1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IPv6 Address [auto]configuration
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
14 * Janos Farkas : delete timer on ifdown
15 * <chexum@bankinf.banki.hu>
16 * Andi Kleen : kill double kfree on module
18 * Maciej W. Rozycki : FDDI support
19 * sekiya@USAGI : Don't send too many RS
21 * yoshfuji@USAGI : Fixed interval between DAD
23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
24 * address validation timer.
25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
28 * address on a same interface.
29 * YOSHIFUJI Hideaki @USAGI : ARCnet support
30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
32 * YOSHIFUJI Hideaki @USAGI : improved source address
33 * selection; consider scope,
37 #define pr_fmt(fmt) "IPv6: " fmt
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
58 #include <linux/sysctl.h>
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
66 #include <net/net_namespace.h>
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
93 #define INFINITY_LIFE_TIME 0xFFFFFFFF
95 #define IPV6_MAX_STRLEN \
96 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98 static inline u32 cstamp_delta(unsigned long cstamp)
100 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
106 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
107 do_div(tmp, 1000000);
111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
114 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
115 do_div(tmp, 1000000);
116 if ((s32)tmp > mrt) {
117 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
118 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
119 do_div(tmp, 1000000);
125 static int addrconf_sysctl_register(struct inet6_dev *idev);
126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
138 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
141 static int ipv6_count_addresses(const struct inet6_dev *idev);
142 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 const struct inet6_dev *idev);
146 #define IN6_ADDR_HSIZE_SHIFT 8
147 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
149 * Configured unicast address hash table
151 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
152 static DEFINE_SPINLOCK(addrconf_hash_lock);
154 static void addrconf_verify(void);
155 static void addrconf_verify_rtnl(void);
156 static void addrconf_verify_work(struct work_struct *);
158 static struct workqueue_struct *addrconf_wq;
159 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
161 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
162 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
164 static void addrconf_type_change(struct net_device *dev,
165 unsigned long event);
166 static int addrconf_ifdown(struct net_device *dev, bool unregister);
168 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
170 const struct net_device *dev,
171 u32 flags, u32 noflags,
174 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
175 static void addrconf_dad_work(struct work_struct *w);
176 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
178 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
179 static void addrconf_rs_timer(struct timer_list *t);
180 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
181 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
183 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
184 struct prefix_info *pinfo);
186 static struct ipv6_devconf ipv6_devconf __read_mostly = {
188 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
189 .mtu6 = IPV6_MIN_MTU,
191 .accept_redirects = 1,
193 .force_mld_version = 0,
194 .mldv1_unsolicited_report_interval = 10 * HZ,
195 .mldv2_unsolicited_report_interval = HZ,
197 .rtr_solicits = MAX_RTR_SOLICITATIONS,
198 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
199 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
200 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
202 .temp_valid_lft = TEMP_VALID_LIFETIME,
203 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
204 .regen_max_retry = REGEN_MAX_RETRY,
205 .max_desync_factor = MAX_DESYNC_FACTOR,
206 .max_addresses = IPV6_MAX_ADDRESSES,
207 .accept_ra_defrtr = 1,
208 .ra_defrtr_metric = IP6_RT_PRIO_USER,
209 .accept_ra_from_local = 0,
210 .accept_ra_min_hop_limit= 1,
211 .accept_ra_pinfo = 1,
212 #ifdef CONFIG_IPV6_ROUTER_PREF
213 .accept_ra_rtr_pref = 1,
214 .rtr_probe_interval = 60 * HZ,
215 #ifdef CONFIG_IPV6_ROUTE_INFO
216 .accept_ra_rt_info_min_plen = 0,
217 .accept_ra_rt_info_max_plen = 0,
221 .accept_source_route = 0, /* we do not accept RH0 by default. */
224 .suppress_frag_ndisc = 1,
227 .initialized = false,
229 .use_oif_addrs_only = 0,
230 .ignore_routes_with_linkdown = 0,
231 .keep_addr_on_down = 0,
233 #ifdef CONFIG_IPV6_SEG6_HMAC
234 .seg6_require_hmac = 0,
237 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
239 .rpl_seg_enabled = 0,
242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
244 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
245 .mtu6 = IPV6_MIN_MTU,
247 .accept_redirects = 1,
249 .force_mld_version = 0,
250 .mldv1_unsolicited_report_interval = 10 * HZ,
251 .mldv2_unsolicited_report_interval = HZ,
253 .rtr_solicits = MAX_RTR_SOLICITATIONS,
254 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
255 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
256 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
258 .temp_valid_lft = TEMP_VALID_LIFETIME,
259 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
260 .regen_max_retry = REGEN_MAX_RETRY,
261 .max_desync_factor = MAX_DESYNC_FACTOR,
262 .max_addresses = IPV6_MAX_ADDRESSES,
263 .accept_ra_defrtr = 1,
264 .ra_defrtr_metric = IP6_RT_PRIO_USER,
265 .accept_ra_from_local = 0,
266 .accept_ra_min_hop_limit= 1,
267 .accept_ra_pinfo = 1,
268 #ifdef CONFIG_IPV6_ROUTER_PREF
269 .accept_ra_rtr_pref = 1,
270 .rtr_probe_interval = 60 * HZ,
271 #ifdef CONFIG_IPV6_ROUTE_INFO
272 .accept_ra_rt_info_min_plen = 0,
273 .accept_ra_rt_info_max_plen = 0,
277 .accept_source_route = 0, /* we do not accept RH0 by default. */
280 .suppress_frag_ndisc = 1,
283 .initialized = false,
285 .use_oif_addrs_only = 0,
286 .ignore_routes_with_linkdown = 0,
287 .keep_addr_on_down = 0,
289 #ifdef CONFIG_IPV6_SEG6_HMAC
290 .seg6_require_hmac = 0,
293 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
295 .rpl_seg_enabled = 0,
298 /* Check if link is ready: is it up and is a valid qdisc available */
299 static inline bool addrconf_link_ready(const struct net_device *dev)
301 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
306 if (del_timer(&idev->rs_timer))
310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
312 if (cancel_delayed_work(&ifp->dad_work))
316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
319 if (!timer_pending(&idev->rs_timer))
321 mod_timer(&idev->rs_timer, jiffies + when);
324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
328 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
332 static int snmp6_alloc_dev(struct inet6_dev *idev)
336 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
337 if (!idev->stats.ipv6)
340 for_each_possible_cpu(i) {
341 struct ipstats_mib *addrconf_stats;
342 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
343 u64_stats_init(&addrconf_stats->syncp);
347 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
349 if (!idev->stats.icmpv6dev)
351 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
353 if (!idev->stats.icmpv6msgdev)
359 kfree(idev->stats.icmpv6dev);
361 free_percpu(idev->stats.ipv6);
366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
368 struct inet6_dev *ndev;
373 if (dev->mtu < IPV6_MIN_MTU)
374 return ERR_PTR(-EINVAL);
376 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
380 rwlock_init(&ndev->lock);
382 INIT_LIST_HEAD(&ndev->addr_list);
383 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
384 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
386 if (ndev->cnf.stable_secret.initialized)
387 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
389 ndev->cnf.mtu6 = dev->mtu;
390 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
391 if (!ndev->nd_parms) {
395 if (ndev->cnf.forwarding)
396 dev_disable_lro(dev);
397 /* We refer to the device */
400 if (snmp6_alloc_dev(ndev) < 0) {
401 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
403 neigh_parms_release(&nd_tbl, ndev->nd_parms);
409 if (snmp6_register_dev(ndev) < 0) {
410 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
411 __func__, dev->name);
415 /* One reference from device. */
416 refcount_set(&ndev->refcnt, 1);
418 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
419 ndev->cnf.accept_dad = -1;
421 #if IS_ENABLED(CONFIG_IPV6_SIT)
422 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
423 pr_info("%s: Disabled Multicast RS\n", dev->name);
424 ndev->cnf.rtr_solicits = 0;
428 INIT_LIST_HEAD(&ndev->tempaddr_list);
429 ndev->desync_factor = U32_MAX;
430 if ((dev->flags&IFF_LOOPBACK) ||
431 dev->type == ARPHRD_TUNNEL ||
432 dev->type == ARPHRD_TUNNEL6 ||
433 dev->type == ARPHRD_SIT ||
434 dev->type == ARPHRD_NONE) {
435 ndev->cnf.use_tempaddr = -1;
438 ndev->token = in6addr_any;
440 if (netif_running(dev) && addrconf_link_ready(dev))
441 ndev->if_flags |= IF_READY;
443 ipv6_mc_init_dev(ndev);
444 ndev->tstamp = jiffies;
445 err = addrconf_sysctl_register(ndev);
447 ipv6_mc_destroy_dev(ndev);
448 snmp6_unregister_dev(ndev);
451 /* protected by rtnl_lock */
452 rcu_assign_pointer(dev->ip6_ptr, ndev);
454 /* Join interface-local all-node multicast group */
455 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
457 /* Join all-node multicast group */
458 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
460 /* Join all-router multicast group if forwarding is set */
461 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
462 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
467 neigh_parms_release(&nd_tbl, ndev->nd_parms);
469 in6_dev_finish_destroy(ndev);
473 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
475 struct inet6_dev *idev;
479 idev = __in6_dev_get(dev);
481 idev = ipv6_add_dev(dev);
486 if (dev->flags&IFF_UP)
491 static int inet6_netconf_msgsize_devconf(int type)
493 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
494 + nla_total_size(4); /* NETCONFA_IFINDEX */
497 if (type == NETCONFA_ALL)
500 if (all || type == NETCONFA_FORWARDING)
501 size += nla_total_size(4);
502 #ifdef CONFIG_IPV6_MROUTE
503 if (all || type == NETCONFA_MC_FORWARDING)
504 size += nla_total_size(4);
506 if (all || type == NETCONFA_PROXY_NEIGH)
507 size += nla_total_size(4);
509 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
510 size += nla_total_size(4);
515 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
516 struct ipv6_devconf *devconf, u32 portid,
517 u32 seq, int event, unsigned int flags,
520 struct nlmsghdr *nlh;
521 struct netconfmsg *ncm;
524 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
529 if (type == NETCONFA_ALL)
532 ncm = nlmsg_data(nlh);
533 ncm->ncm_family = AF_INET6;
535 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
536 goto nla_put_failure;
541 if ((all || type == NETCONFA_FORWARDING) &&
542 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
543 goto nla_put_failure;
544 #ifdef CONFIG_IPV6_MROUTE
545 if ((all || type == NETCONFA_MC_FORWARDING) &&
546 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
547 devconf->mc_forwarding) < 0)
548 goto nla_put_failure;
550 if ((all || type == NETCONFA_PROXY_NEIGH) &&
551 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
552 goto nla_put_failure;
554 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
555 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
556 devconf->ignore_routes_with_linkdown) < 0)
557 goto nla_put_failure;
564 nlmsg_cancel(skb, nlh);
568 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
569 int ifindex, struct ipv6_devconf *devconf)
574 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
578 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
581 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
582 WARN_ON(err == -EMSGSIZE);
586 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
589 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
592 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
593 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
594 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
595 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
596 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
599 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
600 const struct nlmsghdr *nlh,
602 struct netlink_ext_ack *extack)
606 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
607 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
611 if (!netlink_strict_get_check(skb))
612 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
614 devconf_ipv6_policy, extack);
616 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
618 devconf_ipv6_policy, extack);
622 for (i = 0; i <= NETCONFA_MAX; i++) {
627 case NETCONFA_IFINDEX:
630 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
638 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
639 struct nlmsghdr *nlh,
640 struct netlink_ext_ack *extack)
642 struct net *net = sock_net(in_skb->sk);
643 struct nlattr *tb[NETCONFA_MAX+1];
644 struct inet6_dev *in6_dev = NULL;
645 struct net_device *dev = NULL;
647 struct ipv6_devconf *devconf;
651 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
655 if (!tb[NETCONFA_IFINDEX])
659 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
661 case NETCONFA_IFINDEX_ALL:
662 devconf = net->ipv6.devconf_all;
664 case NETCONFA_IFINDEX_DEFAULT:
665 devconf = net->ipv6.devconf_dflt;
668 dev = dev_get_by_index(net, ifindex);
671 in6_dev = in6_dev_get(dev);
674 devconf = &in6_dev->cnf;
679 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
683 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
684 NETLINK_CB(in_skb).portid,
685 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
688 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
689 WARN_ON(err == -EMSGSIZE);
693 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
696 in6_dev_put(in6_dev);
702 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
703 struct netlink_callback *cb)
705 const struct nlmsghdr *nlh = cb->nlh;
706 struct net *net = sock_net(skb->sk);
709 struct net_device *dev;
710 struct inet6_dev *idev;
711 struct hlist_head *head;
713 if (cb->strict_check) {
714 struct netlink_ext_ack *extack = cb->extack;
715 struct netconfmsg *ncm;
717 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
718 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
722 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
723 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
729 s_idx = idx = cb->args[1];
731 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
733 head = &net->dev_index_head[h];
735 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
737 hlist_for_each_entry_rcu(dev, head, index_hlist) {
740 idev = __in6_dev_get(dev);
744 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
746 NETLINK_CB(cb->skb).portid,
754 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
760 if (h == NETDEV_HASHENTRIES) {
761 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
762 net->ipv6.devconf_all,
763 NETLINK_CB(cb->skb).portid,
765 RTM_NEWNETCONF, NLM_F_MULTI,
771 if (h == NETDEV_HASHENTRIES + 1) {
772 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
773 net->ipv6.devconf_dflt,
774 NETLINK_CB(cb->skb).portid,
776 RTM_NEWNETCONF, NLM_F_MULTI,
790 static void dev_forward_change(struct inet6_dev *idev)
792 struct net_device *dev;
793 struct inet6_ifaddr *ifa;
798 if (idev->cnf.forwarding)
799 dev_disable_lro(dev);
800 if (dev->flags & IFF_MULTICAST) {
801 if (idev->cnf.forwarding) {
802 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
803 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
804 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
806 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
807 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
808 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
812 list_for_each_entry(ifa, &idev->addr_list, if_list) {
813 if (ifa->flags&IFA_F_TENTATIVE)
815 if (idev->cnf.forwarding)
816 addrconf_join_anycast(ifa);
818 addrconf_leave_anycast(ifa);
820 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
822 dev->ifindex, &idev->cnf);
826 static void addrconf_forward_change(struct net *net, __s32 newf)
828 struct net_device *dev;
829 struct inet6_dev *idev;
831 for_each_netdev(net, dev) {
832 idev = __in6_dev_get(dev);
834 int changed = (!idev->cnf.forwarding) ^ (!newf);
835 idev->cnf.forwarding = newf;
837 dev_forward_change(idev);
842 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
848 return restart_syscall();
850 net = (struct net *)table->extra2;
854 if (p == &net->ipv6.devconf_dflt->forwarding) {
855 if ((!newf) ^ (!old))
856 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
858 NETCONFA_IFINDEX_DEFAULT,
859 net->ipv6.devconf_dflt);
864 if (p == &net->ipv6.devconf_all->forwarding) {
865 int old_dflt = net->ipv6.devconf_dflt->forwarding;
867 net->ipv6.devconf_dflt->forwarding = newf;
868 if ((!newf) ^ (!old_dflt))
869 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
871 NETCONFA_IFINDEX_DEFAULT,
872 net->ipv6.devconf_dflt);
874 addrconf_forward_change(net, newf);
875 if ((!newf) ^ (!old))
876 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
878 NETCONFA_IFINDEX_ALL,
879 net->ipv6.devconf_all);
880 } else if ((!newf) ^ (!old))
881 dev_forward_change((struct inet6_dev *)table->extra1);
885 rt6_purge_dflt_routers(net);
889 static void addrconf_linkdown_change(struct net *net, __s32 newf)
891 struct net_device *dev;
892 struct inet6_dev *idev;
894 for_each_netdev(net, dev) {
895 idev = __in6_dev_get(dev);
897 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
899 idev->cnf.ignore_routes_with_linkdown = newf;
901 inet6_netconf_notify_devconf(dev_net(dev),
903 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
910 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
916 return restart_syscall();
918 net = (struct net *)table->extra2;
922 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
923 if ((!newf) ^ (!old))
924 inet6_netconf_notify_devconf(net,
926 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
927 NETCONFA_IFINDEX_DEFAULT,
928 net->ipv6.devconf_dflt);
933 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
934 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
935 addrconf_linkdown_change(net, newf);
936 if ((!newf) ^ (!old))
937 inet6_netconf_notify_devconf(net,
939 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
940 NETCONFA_IFINDEX_ALL,
941 net->ipv6.devconf_all);
950 /* Nobody refers to this ifaddr, destroy it */
951 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
953 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
955 #ifdef NET_REFCNT_DEBUG
956 pr_debug("%s\n", __func__);
959 in6_dev_put(ifp->idev);
961 if (cancel_delayed_work(&ifp->dad_work))
962 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
965 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
966 pr_warn("Freeing alive inet6 address %p\n", ifp);
974 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
977 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
980 * Each device address list is sorted in order of scope -
981 * global before linklocal.
