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_regen_rndid(struct inet6_dev *idev);
139 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
145 const struct inet6_dev *idev);
147 #define IN6_ADDR_HSIZE_SHIFT 8
148 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
150 * Configured unicast address hash table
152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
153 static DEFINE_SPINLOCK(addrconf_hash_lock);
155 static void addrconf_verify(void);
156 static void addrconf_verify_rtnl(void);
157 static void addrconf_verify_work(struct work_struct *);
159 static struct workqueue_struct *addrconf_wq;
160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
165 static void addrconf_type_change(struct net_device *dev,
166 unsigned long event);
167 static int addrconf_ifdown(struct net_device *dev, int how);
169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
171 const struct net_device *dev,
172 u32 flags, u32 noflags,
175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
176 static void addrconf_dad_work(struct work_struct *w);
177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
180 static void addrconf_rs_timer(struct timer_list *t);
181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
185 struct prefix_info *pinfo);
187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
189 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
190 .mtu6 = IPV6_MIN_MTU,
192 .accept_redirects = 1,
194 .force_mld_version = 0,
195 .mldv1_unsolicited_report_interval = 10 * HZ,
196 .mldv2_unsolicited_report_interval = HZ,
198 .rtr_solicits = MAX_RTR_SOLICITATIONS,
199 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
200 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
201 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
203 .temp_valid_lft = TEMP_VALID_LIFETIME,
204 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
205 .regen_max_retry = REGEN_MAX_RETRY,
206 .max_desync_factor = MAX_DESYNC_FACTOR,
207 .max_addresses = IPV6_MAX_ADDRESSES,
208 .accept_ra_defrtr = 1,
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,
241 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
243 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
244 .mtu6 = IPV6_MIN_MTU,
246 .accept_redirects = 1,
248 .force_mld_version = 0,
249 .mldv1_unsolicited_report_interval = 10 * HZ,
250 .mldv2_unsolicited_report_interval = HZ,
252 .rtr_solicits = MAX_RTR_SOLICITATIONS,
253 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
254 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
255 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
257 .temp_valid_lft = TEMP_VALID_LIFETIME,
258 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
259 .regen_max_retry = REGEN_MAX_RETRY,
260 .max_desync_factor = MAX_DESYNC_FACTOR,
261 .max_addresses = IPV6_MAX_ADDRESSES,
262 .accept_ra_defrtr = 1,
263 .accept_ra_from_local = 0,
264 .accept_ra_min_hop_limit= 1,
265 .accept_ra_pinfo = 1,
266 #ifdef CONFIG_IPV6_ROUTER_PREF
267 .accept_ra_rtr_pref = 1,
268 .rtr_probe_interval = 60 * HZ,
269 #ifdef CONFIG_IPV6_ROUTE_INFO
270 .accept_ra_rt_info_min_plen = 0,
271 .accept_ra_rt_info_max_plen = 0,
275 .accept_source_route = 0, /* we do not accept RH0 by default. */
278 .suppress_frag_ndisc = 1,
281 .initialized = false,
283 .use_oif_addrs_only = 0,
284 .ignore_routes_with_linkdown = 0,
285 .keep_addr_on_down = 0,
287 #ifdef CONFIG_IPV6_SEG6_HMAC
288 .seg6_require_hmac = 0,
291 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
295 /* Check if link is ready: is it up and is a valid qdisc available */
296 static inline bool addrconf_link_ready(const struct net_device *dev)
298 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
301 static void addrconf_del_rs_timer(struct inet6_dev *idev)
303 if (del_timer(&idev->rs_timer))
307 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
309 if (cancel_delayed_work(&ifp->dad_work))
313 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
316 if (!timer_pending(&idev->rs_timer))
318 mod_timer(&idev->rs_timer, jiffies + when);
321 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
325 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
329 static int snmp6_alloc_dev(struct inet6_dev *idev)
333 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
334 if (!idev->stats.ipv6)
337 for_each_possible_cpu(i) {
338 struct ipstats_mib *addrconf_stats;
339 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
340 u64_stats_init(&addrconf_stats->syncp);
344 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
346 if (!idev->stats.icmpv6dev)
348 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
350 if (!idev->stats.icmpv6msgdev)
356 kfree(idev->stats.icmpv6dev);
358 free_percpu(idev->stats.ipv6);
363 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
365 struct inet6_dev *ndev;
370 if (dev->mtu < IPV6_MIN_MTU)
371 return ERR_PTR(-EINVAL);
373 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
377 rwlock_init(&ndev->lock);
379 INIT_LIST_HEAD(&ndev->addr_list);
380 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
381 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
383 if (ndev->cnf.stable_secret.initialized)
384 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
386 ndev->cnf.mtu6 = dev->mtu;
387 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
388 if (!ndev->nd_parms) {
392 if (ndev->cnf.forwarding)
393 dev_disable_lro(dev);
394 /* We refer to the device */
397 if (snmp6_alloc_dev(ndev) < 0) {
398 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
400 neigh_parms_release(&nd_tbl, ndev->nd_parms);
406 if (snmp6_register_dev(ndev) < 0) {
407 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
408 __func__, dev->name);
412 /* One reference from device. */
413 refcount_set(&ndev->refcnt, 1);
415 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
416 ndev->cnf.accept_dad = -1;
418 #if IS_ENABLED(CONFIG_IPV6_SIT)
419 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
420 pr_info("%s: Disabled Multicast RS\n", dev->name);
421 ndev->cnf.rtr_solicits = 0;
425 INIT_LIST_HEAD(&ndev->tempaddr_list);
426 ndev->desync_factor = U32_MAX;
427 if ((dev->flags&IFF_LOOPBACK) ||
428 dev->type == ARPHRD_TUNNEL ||
429 dev->type == ARPHRD_TUNNEL6 ||
430 dev->type == ARPHRD_SIT ||
431 dev->type == ARPHRD_NONE) {
432 ndev->cnf.use_tempaddr = -1;
434 ipv6_regen_rndid(ndev);
436 ndev->token = in6addr_any;
438 if (netif_running(dev) && addrconf_link_ready(dev))
439 ndev->if_flags |= IF_READY;
441 ipv6_mc_init_dev(ndev);
442 ndev->tstamp = jiffies;
443 err = addrconf_sysctl_register(ndev);
445 ipv6_mc_destroy_dev(ndev);
446 snmp6_unregister_dev(ndev);
449 /* protected by rtnl_lock */
450 rcu_assign_pointer(dev->ip6_ptr, ndev);
452 /* Join interface-local all-node multicast group */
453 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
455 /* Join all-node multicast group */
456 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
458 /* Join all-router multicast group if forwarding is set */
459 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
460 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
465 neigh_parms_release(&nd_tbl, ndev->nd_parms);
467 in6_dev_finish_destroy(ndev);
471 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
473 struct inet6_dev *idev;
477 idev = __in6_dev_get(dev);
479 idev = ipv6_add_dev(dev);
484 if (dev->flags&IFF_UP)
489 static int inet6_netconf_msgsize_devconf(int type)
491 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
492 + nla_total_size(4); /* NETCONFA_IFINDEX */
495 if (type == NETCONFA_ALL)
498 if (all || type == NETCONFA_FORWARDING)
499 size += nla_total_size(4);
500 #ifdef CONFIG_IPV6_MROUTE
501 if (all || type == NETCONFA_MC_FORWARDING)
502 size += nla_total_size(4);
504 if (all || type == NETCONFA_PROXY_NEIGH)
505 size += nla_total_size(4);
507 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
508 size += nla_total_size(4);
513 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
514 struct ipv6_devconf *devconf, u32 portid,
515 u32 seq, int event, unsigned int flags,
518 struct nlmsghdr *nlh;
519 struct netconfmsg *ncm;
522 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
527 if (type == NETCONFA_ALL)
530 ncm = nlmsg_data(nlh);
531 ncm->ncm_family = AF_INET6;
533 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
534 goto nla_put_failure;
539 if ((all || type == NETCONFA_FORWARDING) &&
540 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
541 goto nla_put_failure;
542 #ifdef CONFIG_IPV6_MROUTE
543 if ((all || type == NETCONFA_MC_FORWARDING) &&
544 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
545 devconf->mc_forwarding) < 0)
546 goto nla_put_failure;
548 if ((all || type == NETCONFA_PROXY_NEIGH) &&
549 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
550 goto nla_put_failure;
552 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
553 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
554 devconf->ignore_routes_with_linkdown) < 0)
555 goto nla_put_failure;
562 nlmsg_cancel(skb, nlh);
566 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
567 int ifindex, struct ipv6_devconf *devconf)
572 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
576 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
579 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
580 WARN_ON(err == -EMSGSIZE);
584 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
587 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
590 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
591 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
592 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
593 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
594 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
597 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
598 const struct nlmsghdr *nlh,
600 struct netlink_ext_ack *extack)
604 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
605 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
609 if (!netlink_strict_get_check(skb))
610 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
612 devconf_ipv6_policy, extack);
614 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
616 devconf_ipv6_policy, extack);
620 for (i = 0; i <= NETCONFA_MAX; i++) {
625 case NETCONFA_IFINDEX:
628 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
636 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
637 struct nlmsghdr *nlh,
638 struct netlink_ext_ack *extack)
640 struct net *net = sock_net(in_skb->sk);
641 struct nlattr *tb[NETCONFA_MAX+1];
642 struct inet6_dev *in6_dev = NULL;
643 struct net_device *dev = NULL;
645 struct ipv6_devconf *devconf;
649 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
653 if (!tb[NETCONFA_IFINDEX])
657 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
659 case NETCONFA_IFINDEX_ALL:
660 devconf = net->ipv6.devconf_all;
662 case NETCONFA_IFINDEX_DEFAULT:
663 devconf = net->ipv6.devconf_dflt;
666 dev = dev_get_by_index(net, ifindex);
669 in6_dev = in6_dev_get(dev);
672 devconf = &in6_dev->cnf;
677 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
681 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
682 NETLINK_CB(in_skb).portid,
683 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
686 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
687 WARN_ON(err == -EMSGSIZE);
691 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
694 in6_dev_put(in6_dev);
700 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
701 struct netlink_callback *cb)
703 const struct nlmsghdr *nlh = cb->nlh;
704 struct net *net = sock_net(skb->sk);
707 struct net_device *dev;
708 struct inet6_dev *idev;
709 struct hlist_head *head;
711 if (cb->strict_check) {
712 struct netlink_ext_ack *extack = cb->extack;
713 struct netconfmsg *ncm;
715 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
716 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
720 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
721 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
727 s_idx = idx = cb->args[1];
729 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
731 head = &net->dev_index_head[h];
733 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
735 hlist_for_each_entry_rcu(dev, head, index_hlist) {
738 idev = __in6_dev_get(dev);
742 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
744 NETLINK_CB(cb->skb).portid,
752 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
758 if (h == NETDEV_HASHENTRIES) {
759 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
760 net->ipv6.devconf_all,
761 NETLINK_CB(cb->skb).portid,
763 RTM_NEWNETCONF, NLM_F_MULTI,
769 if (h == NETDEV_HASHENTRIES + 1) {
770 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
771 net->ipv6.devconf_dflt,
772 NETLINK_CB(cb->skb).portid,
774 RTM_NEWNETCONF, NLM_F_MULTI,
788 static void dev_forward_change(struct inet6_dev *idev)
790 struct net_device *dev;
791 struct inet6_ifaddr *ifa;
796 if (idev->cnf.forwarding)
797 dev_disable_lro(dev);
798 if (dev->flags & IFF_MULTICAST) {
799 if (idev->cnf.forwarding) {
800 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
801 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
802 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
804 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
805 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
806 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
810 list_for_each_entry(ifa, &idev->addr_list, if_list) {
811 if (ifa->flags&IFA_F_TENTATIVE)
813 if (idev->cnf.forwarding)
814 addrconf_join_anycast(ifa);
816 addrconf_leave_anycast(ifa);
818 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
820 dev->ifindex, &idev->cnf);
824 static void addrconf_forward_change(struct net *net, __s32 newf)
826 struct net_device *dev;
827 struct inet6_dev *idev;
829 for_each_netdev(net, dev) {
830 idev = __in6_dev_get(dev);
832 int changed = (!idev->cnf.forwarding) ^ (!newf);
833 idev->cnf.forwarding = newf;
835 dev_forward_change(idev);
840 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
846 return restart_syscall();
848 net = (struct net *)table->extra2;
852 if (p == &net->ipv6.devconf_dflt->forwarding) {
853 if ((!newf) ^ (!old))
854 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
856 NETCONFA_IFINDEX_DEFAULT,
857 net->ipv6.devconf_dflt);
862 if (p == &net->ipv6.devconf_all->forwarding) {
863 int old_dflt = net->ipv6.devconf_dflt->forwarding;
865 net->ipv6.devconf_dflt->forwarding = newf;
866 if ((!newf) ^ (!old_dflt))
867 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
869 NETCONFA_IFINDEX_DEFAULT,
870 net->ipv6.devconf_dflt);
872 addrconf_forward_change(net, newf);
873 if ((!newf) ^ (!old))
874 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
876 NETCONFA_IFINDEX_ALL,
877 net->ipv6.devconf_all);
878 } else if ((!newf) ^ (!old))
879 dev_forward_change((struct inet6_dev *)table->extra1);
883 rt6_purge_dflt_routers(net);
887 static void addrconf_linkdown_change(struct net *net, __s32 newf)
889 struct net_device *dev;
890 struct inet6_dev *idev;
892 for_each_netdev(net, dev) {
893 idev = __in6_dev_get(dev);
895 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
897 idev->cnf.ignore_routes_with_linkdown = newf;
899 inet6_netconf_notify_devconf(dev_net(dev),
901 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
908 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
914 return restart_syscall();
916 net = (struct net *)table->extra2;
920 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
921 if ((!newf) ^ (!old))
922 inet6_netconf_notify_devconf(net,
924 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
925 NETCONFA_IFINDEX_DEFAULT,
926 net->ipv6.devconf_dflt);
931 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
932 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
933 addrconf_linkdown_change(net, newf);
934 if ((!newf) ^ (!old))
935 inet6_netconf_notify_devconf(net,
937 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
938 NETCONFA_IFINDEX_ALL,
939 net->ipv6.devconf_all);
948 /* Nobody refers to this ifaddr, destroy it */
949 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
951 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
953 #ifdef NET_REFCNT_DEBUG
954 pr_debug("%s\n", __func__);
957 in6_dev_put(ifp->idev);
959 if (cancel_delayed_work(&ifp->dad_work))
960 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
963 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
964 pr_warn("Freeing alive inet6 address %p\n", ifp);
972 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
975 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
978 * Each device address list is sorted in order of scope -
979 * global before linklocal.
