct netconfmsg *ncm;
623 struct ipv6_devconf *devconf;
627 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
628 devconf_ipv6_policy, extack);
632 if (!tb[NETCONFA_IFINDEX])
636 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
638 case NETCONFA_IFINDEX_ALL:
639 devconf = net->ipv6.devconf_all;
641 case NETCONFA_IFINDEX_DEFAULT:
642 devconf = net->ipv6.devconf_dflt;
645 dev = dev_get_by_index(net, ifindex);
648 in6_dev = in6_dev_get(dev);
651 devconf = &in6_dev->cnf;
656 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
660 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
661 NETLINK_CB(in_skb).portid,
662 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
665 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
666 WARN_ON(err == -EMSGSIZE);
670 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
673 in6_dev_put(in6_dev);
679 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
680 struct netlink_callback *cb)
682 struct net *net = sock_net(skb->sk);
685 struct net_device *dev;
686 struct inet6_dev *idev;
687 struct hlist_head *head;
690 s_idx = idx = cb->args[1];
692 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
694 head = &net->dev_index_head[h];
696 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
698 hlist_for_each_entry_rcu(dev, head, index_hlist) {
701 idev = __in6_dev_get(dev);
705 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
707 NETLINK_CB(cb->skb).portid,
715 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
721 if (h == NETDEV_HASHENTRIES) {
722 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
723 net->ipv6.devconf_all,
724 NETLINK_CB(cb->skb).portid,
726 RTM_NEWNETCONF, NLM_F_MULTI,
732 if (h == NETDEV_HASHENTRIES + 1) {
733 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
734 net->ipv6.devconf_dflt,
735 NETLINK_CB(cb->skb).portid,
737 RTM_NEWNETCONF, NLM_F_MULTI,
751 static void dev_forward_change(struct inet6_dev *idev)
753 struct net_device *dev;
754 struct inet6_ifaddr *ifa;
759 if (idev->cnf.forwarding)
760 dev_disable_lro(dev);
761 if (dev->flags & IFF_MULTICAST) {
762 if (idev->cnf.forwarding) {
763 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
764 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
765 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
767 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
768 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
769 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
773 list_for_each_entry(ifa, &idev->addr_list, if_list) {
774 if (ifa->flags&IFA_F_TENTATIVE)
776 if (idev->cnf.forwarding)
777 addrconf_join_anycast(ifa);
779 addrconf_leave_anycast(ifa);
781 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
783 dev->ifindex, &idev->cnf);
787 static void addrconf_forward_change(struct net *net, __s32 newf)
789 struct net_device *dev;
790 struct inet6_dev *idev;
792 for_each_netdev(net, dev) {
793 idev = __in6_dev_get(dev);
795 int changed = (!idev->cnf.forwarding) ^ (!newf);
796 idev->cnf.forwarding = newf;
798 dev_forward_change(idev);
803 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
809 return restart_syscall();
811 net = (struct net *)table->extra2;
815 if (p == &net->ipv6.devconf_dflt->forwarding) {
816 if ((!newf) ^ (!old))
817 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
819 NETCONFA_IFINDEX_DEFAULT,
820 net->ipv6.devconf_dflt);
825 if (p == &net->ipv6.devconf_all->forwarding) {
826 int old_dflt = net->ipv6.devconf_dflt->forwarding;
828 net->ipv6.devconf_dflt->forwarding = newf;
829 if ((!newf) ^ (!old_dflt))
830 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
832 NETCONFA_IFINDEX_DEFAULT,
833 net->ipv6.devconf_dflt);
835 addrconf_forward_change(net, newf);
836 if ((!newf) ^ (!old))
837 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
839 NETCONFA_IFINDEX_ALL,
840 net->ipv6.devconf_all);
841 } else if ((!newf) ^ (!old))
842 dev_forward_change((struct inet6_dev *)table->extra1);
846 rt6_purge_dflt_routers(net);
850 static void addrconf_linkdown_change(struct net *net, __s32 newf)
852 struct net_device *dev;
853 struct inet6_dev *idev;
855 for_each_netdev(net, dev) {
856 idev = __in6_dev_get(dev);
858 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
860 idev->cnf.ignore_routes_with_linkdown = newf;
862 inet6_netconf_notify_devconf(dev_net(dev),
864 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
871 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
877 return restart_syscall();
879 net = (struct net *)table->extra2;
883 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
884 if ((!newf) ^ (!old))
885 inet6_netconf_notify_devconf(net,
887 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
888 NETCONFA_IFINDEX_DEFAULT,
889 net->ipv6.devconf_dflt);
894 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
895 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
896 addrconf_linkdown_change(net, newf);
897 if ((!newf) ^ (!old))
898 inet6_netconf_notify_devconf(net,
900 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
901 NETCONFA_IFINDEX_ALL,
902 net->ipv6.devconf_all);
911 /* Nobody refers to this ifaddr, destroy it */
912 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
914 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
916 #ifdef NET_REFCNT_DEBUG
917 pr_debug("%s\n", __func__);
920 in6_dev_put(ifp->idev);
922 if (cancel_delayed_work(&ifp->dad_work))
923 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
926 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
927 pr_warn("Freeing alive inet6 address %p\n", ifp);
936 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
939 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
942 * Each device address list is sorted in order of scope -
943 * global before linklocal.
945 list_for_each(p, &idev->addr_list) {
946 struct inet6_ifaddr *ifa
947 = list_entry(p, struct inet6_ifaddr, if_list);
948 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
952 list_add_tail_rcu(&ifp->if_list, p);
955 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
957 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
959 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
962 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
963 struct net_device *dev, unsigned int hash)
965 struct inet6_ifaddr *ifp;
967 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
968 if (!net_eq(dev_net(ifp->idev->dev), net))
970 if (ipv6_addr_equal(&ifp->addr, addr)) {
971 if (!dev || ifp->idev->dev == dev)
978 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
980 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
983 spin_lock(&addrconf_hash_lock);
985 /* Ignore adding duplicate addresses on an interface */
986 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
987 ADBG("ipv6_add_addr: already assigned\n");
990 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
993 spin_unlock(&addrconf_hash_lock);
998 /* On success it returns ifp with increased reference count */
1000 static struct inet6_ifaddr *
1001 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
1002 const struct in6_addr *peer_addr, int pfxlen,
1003 int scope, u32 flags, u32 valid_lft, u32 prefered_lft,
1004 bool can_block, struct netlink_ext_ack *extack)
1006 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1007 struct net *net = dev_net(idev->dev);
1008 struct inet6_ifaddr *ifa = NULL;
1009 struct rt6_info *rt = NULL;
1011 int addr_type = ipv6_addr_type(addr);
1013 if (addr_type == IPV6_ADDR_ANY ||
1014 addr_type & IPV6_ADDR_MULTICAST ||
1015 (!(idev->dev->flags & IFF_LOOPBACK) &&
1016 addr_type & IPV6_ADDR_LOOPBACK))
1017 return ERR_PTR(-EADDRNOTAVAIL);
1020 err = -ENODEV; /*XXX*/
1024 if (idev->cnf.disable_ipv6) {
1029 /* validator notifier needs to be blocking;
1030 * do not call in atomic context
1033 struct in6_validator_info i6vi = {
1039 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1040 err = notifier_to_errno(err);
1045 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1047 ADBG("ipv6_add_addr: malloc failed\n");
1052 rt = addrconf_dst_alloc(idev, addr, false);
1059 if (net->ipv6.devconf_all->disable_policy ||
1060 idev->cnf.disable_policy)
1061 rt->dst.flags |= DST_NOPOLICY;
1063 neigh_parms_data_state_setall(idev->nd_parms);
1067 ifa->peer_addr = *peer_addr;
1069 spin_lock_init(&ifa->lock);
1070 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1071 INIT_HLIST_NODE(&ifa->addr_lst);
1073 ifa->prefix_len = pfxlen;
1075 /* No need to add the TENTATIVE flag for addresses with NODAD */
1076 if (!(flags & IFA_F_NODAD))
1077 ifa->flags |= IFA_F_TENTATIVE;
1078 ifa->valid_lft = valid_lft;
1079 ifa->prefered_lft = prefered_lft;
1080 ifa->cstamp = ifa->tstamp = jiffies;
1081 ifa->tokenized = false;
1089 refcount_set(&ifa->refcnt, 1);
1093 err = ipv6_add_addr_hash(idev->dev, ifa);
1095 rcu_read_unlock_bh();
1099 write_lock(&idev->lock);
1101 /* Add to inet6_dev unicast addr list. */
1102 ipv6_link_dev_addr(idev, ifa);
1104 if (ifa->flags&IFA_F_TEMPORARY) {
1105 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1110 write_unlock(&idev->lock);
1112 rcu_read_unlock_bh();
1114 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1116 if (unlikely(err < 0)) {
1121 in6_dev_put(ifa->idev);
1130 enum cleanup_prefix_rt_t {
1131 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1132 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1133 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1137 * Check, whether the prefix for ifp would still need a prefix route
1138 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1141 * 1) we don't purge prefix if address was not permanent.
1142 * prefix is managed by its own lifetime.
1143 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1144 * 3) if there are no addresses, delete prefix.
1145 * 4) if there are still other permanent address(es),
1146 * corresponding prefix is still permanent.
1147 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1148 * don't purge the prefix, assume user space is managing it.
1149 * 6) otherwise, update prefix lifetime to the
1150 * longest valid lifetime among the corresponding
1151 * addresses on the device.
1152 * Note: subsequent RA will update lifetime.
1154 static enum cleanup_prefix_rt_t
1155 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1157 struct inet6_ifaddr *ifa;
1158 struct inet6_dev *idev = ifp->idev;
1159 unsigned long lifetime;
1160 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1164 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1167 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1170 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1171 return CLEANUP_PREFIX_RT_NOP;
1173 action = CLEANUP_PREFIX_RT_EXPIRE;
1175 spin_lock(&ifa->lock);
1177 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1179 * Note: Because this address is
1180 * not permanent, lifetime <
1181 * LONG_MAX / HZ here.
1183 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1184 *expires = ifa->tstamp + lifetime * HZ;
1185 spin_unlock(&ifa->lock);
1192 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1194 struct rt6_info *rt;
1196 rt = addrconf_get_prefix_route(&ifp->addr,
1199 0, RTF_GATEWAY | RTF_DEFAULT);
1204 if (!(rt->rt6i_flags & RTF_EXPIRES))
1205 rt6_set_expires(rt, expires);
1212 /* This function wants to get referenced ifp and releases it before return */
1214 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1217 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1218 unsigned long expires;
1222 spin_lock_bh(&ifp->lock);
1224 ifp->state = INET6_IFADDR_STATE_DEAD;
1225 spin_unlock_bh(&ifp->lock);
1227 if (state == INET6_IFADDR_STATE_DEAD)
1230 spin_lock_bh(&addrconf_hash_lock);
1231 hlist_del_init_rcu(&ifp->addr_lst);
1232 spin_unlock_bh(&addrconf_hash_lock);
1234 write_lock_bh(&ifp->idev->lock);
1236 if (ifp->flags&IFA_F_TEMPORARY) {
1237 list_del(&ifp->tmp_list);
1239 in6_ifa_put(ifp->ifpub);
1245 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1246 action = check_cleanup_prefix_route(ifp, &expires);
1248 list_del_rcu(&ifp->if_list);
1251 write_unlock_bh(&ifp->idev->lock);
1253 addrconf_del_dad_work(ifp);
1255 ipv6_ifa_notify(RTM_DELADDR, ifp);
1257 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1259 if (action != CLEANUP_PREFIX_RT_NOP) {
1260 cleanup_prefix_route(ifp, expires,
1261 action == CLEANUP_PREFIX_RT_DEL);
1264 /* clean up prefsrc entries */
1265 rt6_remove_prefsrc(ifp);
1270 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1271 struct inet6_ifaddr *ift,
1274 struct inet6_dev *idev = ifp->idev;
1275 struct in6_addr addr, *tmpaddr;
1276 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1277 unsigned long regen_advance;
1281 unsigned long now = jiffies;
1282 long max_desync_factor;
1283 s32 cnf_temp_preferred_lft;
1285 write_lock_bh(&idev->lock);
1287 spin_lock_bh(&ift->lock);
1288 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1289 spin_unlock_bh(&ift->lock);
1296 if (idev->cnf.use_tempaddr <= 0) {
1297 write_unlock_bh(&idev->lock);
1298 pr_info("%s: use_tempaddr is disabled\n", __func__);
1303 spin_lock_bh(&ifp->lock);
1304 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1305 idev->cnf.use_tempaddr = -1; /*XXX*/
1306 spin_unlock_bh(&ifp->lock);
1307 write_unlock_bh(&idev->lock);
1308 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1315 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1316 ipv6_try_regen_rndid(idev, tmpaddr);
1317 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1318 age = (now - ifp->tstamp) / HZ;
1320 regen_advance = idev->cnf.regen_max_retry *
1321 idev->cnf.dad_transmits *
1322 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1324 /* recalculate max_desync_factor each time and update
1325 * idev->desync_factor if it's larger
1327 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1328 max_desync_factor = min_t(__u32,
1329 idev->cnf.max_desync_factor,
1330 cnf_temp_preferred_lft - regen_advance);
1332 if (unlikely(idev->desync_factor > max_desync_factor)) {
1333 if (max_desync_factor > 0) {
1334 get_random_bytes(&idev->desync_factor,
1335 sizeof(idev->desync_factor));
1336 idev->desync_factor %= max_desync_factor;
1338 idev->desync_factor = 0;
1342 tmp_valid_lft = min_t(__u32,
1344 idev->cnf.temp_valid_lft + age);
1345 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1346 idev->desync_factor;
1347 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1348 tmp_plen = ifp->prefix_len;
1349 tmp_tstamp = ifp->tstamp;
1350 spin_unlock_bh(&ifp->lock);
1352 write_unlock_bh(&idev->lock);
1354 /* A temporary address is created only if this calculated Preferred
1355 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1356 * an implementation must not create a temporary address with a zero
1357 * Preferred Lifetime.
1358 * Use age calculation as in addrconf_verify to avoid unnecessary
1359 * temporary addresses being generated.