983 list_for_each(p, &idev->addr_list) {
984 struct inet6_ifaddr *ifa
985 = list_entry(p, struct inet6_ifaddr, if_list);
986 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
990 list_add_tail_rcu(&ifp->if_list, p);
993 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
995 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
997 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1000 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1001 struct net_device *dev, unsigned int hash)
1003 struct inet6_ifaddr *ifp;
1005 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1006 if (!net_eq(dev_net(ifp->idev->dev), net))
1008 if (ipv6_addr_equal(&ifp->addr, addr)) {
1009 if (!dev || ifp->idev->dev == dev)
1016 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1018 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1021 spin_lock(&addrconf_hash_lock);
1023 /* Ignore adding duplicate addresses on an interface */
1024 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1025 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1028 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1031 spin_unlock(&addrconf_hash_lock);
1036 /* On success it returns ifp with increased reference count */
1038 static struct inet6_ifaddr *
1039 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1040 bool can_block, struct netlink_ext_ack *extack)
1042 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1043 int addr_type = ipv6_addr_type(cfg->pfx);
1044 struct net *net = dev_net(idev->dev);
1045 struct inet6_ifaddr *ifa = NULL;
1046 struct fib6_info *f6i = NULL;
1049 if (addr_type == IPV6_ADDR_ANY ||
1050 (addr_type & IPV6_ADDR_MULTICAST &&
1051 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1052 (!(idev->dev->flags & IFF_LOOPBACK) &&
1053 !netif_is_l3_master(idev->dev) &&
1054 addr_type & IPV6_ADDR_LOOPBACK))
1055 return ERR_PTR(-EADDRNOTAVAIL);
1058 err = -ENODEV; /*XXX*/
1062 if (idev->cnf.disable_ipv6) {
1067 /* validator notifier needs to be blocking;
1068 * do not call in atomic context
1071 struct in6_validator_info i6vi = {
1072 .i6vi_addr = *cfg->pfx,
1077 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1078 err = notifier_to_errno(err);
1083 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1089 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1096 if (net->ipv6.devconf_all->disable_policy ||
1097 idev->cnf.disable_policy)
1098 f6i->dst_nopolicy = true;
1100 neigh_parms_data_state_setall(idev->nd_parms);
1102 ifa->addr = *cfg->pfx;
1104 ifa->peer_addr = *cfg->peer_pfx;
1106 spin_lock_init(&ifa->lock);
1107 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1108 INIT_HLIST_NODE(&ifa->addr_lst);
1109 ifa->scope = cfg->scope;
1110 ifa->prefix_len = cfg->plen;
1111 ifa->rt_priority = cfg->rt_priority;
1112 ifa->flags = cfg->ifa_flags;
1113 /* No need to add the TENTATIVE flag for addresses with NODAD */
1114 if (!(cfg->ifa_flags & IFA_F_NODAD))
1115 ifa->flags |= IFA_F_TENTATIVE;
1116 ifa->valid_lft = cfg->valid_lft;
1117 ifa->prefered_lft = cfg->preferred_lft;
1118 ifa->cstamp = ifa->tstamp = jiffies;
1119 ifa->tokenized = false;
1127 refcount_set(&ifa->refcnt, 1);
1131 err = ipv6_add_addr_hash(idev->dev, ifa);
1133 rcu_read_unlock_bh();
1137 write_lock(&idev->lock);
1139 /* Add to inet6_dev unicast addr list. */
1140 ipv6_link_dev_addr(idev, ifa);
1142 if (ifa->flags&IFA_F_TEMPORARY) {
1143 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1148 write_unlock(&idev->lock);
1150 rcu_read_unlock_bh();
1152 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1154 if (unlikely(err < 0)) {
1155 fib6_info_release(f6i);
1159 in6_dev_put(ifa->idev);
1168 enum cleanup_prefix_rt_t {
1169 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1170 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1171 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1175 * Check, whether the prefix for ifp would still need a prefix route
1176 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1179 * 1) we don't purge prefix if address was not permanent.
1180 * prefix is managed by its own lifetime.
1181 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1182 * 3) if there are no addresses, delete prefix.
1183 * 4) if there are still other permanent address(es),
1184 * corresponding prefix is still permanent.
1185 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1186 * don't purge the prefix, assume user space is managing it.
1187 * 6) otherwise, update prefix lifetime to the
1188 * longest valid lifetime among the corresponding
1189 * addresses on the device.
1190 * Note: subsequent RA will update lifetime.
1192 static enum cleanup_prefix_rt_t
1193 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1195 struct inet6_ifaddr *ifa;
1196 struct inet6_dev *idev = ifp->idev;
1197 unsigned long lifetime;
1198 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1202 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1205 if (ifa->prefix_len != ifp->prefix_len ||
1206 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1209 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1210 return CLEANUP_PREFIX_RT_NOP;
1212 action = CLEANUP_PREFIX_RT_EXPIRE;
1214 spin_lock(&ifa->lock);
1216 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1218 * Note: Because this address is
1219 * not permanent, lifetime <
1220 * LONG_MAX / HZ here.
1222 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1223 *expires = ifa->tstamp + lifetime * HZ;
1224 spin_unlock(&ifa->lock);
1231 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1232 bool del_rt, bool del_peer)
1234 struct fib6_info *f6i;
1236 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1238 ifp->idev->dev, 0, RTF_DEFAULT, true);
1241 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1243 if (!(f6i->fib6_flags & RTF_EXPIRES))
1244 fib6_set_expires(f6i, expires);
1245 fib6_info_release(f6i);
1251 /* This function wants to get referenced ifp and releases it before return */
1253 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1256 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1257 unsigned long expires;
1261 spin_lock_bh(&ifp->lock);
1263 ifp->state = INET6_IFADDR_STATE_DEAD;
1264 spin_unlock_bh(&ifp->lock);
1266 if (state == INET6_IFADDR_STATE_DEAD)
1269 spin_lock_bh(&addrconf_hash_lock);
1270 hlist_del_init_rcu(&ifp->addr_lst);
1271 spin_unlock_bh(&addrconf_hash_lock);
1273 write_lock_bh(&ifp->idev->lock);
1275 if (ifp->flags&IFA_F_TEMPORARY) {
1276 list_del(&ifp->tmp_list);
1278 in6_ifa_put(ifp->ifpub);
1284 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1285 action = check_cleanup_prefix_route(ifp, &expires);
1287 list_del_rcu(&ifp->if_list);
1290 write_unlock_bh(&ifp->idev->lock);
1292 addrconf_del_dad_work(ifp);
1294 ipv6_ifa_notify(RTM_DELADDR, ifp);
1296 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1298 if (action != CLEANUP_PREFIX_RT_NOP) {
1299 cleanup_prefix_route(ifp, expires,
1300 action == CLEANUP_PREFIX_RT_DEL, false);
1303 /* clean up prefsrc entries */
1304 rt6_remove_prefsrc(ifp);
1309 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1311 struct inet6_dev *idev = ifp->idev;
1312 unsigned long tmp_tstamp, age;
1313 unsigned long regen_advance;
1314 unsigned long now = jiffies;
1315 s32 cnf_temp_preferred_lft;
1316 struct inet6_ifaddr *ift;
1317 struct ifa6_config cfg;
1318 long max_desync_factor;
1319 struct in6_addr addr;
1322 write_lock_bh(&idev->lock);
1326 if (idev->cnf.use_tempaddr <= 0) {
1327 write_unlock_bh(&idev->lock);
1328 pr_info("%s: use_tempaddr is disabled\n", __func__);
1333 spin_lock_bh(&ifp->lock);
1334 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1335 idev->cnf.use_tempaddr = -1; /*XXX*/
1336 spin_unlock_bh(&ifp->lock);
1337 write_unlock_bh(&idev->lock);
1338 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1345 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1346 ipv6_gen_rnd_iid(&addr);
1348 age = (now - ifp->tstamp) / HZ;
1350 regen_advance = idev->cnf.regen_max_retry *
1351 idev->cnf.dad_transmits *
1352 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1354 /* recalculate max_desync_factor each time and update
1355 * idev->desync_factor if it's larger
1357 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1358 max_desync_factor = min_t(__u32,
1359 idev->cnf.max_desync_factor,
1360 cnf_temp_preferred_lft - regen_advance);
1362 if (unlikely(idev->desync_factor > max_desync_factor)) {
1363 if (max_desync_factor > 0) {
1364 get_random_bytes(&idev->desync_factor,
1365 sizeof(idev->desync_factor));
1366 idev->desync_factor %= max_desync_factor;
1368 idev->desync_factor = 0;
1372 memset(&cfg, 0, sizeof(cfg));
1373 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1374 idev->cnf.temp_valid_lft + age);
1375 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1376 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1378 cfg.plen = ifp->prefix_len;
1379 tmp_tstamp = ifp->tstamp;
1380 spin_unlock_bh(&ifp->lock);
1382 write_unlock_bh(&idev->lock);
1384 /* A temporary address is created only if this calculated Preferred
1385 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1386 * an implementation must not create a temporary address with a zero
1387 * Preferred Lifetime.
1388 * Use age calculation as in addrconf_verify to avoid unnecessary
1389 * temporary addresses being generated.
1391 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1392 if (cfg.preferred_lft <= regen_advance + age) {
1399 cfg.ifa_flags = IFA_F_TEMPORARY;
1400 /* set in addrconf_prefix_rcv() */
1401 if (ifp->flags & IFA_F_OPTIMISTIC)
1402 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1405 cfg.scope = ipv6_addr_scope(cfg.pfx);
1407 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1411 pr_info("%s: retry temporary address regeneration\n", __func__);
1412 write_lock_bh(&idev->lock);
1416 spin_lock_bh(&ift->lock);
1419 ift->tstamp = tmp_tstamp;
1420 spin_unlock_bh(&ift->lock);
1422 addrconf_dad_start(ift);
1430 * Choose an appropriate source address (RFC3484)
1433 IPV6_SADDR_RULE_INIT = 0,
1434 IPV6_SADDR_RULE_LOCAL,
1435 IPV6_SADDR_RULE_SCOPE,
1436 IPV6_SADDR_RULE_PREFERRED,
1437 #ifdef CONFIG_IPV6_MIP6
1438 IPV6_SADDR_RULE_HOA,
1440 IPV6_SADDR_RULE_OIF,
1441 IPV6_SADDR_RULE_LABEL,
1442 IPV6_SADDR_RULE_PRIVACY,
1443 IPV6_SADDR_RULE_ORCHID,
1444 IPV6_SADDR_RULE_PREFIX,
1445 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1446 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1451 struct ipv6_saddr_score {
1454 struct inet6_ifaddr *ifa;
1455 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1460 struct ipv6_saddr_dst {
1461 const struct in6_addr *addr;
1468 static inline int ipv6_saddr_preferred(int type)
1470 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1475 static bool ipv6_use_optimistic_addr(struct net *net,
1476 struct inet6_dev *idev)
1478 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1481 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1483 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1492 static bool ipv6_allow_optimistic_dad(struct net *net,
1493 struct inet6_dev *idev)
1495 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1498 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1507 static int ipv6_get_saddr_eval(struct net *net,
1508 struct ipv6_saddr_score *score,
1509 struct ipv6_saddr_dst *dst,
1514 if (i <= score->rule) {
1516 case IPV6_SADDR_RULE_SCOPE:
1517 ret = score->scopedist;
1519 case IPV6_SADDR_RULE_PREFIX:
1520 ret = score->matchlen;
1523 ret = !!test_bit(i, score->scorebits);
1529 case IPV6_SADDR_RULE_INIT:
1530 /* Rule 0: remember if hiscore is not ready yet */
1533 case IPV6_SADDR_RULE_LOCAL:
1534 /* Rule 1: Prefer same address */
1535 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1537 case IPV6_SADDR_RULE_SCOPE:
1538 /* Rule 2: Prefer appropriate scope
1543 * ---+--+-+---> scope
1545 * | d is scope of the destination.
1547 * | \ <- smaller scope is better if
1548 * B-15 | \ if scope is enough for destination.
1549 * | ret = B - scope (-1 <= scope >= d <= 15).
1551 * |/ <- greater is better
1552 * -C / if scope is not enough for destination.
1553 * /| ret = scope - C (-1 <= d < scope <= 15).
1555 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1556 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1557 * Assume B = 0 and we get C > 29.
1559 ret = __ipv6_addr_src_scope(score->addr_type);
1560 if (ret >= dst->scope)
1563 ret -= 128; /* 30 is enough */
1564 score->scopedist = ret;
1566 case IPV6_SADDR_RULE_PREFERRED:
1568 /* Rule 3: Avoid deprecated and optimistic addresses */
1569 u8 avoid = IFA_F_DEPRECATED;
1571 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1572 avoid |= IFA_F_OPTIMISTIC;
1573 ret = ipv6_saddr_preferred(score->addr_type) ||
1574 !(score->ifa->flags & avoid);
1577 #ifdef CONFIG_IPV6_MIP6
1578 case IPV6_SADDR_RULE_HOA:
1580 /* Rule 4: Prefer home address */
1581 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1582 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1586 case IPV6_SADDR_RULE_OIF:
1587 /* Rule 5: Prefer outgoing interface */
1588 ret = (!dst->ifindex ||
1589 dst->ifindex == score->ifa->idev->dev->ifindex);
1591 case IPV6_SADDR_RULE_LABEL:
1592 /* Rule 6: Prefer matching label */
1593 ret = ipv6_addr_label(net,
1594 &score->ifa->addr, score->addr_type,
1595 score->ifa->idev->dev->ifindex) == dst->label;
1597 case IPV6_SADDR_RULE_PRIVACY:
1599 /* Rule 7: Prefer public address
1600 * Note: prefer temporary address if use_tempaddr >= 2
1602 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1603 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1604 score->ifa->idev->cnf.use_tempaddr >= 2;
1605 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1608 case IPV6_SADDR_RULE_ORCHID:
1609 /* Rule 8-: Prefer ORCHID vs ORCHID or
1610 * non-ORCHID vs non-ORCHID
1612 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1613 ipv6_addr_orchid(dst->addr));
1615 case IPV6_SADDR_RULE_PREFIX:
1616 /* Rule 8: Use longest matching prefix */
1617 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1618 if (ret > score->ifa->prefix_len)
1619 ret = score->ifa->prefix_len;
1620 score->matchlen = ret;
1622 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1623 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1624 /* Optimistic addresses still have lower precedence than other
1625 * preferred addresses.
1627 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1635 __set_bit(i, score->scorebits);
1641 static int __ipv6_dev_get_saddr(struct net *net,
1642 struct ipv6_saddr_dst *dst,
1643 struct inet6_dev *idev,
1644 struct ipv6_saddr_score *scores,
1647 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1649 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1653 * - Tentative Address (RFC2462 section 5.4)
1654 * - A tentative address is not considered
1655 * "assigned to an interface" in the traditional
1656 * sense, unless it is also flagged as optimistic.
1657 * - Candidate Source Address (section 4)
1658 * - In any case, anycast addresses, multicast
1659 * addresses, and the unspecified address MUST
1660 * NOT be included in a candidate set.
1662 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1663 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1666 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1668 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1669 score->addr_type & IPV6_ADDR_MULTICAST)) {
1670 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1676 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1678 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1679 int minihiscore, miniscore;
1681 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1682 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1684 if (minihiscore > miniscore) {
1685 if (i == IPV6_SADDR_RULE_SCOPE &&
1686 score->scopedist > 0) {
1689 * each remaining entry
1690 * has too small (not enough)
1691 * scope, because ifa entries
1692 * are sorted by their scope
1698 } else if (minihiscore < miniscore) {
1699 swap(hiscore, score);
1700 hiscore_idx = 1 - hiscore_idx;
1702 /* restore our iterator */
1703 score->ifa = hiscore->ifa;
1713 static int ipv6_get_saddr_master(struct net *net,
1714 const struct net_device *dst_dev,
1715 const struct net_device *master,
1716 struct ipv6_saddr_dst *dst,
1717 struct ipv6_saddr_score *scores,
1720 struct inet6_dev *idev;
1722 idev = __in6_dev_get(dst_dev);
1724 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1725 scores, hiscore_idx);
1727 idev = __in6_dev_get(master);
1729 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1730 scores, hiscore_idx);
1735 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1736 const struct in6_addr *daddr, unsigned int prefs,
1737 struct in6_addr *saddr)
1739 struct ipv6_saddr_score scores[2], *hiscore;
1740 struct ipv6_saddr_dst dst;
1741 struct inet6_dev *idev;
1742 struct net_device *dev;
1744 bool use_oif_addr = false;
1745 int hiscore_idx = 0;
1748 dst_type = __ipv6_addr_type(daddr);
1750 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1751 dst.scope = __ipv6_addr_src_scope(dst_type);
1752 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1755 scores[hiscore_idx].rule = -1;
1756 scores[hiscore_idx].ifa = NULL;
1760 /* Candidate Source Address (section 4)
1761 * - multicast and link-local destination address,
1762 * the set of candidate source address MUST only
1763 * include addresses assigned to interfaces
1764 * belonging to the same link as the outgoing
1766 * (- For site-local destination addresses, the
1767 * set of candidate source addresses MUST only
1768 * include addresses assigned to interfaces
1769 * belonging to the same site as the outgoing
1771 * - "It is RECOMMENDED that the candidate source addresses
1772 * be the set of unicast addresses assigned to the
1773 * interface that will be used to send to the destination
1774 * (the 'outgoing' interface)." (RFC 6724)
1777 idev = __in6_dev_get(dst_dev);
1778 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1779 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1780 (idev && idev->cnf.use_oif_addrs_only)) {
1781 use_oif_addr = true;
1787 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1789 const struct net_device *master;
1792 /* if dst_dev exists and is enslaved to an L3 device, then
1793 * prefer addresses from dst_dev and then the master over
1794 * any other enslaved devices in the L3 domain.
1796 master = l3mdev_master_dev_rcu(dst_dev);
1798 master_idx = master->ifindex;
1800 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1802 scores, hiscore_idx);
1804 if (scores[hiscore_idx].ifa)
1808 for_each_netdev_rcu(net, dev) {
1809 /* only consider addresses on devices in the
1812 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1814 idev = __in6_dev_get(dev);
1817 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1822 hiscore = &scores[hiscore_idx];
1824 ret = -EADDRNOTAVAIL;
1826 *saddr = hiscore->ifa->addr;
1831 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1833 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1836 struct inet6_ifaddr *ifp;
1837 int err = -EADDRNOTAVAIL;
1839 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1840 if (ifp->scope > IFA_LINK)
1842 if (ifp->scope == IFA_LINK &&
1843 !(ifp->flags & banned_flags)) {
1852 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1855 struct inet6_dev *idev;
1856 int err = -EADDRNOTAVAIL;
1859 idev = __in6_dev_get(dev);
1861 read_lock_bh(&idev->lock);
1862 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1863 read_unlock_bh(&idev->lock);
1869 static int ipv6_count_addresses(const struct inet6_dev *idev)
1871 const struct inet6_ifaddr *ifp;
1875 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1881 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1882 const struct net_device *dev, int strict)
1884 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1885 strict, IFA_F_TENTATIVE);
1887 EXPORT_SYMBOL(ipv6_chk_addr);
1889 /* device argument is used to find the L3 domain of interest. If
1890 * skip_dev_check is set, then the ifp device is not checked against
1891 * the passed in dev argument. So the 2 cases for addresses checks are:
1892 * 1. does the address exist in the L3 domain that dev is part of
1893 * (skip_dev_check = true), or
1895 * 2. does the address exist on the specific device
1896 * (skip_dev_check = false)
1898 static struct net_device *
1899 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1900 const struct net_device *dev, bool skip_dev_check,
1901 int strict, u32 banned_flags)
1903 unsigned int hash = inet6_addr_hash(net, addr);
1904 struct net_device *l3mdev, *ndev;
1905 struct inet6_ifaddr *ifp;
1910 l3mdev = l3mdev_master_dev_rcu(dev);
1914 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1915 ndev = ifp->idev->dev;
1916 if (!net_eq(dev_net(ndev), net))
1919 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1922 /* Decouple optimistic from tentative for evaluation here.