981 list_for_each(p, &idev->addr_list) {
982 struct inet6_ifaddr *ifa
983 = list_entry(p, struct inet6_ifaddr, if_list);
984 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
988 list_add_tail_rcu(&ifp->if_list, p);
991 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
993 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
995 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
998 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
999 struct net_device *dev, unsigned int hash)
1001 struct inet6_ifaddr *ifp;
1003 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1004 if (!net_eq(dev_net(ifp->idev->dev), net))
1006 if (ipv6_addr_equal(&ifp->addr, addr)) {
1007 if (!dev || ifp->idev->dev == dev)
1014 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1016 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1019 spin_lock(&addrconf_hash_lock);
1021 /* Ignore adding duplicate addresses on an interface */
1022 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1023 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1026 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1029 spin_unlock(&addrconf_hash_lock);
1034 /* On success it returns ifp with increased reference count */
1036 static struct inet6_ifaddr *
1037 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1038 bool can_block, struct netlink_ext_ack *extack)
1040 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1041 int addr_type = ipv6_addr_type(cfg->pfx);
1042 struct net *net = dev_net(idev->dev);
1043 struct inet6_ifaddr *ifa = NULL;
1044 struct fib6_info *f6i = NULL;
1047 if (addr_type == IPV6_ADDR_ANY ||
1048 (addr_type & IPV6_ADDR_MULTICAST &&
1049 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1050 (!(idev->dev->flags & IFF_LOOPBACK) &&
1051 !netif_is_l3_master(idev->dev) &&
1052 addr_type & IPV6_ADDR_LOOPBACK))
1053 return ERR_PTR(-EADDRNOTAVAIL);
1056 err = -ENODEV; /*XXX*/
1060 if (idev->cnf.disable_ipv6) {
1065 /* validator notifier needs to be blocking;
1066 * do not call in atomic context
1069 struct in6_validator_info i6vi = {
1070 .i6vi_addr = *cfg->pfx,
1075 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1076 err = notifier_to_errno(err);
1081 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1087 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1094 if (net->ipv6.devconf_all->disable_policy ||
1095 idev->cnf.disable_policy)
1096 f6i->dst_nopolicy = true;
1098 neigh_parms_data_state_setall(idev->nd_parms);
1100 ifa->addr = *cfg->pfx;
1102 ifa->peer_addr = *cfg->peer_pfx;
1104 spin_lock_init(&ifa->lock);
1105 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1106 INIT_HLIST_NODE(&ifa->addr_lst);
1107 ifa->scope = cfg->scope;
1108 ifa->prefix_len = cfg->plen;
1109 ifa->rt_priority = cfg->rt_priority;
1110 ifa->flags = cfg->ifa_flags;
1111 /* No need to add the TENTATIVE flag for addresses with NODAD */
1112 if (!(cfg->ifa_flags & IFA_F_NODAD))
1113 ifa->flags |= IFA_F_TENTATIVE;
1114 ifa->valid_lft = cfg->valid_lft;
1115 ifa->prefered_lft = cfg->preferred_lft;
1116 ifa->cstamp = ifa->tstamp = jiffies;
1117 ifa->tokenized = false;
1125 refcount_set(&ifa->refcnt, 1);
1129 err = ipv6_add_addr_hash(idev->dev, ifa);
1131 rcu_read_unlock_bh();
1135 write_lock(&idev->lock);
1137 /* Add to inet6_dev unicast addr list. */
1138 ipv6_link_dev_addr(idev, ifa);
1140 if (ifa->flags&IFA_F_TEMPORARY) {
1141 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1146 write_unlock(&idev->lock);
1148 rcu_read_unlock_bh();
1150 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1152 if (unlikely(err < 0)) {
1153 fib6_info_release(f6i);
1157 in6_dev_put(ifa->idev);
1166 enum cleanup_prefix_rt_t {
1167 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1168 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1169 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1173 * Check, whether the prefix for ifp would still need a prefix route
1174 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1177 * 1) we don't purge prefix if address was not permanent.
1178 * prefix is managed by its own lifetime.
1179 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1180 * 3) if there are no addresses, delete prefix.
1181 * 4) if there are still other permanent address(es),
1182 * corresponding prefix is still permanent.
1183 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1184 * don't purge the prefix, assume user space is managing it.
1185 * 6) otherwise, update prefix lifetime to the
1186 * longest valid lifetime among the corresponding
1187 * addresses on the device.
1188 * Note: subsequent RA will update lifetime.
1190 static enum cleanup_prefix_rt_t
1191 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1193 struct inet6_ifaddr *ifa;
1194 struct inet6_dev *idev = ifp->idev;
1195 unsigned long lifetime;
1196 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1200 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1203 if (ifa->prefix_len != ifp->prefix_len ||
1204 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1207 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1208 return CLEANUP_PREFIX_RT_NOP;
1210 action = CLEANUP_PREFIX_RT_EXPIRE;
1212 spin_lock(&ifa->lock);
1214 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1216 * Note: Because this address is
1217 * not permanent, lifetime <
1218 * LONG_MAX / HZ here.
1220 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1221 *expires = ifa->tstamp + lifetime * HZ;
1222 spin_unlock(&ifa->lock);
1229 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1231 struct fib6_info *f6i;
1233 f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
1234 ifp->idev->dev, 0, RTF_DEFAULT, true);
1237 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1239 if (!(f6i->fib6_flags & RTF_EXPIRES))
1240 fib6_set_expires(f6i, expires);
1241 fib6_info_release(f6i);
1247 /* This function wants to get referenced ifp and releases it before return */
1249 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1252 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1253 unsigned long expires;
1257 spin_lock_bh(&ifp->lock);
1259 ifp->state = INET6_IFADDR_STATE_DEAD;
1260 spin_unlock_bh(&ifp->lock);
1262 if (state == INET6_IFADDR_STATE_DEAD)
1265 spin_lock_bh(&addrconf_hash_lock);
1266 hlist_del_init_rcu(&ifp->addr_lst);
1267 spin_unlock_bh(&addrconf_hash_lock);
1269 write_lock_bh(&ifp->idev->lock);
1271 if (ifp->flags&IFA_F_TEMPORARY) {
1272 list_del(&ifp->tmp_list);
1274 in6_ifa_put(ifp->ifpub);
1280 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1281 action = check_cleanup_prefix_route(ifp, &expires);
1283 list_del_rcu(&ifp->if_list);
1286 write_unlock_bh(&ifp->idev->lock);
1288 addrconf_del_dad_work(ifp);
1290 ipv6_ifa_notify(RTM_DELADDR, ifp);
1292 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1294 if (action != CLEANUP_PREFIX_RT_NOP) {
1295 cleanup_prefix_route(ifp, expires,
1296 action == CLEANUP_PREFIX_RT_DEL);
1299 /* clean up prefsrc entries */
1300 rt6_remove_prefsrc(ifp);
1305 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1306 struct inet6_ifaddr *ift,
1309 struct inet6_dev *idev = ifp->idev;
1310 struct in6_addr addr, *tmpaddr;
1311 unsigned long tmp_tstamp, age;
1312 unsigned long regen_advance;
1313 struct ifa6_config cfg;
1315 unsigned long now = jiffies;
1316 long max_desync_factor;
1317 s32 cnf_temp_preferred_lft;
1319 write_lock_bh(&idev->lock);
1321 spin_lock_bh(&ift->lock);
1322 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1323 spin_unlock_bh(&ift->lock);
1330 if (idev->cnf.use_tempaddr <= 0) {
1331 write_unlock_bh(&idev->lock);
1332 pr_info("%s: use_tempaddr is disabled\n", __func__);
1337 spin_lock_bh(&ifp->lock);
1338 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1339 idev->cnf.use_tempaddr = -1; /*XXX*/
1340 spin_unlock_bh(&ifp->lock);
1341 write_unlock_bh(&idev->lock);
1342 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1349 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1350 ipv6_try_regen_rndid(idev, tmpaddr);
1351 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1352 age = (now - ifp->tstamp) / HZ;
1354 regen_advance = idev->cnf.regen_max_retry *
1355 idev->cnf.dad_transmits *
1356 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1358 /* recalculate max_desync_factor each time and update
1359 * idev->desync_factor if it's larger
1361 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1362 max_desync_factor = min_t(__u32,
1363 idev->cnf.max_desync_factor,
1364 cnf_temp_preferred_lft - regen_advance);
1366 if (unlikely(idev->desync_factor > max_desync_factor)) {
1367 if (max_desync_factor > 0) {
1368 get_random_bytes(&idev->desync_factor,
1369 sizeof(idev->desync_factor));
1370 idev->desync_factor %= max_desync_factor;
1372 idev->desync_factor = 0;
1376 memset(&cfg, 0, sizeof(cfg));
1377 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1378 idev->cnf.temp_valid_lft + age);
1379 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1380 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1382 cfg.plen = ifp->prefix_len;
1383 tmp_tstamp = ifp->tstamp;
1384 spin_unlock_bh(&ifp->lock);
1386 write_unlock_bh(&idev->lock);
1388 /* A temporary address is created only if this calculated Preferred
1389 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1390 * an implementation must not create a temporary address with a zero
1391 * Preferred Lifetime.
1392 * Use age calculation as in addrconf_verify to avoid unnecessary
1393 * temporary addresses being generated.