1361 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1362 if (tmp_prefered_lft <= regen_advance + age) {
1369 addr_flags = IFA_F_TEMPORARY;
1370 /* set in addrconf_prefix_rcv() */
1371 if (ifp->flags & IFA_F_OPTIMISTIC)
1372 addr_flags |= IFA_F_OPTIMISTIC;
1374 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1375 ipv6_addr_scope(&addr), addr_flags,
1376 tmp_valid_lft, tmp_prefered_lft, block, NULL);
1380 pr_info("%s: retry temporary address regeneration\n", __func__);
1382 write_lock_bh(&idev->lock);
1386 spin_lock_bh(&ift->lock);
1389 ift->tstamp = tmp_tstamp;
1390 spin_unlock_bh(&ift->lock);
1392 addrconf_dad_start(ift);
1400 * Choose an appropriate source address (RFC3484)
1403 IPV6_SADDR_RULE_INIT = 0,
1404 IPV6_SADDR_RULE_LOCAL,
1405 IPV6_SADDR_RULE_SCOPE,
1406 IPV6_SADDR_RULE_PREFERRED,
1407 #ifdef CONFIG_IPV6_MIP6
1408 IPV6_SADDR_RULE_HOA,
1410 IPV6_SADDR_RULE_OIF,
1411 IPV6_SADDR_RULE_LABEL,
1412 IPV6_SADDR_RULE_PRIVACY,
1413 IPV6_SADDR_RULE_ORCHID,
1414 IPV6_SADDR_RULE_PREFIX,
1415 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1416 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1421 struct ipv6_saddr_score {
1424 struct inet6_ifaddr *ifa;
1425 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1430 struct ipv6_saddr_dst {
1431 const struct in6_addr *addr;
1438 static inline int ipv6_saddr_preferred(int type)
1440 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1445 static bool ipv6_use_optimistic_addr(struct net *net,
1446 struct inet6_dev *idev)
1448 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1451 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1453 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1462 static bool ipv6_allow_optimistic_dad(struct net *net,
1463 struct inet6_dev *idev)
1465 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1468 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1477 static int ipv6_get_saddr_eval(struct net *net,
1478 struct ipv6_saddr_score *score,
1479 struct ipv6_saddr_dst *dst,
1484 if (i <= score->rule) {
1486 case IPV6_SADDR_RULE_SCOPE:
1487 ret = score->scopedist;
1489 case IPV6_SADDR_RULE_PREFIX:
1490 ret = score->matchlen;
1493 ret = !!test_bit(i, score->scorebits);
1499 case IPV6_SADDR_RULE_INIT:
1500 /* Rule 0: remember if hiscore is not ready yet */
1503 case IPV6_SADDR_RULE_LOCAL:
1504 /* Rule 1: Prefer same address */
1505 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1507 case IPV6_SADDR_RULE_SCOPE:
1508 /* Rule 2: Prefer appropriate scope
1513 * ---+--+-+---> scope
1515 * | d is scope of the destination.
1517 * | \ <- smaller scope is better if
1518 * B-15 | \ if scope is enough for destination.
1519 * | ret = B - scope (-1 <= scope >= d <= 15).
1521 * |/ <- greater is better
1522 * -C / if scope is not enough for destination.
1523 * /| ret = scope - C (-1 <= d < scope <= 15).
1525 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1526 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1527 * Assume B = 0 and we get C > 29.
1529 ret = __ipv6_addr_src_scope(score->addr_type);
1530 if (ret >= dst->scope)
1533 ret -= 128; /* 30 is enough */
1534 score->scopedist = ret;
1536 case IPV6_SADDR_RULE_PREFERRED:
1538 /* Rule 3: Avoid deprecated and optimistic addresses */
1539 u8 avoid = IFA_F_DEPRECATED;
1541 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1542 avoid |= IFA_F_OPTIMISTIC;
1543 ret = ipv6_saddr_preferred(score->addr_type) ||
1544 !(score->ifa->flags & avoid);
1547 #ifdef CONFIG_IPV6_MIP6
1548 case IPV6_SADDR_RULE_HOA:
1550 /* Rule 4: Prefer home address */
1551 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1552 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1556 case IPV6_SADDR_RULE_OIF:
1557 /* Rule 5: Prefer outgoing interface */
1558 ret = (!dst->ifindex ||
1559 dst->ifindex == score->ifa->idev->dev->ifindex);
1561 case IPV6_SADDR_RULE_LABEL:
1562 /* Rule 6: Prefer matching label */
1563 ret = ipv6_addr_label(net,
1564 &score->ifa->addr, score->addr_type,
1565 score->ifa->idev->dev->ifindex) == dst->label;
1567 case IPV6_SADDR_RULE_PRIVACY:
1569 /* Rule 7: Prefer public address
1570 * Note: prefer temporary address if use_tempaddr >= 2
1572 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1573 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1574 score->ifa->idev->cnf.use_tempaddr >= 2;
1575 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1578 case IPV6_SADDR_RULE_ORCHID:
1579 /* Rule 8-: Prefer ORCHID vs ORCHID or
1580 * non-ORCHID vs non-ORCHID
1582 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1583 ipv6_addr_orchid(dst->addr));
1585 case IPV6_SADDR_RULE_PREFIX:
1586 /* Rule 8: Use longest matching prefix */
1587 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1588 if (ret > score->ifa->prefix_len)
1589 ret = score->ifa->prefix_len;
1590 score->matchlen = ret;
1592 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1593 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1594 /* Optimistic addresses still have lower precedence than other
1595 * preferred addresses.
1597 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1605 __set_bit(i, score->scorebits);
1611 static int __ipv6_dev_get_saddr(struct net *net,
1612 struct ipv6_saddr_dst *dst,
1613 struct inet6_dev *idev,
1614 struct ipv6_saddr_score *scores,
1617 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1619 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1623 * - Tentative Address (RFC2462 section 5.4)
1624 * - A tentative address is not considered
1625 * "assigned to an interface" in the traditional
1626 * sense, unless it is also flagged as optimistic.
1627 * - Candidate Source Address (section 4)
1628 * - In any case, anycast addresses, multicast
1629 * addresses, and the unspecified address MUST
1630 * NOT be included in a candidate set.
1632 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1633 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1636 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1638 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1639 score->addr_type & IPV6_ADDR_MULTICAST)) {
1640 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1646 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1648 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1649 int minihiscore, miniscore;
1651 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1652 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1654 if (minihiscore > miniscore) {
1655 if (i == IPV6_SADDR_RULE_SCOPE &&
1656 score->scopedist > 0) {
1659 * each remaining entry
1660 * has too small (not enough)
1661 * scope, because ifa entries
1662 * are sorted by their scope
1668 } else if (minihiscore < miniscore) {
1669 swap(hiscore, score);
1670 hiscore_idx = 1 - hiscore_idx;
1672 /* restore our iterator */
1673 score->ifa = hiscore->ifa;
1683 static int ipv6_get_saddr_master(struct net *net,
1684 const struct net_device *dst_dev,
1685 const struct net_device *master,
1686 struct ipv6_saddr_dst *dst,
1687 struct ipv6_saddr_score *scores,
1690 struct inet6_dev *idev;
1692 idev = __in6_dev_get(dst_dev);
1694 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1695 scores, hiscore_idx);
1697 idev = __in6_dev_get(master);
1699 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1700 scores, hiscore_idx);
1705 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1706 const struct in6_addr *daddr, unsigned int prefs,
1707 struct in6_addr *saddr)
1709 struct ipv6_saddr_score scores[2], *hiscore;
1710 struct ipv6_saddr_dst dst;
1711 struct inet6_dev *idev;
1712 struct net_device *dev;
1714 bool use_oif_addr = false;
1715 int hiscore_idx = 0;
1718 dst_type = __ipv6_addr_type(daddr);
1720 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1721 dst.scope = __ipv6_addr_src_scope(dst_type);
1722 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1725 scores[hiscore_idx].rule = -1;
1726 scores[hiscore_idx].ifa = NULL;
1730 /* Candidate Source Address (section 4)
1731 * - multicast and link-local destination address,
1732 * the set of candidate source address MUST only
1733 * include addresses assigned to interfaces
1734 * belonging to the same link as the outgoing
1736 * (- For site-local destination addresses, the
1737 * set of candidate source addresses MUST only
1738 * include addresses assigned to interfaces
1739 * belonging to the same site as the outgoing
1741 * - "It is RECOMMENDED that the candidate source addresses
1742 * be the set of unicast addresses assigned to the
1743 * interface that will be used to send to the destination
1744 * (the 'outgoing' interface)." (RFC 6724)
1747 idev = __in6_dev_get(dst_dev);
1748 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1749 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1750 (idev && idev->cnf.use_oif_addrs_only)) {
1751 use_oif_addr = true;
1757 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1759 const struct net_device *master;
1762 /* if dst_dev exists and is enslaved to an L3 device, then
1763 * prefer addresses from dst_dev and then the master over
1764 * any other enslaved devices in the L3 domain.
1766 master = l3mdev_master_dev_rcu(dst_dev);
1768 master_idx = master->ifindex;
1770 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1772 scores, hiscore_idx);
1774 if (scores[hiscore_idx].ifa)
1778 for_each_netdev_rcu(net, dev) {
1779 /* only consider addresses on devices in the
1782 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1784 idev = __in6_dev_get(dev);
1787 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1792 hiscore = &scores[hiscore_idx];
1794 ret = -EADDRNOTAVAIL;
1796 *saddr = hiscore->ifa->addr;
1801 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1803 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1806 struct inet6_ifaddr *ifp;
1807 int err = -EADDRNOTAVAIL;
1809 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1810 if (ifp->scope > IFA_LINK)
1812 if (ifp->scope == IFA_LINK &&
1813 !(ifp->flags & banned_flags)) {
1822 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1825 struct inet6_dev *idev;
1826 int err = -EADDRNOTAVAIL;
1829 idev = __in6_dev_get(dev);
1831 read_lock_bh(&idev->lock);
1832 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1833 read_unlock_bh(&idev->lock);
1839 static int ipv6_count_addresses(const struct inet6_dev *idev)
1841 const struct inet6_ifaddr *ifp;
1845 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1851 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1852 const struct net_device *dev, int strict)
1854 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1855 strict, IFA_F_TENTATIVE);
1857 EXPORT_SYMBOL(ipv6_chk_addr);
1859 /* device argument is used to find the L3 domain of interest. If
1860 * skip_dev_check is set, then the ifp device is not checked against
1861 * the passed in dev argument. So the 2 cases for addresses checks are:
1862 * 1. does the address exist in the L3 domain that dev is part of
1863 * (skip_dev_check = true), or
1865 * 2. does the address exist on the specific device
1866 * (skip_dev_check = false)
1868 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1869 const struct net_device *dev, bool skip_dev_check,
1870 int strict, u32 banned_flags)
1872 unsigned int hash = inet6_addr_hash(net, addr);
1873 const struct net_device *l3mdev;
1874 struct inet6_ifaddr *ifp;
1879 l3mdev = l3mdev_master_dev_rcu(dev);
1883 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1884 if (!net_eq(dev_net(ifp->idev->dev), net))
1887 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1890 /* Decouple optimistic from tentative for evaluation here.
1891 * Ban optimistic addresses explicitly, when required.
1893 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1894 ? (ifp->flags&~IFA_F_TENTATIVE)
1896 if (ipv6_addr_equal(&ifp->addr, addr) &&
1897 !(ifp_flags&banned_flags) &&
1898 (!dev || ifp->idev->dev == dev ||
1899 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1908 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1911 /* Compares an address/prefix_len with addresses on device @dev.
1912 * If one is found it returns true.
1914 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1915 const unsigned int prefix_len, struct net_device *dev)
1917 const struct inet6_ifaddr *ifa;
1918 const struct inet6_dev *idev;
1922 idev = __in6_dev_get(dev);
1924 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1925 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1934 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1936 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1938 const struct inet6_ifaddr *ifa;
1939 const struct inet6_dev *idev;
1944 idev = __in6_dev_get(dev);
1946 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1947 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1956 EXPORT_SYMBOL(ipv6_chk_prefix);
1958 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1959 struct net_device *dev, int strict)
1961 unsigned int hash = inet6_addr_hash(net, addr);
1962 struct inet6_ifaddr *ifp, *result = NULL;
1965 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1966 if (!net_eq(dev_net(ifp->idev->dev), net))
1968 if (ipv6_addr_equal(&ifp->addr, addr)) {
1969 if (!dev || ifp->idev->dev == dev ||
1970 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1982 /* Gets referenced address, destroys ifaddr */
1984 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1987 ifp->flags |= IFA_F_DADFAILED;
1989 if (ifp->flags&IFA_F_TEMPORARY) {
1990 struct inet6_ifaddr *ifpub;
1991 spin_lock_bh(&ifp->lock);
1994 in6_ifa_hold(ifpub);
1995 spin_unlock_bh(&ifp->lock);
1996 ipv6_create_tempaddr(ifpub, ifp, true);
1999 spin_unlock_bh(&ifp->lock);
2002 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2003 spin_lock_bh(&ifp->lock);
2004 addrconf_del_dad_work(ifp);
2005 ifp->flags |= IFA_F_TENTATIVE;
2007 ifp->flags &= ~IFA_F_OPTIMISTIC;
2008 spin_unlock_bh(&ifp->lock);
2010 ipv6_ifa_notify(0, ifp);
2017 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2021 spin_lock_bh(&ifp->lock);
2022 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2023 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2026 spin_unlock_bh(&ifp->lock);
2031 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2033 struct inet6_dev *idev = ifp->idev;
2034 struct net *net = dev_net(ifp->idev->dev);
2036 if (addrconf_dad_end(ifp)) {
2041 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2042 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2044 spin_lock_bh(&ifp->lock);
2046 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2047 int scope = ifp->scope;
2048 u32 flags = ifp->flags;
2049 struct in6_addr new_addr;
2050 struct inet6_ifaddr *ifp2;
2051 u32 valid_lft, preferred_lft;
2052 int pfxlen = ifp->prefix_len;
2053 int retries = ifp->stable_privacy_retry + 1;
2055 if (retries > net->ipv6.sysctl.idgen_retries) {
2056 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2057 ifp->idev->dev->name);
2061 new_addr = ifp->addr;
2062 if (ipv6_generate_stable_address(&new_addr, retries,
2066 valid_lft = ifp->valid_lft;
2067 preferred_lft = ifp->prefered_lft;
2069 spin_unlock_bh(&ifp->lock);
2071 if (idev->cnf.max_addresses &&
2072 ipv6_count_addresses(idev) >=
2073 idev->cnf.max_addresses)
2076 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2077 ifp->idev->dev->name);
2079 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2080 scope, flags, valid_lft,
2081 preferred_lft, false, NULL);
2085 spin_lock_bh(&ifp2->lock);
2086 ifp2->stable_privacy_retry = retries;
2087 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2088 spin_unlock_bh(&ifp2->lock);
2090 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2093 spin_lock_bh(&ifp->lock);
2097 /* transition from _POSTDAD to _ERRDAD */
2098 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2099 spin_unlock_bh(&ifp->lock);
2101 addrconf_mod_dad_work(ifp, 0);
2105 /* Join to solicited addr multicast group.