1923 * Ban optimistic addresses explicitly, when required.
1925 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1926 ? (ifp->flags&~IFA_F_TENTATIVE)
1928 if (ipv6_addr_equal(&ifp->addr, addr) &&
1929 !(ifp_flags&banned_flags) &&
1930 (!dev || ndev == dev ||
1931 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1941 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1942 const struct net_device *dev, bool skip_dev_check,
1943 int strict, u32 banned_flags)
1945 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1946 strict, banned_flags) ? 1 : 0;
1948 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1951 /* Compares an address/prefix_len with addresses on device @dev.
1952 * If one is found it returns true.
1954 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1955 const unsigned int prefix_len, struct net_device *dev)
1957 const struct inet6_ifaddr *ifa;
1958 const struct inet6_dev *idev;
1962 idev = __in6_dev_get(dev);
1964 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1965 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1974 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1976 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1978 const struct inet6_ifaddr *ifa;
1979 const struct inet6_dev *idev;
1984 idev = __in6_dev_get(dev);
1986 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1987 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1996 EXPORT_SYMBOL(ipv6_chk_prefix);
1999 * ipv6_dev_find - find the first device with a given source address.
2000 * @net: the net namespace
2001 * @addr: the source address
2002 * @dev: used to find the L3 domain of interest
2004 * The caller should be protected by RCU, or RTNL.
2006 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2007 struct net_device *dev)
2009 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2012 EXPORT_SYMBOL(ipv6_dev_find);
2014 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2015 struct net_device *dev, int strict)
2017 unsigned int hash = inet6_addr_hash(net, addr);
2018 struct inet6_ifaddr *ifp, *result = NULL;
2021 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2022 if (!net_eq(dev_net(ifp->idev->dev), net))
2024 if (ipv6_addr_equal(&ifp->addr, addr)) {
2025 if (!dev || ifp->idev->dev == dev ||
2026 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2038 /* Gets referenced address, destroys ifaddr */
2040 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2043 ifp->flags |= IFA_F_DADFAILED;
2045 if (ifp->flags&IFA_F_TEMPORARY) {
2046 struct inet6_ifaddr *ifpub;
2047 spin_lock_bh(&ifp->lock);
2050 in6_ifa_hold(ifpub);
2051 spin_unlock_bh(&ifp->lock);
2052 ipv6_create_tempaddr(ifpub, true);
2055 spin_unlock_bh(&ifp->lock);
2058 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2059 spin_lock_bh(&ifp->lock);
2060 addrconf_del_dad_work(ifp);
2061 ifp->flags |= IFA_F_TENTATIVE;
2063 ifp->flags &= ~IFA_F_OPTIMISTIC;
2064 spin_unlock_bh(&ifp->lock);
2066 ipv6_ifa_notify(0, ifp);
2073 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2077 spin_lock_bh(&ifp->lock);
2078 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2079 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2082 spin_unlock_bh(&ifp->lock);
2087 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2089 struct inet6_dev *idev = ifp->idev;
2090 struct net *net = dev_net(ifp->idev->dev);
2092 if (addrconf_dad_end(ifp)) {
2097 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2098 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2100 spin_lock_bh(&ifp->lock);
2102 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2103 struct in6_addr new_addr;
2104 struct inet6_ifaddr *ifp2;
2105 int retries = ifp->stable_privacy_retry + 1;
2106 struct ifa6_config cfg = {
2108 .plen = ifp->prefix_len,
2109 .ifa_flags = ifp->flags,
2110 .valid_lft = ifp->valid_lft,
2111 .preferred_lft = ifp->prefered_lft,
2112 .scope = ifp->scope,
2115 if (retries > net->ipv6.sysctl.idgen_retries) {
2116 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2117 ifp->idev->dev->name);
2121 new_addr = ifp->addr;
2122 if (ipv6_generate_stable_address(&new_addr, retries,
2126 spin_unlock_bh(&ifp->lock);
2128 if (idev->cnf.max_addresses &&
2129 ipv6_count_addresses(idev) >=
2130 idev->cnf.max_addresses)
2133 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2134 ifp->idev->dev->name);
2136 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2140 spin_lock_bh(&ifp2->lock);
2141 ifp2->stable_privacy_retry = retries;
2142 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2143 spin_unlock_bh(&ifp2->lock);
2145 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2148 spin_lock_bh(&ifp->lock);
2152 /* transition from _POSTDAD to _ERRDAD */
2153 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2154 spin_unlock_bh(&ifp->lock);
2156 addrconf_mod_dad_work(ifp, 0);
2160 /* Join to solicited addr multicast group.
2161 * caller must hold RTNL */
2162 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2164 struct in6_addr maddr;
2166 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2169 addrconf_addr_solict_mult(addr, &maddr);
2170 ipv6_dev_mc_inc(dev, &maddr);
2173 /* caller must hold RTNL */
2174 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2176 struct in6_addr maddr;
2178 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2181 addrconf_addr_solict_mult(addr, &maddr);
2182 __ipv6_dev_mc_dec(idev, &maddr);
2185 /* caller must hold RTNL */
2186 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2188 struct in6_addr addr;
2190 if (ifp->prefix_len >= 127) /* RFC 6164 */
2192 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2193 if (ipv6_addr_any(&addr))
2195 __ipv6_dev_ac_inc(ifp->idev, &addr);
2198 /* caller must hold RTNL */
2199 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2201 struct in6_addr addr;
2203 if (ifp->prefix_len >= 127) /* RFC 6164 */
2205 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2206 if (ipv6_addr_any(&addr))
2208 __ipv6_dev_ac_dec(ifp->idev, &addr);
2211 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2213 switch (dev->addr_len) {
2215 memcpy(eui, dev->dev_addr, 3);
2218 memcpy(eui + 5, dev->dev_addr + 3, 3);
2220 case EUI64_ADDR_LEN:
2221 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2231 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2233 union fwnet_hwaddr *ha;
2235 if (dev->addr_len != FWNET_ALEN)
2238 ha = (union fwnet_hwaddr *)dev->dev_addr;
2240 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2245 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2247 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2248 if (dev->addr_len != ARCNET_ALEN)
2251 eui[7] = *(u8 *)dev->dev_addr;
2255 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2257 if (dev->addr_len != INFINIBAND_ALEN)
2259 memcpy(eui, dev->dev_addr + 12, 8);
2264 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2268 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2269 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2270 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2271 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2272 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2273 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2277 memcpy(eui + 4, &addr, 4);
2281 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2283 if (dev->priv_flags & IFF_ISATAP)
2284 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2288 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2290 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2293 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2295 memcpy(eui, dev->perm_addr, 3);
2296 memcpy(eui + 5, dev->perm_addr + 3, 3);
2303 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2305 switch (dev->type) {
2308 return addrconf_ifid_eui48(eui, dev);
2310 return addrconf_ifid_arcnet(eui, dev);
2311 case ARPHRD_INFINIBAND:
2312 return addrconf_ifid_infiniband(eui, dev);
2314 return addrconf_ifid_sit(eui, dev);
2317 return addrconf_ifid_gre(eui, dev);
2318 case ARPHRD_6LOWPAN:
2319 return addrconf_ifid_6lowpan(eui, dev);
2320 case ARPHRD_IEEE1394:
2321 return addrconf_ifid_ieee1394(eui, dev);
2322 case ARPHRD_TUNNEL6:
2325 return addrconf_ifid_ip6tnl(eui, dev);
2330 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2333 struct inet6_ifaddr *ifp;
2335 read_lock_bh(&idev->lock);
2336 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2337 if (ifp->scope > IFA_LINK)
2339 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2340 memcpy(eui, ifp->addr.s6_addr+8, 8);
2345 read_unlock_bh(&idev->lock);
2349 /* Generation of a randomized Interface Identifier
2350 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2353 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2356 get_random_bytes(&addr->s6_addr[8], 8);
2358 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2359 * check if generated address is not inappropriate:
2361 * - Reserved IPv6 Interface Identifers
2362 * - XXX: already assigned to an address on the device
2365 /* Subnet-router anycast: 0000:0000:0000:0000 */
2366 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2369 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2370 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2371 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2373 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2374 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2377 /* Reserved subnet anycast addresses */
2378 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2379 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2388 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2389 struct net_device *dev, unsigned long expires,
2390 u32 flags, gfp_t gfp_flags)
2392 struct fib6_config cfg = {
2393 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2394 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2395 .fc_ifindex = dev->ifindex,
2396 .fc_expires = expires,
2398 .fc_flags = RTF_UP | flags,
2399 .fc_nlinfo.nl_net = dev_net(dev),
2400 .fc_protocol = RTPROT_KERNEL,
2401 .fc_type = RTN_UNICAST,
2406 /* Prevent useless cloning on PtP SIT.
2407 This thing is done here expecting that the whole
2408 class of non-broadcast devices need not cloning.
2410 #if IS_ENABLED(CONFIG_IPV6_SIT)
2411 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2412 cfg.fc_flags |= RTF_NONEXTHOP;
2415 ip6_route_add(&cfg, gfp_flags, NULL);
2419 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2421 const struct net_device *dev,
2422 u32 flags, u32 noflags,
2425 struct fib6_node *fn;
2426 struct fib6_info *rt = NULL;
2427 struct fib6_table *table;
2428 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2430 table = fib6_get_table(dev_net(dev), tb_id);
2435 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2439 for_each_fib6_node_rt_rcu(fn) {
2440 /* prefix routes only use builtin fib6_nh */
2444 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2446 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2448 if ((rt->fib6_flags & flags) != flags)
2450 if ((rt->fib6_flags & noflags) != 0)
2452 if (!fib6_info_hold_safe(rt))
2462 /* Create "default" multicast route to the interface */
2464 static void addrconf_add_mroute(struct net_device *dev)
2466 struct fib6_config cfg = {
2467 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2468 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2469 .fc_ifindex = dev->ifindex,
2472 .fc_type = RTN_MULTICAST,
2473 .fc_nlinfo.nl_net = dev_net(dev),
2474 .fc_protocol = RTPROT_KERNEL,
2477 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2479 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2482 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2484 struct inet6_dev *idev;
2488 idev = ipv6_find_idev(dev);
2492 if (idev->cnf.disable_ipv6)
2493 return ERR_PTR(-EACCES);
2495 /* Add default multicast route */
2496 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2497 addrconf_add_mroute(dev);
2502 static void manage_tempaddrs(struct inet6_dev *idev,
2503 struct inet6_ifaddr *ifp,
2504 __u32 valid_lft, __u32 prefered_lft,
2505 bool create, unsigned long now)
2508 struct inet6_ifaddr *ift;
2510 read_lock_bh(&idev->lock);
2511 /* update all temporary addresses in the list */
2512 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2513 int age, max_valid, max_prefered;
2515 if (ifp != ift->ifpub)
2518 /* RFC 4941 section 3.3:
2519 * If a received option will extend the lifetime of a public
2520 * address, the lifetimes of temporary addresses should
2521 * be extended, subject to the overall constraint that no
2522 * temporary addresses should ever remain "valid" or "preferred"
2523 * for a time longer than (TEMP_VALID_LIFETIME) or
2524 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2526 age = (now - ift->cstamp) / HZ;
2527 max_valid = idev->cnf.temp_valid_lft - age;
2531 max_prefered = idev->cnf.temp_prefered_lft -
2532 idev->desync_factor - age;
2533 if (max_prefered < 0)
2536 if (valid_lft > max_valid)
2537 valid_lft = max_valid;
2539 if (prefered_lft > max_prefered)
2540 prefered_lft = max_prefered;
2542 spin_lock(&ift->lock);
2544 ift->valid_lft = valid_lft;
2545 ift->prefered_lft = prefered_lft;
2547 if (prefered_lft > 0)
2548 ift->flags &= ~IFA_F_DEPRECATED;
2550 spin_unlock(&ift->lock);
2551 if (!(flags&IFA_F_TENTATIVE))
2552 ipv6_ifa_notify(0, ift);
2555 if ((create || list_empty(&idev->tempaddr_list)) &&
2556 idev->cnf.use_tempaddr > 0) {
2557 /* When a new public address is created as described
2558 * in [ADDRCONF], also create a new temporary address.
2559 * Also create a temporary address if it's enabled but
2560 * no temporary address currently exists.
2562 read_unlock_bh(&idev->lock);
2563 ipv6_create_tempaddr(ifp, false);
2565 read_unlock_bh(&idev->lock);
2569 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2571 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2572 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2575 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2576 const struct prefix_info *pinfo,
2577 struct inet6_dev *in6_dev,
2578 const struct in6_addr *addr, int addr_type,
2579 u32 addr_flags, bool sllao, bool tokenized,
2580 __u32 valid_lft, u32 prefered_lft)
2582 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2585 if (!ifp && valid_lft) {
2586 int max_addresses = in6_dev->cnf.max_addresses;
2587 struct ifa6_config cfg = {
2589 .plen = pinfo->prefix_len,
2590 .ifa_flags = addr_flags,
2591 .valid_lft = valid_lft,
2592 .preferred_lft = prefered_lft,
2593 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2596 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2597 if ((net->ipv6.devconf_all->optimistic_dad ||
2598 in6_dev->cnf.optimistic_dad) &&
2599 !net->ipv6.devconf_all->forwarding && sllao)
2600 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2603 /* Do not allow to create too much of autoconfigured
2604 * addresses; this would be too easy way to crash kernel.
2606 if (!max_addresses ||
2607 ipv6_count_addresses(in6_dev) < max_addresses)
2608 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2610 if (IS_ERR_OR_NULL(ifp))
2614 spin_lock_bh(&ifp->lock);
2615 ifp->flags |= IFA_F_MANAGETEMPADDR;
2616 ifp->cstamp = jiffies;
2617 ifp->tokenized = tokenized;
2618 spin_unlock_bh(&ifp->lock);
2619 addrconf_dad_start(ifp);
2627 /* Update lifetime (RFC4862 5.5.3 e)
2628 * We deviate from RFC4862 by honoring all Valid Lifetimes to
2629 * improve the reaction of SLAAC to renumbering events
2630 * (draft-gont-6man-slaac-renum-06, Section 4.2)
2632 spin_lock_bh(&ifp->lock);
2634 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2635 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2639 if (!create && stored_lft) {
2640 ifp->valid_lft = valid_lft;
2641 ifp->prefered_lft = prefered_lft;
2644 ifp->flags &= ~IFA_F_DEPRECATED;
2645 spin_unlock_bh(&ifp->lock);
2647 if (!(flags&IFA_F_TENTATIVE))
2648 ipv6_ifa_notify(0, ifp);
2650 spin_unlock_bh(&ifp->lock);
2652 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2661 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2663 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2665 struct prefix_info *pinfo;
2670 struct inet6_dev *in6_dev;
2671 struct net *net = dev_net(dev);
2673 pinfo = (struct prefix_info *) opt;
2675 if (len < sizeof(struct prefix_info)) {
2676 netdev_dbg(dev, "addrconf: prefix option too short\n");
2681 * Validation checks ([ADDRCONF], page 19)
2684 addr_type = ipv6_addr_type(&pinfo->prefix);
2686 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2689 valid_lft = ntohl(pinfo->valid);
2690 prefered_lft = ntohl(pinfo->prefered);
2692 if (prefered_lft > valid_lft) {
2693 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2697 in6_dev = in6_dev_get(dev);
2700 net_dbg_ratelimited("addrconf: device %s not configured\n",
2706 * Two things going on here:
2707 * 1) Add routes for on-link prefixes
2708 * 2) Configure prefixes with the auto flag set
2711 if (pinfo->onlink) {
2712 struct fib6_info *rt;
2713 unsigned long rt_expires;
2715 /* Avoid arithmetic overflow. Really, we could
2716 * save rt_expires in seconds, likely valid_lft,
2717 * but it would require division in fib gc, that it
2721 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2723 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2725 if (addrconf_finite_timeout(rt_expires))
2728 rt = addrconf_get_prefix_route(&pinfo->prefix,
2731 RTF_ADDRCONF | RTF_PREFIX_RT,
2735 /* Autoconf prefix route */
2736 if (valid_lft == 0) {
2737 ip6_del_rt(net, rt, false);
2739 } else if (addrconf_finite_timeout(rt_expires)) {
2741 fib6_set_expires(rt, jiffies + rt_expires);
2743 fib6_clean_expires(rt);
2745 } else if (valid_lft) {
2746 clock_t expires = 0;
2747 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2748 if (addrconf_finite_timeout(rt_expires)) {
2750 flags |= RTF_EXPIRES;
2751 expires = jiffies_to_clock_t(rt_expires);
2753 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2754 0, dev, expires, flags,
2757 fib6_info_release(rt);
2760 /* Try to figure out our local address for this prefix */
2762 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2763 struct in6_addr addr;
2764 bool tokenized = false, dev_addr_generated = false;
2766 if (pinfo->prefix_len == 64) {
2767 memcpy(&addr, &pinfo->prefix, 8);
2769 if (!ipv6_addr_any(&in6_dev->token)) {
2770 read_lock_bh(&in6_dev->lock);
2771 memcpy(addr.s6_addr + 8,
2772 in6_dev->token.s6_addr + 8, 8);
2773 read_unlock_bh(&in6_dev->lock);
2775 } else if (is_addr_mode_generate_stable(in6_dev) &&
2776 !ipv6_generate_stable_address(&addr, 0,
2778 addr_flags |= IFA_F_STABLE_PRIVACY;
2780 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2781 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2784 dev_addr_generated = true;
2788 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2793 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2796 tokenized, valid_lft,
2801 /* Ignore error case here because previous prefix add addr was
2802 * successful which will be notified.