1395 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1396 if (cfg.preferred_lft <= regen_advance + age) {
1403 cfg.ifa_flags = IFA_F_TEMPORARY;
1404 /* set in addrconf_prefix_rcv() */
1405 if (ifp->flags & IFA_F_OPTIMISTIC)
1406 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1409 cfg.scope = ipv6_addr_scope(cfg.pfx);
1411 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1415 pr_info("%s: retry temporary address regeneration\n", __func__);
1417 write_lock_bh(&idev->lock);
1421 spin_lock_bh(&ift->lock);
1424 ift->tstamp = tmp_tstamp;
1425 spin_unlock_bh(&ift->lock);
1427 addrconf_dad_start(ift);
1435 * Choose an appropriate source address (RFC3484)
1438 IPV6_SADDR_RULE_INIT = 0,
1439 IPV6_SADDR_RULE_LOCAL,
1440 IPV6_SADDR_RULE_SCOPE,
1441 IPV6_SADDR_RULE_PREFERRED,
1442 #ifdef CONFIG_IPV6_MIP6
1443 IPV6_SADDR_RULE_HOA,
1445 IPV6_SADDR_RULE_OIF,
1446 IPV6_SADDR_RULE_LABEL,
1447 IPV6_SADDR_RULE_PRIVACY,
1448 IPV6_SADDR_RULE_ORCHID,
1449 IPV6_SADDR_RULE_PREFIX,
1450 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1451 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1456 struct ipv6_saddr_score {
1459 struct inet6_ifaddr *ifa;
1460 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1465 struct ipv6_saddr_dst {
1466 const struct in6_addr *addr;
1473 static inline int ipv6_saddr_preferred(int type)
1475 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1480 static bool ipv6_use_optimistic_addr(struct net *net,
1481 struct inet6_dev *idev)
1483 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1488 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1497 static bool ipv6_allow_optimistic_dad(struct net *net,
1498 struct inet6_dev *idev)
1500 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1503 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1512 static int ipv6_get_saddr_eval(struct net *net,
1513 struct ipv6_saddr_score *score,
1514 struct ipv6_saddr_dst *dst,
1519 if (i <= score->rule) {
1521 case IPV6_SADDR_RULE_SCOPE:
1522 ret = score->scopedist;
1524 case IPV6_SADDR_RULE_PREFIX:
1525 ret = score->matchlen;
1528 ret = !!test_bit(i, score->scorebits);
1534 case IPV6_SADDR_RULE_INIT:
1535 /* Rule 0: remember if hiscore is not ready yet */
1538 case IPV6_SADDR_RULE_LOCAL:
1539 /* Rule 1: Prefer same address */
1540 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1542 case IPV6_SADDR_RULE_SCOPE:
1543 /* Rule 2: Prefer appropriate scope
1548 * ---+--+-+---> scope
1550 * | d is scope of the destination.
1552 * | \ <- smaller scope is better if
1553 * B-15 | \ if scope is enough for destination.
1554 * | ret = B - scope (-1 <= scope >= d <= 15).
1556 * |/ <- greater is better
1557 * -C / if scope is not enough for destination.
1558 * /| ret = scope - C (-1 <= d < scope <= 15).
1560 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1561 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1562 * Assume B = 0 and we get C > 29.
1564 ret = __ipv6_addr_src_scope(score->addr_type);
1565 if (ret >= dst->scope)
1568 ret -= 128; /* 30 is enough */
1569 score->scopedist = ret;
1571 case IPV6_SADDR_RULE_PREFERRED:
1573 /* Rule 3: Avoid deprecated and optimistic addresses */
1574 u8 avoid = IFA_F_DEPRECATED;
1576 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1577 avoid |= IFA_F_OPTIMISTIC;
1578 ret = ipv6_saddr_preferred(score->addr_type) ||
1579 !(score->ifa->flags & avoid);
1582 #ifdef CONFIG_IPV6_MIP6
1583 case IPV6_SADDR_RULE_HOA:
1585 /* Rule 4: Prefer home address */
1586 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1587 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1591 case IPV6_SADDR_RULE_OIF:
1592 /* Rule 5: Prefer outgoing interface */
1593 ret = (!dst->ifindex ||
1594 dst->ifindex == score->ifa->idev->dev->ifindex);
1596 case IPV6_SADDR_RULE_LABEL:
1597 /* Rule 6: Prefer matching label */
1598 ret = ipv6_addr_label(net,
1599 &score->ifa->addr, score->addr_type,
1600 score->ifa->idev->dev->ifindex) == dst->label;
1602 case IPV6_SADDR_RULE_PRIVACY:
1604 /* Rule 7: Prefer public address
1605 * Note: prefer temporary address if use_tempaddr >= 2
1607 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1608 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1609 score->ifa->idev->cnf.use_tempaddr >= 2;
1610 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1613 case IPV6_SADDR_RULE_ORCHID:
1614 /* Rule 8-: Prefer ORCHID vs ORCHID or
1615 * non-ORCHID vs non-ORCHID
1617 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1618 ipv6_addr_orchid(dst->addr));
1620 case IPV6_SADDR_RULE_PREFIX:
1621 /* Rule 8: Use longest matching prefix */
1622 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1623 if (ret > score->ifa->prefix_len)
1624 ret = score->ifa->prefix_len;
1625 score->matchlen = ret;
1627 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1628 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1629 /* Optimistic addresses still have lower precedence than other
1630 * preferred addresses.
1632 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1640 __set_bit(i, score->scorebits);
1646 static int __ipv6_dev_get_saddr(struct net *net,
1647 struct ipv6_saddr_dst *dst,
1648 struct inet6_dev *idev,
1649 struct ipv6_saddr_score *scores,
1652 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1654 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1658 * - Tentative Address (RFC2462 section 5.4)
1659 * - A tentative address is not considered
1660 * "assigned to an interface" in the traditional
1661 * sense, unless it is also flagged as optimistic.
1662 * - Candidate Source Address (section 4)
1663 * - In any case, anycast addresses, multicast
1664 * addresses, and the unspecified address MUST
1665 * NOT be included in a candidate set.
1667 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1668 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1671 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1673 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1674 score->addr_type & IPV6_ADDR_MULTICAST)) {
1675 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1681 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1683 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1684 int minihiscore, miniscore;
1686 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1687 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1689 if (minihiscore > miniscore) {
1690 if (i == IPV6_SADDR_RULE_SCOPE &&
1691 score->scopedist > 0) {
1694 * each remaining entry
1695 * has too small (not enough)
1696 * scope, because ifa entries
1697 * are sorted by their scope
1703 } else if (minihiscore < miniscore) {
1704 swap(hiscore, score);
1705 hiscore_idx = 1 - hiscore_idx;
1707 /* restore our iterator */
1708 score->ifa = hiscore->ifa;
1718 static int ipv6_get_saddr_master(struct net *net,
1719 const struct net_device *dst_dev,
1720 const struct net_device *master,
1721 struct ipv6_saddr_dst *dst,
1722 struct ipv6_saddr_score *scores,
1725 struct inet6_dev *idev;
1727 idev = __in6_dev_get(dst_dev);
1729 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1730 scores, hiscore_idx);
1732 idev = __in6_dev_get(master);
1734 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1735 scores, hiscore_idx);
1740 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1741 const struct in6_addr *daddr, unsigned int prefs,
1742 struct in6_addr *saddr)
1744 struct ipv6_saddr_score scores[2], *hiscore;
1745 struct ipv6_saddr_dst dst;
1746 struct inet6_dev *idev;
1747 struct net_device *dev;
1749 bool use_oif_addr = false;
1750 int hiscore_idx = 0;
1753 dst_type = __ipv6_addr_type(daddr);
1755 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1756 dst.scope = __ipv6_addr_src_scope(dst_type);
1757 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1760 scores[hiscore_idx].rule = -1;
1761 scores[hiscore_idx].ifa = NULL;
1765 /* Candidate Source Address (section 4)
1766 * - multicast and link-local destination address,
1767 * the set of candidate source address MUST only
1768 * include addresses assigned to interfaces
1769 * belonging to the same link as the outgoing
1771 * (- For site-local destination addresses, the
1772 * set of candidate source addresses MUST only
1773 * include addresses assigned to interfaces
1774 * belonging to the same site as the outgoing
1776 * - "It is RECOMMENDED that the candidate source addresses
1777 * be the set of unicast addresses assigned to the
1778 * interface that will be used to send to the destination
1779 * (the 'outgoing' interface)." (RFC 6724)
1782 idev = __in6_dev_get(dst_dev);
1783 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1784 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1785 (idev && idev->cnf.use_oif_addrs_only)) {
1786 use_oif_addr = true;
1792 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1794 const struct net_device *master;
1797 /* if dst_dev exists and is enslaved to an L3 device, then
1798 * prefer addresses from dst_dev and then the master over
1799 * any other enslaved devices in the L3 domain.
1801 master = l3mdev_master_dev_rcu(dst_dev);
1803 master_idx = master->ifindex;
1805 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1807 scores, hiscore_idx);
1809 if (scores[hiscore_idx].ifa)
1813 for_each_netdev_rcu(net, dev) {
1814 /* only consider addresses on devices in the
1817 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1819 idev = __in6_dev_get(dev);
1822 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1827 hiscore = &scores[hiscore_idx];
1829 ret = -EADDRNOTAVAIL;
1831 *saddr = hiscore->ifa->addr;
1836 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1838 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1841 struct inet6_ifaddr *ifp;
1842 int err = -EADDRNOTAVAIL;
1844 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1845 if (ifp->scope > IFA_LINK)
1847 if (ifp->scope == IFA_LINK &&
1848 !(ifp->flags & banned_flags)) {
1857 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1860 struct inet6_dev *idev;
1861 int err = -EADDRNOTAVAIL;
1864 idev = __in6_dev_get(dev);
1866 read_lock_bh(&idev->lock);
1867 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1868 read_unlock_bh(&idev->lock);
1874 static int ipv6_count_addresses(const struct inet6_dev *idev)
1876 const struct inet6_ifaddr *ifp;
1880 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1886 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1887 const struct net_device *dev, int strict)
1889 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1890 strict, IFA_F_TENTATIVE);
1892 EXPORT_SYMBOL(ipv6_chk_addr);
1894 /* device argument is used to find the L3 domain of interest. If
1895 * skip_dev_check is set, then the ifp device is not checked against
1896 * the passed in dev argument. So the 2 cases for addresses checks are:
1897 * 1. does the address exist in the L3 domain that dev is part of
1898 * (skip_dev_check = true), or
1900 * 2. does the address exist on the specific device
1901 * (skip_dev_check = false)
1903 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1904 const struct net_device *dev, bool skip_dev_check,
1905 int strict, u32 banned_flags)
1907 unsigned int hash = inet6_addr_hash(net, addr);
1908 const struct net_device *l3mdev;
1909 struct inet6_ifaddr *ifp;
1914 l3mdev = l3mdev_master_dev_rcu(dev);
1918 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1919 if (!net_eq(dev_net(ifp->idev->dev), net))
1922 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1925 /* Decouple optimistic from tentative for evaluation here.
1926 * Ban optimistic addresses explicitly, when required.
1928 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1929 ? (ifp->flags&~IFA_F_TENTATIVE)
1931 if (ipv6_addr_equal(&ifp->addr, addr) &&
1932 !(ifp_flags&banned_flags) &&
1933 (!dev || ifp->idev->dev == dev ||
1934 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1943 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1946 /* Compares an address/prefix_len with addresses on device @dev.
1947 * If one is found it returns true.
1949 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1950 const unsigned int prefix_len, struct net_device *dev)
1952 const struct inet6_ifaddr *ifa;
1953 const struct inet6_dev *idev;
1957 idev = __in6_dev_get(dev);
1959 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1960 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1969 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1971 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1973 const struct inet6_ifaddr *ifa;
1974 const struct inet6_dev *idev;
1979 idev = __in6_dev_get(dev);
1981 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1982 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1991 EXPORT_SYMBOL(ipv6_chk_prefix);
1993 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1994 struct net_device *dev, int strict)
1996 unsigned int hash = inet6_addr_hash(net, addr);
1997 struct inet6_ifaddr *ifp, *result = NULL;
2000 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2001 if (!net_eq(dev_net(ifp->idev->dev), net))
2003 if (ipv6_addr_equal(&ifp->addr, addr)) {
2004 if (!dev || ifp->idev->dev == dev ||
2005 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2017 /* Gets referenced address, destroys ifaddr */
2019 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2022 ifp->flags |= IFA_F_DADFAILED;
2024 if (ifp->flags&IFA_F_TEMPORARY) {
2025 struct inet6_ifaddr *ifpub;
2026 spin_lock_bh(&ifp->lock);
2029 in6_ifa_hold(ifpub);
2030 spin_unlock_bh(&ifp->lock);
2031 ipv6_create_tempaddr(ifpub, ifp, true);
2034 spin_unlock_bh(&ifp->lock);
2037 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2038 spin_lock_bh(&ifp->lock);
2039 addrconf_del_dad_work(ifp);
2040 ifp->flags |= IFA_F_TENTATIVE;
2042 ifp->flags &= ~IFA_F_OPTIMISTIC;
2043 spin_unlock_bh(&ifp->lock);
2045 ipv6_ifa_notify(0, ifp);
2052 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2056 spin_lock_bh(&ifp->lock);
2057 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2058 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2061 spin_unlock_bh(&ifp->lock);
2066 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2068 struct inet6_dev *idev = ifp->idev;
2069 struct net *net = dev_net(ifp->idev->dev);
2071 if (addrconf_dad_end(ifp)) {
2076 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2077 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2079 spin_lock_bh(&ifp->lock);
2081 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2082 struct in6_addr new_addr;
2083 struct inet6_ifaddr *ifp2;
2084 int retries = ifp->stable_privacy_retry + 1;
2085 struct ifa6_config cfg = {
2087 .plen = ifp->prefix_len,
2088 .ifa_flags = ifp->flags,
2089 .valid_lft = ifp->valid_lft,
2090 .preferred_lft = ifp->prefered_lft,
2091 .scope = ifp->scope,
2094 if (retries > net->ipv6.sysctl.idgen_retries) {
2095 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2096 ifp->idev->dev->name);
2100 new_addr = ifp->addr;
2101 if (ipv6_generate_stable_address(&new_addr, retries,
2105 spin_unlock_bh(&ifp->lock);
2107 if (idev->cnf.max_addresses &&
2108 ipv6_count_addresses(idev) >=
2109 idev->cnf.max_addresses)
2112 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2113 ifp->idev->dev->name);
2115 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2119 spin_lock_bh(&ifp2->lock);
2120 ifp2->stable_privacy_retry = retries;
2121 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2122 spin_unlock_bh(&ifp2->lock);
2124 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2127 spin_lock_bh(&ifp->lock);
2131 /* transition from _POSTDAD to _ERRDAD */
2132 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2133 spin_unlock_bh(&ifp->lock);
2135 addrconf_mod_dad_work(ifp, 0);
2139 /* Join to solicited addr multicast group.