2106 * caller must hold RTNL */
2107 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2109 struct in6_addr maddr;
2111 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2114 addrconf_addr_solict_mult(addr, &maddr);
2115 ipv6_dev_mc_inc(dev, &maddr);
2118 /* caller must hold RTNL */
2119 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2121 struct in6_addr maddr;
2123 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2126 addrconf_addr_solict_mult(addr, &maddr);
2127 __ipv6_dev_mc_dec(idev, &maddr);
2130 /* caller must hold RTNL */
2131 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2133 struct in6_addr addr;
2135 if (ifp->prefix_len >= 127) /* RFC 6164 */
2137 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2138 if (ipv6_addr_any(&addr))
2140 __ipv6_dev_ac_inc(ifp->idev, &addr);
2143 /* caller must hold RTNL */
2144 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2146 struct in6_addr addr;
2148 if (ifp->prefix_len >= 127) /* RFC 6164 */
2150 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2151 if (ipv6_addr_any(&addr))
2153 __ipv6_dev_ac_dec(ifp->idev, &addr);
2156 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2158 switch (dev->addr_len) {
2160 memcpy(eui, dev->dev_addr, 3);
2163 memcpy(eui + 5, dev->dev_addr + 3, 3);
2165 case EUI64_ADDR_LEN:
2166 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2176 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2178 union fwnet_hwaddr *ha;
2180 if (dev->addr_len != FWNET_ALEN)
2183 ha = (union fwnet_hwaddr *)dev->dev_addr;
2185 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2190 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2192 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2193 if (dev->addr_len != ARCNET_ALEN)
2196 eui[7] = *(u8 *)dev->dev_addr;
2200 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2202 if (dev->addr_len != INFINIBAND_ALEN)
2204 memcpy(eui, dev->dev_addr + 12, 8);
2209 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2213 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2214 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2215 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2216 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2217 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2218 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2222 memcpy(eui + 4, &addr, 4);
2226 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2228 if (dev->priv_flags & IFF_ISATAP)
2229 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2233 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2235 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2238 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2240 memcpy(eui, dev->perm_addr, 3);
2241 memcpy(eui + 5, dev->perm_addr + 3, 3);
2248 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2250 switch (dev->type) {
2253 return addrconf_ifid_eui48(eui, dev);
2255 return addrconf_ifid_arcnet(eui, dev);
2256 case ARPHRD_INFINIBAND:
2257 return addrconf_ifid_infiniband(eui, dev);
2259 return addrconf_ifid_sit(eui, dev);
2262 return addrconf_ifid_gre(eui, dev);
2263 case ARPHRD_6LOWPAN:
2264 return addrconf_ifid_6lowpan(eui, dev);
2265 case ARPHRD_IEEE1394:
2266 return addrconf_ifid_ieee1394(eui, dev);
2267 case ARPHRD_TUNNEL6:
2269 return addrconf_ifid_ip6tnl(eui, dev);
2274 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2277 struct inet6_ifaddr *ifp;
2279 read_lock_bh(&idev->lock);
2280 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2281 if (ifp->scope > IFA_LINK)
2283 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2284 memcpy(eui, ifp->addr.s6_addr+8, 8);
2289 read_unlock_bh(&idev->lock);
2293 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2294 static void ipv6_regen_rndid(struct inet6_dev *idev)
2297 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2298 idev->rndid[0] &= ~0x02;
2301 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2302 * check if generated address is not inappropriate
2304 * - Reserved subnet anycast (RFC 2526)
2305 * 11111101 11....11 1xxxxxxx
2306 * - ISATAP (RFC4214) 6.1
2307 * 00-00-5E-FE-xx-xx-xx-xx
2309 * - XXX: already assigned to an address on the device
2311 if (idev->rndid[0] == 0xfd &&
2312 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2313 (idev->rndid[7]&0x80))
2315 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2316 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2318 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2323 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2325 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2326 ipv6_regen_rndid(idev);
2334 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2335 unsigned long expires, u32 flags)
2337 struct fib6_config cfg = {
2338 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2339 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2340 .fc_ifindex = dev->ifindex,
2341 .fc_expires = expires,
2343 .fc_flags = RTF_UP | flags,
2344 .fc_nlinfo.nl_net = dev_net(dev),
2345 .fc_protocol = RTPROT_KERNEL,
2350 /* Prevent useless cloning on PtP SIT.
2351 This thing is done here expecting that the whole
2352 class of non-broadcast devices need not cloning.
2354 #if IS_ENABLED(CONFIG_IPV6_SIT)
2355 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2356 cfg.fc_flags |= RTF_NONEXTHOP;
2359 ip6_route_add(&cfg, NULL);
2363 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2365 const struct net_device *dev,
2366 u32 flags, u32 noflags)
2368 struct fib6_node *fn;
2369 struct rt6_info *rt = NULL;
2370 struct fib6_table *table;
2371 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2373 table = fib6_get_table(dev_net(dev), tb_id);
2378 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2382 for_each_fib6_node_rt_rcu(fn) {
2383 if (rt->dst.dev->ifindex != dev->ifindex)
2385 if ((rt->rt6i_flags & flags) != flags)
2387 if ((rt->rt6i_flags & noflags) != 0)
2389 if (!dst_hold_safe(&rt->dst))
2399 /* Create "default" multicast route to the interface */
2401 static void addrconf_add_mroute(struct net_device *dev)
2403 struct fib6_config cfg = {
2404 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2405 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2406 .fc_ifindex = dev->ifindex,
2409 .fc_nlinfo.nl_net = dev_net(dev),
2412 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2414 ip6_route_add(&cfg, NULL);
2417 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2419 struct inet6_dev *idev;
2423 idev = ipv6_find_idev(dev);
2425 return ERR_PTR(-ENOBUFS);
2427 if (idev->cnf.disable_ipv6)
2428 return ERR_PTR(-EACCES);
2430 /* Add default multicast route */
2431 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2432 addrconf_add_mroute(dev);
2437 static void manage_tempaddrs(struct inet6_dev *idev,
2438 struct inet6_ifaddr *ifp,
2439 __u32 valid_lft, __u32 prefered_lft,
2440 bool create, unsigned long now)
2443 struct inet6_ifaddr *ift;
2445 read_lock_bh(&idev->lock);
2446 /* update all temporary addresses in the list */
2447 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2448 int age, max_valid, max_prefered;
2450 if (ifp != ift->ifpub)
2453 /* RFC 4941 section 3.3:
2454 * If a received option will extend the lifetime of a public
2455 * address, the lifetimes of temporary addresses should
2456 * be extended, subject to the overall constraint that no
2457 * temporary addresses should ever remain "valid" or "preferred"
2458 * for a time longer than (TEMP_VALID_LIFETIME) or
2459 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2461 age = (now - ift->cstamp) / HZ;
2462 max_valid = idev->cnf.temp_valid_lft - age;
2466 max_prefered = idev->cnf.temp_prefered_lft -
2467 idev->desync_factor - age;
2468 if (max_prefered < 0)
2471 if (valid_lft > max_valid)
2472 valid_lft = max_valid;
2474 if (prefered_lft > max_prefered)
2475 prefered_lft = max_prefered;
2477 spin_lock(&ift->lock);
2479 ift->valid_lft = valid_lft;
2480 ift->prefered_lft = prefered_lft;
2482 if (prefered_lft > 0)
2483 ift->flags &= ~IFA_F_DEPRECATED;
2485 spin_unlock(&ift->lock);
2486 if (!(flags&IFA_F_TENTATIVE))
2487 ipv6_ifa_notify(0, ift);
2490 if ((create || list_empty(&idev->tempaddr_list)) &&
2491 idev->cnf.use_tempaddr > 0) {
2492 /* When a new public address is created as described
2493 * in [ADDRCONF], also create a new temporary address.
2494 * Also create a temporary address if it's enabled but
2495 * no temporary address currently exists.
2497 read_unlock_bh(&idev->lock);
2498 ipv6_create_tempaddr(ifp, NULL, false);
2500 read_unlock_bh(&idev->lock);
2504 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2506 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2507 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2510 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2511 const struct prefix_info *pinfo,
2512 struct inet6_dev *in6_dev,
2513 const struct in6_addr *addr, int addr_type,
2514 u32 addr_flags, bool sllao, bool tokenized,
2515 __u32 valid_lft, u32 prefered_lft)
2517 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2518 int create = 0, update_lft = 0;
2520 if (!ifp && valid_lft) {
2521 int max_addresses = in6_dev->cnf.max_addresses;
2523 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2524 if ((net->ipv6.devconf_all->optimistic_dad ||
2525 in6_dev->cnf.optimistic_dad) &&
2526 !net->ipv6.devconf_all->forwarding && sllao)
2527 addr_flags |= IFA_F_OPTIMISTIC;
2530 /* Do not allow to create too much of autoconfigured
2531 * addresses; this would be too easy way to crash kernel.
2533 if (!max_addresses ||
2534 ipv6_count_addresses(in6_dev) < max_addresses)
2535 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2537 addr_type&IPV6_ADDR_SCOPE_MASK,
2538 addr_flags, valid_lft,
2539 prefered_lft, false, NULL);
2541 if (IS_ERR_OR_NULL(ifp))
2546 spin_lock_bh(&ifp->lock);
2547 ifp->flags |= IFA_F_MANAGETEMPADDR;
2548 ifp->cstamp = jiffies;
2549 ifp->tokenized = tokenized;
2550 spin_unlock_bh(&ifp->lock);
2551 addrconf_dad_start(ifp);
2559 /* update lifetime (RFC2462 5.5.3 e) */
2560 spin_lock_bh(&ifp->lock);
2562 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2563 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2566 if (!update_lft && !create && stored_lft) {
2567 const u32 minimum_lft = min_t(u32,
2568 stored_lft, MIN_VALID_LIFETIME);
2569 valid_lft = max(valid_lft, minimum_lft);
2571 /* RFC4862 Section 5.5.3e:
2572 * "Note that the preferred lifetime of the
2573 * corresponding address is always reset to
2574 * the Preferred Lifetime in the received
2575 * Prefix Information option, regardless of
2576 * whether the valid lifetime is also reset or
2579 * So we should always update prefered_lft here.
2585 ifp->valid_lft = valid_lft;
2586 ifp->prefered_lft = prefered_lft;
2589 ifp->flags &= ~IFA_F_DEPRECATED;
2590 spin_unlock_bh(&ifp->lock);
2592 if (!(flags&IFA_F_TENTATIVE))
2593 ipv6_ifa_notify(0, ifp);
2595 spin_unlock_bh(&ifp->lock);
2597 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2606 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2608 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2610 struct prefix_info *pinfo;
2615 struct inet6_dev *in6_dev;
2616 struct net *net = dev_net(dev);
2618 pinfo = (struct prefix_info *) opt;
2620 if (len < sizeof(struct prefix_info)) {
2621 ADBG("addrconf: prefix option too short\n");
2626 * Validation checks ([ADDRCONF], page 19)
2629 addr_type = ipv6_addr_type(&pinfo->prefix);
2631 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2634 valid_lft = ntohl(pinfo->valid);
2635 prefered_lft = ntohl(pinfo->prefered);
2637 if (prefered_lft > valid_lft) {
2638 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2642 in6_dev = in6_dev_get(dev);
2645 net_dbg_ratelimited("addrconf: device %s not configured\n",
2651 * Two things going on here:
2652 * 1) Add routes for on-link prefixes
2653 * 2) Configure prefixes with the auto flag set
2656 if (pinfo->onlink) {
2657 struct rt6_info *rt;
2658 unsigned long rt_expires;
2660 /* Avoid arithmetic overflow. Really, we could
2661 * save rt_expires in seconds, likely valid_lft,
2662 * but it would require division in fib gc, that it
2666 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2668 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2670 if (addrconf_finite_timeout(rt_expires))
2673 rt = addrconf_get_prefix_route(&pinfo->prefix,
2676 RTF_ADDRCONF | RTF_PREFIX_RT,
2677 RTF_GATEWAY | RTF_DEFAULT);
2680 /* Autoconf prefix route */
2681 if (valid_lft == 0) {
2684 } else if (addrconf_finite_timeout(rt_expires)) {
2686 rt6_set_expires(rt, jiffies + rt_expires);
2688 rt6_clean_expires(rt);
2690 } else if (valid_lft) {
2691 clock_t expires = 0;
2692 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2693 if (addrconf_finite_timeout(rt_expires)) {
2695 flags |= RTF_EXPIRES;
2696 expires = jiffies_to_clock_t(rt_expires);
2698 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2699 dev, expires, flags);
2704 /* Try to figure out our local address for this prefix */
2706 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2707 struct in6_addr addr;
2708 bool tokenized = false, dev_addr_generated = false;
2710 if (pinfo->prefix_len == 64) {
2711 memcpy(&addr, &pinfo->prefix, 8);
2713 if (!ipv6_addr_any(&in6_dev->token)) {
2714 read_lock_bh(&in6_dev->lock);
2715 memcpy(addr.s6_addr + 8,
2716 in6_dev->token.s6_addr + 8, 8);
2717 read_unlock_bh(&in6_dev->lock);
2719 } else if (is_addr_mode_generate_stable(in6_dev) &&
2720 !ipv6_generate_stable_address(&addr, 0,
2722 addr_flags |= IFA_F_STABLE_PRIVACY;
2724 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2725 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2728 dev_addr_generated = true;
2732 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2737 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2740 tokenized, valid_lft,
2745 /* Ignore error case here because previous prefix add addr was
2746 * successful which will be notified.
2748 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2749 addr_type, addr_flags, sllao,
2750 tokenized, valid_lft,
2752 dev_addr_generated);
2754 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2756 in6_dev_put(in6_dev);
2760 * Set destination address.