2804 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2805 addr_type, addr_flags, sllao,
2806 tokenized, valid_lft,
2808 dev_addr_generated);
2810 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2812 in6_dev_put(in6_dev);
2815 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2816 struct in6_ifreq *ireq)
2818 struct ip_tunnel_parm p = { };
2821 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2822 return -EADDRNOTAVAIL;
2824 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2827 p.iph.protocol = IPPROTO_IPV6;
2830 if (!dev->netdev_ops->ndo_tunnel_ctl)
2832 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2836 dev = __dev_get_by_name(net, p.name);
2839 return dev_open(dev, NULL);
2843 * Set destination address.
2844 * Special case for SIT interfaces where we create a new "virtual"
2847 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2849 struct net_device *dev;
2850 struct in6_ifreq ireq;
2853 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2855 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2859 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2860 if (dev && dev->type == ARPHRD_SIT)
2861 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2866 static int ipv6_mc_config(struct sock *sk, bool join,
2867 const struct in6_addr *addr, int ifindex)
2875 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2877 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2884 * Manual configuration of address on an interface
2886 static int inet6_addr_add(struct net *net, int ifindex,
2887 struct ifa6_config *cfg,
2888 struct netlink_ext_ack *extack)
2890 struct inet6_ifaddr *ifp;
2891 struct inet6_dev *idev;
2892 struct net_device *dev;
2893 unsigned long timeout;
2899 if (cfg->plen > 128)
2902 /* check the lifetime */
2903 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2906 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2909 dev = __dev_get_by_index(net, ifindex);
2913 idev = addrconf_add_dev(dev);
2915 return PTR_ERR(idev);
2917 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2918 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2919 true, cfg->pfx, ifindex);
2925 cfg->scope = ipv6_addr_scope(cfg->pfx);
2927 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2928 if (addrconf_finite_timeout(timeout)) {
2929 expires = jiffies_to_clock_t(timeout * HZ);
2930 cfg->valid_lft = timeout;
2931 flags = RTF_EXPIRES;
2935 cfg->ifa_flags |= IFA_F_PERMANENT;
2938 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2939 if (addrconf_finite_timeout(timeout)) {
2941 cfg->ifa_flags |= IFA_F_DEPRECATED;
2942 cfg->preferred_lft = timeout;
2945 ifp = ipv6_add_addr(idev, cfg, true, extack);
2947 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2948 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2949 ifp->rt_priority, dev, expires,
2953 /* Send a netlink notification if DAD is enabled and
2954 * optimistic flag is not set
2956 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2957 ipv6_ifa_notify(0, ifp);
2959 * Note that section 3.1 of RFC 4429 indicates
2960 * that the Optimistic flag should not be set for
2961 * manually configured addresses
2963 addrconf_dad_start(ifp);
2964 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2965 manage_tempaddrs(idev, ifp, cfg->valid_lft,
2966 cfg->preferred_lft, true, jiffies);
2968 addrconf_verify_rtnl();
2970 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2971 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2975 return PTR_ERR(ifp);
2978 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2979 const struct in6_addr *pfx, unsigned int plen)
2981 struct inet6_ifaddr *ifp;
2982 struct inet6_dev *idev;
2983 struct net_device *dev;
2988 dev = __dev_get_by_index(net, ifindex);
2992 idev = __in6_dev_get(dev);
2996 read_lock_bh(&idev->lock);
2997 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2998 if (ifp->prefix_len == plen &&
2999 ipv6_addr_equal(pfx, &ifp->addr)) {
3001 read_unlock_bh(&idev->lock);
3003 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3004 (ifa_flags & IFA_F_MANAGETEMPADDR))
3005 manage_tempaddrs(idev, ifp, 0, 0, false,
3008 addrconf_verify_rtnl();
3009 if (ipv6_addr_is_multicast(pfx)) {
3010 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3011 false, pfx, dev->ifindex);
3016 read_unlock_bh(&idev->lock);
3017 return -EADDRNOTAVAIL;
3021 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3023 struct ifa6_config cfg = {
3024 .ifa_flags = IFA_F_PERMANENT,
3025 .preferred_lft = INFINITY_LIFE_TIME,
3026 .valid_lft = INFINITY_LIFE_TIME,
3028 struct in6_ifreq ireq;
3031 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3034 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3037 cfg.pfx = &ireq.ifr6_addr;
3038 cfg.plen = ireq.ifr6_prefixlen;
3041 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3046 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3048 struct in6_ifreq ireq;
3051 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3054 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3058 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3059 ireq.ifr6_prefixlen);
3064 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3065 int plen, int scope)
3067 struct inet6_ifaddr *ifp;
3068 struct ifa6_config cfg = {
3071 .ifa_flags = IFA_F_PERMANENT,
3072 .valid_lft = INFINITY_LIFE_TIME,
3073 .preferred_lft = INFINITY_LIFE_TIME,
3077 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3079 spin_lock_bh(&ifp->lock);
3080 ifp->flags &= ~IFA_F_TENTATIVE;
3081 spin_unlock_bh(&ifp->lock);
3082 rt_genid_bump_ipv6(dev_net(idev->dev));
3083 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3088 #if IS_ENABLED(CONFIG_IPV6_SIT)
3089 static void sit_add_v4_addrs(struct inet6_dev *idev)
3091 struct in6_addr addr;
3092 struct net_device *dev;
3093 struct net *net = dev_net(idev->dev);
3099 memset(&addr, 0, sizeof(struct in6_addr));
3100 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3102 if (idev->dev->flags&IFF_POINTOPOINT) {
3103 addr.s6_addr32[0] = htonl(0xfe800000);
3107 scope = IPV6_ADDR_COMPATv4;
3109 pflags |= RTF_NONEXTHOP;
3112 if (addr.s6_addr32[3]) {
3113 add_addr(idev, &addr, plen, scope);
3114 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3119 for_each_netdev(net, dev) {
3120 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3121 if (in_dev && (dev->flags & IFF_UP)) {
3122 struct in_ifaddr *ifa;
3125 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3126 addr.s6_addr32[3] = ifa->ifa_local;
3128 if (ifa->ifa_scope == RT_SCOPE_LINK)
3130 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3131 if (idev->dev->flags&IFF_POINTOPOINT)
3136 add_addr(idev, &addr, plen, flag);
3137 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3138 0, pflags, GFP_KERNEL);
3145 static void init_loopback(struct net_device *dev)
3147 struct inet6_dev *idev;
3153 idev = ipv6_find_idev(dev);
3155 pr_debug("%s: add_dev failed\n", __func__);
3159 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3162 void addrconf_add_linklocal(struct inet6_dev *idev,
3163 const struct in6_addr *addr, u32 flags)
3165 struct ifa6_config cfg = {
3168 .ifa_flags = flags | IFA_F_PERMANENT,
3169 .valid_lft = INFINITY_LIFE_TIME,
3170 .preferred_lft = INFINITY_LIFE_TIME,
3173 struct inet6_ifaddr *ifp;
3175 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3176 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3177 idev->cnf.optimistic_dad) &&
3178 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3179 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3182 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3184 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3186 addrconf_dad_start(ifp);
3190 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3192 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3194 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3197 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3198 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3201 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3202 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3208 static int ipv6_generate_stable_address(struct in6_addr *address,
3210 const struct inet6_dev *idev)
3212 static DEFINE_SPINLOCK(lock);
3213 static __u32 digest[SHA1_DIGEST_WORDS];
3214 static __u32 workspace[SHA1_WORKSPACE_WORDS];
3217 char __data[SHA1_BLOCK_SIZE];
3219 struct in6_addr secret;
3221 unsigned char hwaddr[MAX_ADDR_LEN];
3226 struct in6_addr secret;
3227 struct in6_addr temp;
3228 struct net *net = dev_net(idev->dev);
3230 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3232 if (idev->cnf.stable_secret.initialized)
3233 secret = idev->cnf.stable_secret.secret;
3234 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3235 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3240 spin_lock_bh(&lock);
3243 memset(&data, 0, sizeof(data));
3244 memset(workspace, 0, sizeof(workspace));
3245 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3246 data.prefix[0] = address->s6_addr32[0];
3247 data.prefix[1] = address->s6_addr32[1];
3248 data.secret = secret;
3249 data.dad_count = dad_count;
3251 sha1_transform(digest, data.__data, workspace);
3254 temp.s6_addr32[2] = (__force __be32)digest[0];
3255 temp.s6_addr32[3] = (__force __be32)digest[1];
3257 spin_unlock_bh(&lock);
3259 if (ipv6_reserved_interfaceid(temp)) {
3261 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3270 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3272 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3276 s = &idev->cnf.stable_secret;
3277 get_random_bytes(&s->secret, sizeof(s->secret));
3278 s->initialized = true;
3281 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3283 struct in6_addr addr;
3285 /* no link local addresses on L3 master devices */
3286 if (netif_is_l3_master(idev->dev))
3289 /* no link local addresses on devices flagged as slaves */
3290 if (idev->dev->flags & IFF_SLAVE)
3293 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3295 switch (idev->cnf.addr_gen_mode) {
3296 case IN6_ADDR_GEN_MODE_RANDOM:
3297 ipv6_gen_mode_random_init(idev);
3299 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3300 if (!ipv6_generate_stable_address(&addr, 0, idev))
3301 addrconf_add_linklocal(idev, &addr,
3302 IFA_F_STABLE_PRIVACY);
3303 else if (prefix_route)
3304 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3307 case IN6_ADDR_GEN_MODE_EUI64:
3308 /* addrconf_add_linklocal also adds a prefix_route and we
3309 * only need to care about prefix routes if ipv6_generate_eui64
3310 * couldn't generate one.
3312 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3313 addrconf_add_linklocal(idev, &addr, 0);
3314 else if (prefix_route)
3315 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3318 case IN6_ADDR_GEN_MODE_NONE:
3320 /* will not add any link local address */
3325 static void addrconf_dev_config(struct net_device *dev)
3327 struct inet6_dev *idev;
3331 if ((dev->type != ARPHRD_ETHER) &&
3332 (dev->type != ARPHRD_FDDI) &&
3333 (dev->type != ARPHRD_ARCNET) &&
3334 (dev->type != ARPHRD_INFINIBAND) &&
3335 (dev->type != ARPHRD_IEEE1394) &&
3336 (dev->type != ARPHRD_TUNNEL6) &&
3337 (dev->type != ARPHRD_6LOWPAN) &&
3338 (dev->type != ARPHRD_IP6GRE) &&
3339 (dev->type != ARPHRD_IPGRE) &&
3340 (dev->type != ARPHRD_TUNNEL) &&
3341 (dev->type != ARPHRD_NONE) &&
3342 (dev->type != ARPHRD_RAWIP)) {
3343 /* Alas, we support only Ethernet autoconfiguration. */
3344 idev = __in6_dev_get(dev);
3345 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3346 dev->flags & IFF_MULTICAST)
3351 idev = addrconf_add_dev(dev);
3355 /* this device type has no EUI support */
3356 if (dev->type == ARPHRD_NONE &&
3357 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3358 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3360 addrconf_addr_gen(idev, false);
3363 #if IS_ENABLED(CONFIG_IPV6_SIT)
3364 static void addrconf_sit_config(struct net_device *dev)
3366 struct inet6_dev *idev;
3371 * Configure the tunnel with one of our IPv4
3372 * addresses... we should configure all of
3373 * our v4 addrs in the tunnel
3376 idev = ipv6_find_idev(dev);
3378 pr_debug("%s: add_dev failed\n", __func__);
3382 if (dev->priv_flags & IFF_ISATAP) {
3383 addrconf_addr_gen(idev, false);
3387 sit_add_v4_addrs(idev);
3389 if (dev->flags&IFF_POINTOPOINT)
3390 addrconf_add_mroute(dev);
3394 #if IS_ENABLED(CONFIG_NET_IPGRE)
3395 static void addrconf_gre_config(struct net_device *dev)
3397 struct inet6_dev *idev;
3401 idev = ipv6_find_idev(dev);
3403 pr_debug("%s: add_dev failed\n", __func__);
3407 addrconf_addr_gen(idev, true);
3408 if (dev->flags & IFF_POINTOPOINT)
3409 addrconf_add_mroute(dev);
3413 static int fixup_permanent_addr(struct net *net,
3414 struct inet6_dev *idev,
3415 struct inet6_ifaddr *ifp)
3417 /* !fib6_node means the host route was removed from the
3418 * FIB, for example, if 'lo' device is taken down. In that
3419 * case regenerate the host route.
3421 if (!ifp->rt || !ifp->rt->fib6_node) {
3422 struct fib6_info *f6i, *prev;
3424 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3427 return PTR_ERR(f6i);
3429 /* ifp->rt can be accessed outside of rtnl */
3430 spin_lock(&ifp->lock);
3433 spin_unlock(&ifp->lock);
3435 fib6_info_release(prev);
3438 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3439 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3440 ifp->rt_priority, idev->dev, 0, 0,
3444 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3445 addrconf_dad_start(ifp);
3450 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3452 struct inet6_ifaddr *ifp, *tmp;
3453 struct inet6_dev *idev;
3455 idev = __in6_dev_get(dev);
3459 write_lock_bh(&idev->lock);
3461 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3462 if ((ifp->flags & IFA_F_PERMANENT) &&
3463 fixup_permanent_addr(net, idev, ifp) < 0) {
3464 write_unlock_bh(&idev->lock);
3467 write_lock_bh(&idev->lock);
3469 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3470 idev->dev->name, &ifp->addr);
3474 write_unlock_bh(&idev->lock);
3477 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3480 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3481 struct netdev_notifier_change_info *change_info;
3482 struct netdev_notifier_changeupper_info *info;
3483 struct inet6_dev *idev = __in6_dev_get(dev);
3484 struct net *net = dev_net(dev);
3485 int run_pending = 0;
3489 case NETDEV_REGISTER:
3490 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3491 idev = ipv6_add_dev(dev);
3493 return notifier_from_errno(PTR_ERR(idev));
3497 case NETDEV_CHANGEMTU:
3498 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3499 if (dev->mtu < IPV6_MIN_MTU) {
3500 addrconf_ifdown(dev, dev != net->loopback_dev);
3505 rt6_mtu_change(dev, dev->mtu);
3506 idev->cnf.mtu6 = dev->mtu;
3510 /* allocate new idev */
3511 idev = ipv6_add_dev(dev);
3515 /* device is still not ready */
3516 if (!(idev->if_flags & IF_READY))
3523 if (dev->flags & IFF_SLAVE)
3526 if (idev && idev->cnf.disable_ipv6)
3529 if (event == NETDEV_UP) {
3530 /* restore routes for permanent addresses */
3531 addrconf_permanent_addr(net, dev);
3533 if (!addrconf_link_ready(dev)) {
3534 /* device is not ready yet. */
3535 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3540 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3541 idev = ipv6_add_dev(dev);
3543 if (!IS_ERR_OR_NULL(idev)) {
3544 idev->if_flags |= IF_READY;
3547 } else if (event == NETDEV_CHANGE) {
3548 if (!addrconf_link_ready(dev)) {
3549 /* device is still not ready. */
3550 rt6_sync_down_dev(dev, event);
3554 if (!IS_ERR_OR_NULL(idev)) {
3555 if (idev->if_flags & IF_READY) {
3556 /* device is already configured -
3557 * but resend MLD reports, we might
3558 * have roamed and need to update
3559 * multicast snooping switches
3563 if (change_info->flags_changed & IFF_NOARP)
3564 addrconf_dad_run(idev, true);
3565 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3568 idev->if_flags |= IF_READY;
3571 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3577 switch (dev->type) {
3578 #if IS_ENABLED(CONFIG_IPV6_SIT)
3580 addrconf_sit_config(dev);
3583 #if IS_ENABLED(CONFIG_NET_IPGRE)
3585 addrconf_gre_config(dev);
3588 case ARPHRD_LOOPBACK:
3593 addrconf_dev_config(dev);
3597 if (!IS_ERR_OR_NULL(idev)) {
3599 addrconf_dad_run(idev, false);
3601 /* Device has an address by now */
3602 rt6_sync_up(dev, RTNH_F_DEAD);
3605 * If the MTU changed during the interface down,
3606 * when the interface up, the changed MTU must be
3607 * reflected in the idev as well as routers.
3609 if (idev->cnf.mtu6 != dev->mtu &&
3610 dev->mtu >= IPV6_MIN_MTU) {
3611 rt6_mtu_change(dev, dev->mtu);
3612 idev->cnf.mtu6 = dev->mtu;
3614 idev->tstamp = jiffies;
3615 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3618 * If the changed mtu during down is lower than
3619 * IPV6_MIN_MTU stop IPv6 on this interface.
3621 if (dev->mtu < IPV6_MIN_MTU)
3622 addrconf_ifdown(dev, dev != net->loopback_dev);
3627 case NETDEV_UNREGISTER:
3629 * Remove all addresses from this interface.