2140 * caller must hold RTNL */
2141 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2143 struct in6_addr maddr;
2145 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2148 addrconf_addr_solict_mult(addr, &maddr);
2149 ipv6_dev_mc_inc(dev, &maddr);
2152 /* caller must hold RTNL */
2153 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2155 struct in6_addr maddr;
2157 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2160 addrconf_addr_solict_mult(addr, &maddr);
2161 __ipv6_dev_mc_dec(idev, &maddr);
2164 /* caller must hold RTNL */
2165 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2167 struct in6_addr addr;
2169 if (ifp->prefix_len >= 127) /* RFC 6164 */
2171 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2172 if (ipv6_addr_any(&addr))
2174 __ipv6_dev_ac_inc(ifp->idev, &addr);
2177 /* caller must hold RTNL */
2178 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2180 struct in6_addr addr;
2182 if (ifp->prefix_len >= 127) /* RFC 6164 */
2184 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2185 if (ipv6_addr_any(&addr))
2187 __ipv6_dev_ac_dec(ifp->idev, &addr);
2190 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2192 switch (dev->addr_len) {
2194 memcpy(eui, dev->dev_addr, 3);
2197 memcpy(eui + 5, dev->dev_addr + 3, 3);
2199 case EUI64_ADDR_LEN:
2200 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2210 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2212 union fwnet_hwaddr *ha;
2214 if (dev->addr_len != FWNET_ALEN)
2217 ha = (union fwnet_hwaddr *)dev->dev_addr;
2219 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2224 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2226 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2227 if (dev->addr_len != ARCNET_ALEN)
2230 eui[7] = *(u8 *)dev->dev_addr;
2234 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2236 if (dev->addr_len != INFINIBAND_ALEN)
2238 memcpy(eui, dev->dev_addr + 12, 8);
2243 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2247 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2248 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2249 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2250 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2251 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2252 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2256 memcpy(eui + 4, &addr, 4);
2260 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2262 if (dev->priv_flags & IFF_ISATAP)
2263 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2267 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2269 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2272 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2274 memcpy(eui, dev->perm_addr, 3);
2275 memcpy(eui + 5, dev->perm_addr + 3, 3);
2282 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2284 switch (dev->type) {
2287 return addrconf_ifid_eui48(eui, dev);
2289 return addrconf_ifid_arcnet(eui, dev);
2290 case ARPHRD_INFINIBAND:
2291 return addrconf_ifid_infiniband(eui, dev);
2293 return addrconf_ifid_sit(eui, dev);
2296 return addrconf_ifid_gre(eui, dev);
2297 case ARPHRD_6LOWPAN:
2298 return addrconf_ifid_6lowpan(eui, dev);
2299 case ARPHRD_IEEE1394:
2300 return addrconf_ifid_ieee1394(eui, dev);
2301 case ARPHRD_TUNNEL6:
2304 return addrconf_ifid_ip6tnl(eui, dev);
2309 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2312 struct inet6_ifaddr *ifp;
2314 read_lock_bh(&idev->lock);
2315 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2316 if (ifp->scope > IFA_LINK)
2318 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2319 memcpy(eui, ifp->addr.s6_addr+8, 8);
2324 read_unlock_bh(&idev->lock);
2328 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2329 static void ipv6_regen_rndid(struct inet6_dev *idev)
2332 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2333 idev->rndid[0] &= ~0x02;
2336 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2337 * check if generated address is not inappropriate
2339 * - Reserved subnet anycast (RFC 2526)
2340 * 11111101 11....11 1xxxxxxx
2341 * - ISATAP (RFC4214) 6.1
2342 * 00-00-5E-FE-xx-xx-xx-xx
2344 * - XXX: already assigned to an address on the device
2346 if (idev->rndid[0] == 0xfd &&
2347 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2348 (idev->rndid[7]&0x80))
2350 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2351 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2353 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2358 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2360 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2361 ipv6_regen_rndid(idev);
2369 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2370 struct net_device *dev, unsigned long expires,
2371 u32 flags, gfp_t gfp_flags)
2373 struct fib6_config cfg = {
2374 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2375 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2376 .fc_ifindex = dev->ifindex,
2377 .fc_expires = expires,
2379 .fc_flags = RTF_UP | flags,
2380 .fc_nlinfo.nl_net = dev_net(dev),
2381 .fc_protocol = RTPROT_KERNEL,
2382 .fc_type = RTN_UNICAST,
2387 /* Prevent useless cloning on PtP SIT.
2388 This thing is done here expecting that the whole
2389 class of non-broadcast devices need not cloning.
2391 #if IS_ENABLED(CONFIG_IPV6_SIT)
2392 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2393 cfg.fc_flags |= RTF_NONEXTHOP;
2396 ip6_route_add(&cfg, gfp_flags, NULL);
2400 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2402 const struct net_device *dev,
2403 u32 flags, u32 noflags,
2406 struct fib6_node *fn;
2407 struct fib6_info *rt = NULL;
2408 struct fib6_table *table;
2409 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2411 table = fib6_get_table(dev_net(dev), tb_id);
2416 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2420 for_each_fib6_node_rt_rcu(fn) {
2421 /* prefix routes only use builtin fib6_nh */
2425 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2427 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2429 if ((rt->fib6_flags & flags) != flags)
2431 if ((rt->fib6_flags & noflags) != 0)
2433 if (!fib6_info_hold_safe(rt))
2443 /* Create "default" multicast route to the interface */
2445 static void addrconf_add_mroute(struct net_device *dev)
2447 struct fib6_config cfg = {
2448 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2449 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2450 .fc_ifindex = dev->ifindex,
2453 .fc_type = RTN_UNICAST,
2454 .fc_nlinfo.nl_net = dev_net(dev),
2457 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2459 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2462 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2464 struct inet6_dev *idev;
2468 idev = ipv6_find_idev(dev);
2472 if (idev->cnf.disable_ipv6)
2473 return ERR_PTR(-EACCES);
2475 /* Add default multicast route */
2476 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2477 addrconf_add_mroute(dev);
2482 static void manage_tempaddrs(struct inet6_dev *idev,
2483 struct inet6_ifaddr *ifp,
2484 __u32 valid_lft, __u32 prefered_lft,
2485 bool create, unsigned long now)
2488 struct inet6_ifaddr *ift;
2490 read_lock_bh(&idev->lock);
2491 /* update all temporary addresses in the list */
2492 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2493 int age, max_valid, max_prefered;
2495 if (ifp != ift->ifpub)
2498 /* RFC 4941 section 3.3:
2499 * If a received option will extend the lifetime of a public
2500 * address, the lifetimes of temporary addresses should
2501 * be extended, subject to the overall constraint that no
2502 * temporary addresses should ever remain "valid" or "preferred"
2503 * for a time longer than (TEMP_VALID_LIFETIME) or
2504 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2506 age = (now - ift->cstamp) / HZ;
2507 max_valid = idev->cnf.temp_valid_lft - age;
2511 max_prefered = idev->cnf.temp_prefered_lft -
2512 idev->desync_factor - age;
2513 if (max_prefered < 0)
2516 if (valid_lft > max_valid)
2517 valid_lft = max_valid;
2519 if (prefered_lft > max_prefered)
2520 prefered_lft = max_prefered;
2522 spin_lock(&ift->lock);
2524 ift->valid_lft = valid_lft;
2525 ift->prefered_lft = prefered_lft;
2527 if (prefered_lft > 0)
2528 ift->flags &= ~IFA_F_DEPRECATED;
2530 spin_unlock(&ift->lock);
2531 if (!(flags&IFA_F_TENTATIVE))
2532 ipv6_ifa_notify(0, ift);
2535 if ((create || list_empty(&idev->tempaddr_list)) &&
2536 idev->cnf.use_tempaddr > 0) {
2537 /* When a new public address is created as described
2538 * in [ADDRCONF], also create a new temporary address.
2539 * Also create a temporary address if it's enabled but
2540 * no temporary address currently exists.
2542 read_unlock_bh(&idev->lock);
2543 ipv6_create_tempaddr(ifp, NULL, false);
2545 read_unlock_bh(&idev->lock);
2549 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2551 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2552 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2555 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2556 const struct prefix_info *pinfo,
2557 struct inet6_dev *in6_dev,
2558 const struct in6_addr *addr, int addr_type,
2559 u32 addr_flags, bool sllao, bool tokenized,
2560 __u32 valid_lft, u32 prefered_lft)
2562 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2563 int create = 0, update_lft = 0;
2565 if (!ifp && valid_lft) {
2566 int max_addresses = in6_dev->cnf.max_addresses;
2567 struct ifa6_config cfg = {
2569 .plen = pinfo->prefix_len,
2570 .ifa_flags = addr_flags,
2571 .valid_lft = valid_lft,
2572 .preferred_lft = prefered_lft,
2573 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2576 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2577 if ((net->ipv6.devconf_all->optimistic_dad ||
2578 in6_dev->cnf.optimistic_dad) &&
2579 !net->ipv6.devconf_all->forwarding && sllao)
2580 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2583 /* Do not allow to create too much of autoconfigured
2584 * addresses; this would be too easy way to crash kernel.
2586 if (!max_addresses ||
2587 ipv6_count_addresses(in6_dev) < max_addresses)
2588 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2590 if (IS_ERR_OR_NULL(ifp))
2594 spin_lock_bh(&ifp->lock);
2595 ifp->flags |= IFA_F_MANAGETEMPADDR;
2596 ifp->cstamp = jiffies;
2597 ifp->tokenized = tokenized;
2598 spin_unlock_bh(&ifp->lock);
2599 addrconf_dad_start(ifp);
2607 /* update lifetime (RFC2462 5.5.3 e) */
2608 spin_lock_bh(&ifp->lock);
2610 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2611 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2614 if (!create && stored_lft) {
2615 const u32 minimum_lft = min_t(u32,
2616 stored_lft, MIN_VALID_LIFETIME);
2617 valid_lft = max(valid_lft, minimum_lft);
2619 /* RFC4862 Section 5.5.3e:
2620 * "Note that the preferred lifetime of the
2621 * corresponding address is always reset to
2622 * the Preferred Lifetime in the received
2623 * Prefix Information option, regardless of
2624 * whether the valid lifetime is also reset or
2627 * So we should always update prefered_lft here.