2761 * Special case for SIT interfaces where we create a new "virtual"
2764 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2766 struct in6_ifreq ireq;
2767 struct net_device *dev;
2773 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2776 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2782 #if IS_ENABLED(CONFIG_IPV6_SIT)
2783 if (dev->type == ARPHRD_SIT) {
2784 const struct net_device_ops *ops = dev->netdev_ops;
2786 struct ip_tunnel_parm p;
2788 err = -EADDRNOTAVAIL;
2789 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2792 memset(&p, 0, sizeof(p));
2793 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2797 p.iph.protocol = IPPROTO_IPV6;
2799 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2801 if (ops->ndo_do_ioctl) {
2802 mm_segment_t oldfs = get_fs();
2805 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2812 dev = __dev_get_by_name(net, p.name);
2815 err = dev_open(dev);
2825 static int ipv6_mc_config(struct sock *sk, bool join,
2826 const struct in6_addr *addr, int ifindex)
2834 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2836 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2843 * Manual configuration of address on an interface
2845 static int inet6_addr_add(struct net *net, int ifindex,
2846 const struct in6_addr *pfx,
2847 const struct in6_addr *peer_pfx,
2848 unsigned int plen, __u32 ifa_flags,
2849 __u32 prefered_lft, __u32 valid_lft,
2850 struct netlink_ext_ack *extack)
2852 struct inet6_ifaddr *ifp;
2853 struct inet6_dev *idev;
2854 struct net_device *dev;
2855 unsigned long timeout;
2865 /* check the lifetime */
2866 if (!valid_lft || prefered_lft > valid_lft)
2869 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2872 dev = __dev_get_by_index(net, ifindex);
2876 idev = addrconf_add_dev(dev);
2878 return PTR_ERR(idev);
2880 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2881 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2882 true, pfx, ifindex);
2888 scope = ipv6_addr_scope(pfx);
2890 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2891 if (addrconf_finite_timeout(timeout)) {
2892 expires = jiffies_to_clock_t(timeout * HZ);
2893 valid_lft = timeout;
2894 flags = RTF_EXPIRES;
2898 ifa_flags |= IFA_F_PERMANENT;
2901 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2902 if (addrconf_finite_timeout(timeout)) {
2904 ifa_flags |= IFA_F_DEPRECATED;
2905 prefered_lft = timeout;
2908 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2909 valid_lft, prefered_lft, true, extack);
2912 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2913 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2918 * Note that section 3.1 of RFC 4429 indicates
2919 * that the Optimistic flag should not be set for
2920 * manually configured addresses
2922 addrconf_dad_start(ifp);
2923 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2924 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2927 addrconf_verify_rtnl();
2929 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2930 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2931 false, pfx, ifindex);
2934 return PTR_ERR(ifp);
2937 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2938 const struct in6_addr *pfx, unsigned int plen)
2940 struct inet6_ifaddr *ifp;
2941 struct inet6_dev *idev;
2942 struct net_device *dev;
2947 dev = __dev_get_by_index(net, ifindex);
2951 idev = __in6_dev_get(dev);
2955 read_lock_bh(&idev->lock);
2956 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2957 if (ifp->prefix_len == plen &&
2958 ipv6_addr_equal(pfx, &ifp->addr)) {
2960 read_unlock_bh(&idev->lock);
2962 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2963 (ifa_flags & IFA_F_MANAGETEMPADDR))
2964 manage_tempaddrs(idev, ifp, 0, 0, false,
2967 addrconf_verify_rtnl();
2968 if (ipv6_addr_is_multicast(pfx)) {
2969 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2970 false, pfx, dev->ifindex);
2975 read_unlock_bh(&idev->lock);
2976 return -EADDRNOTAVAIL;
2980 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2982 struct in6_ifreq ireq;
2985 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2988 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2992 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2993 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2994 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2999 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3001 struct in6_ifreq ireq;
3004 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3007 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3011 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3012 ireq.ifr6_prefixlen);
3017 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3018 int plen, int scope)
3020 struct inet6_ifaddr *ifp;
3022 ifp = ipv6_add_addr(idev, addr, NULL, plen,
3023 scope, IFA_F_PERMANENT,
3024 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
3027 spin_lock_bh(&ifp->lock);
3028 ifp->flags &= ~IFA_F_TENTATIVE;
3029 spin_unlock_bh(&ifp->lock);
3030 rt_genid_bump_ipv6(dev_net(idev->dev));
3031 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3036 #if IS_ENABLED(CONFIG_IPV6_SIT)
3037 static void sit_add_v4_addrs(struct inet6_dev *idev)
3039 struct in6_addr addr;
3040 struct net_device *dev;
3041 struct net *net = dev_net(idev->dev);
3047 memset(&addr, 0, sizeof(struct in6_addr));
3048 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3050 if (idev->dev->flags&IFF_POINTOPOINT) {
3051 addr.s6_addr32[0] = htonl(0xfe800000);
3055 scope = IPV6_ADDR_COMPATv4;
3057 pflags |= RTF_NONEXTHOP;
3060 if (addr.s6_addr32[3]) {
3061 add_addr(idev, &addr, plen, scope);
3062 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3066 for_each_netdev(net, dev) {
3067 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3068 if (in_dev && (dev->flags & IFF_UP)) {
3069 struct in_ifaddr *ifa;
3073 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3075 addr.s6_addr32[3] = ifa->ifa_local;
3077 if (ifa->ifa_scope == RT_SCOPE_LINK)
3079 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3080 if (idev->dev->flags&IFF_POINTOPOINT)
3085 add_addr(idev, &addr, plen, flag);
3086 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3094 static void init_loopback(struct net_device *dev)
3096 struct inet6_dev *idev;
3102 idev = ipv6_find_idev(dev);
3104 pr_debug("%s: add_dev failed\n", __func__);
3108 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3111 void addrconf_add_linklocal(struct inet6_dev *idev,
3112 const struct in6_addr *addr, u32 flags)
3114 struct inet6_ifaddr *ifp;
3115 u32 addr_flags = flags | IFA_F_PERMANENT;
3117 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3118 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3119 idev->cnf.optimistic_dad) &&
3120 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3121 addr_flags |= IFA_F_OPTIMISTIC;
3124 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3125 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3127 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3128 addrconf_dad_start(ifp);
3132 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3134 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3136 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3139 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3140 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3143 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3144 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3150 static int ipv6_generate_stable_address(struct in6_addr *address,
3152 const struct inet6_dev *idev)
3154 static DEFINE_SPINLOCK(lock);
3155 static __u32 digest[SHA_DIGEST_WORDS];
3156 static __u32 workspace[SHA_WORKSPACE_WORDS];
3159 char __data[SHA_MESSAGE_BYTES];
3161 struct in6_addr secret;
3163 unsigned char hwaddr[MAX_ADDR_LEN];
3168 struct in6_addr secret;
3169 struct in6_addr temp;
3170 struct net *net = dev_net(idev->dev);
3172 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3174 if (idev->cnf.stable_secret.initialized)
3175 secret = idev->cnf.stable_secret.secret;
3176 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3177 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3182 spin_lock_bh(&lock);
3185 memset(&data, 0, sizeof(data));
3186 memset(workspace, 0, sizeof(workspace));
3187 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3188 data.prefix[0] = address->s6_addr32[0];
3189 data.prefix[1] = address->s6_addr32[1];
3190 data.secret = secret;
3191 data.dad_count = dad_count;
3193 sha_transform(digest, data.__data, workspace);
3196 temp.s6_addr32[2] = (__force __be32)digest[0];
3197 temp.s6_addr32[3] = (__force __be32)digest[1];
3199 spin_unlock_bh(&lock);
3201 if (ipv6_reserved_interfaceid(temp)) {
3203 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3212 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3214 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3218 s = &idev->cnf.stable_secret;
3219 get_random_bytes(&s->secret, sizeof(s->secret));
3220 s->initialized = true;
3223 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3225 struct in6_addr addr;
3227 /* no link local addresses on L3 master devices */
3228 if (netif_is_l3_master(idev->dev))
3231 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3233 switch (idev->cnf.addr_gen_mode) {
3234 case IN6_ADDR_GEN_MODE_RANDOM:
3235 ipv6_gen_mode_random_init(idev);
3237 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3238 if (!ipv6_generate_stable_address(&addr, 0, idev))
3239 addrconf_add_linklocal(idev, &addr,
3240 IFA_F_STABLE_PRIVACY);
3241 else if (prefix_route)
3242 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3244 case IN6_ADDR_GEN_MODE_EUI64:
3245 /* addrconf_add_linklocal also adds a prefix_route and we
3246 * only need to care about prefix routes if ipv6_generate_eui64
3247 * couldn't generate one.
3249 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3250 addrconf_add_linklocal(idev, &addr, 0);
3251 else if (prefix_route)
3252 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3254 case IN6_ADDR_GEN_MODE_NONE:
3256 /* will not add any link local address */
3261 static void addrconf_dev_config(struct net_device *dev)
3263 struct inet6_dev *idev;
3267 if ((dev->type != ARPHRD_ETHER) &&
3268 (dev->type != ARPHRD_FDDI) &&
3269 (dev->type != ARPHRD_ARCNET) &&
3270 (dev->type != ARPHRD_INFINIBAND) &&
3271 (dev->type != ARPHRD_IEEE1394) &&
3272 (dev->type != ARPHRD_TUNNEL6) &&
3273 (dev->type != ARPHRD_6LOWPAN) &&
3274 (dev->type != ARPHRD_IP6GRE) &&
3275 (dev->type != ARPHRD_IPGRE) &&
3276 (dev->type != ARPHRD_TUNNEL) &&
3277 (dev->type != ARPHRD_NONE)) {
3278 /* Alas, we support only Ethernet autoconfiguration. */
3282 idev = addrconf_add_dev(dev);
3286 /* this device type has no EUI support */
3287 if (dev->type == ARPHRD_NONE &&
3288 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3289 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3291 addrconf_addr_gen(idev, false);
3294 #if IS_ENABLED(CONFIG_IPV6_SIT)
3295 static void addrconf_sit_config(struct net_device *dev)
3297 struct inet6_dev *idev;
3302 * Configure the tunnel with one of our IPv4
3303 * addresses... we should configure all of
3304 * our v4 addrs in the tunnel
3307 idev = ipv6_find_idev(dev);
3309 pr_debug("%s: add_dev failed\n", __func__);
3313 if (dev->priv_flags & IFF_ISATAP) {
3314 addrconf_addr_gen(idev, false);
3318 sit_add_v4_addrs(idev);
3320 if (dev->flags&IFF_POINTOPOINT)
3321 addrconf_add_mroute(dev);
3325 #if IS_ENABLED(CONFIG_NET_IPGRE)
3326 static void addrconf_gre_config(struct net_device *dev)
3328 struct inet6_dev *idev;
3332 idev = ipv6_find_idev(dev);
3334 pr_debug("%s: add_dev failed\n", __func__);
3338 addrconf_addr_gen(idev, true);
3339 if (dev->flags & IFF_POINTOPOINT)
3340 addrconf_add_mroute(dev);
3344 static int fixup_permanent_addr(struct inet6_dev *idev,
3345 struct inet6_ifaddr *ifp)
3347 /* !rt6i_node means the host route was removed from the
3348 * FIB, for example, if 'lo' device is taken down. In that
3349 * case regenerate the host route.
3351 if (!ifp->rt || !ifp->rt->rt6i_node) {
3352 struct rt6_info *rt, *prev;
3354 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3358 /* ifp->rt can be accessed outside of rtnl */
3359 spin_lock(&ifp->lock);
3362 spin_unlock(&ifp->lock);
3367 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3368 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3372 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3373 addrconf_dad_start(ifp);
3378 static void addrconf_permanent_addr(struct net_device *dev)
3380 struct inet6_ifaddr *ifp, *tmp;
3381 struct inet6_dev *idev;
3383 idev = __in6_dev_get(dev);
3387 write_lock_bh(&idev->lock);
3389 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3390 if ((ifp->flags & IFA_F_PERMANENT) &&
3391 fixup_permanent_addr(idev, ifp) < 0) {
3392 write_unlock_bh(&idev->lock);
3395 write_lock_bh(&idev->lock);
3397 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3398 idev->dev->name, &ifp->addr);
3402 write_unlock_bh(&idev->lock);
3405 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3408 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3409 struct netdev_notifier_changeupper_info *info;
3410 struct inet6_dev *idev = __in6_dev_get(dev);
3411 struct net *net = dev_net(dev);
3412 int run_pending = 0;
3416 case NETDEV_REGISTER:
3417 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3418 idev = ipv6_add_dev(dev);
3420 return notifier_from_errno(PTR_ERR(idev));
3424 case NETDEV_CHANGEMTU:
3425 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3426 if (dev->mtu < IPV6_MIN_MTU) {
3427 addrconf_ifdown(dev, dev != net->loopback_dev);
3432 rt6_mtu_change(dev, dev->mtu);
3433 idev->cnf.mtu6 = dev->mtu;
3437 /* allocate new idev */
3438 idev = ipv6_add_dev(dev);
3442 /* device is still not ready */
3443 if (!(idev->if_flags & IF_READY))
3452 if (dev->flags & IFF_SLAVE)
3455 if (idev && idev->cnf.disable_ipv6)
3458 if (event == NETDEV_UP) {
3459 /* restore routes for permanent addresses */
3460 addrconf_permanent_addr(dev);
3462 if (!addrconf_link_ready(dev)) {
3463 /* device is not ready yet. */
3464 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3469 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3470 idev = ipv6_add_dev(dev);
3472 if (!IS_ERR_OR_NULL(idev)) {
3473 idev->if_flags |= IF_READY;
3476 } else if (event == NETDEV_CHANGE) {
3477 if (!addrconf_link_ready(dev)) {
3478 /* device is still not ready. */
3479 rt6_sync_down_dev(dev, event);
3484 if (idev->if_flags & IF_READY) {
3485 /* device is already configured -
3486 * but resend MLD reports, we might
3487 * have roamed and need to update
3488 * multicast snooping switches
3491 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3494 idev->if_flags |= IF_READY;
3497 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3503 switch (dev->type) {
3504 #if IS_ENABLED(CONFIG_IPV6_SIT)
3506 addrconf_sit_config(dev);
3509 #if IS_ENABLED(CONFIG_NET_IPGRE)
3511 addrconf_gre_config(dev);
3514 case ARPHRD_LOOPBACK:
3519 addrconf_dev_config(dev);
3523 if (!IS_ERR_OR_NULL(idev)) {
3525 addrconf_dad_run(idev);
3527 /* Device has an address by now */
3528 rt6_sync_up(dev, RTNH_F_DEAD);
3531 * If the MTU changed during the interface down,
3532 * when the interface up, the changed MTU must be
3533 * reflected in the idev as well as routers.
3535 if (idev->cnf.mtu6 != dev->mtu &&
3536 dev->mtu >= IPV6_MIN_MTU) {
3537 rt6_mtu_change(dev, dev->mtu);
3538 idev->cnf.mtu6 = dev->mtu;
3540 idev->tstamp = jiffies;
3541 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3544 * If the changed mtu during down is lower than
3545 * IPV6_MIN_MTU stop IPv6 on this interface.
3547 if (dev->mtu < IPV6_MIN_MTU)
3548 addrconf_ifdown(dev, dev != net->loopback_dev);
3553 case NETDEV_UNREGISTER:
3555 * Remove all addresses from this interface.
3557 addrconf_ifdown(dev, event != NETDEV_DOWN);
3560 case NETDEV_CHANGENAME:
3562 snmp6_unregister_dev(idev);
3563 addrconf_sysctl_unregister(idev);
3564 err = addrconf_sysctl_register(idev);
3566 return notifier_from_errno(err);
3567 err = snmp6_register_dev(idev);
3569 addrconf_sysctl_unregister(idev);
3570 return notifier_from_errno(err);
3575 case NETDEV_PRE_TYPE_CHANGE:
3576 case NETDEV_POST_TYPE_CHANGE:
3578 addrconf_type_change(dev, event);
3581 case NETDEV_CHANGEUPPER:
3584 /* flush all routes if dev is linked to or unlinked from
3585 * an L3 master device (e.g., VRF)
3587 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3588 addrconf_ifdown(dev, 0);
3595 * addrconf module should be notified of a device going up
3597 static struct notifier_block ipv6_dev_notf = {
3598 .notifier_call = addrconf_notify,
3599 .priority = ADDRCONF_NOTIFY_PRIORITY,
3602 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3604 struct inet6_dev *idev;
3607 idev = __in6_dev_get(dev);
3609 if (event == NETDEV_POST_TYPE_CHANGE)
3610 ipv6_mc_remap(idev);
3611 else if (event == NETDEV_PRE_TYPE_CHANGE)
3612 ipv6_mc_unmap(idev);
3615 static bool addr_is_local(const struct in6_addr *addr)
3617 return ipv6_addr_type(addr) &
3618 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3621 static int addrconf_ifdown(struct net_device *dev, int how)
3623 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3624 struct net *net = dev_net(dev);
3625 struct inet6_dev *idev;
3626 struct inet6_ifaddr *ifa, *tmp;
3633 rt6_disable_ip(dev, event);
3635 idev = __in6_dev_get(dev);
3640 * Step 1: remove reference to ipv6 device from parent device.