3631 addrconf_ifdown(dev, event != NETDEV_DOWN);
3634 case NETDEV_CHANGENAME:
3636 snmp6_unregister_dev(idev);
3637 addrconf_sysctl_unregister(idev);
3638 err = addrconf_sysctl_register(idev);
3640 return notifier_from_errno(err);
3641 err = snmp6_register_dev(idev);
3643 addrconf_sysctl_unregister(idev);
3644 return notifier_from_errno(err);
3649 case NETDEV_PRE_TYPE_CHANGE:
3650 case NETDEV_POST_TYPE_CHANGE:
3652 addrconf_type_change(dev, event);
3655 case NETDEV_CHANGEUPPER:
3658 /* flush all routes if dev is linked to or unlinked from
3659 * an L3 master device (e.g., VRF)
3661 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3662 addrconf_ifdown(dev, false);
3669 * addrconf module should be notified of a device going up
3671 static struct notifier_block ipv6_dev_notf = {
3672 .notifier_call = addrconf_notify,
3673 .priority = ADDRCONF_NOTIFY_PRIORITY,
3676 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3678 struct inet6_dev *idev;
3681 idev = __in6_dev_get(dev);
3683 if (event == NETDEV_POST_TYPE_CHANGE)
3684 ipv6_mc_remap(idev);
3685 else if (event == NETDEV_PRE_TYPE_CHANGE)
3686 ipv6_mc_unmap(idev);
3689 static bool addr_is_local(const struct in6_addr *addr)
3691 return ipv6_addr_type(addr) &
3692 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3695 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3697 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3698 struct net *net = dev_net(dev);
3699 struct inet6_dev *idev;
3700 struct inet6_ifaddr *ifa, *tmp;
3701 bool keep_addr = false;
3706 rt6_disable_ip(dev, event);
3708 idev = __in6_dev_get(dev);
3713 * Step 1: remove reference to ipv6 device from parent device.
3719 /* protected by rtnl_lock */
3720 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3722 /* Step 1.5: remove snmp6 entry */
3723 snmp6_unregister_dev(idev);
3727 /* combine the user config with event to determine if permanent
3728 * addresses are to be removed from address hash table
3730 if (!unregister && !idev->cnf.disable_ipv6) {
3731 /* aggregate the system setting and interface setting */
3732 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3735 _keep_addr = idev->cnf.keep_addr_on_down;
3737 keep_addr = (_keep_addr > 0);
3740 /* Step 2: clear hash table */
3741 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3742 struct hlist_head *h = &inet6_addr_lst[i];
3744 spin_lock_bh(&addrconf_hash_lock);
3746 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3747 if (ifa->idev == idev) {
3748 addrconf_del_dad_work(ifa);
3749 /* combined flag + permanent flag decide if
3750 * address is retained on a down event
3753 !(ifa->flags & IFA_F_PERMANENT) ||
3754 addr_is_local(&ifa->addr)) {
3755 hlist_del_init_rcu(&ifa->addr_lst);
3760 spin_unlock_bh(&addrconf_hash_lock);
3763 write_lock_bh(&idev->lock);
3765 addrconf_del_rs_timer(idev);
3767 /* Step 2: clear flags for stateless addrconf */
3769 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3771 /* Step 3: clear tempaddr list */
3772 while (!list_empty(&idev->tempaddr_list)) {
3773 ifa = list_first_entry(&idev->tempaddr_list,
3774 struct inet6_ifaddr, tmp_list);
3775 list_del(&ifa->tmp_list);
3776 write_unlock_bh(&idev->lock);
3777 spin_lock_bh(&ifa->lock);
3780 in6_ifa_put(ifa->ifpub);
3783 spin_unlock_bh(&ifa->lock);
3785 write_lock_bh(&idev->lock);
3788 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3789 struct fib6_info *rt = NULL;
3792 addrconf_del_dad_work(ifa);
3794 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3795 !addr_is_local(&ifa->addr);
3797 write_unlock_bh(&idev->lock);
3798 spin_lock_bh(&ifa->lock);
3801 /* set state to skip the notifier below */
3802 state = INET6_IFADDR_STATE_DEAD;
3803 ifa->state = INET6_IFADDR_STATE_PREDAD;
3804 if (!(ifa->flags & IFA_F_NODAD))
3805 ifa->flags |= IFA_F_TENTATIVE;
3811 ifa->state = INET6_IFADDR_STATE_DEAD;
3814 spin_unlock_bh(&ifa->lock);
3817 ip6_del_rt(net, rt, false);
3819 if (state != INET6_IFADDR_STATE_DEAD) {
3820 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3821 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3823 if (idev->cnf.forwarding)
3824 addrconf_leave_anycast(ifa);
3825 addrconf_leave_solict(ifa->idev, &ifa->addr);
3828 write_lock_bh(&idev->lock);
3830 list_del_rcu(&ifa->if_list);
3835 write_unlock_bh(&idev->lock);
3837 /* Step 5: Discard anycast and multicast list */
3839 ipv6_ac_destroy_dev(idev);
3840 ipv6_mc_destroy_dev(idev);
3845 idev->tstamp = jiffies;
3847 /* Last: Shot the device (if unregistered) */
3849 addrconf_sysctl_unregister(idev);
3850 neigh_parms_release(&nd_tbl, idev->nd_parms);
3851 neigh_ifdown(&nd_tbl, dev);
3857 static void addrconf_rs_timer(struct timer_list *t)
3859 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3860 struct net_device *dev = idev->dev;
3861 struct in6_addr lladdr;
3863 write_lock(&idev->lock);
3864 if (idev->dead || !(idev->if_flags & IF_READY))
3867 if (!ipv6_accept_ra(idev))
3870 /* Announcement received after solicitation was sent */
3871 if (idev->if_flags & IF_RA_RCVD)
3874 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3875 write_unlock(&idev->lock);
3876 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3877 ndisc_send_rs(dev, &lladdr,
3878 &in6addr_linklocal_allrouters);
3882 write_lock(&idev->lock);
3883 idev->rs_interval = rfc3315_s14_backoff_update(
3884 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3885 /* The wait after the last probe can be shorter */
3886 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3887 idev->cnf.rtr_solicits) ?
3888 idev->cnf.rtr_solicit_delay :
3892 * Note: we do not support deprecated "all on-link"
3893 * assumption any longer.
3895 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3899 write_unlock(&idev->lock);
3905 * Duplicate Address Detection
3907 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3909 unsigned long rand_num;
3910 struct inet6_dev *idev = ifp->idev;
3913 if (ifp->flags & IFA_F_OPTIMISTIC)
3916 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3919 if (idev->cnf.enhanced_dad ||
3920 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3922 get_random_bytes(&nonce, 6);
3925 ifp->dad_nonce = nonce;
3926 ifp->dad_probes = idev->cnf.dad_transmits;
3927 addrconf_mod_dad_work(ifp, rand_num);
3930 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3932 struct inet6_dev *idev = ifp->idev;
3933 struct net_device *dev = idev->dev;
3934 bool bump_id, notify = false;
3937 addrconf_join_solict(dev, &ifp->addr);
3939 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3941 read_lock_bh(&idev->lock);
3942 spin_lock(&ifp->lock);
3943 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3947 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3948 (net->ipv6.devconf_all->accept_dad < 1 &&
3949 idev->cnf.accept_dad < 1) ||
3950 !(ifp->flags&IFA_F_TENTATIVE) ||
3951 ifp->flags & IFA_F_NODAD) {
3952 bool send_na = false;
3954 if (ifp->flags & IFA_F_TENTATIVE &&
3955 !(ifp->flags & IFA_F_OPTIMISTIC))
3957 bump_id = ifp->flags & IFA_F_TENTATIVE;
3958 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3959 spin_unlock(&ifp->lock);
3960 read_unlock_bh(&idev->lock);
3962 addrconf_dad_completed(ifp, bump_id, send_na);
3966 if (!(idev->if_flags & IF_READY)) {
3967 spin_unlock(&ifp->lock);
3968 read_unlock_bh(&idev->lock);
3970 * If the device is not ready:
3971 * - keep it tentative if it is a permanent address.
3972 * - otherwise, kill it.
3975 addrconf_dad_stop(ifp, 0);
3980 * Optimistic nodes can start receiving
3983 if (ifp->flags & IFA_F_OPTIMISTIC) {
3984 ip6_ins_rt(net, ifp->rt);
3985 if (ipv6_use_optimistic_addr(net, idev)) {
3986 /* Because optimistic nodes can use this address,
3987 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3993 addrconf_dad_kick(ifp);
3995 spin_unlock(&ifp->lock);
3996 read_unlock_bh(&idev->lock);
3998 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4001 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4003 bool begin_dad = false;
4005 spin_lock_bh(&ifp->lock);
4006 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4007 ifp->state = INET6_IFADDR_STATE_PREDAD;
4010 spin_unlock_bh(&ifp->lock);
4013 addrconf_mod_dad_work(ifp, 0);
4016 static void addrconf_dad_work(struct work_struct *w)
4018 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4019 struct inet6_ifaddr,
4021 struct inet6_dev *idev = ifp->idev;
4022 bool bump_id, disable_ipv6 = false;
4023 struct in6_addr mcaddr;
4029 } action = DAD_PROCESS;
4033 spin_lock_bh(&ifp->lock);
4034 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4036 ifp->state = INET6_IFADDR_STATE_DAD;
4037 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4039 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4041 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4042 idev->cnf.accept_dad > 1) &&
4043 !idev->cnf.disable_ipv6 &&
4044 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4045 struct in6_addr addr;
4047 addr.s6_addr32[0] = htonl(0xfe800000);
4048 addr.s6_addr32[1] = 0;
4050 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4051 ipv6_addr_equal(&ifp->addr, &addr)) {
4052 /* DAD failed for link-local based on MAC */
4053 idev->cnf.disable_ipv6 = 1;
4055 pr_info("%s: IPv6 being disabled!\n",
4056 ifp->idev->dev->name);
4057 disable_ipv6 = true;
4061 spin_unlock_bh(&ifp->lock);
4063 if (action == DAD_BEGIN) {
4064 addrconf_dad_begin(ifp);
4066 } else if (action == DAD_ABORT) {
4068 addrconf_dad_stop(ifp, 1);
4070 addrconf_ifdown(idev->dev, false);
4074 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4077 write_lock_bh(&idev->lock);
4078 if (idev->dead || !(idev->if_flags & IF_READY)) {
4079 write_unlock_bh(&idev->lock);
4083 spin_lock(&ifp->lock);
4084 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4085 spin_unlock(&ifp->lock);
4086 write_unlock_bh(&idev->lock);
4090 if (ifp->dad_probes == 0) {
4091 bool send_na = false;
4094 * DAD was successful
4097 if (ifp->flags & IFA_F_TENTATIVE &&
4098 !(ifp->flags & IFA_F_OPTIMISTIC))
4100 bump_id = ifp->flags & IFA_F_TENTATIVE;
4101 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4102 spin_unlock(&ifp->lock);
4103 write_unlock_bh(&idev->lock);
4105 addrconf_dad_completed(ifp, bump_id, send_na);
4111 addrconf_mod_dad_work(ifp,
4112 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4114 spin_unlock(&ifp->lock);
4115 write_unlock_bh(&idev->lock);
4117 /* send a neighbour solicitation for our addr */
4118 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4119 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4126 /* ifp->idev must be at least read locked */
4127 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4129 struct inet6_ifaddr *ifpiter;
4130 struct inet6_dev *idev = ifp->idev;
4132 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4133 if (ifpiter->scope > IFA_LINK)
4135 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4136 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4137 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4147 struct net_device *dev = ifp->idev->dev;
4148 struct in6_addr lladdr;
4149 bool send_rs, send_mld;
4151 addrconf_del_dad_work(ifp);
4154 * Configure the address for reception. Now it is valid.
4157 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4159 /* If added prefix is link local and we are prepared to process
4160 router advertisements, start sending router solicitations.
4163 read_lock_bh(&ifp->idev->lock);
4164 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4165 send_rs = send_mld &&
4166 ipv6_accept_ra(ifp->idev) &&
4167 ifp->idev->cnf.rtr_solicits != 0 &&
4168 (dev->flags&IFF_LOOPBACK) == 0;
4169 read_unlock_bh(&ifp->idev->lock);
4171 /* While dad is in progress mld report's source address is in6_addrany.
4172 * Resend with proper ll now.
4175 ipv6_mc_dad_complete(ifp->idev);
4177 /* send unsolicited NA if enabled */
4179 (ifp->idev->cnf.ndisc_notify ||
4180 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4181 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4182 /*router=*/ !!ifp->idev->cnf.forwarding,
4183 /*solicited=*/ false, /*override=*/ true,
4189 * If a host as already performed a random delay
4190 * [...] as part of DAD [...] there is no need
4191 * to delay again before sending the first RS
4193 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4195 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4197 write_lock_bh(&ifp->idev->lock);
4198 spin_lock(&ifp->lock);
4199 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4200 ifp->idev->cnf.rtr_solicit_interval);
4201 ifp->idev->rs_probes = 1;
4202 ifp->idev->if_flags |= IF_RS_SENT;
4203 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4204 spin_unlock(&ifp->lock);
4205 write_unlock_bh(&ifp->idev->lock);
4209 rt_genid_bump_ipv6(dev_net(dev));
4211 /* Make sure that a new temporary address will be created
4212 * before this temporary address becomes deprecated.
4214 if (ifp->flags & IFA_F_TEMPORARY)
4215 addrconf_verify_rtnl();
4218 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4220 struct inet6_ifaddr *ifp;
4222 read_lock_bh(&idev->lock);
4223 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4224 spin_lock(&ifp->lock);
4225 if ((ifp->flags & IFA_F_TENTATIVE &&
4226 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4228 ifp->state = INET6_IFADDR_STATE_PREDAD;
4229 addrconf_dad_kick(ifp);
4231 spin_unlock(&ifp->lock);
4233 read_unlock_bh(&idev->lock);
4236 #ifdef CONFIG_PROC_FS
4237 struct if6_iter_state {
4238 struct seq_net_private p;
4243 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4245 struct if6_iter_state *state = seq->private;
4246 struct net *net = seq_file_net(seq);
4247 struct inet6_ifaddr *ifa = NULL;
4250 /* initial bucket if pos is 0 */
4256 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4257 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4259 if (!net_eq(dev_net(ifa->idev->dev), net))
4261 /* sync with offset */
4262 if (p < state->offset) {
4269 /* prepare for next bucket */
4276 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4277 struct inet6_ifaddr *ifa)
4279 struct if6_iter_state *state = seq->private;
4280 struct net *net = seq_file_net(seq);
4282 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4283 if (!net_eq(dev_net(ifa->idev->dev), net))
4290 while (++state->bucket < IN6_ADDR_HSIZE) {
4291 hlist_for_each_entry_rcu(ifa,
4292 &inet6_addr_lst[state->bucket], addr_lst) {
4293 if (!net_eq(dev_net(ifa->idev->dev), net))
4302 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4306 return if6_get_first(seq, *pos);
4309 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4311 struct inet6_ifaddr *ifa;
4313 ifa = if6_get_next(seq, v);
4318 static void if6_seq_stop(struct seq_file *seq, void *v)
4324 static int if6_seq_show(struct seq_file *seq, void *v)
4326 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4327 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4329 ifp->idev->dev->ifindex,
4333 ifp->idev->dev->name);
4337 static const struct seq_operations if6_seq_ops = {
4338 .start = if6_seq_start,
4339 .next = if6_seq_next,
4340 .show = if6_seq_show,
4341 .stop = if6_seq_stop,
4344 static int __net_init if6_proc_net_init(struct net *net)
4346 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4347 sizeof(struct if6_iter_state)))
4352 static void __net_exit if6_proc_net_exit(struct net *net)
4354 remove_proc_entry("if_inet6", net->proc_net);
4357 static struct pernet_operations if6_proc_net_ops = {
4358 .init = if6_proc_net_init,
4359 .exit = if6_proc_net_exit,
4362 int __init if6_proc_init(void)
4364 return register_pernet_subsys(&if6_proc_net_ops);
4367 void if6_proc_exit(void)
4369 unregister_pernet_subsys(&if6_proc_net_ops);
4371 #endif /* CONFIG_PROC_FS */
4373 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4374 /* Check if address is a home address configured on any interface. */
4375 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4377 unsigned int hash = inet6_addr_hash(net, addr);
4378 struct inet6_ifaddr *ifp = NULL;
4382 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4383 if (!net_eq(dev_net(ifp->idev->dev), net))
4385 if (ipv6_addr_equal(&ifp->addr, addr) &&
4386 (ifp->flags & IFA_F_HOMEADDRESS)) {
4396 /* RFC6554 has some algorithm to avoid loops in segment routing by
4397 * checking if the segments contains any of a local interface address.
4401 * To detect loops in the SRH, a router MUST determine if the SRH
4402 * includes multiple addresses assigned to any interface on that router.
4403 * If such addresses appear more than once and are separated by at least
4404 * one address not assigned to that router.
4406 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4407 unsigned char nsegs)
4409 const struct in6_addr *addr;
4410 int i, ret = 0, found = 0;
4411 struct inet6_ifaddr *ifp;
4412 bool separated = false;
4417 for (i = 0; i < nsegs; i++) {
4419 hash = inet6_addr_hash(net, addr);
4422 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4423 if (!net_eq(dev_net(ifp->idev->dev), net))
4426 if (ipv6_addr_equal(&ifp->addr, addr)) {
4433 if (found > 1 && separated) {
4450 * Periodic address status verification
4453 static void addrconf_verify_rtnl(void)
4455 unsigned long now, next, next_sec, next_sched;
4456 struct inet6_ifaddr *ifp;
4463 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4465 cancel_delayed_work(&addr_chk_work);
4467 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4469 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4472 /* When setting preferred_lft to a value not zero or
4473 * infinity, while valid_lft is infinity
4474 * IFA_F_PERMANENT has a non-infinity life time.