2633 ifp->valid_lft = valid_lft;
2634 ifp->prefered_lft = prefered_lft;
2637 ifp->flags &= ~IFA_F_DEPRECATED;
2638 spin_unlock_bh(&ifp->lock);
2640 if (!(flags&IFA_F_TENTATIVE))
2641 ipv6_ifa_notify(0, ifp);
2643 spin_unlock_bh(&ifp->lock);
2645 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2654 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2656 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2658 struct prefix_info *pinfo;
2663 struct inet6_dev *in6_dev;
2664 struct net *net = dev_net(dev);
2666 pinfo = (struct prefix_info *) opt;
2668 if (len < sizeof(struct prefix_info)) {
2669 netdev_dbg(dev, "addrconf: prefix option too short\n");
2674 * Validation checks ([ADDRCONF], page 19)
2677 addr_type = ipv6_addr_type(&pinfo->prefix);
2679 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2682 valid_lft = ntohl(pinfo->valid);
2683 prefered_lft = ntohl(pinfo->prefered);
2685 if (prefered_lft > valid_lft) {
2686 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2690 in6_dev = in6_dev_get(dev);
2693 net_dbg_ratelimited("addrconf: device %s not configured\n",
2699 * Two things going on here:
2700 * 1) Add routes for on-link prefixes
2701 * 2) Configure prefixes with the auto flag set
2704 if (pinfo->onlink) {
2705 struct fib6_info *rt;
2706 unsigned long rt_expires;
2708 /* Avoid arithmetic overflow. Really, we could
2709 * save rt_expires in seconds, likely valid_lft,
2710 * but it would require division in fib gc, that it
2714 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2716 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2718 if (addrconf_finite_timeout(rt_expires))
2721 rt = addrconf_get_prefix_route(&pinfo->prefix,
2724 RTF_ADDRCONF | RTF_PREFIX_RT,
2728 /* Autoconf prefix route */
2729 if (valid_lft == 0) {
2730 ip6_del_rt(net, rt);
2732 } else if (addrconf_finite_timeout(rt_expires)) {
2734 fib6_set_expires(rt, jiffies + rt_expires);
2736 fib6_clean_expires(rt);
2738 } else if (valid_lft) {
2739 clock_t expires = 0;
2740 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2741 if (addrconf_finite_timeout(rt_expires)) {
2743 flags |= RTF_EXPIRES;
2744 expires = jiffies_to_clock_t(rt_expires);
2746 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2747 0, dev, expires, flags,
2750 fib6_info_release(rt);
2753 /* Try to figure out our local address for this prefix */
2755 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2756 struct in6_addr addr;
2757 bool tokenized = false, dev_addr_generated = false;
2759 if (pinfo->prefix_len == 64) {
2760 memcpy(&addr, &pinfo->prefix, 8);
2762 if (!ipv6_addr_any(&in6_dev->token)) {
2763 read_lock_bh(&in6_dev->lock);
2764 memcpy(addr.s6_addr + 8,
2765 in6_dev->token.s6_addr + 8, 8);
2766 read_unlock_bh(&in6_dev->lock);
2768 } else if (is_addr_mode_generate_stable(in6_dev) &&
2769 !ipv6_generate_stable_address(&addr, 0,
2771 addr_flags |= IFA_F_STABLE_PRIVACY;
2773 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2774 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2777 dev_addr_generated = true;
2781 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2786 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2789 tokenized, valid_lft,
2794 /* Ignore error case here because previous prefix add addr was
2795 * successful which will be notified.
2797 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2798 addr_type, addr_flags, sllao,
2799 tokenized, valid_lft,
2801 dev_addr_generated);
2803 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2805 in6_dev_put(in6_dev);
2809 * Set destination address.
2810 * Special case for SIT interfaces where we create a new "virtual"
2813 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2815 struct in6_ifreq ireq;
2816 struct net_device *dev;
2822 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2825 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2831 #if IS_ENABLED(CONFIG_IPV6_SIT)
2832 if (dev->type == ARPHRD_SIT) {
2833 const struct net_device_ops *ops = dev->netdev_ops;
2835 struct ip_tunnel_parm p;
2837 err = -EADDRNOTAVAIL;
2838 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2841 memset(&p, 0, sizeof(p));
2842 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2846 p.iph.protocol = IPPROTO_IPV6;
2848 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2850 if (ops->ndo_do_ioctl) {
2851 mm_segment_t oldfs = get_fs();
2854 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2861 dev = __dev_get_by_name(net, p.name);
2864 err = dev_open(dev, NULL);
2874 static int ipv6_mc_config(struct sock *sk, bool join,
2875 const struct in6_addr *addr, int ifindex)
2883 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2885 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2892 * Manual configuration of address on an interface
2894 static int inet6_addr_add(struct net *net, int ifindex,
2895 struct ifa6_config *cfg,
2896 struct netlink_ext_ack *extack)
2898 struct inet6_ifaddr *ifp;
2899 struct inet6_dev *idev;
2900 struct net_device *dev;
2901 unsigned long timeout;
2907 if (cfg->plen > 128)
2910 /* check the lifetime */
2911 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2914 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2917 dev = __dev_get_by_index(net, ifindex);
2921 idev = addrconf_add_dev(dev);
2923 return PTR_ERR(idev);
2925 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2926 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2927 true, cfg->pfx, ifindex);
2933 cfg->scope = ipv6_addr_scope(cfg->pfx);
2935 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2936 if (addrconf_finite_timeout(timeout)) {
2937 expires = jiffies_to_clock_t(timeout * HZ);
2938 cfg->valid_lft = timeout;
2939 flags = RTF_EXPIRES;
2943 cfg->ifa_flags |= IFA_F_PERMANENT;
2946 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2947 if (addrconf_finite_timeout(timeout)) {
2949 cfg->ifa_flags |= IFA_F_DEPRECATED;
2950 cfg->preferred_lft = timeout;
2953 ifp = ipv6_add_addr(idev, cfg, true, extack);
2955 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2956 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2957 ifp->rt_priority, dev, expires,
2961 /* Send a netlink notification if DAD is enabled and
2962 * optimistic flag is not set
2964 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2965 ipv6_ifa_notify(0, ifp);
2967 * Note that section 3.1 of RFC 4429 indicates
2968 * that the Optimistic flag should not be set for
2969 * manually configured addresses
2971 addrconf_dad_start(ifp);
2972 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2973 manage_tempaddrs(idev, ifp, cfg->valid_lft,
2974 cfg->preferred_lft, true, jiffies);
2976 addrconf_verify_rtnl();
2978 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2979 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2983 return PTR_ERR(ifp);
2986 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2987 const struct in6_addr *pfx, unsigned int plen)
2989 struct inet6_ifaddr *ifp;
2990 struct inet6_dev *idev;
2991 struct net_device *dev;
2996 dev = __dev_get_by_index(net, ifindex);
3000 idev = __in6_dev_get(dev);
3004 read_lock_bh(&idev->lock);
3005 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3006 if (ifp->prefix_len == plen &&
3007 ipv6_addr_equal(pfx, &ifp->addr)) {
3009 read_unlock_bh(&idev->lock);
3011 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3012 (ifa_flags & IFA_F_MANAGETEMPADDR))
3013 manage_tempaddrs(idev, ifp, 0, 0, false,
3016 addrconf_verify_rtnl();
3017 if (ipv6_addr_is_multicast(pfx)) {
3018 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3019 false, pfx, dev->ifindex);
3024 read_unlock_bh(&idev->lock);
3025 return -EADDRNOTAVAIL;
3029 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3031 struct ifa6_config cfg = {
3032 .ifa_flags = IFA_F_PERMANENT,
3033 .preferred_lft = INFINITY_LIFE_TIME,
3034 .valid_lft = INFINITY_LIFE_TIME,
3036 struct in6_ifreq ireq;
3039 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3042 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3045 cfg.pfx = &ireq.ifr6_addr;
3046 cfg.plen = ireq.ifr6_prefixlen;
3049 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3054 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3056 struct in6_ifreq ireq;
3059 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3062 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3066 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3067 ireq.ifr6_prefixlen);
3072 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3073 int plen, int scope)
3075 struct inet6_ifaddr *ifp;
3076 struct ifa6_config cfg = {
3079 .ifa_flags = IFA_F_PERMANENT,
3080 .valid_lft = INFINITY_LIFE_TIME,
3081 .preferred_lft = INFINITY_LIFE_TIME,
3085 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3087 spin_lock_bh(&ifp->lock);
3088 ifp->flags &= ~IFA_F_TENTATIVE;
3089 spin_unlock_bh(&ifp->lock);
3090 rt_genid_bump_ipv6(dev_net(idev->dev));
3091 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3096 #if IS_ENABLED(CONFIG_IPV6_SIT)
3097 static void sit_add_v4_addrs(struct inet6_dev *idev)
3099 struct in6_addr addr;
3100 struct net_device *dev;
3101 struct net *net = dev_net(idev->dev);
3107 memset(&addr, 0, sizeof(struct in6_addr));
3108 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3110 if (idev->dev->flags&IFF_POINTOPOINT) {
3111 addr.s6_addr32[0] = htonl(0xfe800000);
3115 scope = IPV6_ADDR_COMPATv4;
3117 pflags |= RTF_NONEXTHOP;
3120 if (addr.s6_addr32[3]) {
3121 add_addr(idev, &addr, plen, scope);
3122 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3127 for_each_netdev(net, dev) {
3128 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3129 if (in_dev && (dev->flags & IFF_UP)) {
3130 struct in_ifaddr *ifa;
3133 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3134 addr.s6_addr32[3] = ifa->ifa_local;
3136 if (ifa->ifa_scope == RT_SCOPE_LINK)
3138 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3139 if (idev->dev->flags&IFF_POINTOPOINT)
3144 add_addr(idev, &addr, plen, flag);
3145 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3146 0, pflags, GFP_KERNEL);
3153 static void init_loopback(struct net_device *dev)
3155 struct inet6_dev *idev;
3161 idev = ipv6_find_idev(dev);
3163 pr_debug("%s: add_dev failed\n", __func__);
3167 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3170 void addrconf_add_linklocal(struct inet6_dev *idev,
3171 const struct in6_addr *addr, u32 flags)
3173 struct ifa6_config cfg = {
3176 .ifa_flags = flags | IFA_F_PERMANENT,
3177 .valid_lft = INFINITY_LIFE_TIME,
3178 .preferred_lft = INFINITY_LIFE_TIME,
3181 struct inet6_ifaddr *ifp;
3183 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3184 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3185 idev->cnf.optimistic_dad) &&
3186 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3187 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3190 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3192 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3194 addrconf_dad_start(ifp);
3198 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3200 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3202 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3205 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3206 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3209 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3210 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3216 static int ipv6_generate_stable_address(struct in6_addr *address,
3218 const struct inet6_dev *idev)
3220 static DEFINE_SPINLOCK(lock);
3221 static __u32 digest[SHA_DIGEST_WORDS];
3222 static __u32 workspace[SHA_WORKSPACE_WORDS];
3225 char __data[SHA_MESSAGE_BYTES];
3227 struct in6_addr secret;
3229 unsigned char hwaddr[MAX_ADDR_LEN];
3234 struct in6_addr secret;
3235 struct in6_addr temp;
3236 struct net *net = dev_net(idev->dev);
3238 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3240 if (idev->cnf.stable_secret.initialized)
3241 secret = idev->cnf.stable_secret.secret;
3242 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3243 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3248 spin_lock_bh(&lock);
3251 memset(&data, 0, sizeof(data));
3252 memset(workspace, 0, sizeof(workspace));
3253 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3254 data.prefix[0] = address->s6_addr32[0];
3255 data.prefix[1] = address->s6_addr32[1];
3256 data.secret = secret;
3257 data.dad_count = dad_count;
3259 sha_transform(digest, data.__data, workspace);
3262 temp.s6_addr32[2] = (__force __be32)digest[0];
3263 temp.s6_addr32[3] = (__force __be32)digest[1];
3265 spin_unlock_bh(&lock);
3267 if (ipv6_reserved_interfaceid(temp)) {
3269 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3278 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3280 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3284 s = &idev->cnf.stable_secret;
3285 get_random_bytes(&s->secret, sizeof(s->secret));
3286 s->initialized = true;
3289 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3291 struct in6_addr addr;
3293 /* no link local addresses on L3 master devices */
3294 if (netif_is_l3_master(idev->dev))
3297 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3299 switch (idev->cnf.addr_gen_mode) {
3300 case IN6_ADDR_GEN_MODE_RANDOM:
3301 ipv6_gen_mode_random_init(idev);
3303 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3304 if (!ipv6_generate_stable_address(&addr, 0, idev))
3305 addrconf_add_linklocal(idev, &addr,
3306 IFA_F_STABLE_PRIVACY);
3307 else if (prefix_route)
3308 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3311 case IN6_ADDR_GEN_MODE_EUI64:
3312 /* addrconf_add_linklocal also adds a prefix_route and we
3313 * only need to care about prefix routes if ipv6_generate_eui64
3314 * couldn't generate one.