3646 /* protected by rtnl_lock */
3647 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3649 /* Step 1.5: remove snmp6 entry */
3650 snmp6_unregister_dev(idev);
3654 /* aggregate the system setting and interface setting */
3655 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3657 _keep_addr = idev->cnf.keep_addr_on_down;
3659 /* combine the user config with event to determine if permanent
3660 * addresses are to be removed from address hash table
3662 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3664 /* Step 2: clear hash table */
3665 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3666 struct hlist_head *h = &inet6_addr_lst[i];
3668 spin_lock_bh(&addrconf_hash_lock);
3670 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3671 if (ifa->idev == idev) {
3672 addrconf_del_dad_work(ifa);
3673 /* combined flag + permanent flag decide if
3674 * address is retained on a down event
3677 !(ifa->flags & IFA_F_PERMANENT) ||
3678 addr_is_local(&ifa->addr)) {
3679 hlist_del_init_rcu(&ifa->addr_lst);
3684 spin_unlock_bh(&addrconf_hash_lock);
3687 write_lock_bh(&idev->lock);
3689 addrconf_del_rs_timer(idev);
3691 /* Step 2: clear flags for stateless addrconf */
3693 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3695 /* Step 3: clear tempaddr list */
3696 while (!list_empty(&idev->tempaddr_list)) {
3697 ifa = list_first_entry(&idev->tempaddr_list,
3698 struct inet6_ifaddr, tmp_list);
3699 list_del(&ifa->tmp_list);
3700 write_unlock_bh(&idev->lock);
3701 spin_lock_bh(&ifa->lock);
3704 in6_ifa_put(ifa->ifpub);
3707 spin_unlock_bh(&ifa->lock);
3709 write_lock_bh(&idev->lock);
3712 /* re-combine the user config with event to determine if permanent
3713 * addresses are to be removed from the interface list
3715 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3717 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3718 struct rt6_info *rt = NULL;
3721 addrconf_del_dad_work(ifa);
3723 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3724 !addr_is_local(&ifa->addr);
3726 write_unlock_bh(&idev->lock);
3727 spin_lock_bh(&ifa->lock);
3730 /* set state to skip the notifier below */
3731 state = INET6_IFADDR_STATE_DEAD;
3732 ifa->state = INET6_IFADDR_STATE_PREDAD;
3733 if (!(ifa->flags & IFA_F_NODAD))
3734 ifa->flags |= IFA_F_TENTATIVE;
3740 ifa->state = INET6_IFADDR_STATE_DEAD;
3743 spin_unlock_bh(&ifa->lock);
3748 if (state != INET6_IFADDR_STATE_DEAD) {
3749 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3750 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3752 if (idev->cnf.forwarding)
3753 addrconf_leave_anycast(ifa);
3754 addrconf_leave_solict(ifa->idev, &ifa->addr);
3757 write_lock_bh(&idev->lock);
3759 list_del_rcu(&ifa->if_list);
3764 write_unlock_bh(&idev->lock);
3766 /* Step 5: Discard anycast and multicast list */
3768 ipv6_ac_destroy_dev(idev);
3769 ipv6_mc_destroy_dev(idev);
3774 idev->tstamp = jiffies;
3776 /* Last: Shot the device (if unregistered) */
3778 addrconf_sysctl_unregister(idev);
3779 neigh_parms_release(&nd_tbl, idev->nd_parms);
3780 neigh_ifdown(&nd_tbl, dev);
3786 static void addrconf_rs_timer(struct timer_list *t)
3788 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3789 struct net_device *dev = idev->dev;
3790 struct in6_addr lladdr;
3792 write_lock(&idev->lock);
3793 if (idev->dead || !(idev->if_flags & IF_READY))
3796 if (!ipv6_accept_ra(idev))
3799 /* Announcement received after solicitation was sent */
3800 if (idev->if_flags & IF_RA_RCVD)
3803 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3804 write_unlock(&idev->lock);
3805 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3806 ndisc_send_rs(dev, &lladdr,
3807 &in6addr_linklocal_allrouters);
3811 write_lock(&idev->lock);
3812 idev->rs_interval = rfc3315_s14_backoff_update(
3813 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3814 /* The wait after the last probe can be shorter */
3815 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3816 idev->cnf.rtr_solicits) ?
3817 idev->cnf.rtr_solicit_delay :
3821 * Note: we do not support deprecated "all on-link"
3822 * assumption any longer.
3824 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3828 write_unlock(&idev->lock);
3834 * Duplicate Address Detection
3836 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3838 unsigned long rand_num;
3839 struct inet6_dev *idev = ifp->idev;
3842 if (ifp->flags & IFA_F_OPTIMISTIC)
3845 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3848 if (idev->cnf.enhanced_dad ||
3849 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3851 get_random_bytes(&nonce, 6);
3854 ifp->dad_nonce = nonce;
3855 ifp->dad_probes = idev->cnf.dad_transmits;
3856 addrconf_mod_dad_work(ifp, rand_num);
3859 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3861 struct inet6_dev *idev = ifp->idev;
3862 struct net_device *dev = idev->dev;
3863 bool bump_id, notify = false;
3865 addrconf_join_solict(dev, &ifp->addr);
3867 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3869 read_lock_bh(&idev->lock);
3870 spin_lock(&ifp->lock);
3871 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3874 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3875 (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3876 idev->cnf.accept_dad < 1) ||
3877 !(ifp->flags&IFA_F_TENTATIVE) ||
3878 ifp->flags & IFA_F_NODAD) {
3879 bool send_na = false;
3881 if (ifp->flags & IFA_F_TENTATIVE &&
3882 !(ifp->flags & IFA_F_OPTIMISTIC))
3884 bump_id = ifp->flags & IFA_F_TENTATIVE;
3885 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3886 spin_unlock(&ifp->lock);
3887 read_unlock_bh(&idev->lock);
3889 addrconf_dad_completed(ifp, bump_id, send_na);
3893 if (!(idev->if_flags & IF_READY)) {
3894 spin_unlock(&ifp->lock);
3895 read_unlock_bh(&idev->lock);
3897 * If the device is not ready:
3898 * - keep it tentative if it is a permanent address.
3899 * - otherwise, kill it.
3902 addrconf_dad_stop(ifp, 0);
3907 * Optimistic nodes can start receiving
3910 if (ifp->flags & IFA_F_OPTIMISTIC) {
3911 ip6_ins_rt(ifp->rt);
3912 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3913 /* Because optimistic nodes can use this address,
3914 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3920 addrconf_dad_kick(ifp);
3922 spin_unlock(&ifp->lock);
3923 read_unlock_bh(&idev->lock);
3925 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3928 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3930 bool begin_dad = false;
3932 spin_lock_bh(&ifp->lock);
3933 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3934 ifp->state = INET6_IFADDR_STATE_PREDAD;
3937 spin_unlock_bh(&ifp->lock);
3940 addrconf_mod_dad_work(ifp, 0);
3943 static void addrconf_dad_work(struct work_struct *w)
3945 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3946 struct inet6_ifaddr,
3948 struct inet6_dev *idev = ifp->idev;
3949 bool bump_id, disable_ipv6 = false;
3950 struct in6_addr mcaddr;
3956 } action = DAD_PROCESS;
3960 spin_lock_bh(&ifp->lock);
3961 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3963 ifp->state = INET6_IFADDR_STATE_DAD;
3964 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3966 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3968 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3969 idev->cnf.accept_dad > 1) &&
3970 !idev->cnf.disable_ipv6 &&
3971 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3972 struct in6_addr addr;
3974 addr.s6_addr32[0] = htonl(0xfe800000);
3975 addr.s6_addr32[1] = 0;
3977 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3978 ipv6_addr_equal(&ifp->addr, &addr)) {
3979 /* DAD failed for link-local based on MAC */
3980 idev->cnf.disable_ipv6 = 1;
3982 pr_info("%s: IPv6 being disabled!\n",
3983 ifp->idev->dev->name);
3984 disable_ipv6 = true;
3988 spin_unlock_bh(&ifp->lock);
3990 if (action == DAD_BEGIN) {
3991 addrconf_dad_begin(ifp);
3993 } else if (action == DAD_ABORT) {
3995 addrconf_dad_stop(ifp, 1);
3997 addrconf_ifdown(idev->dev, 0);
4001 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4004 write_lock_bh(&idev->lock);
4005 if (idev->dead || !(idev->if_flags & IF_READY)) {
4006 write_unlock_bh(&idev->lock);
4010 spin_lock(&ifp->lock);
4011 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4012 spin_unlock(&ifp->lock);
4013 write_unlock_bh(&idev->lock);
4017 if (ifp->dad_probes == 0) {
4018 bool send_na = false;
4021 * DAD was successful
4024 if (ifp->flags & IFA_F_TENTATIVE &&
4025 !(ifp->flags & IFA_F_OPTIMISTIC))
4027 bump_id = ifp->flags & IFA_F_TENTATIVE;
4028 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4029 spin_unlock(&ifp->lock);
4030 write_unlock_bh(&idev->lock);
4032 addrconf_dad_completed(ifp, bump_id, send_na);
4038 addrconf_mod_dad_work(ifp,
4039 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4040 spin_unlock(&ifp->lock);
4041 write_unlock_bh(&idev->lock);
4043 /* send a neighbour solicitation for our addr */
4044 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4045 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4052 /* ifp->idev must be at least read locked */
4053 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4055 struct inet6_ifaddr *ifpiter;
4056 struct inet6_dev *idev = ifp->idev;
4058 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4059 if (ifpiter->scope > IFA_LINK)
4061 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4062 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4063 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4070 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4073 struct net_device *dev = ifp->idev->dev;
4074 struct in6_addr lladdr;
4075 bool send_rs, send_mld;
4077 addrconf_del_dad_work(ifp);
4080 * Configure the address for reception. Now it is valid.
4083 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4085 /* If added prefix is link local and we are prepared to process
4086 router advertisements, start sending router solicitations.
4089 read_lock_bh(&ifp->idev->lock);
4090 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4091 send_rs = send_mld &&
4092 ipv6_accept_ra(ifp->idev) &&
4093 ifp->idev->cnf.rtr_solicits != 0 &&
4094 (dev->flags&IFF_LOOPBACK) == 0;
4095 read_unlock_bh(&ifp->idev->lock);
4097 /* While dad is in progress mld report's source address is in6_addrany.
4098 * Resend with proper ll now.
4101 ipv6_mc_dad_complete(ifp->idev);
4103 /* send unsolicited NA if enabled */
4105 (ifp->idev->cnf.ndisc_notify ||
4106 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4107 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4108 /*router=*/ !!ifp->idev->cnf.forwarding,
4109 /*solicited=*/ false, /*override=*/ true,
4115 * If a host as already performed a random delay
4116 * [...] as part of DAD [...] there is no need
4117 * to delay again before sending the first RS
4119 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4121 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4123 write_lock_bh(&ifp->idev->lock);
4124 spin_lock(&ifp->lock);
4125 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4126 ifp->idev->cnf.rtr_solicit_interval);
4127 ifp->idev->rs_probes = 1;
4128 ifp->idev->if_flags |= IF_RS_SENT;
4129 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4130 spin_unlock(&ifp->lock);
4131 write_unlock_bh(&ifp->idev->lock);
4135 rt_genid_bump_ipv6(dev_net(dev));
4137 /* Make sure that a new temporary address will be created
4138 * before this temporary address becomes deprecated.
4140 if (ifp->flags & IFA_F_TEMPORARY)
4141 addrconf_verify_rtnl();
4144 static void addrconf_dad_run(struct inet6_dev *idev)
4146 struct inet6_ifaddr *ifp;
4148 read_lock_bh(&idev->lock);
4149 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4150 spin_lock(&ifp->lock);
4151 if (ifp->flags & IFA_F_TENTATIVE &&
4152 ifp->state == INET6_IFADDR_STATE_DAD)
4153 addrconf_dad_kick(ifp);
4154 spin_unlock(&ifp->lock);
4156 read_unlock_bh(&idev->lock);
4159 #ifdef CONFIG_PROC_FS
4160 struct if6_iter_state {
4161 struct seq_net_private p;
4166 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4168 struct if6_iter_state *state = seq->private;
4169 struct net *net = seq_file_net(seq);
4170 struct inet6_ifaddr *ifa = NULL;
4173 /* initial bucket if pos is 0 */
4179 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4180 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4182 if (!net_eq(dev_net(ifa->idev->dev), net))
4184 /* sync with offset */
4185 if (p < state->offset) {
4193 /* prepare for next bucket */
4200 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4201 struct inet6_ifaddr *ifa)
4203 struct if6_iter_state *state = seq->private;
4204 struct net *net = seq_file_net(seq);
4206 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4207 if (!net_eq(dev_net(ifa->idev->dev), net))
4213 while (++state->bucket < IN6_ADDR_HSIZE) {
4215 hlist_for_each_entry_rcu(ifa,
4216 &inet6_addr_lst[state->bucket], addr_lst) {
4217 if (!net_eq(dev_net(ifa->idev->dev), net))
4227 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4231 return if6_get_first(seq, *pos);
4234 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4236 struct inet6_ifaddr *ifa;
4238 ifa = if6_get_next(seq, v);
4243 static void if6_seq_stop(struct seq_file *seq, void *v)
4249 static int if6_seq_show(struct seq_file *seq, void *v)
4251 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4252 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4254 ifp->idev->dev->ifindex,
4258 ifp->idev->dev->name);
4262 static const struct seq_operations if6_seq_ops = {
4263 .start = if6_seq_start,
4264 .next = if6_seq_next,
4265 .show = if6_seq_show,
4266 .stop = if6_seq_stop,
4269 static int if6_seq_open(struct inode *inode, struct file *file)
4271 return seq_open_net(inode, file, &if6_seq_ops,
4272 sizeof(struct if6_iter_state));
4275 static const struct file_operations if6_fops = {
4276 .open = if6_seq_open,
4278 .llseek = seq_lseek,
4279 .release = seq_release_net,
4282 static int __net_init if6_proc_net_init(struct net *net)
4284 if (!proc_create("if_inet6", 0444, net->proc_net, &if6_fops))
4289 static void __net_exit if6_proc_net_exit(struct net *net)
4291 remove_proc_entry("if_inet6", net->proc_net);
4294 static struct pernet_operations if6_proc_net_ops = {
4295 .init = if6_proc_net_init,
4296 .exit = if6_proc_net_exit,
4300 int __init if6_proc_init(void)
4302 return register_pernet_subsys(&if6_proc_net_ops);
4305 void if6_proc_exit(void)
4307 unregister_pernet_subsys(&if6_proc_net_ops);
4309 #endif /* CONFIG_PROC_FS */
4311 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4312 /* Check if address is a home address configured on any interface. */
4313 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4315 unsigned int hash = inet6_addr_hash(net, addr);
4316 struct inet6_ifaddr *ifp = NULL;
4320 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4321 if (!net_eq(dev_net(ifp->idev->dev), net))
4323 if (ipv6_addr_equal(&ifp->addr, addr) &&
4324 (ifp->flags & IFA_F_HOMEADDRESS)) {
4335 * Periodic address status verification
4338 static void addrconf_verify_rtnl(void)
4340 unsigned long now, next, next_sec, next_sched;
4341 struct inet6_ifaddr *ifp;
4348 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4350 cancel_delayed_work(&addr_chk_work);
4352 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4354 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4357 /* When setting preferred_lft to a value not zero or
4358 * infinity, while valid_lft is infinity
4359 * IFA_F_PERMANENT has a non-infinity life time.