4476 if ((ifp->flags & IFA_F_PERMANENT) &&
4477 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4480 spin_lock(&ifp->lock);
4481 /* We try to batch several events at once. */
4482 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4484 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4485 age >= ifp->valid_lft) {
4486 spin_unlock(&ifp->lock);
4490 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4491 spin_unlock(&ifp->lock);
4493 } else if (age >= ifp->prefered_lft) {
4494 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4497 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4499 ifp->flags |= IFA_F_DEPRECATED;
4502 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4503 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4504 next = ifp->tstamp + ifp->valid_lft * HZ;
4506 spin_unlock(&ifp->lock);
4511 ipv6_ifa_notify(0, ifp);
4515 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4516 !(ifp->flags&IFA_F_TENTATIVE)) {
4517 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4518 ifp->idev->cnf.dad_transmits *
4519 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4521 if (age >= ifp->prefered_lft - regen_advance) {
4522 struct inet6_ifaddr *ifpub = ifp->ifpub;
4523 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4524 next = ifp->tstamp + ifp->prefered_lft * HZ;
4525 if (!ifp->regen_count && ifpub) {
4528 in6_ifa_hold(ifpub);
4529 spin_unlock(&ifp->lock);
4531 spin_lock(&ifpub->lock);
4532 ifpub->regen_count = 0;
4533 spin_unlock(&ifpub->lock);
4534 rcu_read_unlock_bh();
4535 ipv6_create_tempaddr(ifpub, true);
4541 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4542 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4543 spin_unlock(&ifp->lock);
4545 /* ifp->prefered_lft <= ifp->valid_lft */
4546 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4547 next = ifp->tstamp + ifp->prefered_lft * HZ;
4548 spin_unlock(&ifp->lock);
4553 next_sec = round_jiffies_up(next);
4556 /* If rounded timeout is accurate enough, accept it. */
4557 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4558 next_sched = next_sec;
4560 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4561 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4562 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4564 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4565 now, next, next_sec, next_sched);
4566 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4567 rcu_read_unlock_bh();
4570 static void addrconf_verify_work(struct work_struct *w)
4573 addrconf_verify_rtnl();
4577 static void addrconf_verify(void)
4579 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4582 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4583 struct in6_addr **peer_pfx)
4585 struct in6_addr *pfx = NULL;
4590 pfx = nla_data(addr);
4593 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4595 pfx = nla_data(local);
4601 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4602 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4603 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4604 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4605 [IFA_FLAGS] = { .len = sizeof(u32) },
4606 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4607 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4611 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4612 struct netlink_ext_ack *extack)
4614 struct net *net = sock_net(skb->sk);
4615 struct ifaddrmsg *ifm;
4616 struct nlattr *tb[IFA_MAX+1];
4617 struct in6_addr *pfx, *peer_pfx;
4621 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4622 ifa_ipv6_policy, extack);
4626 ifm = nlmsg_data(nlh);
4627 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4631 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4633 /* We ignore other flags so far. */
4634 ifa_flags &= IFA_F_MANAGETEMPADDR;
4636 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4637 ifm->ifa_prefixlen);
4640 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4641 unsigned long expires, u32 flags,
4644 struct fib6_info *f6i;
4647 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4649 ifp->idev->dev, 0, RTF_DEFAULT, true);
4653 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4654 if (f6i->fib6_metric != prio) {
4655 /* delete old one */
4656 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4659 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4661 ifp->rt_priority, ifp->idev->dev,
4662 expires, flags, GFP_KERNEL);
4665 fib6_clean_expires(f6i);
4667 fib6_set_expires(f6i, expires);
4669 fib6_info_release(f6i);
4675 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4679 unsigned long timeout;
4680 bool was_managetempaddr;
4681 bool had_prefixroute;
4682 bool new_peer = false;
4686 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4689 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4690 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4693 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4694 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4696 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4697 if (addrconf_finite_timeout(timeout)) {
4698 expires = jiffies_to_clock_t(timeout * HZ);
4699 cfg->valid_lft = timeout;
4700 flags = RTF_EXPIRES;
4704 cfg->ifa_flags |= IFA_F_PERMANENT;
4707 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4708 if (addrconf_finite_timeout(timeout)) {
4710 cfg->ifa_flags |= IFA_F_DEPRECATED;
4711 cfg->preferred_lft = timeout;
4714 if (cfg->peer_pfx &&
4715 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4716 if (!ipv6_addr_any(&ifp->peer_addr))
4717 cleanup_prefix_route(ifp, expires, true, true);
4721 spin_lock_bh(&ifp->lock);
4722 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4723 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4724 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4725 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4726 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4727 IFA_F_NOPREFIXROUTE);
4728 ifp->flags |= cfg->ifa_flags;
4729 ifp->tstamp = jiffies;
4730 ifp->valid_lft = cfg->valid_lft;
4731 ifp->prefered_lft = cfg->preferred_lft;
4733 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4734 ifp->rt_priority = cfg->rt_priority;
4737 ifp->peer_addr = *cfg->peer_pfx;
4739 spin_unlock_bh(&ifp->lock);
4740 if (!(ifp->flags&IFA_F_TENTATIVE))
4741 ipv6_ifa_notify(0, ifp);
4743 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4746 if (had_prefixroute)
4747 rc = modify_prefix_route(ifp, expires, flags, false);
4749 /* prefix route could have been deleted; if so restore it */
4750 if (rc == -ENOENT) {
4751 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4752 ifp->rt_priority, ifp->idev->dev,
4753 expires, flags, GFP_KERNEL);
4756 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4757 rc = modify_prefix_route(ifp, expires, flags, true);
4759 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4760 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4761 ifp->rt_priority, ifp->idev->dev,
4762 expires, flags, GFP_KERNEL);
4764 } else if (had_prefixroute) {
4765 enum cleanup_prefix_rt_t action;
4766 unsigned long rt_expires;
4768 write_lock_bh(&ifp->idev->lock);
4769 action = check_cleanup_prefix_route(ifp, &rt_expires);
4770 write_unlock_bh(&ifp->idev->lock);
4772 if (action != CLEANUP_PREFIX_RT_NOP) {
4773 cleanup_prefix_route(ifp, rt_expires,
4774 action == CLEANUP_PREFIX_RT_DEL, false);
4778 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4779 if (was_managetempaddr &&
4780 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4782 cfg->preferred_lft = 0;
4784 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4785 cfg->preferred_lft, !was_managetempaddr,
4789 addrconf_verify_rtnl();
4795 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4796 struct netlink_ext_ack *extack)
4798 struct net *net = sock_net(skb->sk);
4799 struct ifaddrmsg *ifm;
4800 struct nlattr *tb[IFA_MAX+1];
4801 struct in6_addr *peer_pfx;
4802 struct inet6_ifaddr *ifa;
4803 struct net_device *dev;
4804 struct inet6_dev *idev;
4805 struct ifa6_config cfg;
4808 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4809 ifa_ipv6_policy, extack);
4813 memset(&cfg, 0, sizeof(cfg));
4815 ifm = nlmsg_data(nlh);
4816 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4820 cfg.peer_pfx = peer_pfx;
4821 cfg.plen = ifm->ifa_prefixlen;
4822 if (tb[IFA_RT_PRIORITY])
4823 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4825 cfg.valid_lft = INFINITY_LIFE_TIME;
4826 cfg.preferred_lft = INFINITY_LIFE_TIME;
4828 if (tb[IFA_CACHEINFO]) {
4829 struct ifa_cacheinfo *ci;
4831 ci = nla_data(tb[IFA_CACHEINFO]);
4832 cfg.valid_lft = ci->ifa_valid;
4833 cfg.preferred_lft = ci->ifa_prefered;
4836 dev = __dev_get_by_index(net, ifm->ifa_index);
4841 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4843 cfg.ifa_flags = ifm->ifa_flags;
4845 /* We ignore other flags so far. */
4846 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4847 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4848 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4850 idev = ipv6_find_idev(dev);
4852 return PTR_ERR(idev);
4854 if (!ipv6_allow_optimistic_dad(net, idev))
4855 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4857 if (cfg.ifa_flags & IFA_F_NODAD &&
4858 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4859 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4863 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4866 * It would be best to check for !NLM_F_CREATE here but
4867 * userspace already relies on not having to provide this.
4869 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4872 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4873 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4876 err = inet6_addr_modify(ifa, &cfg);
4883 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4884 u8 scope, int ifindex)
4886 struct ifaddrmsg *ifm;
4888 ifm = nlmsg_data(nlh);
4889 ifm->ifa_family = AF_INET6;
4890 ifm->ifa_prefixlen = prefixlen;
4891 ifm->ifa_flags = flags;
4892 ifm->ifa_scope = scope;
4893 ifm->ifa_index = ifindex;
4896 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4897 unsigned long tstamp, u32 preferred, u32 valid)
4899 struct ifa_cacheinfo ci;
4901 ci.cstamp = cstamp_delta(cstamp);
4902 ci.tstamp = cstamp_delta(tstamp);
4903 ci.ifa_prefered = preferred;
4904 ci.ifa_valid = valid;
4906 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4909 static inline int rt_scope(int ifa_scope)
4911 if (ifa_scope & IFA_HOST)
4912 return RT_SCOPE_HOST;
4913 else if (ifa_scope & IFA_LINK)
4914 return RT_SCOPE_LINK;
4915 else if (ifa_scope & IFA_SITE)
4916 return RT_SCOPE_SITE;
4918 return RT_SCOPE_UNIVERSE;
4921 static inline int inet6_ifaddr_msgsize(void)
4923 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4924 + nla_total_size(16) /* IFA_LOCAL */
4925 + nla_total_size(16) /* IFA_ADDRESS */
4926 + nla_total_size(sizeof(struct ifa_cacheinfo))
4927 + nla_total_size(4) /* IFA_FLAGS */
4928 + nla_total_size(4) /* IFA_RT_PRIORITY */;
4937 struct inet6_fill_args {
4944 enum addr_type_t type;
4947 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4948 struct inet6_fill_args *args)
4950 struct nlmsghdr *nlh;
4951 u32 preferred, valid;
4953 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4954 sizeof(struct ifaddrmsg), args->flags);
4958 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4959 ifa->idev->dev->ifindex);
4961 if (args->netnsid >= 0 &&
4962 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4965 if (!((ifa->flags&IFA_F_PERMANENT) &&
4966 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4967 preferred = ifa->prefered_lft;
4968 valid = ifa->valid_lft;
4969 if (preferred != INFINITY_LIFE_TIME) {
4970 long tval = (jiffies - ifa->tstamp)/HZ;
4971 if (preferred > tval)
4975 if (valid != INFINITY_LIFE_TIME) {
4983 preferred = INFINITY_LIFE_TIME;
4984 valid = INFINITY_LIFE_TIME;
4987 if (!ipv6_addr_any(&ifa->peer_addr)) {
4988 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4989 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4992 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4995 if (ifa->rt_priority &&
4996 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4999 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5002 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5005 nlmsg_end(skb, nlh);
5009 nlmsg_cancel(skb, nlh);
5013 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5014 struct inet6_fill_args *args)
5016 struct nlmsghdr *nlh;
5017 u8 scope = RT_SCOPE_UNIVERSE;
5018 int ifindex = ifmca->idev->dev->ifindex;
5020 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5021 scope = RT_SCOPE_SITE;
5023 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5024 sizeof(struct ifaddrmsg), args->flags);
5028 if (args->netnsid >= 0 &&
5029 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5030 nlmsg_cancel(skb, nlh);
5034 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5035 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5036 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5037 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5038 nlmsg_cancel(skb, nlh);
5042 nlmsg_end(skb, nlh);
5046 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5047 struct inet6_fill_args *args)
5049 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5050 int ifindex = dev ? dev->ifindex : 1;
5051 struct nlmsghdr *nlh;
5052 u8 scope = RT_SCOPE_UNIVERSE;
5054 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5055 scope = RT_SCOPE_SITE;
5057 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5058 sizeof(struct ifaddrmsg), args->flags);
5062 if (args->netnsid >= 0 &&
5063 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5064 nlmsg_cancel(skb, nlh);
5068 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5069 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5070 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5071 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5072 nlmsg_cancel(skb, nlh);
5076 nlmsg_end(skb, nlh);
5080 /* called with rcu_read_lock() */
5081 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5082 struct netlink_callback *cb, int s_ip_idx,
5083 struct inet6_fill_args *fillargs)
5085 struct ifmcaddr6 *ifmca;
5086 struct ifacaddr6 *ifaca;
5090 read_lock_bh(&idev->lock);
5091 switch (fillargs->type) {
5092 case UNICAST_ADDR: {
5093 struct inet6_ifaddr *ifa;
5094 fillargs->event = RTM_NEWADDR;
5096 /* unicast address incl. temp addr */
5097 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5098 if (ip_idx < s_ip_idx)
5100 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5103 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5109 case MULTICAST_ADDR:
5110 fillargs->event = RTM_GETMULTICAST;
5112 /* multicast address */
5113 for (ifmca = idev->mc_list; ifmca;
5114 ifmca = ifmca->next, ip_idx++) {
5115 if (ip_idx < s_ip_idx)
5117 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5123 fillargs->event = RTM_GETANYCAST;
5124 /* anycast address */
5125 for (ifaca = idev->ac_list; ifaca;
5126 ifaca = ifaca->aca_next, ip_idx++) {
5127 if (ip_idx < s_ip_idx)
5129 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5137 read_unlock_bh(&idev->lock);
5138 cb->args[2] = ip_idx;
5142 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5143 struct inet6_fill_args *fillargs,
5144 struct net **tgt_net, struct sock *sk,
5145 struct netlink_callback *cb)
5147 struct netlink_ext_ack *extack = cb->extack;
5148 struct nlattr *tb[IFA_MAX+1];
5149 struct ifaddrmsg *ifm;
5152 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5153 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5157 ifm = nlmsg_data(nlh);
5158 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5159 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5163 fillargs->ifindex = ifm->ifa_index;
5164 if (fillargs->ifindex) {
5165 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5166 fillargs->flags |= NLM_F_DUMP_FILTERED;
5169 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5170 ifa_ipv6_policy, extack);
5174 for (i = 0; i <= IFA_MAX; ++i) {
5178 if (i == IFA_TARGET_NETNSID) {
5181 fillargs->netnsid = nla_get_s32(tb[i]);
5182 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5184 fillargs->netnsid = -1;
5185 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5186 return PTR_ERR(net);
5190 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5198 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5199 enum addr_type_t type)
5201 const struct nlmsghdr *nlh = cb->nlh;
5202 struct inet6_fill_args fillargs = {
5203 .portid = NETLINK_CB(cb->skb).portid,
5204 .seq = cb->nlh->nlmsg_seq,
5205 .flags = NLM_F_MULTI,
5209 struct net *net = sock_net(skb->sk);
5210 struct net *tgt_net = net;
5211 int idx, s_idx, s_ip_idx;
5213 struct net_device *dev;
5214 struct inet6_dev *idev;
5215 struct hlist_head *head;
5219 s_idx = idx = cb->args[1];
5220 s_ip_idx = cb->args[2];
5222 if (cb->strict_check) {
5223 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5229 if (fillargs.ifindex) {
5230 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5235 idev = __in6_dev_get(dev);
5237 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5247 cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5248 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5250 head = &tgt_net->dev_index_head[h];
5251 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5254 if (h > s_h || idx > s_idx)
5256 idev = __in6_dev_get(dev);
5260 if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5272 if (fillargs.netnsid >= 0)
5275 return skb->len ? : err;
5278 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5280 enum addr_type_t type = UNICAST_ADDR;
5282 return inet6_dump_addr(skb, cb, type);
5285 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5287 enum addr_type_t type = MULTICAST_ADDR;
5289 return inet6_dump_addr(skb, cb, type);
5293 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5295 enum addr_type_t type = ANYCAST_ADDR;
5297 return inet6_dump_addr(skb, cb, type);
5300 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5301 const struct nlmsghdr *nlh,
5303 struct netlink_ext_ack *extack)
5305 struct ifaddrmsg *ifm;
5308 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5309 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5313 if (!netlink_strict_get_check(skb))
5314 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5315 ifa_ipv6_policy, extack);
5317 ifm = nlmsg_data(nlh);
5318 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5319 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5323 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5324 ifa_ipv6_policy, extack);
5328 for (i = 0; i <= IFA_MAX; i++) {
5333 case IFA_TARGET_NETNSID:
5338 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5346 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5347 struct netlink_ext_ack *extack)
5349 struct net *net = sock_net(in_skb->sk);
5350 struct inet6_fill_args fillargs = {
5351 .portid = NETLINK_CB(in_skb).portid,
5352 .seq = nlh->nlmsg_seq,
5353 .event = RTM_NEWADDR,
5357 struct net *tgt_net = net;
5358 struct ifaddrmsg *ifm;
5359 struct nlattr *tb[IFA_MAX+1];
5360 struct in6_addr *addr = NULL, *peer;
5361 struct net_device *dev = NULL;
5362 struct inet6_ifaddr *ifa;
5363 struct sk_buff *skb;
5366 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5370 if (tb[IFA_TARGET_NETNSID]) {
5371 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5373 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5375 if (IS_ERR(tgt_net))
5376 return PTR_ERR(tgt_net);
5379 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5383 ifm = nlmsg_data(nlh);
5385 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5387 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5389 err = -EADDRNOTAVAIL;
5393 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5399 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5401 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5402 WARN_ON(err == -EMSGSIZE);
5406 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5412 if (fillargs.netnsid >= 0)
5418 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5420 struct sk_buff *skb;
5421 struct net *net = dev_net(ifa->idev->dev);
5422 struct inet6_fill_args fillargs = {
5431 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5435 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5437 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5438 WARN_ON(err == -EMSGSIZE);
5442 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5446 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5449 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5450 __s32 *array, int bytes)
5452 BUG_ON(bytes < (DEVCONF_MAX * 4));
5454 memset(array, 0, bytes);
5455 array[DEVCONF_FORWARDING] = cnf->forwarding;
5456 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5457 array[DEVCONF_MTU6] = cnf->mtu6;
5458 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5459 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5460 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5461 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5462 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5463 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5464 jiffies_to_msecs(cnf->rtr_solicit_interval);
5465 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5466 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5467 array[DEVCONF_RTR_SOLICIT_DELAY] =
5468 jiffies_to_msecs(cnf->rtr_solicit_delay);
5469 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5470 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5471 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5472 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5473 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5474 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5475 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5476 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5477 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5478 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5479 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5480 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5481 array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5482 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5483 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5484 #ifdef CONFIG_IPV6_ROUTER_PREF
5485 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5486 array[DEVCONF_RTR_PROBE_INTERVAL] =
5487 jiffies_to_msecs(cnf->rtr_probe_interval);
5488 #ifdef CONFIG_IPV6_ROUTE_INFO
5489 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5490 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5493 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5494 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5495 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5496 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5497 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5499 #ifdef CONFIG_IPV6_MROUTE
5500 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5502 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5503 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5504 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5505 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5506 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5507 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5508 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5509 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5510 /* we omit DEVCONF_STABLE_SECRET for now */
5511 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5512 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5513 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5514 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5515 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5516 #ifdef CONFIG_IPV6_SEG6_HMAC
5517 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5519 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5520 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5521 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5522 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5523 array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5526 static inline size_t inet6_ifla6_size(void)
5528 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5529 + nla_total_size(sizeof(struct ifla_cacheinfo))
5530 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5531 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5532 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5533 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5534 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5538 static inline size_t inet6_if_nlmsg_size(void)
5540 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5541 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5542 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5543 + nla_total_size(4) /* IFLA_MTU */
5544 + nla_total_size(4) /* IFLA_LINK */
5545 + nla_total_size(1) /* IFLA_OPERSTATE */
5546 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5549 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5553 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5556 /* Use put_unaligned() because stats may not be aligned for u64. */
5557 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5558 for (i = 1; i < ICMP6_MIB_MAX; i++)
5559 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5561 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5564 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5565 int bytes, size_t syncpoff)
5568 u64 buff[IPSTATS_MIB_MAX];
5569 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5573 memset(buff, 0, sizeof(buff));
5574 buff[0] = IPSTATS_MIB_MAX;
5576 for_each_possible_cpu(c) {
5577 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5578 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5581 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5582 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5585 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5589 case IFLA_INET6_STATS:
5590 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5591 offsetof(struct ipstats_mib, syncp));
5593 case IFLA_INET6_ICMP6STATS:
5594 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5599 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5600 u32 ext_filter_mask)
5603 struct ifla_cacheinfo ci;
5605 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5606 goto nla_put_failure;
5607 ci.max_reasm_len = IPV6_MAXPLEN;
5608 ci.tstamp = cstamp_delta(idev->tstamp);
5609 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5610 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5611 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5612 goto nla_put_failure;
5613 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5615 goto nla_put_failure;
5616 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5618 /* XXX - MC not implemented */
5620 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5623 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5625 goto nla_put_failure;
5626 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5628 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5630 goto nla_put_failure;
5631 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5633 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5635 goto nla_put_failure;
5636 read_lock_bh(&idev->lock);
5637 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5638 read_unlock_bh(&idev->lock);
5640 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5641 goto nla_put_failure;
5649 static size_t inet6_get_link_af_size(const struct net_device *dev,
5650 u32 ext_filter_mask)
5652 if (!__in6_dev_get(dev))
5655 return inet6_ifla6_size();
5658 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5659 u32 ext_filter_mask)
5661 struct inet6_dev *idev = __in6_dev_get(dev);
5666 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5672 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5673 struct netlink_ext_ack *extack)
5675 struct inet6_ifaddr *ifp;
5676 struct net_device *dev = idev->dev;
5677 bool clear_token, update_rs = false;
5678 struct in6_addr ll_addr;
5685 if (dev->flags & IFF_LOOPBACK) {
5686 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5690 if (dev->flags & IFF_NOARP) {
5691 NL_SET_ERR_MSG_MOD(extack,
5692 "Device does not do neighbour discovery");
5696 if (!ipv6_accept_ra(idev)) {
5697 NL_SET_ERR_MSG_MOD(extack,
5698 "Router advertisement is disabled on device");
5702 if (idev->cnf.rtr_solicits == 0) {
5703 NL_SET_ERR_MSG(extack,
5704 "Router solicitation is disabled on device");
5708 write_lock_bh(&idev->lock);
5710 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5711 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5713 write_unlock_bh(&idev->lock);
5715 clear_token = ipv6_addr_any(token);
5719 if (!idev->dead && (idev->if_flags & IF_READY) &&
5720 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5721 IFA_F_OPTIMISTIC)) {
5722 /* If we're not ready, then normal ifup will take care
5723 * of this. Otherwise, we need to request our rs here.