3316 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3317 addrconf_add_linklocal(idev, &addr, 0);
3318 else if (prefix_route)
3319 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3322 case IN6_ADDR_GEN_MODE_NONE:
3324 /* will not add any link local address */
3329 static void addrconf_dev_config(struct net_device *dev)
3331 struct inet6_dev *idev;
3335 if ((dev->type != ARPHRD_ETHER) &&
3336 (dev->type != ARPHRD_FDDI) &&
3337 (dev->type != ARPHRD_ARCNET) &&
3338 (dev->type != ARPHRD_INFINIBAND) &&
3339 (dev->type != ARPHRD_IEEE1394) &&
3340 (dev->type != ARPHRD_TUNNEL6) &&
3341 (dev->type != ARPHRD_6LOWPAN) &&
3342 (dev->type != ARPHRD_IP6GRE) &&
3343 (dev->type != ARPHRD_IPGRE) &&
3344 (dev->type != ARPHRD_TUNNEL) &&
3345 (dev->type != ARPHRD_NONE) &&
3346 (dev->type != ARPHRD_RAWIP)) {
3347 /* Alas, we support only Ethernet autoconfiguration. */
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))
3525 if (dev->flags & IFF_SLAVE)
3528 if (idev && idev->cnf.disable_ipv6)
3531 if (event == NETDEV_UP) {
3532 /* restore routes for permanent addresses */
3533 addrconf_permanent_addr(net, dev);
3535 if (!addrconf_link_ready(dev)) {
3536 /* device is not ready yet. */
3537 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3542 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3543 idev = ipv6_add_dev(dev);
3545 if (!IS_ERR_OR_NULL(idev)) {
3546 idev->if_flags |= IF_READY;
3549 } else if (event == NETDEV_CHANGE) {
3550 if (!addrconf_link_ready(dev)) {
3551 /* device is still not ready. */
3552 rt6_sync_down_dev(dev, event);
3556 if (!IS_ERR_OR_NULL(idev)) {
3557 if (idev->if_flags & IF_READY) {
3558 /* device is already configured -
3559 * but resend MLD reports, we might
3560 * have roamed and need to update
3561 * multicast snooping switches
3565 if (change_info->flags_changed & IFF_NOARP)
3566 addrconf_dad_run(idev, true);
3567 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3570 idev->if_flags |= IF_READY;
3573 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3579 switch (dev->type) {
3580 #if IS_ENABLED(CONFIG_IPV6_SIT)
3582 addrconf_sit_config(dev);
3585 #if IS_ENABLED(CONFIG_NET_IPGRE)
3587 addrconf_gre_config(dev);
3590 case ARPHRD_LOOPBACK:
3595 addrconf_dev_config(dev);
3599 if (!IS_ERR_OR_NULL(idev)) {
3601 addrconf_dad_run(idev, false);
3603 /* Device has an address by now */
3604 rt6_sync_up(dev, RTNH_F_DEAD);
3607 * If the MTU changed during the interface down,
3608 * when the interface up, the changed MTU must be
3609 * reflected in the idev as well as routers.
3611 if (idev->cnf.mtu6 != dev->mtu &&
3612 dev->mtu >= IPV6_MIN_MTU) {
3613 rt6_mtu_change(dev, dev->mtu);
3614 idev->cnf.mtu6 = dev->mtu;
3616 idev->tstamp = jiffies;
3617 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3620 * If the changed mtu during down is lower than
3621 * IPV6_MIN_MTU stop IPv6 on this interface.
3623 if (dev->mtu < IPV6_MIN_MTU)
3624 addrconf_ifdown(dev, dev != net->loopback_dev);
3629 case NETDEV_UNREGISTER:
3631 * Remove all addresses from this interface.
3633 addrconf_ifdown(dev, event != NETDEV_DOWN);
3636 case NETDEV_CHANGENAME:
3638 snmp6_unregister_dev(idev);
3639 addrconf_sysctl_unregister(idev);
3640 err = addrconf_sysctl_register(idev);
3642 return notifier_from_errno(err);
3643 err = snmp6_register_dev(idev);
3645 addrconf_sysctl_unregister(idev);
3646 return notifier_from_errno(err);
3651 case NETDEV_PRE_TYPE_CHANGE:
3652 case NETDEV_POST_TYPE_CHANGE:
3654 addrconf_type_change(dev, event);
3657 case NETDEV_CHANGEUPPER:
3660 /* flush all routes if dev is linked to or unlinked from
3661 * an L3 master device (e.g., VRF)
3663 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3664 addrconf_ifdown(dev, 0);
3671 * addrconf module should be notified of a device going up
3673 static struct notifier_block ipv6_dev_notf = {
3674 .notifier_call = addrconf_notify,
3675 .priority = ADDRCONF_NOTIFY_PRIORITY,
3678 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3680 struct inet6_dev *idev;
3683 idev = __in6_dev_get(dev);
3685 if (event == NETDEV_POST_TYPE_CHANGE)
3686 ipv6_mc_remap(idev);
3687 else if (event == NETDEV_PRE_TYPE_CHANGE)
3688 ipv6_mc_unmap(idev);
3691 static bool addr_is_local(const struct in6_addr *addr)
3693 return ipv6_addr_type(addr) &
3694 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3697 static int addrconf_ifdown(struct net_device *dev, int how)
3699 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3700 struct net *net = dev_net(dev);
3701 struct inet6_dev *idev;
3702 struct inet6_ifaddr *ifa, *tmp;
3703 bool keep_addr = false;
3708 rt6_disable_ip(dev, event);
3710 idev = __in6_dev_get(dev);
3715 * Step 1: remove reference to ipv6 device from parent device.
3721 /* protected by rtnl_lock */
3722 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3724 /* Step 1.5: remove snmp6 entry */
3725 snmp6_unregister_dev(idev);
3729 /* combine the user config with event to determine if permanent
3730 * addresses are to be removed from address hash table
3732 if (!how && !idev->cnf.disable_ipv6) {
3733 /* aggregate the system setting and interface setting */
3734 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3737 _keep_addr = idev->cnf.keep_addr_on_down;
3739 keep_addr = (_keep_addr > 0);
3742 /* Step 2: clear hash table */
3743 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3744 struct hlist_head *h = &inet6_addr_lst[i];
3746 spin_lock_bh(&addrconf_hash_lock);
3748 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3749 if (ifa->idev == idev) {
3750 addrconf_del_dad_work(ifa);
3751 /* combined flag + permanent flag decide if
3752 * address is retained on a down event
3755 !(ifa->flags & IFA_F_PERMANENT) ||
3756 addr_is_local(&ifa->addr)) {
3757 hlist_del_init_rcu(&ifa->addr_lst);
3762 spin_unlock_bh(&addrconf_hash_lock);
3765 write_lock_bh(&idev->lock);
3767 addrconf_del_rs_timer(idev);
3769 /* Step 2: clear flags for stateless addrconf */
3771 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3773 /* Step 3: clear tempaddr list */
3774 while (!list_empty(&idev->tempaddr_list)) {
3775 ifa = list_first_entry(&idev->tempaddr_list,
3776 struct inet6_ifaddr, tmp_list);
3777 list_del(&ifa->tmp_list);
3778 write_unlock_bh(&idev->lock);
3779 spin_lock_bh(&ifa->lock);
3782 in6_ifa_put(ifa->ifpub);
3785 spin_unlock_bh(&ifa->lock);
3787 write_lock_bh(&idev->lock);
3790 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3791 struct fib6_info *rt = NULL;
3794 addrconf_del_dad_work(ifa);
3796 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3797 !addr_is_local(&ifa->addr);
3799 write_unlock_bh(&idev->lock);
3800 spin_lock_bh(&ifa->lock);
3803 /* set state to skip the notifier below */
3804 state = INET6_IFADDR_STATE_DEAD;
3805 ifa->state = INET6_IFADDR_STATE_PREDAD;
3806 if (!(ifa->flags & IFA_F_NODAD))
3807 ifa->flags |= IFA_F_TENTATIVE;
3813 ifa->state = INET6_IFADDR_STATE_DEAD;
3816 spin_unlock_bh(&ifa->lock);
3819 ip6_del_rt(net, rt);
3821 if (state != INET6_IFADDR_STATE_DEAD) {
3822 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3823 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3825 if (idev->cnf.forwarding)
3826 addrconf_leave_anycast(ifa);
3827 addrconf_leave_solict(ifa->idev, &ifa->addr);
3830 write_lock_bh(&idev->lock);
3832 list_del_rcu(&ifa->if_list);
3837 write_unlock_bh(&idev->lock);
3839 /* Step 5: Discard anycast and multicast list */
3841 ipv6_ac_destroy_dev(idev);
3842 ipv6_mc_destroy_dev(idev);
3847 idev->tstamp = jiffies;
3849 /* Last: Shot the device (if unregistered) */
3851 addrconf_sysctl_unregister(idev);
3852 neigh_parms_release(&nd_tbl, idev->nd_parms);
3853 neigh_ifdown(&nd_tbl, dev);
3859 static void addrconf_rs_timer(struct timer_list *t)
3861 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3862 struct net_device *dev = idev->dev;
3863 struct in6_addr lladdr;
3865 write_lock(&idev->lock);
3866 if (idev->dead || !(idev->if_flags & IF_READY))
3869 if (!ipv6_accept_ra(idev))
3872 /* Announcement received after solicitation was sent */
3873 if (idev->if_flags & IF_RA_RCVD)
3876 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3877 write_unlock(&idev->lock);
3878 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3879 ndisc_send_rs(dev, &lladdr,
3880 &in6addr_linklocal_allrouters);
3884 write_lock(&idev->lock);
3885 idev->rs_interval = rfc3315_s14_backoff_update(
3886 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3887 /* The wait after the last probe can be shorter */
3888 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3889 idev->cnf.rtr_solicits) ?
3890 idev->cnf.rtr_solicit_delay :
3894 * Note: we do not support deprecated "all on-link"
3895 * assumption any longer.
3897 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3901 write_unlock(&idev->lock);
3907 * Duplicate Address Detection
3909 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3911 unsigned long rand_num;
3912 struct inet6_dev *idev = ifp->idev;
3915 if (ifp->flags & IFA_F_OPTIMISTIC)
3918 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3921 if (idev->cnf.enhanced_dad ||
3922 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3924 get_random_bytes(&nonce, 6);
3927 ifp->dad_nonce = nonce;
3928 ifp->dad_probes = idev->cnf.dad_transmits;
3929 addrconf_mod_dad_work(ifp, rand_num);
3932 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3934 struct inet6_dev *idev = ifp->idev;
3935 struct net_device *dev = idev->dev;
3936 bool bump_id, notify = false;
3939 addrconf_join_solict(dev, &ifp->addr);
3941 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3943 read_lock_bh(&idev->lock);
3944 spin_lock(&ifp->lock);
3945 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3949 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3950 (net->ipv6.devconf_all->accept_dad < 1 &&
3951 idev->cnf.accept_dad < 1) ||
3952 !(ifp->flags&IFA_F_TENTATIVE) ||
3953 ifp->flags & IFA_F_NODAD) {
3954 bool send_na = false;
3956 if (ifp->flags & IFA_F_TENTATIVE &&
3957 !(ifp->flags & IFA_F_OPTIMISTIC))
3959 bump_id = ifp->flags & IFA_F_TENTATIVE;
3960 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3961 spin_unlock(&ifp->lock);
3962 read_unlock_bh(&idev->lock);
3964 addrconf_dad_completed(ifp, bump_id, send_na);
3968 if (!(idev->if_flags & IF_READY)) {
3969 spin_unlock(&ifp->lock);
3970 read_unlock_bh(&idev->lock);
3972 * If the device is not ready:
3973 * - keep it tentative if it is a permanent address.
3974 * - otherwise, kill it.