4361 if ((ifp->flags & IFA_F_PERMANENT) &&
4362 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4365 spin_lock(&ifp->lock);
4366 /* We try to batch several events at once. */
4367 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4369 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4370 age >= ifp->valid_lft) {
4371 spin_unlock(&ifp->lock);
4375 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4376 spin_unlock(&ifp->lock);
4378 } else if (age >= ifp->prefered_lft) {
4379 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4382 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4384 ifp->flags |= IFA_F_DEPRECATED;
4387 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4388 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4389 next = ifp->tstamp + ifp->valid_lft * HZ;
4391 spin_unlock(&ifp->lock);
4396 ipv6_ifa_notify(0, ifp);
4400 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4401 !(ifp->flags&IFA_F_TENTATIVE)) {
4402 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4403 ifp->idev->cnf.dad_transmits *
4404 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4406 if (age >= ifp->prefered_lft - regen_advance) {
4407 struct inet6_ifaddr *ifpub = ifp->ifpub;
4408 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4409 next = ifp->tstamp + ifp->prefered_lft * HZ;
4410 if (!ifp->regen_count && ifpub) {
4413 in6_ifa_hold(ifpub);
4414 spin_unlock(&ifp->lock);
4416 spin_lock(&ifpub->lock);
4417 ifpub->regen_count = 0;
4418 spin_unlock(&ifpub->lock);
4419 rcu_read_unlock_bh();
4420 ipv6_create_tempaddr(ifpub, ifp, true);
4426 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4427 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4428 spin_unlock(&ifp->lock);
4430 /* ifp->prefered_lft <= ifp->valid_lft */
4431 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4432 next = ifp->tstamp + ifp->prefered_lft * HZ;
4433 spin_unlock(&ifp->lock);
4438 next_sec = round_jiffies_up(next);
4441 /* If rounded timeout is accurate enough, accept it. */
4442 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4443 next_sched = next_sec;
4445 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4446 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4447 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4449 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4450 now, next, next_sec, next_sched);
4451 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4452 rcu_read_unlock_bh();
4455 static void addrconf_verify_work(struct work_struct *w)
4458 addrconf_verify_rtnl();
4462 static void addrconf_verify(void)
4464 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4467 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4468 struct in6_addr **peer_pfx)
4470 struct in6_addr *pfx = NULL;
4475 pfx = nla_data(addr);
4478 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4480 pfx = nla_data(local);
4486 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4487 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4488 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4489 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4490 [IFA_FLAGS] = { .len = sizeof(u32) },
4494 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4495 struct netlink_ext_ack *extack)
4497 struct net *net = sock_net(skb->sk);
4498 struct ifaddrmsg *ifm;
4499 struct nlattr *tb[IFA_MAX+1];
4500 struct in6_addr *pfx, *peer_pfx;
4504 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4509 ifm = nlmsg_data(nlh);
4510 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4514 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4516 /* We ignore other flags so far. */
4517 ifa_flags &= IFA_F_MANAGETEMPADDR;
4519 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4520 ifm->ifa_prefixlen);
4523 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4524 u32 prefered_lft, u32 valid_lft)
4528 unsigned long timeout;
4529 bool was_managetempaddr;
4530 bool had_prefixroute;
4534 if (!valid_lft || (prefered_lft > valid_lft))
4537 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4538 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4541 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4542 ifa_flags &= ~IFA_F_OPTIMISTIC;
4544 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4545 if (addrconf_finite_timeout(timeout)) {
4546 expires = jiffies_to_clock_t(timeout * HZ);
4547 valid_lft = timeout;
4548 flags = RTF_EXPIRES;
4552 ifa_flags |= IFA_F_PERMANENT;
4555 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4556 if (addrconf_finite_timeout(timeout)) {
4558 ifa_flags |= IFA_F_DEPRECATED;
4559 prefered_lft = timeout;
4562 spin_lock_bh(&ifp->lock);
4563 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4564 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4565 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4566 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4567 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4568 IFA_F_NOPREFIXROUTE);
4569 ifp->flags |= ifa_flags;
4570 ifp->tstamp = jiffies;
4571 ifp->valid_lft = valid_lft;
4572 ifp->prefered_lft = prefered_lft;
4574 spin_unlock_bh(&ifp->lock);
4575 if (!(ifp->flags&IFA_F_TENTATIVE))
4576 ipv6_ifa_notify(0, ifp);
4578 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4579 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4581 } else if (had_prefixroute) {
4582 enum cleanup_prefix_rt_t action;
4583 unsigned long rt_expires;
4585 write_lock_bh(&ifp->idev->lock);
4586 action = check_cleanup_prefix_route(ifp, &rt_expires);
4587 write_unlock_bh(&ifp->idev->lock);
4589 if (action != CLEANUP_PREFIX_RT_NOP) {
4590 cleanup_prefix_route(ifp, rt_expires,
4591 action == CLEANUP_PREFIX_RT_DEL);
4595 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4596 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4597 valid_lft = prefered_lft = 0;
4598 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4599 !was_managetempaddr, jiffies);
4602 addrconf_verify_rtnl();
4608 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4609 struct netlink_ext_ack *extack)
4611 struct net *net = sock_net(skb->sk);
4612 struct ifaddrmsg *ifm;
4613 struct nlattr *tb[IFA_MAX+1];
4614 struct in6_addr *pfx, *peer_pfx;
4615 struct inet6_ifaddr *ifa;
4616 struct net_device *dev;
4617 struct inet6_dev *idev;
4618 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4622 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4627 ifm = nlmsg_data(nlh);
4628 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4632 if (tb[IFA_CACHEINFO]) {
4633 struct ifa_cacheinfo *ci;
4635 ci = nla_data(tb[IFA_CACHEINFO]);
4636 valid_lft = ci->ifa_valid;
4637 preferred_lft = ci->ifa_prefered;
4639 preferred_lft = INFINITY_LIFE_TIME;
4640 valid_lft = INFINITY_LIFE_TIME;
4643 dev = __dev_get_by_index(net, ifm->ifa_index);
4647 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4649 /* We ignore other flags so far. */
4650 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4651 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4653 idev = ipv6_find_idev(dev);
4655 return PTR_ERR(idev);
4657 if (!ipv6_allow_optimistic_dad(net, idev))
4658 ifa_flags &= ~IFA_F_OPTIMISTIC;
4660 if (ifa_flags & IFA_F_NODAD && ifa_flags & IFA_F_OPTIMISTIC) {
4661 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4665 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4668 * It would be best to check for !NLM_F_CREATE here but
4669 * userspace already relies on not having to provide this.
4671 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4672 ifm->ifa_prefixlen, ifa_flags,
4673 preferred_lft, valid_lft, extack);
4676 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4677 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4680 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4687 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4688 u8 scope, int ifindex)
4690 struct ifaddrmsg *ifm;
4692 ifm = nlmsg_data(nlh);
4693 ifm->ifa_family = AF_INET6;
4694 ifm->ifa_prefixlen = prefixlen;
4695 ifm->ifa_flags = flags;
4696 ifm->ifa_scope = scope;
4697 ifm->ifa_index = ifindex;
4700 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4701 unsigned long tstamp, u32 preferred, u32 valid)
4703 struct ifa_cacheinfo ci;
4705 ci.cstamp = cstamp_delta(cstamp);
4706 ci.tstamp = cstamp_delta(tstamp);
4707 ci.ifa_prefered = preferred;
4708 ci.ifa_valid = valid;
4710 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4713 static inline int rt_scope(int ifa_scope)
4715 if (ifa_scope & IFA_HOST)
4716 return RT_SCOPE_HOST;
4717 else if (ifa_scope & IFA_LINK)
4718 return RT_SCOPE_LINK;
4719 else if (ifa_scope & IFA_SITE)
4720 return RT_SCOPE_SITE;
4722 return RT_SCOPE_UNIVERSE;
4725 static inline int inet6_ifaddr_msgsize(void)
4727 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4728 + nla_total_size(16) /* IFA_LOCAL */
4729 + nla_total_size(16) /* IFA_ADDRESS */
4730 + nla_total_size(sizeof(struct ifa_cacheinfo))
4731 + nla_total_size(4) /* IFA_FLAGS */;
4734 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4735 u32 portid, u32 seq, int event, unsigned int flags)
4737 struct nlmsghdr *nlh;
4738 u32 preferred, valid;
4740 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4744 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4745 ifa->idev->dev->ifindex);
4747 if (!((ifa->flags&IFA_F_PERMANENT) &&
4748 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4749 preferred = ifa->prefered_lft;
4750 valid = ifa->valid_lft;
4751 if (preferred != INFINITY_LIFE_TIME) {
4752 long tval = (jiffies - ifa->tstamp)/HZ;
4753 if (preferred > tval)
4757 if (valid != INFINITY_LIFE_TIME) {
4765 preferred = INFINITY_LIFE_TIME;
4766 valid = INFINITY_LIFE_TIME;
4769 if (!ipv6_addr_any(&ifa->peer_addr)) {
4770 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4771 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4774 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4777 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4780 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4783 nlmsg_end(skb, nlh);
4787 nlmsg_cancel(skb, nlh);
4791 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4792 u32 portid, u32 seq, int event, u16 flags)
4794 struct nlmsghdr *nlh;
4795 u8 scope = RT_SCOPE_UNIVERSE;
4796 int ifindex = ifmca->idev->dev->ifindex;
4798 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4799 scope = RT_SCOPE_SITE;
4801 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4805 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4806 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4807 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4808 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4809 nlmsg_cancel(skb, nlh);
4813 nlmsg_end(skb, nlh);
4817 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4818 u32 portid, u32 seq, int event, unsigned int flags)
4820 struct nlmsghdr *nlh;
4821 u8 scope = RT_SCOPE_UNIVERSE;
4822 int ifindex = ifaca->aca_idev->dev->ifindex;
4824 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4825 scope = RT_SCOPE_SITE;
4827 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4831 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4832 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4833 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4834 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4835 nlmsg_cancel(skb, nlh);
4839 nlmsg_end(skb, nlh);
4849 /* called with rcu_read_lock() */
4850 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4851 struct netlink_callback *cb, enum addr_type_t type,
4852 int s_ip_idx, int *p_ip_idx)
4854 struct ifmcaddr6 *ifmca;
4855 struct ifacaddr6 *ifaca;
4857 int ip_idx = *p_ip_idx;
4859 read_lock_bh(&idev->lock);
4861 case UNICAST_ADDR: {
4862 struct inet6_ifaddr *ifa;
4864 /* unicast address incl. temp addr */
4865 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4866 if (++ip_idx < s_ip_idx)
4868 err = inet6_fill_ifaddr(skb, ifa,
4869 NETLINK_CB(cb->skb).portid,
4875 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4879 case MULTICAST_ADDR:
4880 /* multicast address */
4881 for (ifmca = idev->mc_list; ifmca;
4882 ifmca = ifmca->next, ip_idx++) {
4883 if (ip_idx < s_ip_idx)
4885 err = inet6_fill_ifmcaddr(skb, ifmca,
4886 NETLINK_CB(cb->skb).portid,
4895 /* anycast address */
4896 for (ifaca = idev->ac_list; ifaca;
4897 ifaca = ifaca->aca_next, ip_idx++) {
4898 if (ip_idx < s_ip_idx)
4900 err = inet6_fill_ifacaddr(skb, ifaca,
4901 NETLINK_CB(cb->skb).portid,
4912 read_unlock_bh(&idev->lock);
4917 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4918 enum addr_type_t type)
4920 struct net *net = sock_net(skb->sk);
4923 int s_idx, s_ip_idx;
4924 struct net_device *dev;
4925 struct inet6_dev *idev;
4926 struct hlist_head *head;
4929 s_idx = idx = cb->args[1];
4930 s_ip_idx = ip_idx = cb->args[2];
4933 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4934 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4936 head = &net->dev_index_head[h];
4937 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4940 if (h > s_h || idx > s_idx)
4943 idev = __in6_dev_get(dev);
4947 if (in6_dump_addrs(idev, skb, cb, type,
4948 s_ip_idx, &ip_idx) < 0)
4958 cb->args[2] = ip_idx;
4963 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4965 enum addr_type_t type = UNICAST_ADDR;
4967 return inet6_dump_addr(skb, cb, type);
4970 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4972 enum addr_type_t type = MULTICAST_ADDR;
4974 return inet6_dump_addr(skb, cb, type);
4978 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4980 enum addr_type_t type = ANYCAST_ADDR;
4982 return inet6_dump_addr(skb, cb, type);
4985 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4986 struct netlink_ext_ack *extack)
4988 struct net *net = sock_net(in_skb->sk);
4989 struct ifaddrmsg *ifm;
4990 struct nlattr *tb[IFA_MAX+1];
4991 struct in6_addr *addr = NULL, *peer;
4992 struct net_device *dev = NULL;
4993 struct inet6_ifaddr *ifa;
4994 struct sk_buff *skb;
4997 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5002 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5006 ifm = nlmsg_data(nlh);
5008 dev = dev_get_by_index(net, ifm->ifa_index);
5010 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5012 err = -EADDRNOTAVAIL;
5016 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5022 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5023 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5025 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5026 WARN_ON(err == -EMSGSIZE);
5030 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5039 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5041 struct sk_buff *skb;
5042 struct net *net = dev_net(ifa->idev->dev);
5045 /* Don't send DELADDR notification for TENTATIVE address,
5046 * since NEWADDR notification is sent only after removing
5047 * TENTATIVE flag, if DAD has not failed.
5049 if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
5050 event == RTM_DELADDR)
5053 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5057 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5059 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5060 WARN_ON(err == -EMSGSIZE);
5064 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5068 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5071 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5072 __s32 *array, int bytes)
5074 BUG_ON(bytes < (DEVCONF_MAX * 4));
5076 memset(array, 0, bytes);
5077 array[DEVCONF_FORWARDING] = cnf->forwarding;
5078 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5079 array[DEVCONF_MTU6] = cnf->mtu6;
5080 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5081 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5082 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5083 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5084 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5085 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5086 jiffies_to_msecs(cnf->rtr_solicit_interval);
5087 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5088 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5089 array[DEVCONF_RTR_SOLICIT_DELAY] =
5090 jiffies_to_msecs(cnf->rtr_solicit_delay);
5091 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5092 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5093 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5094 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5095 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5096 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5097 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5098 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5099 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5100 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5101 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5102 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5103 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5104 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5105 #ifdef CONFIG_IPV6_ROUTER_PREF
5106 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5107 array[DEVCONF_RTR_PROBE_INTERVAL] =
5108 jiffies_to_msecs(cnf->rtr_probe_interval);
5109 #ifdef CONFIG_IPV6_ROUTE_INFO
5110 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5111 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5114 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5115 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5116 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5117 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5118 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5120 #ifdef CONFIG_IPV6_MROUTE
5121 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5123 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5124 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5125 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5126 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5127 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5128 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5129 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5130 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5131 /* we omit DEVCONF_STABLE_SECRET for now */
5132 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5133 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5134 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5135 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5136 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5137 #ifdef CONFIG_IPV6_SEG6_HMAC
5138 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5140 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5141 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5142 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5143 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5146 static inline size_t inet6_ifla6_size(void)
5148 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5149 + nla_total_size(sizeof(struct ifla_cacheinfo))
5150 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5151 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5152 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5153 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5156 static inline size_t inet6_if_nlmsg_size(void)
5158 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5159 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5160 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5161 + nla_total_size(4) /* IFLA_MTU */
5162 + nla_total_size(4) /* IFLA_LINK */
5163 + nla_total_size(1) /* IFLA_OPERSTATE */
5164 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5167 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5171 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5174 /* Use put_unaligned() because stats may not be aligned for u64. */
5175 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5176 for (i = 1; i < ICMP6_MIB_MAX; i++)
5177 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5179 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5182 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5183 int bytes, size_t syncpoff)
5186 u64 buff[IPSTATS_MIB_MAX];
5187 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5191 memset(buff, 0, sizeof(buff));
5192 buff[0] = IPSTATS_MIB_MAX;
5194 for_each_possible_cpu(c) {
5195 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5196 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5199 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5200 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5203 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5207 case IFLA_INET6_STATS:
5208 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5209 offsetof(struct ipstats_mib, syncp));
5211 case IFLA_INET6_ICMP6STATS:
5212 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5217 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5218 u32 ext_filter_mask)
5221 struct ifla_cacheinfo ci;
5223 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5224 goto nla_put_failure;
5225 ci.max_reasm_len = IPV6_MAXPLEN;
5226 ci.tstamp = cstamp_delta(idev->tstamp);
5227 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5228 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5229 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5230 goto nla_put_failure;
5231 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5233 goto nla_put_failure;
5234 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5236 /* XXX - MC not implemented */
5238 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5241 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5243 goto nla_put_failure;
5244 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5246 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5248 goto nla_put_failure;
5249 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5251 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5253 goto nla_put_failure;
5255 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5256 goto nla_put_failure;
5258 read_lock_bh(&idev->lock);
5259 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5260 read_unlock_bh(&idev->lock);
5268 static size_t inet6_get_link_af_size(const struct net_device *dev,
5269 u32 ext_filter_mask)
5271 if (!__in6_dev_get(dev))
5274 return inet6_ifla6_size();
5277 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5278 u32 ext_filter_mask)
5280 struct inet6_dev *idev = __in6_dev_get(dev);
5285 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5291 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5293 struct inet6_ifaddr *ifp;
5294 struct net_device *dev = idev->dev;
5295 bool clear_token, update_rs = false;
5296 struct in6_addr ll_addr;
5302 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5304 if (!ipv6_accept_ra(idev))
5306 if (idev->cnf.rtr_solicits == 0)
5309 write_lock_bh(&idev->lock);
5311 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5312 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5314 write_unlock_bh(&idev->lock);
5316 clear_token = ipv6_addr_any(token);
5320 if (!idev->dead && (idev->if_flags & IF_READY) &&
5321 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5322 IFA_F_OPTIMISTIC)) {
5323 /* If we're not ready, then normal ifup will take care
5324 * of this. Otherwise, we need to request our rs here.