5725 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5730 write_lock_bh(&idev->lock);
5733 idev->if_flags |= IF_RS_SENT;
5734 idev->rs_interval = rfc3315_s14_backoff_init(
5735 idev->cnf.rtr_solicit_interval);
5736 idev->rs_probes = 1;
5737 addrconf_mod_rs_timer(idev, idev->rs_interval);
5740 /* Well, that's kinda nasty ... */
5741 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5742 spin_lock(&ifp->lock);
5743 if (ifp->tokenized) {
5745 ifp->prefered_lft = 0;
5747 spin_unlock(&ifp->lock);
5750 write_unlock_bh(&idev->lock);
5751 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5752 addrconf_verify_rtnl();
5756 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5757 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5758 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5761 static int check_addr_gen_mode(int mode)
5763 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5764 mode != IN6_ADDR_GEN_MODE_NONE &&
5765 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5766 mode != IN6_ADDR_GEN_MODE_RANDOM)
5771 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5774 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5775 !idev->cnf.stable_secret.initialized &&
5776 !net->ipv6.devconf_dflt->stable_secret.initialized)
5781 static int inet6_validate_link_af(const struct net_device *dev,
5782 const struct nlattr *nla)
5784 struct nlattr *tb[IFLA_INET6_MAX + 1];
5785 struct inet6_dev *idev = NULL;
5789 idev = __in6_dev_get(dev);
5791 return -EAFNOSUPPORT;
5794 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5795 inet6_af_policy, NULL);
5799 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5802 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5803 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5805 if (check_addr_gen_mode(mode) < 0)
5807 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5814 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5815 struct netlink_ext_ack *extack)
5817 struct inet6_dev *idev = __in6_dev_get(dev);
5818 struct nlattr *tb[IFLA_INET6_MAX + 1];
5822 return -EAFNOSUPPORT;
5824 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5827 if (tb[IFLA_INET6_TOKEN]) {
5828 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5834 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5835 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5837 idev->cnf.addr_gen_mode = mode;
5843 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5844 u32 portid, u32 seq, int event, unsigned int flags)
5846 struct net_device *dev = idev->dev;
5847 struct ifinfomsg *hdr;
5848 struct nlmsghdr *nlh;
5851 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5855 hdr = nlmsg_data(nlh);
5856 hdr->ifi_family = AF_INET6;
5858 hdr->ifi_type = dev->type;
5859 hdr->ifi_index = dev->ifindex;
5860 hdr->ifi_flags = dev_get_flags(dev);
5861 hdr->ifi_change = 0;
5863 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5865 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5866 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5867 (dev->ifindex != dev_get_iflink(dev) &&
5868 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5869 nla_put_u8(skb, IFLA_OPERSTATE,
5870 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5871 goto nla_put_failure;
5872 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5874 goto nla_put_failure;
5876 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5877 goto nla_put_failure;
5879 nla_nest_end(skb, protoinfo);
5880 nlmsg_end(skb, nlh);
5884 nlmsg_cancel(skb, nlh);
5888 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5889 struct netlink_ext_ack *extack)
5891 struct ifinfomsg *ifm;
5893 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5894 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5898 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5899 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5903 ifm = nlmsg_data(nlh);
5904 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5905 ifm->ifi_change || ifm->ifi_index) {
5906 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5913 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5915 struct net *net = sock_net(skb->sk);
5918 struct net_device *dev;
5919 struct inet6_dev *idev;
5920 struct hlist_head *head;
5922 /* only requests using strict checking can pass data to
5923 * influence the dump
5925 if (cb->strict_check) {
5926 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5933 s_idx = cb->args[1];
5936 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5938 head = &net->dev_index_head[h];
5939 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5942 idev = __in6_dev_get(dev);
5945 if (inet6_fill_ifinfo(skb, idev,
5946 NETLINK_CB(cb->skb).portid,
5948 RTM_NEWLINK, NLM_F_MULTI) < 0)
5962 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5964 struct sk_buff *skb;
5965 struct net *net = dev_net(idev->dev);
5968 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5972 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5974 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5975 WARN_ON(err == -EMSGSIZE);
5979 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5983 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5986 static inline size_t inet6_prefix_nlmsg_size(void)
5988 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5989 + nla_total_size(sizeof(struct in6_addr))
5990 + nla_total_size(sizeof(struct prefix_cacheinfo));
5993 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5994 struct prefix_info *pinfo, u32 portid, u32 seq,
5995 int event, unsigned int flags)
5997 struct prefixmsg *pmsg;
5998 struct nlmsghdr *nlh;
5999 struct prefix_cacheinfo ci;
6001 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6005 pmsg = nlmsg_data(nlh);
6006 pmsg->prefix_family = AF_INET6;
6007 pmsg->prefix_pad1 = 0;
6008 pmsg->prefix_pad2 = 0;
6009 pmsg->prefix_ifindex = idev->dev->ifindex;
6010 pmsg->prefix_len = pinfo->prefix_len;
6011 pmsg->prefix_type = pinfo->type;
6012 pmsg->prefix_pad3 = 0;
6013 pmsg->prefix_flags = 0;
6015 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6016 if (pinfo->autoconf)
6017 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6019 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6020 goto nla_put_failure;
6021 ci.preferred_time = ntohl(pinfo->prefered);
6022 ci.valid_time = ntohl(pinfo->valid);
6023 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6024 goto nla_put_failure;
6025 nlmsg_end(skb, nlh);
6029 nlmsg_cancel(skb, nlh);
6033 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6034 struct prefix_info *pinfo)
6036 struct sk_buff *skb;
6037 struct net *net = dev_net(idev->dev);
6040 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6044 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6046 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6047 WARN_ON(err == -EMSGSIZE);
6051 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6055 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6058 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6060 struct net *net = dev_net(ifp->idev->dev);
6065 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6070 * If the address was optimistic we inserted the route at the
6071 * start of our DAD process, so we don't need to do it again.
6072 * If the device was taken down in the middle of the DAD
6073 * cycle there is a race where we could get here without a
6074 * host route, so nothing to insert. That will be fixed when
6075 * the device is brought up.
6077 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6078 ip6_ins_rt(net, ifp->rt);
6079 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6080 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6081 &ifp->addr, ifp->idev->dev->name);
6084 if (ifp->idev->cnf.forwarding)
6085 addrconf_join_anycast(ifp);
6086 if (!ipv6_addr_any(&ifp->peer_addr))
6087 addrconf_prefix_route(&ifp->peer_addr, 128,
6088 ifp->rt_priority, ifp->idev->dev,
6092 if (ifp->idev->cnf.forwarding)
6093 addrconf_leave_anycast(ifp);
6094 addrconf_leave_solict(ifp->idev, &ifp->addr);
6095 if (!ipv6_addr_any(&ifp->peer_addr)) {
6096 struct fib6_info *rt;
6098 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6099 ifp->idev->dev, 0, 0,
6102 ip6_del_rt(net, rt, false);
6105 ip6_del_rt(net, ifp->rt, false);
6108 rt_genid_bump_ipv6(net);
6111 atomic_inc(&net->ipv6.dev_addr_genid);
6114 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6117 if (likely(ifp->idev->dead == 0))
6118 __ipv6_ifa_notify(event, ifp);
6119 rcu_read_unlock_bh();
6122 #ifdef CONFIG_SYSCTL
6124 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6125 void *buffer, size_t *lenp, loff_t *ppos)
6127 int *valp = ctl->data;
6130 struct ctl_table lctl;
6134 * ctl->data points to idev->cnf.forwarding, we should
6135 * not modify it until we get the rtnl lock.
6140 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6143 ret = addrconf_fixup_forwarding(ctl, valp, val);
6149 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6150 void *buffer, size_t *lenp, loff_t *ppos)
6152 struct inet6_dev *idev = ctl->extra1;
6153 int min_mtu = IPV6_MIN_MTU;
6154 struct ctl_table lctl;
6157 lctl.extra1 = &min_mtu;
6158 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6160 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6163 static void dev_disable_change(struct inet6_dev *idev)
6165 struct netdev_notifier_info info;
6167 if (!idev || !idev->dev)
6170 netdev_notifier_info_init(&info, idev->dev);
6171 if (idev->cnf.disable_ipv6)
6172 addrconf_notify(NULL, NETDEV_DOWN, &info);
6174 addrconf_notify(NULL, NETDEV_UP, &info);
6177 static void addrconf_disable_change(struct net *net, __s32 newf)
6179 struct net_device *dev;
6180 struct inet6_dev *idev;
6182 for_each_netdev(net, dev) {
6183 idev = __in6_dev_get(dev);
6185 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6186 idev->cnf.disable_ipv6 = newf;
6188 dev_disable_change(idev);
6193 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6198 if (!rtnl_trylock())
6199 return restart_syscall();
6201 net = (struct net *)table->extra2;
6205 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6210 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6211 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6212 addrconf_disable_change(net, newf);
6213 } else if ((!newf) ^ (!old))
6214 dev_disable_change((struct inet6_dev *)table->extra1);
6220 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6221 void *buffer, size_t *lenp, loff_t *ppos)
6223 int *valp = ctl->data;
6226 struct ctl_table lctl;
6230 * ctl->data points to idev->cnf.disable_ipv6, we should
6231 * not modify it until we get the rtnl lock.
6236 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6239 ret = addrconf_disable_ipv6(ctl, valp, val);
6245 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6246 void *buffer, size_t *lenp, loff_t *ppos)
6248 int *valp = ctl->data;
6253 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6256 if (write && old != new) {
6257 struct net *net = ctl->extra2;
6259 if (!rtnl_trylock())
6260 return restart_syscall();
6262 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6263 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6264 NETCONFA_PROXY_NEIGH,
6265 NETCONFA_IFINDEX_DEFAULT,
6266 net->ipv6.devconf_dflt);
6267 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6268 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6269 NETCONFA_PROXY_NEIGH,
6270 NETCONFA_IFINDEX_ALL,
6271 net->ipv6.devconf_all);
6273 struct inet6_dev *idev = ctl->extra1;
6275 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6276 NETCONFA_PROXY_NEIGH,
6286 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6287 void *buffer, size_t *lenp,
6292 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6293 struct net *net = (struct net *)ctl->extra2;
6294 struct ctl_table tmp = {
6296 .maxlen = sizeof(new_val),
6300 if (!rtnl_trylock())
6301 return restart_syscall();
6303 new_val = *((u32 *)ctl->data);
6305 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6310 if (check_addr_gen_mode(new_val) < 0) {
6316 if (check_stable_privacy(idev, net, new_val) < 0) {
6321 if (idev->cnf.addr_gen_mode != new_val) {
6322 idev->cnf.addr_gen_mode = new_val;
6323 addrconf_dev_config(idev->dev);
6325 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6326 struct net_device *dev;
6328 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6329 for_each_netdev(net, dev) {
6330 idev = __in6_dev_get(dev);
6332 idev->cnf.addr_gen_mode != new_val) {
6333 idev->cnf.addr_gen_mode = new_val;
6334 addrconf_dev_config(idev->dev);
6339 *((u32 *)ctl->data) = new_val;
6348 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6349 void *buffer, size_t *lenp,
6353 struct in6_addr addr;
6354 char str[IPV6_MAX_STRLEN];
6355 struct ctl_table lctl = *ctl;
6356 struct net *net = ctl->extra2;
6357 struct ipv6_stable_secret *secret = ctl->data;
6359 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6362 lctl.maxlen = IPV6_MAX_STRLEN;
6365 if (!rtnl_trylock())
6366 return restart_syscall();
6368 if (!write && !secret->initialized) {
6373 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6374 if (err >= sizeof(str)) {
6379 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6383 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6388 secret->initialized = true;
6389 secret->secret = addr;
6391 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6392 struct net_device *dev;
6394 for_each_netdev(net, dev) {
6395 struct inet6_dev *idev = __in6_dev_get(dev);
6398 idev->cnf.addr_gen_mode =
6399 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6403 struct inet6_dev *idev = ctl->extra1;
6405 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6415 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6416 int write, void *buffer,
6420 int *valp = ctl->data;
6423 struct ctl_table lctl;
6426 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6427 * we should not modify it until we get the rtnl lock.