3977 addrconf_dad_stop(ifp, 0);
3982 * Optimistic nodes can start receiving
3985 if (ifp->flags & IFA_F_OPTIMISTIC) {
3986 ip6_ins_rt(net, ifp->rt);
3987 if (ipv6_use_optimistic_addr(net, idev)) {
3988 /* Because optimistic nodes can use this address,
3989 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3995 addrconf_dad_kick(ifp);
3997 spin_unlock(&ifp->lock);
3998 read_unlock_bh(&idev->lock);
4000 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4003 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4005 bool begin_dad = false;
4007 spin_lock_bh(&ifp->lock);
4008 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4009 ifp->state = INET6_IFADDR_STATE_PREDAD;
4012 spin_unlock_bh(&ifp->lock);
4015 addrconf_mod_dad_work(ifp, 0);
4018 static void addrconf_dad_work(struct work_struct *w)
4020 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4021 struct inet6_ifaddr,
4023 struct inet6_dev *idev = ifp->idev;
4024 bool bump_id, disable_ipv6 = false;
4025 struct in6_addr mcaddr;
4031 } action = DAD_PROCESS;
4035 spin_lock_bh(&ifp->lock);
4036 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4038 ifp->state = INET6_IFADDR_STATE_DAD;
4039 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4041 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4043 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4044 idev->cnf.accept_dad > 1) &&
4045 !idev->cnf.disable_ipv6 &&
4046 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4047 struct in6_addr addr;
4049 addr.s6_addr32[0] = htonl(0xfe800000);
4050 addr.s6_addr32[1] = 0;
4052 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4053 ipv6_addr_equal(&ifp->addr, &addr)) {
4054 /* DAD failed for link-local based on MAC */
4055 idev->cnf.disable_ipv6 = 1;
4057 pr_info("%s: IPv6 being disabled!\n",
4058 ifp->idev->dev->name);
4059 disable_ipv6 = true;
4063 spin_unlock_bh(&ifp->lock);
4065 if (action == DAD_BEGIN) {
4066 addrconf_dad_begin(ifp);
4068 } else if (action == DAD_ABORT) {
4070 addrconf_dad_stop(ifp, 1);
4072 addrconf_ifdown(idev->dev, 0);
4076 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4079 write_lock_bh(&idev->lock);
4080 if (idev->dead || !(idev->if_flags & IF_READY)) {
4081 write_unlock_bh(&idev->lock);
4085 spin_lock(&ifp->lock);
4086 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4087 spin_unlock(&ifp->lock);
4088 write_unlock_bh(&idev->lock);
4092 if (ifp->dad_probes == 0) {
4093 bool send_na = false;
4096 * DAD was successful
4099 if (ifp->flags & IFA_F_TENTATIVE &&
4100 !(ifp->flags & IFA_F_OPTIMISTIC))
4102 bump_id = ifp->flags & IFA_F_TENTATIVE;
4103 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4104 spin_unlock(&ifp->lock);
4105 write_unlock_bh(&idev->lock);
4107 addrconf_dad_completed(ifp, bump_id, send_na);
4113 addrconf_mod_dad_work(ifp,
4114 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4115 spin_unlock(&ifp->lock);
4116 write_unlock_bh(&idev->lock);
4118 /* send a neighbour solicitation for our addr */
4119 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4120 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4127 /* ifp->idev must be at least read locked */
4128 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4130 struct inet6_ifaddr *ifpiter;
4131 struct inet6_dev *idev = ifp->idev;
4133 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4134 if (ifpiter->scope > IFA_LINK)
4136 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4137 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4138 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4145 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4148 struct net_device *dev = ifp->idev->dev;
4149 struct in6_addr lladdr;
4150 bool send_rs, send_mld;
4152 addrconf_del_dad_work(ifp);
4155 * Configure the address for reception. Now it is valid.
4158 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4160 /* If added prefix is link local and we are prepared to process
4161 router advertisements, start sending router solicitations.
4164 read_lock_bh(&ifp->idev->lock);
4165 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4166 send_rs = send_mld &&
4167 ipv6_accept_ra(ifp->idev) &&
4168 ifp->idev->cnf.rtr_solicits != 0 &&
4169 (dev->flags&IFF_LOOPBACK) == 0;
4170 read_unlock_bh(&ifp->idev->lock);
4172 /* While dad is in progress mld report's source address is in6_addrany.
4173 * Resend with proper ll now.
4176 ipv6_mc_dad_complete(ifp->idev);
4178 /* send unsolicited NA if enabled */
4180 (ifp->idev->cnf.ndisc_notify ||
4181 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4182 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4183 /*router=*/ !!ifp->idev->cnf.forwarding,
4184 /*solicited=*/ false, /*override=*/ true,
4190 * If a host as already performed a random delay
4191 * [...] as part of DAD [...] there is no need
4192 * to delay again before sending the first RS
4194 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4196 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4198 write_lock_bh(&ifp->idev->lock);
4199 spin_lock(&ifp->lock);
4200 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4201 ifp->idev->cnf.rtr_solicit_interval);
4202 ifp->idev->rs_probes = 1;
4203 ifp->idev->if_flags |= IF_RS_SENT;
4204 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4205 spin_unlock(&ifp->lock);
4206 write_unlock_bh(&ifp->idev->lock);
4210 rt_genid_bump_ipv6(dev_net(dev));
4212 /* Make sure that a new temporary address will be created
4213 * before this temporary address becomes deprecated.
4215 if (ifp->flags & IFA_F_TEMPORARY)
4216 addrconf_verify_rtnl();
4219 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4221 struct inet6_ifaddr *ifp;
4223 read_lock_bh(&idev->lock);
4224 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4225 spin_lock(&ifp->lock);
4226 if ((ifp->flags & IFA_F_TENTATIVE &&
4227 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4229 ifp->state = INET6_IFADDR_STATE_PREDAD;
4230 addrconf_dad_kick(ifp);
4232 spin_unlock(&ifp->lock);
4234 read_unlock_bh(&idev->lock);
4237 #ifdef CONFIG_PROC_FS
4238 struct if6_iter_state {
4239 struct seq_net_private p;
4244 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4246 struct if6_iter_state *state = seq->private;
4247 struct net *net = seq_file_net(seq);
4248 struct inet6_ifaddr *ifa = NULL;
4251 /* initial bucket if pos is 0 */
4257 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4258 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4260 if (!net_eq(dev_net(ifa->idev->dev), net))
4262 /* sync with offset */
4263 if (p < state->offset) {
4270 /* prepare for next bucket */
4277 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4278 struct inet6_ifaddr *ifa)
4280 struct if6_iter_state *state = seq->private;
4281 struct net *net = seq_file_net(seq);
4283 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4284 if (!net_eq(dev_net(ifa->idev->dev), net))
4291 while (++state->bucket < IN6_ADDR_HSIZE) {
4292 hlist_for_each_entry_rcu(ifa,
4293 &inet6_addr_lst[state->bucket], addr_lst) {
4294 if (!net_eq(dev_net(ifa->idev->dev), net))
4303 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4307 return if6_get_first(seq, *pos);
4310 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4312 struct inet6_ifaddr *ifa;
4314 ifa = if6_get_next(seq, v);
4319 static void if6_seq_stop(struct seq_file *seq, void *v)
4325 static int if6_seq_show(struct seq_file *seq, void *v)
4327 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4328 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4330 ifp->idev->dev->ifindex,
4334 ifp->idev->dev->name);
4338 static const struct seq_operations if6_seq_ops = {
4339 .start = if6_seq_start,
4340 .next = if6_seq_next,
4341 .show = if6_seq_show,
4342 .stop = if6_seq_stop,
4345 static int __net_init if6_proc_net_init(struct net *net)
4347 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4348 sizeof(struct if6_iter_state)))
4353 static void __net_exit if6_proc_net_exit(struct net *net)
4355 remove_proc_entry("if_inet6", net->proc_net);
4358 static struct pernet_operations if6_proc_net_ops = {
4359 .init = if6_proc_net_init,
4360 .exit = if6_proc_net_exit,
4363 int __init if6_proc_init(void)
4365 return register_pernet_subsys(&if6_proc_net_ops);
4368 void if6_proc_exit(void)
4370 unregister_pernet_subsys(&if6_proc_net_ops);
4372 #endif /* CONFIG_PROC_FS */
4374 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4375 /* Check if address is a home address configured on any interface. */
4376 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4378 unsigned int hash = inet6_addr_hash(net, addr);
4379 struct inet6_ifaddr *ifp = NULL;
4383 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4384 if (!net_eq(dev_net(ifp->idev->dev), net))
4386 if (ipv6_addr_equal(&ifp->addr, addr) &&
4387 (ifp->flags & IFA_F_HOMEADDRESS)) {
4398 * Periodic address status verification
4401 static void addrconf_verify_rtnl(void)
4403 unsigned long now, next, next_sec, next_sched;
4404 struct inet6_ifaddr *ifp;
4411 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4413 cancel_delayed_work(&addr_chk_work);
4415 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4417 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4420 /* When setting preferred_lft to a value not zero or
4421 * infinity, while valid_lft is infinity
4422 * IFA_F_PERMANENT has a non-infinity life time.
4424 if ((ifp->flags & IFA_F_PERMANENT) &&
4425 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4428 spin_lock(&ifp->lock);
4429 /* We try to batch several events at once. */
4430 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4432 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4433 age >= ifp->valid_lft) {
4434 spin_unlock(&ifp->lock);
4438 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4439 spin_unlock(&ifp->lock);
4441 } else if (age >= ifp->prefered_lft) {
4442 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4445 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4447 ifp->flags |= IFA_F_DEPRECATED;
4450 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4451 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4452 next = ifp->tstamp + ifp->valid_lft * HZ;
4454 spin_unlock(&ifp->lock);
4459 ipv6_ifa_notify(0, ifp);
4463 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4464 !(ifp->flags&IFA_F_TENTATIVE)) {
4465 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4466 ifp->idev->cnf.dad_transmits *
4467 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4469 if (age >= ifp->prefered_lft - regen_advance) {
4470 struct inet6_ifaddr *ifpub = ifp->ifpub;
4471 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4472 next = ifp->tstamp + ifp->prefered_lft * HZ;
4473 if (!ifp->regen_count && ifpub) {
4476 in6_ifa_hold(ifpub);
4477 spin_unlock(&ifp->lock);
4479 spin_lock(&ifpub->lock);
4480 ifpub->regen_count = 0;
4481 spin_unlock(&ifpub->lock);
4482 rcu_read_unlock_bh();
4483 ipv6_create_tempaddr(ifpub, ifp, true);
4489 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4490 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4491 spin_unlock(&ifp->lock);
4493 /* ifp->prefered_lft <= ifp->valid_lft */
4494 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4495 next = ifp->tstamp + ifp->prefered_lft * HZ;
4496 spin_unlock(&ifp->lock);
4501 next_sec = round_jiffies_up(next);
4504 /* If rounded timeout is accurate enough, accept it. */
4505 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4506 next_sched = next_sec;
4508 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4509 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4510 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4512 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4513 now, next, next_sec, next_sched);
4514 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4515 rcu_read_unlock_bh();
4518 static void addrconf_verify_work(struct work_struct *w)
4521 addrconf_verify_rtnl();
4525 static void addrconf_verify(void)
4527 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4530 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4531 struct in6_addr **peer_pfx)
4533 struct in6_addr *pfx = NULL;
4538 pfx = nla_data(addr);
4541 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4543 pfx = nla_data(local);
4549 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4550 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4551 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4552 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4553 [IFA_FLAGS] = { .len = sizeof(u32) },
4554 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4555 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4559 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4560 struct netlink_ext_ack *extack)
4562 struct net *net = sock_net(skb->sk);
4563 struct ifaddrmsg *ifm;
4564 struct nlattr *tb[IFA_MAX+1];
4565 struct in6_addr *pfx, *peer_pfx;
4569 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4570 ifa_ipv6_policy, extack);
4574 ifm = nlmsg_data(nlh);
4575 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4579 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4581 /* We ignore other flags so far. */
4582 ifa_flags &= IFA_F_MANAGETEMPADDR;
4584 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4585 ifm->ifa_prefixlen);
4588 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4589 unsigned long expires, u32 flags)
4591 struct fib6_info *f6i;
4594 f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
4595 ifp->idev->dev, 0, RTF_DEFAULT, true);
4599 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4600 if (f6i->fib6_metric != prio) {
4601 /* delete old one */
4602 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4605 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4606 ifp->rt_priority, ifp->idev->dev,
4607 expires, flags, GFP_KERNEL);
4610 fib6_clean_expires(f6i);
4612 fib6_set_expires(f6i, expires);
4614 fib6_info_release(f6i);
4620 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4624 unsigned long timeout;
4625 bool was_managetempaddr;
4626 bool had_prefixroute;
4630 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4633 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4634 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4637 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4638 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4640 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4641 if (addrconf_finite_timeout(timeout)) {
4642 expires = jiffies_to_clock_t(timeout * HZ);
4643 cfg->valid_lft = timeout;
4644 flags = RTF_EXPIRES;
4648 cfg->ifa_flags |= IFA_F_PERMANENT;
4651 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4652 if (addrconf_finite_timeout(timeout)) {
4654 cfg->ifa_flags |= IFA_F_DEPRECATED;
4655 cfg->preferred_lft = timeout;
4658 spin_lock_bh(&ifp->lock);
4659 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4660 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4661 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4662 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4663 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4664 IFA_F_NOPREFIXROUTE);
4665 ifp->flags |= cfg->ifa_flags;
4666 ifp->tstamp = jiffies;
4667 ifp->valid_lft = cfg->valid_lft;
4668 ifp->prefered_lft = cfg->preferred_lft;
4670 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4671 ifp->rt_priority = cfg->rt_priority;
4673 spin_unlock_bh(&ifp->lock);
4674 if (!(ifp->flags&IFA_F_TENTATIVE))
4675 ipv6_ifa_notify(0, ifp);
4677 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4680 if (had_prefixroute)
4681 rc = modify_prefix_route(ifp, expires, flags);
4683 /* prefix route could have been deleted; if so restore it */
4684 if (rc == -ENOENT) {
4685 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4686 ifp->rt_priority, ifp->idev->dev,
4687 expires, flags, GFP_KERNEL);
4689 } else if (had_prefixroute) {
4690 enum cleanup_prefix_rt_t action;
4691 unsigned long rt_expires;
4693 write_lock_bh(&ifp->idev->lock);
4694 action = check_cleanup_prefix_route(ifp, &rt_expires);
4695 write_unlock_bh(&ifp->idev->lock);
4697 if (action != CLEANUP_PREFIX_RT_NOP) {
4698 cleanup_prefix_route(ifp, rt_expires,
4699 action == CLEANUP_PREFIX_RT_DEL);
4703 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4704 if (was_managetempaddr &&
4705 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4707 cfg->preferred_lft = 0;
4709 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4710 cfg->preferred_lft, !was_managetempaddr,
4714 addrconf_verify_rtnl();
4720 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4721 struct netlink_ext_ack *extack)
4723 struct net *net = sock_net(skb->sk);
4724 struct ifaddrmsg *ifm;
4725 struct nlattr *tb[IFA_MAX+1];
4726 struct in6_addr *peer_pfx;
4727 struct inet6_ifaddr *ifa;
4728 struct net_device *dev;
4729 struct inet6_dev *idev;
4730 struct ifa6_config cfg;
4733 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4734 ifa_ipv6_policy, extack);
4738 memset(&cfg, 0, sizeof(cfg));
4740 ifm = nlmsg_data(nlh);
4741 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4745 cfg.peer_pfx = peer_pfx;
4746 cfg.plen = ifm->ifa_prefixlen;
4747 if (tb[IFA_RT_PRIORITY])
4748 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4750 cfg.valid_lft = INFINITY_LIFE_TIME;
4751 cfg.preferred_lft = INFINITY_LIFE_TIME;
4753 if (tb[IFA_CACHEINFO]) {
4754 struct ifa_cacheinfo *ci;
4756 ci = nla_data(tb[IFA_CACHEINFO]);
4757 cfg.valid_lft = ci->ifa_valid;
4758 cfg.preferred_lft = ci->ifa_prefered;
4761 dev = __dev_get_by_index(net, ifm->ifa_index);
4766 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4768 cfg.ifa_flags = ifm->ifa_flags;
4770 /* We ignore other flags so far. */
4771 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4772 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4773 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4775 idev = ipv6_find_idev(dev);
4777 return PTR_ERR(idev);
4779 if (!ipv6_allow_optimistic_dad(net, idev))
4780 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4782 if (cfg.ifa_flags & IFA_F_NODAD &&
4783 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4784 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4788 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4791 * It would be best to check for !NLM_F_CREATE here but
4792 * userspace already relies on not having to provide this.