5326 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5331 write_lock_bh(&idev->lock);
5334 idev->if_flags |= IF_RS_SENT;
5335 idev->rs_interval = rfc3315_s14_backoff_init(
5336 idev->cnf.rtr_solicit_interval);
5337 idev->rs_probes = 1;
5338 addrconf_mod_rs_timer(idev, idev->rs_interval);
5341 /* Well, that's kinda nasty ... */
5342 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5343 spin_lock(&ifp->lock);
5344 if (ifp->tokenized) {
5346 ifp->prefered_lft = 0;
5348 spin_unlock(&ifp->lock);
5351 write_unlock_bh(&idev->lock);
5352 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5353 addrconf_verify_rtnl();
5357 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5358 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5359 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5362 static int inet6_validate_link_af(const struct net_device *dev,
5363 const struct nlattr *nla)
5365 struct nlattr *tb[IFLA_INET6_MAX + 1];
5367 if (dev && !__in6_dev_get(dev))
5368 return -EAFNOSUPPORT;
5370 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5374 static int check_addr_gen_mode(int mode)
5376 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5377 mode != IN6_ADDR_GEN_MODE_NONE &&
5378 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5379 mode != IN6_ADDR_GEN_MODE_RANDOM)
5384 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5387 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5388 !idev->cnf.stable_secret.initialized &&
5389 !net->ipv6.devconf_dflt->stable_secret.initialized)
5394 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5397 struct inet6_dev *idev = __in6_dev_get(dev);
5398 struct nlattr *tb[IFLA_INET6_MAX + 1];
5401 return -EAFNOSUPPORT;
5403 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5406 if (tb[IFLA_INET6_TOKEN]) {
5407 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5412 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5413 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5415 if (check_addr_gen_mode(mode) < 0 ||
5416 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5419 idev->cnf.addr_gen_mode = mode;
5426 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5427 u32 portid, u32 seq, int event, unsigned int flags)
5429 struct net_device *dev = idev->dev;
5430 struct ifinfomsg *hdr;
5431 struct nlmsghdr *nlh;
5434 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5438 hdr = nlmsg_data(nlh);
5439 hdr->ifi_family = AF_INET6;
5441 hdr->ifi_type = dev->type;
5442 hdr->ifi_index = dev->ifindex;
5443 hdr->ifi_flags = dev_get_flags(dev);
5444 hdr->ifi_change = 0;
5446 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5448 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5449 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5450 (dev->ifindex != dev_get_iflink(dev) &&
5451 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5452 nla_put_u8(skb, IFLA_OPERSTATE,
5453 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5454 goto nla_put_failure;
5455 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5457 goto nla_put_failure;
5459 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5460 goto nla_put_failure;
5462 nla_nest_end(skb, protoinfo);
5463 nlmsg_end(skb, nlh);
5467 nlmsg_cancel(skb, nlh);
5471 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5473 struct net *net = sock_net(skb->sk);
5476 struct net_device *dev;
5477 struct inet6_dev *idev;
5478 struct hlist_head *head;
5481 s_idx = cb->args[1];
5484 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5486 head = &net->dev_index_head[h];
5487 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5490 idev = __in6_dev_get(dev);
5493 if (inet6_fill_ifinfo(skb, idev,
5494 NETLINK_CB(cb->skb).portid,
5496 RTM_NEWLINK, NLM_F_MULTI) < 0)
5510 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5512 struct sk_buff *skb;
5513 struct net *net = dev_net(idev->dev);
5516 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5520 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5522 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5523 WARN_ON(err == -EMSGSIZE);
5527 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5531 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5534 static inline size_t inet6_prefix_nlmsg_size(void)
5536 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5537 + nla_total_size(sizeof(struct in6_addr))
5538 + nla_total_size(sizeof(struct prefix_cacheinfo));
5541 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5542 struct prefix_info *pinfo, u32 portid, u32 seq,
5543 int event, unsigned int flags)
5545 struct prefixmsg *pmsg;
5546 struct nlmsghdr *nlh;
5547 struct prefix_cacheinfo ci;
5549 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5553 pmsg = nlmsg_data(nlh);
5554 pmsg->prefix_family = AF_INET6;
5555 pmsg->prefix_pad1 = 0;
5556 pmsg->prefix_pad2 = 0;
5557 pmsg->prefix_ifindex = idev->dev->ifindex;
5558 pmsg->prefix_len = pinfo->prefix_len;
5559 pmsg->prefix_type = pinfo->type;
5560 pmsg->prefix_pad3 = 0;
5561 pmsg->prefix_flags = 0;
5563 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5564 if (pinfo->autoconf)
5565 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5567 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5568 goto nla_put_failure;
5569 ci.preferred_time = ntohl(pinfo->prefered);
5570 ci.valid_time = ntohl(pinfo->valid);
5571 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5572 goto nla_put_failure;
5573 nlmsg_end(skb, nlh);
5577 nlmsg_cancel(skb, nlh);
5581 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5582 struct prefix_info *pinfo)
5584 struct sk_buff *skb;
5585 struct net *net = dev_net(idev->dev);
5588 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5592 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5594 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5595 WARN_ON(err == -EMSGSIZE);
5599 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5603 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5606 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5608 struct net *net = dev_net(ifp->idev->dev);
5613 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5618 * If the address was optimistic
5619 * we inserted the route at the start of
5620 * our DAD process, so we don't need
5623 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5624 ip6_ins_rt(ifp->rt);
5625 if (ifp->idev->cnf.forwarding)
5626 addrconf_join_anycast(ifp);
5627 if (!ipv6_addr_any(&ifp->peer_addr))
5628 addrconf_prefix_route(&ifp->peer_addr, 128,
5629 ifp->idev->dev, 0, 0);
5632 if (ifp->idev->cnf.forwarding)
5633 addrconf_leave_anycast(ifp);
5634 addrconf_leave_solict(ifp->idev, &ifp->addr);
5635 if (!ipv6_addr_any(&ifp->peer_addr)) {
5636 struct rt6_info *rt;
5638 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5639 ifp->idev->dev, 0, 0);
5644 if (dst_hold_safe(&ifp->rt->dst))
5645 ip6_del_rt(ifp->rt);
5647 rt_genid_bump_ipv6(net);
5650 atomic_inc(&net->ipv6.dev_addr_genid);
5653 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5656 if (likely(ifp->idev->dead == 0))
5657 __ipv6_ifa_notify(event, ifp);
5658 rcu_read_unlock_bh();
5661 #ifdef CONFIG_SYSCTL
5664 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5665 void __user *buffer, size_t *lenp, loff_t *ppos)
5667 int *valp = ctl->data;
5670 struct ctl_table lctl;
5674 * ctl->data points to idev->cnf.forwarding, we should
5675 * not modify it until we get the rtnl lock.
5680 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5683 ret = addrconf_fixup_forwarding(ctl, valp, val);
5690 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5691 void __user *buffer, size_t *lenp, loff_t *ppos)
5693 struct inet6_dev *idev = ctl->extra1;
5694 int min_mtu = IPV6_MIN_MTU;
5695 struct ctl_table lctl;
5698 lctl.extra1 = &min_mtu;
5699 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5701 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5704 static void dev_disable_change(struct inet6_dev *idev)
5706 struct netdev_notifier_info info;
5708 if (!idev || !idev->dev)
5711 netdev_notifier_info_init(&info, idev->dev);
5712 if (idev->cnf.disable_ipv6)
5713 addrconf_notify(NULL, NETDEV_DOWN, &info);
5715 addrconf_notify(NULL, NETDEV_UP, &info);
5718 static void addrconf_disable_change(struct net *net, __s32 newf)
5720 struct net_device *dev;
5721 struct inet6_dev *idev;
5723 for_each_netdev(net, dev) {
5724 idev = __in6_dev_get(dev);
5726 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5727 idev->cnf.disable_ipv6 = newf;
5729 dev_disable_change(idev);
5734 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5739 if (!rtnl_trylock())
5740 return restart_syscall();
5742 net = (struct net *)table->extra2;
5746 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5751 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5752 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5753 addrconf_disable_change(net, newf);
5754 } else if ((!newf) ^ (!old))
5755 dev_disable_change((struct inet6_dev *)table->extra1);
5762 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5763 void __user *buffer, size_t *lenp, loff_t *ppos)
5765 int *valp = ctl->data;
5768 struct ctl_table lctl;
5772 * ctl->data points to idev->cnf.disable_ipv6, we should
5773 * not modify it until we get the rtnl lock.
5778 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5781 ret = addrconf_disable_ipv6(ctl, valp, val);
5788 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5789 void __user *buffer, size_t *lenp, loff_t *ppos)
5791 int *valp = ctl->data;
5796 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5799 if (write && old != new) {
5800 struct net *net = ctl->extra2;
5802 if (!rtnl_trylock())
5803 return restart_syscall();
5805 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5806 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5807 NETCONFA_PROXY_NEIGH,
5808 NETCONFA_IFINDEX_DEFAULT,
5809 net->ipv6.devconf_dflt);
5810 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5811 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5812 NETCONFA_PROXY_NEIGH,
5813 NETCONFA_IFINDEX_ALL,
5814 net->ipv6.devconf_all);
5816 struct inet6_dev *idev = ctl->extra1;
5818 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5819 NETCONFA_PROXY_NEIGH,
5829 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5830 void __user *buffer, size_t *lenp,
5835 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5836 struct net *net = (struct net *)ctl->extra2;
5838 if (!rtnl_trylock())
5839 return restart_syscall();
5841 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5844 new_val = *((int *)ctl->data);
5846 if (check_addr_gen_mode(new_val) < 0) {
5851 /* request for default */
5852 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5853 ipv6_devconf_dflt.addr_gen_mode = new_val;
5855 /* request for individual net device */
5860 if (check_stable_privacy(idev, net, new_val) < 0) {
5865 if (idev->cnf.addr_gen_mode != new_val) {
5866 idev->cnf.addr_gen_mode = new_val;
5867 addrconf_dev_config(idev->dev);
5878 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5879 void __user *buffer, size_t *lenp,
5883 struct in6_addr addr;
5884 char str[IPV6_MAX_STRLEN];
5885 struct ctl_table lctl = *ctl;
5886 struct net *net = ctl->extra2;
5887 struct ipv6_stable_secret *secret = ctl->data;
5889 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5892 lctl.maxlen = IPV6_MAX_STRLEN;
5895 if (!rtnl_trylock())
5896 return restart_syscall();
5898 if (!write && !secret->initialized) {
5903 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5904 if (err >= sizeof(str)) {
5909 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5913 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5918 secret->initialized = true;
5919 secret->secret = addr;
5921 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5922 struct net_device *dev;
5924 for_each_netdev(net, dev) {
5925 struct inet6_dev *idev = __in6_dev_get(dev);
5928 idev->cnf.addr_gen_mode =
5929 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5933 struct inet6_dev *idev = ctl->extra1;
5935 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5945 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5947 void __user *buffer,
5951 int *valp = ctl->data;
5954 struct ctl_table lctl;
5957 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5958 * we should not modify it until we get the rtnl lock.