6432 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6435 ret = addrconf_fixup_linkdown(ctl, valp, val);
6442 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6446 rt->dst.flags |= DST_NOPOLICY;
6448 rt->dst.flags &= ~DST_NOPOLICY;
6453 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6455 struct inet6_ifaddr *ifa;
6457 read_lock_bh(&idev->lock);
6458 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6459 spin_lock(&ifa->lock);
6461 /* host routes only use builtin fib6_nh */
6462 struct fib6_nh *nh = ifa->rt->fib6_nh;
6466 ifa->rt->dst_nopolicy = val ? true : false;
6467 if (nh->rt6i_pcpu) {
6468 for_each_possible_cpu(cpu) {
6469 struct rt6_info **rtp;
6471 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6472 addrconf_set_nopolicy(*rtp, val);
6477 spin_unlock(&ifa->lock);
6479 read_unlock_bh(&idev->lock);
6483 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6485 struct inet6_dev *idev;
6488 if (!rtnl_trylock())
6489 return restart_syscall();
6493 net = (struct net *)ctl->extra2;
6494 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6499 if (valp == &net->ipv6.devconf_all->disable_policy) {
6500 struct net_device *dev;
6502 for_each_netdev(net, dev) {
6503 idev = __in6_dev_get(dev);
6505 addrconf_disable_policy_idev(idev, val);
6508 idev = (struct inet6_dev *)ctl->extra1;
6509 addrconf_disable_policy_idev(idev, val);
6516 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6517 void *buffer, size_t *lenp, loff_t *ppos)
6519 int *valp = ctl->data;
6522 struct ctl_table lctl;
6527 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6529 if (write && (*valp != val))
6530 ret = addrconf_disable_policy(ctl, valp, val);
6538 static int minus_one = -1;
6539 static const int two_five_five = 255;
6541 static const struct ctl_table addrconf_sysctl[] = {
6543 .procname = "forwarding",
6544 .data = &ipv6_devconf.forwarding,
6545 .maxlen = sizeof(int),
6547 .proc_handler = addrconf_sysctl_forward,
6550 .procname = "hop_limit",
6551 .data = &ipv6_devconf.hop_limit,
6552 .maxlen = sizeof(int),
6554 .proc_handler = proc_dointvec_minmax,
6555 .extra1 = (void *)SYSCTL_ONE,
6556 .extra2 = (void *)&two_five_five,
6560 .data = &ipv6_devconf.mtu6,
6561 .maxlen = sizeof(int),
6563 .proc_handler = addrconf_sysctl_mtu,
6566 .procname = "accept_ra",
6567 .data = &ipv6_devconf.accept_ra,
6568 .maxlen = sizeof(int),
6570 .proc_handler = proc_dointvec,
6573 .procname = "accept_redirects",
6574 .data = &ipv6_devconf.accept_redirects,
6575 .maxlen = sizeof(int),
6577 .proc_handler = proc_dointvec,
6580 .procname = "autoconf",
6581 .data = &ipv6_devconf.autoconf,
6582 .maxlen = sizeof(int),
6584 .proc_handler = proc_dointvec,
6587 .procname = "dad_transmits",
6588 .data = &ipv6_devconf.dad_transmits,
6589 .maxlen = sizeof(int),
6591 .proc_handler = proc_dointvec,
6594 .procname = "router_solicitations",
6595 .data = &ipv6_devconf.rtr_solicits,
6596 .maxlen = sizeof(int),
6598 .proc_handler = proc_dointvec_minmax,
6599 .extra1 = &minus_one,
6602 .procname = "router_solicitation_interval",
6603 .data = &ipv6_devconf.rtr_solicit_interval,
6604 .maxlen = sizeof(int),
6606 .proc_handler = proc_dointvec_jiffies,
6609 .procname = "router_solicitation_max_interval",
6610 .data = &ipv6_devconf.rtr_solicit_max_interval,
6611 .maxlen = sizeof(int),
6613 .proc_handler = proc_dointvec_jiffies,
6616 .procname = "router_solicitation_delay",
6617 .data = &ipv6_devconf.rtr_solicit_delay,
6618 .maxlen = sizeof(int),
6620 .proc_handler = proc_dointvec_jiffies,
6623 .procname = "force_mld_version",
6624 .data = &ipv6_devconf.force_mld_version,
6625 .maxlen = sizeof(int),
6627 .proc_handler = proc_dointvec,
6630 .procname = "mldv1_unsolicited_report_interval",
6632 &ipv6_devconf.mldv1_unsolicited_report_interval,
6633 .maxlen = sizeof(int),
6635 .proc_handler = proc_dointvec_ms_jiffies,
6638 .procname = "mldv2_unsolicited_report_interval",
6640 &ipv6_devconf.mldv2_unsolicited_report_interval,
6641 .maxlen = sizeof(int),
6643 .proc_handler = proc_dointvec_ms_jiffies,
6646 .procname = "use_tempaddr",
6647 .data = &ipv6_devconf.use_tempaddr,
6648 .maxlen = sizeof(int),
6650 .proc_handler = proc_dointvec,
6653 .procname = "temp_valid_lft",
6654 .data = &ipv6_devconf.temp_valid_lft,
6655 .maxlen = sizeof(int),
6657 .proc_handler = proc_dointvec,
6660 .procname = "temp_prefered_lft",
6661 .data = &ipv6_devconf.temp_prefered_lft,
6662 .maxlen = sizeof(int),
6664 .proc_handler = proc_dointvec,
6667 .procname = "regen_max_retry",
6668 .data = &ipv6_devconf.regen_max_retry,
6669 .maxlen = sizeof(int),
6671 .proc_handler = proc_dointvec,
6674 .procname = "max_desync_factor",
6675 .data = &ipv6_devconf.max_desync_factor,
6676 .maxlen = sizeof(int),
6678 .proc_handler = proc_dointvec,
6681 .procname = "max_addresses",
6682 .data = &ipv6_devconf.max_addresses,
6683 .maxlen = sizeof(int),
6685 .proc_handler = proc_dointvec,
6688 .procname = "accept_ra_defrtr",
6689 .data = &ipv6_devconf.accept_ra_defrtr,
6690 .maxlen = sizeof(int),
6692 .proc_handler = proc_dointvec,
6695 .procname = "ra_defrtr_metric",
6696 .data = &ipv6_devconf.ra_defrtr_metric,
6697 .maxlen = sizeof(u32),
6699 .proc_handler = proc_douintvec_minmax,
6700 .extra1 = (void *)SYSCTL_ONE,
6703 .procname = "accept_ra_min_hop_limit",
6704 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6705 .maxlen = sizeof(int),
6707 .proc_handler = proc_dointvec,
6710 .procname = "accept_ra_pinfo",
6711 .data = &ipv6_devconf.accept_ra_pinfo,
6712 .maxlen = sizeof(int),
6714 .proc_handler = proc_dointvec,
6716 #ifdef CONFIG_IPV6_ROUTER_PREF
6718 .procname = "accept_ra_rtr_pref",
6719 .data = &ipv6_devconf.accept_ra_rtr_pref,
6720 .maxlen = sizeof(int),
6722 .proc_handler = proc_dointvec,
6725 .procname = "router_probe_interval",
6726 .data = &ipv6_devconf.rtr_probe_interval,
6727 .maxlen = sizeof(int),
6729 .proc_handler = proc_dointvec_jiffies,
6731 #ifdef CONFIG_IPV6_ROUTE_INFO
6733 .procname = "accept_ra_rt_info_min_plen",
6734 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6735 .maxlen = sizeof(int),
6737 .proc_handler = proc_dointvec,
6740 .procname = "accept_ra_rt_info_max_plen",
6741 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6742 .maxlen = sizeof(int),
6744 .proc_handler = proc_dointvec,
6749 .procname = "proxy_ndp",
6750 .data = &ipv6_devconf.proxy_ndp,
6751 .maxlen = sizeof(int),
6753 .proc_handler = addrconf_sysctl_proxy_ndp,
6756 .procname = "accept_source_route",
6757 .data = &ipv6_devconf.accept_source_route,
6758 .maxlen = sizeof(int),
6760 .proc_handler = proc_dointvec,
6762 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6764 .procname = "optimistic_dad",
6765 .data = &ipv6_devconf.optimistic_dad,
6766 .maxlen = sizeof(int),
6768 .proc_handler = proc_dointvec,
6771 .procname = "use_optimistic",
6772 .data = &ipv6_devconf.use_optimistic,
6773 .maxlen = sizeof(int),
6775 .proc_handler = proc_dointvec,
6778 #ifdef CONFIG_IPV6_MROUTE
6780 .procname = "mc_forwarding",
6781 .data = &ipv6_devconf.mc_forwarding,
6782 .maxlen = sizeof(int),
6784 .proc_handler = proc_dointvec,
6788 .procname = "disable_ipv6",
6789 .data = &ipv6_devconf.disable_ipv6,
6790 .maxlen = sizeof(int),
6792 .proc_handler = addrconf_sysctl_disable,
6795 .procname = "accept_dad",
6796 .data = &ipv6_devconf.accept_dad,
6797 .maxlen = sizeof(int),
6799 .proc_handler = proc_dointvec,
6802 .procname = "force_tllao",
6803 .data = &ipv6_devconf.force_tllao,
6804 .maxlen = sizeof(int),
6806 .proc_handler = proc_dointvec
6809 .procname = "ndisc_notify",
6810 .data = &ipv6_devconf.ndisc_notify,
6811 .maxlen = sizeof(int),
6813 .proc_handler = proc_dointvec
6816 .procname = "suppress_frag_ndisc",
6817 .data = &ipv6_devconf.suppress_frag_ndisc,
6818 .maxlen = sizeof(int),
6820 .proc_handler = proc_dointvec
6823 .procname = "accept_ra_from_local",
6824 .data = &ipv6_devconf.accept_ra_from_local,
6825 .maxlen = sizeof(int),
6827 .proc_handler = proc_dointvec,
6830 .procname = "accept_ra_mtu",
6831 .data = &ipv6_devconf.accept_ra_mtu,
6832 .maxlen = sizeof(int),
6834 .proc_handler = proc_dointvec,
6837 .procname = "stable_secret",
6838 .data = &ipv6_devconf.stable_secret,
6839 .maxlen = IPV6_MAX_STRLEN,
6841 .proc_handler = addrconf_sysctl_stable_secret,
6844 .procname = "use_oif_addrs_only",
6845 .data = &ipv6_devconf.use_oif_addrs_only,
6846 .maxlen = sizeof(int),
6848 .proc_handler = proc_dointvec,
6851 .procname = "ignore_routes_with_linkdown",
6852 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6853 .maxlen = sizeof(int),
6855 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6858 .procname = "drop_unicast_in_l2_multicast",
6859 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6860 .maxlen = sizeof(int),
6862 .proc_handler = proc_dointvec,
6865 .procname = "drop_unsolicited_na",
6866 .data = &ipv6_devconf.drop_unsolicited_na,
6867 .maxlen = sizeof(int),
6869 .proc_handler = proc_dointvec,
6872 .procname = "keep_addr_on_down",
6873 .data = &ipv6_devconf.keep_addr_on_down,
6874 .maxlen = sizeof(int),
6876 .proc_handler = proc_dointvec,
6880 .procname = "seg6_enabled",
6881 .data = &ipv6_devconf.seg6_enabled,
6882 .maxlen = sizeof(int),
6884 .proc_handler = proc_dointvec,
6886 #ifdef CONFIG_IPV6_SEG6_HMAC
6888 .procname = "seg6_require_hmac",
6889 .data = &ipv6_devconf.seg6_require_hmac,
6890 .maxlen = sizeof(int),
6892 .proc_handler = proc_dointvec,
6896 .procname = "enhanced_dad",
6897 .data = &ipv6_devconf.enhanced_dad,
6898 .maxlen = sizeof(int),
6900 .proc_handler = proc_dointvec,
6903 .procname = "addr_gen_mode",
6904 .data = &ipv6_devconf.addr_gen_mode,
6905 .maxlen = sizeof(int),
6907 .proc_handler = addrconf_sysctl_addr_gen_mode,
6910 .procname = "disable_policy",
6911 .data = &ipv6_devconf.disable_policy,
6912 .maxlen = sizeof(int),
6914 .proc_handler = addrconf_sysctl_disable_policy,
6917 .procname = "ndisc_tclass",
6918 .data = &ipv6_devconf.ndisc_tclass,
6919 .maxlen = sizeof(int),
6921 .proc_handler = proc_dointvec_minmax,
6922 .extra1 = (void *)SYSCTL_ZERO,
6923 .extra2 = (void *)&two_five_five,
6926 .procname = "rpl_seg_enabled",
6927 .data = &ipv6_devconf.rpl_seg_enabled,
6928 .maxlen = sizeof(int),
6930 .proc_handler = proc_dointvec,
6937 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6938 struct inet6_dev *idev, struct ipv6_devconf *p)
6941 struct ctl_table *table;
6942 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6944 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6948 for (i = 0; table[i].data; i++) {
6949 table[i].data += (char *)p - (char *)&ipv6_devconf;
6950 /* If one of these is already set, then it is not safe to
6951 * overwrite either of them: this makes proc_dointvec_minmax
6954 if (!table[i].extra1 && !table[i].extra2) {
6955 table[i].extra1 = idev; /* embedded; no ref */
6956 table[i].extra2 = net;
6960 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6962 p->sysctl_header = register_net_sysctl(net, path, table);
6963 if (!p->sysctl_header)
6966 if (!strcmp(dev_name, "all"))
6967 ifindex = NETCONFA_IFINDEX_ALL;
6968 else if (!strcmp(dev_name, "default"))
6969 ifindex = NETCONFA_IFINDEX_DEFAULT;
6971 ifindex = idev->dev->ifindex;
6972 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6982 static void __addrconf_sysctl_unregister(struct net *net,
6983 struct ipv6_devconf *p, int ifindex)
6985 struct ctl_table *table;
6987 if (!p->sysctl_header)
6990 table = p->sysctl_header->ctl_table_arg;
6991 unregister_net_sysctl_table(p->sysctl_header);
6992 p->sysctl_header = NULL;
6995 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6998 static int addrconf_sysctl_register(struct inet6_dev *idev)
7002 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7005 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7006 &ndisc_ifinfo_sysctl_change);
7009 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7012 neigh_sysctl_unregister(idev->nd_parms);
7017 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7019 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7020 idev->dev->ifindex);
7021 neigh_sysctl_unregister(idev->nd_parms);
7027 static int __net_init addrconf_init_net(struct net *net)
7030 struct ipv6_devconf *all, *dflt;
7032 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7036 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7038 goto err_alloc_dflt;
7040 if (IS_ENABLED(CONFIG_SYSCTL) &&
7041 !net_eq(net, &init_net)) {
7042 switch (sysctl_devconf_inherit_init_net) {
7043 case 1: /* copy from init_net */
7044 memcpy(all, init_net.ipv6.devconf_all,
7045 sizeof(ipv6_devconf));
7046 memcpy(dflt, init_net.ipv6.devconf_dflt,
7047 sizeof(ipv6_devconf_dflt));
7049 case 3: /* copy from the current netns */
7050 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7051 sizeof(ipv6_devconf));
7053 current->nsproxy->net_ns->ipv6.devconf_dflt,
7054 sizeof(ipv6_devconf_dflt));
7058 /* use compiled values */
7063 /* these will be inherited by all namespaces */
7064 dflt->autoconf = ipv6_defaults.autoconf;
7065 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7067 dflt->stable_secret.initialized = false;
7068 all->stable_secret.initialized = false;
7070 net->ipv6.devconf_all = all;
7071 net->ipv6.devconf_dflt = dflt;
7073 #ifdef CONFIG_SYSCTL
7074 err = __addrconf_sysctl_register(net, "all", NULL, all);
7078 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7084 #ifdef CONFIG_SYSCTL
7086 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7096 static void __net_exit addrconf_exit_net(struct net *net)
7098 #ifdef CONFIG_SYSCTL
7099 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7100 NETCONFA_IFINDEX_DEFAULT);
7101 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7102 NETCONFA_IFINDEX_ALL);
7104 kfree(net->ipv6.devconf_dflt);
7105 kfree(net->ipv6.devconf_all);
7108 static struct pernet_operations addrconf_ops = {
7109 .init = addrconf_init_net,
7110 .exit = addrconf_exit_net,
7113 static struct rtnl_af_ops inet6_ops __read_mostly = {
7115 .fill_link_af = inet6_fill_link_af,
7116 .get_link_af_size = inet6_get_link_af_size,
7117 .validate_link_af = inet6_validate_link_af,
7118 .set_link_af = inet6_set_link_af,
7122 * Init / cleanup code
7125 int __init addrconf_init(void)
7127 struct inet6_dev *idev;
7130 err = ipv6_addr_label_init();
7132 pr_crit("%s: cannot initialize default policy table: %d\n",
7137 err = register_pernet_subsys(&addrconf_ops);
7141 addrconf_wq = create_workqueue("ipv6_addrconf");
7147 /* The addrconf netdev notifier requires that loopback_dev
7148 * has it's ipv6 private information allocated and setup
7149 * before it can bring up and give link-local addresses
7150 * to other devices which are up.
7152 * Unfortunately, loopback_dev is not necessarily the first
7153 * entry in the global dev_base list of net devices. In fact,
7154 * it is likely to be the very last entry on that list.
7155 * So this causes the notifier registry below to try and
7156 * give link-local addresses to all devices besides loopback_dev
7157 * first, then loopback_dev, which cases all the non-loopback_dev
7158 * devices to fail to get a link-local address.
7160 * So, as a temporary fix, allocate the ipv6 structure for
7161 * loopback_dev first by hand.
7162 * Longer term, all of the dependencies ipv6 has upon the loopback
7163 * device and it being up should be removed.
7166 idev = ipv6_add_dev(init_net.loopback_dev);
7169 err = PTR_ERR(idev);
7173 ip6_route_init_special_entries();
7175 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7176 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7178 register_netdevice_notifier(&ipv6_dev_notf);
7182 rtnl_af_register(&inet6_ops);
7184 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7185 NULL, inet6_dump_ifinfo, 0);
7189 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7190 inet6_rtm_newaddr, NULL, 0);
7193 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7194 inet6_rtm_deladdr, NULL, 0);
7197 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7198 inet6_rtm_getaddr, inet6_dump_ifaddr,
7199 RTNL_FLAG_DOIT_UNLOCKED);
7202 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7203 NULL, inet6_dump_ifmcaddr, 0);
7206 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7207 NULL, inet6_dump_ifacaddr, 0);
7210 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7211 inet6_netconf_get_devconf,
7212 inet6_netconf_dump_devconf,
7213 RTNL_FLAG_DOIT_UNLOCKED);
7216 err = ipv6_addr_label_rtnl_register();
7222 rtnl_unregister_all(PF_INET6);
7223 rtnl_af_unregister(&inet6_ops);
7224 unregister_netdevice_notifier(&ipv6_dev_notf);
7226 destroy_workqueue(addrconf_wq);
7228 unregister_pernet_subsys(&addrconf_ops);
7230 ipv6_addr_label_cleanup();
7235 void addrconf_cleanup(void)
7237 struct net_device *dev;
7240 unregister_netdevice_notifier(&ipv6_dev_notf);
7241 unregister_pernet_subsys(&addrconf_ops);
7242 ipv6_addr_label_cleanup();
7244 rtnl_af_unregister(&inet6_ops);
7248 /* clean dev list */
7249 for_each_netdev(&init_net, dev) {
7250 if (__in6_dev_get(dev) == NULL)
7252 addrconf_ifdown(dev, true);
7254 addrconf_ifdown(init_net.loopback_dev, true);
7259 spin_lock_bh(&addrconf_hash_lock);
7260 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7261 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7262 spin_unlock_bh(&addrconf_hash_lock);
7263 cancel_delayed_work(&addr_chk_work);
7266 destroy_workqueue(addrconf_wq);