5963 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5966 ret = addrconf_fixup_linkdown(ctl, valp, val);
5973 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5977 rt->dst.flags |= DST_NOPOLICY;
5979 rt->dst.flags &= ~DST_NOPOLICY;
5984 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5986 struct inet6_ifaddr *ifa;
5988 read_lock_bh(&idev->lock);
5989 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5990 spin_lock(&ifa->lock);
5992 struct rt6_info *rt = ifa->rt;
5996 addrconf_set_nopolicy(ifa->rt, val);
5997 if (rt->rt6i_pcpu) {
5998 for_each_possible_cpu(cpu) {
5999 struct rt6_info **rtp;
6001 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6002 addrconf_set_nopolicy(*rtp, val);
6007 spin_unlock(&ifa->lock);
6009 read_unlock_bh(&idev->lock);
6013 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6015 struct inet6_dev *idev;
6018 if (!rtnl_trylock())
6019 return restart_syscall();
6023 net = (struct net *)ctl->extra2;
6024 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6029 if (valp == &net->ipv6.devconf_all->disable_policy) {
6030 struct net_device *dev;
6032 for_each_netdev(net, dev) {
6033 idev = __in6_dev_get(dev);
6035 addrconf_disable_policy_idev(idev, val);
6038 idev = (struct inet6_dev *)ctl->extra1;
6039 addrconf_disable_policy_idev(idev, val);
6047 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6048 void __user *buffer, size_t *lenp,
6051 int *valp = ctl->data;
6054 struct ctl_table lctl;
6059 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6061 if (write && (*valp != val))
6062 ret = addrconf_disable_policy(ctl, valp, val);
6070 static int minus_one = -1;
6071 static const int zero = 0;
6072 static const int one = 1;
6073 static const int two_five_five = 255;
6075 static const struct ctl_table addrconf_sysctl[] = {
6077 .procname = "forwarding",
6078 .data = &ipv6_devconf.forwarding,
6079 .maxlen = sizeof(int),
6081 .proc_handler = addrconf_sysctl_forward,
6084 .procname = "hop_limit",
6085 .data = &ipv6_devconf.hop_limit,
6086 .maxlen = sizeof(int),
6088 .proc_handler = proc_dointvec_minmax,
6089 .extra1 = (void *)&one,
6090 .extra2 = (void *)&two_five_five,
6094 .data = &ipv6_devconf.mtu6,
6095 .maxlen = sizeof(int),
6097 .proc_handler = addrconf_sysctl_mtu,
6100 .procname = "accept_ra",
6101 .data = &ipv6_devconf.accept_ra,
6102 .maxlen = sizeof(int),
6104 .proc_handler = proc_dointvec,
6107 .procname = "accept_redirects",
6108 .data = &ipv6_devconf.accept_redirects,
6109 .maxlen = sizeof(int),
6111 .proc_handler = proc_dointvec,
6114 .procname = "autoconf",
6115 .data = &ipv6_devconf.autoconf,
6116 .maxlen = sizeof(int),
6118 .proc_handler = proc_dointvec,
6121 .procname = "dad_transmits",
6122 .data = &ipv6_devconf.dad_transmits,
6123 .maxlen = sizeof(int),
6125 .proc_handler = proc_dointvec,
6128 .procname = "router_solicitations",
6129 .data = &ipv6_devconf.rtr_solicits,
6130 .maxlen = sizeof(int),
6132 .proc_handler = proc_dointvec_minmax,
6133 .extra1 = &minus_one,
6136 .procname = "router_solicitation_interval",
6137 .data = &ipv6_devconf.rtr_solicit_interval,
6138 .maxlen = sizeof(int),
6140 .proc_handler = proc_dointvec_jiffies,
6143 .procname = "router_solicitation_max_interval",
6144 .data = &ipv6_devconf.rtr_solicit_max_interval,
6145 .maxlen = sizeof(int),
6147 .proc_handler = proc_dointvec_jiffies,
6150 .procname = "router_solicitation_delay",
6151 .data = &ipv6_devconf.rtr_solicit_delay,
6152 .maxlen = sizeof(int),
6154 .proc_handler = proc_dointvec_jiffies,
6157 .procname = "force_mld_version",
6158 .data = &ipv6_devconf.force_mld_version,
6159 .maxlen = sizeof(int),
6161 .proc_handler = proc_dointvec,
6164 .procname = "mldv1_unsolicited_report_interval",
6166 &ipv6_devconf.mldv1_unsolicited_report_interval,
6167 .maxlen = sizeof(int),
6169 .proc_handler = proc_dointvec_ms_jiffies,
6172 .procname = "mldv2_unsolicited_report_interval",
6174 &ipv6_devconf.mldv2_unsolicited_report_interval,
6175 .maxlen = sizeof(int),
6177 .proc_handler = proc_dointvec_ms_jiffies,
6180 .procname = "use_tempaddr",
6181 .data = &ipv6_devconf.use_tempaddr,
6182 .maxlen = sizeof(int),
6184 .proc_handler = proc_dointvec,
6187 .procname = "temp_valid_lft",
6188 .data = &ipv6_devconf.temp_valid_lft,
6189 .maxlen = sizeof(int),
6191 .proc_handler = proc_dointvec,
6194 .procname = "temp_prefered_lft",
6195 .data = &ipv6_devconf.temp_prefered_lft,
6196 .maxlen = sizeof(int),
6198 .proc_handler = proc_dointvec,
6201 .procname = "regen_max_retry",
6202 .data = &ipv6_devconf.regen_max_retry,
6203 .maxlen = sizeof(int),
6205 .proc_handler = proc_dointvec,
6208 .procname = "max_desync_factor",
6209 .data = &ipv6_devconf.max_desync_factor,
6210 .maxlen = sizeof(int),
6212 .proc_handler = proc_dointvec,
6215 .procname = "max_addresses",
6216 .data = &ipv6_devconf.max_addresses,
6217 .maxlen = sizeof(int),
6219 .proc_handler = proc_dointvec,
6222 .procname = "accept_ra_defrtr",
6223 .data = &ipv6_devconf.accept_ra_defrtr,
6224 .maxlen = sizeof(int),
6226 .proc_handler = proc_dointvec,
6229 .procname = "accept_ra_min_hop_limit",
6230 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6231 .maxlen = sizeof(int),
6233 .proc_handler = proc_dointvec,
6236 .procname = "accept_ra_pinfo",
6237 .data = &ipv6_devconf.accept_ra_pinfo,
6238 .maxlen = sizeof(int),
6240 .proc_handler = proc_dointvec,
6242 #ifdef CONFIG_IPV6_ROUTER_PREF
6244 .procname = "accept_ra_rtr_pref",
6245 .data = &ipv6_devconf.accept_ra_rtr_pref,
6246 .maxlen = sizeof(int),
6248 .proc_handler = proc_dointvec,
6251 .procname = "router_probe_interval",
6252 .data = &ipv6_devconf.rtr_probe_interval,
6253 .maxlen = sizeof(int),
6255 .proc_handler = proc_dointvec_jiffies,
6257 #ifdef CONFIG_IPV6_ROUTE_INFO
6259 .procname = "accept_ra_rt_info_min_plen",
6260 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6261 .maxlen = sizeof(int),
6263 .proc_handler = proc_dointvec,
6266 .procname = "accept_ra_rt_info_max_plen",
6267 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6268 .maxlen = sizeof(int),
6270 .proc_handler = proc_dointvec,
6275 .procname = "proxy_ndp",
6276 .data = &ipv6_devconf.proxy_ndp,
6277 .maxlen = sizeof(int),
6279 .proc_handler = addrconf_sysctl_proxy_ndp,
6282 .procname = "accept_source_route",
6283 .data = &ipv6_devconf.accept_source_route,
6284 .maxlen = sizeof(int),
6286 .proc_handler = proc_dointvec,
6288 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6290 .procname = "optimistic_dad",
6291 .data = &ipv6_devconf.optimistic_dad,
6292 .maxlen = sizeof(int),
6294 .proc_handler = proc_dointvec,
6297 .procname = "use_optimistic",
6298 .data = &ipv6_devconf.use_optimistic,
6299 .maxlen = sizeof(int),
6301 .proc_handler = proc_dointvec,
6304 #ifdef CONFIG_IPV6_MROUTE
6306 .procname = "mc_forwarding",
6307 .data = &ipv6_devconf.mc_forwarding,
6308 .maxlen = sizeof(int),
6310 .proc_handler = proc_dointvec,
6314 .procname = "disable_ipv6",
6315 .data = &ipv6_devconf.disable_ipv6,
6316 .maxlen = sizeof(int),
6318 .proc_handler = addrconf_sysctl_disable,
6321 .procname = "accept_dad",
6322 .data = &ipv6_devconf.accept_dad,
6323 .maxlen = sizeof(int),
6325 .proc_handler = proc_dointvec,
6328 .procname = "force_tllao",
6329 .data = &ipv6_devconf.force_tllao,
6330 .maxlen = sizeof(int),
6332 .proc_handler = proc_dointvec
6335 .procname = "ndisc_notify",
6336 .data = &ipv6_devconf.ndisc_notify,
6337 .maxlen = sizeof(int),
6339 .proc_handler = proc_dointvec
6342 .procname = "suppress_frag_ndisc",
6343 .data = &ipv6_devconf.suppress_frag_ndisc,
6344 .maxlen = sizeof(int),
6346 .proc_handler = proc_dointvec
6349 .procname = "accept_ra_from_local",
6350 .data = &ipv6_devconf.accept_ra_from_local,
6351 .maxlen = sizeof(int),
6353 .proc_handler = proc_dointvec,
6356 .procname = "accept_ra_mtu",
6357 .data = &ipv6_devconf.accept_ra_mtu,
6358 .maxlen = sizeof(int),
6360 .proc_handler = proc_dointvec,
6363 .procname = "stable_secret",
6364 .data = &ipv6_devconf.stable_secret,
6365 .maxlen = IPV6_MAX_STRLEN,
6367 .proc_handler = addrconf_sysctl_stable_secret,
6370 .procname = "use_oif_addrs_only",
6371 .data = &ipv6_devconf.use_oif_addrs_only,
6372 .maxlen = sizeof(int),
6374 .proc_handler = proc_dointvec,
6377 .procname = "ignore_routes_with_linkdown",
6378 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6379 .maxlen = sizeof(int),
6381 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6384 .procname = "drop_unicast_in_l2_multicast",
6385 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6386 .maxlen = sizeof(int),
6388 .proc_handler = proc_dointvec,
6391 .procname = "drop_unsolicited_na",
6392 .data = &ipv6_devconf.drop_unsolicited_na,
6393 .maxlen = sizeof(int),
6395 .proc_handler = proc_dointvec,
6398 .procname = "keep_addr_on_down",
6399 .data = &ipv6_devconf.keep_addr_on_down,
6400 .maxlen = sizeof(int),
6402 .proc_handler = proc_dointvec,
6406 .procname = "seg6_enabled",
6407 .data = &ipv6_devconf.seg6_enabled,
6408 .maxlen = sizeof(int),
6410 .proc_handler = proc_dointvec,
6412 #ifdef CONFIG_IPV6_SEG6_HMAC
6414 .procname = "seg6_require_hmac",
6415 .data = &ipv6_devconf.seg6_require_hmac,
6416 .maxlen = sizeof(int),
6418 .proc_handler = proc_dointvec,
6422 .procname = "enhanced_dad",
6423 .data = &ipv6_devconf.enhanced_dad,
6424 .maxlen = sizeof(int),
6426 .proc_handler = proc_dointvec,
6429 .procname = "addr_gen_mode",
6430 .data = &ipv6_devconf.addr_gen_mode,
6431 .maxlen = sizeof(int),
6433 .proc_handler = addrconf_sysctl_addr_gen_mode,
6436 .procname = "disable_policy",
6437 .data = &ipv6_devconf.disable_policy,
6438 .maxlen = sizeof(int),
6440 .proc_handler = addrconf_sysctl_disable_policy,
6443 .procname = "ndisc_tclass",
6444 .data = &ipv6_devconf.ndisc_tclass,
6445 .maxlen = sizeof(int),
6447 .proc_handler = proc_dointvec_minmax,
6448 .extra1 = (void *)&zero,
6449 .extra2 = (void *)&two_five_five,
6456 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6457 struct inet6_dev *idev, struct ipv6_devconf *p)
6460 struct ctl_table *table;
6461 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6463 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6467 for (i = 0; table[i].data; i++) {
6468 table[i].data += (char *)p - (char *)&ipv6_devconf;
6469 /* If one of these is already set, then it is not safe to
6470 * overwrite either of them: this makes proc_dointvec_minmax
6473 if (!table[i].extra1 && !table[i].extra2) {
6474 table[i].extra1 = idev; /* embedded; no ref */
6475 table[i].extra2 = net;
6479 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6481 p->sysctl_header = register_net_sysctl(net, path, table);
6482 if (!p->sysctl_header)
6485 if (!strcmp(dev_name, "all"))
6486 ifindex = NETCONFA_IFINDEX_ALL;
6487 else if (!strcmp(dev_name, "default"))
6488 ifindex = NETCONFA_IFINDEX_DEFAULT;
6490 ifindex = idev->dev->ifindex;
6491 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6501 static void __addrconf_sysctl_unregister(struct net *net,
6502 struct ipv6_devconf *p, int ifindex)
6504 struct ctl_table *table;
6506 if (!p->sysctl_header)
6509 table = p->sysctl_header->ctl_table_arg;
6510 unregister_net_sysctl_table(p->sysctl_header);
6511 p->sysctl_header = NULL;
6514 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6517 static int addrconf_sysctl_register(struct inet6_dev *idev)
6521 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6524 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6525 &ndisc_ifinfo_sysctl_change);
6528 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6531 neigh_sysctl_unregister(idev->nd_parms);
6536 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6538 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6539 idev->dev->ifindex);
6540 neigh_sysctl_unregister(idev->nd_parms);
6546 static int __net_init addrconf_init_net(struct net *net)
6549 struct ipv6_devconf *all, *dflt;
6551 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6555 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6557 goto err_alloc_dflt;
6559 /* these will be inherited by all namespaces */
6560 dflt->autoconf = ipv6_defaults.autoconf;
6561 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6563 dflt->stable_secret.initialized = false;
6564 all->stable_secret.initialized = false;
6566 net->ipv6.devconf_all = all;
6567 net->ipv6.devconf_dflt = dflt;
6569 #ifdef CONFIG_SYSCTL
6570 err = __addrconf_sysctl_register(net, "all", NULL, all);
6574 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6580 #ifdef CONFIG_SYSCTL
6582 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6592 static void __net_exit addrconf_exit_net(struct net *net)
6594 #ifdef CONFIG_SYSCTL
6595 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6596 NETCONFA_IFINDEX_DEFAULT);
6597 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6598 NETCONFA_IFINDEX_ALL);
6600 kfree(net->ipv6.devconf_dflt);
6601 kfree(net->ipv6.devconf_all);
6604 static struct pernet_operations addrconf_ops = {
6605 .init = addrconf_init_net,
6606 .exit = addrconf_exit_net,
6610 static struct rtnl_af_ops inet6_ops __read_mostly = {
6612 .fill_link_af = inet6_fill_link_af,
6613 .get_link_af_size = inet6_get_link_af_size,
6614 .validate_link_af = inet6_validate_link_af,
6615 .set_link_af = inet6_set_link_af,
6619 * Init / cleanup code
6622 int __init addrconf_init(void)
6624 struct inet6_dev *idev;
6627 err = ipv6_addr_label_init();
6629 pr_crit("%s: cannot initialize default policy table: %d\n",
6634 err = register_pernet_subsys(&addrconf_ops);
6638 addrconf_wq = create_workqueue("ipv6_addrconf");
6644 /* The addrconf netdev notifier requires that loopback_dev
6645 * has it's ipv6 private information allocated and setup
6646 * before it can bring up and give link-local addresses
6647 * to other devices which are up.
6649 * Unfortunately, loopback_dev is not necessarily the first
6650 * entry in the global dev_base list of net devices. In fact,
6651 * it is likely to be the very last entry on that list.
6652 * So this causes the notifier registry below to try and
6653 * give link-local addresses to all devices besides loopback_dev
6654 * first, then loopback_dev, which cases all the non-loopback_dev
6655 * devices to fail to get a link-local address.
6657 * So, as a temporary fix, allocate the ipv6 structure for
6658 * loopback_dev first by hand.
6659 * Longer term, all of the dependencies ipv6 has upon the loopback
6660 * device and it being up should be removed.
6663 idev = ipv6_add_dev(init_net.loopback_dev);
6666 err = PTR_ERR(idev);
6670 ip6_route_init_special_entries();
6672 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6673 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6675 register_netdevice_notifier(&ipv6_dev_notf);
6679 rtnl_af_register(&inet6_ops);
6681 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6682 NULL, inet6_dump_ifinfo, 0);
6686 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6687 inet6_rtm_newaddr, NULL, 0);
6690 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6691 inet6_rtm_deladdr, NULL, 0);
6694 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6695 inet6_rtm_getaddr, inet6_dump_ifaddr,
6696 RTNL_FLAG_DOIT_UNLOCKED);
6699 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6700 NULL, inet6_dump_ifmcaddr, 0);
6703 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6704 NULL, inet6_dump_ifacaddr, 0);
6707 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6708 inet6_netconf_get_devconf,
6709 inet6_netconf_dump_devconf,
6710 RTNL_FLAG_DOIT_UNLOCKED);
6713 err = ipv6_addr_label_rtnl_register();
6719 rtnl_unregister_all(PF_INET6);
6720 rtnl_af_unregister(&inet6_ops);
6721 unregister_netdevice_notifier(&ipv6_dev_notf);
6723 destroy_workqueue(addrconf_wq);
6725 unregister_pernet_subsys(&addrconf_ops);
6727 ipv6_addr_label_cleanup();
6732 void addrconf_cleanup(void)
6734 struct net_device *dev;
6737 unregister_netdevice_notifier(&ipv6_dev_notf);
6738 unregister_pernet_subsys(&addrconf_ops);
6739 ipv6_addr_label_cleanup();
6741 rtnl_af_unregister(&inet6_ops);
6745 /* clean dev list */
6746 for_each_netdev(&init_net, dev) {
6747 if (__in6_dev_get(dev) == NULL)
6749 addrconf_ifdown(dev, 1);
6751 addrconf_ifdown(init_net.loopback_dev, 2);
6756 spin_lock_bh(&addrconf_hash_lock);
6757 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6758 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6759 spin_unlock_bh(&addrconf_hash_lock);
6760 cancel_delayed_work(&addr_chk_work);
6763 destroy_workqueue(addrconf_wq);