4794 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4797 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4798 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4801 err = inet6_addr_modify(ifa, &cfg);
4808 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4809 u8 scope, int ifindex)
4811 struct ifaddrmsg *ifm;
4813 ifm = nlmsg_data(nlh);
4814 ifm->ifa_family = AF_INET6;
4815 ifm->ifa_prefixlen = prefixlen;
4816 ifm->ifa_flags = flags;
4817 ifm->ifa_scope = scope;
4818 ifm->ifa_index = ifindex;
4821 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4822 unsigned long tstamp, u32 preferred, u32 valid)
4824 struct ifa_cacheinfo ci;
4826 ci.cstamp = cstamp_delta(cstamp);
4827 ci.tstamp = cstamp_delta(tstamp);
4828 ci.ifa_prefered = preferred;
4829 ci.ifa_valid = valid;
4831 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4834 static inline int rt_scope(int ifa_scope)
4836 if (ifa_scope & IFA_HOST)
4837 return RT_SCOPE_HOST;
4838 else if (ifa_scope & IFA_LINK)
4839 return RT_SCOPE_LINK;
4840 else if (ifa_scope & IFA_SITE)
4841 return RT_SCOPE_SITE;
4843 return RT_SCOPE_UNIVERSE;
4846 static inline int inet6_ifaddr_msgsize(void)
4848 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4849 + nla_total_size(16) /* IFA_LOCAL */
4850 + nla_total_size(16) /* IFA_ADDRESS */
4851 + nla_total_size(sizeof(struct ifa_cacheinfo))
4852 + nla_total_size(4) /* IFA_FLAGS */
4853 + nla_total_size(4) /* IFA_RT_PRIORITY */;
4862 struct inet6_fill_args {
4869 enum addr_type_t type;
4872 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4873 struct inet6_fill_args *args)
4875 struct nlmsghdr *nlh;
4876 u32 preferred, valid;
4878 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4879 sizeof(struct ifaddrmsg), args->flags);
4883 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4884 ifa->idev->dev->ifindex);
4886 if (args->netnsid >= 0 &&
4887 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4890 if (!((ifa->flags&IFA_F_PERMANENT) &&
4891 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4892 preferred = ifa->prefered_lft;
4893 valid = ifa->valid_lft;
4894 if (preferred != INFINITY_LIFE_TIME) {
4895 long tval = (jiffies - ifa->tstamp)/HZ;
4896 if (preferred > tval)
4900 if (valid != INFINITY_LIFE_TIME) {
4908 preferred = INFINITY_LIFE_TIME;
4909 valid = INFINITY_LIFE_TIME;
4912 if (!ipv6_addr_any(&ifa->peer_addr)) {
4913 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4914 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4917 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4920 if (ifa->rt_priority &&
4921 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4924 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4927 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4930 nlmsg_end(skb, nlh);
4934 nlmsg_cancel(skb, nlh);
4938 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4939 struct inet6_fill_args *args)
4941 struct nlmsghdr *nlh;
4942 u8 scope = RT_SCOPE_UNIVERSE;
4943 int ifindex = ifmca->idev->dev->ifindex;
4945 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4946 scope = RT_SCOPE_SITE;
4948 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4949 sizeof(struct ifaddrmsg), args->flags);
4953 if (args->netnsid >= 0 &&
4954 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4957 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4958 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4959 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4960 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4961 nlmsg_cancel(skb, nlh);
4965 nlmsg_end(skb, nlh);
4969 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4970 struct inet6_fill_args *args)
4972 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4973 int ifindex = dev ? dev->ifindex : 1;
4974 struct nlmsghdr *nlh;
4975 u8 scope = RT_SCOPE_UNIVERSE;
4977 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4978 scope = RT_SCOPE_SITE;
4980 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4981 sizeof(struct ifaddrmsg), args->flags);
4985 if (args->netnsid >= 0 &&
4986 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4989 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4990 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4991 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4992 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4993 nlmsg_cancel(skb, nlh);
4997 nlmsg_end(skb, nlh);
5001 /* called with rcu_read_lock() */
5002 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5003 struct netlink_callback *cb, int s_ip_idx,
5004 struct inet6_fill_args *fillargs)
5006 struct ifmcaddr6 *ifmca;
5007 struct ifacaddr6 *ifaca;
5011 read_lock_bh(&idev->lock);
5012 switch (fillargs->type) {
5013 case UNICAST_ADDR: {
5014 struct inet6_ifaddr *ifa;
5015 fillargs->event = RTM_NEWADDR;
5017 /* unicast address incl. temp addr */
5018 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5019 if (ip_idx < s_ip_idx)
5021 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5024 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5030 case MULTICAST_ADDR:
5031 fillargs->event = RTM_GETMULTICAST;
5033 /* multicast address */
5034 for (ifmca = idev->mc_list; ifmca;
5035 ifmca = ifmca->next, ip_idx++) {
5036 if (ip_idx < s_ip_idx)
5038 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5044 fillargs->event = RTM_GETANYCAST;
5045 /* anycast address */
5046 for (ifaca = idev->ac_list; ifaca;
5047 ifaca = ifaca->aca_next, ip_idx++) {
5048 if (ip_idx < s_ip_idx)
5050 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5058 read_unlock_bh(&idev->lock);
5059 cb->args[2] = ip_idx;
5063 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5064 struct inet6_fill_args *fillargs,
5065 struct net **tgt_net, struct sock *sk,
5066 struct netlink_callback *cb)
5068 struct netlink_ext_ack *extack = cb->extack;
5069 struct nlattr *tb[IFA_MAX+1];
5070 struct ifaddrmsg *ifm;
5073 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5074 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5078 ifm = nlmsg_data(nlh);
5079 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5080 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5084 fillargs->ifindex = ifm->ifa_index;
5085 if (fillargs->ifindex) {
5086 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5087 fillargs->flags |= NLM_F_DUMP_FILTERED;
5090 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5091 ifa_ipv6_policy, extack);
5095 for (i = 0; i <= IFA_MAX; ++i) {
5099 if (i == IFA_TARGET_NETNSID) {
5102 fillargs->netnsid = nla_get_s32(tb[i]);
5103 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5105 fillargs->netnsid = -1;
5106 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5107 return PTR_ERR(net);
5111 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5119 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5120 enum addr_type_t type)
5122 const struct nlmsghdr *nlh = cb->nlh;
5123 struct inet6_fill_args fillargs = {
5124 .portid = NETLINK_CB(cb->skb).portid,
5125 .seq = cb->nlh->nlmsg_seq,
5126 .flags = NLM_F_MULTI,
5130 struct net *net = sock_net(skb->sk);
5131 struct net *tgt_net = net;
5132 int idx, s_idx, s_ip_idx;
5134 struct net_device *dev;
5135 struct inet6_dev *idev;
5136 struct hlist_head *head;
5140 s_idx = idx = cb->args[1];
5141 s_ip_idx = cb->args[2];
5143 if (cb->strict_check) {
5144 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5150 if (fillargs.ifindex) {
5151 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5156 idev = __in6_dev_get(dev);
5158 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5168 cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5169 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5171 head = &tgt_net->dev_index_head[h];
5172 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5175 if (h > s_h || idx > s_idx)
5177 idev = __in6_dev_get(dev);
5181 if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5193 if (fillargs.netnsid >= 0)
5196 return skb->len ? : err;
5199 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5201 enum addr_type_t type = UNICAST_ADDR;
5203 return inet6_dump_addr(skb, cb, type);
5206 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5208 enum addr_type_t type = MULTICAST_ADDR;
5210 return inet6_dump_addr(skb, cb, type);
5214 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5216 enum addr_type_t type = ANYCAST_ADDR;
5218 return inet6_dump_addr(skb, cb, type);
5221 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5222 const struct nlmsghdr *nlh,
5224 struct netlink_ext_ack *extack)
5226 struct ifaddrmsg *ifm;
5229 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5230 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5234 ifm = nlmsg_data(nlh);
5235 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5236 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5240 if (!netlink_strict_get_check(skb))
5241 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5242 ifa_ipv6_policy, extack);
5244 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5245 ifa_ipv6_policy, extack);
5249 for (i = 0; i <= IFA_MAX; i++) {
5254 case IFA_TARGET_NETNSID:
5259 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5267 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5268 struct netlink_ext_ack *extack)
5270 struct net *net = sock_net(in_skb->sk);
5271 struct inet6_fill_args fillargs = {
5272 .portid = NETLINK_CB(in_skb).portid,
5273 .seq = nlh->nlmsg_seq,
5274 .event = RTM_NEWADDR,
5278 struct net *tgt_net = net;
5279 struct ifaddrmsg *ifm;
5280 struct nlattr *tb[IFA_MAX+1];
5281 struct in6_addr *addr = NULL, *peer;
5282 struct net_device *dev = NULL;
5283 struct inet6_ifaddr *ifa;
5284 struct sk_buff *skb;
5287 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5291 if (tb[IFA_TARGET_NETNSID]) {
5292 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5294 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5296 if (IS_ERR(tgt_net))
5297 return PTR_ERR(tgt_net);
5300 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5304 ifm = nlmsg_data(nlh);
5306 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5308 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5310 err = -EADDRNOTAVAIL;
5314 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5320 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5322 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5323 WARN_ON(err == -EMSGSIZE);
5327 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5333 if (fillargs.netnsid >= 0)
5339 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5341 struct sk_buff *skb;
5342 struct net *net = dev_net(ifa->idev->dev);
5343 struct inet6_fill_args fillargs = {
5352 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5356 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5358 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5359 WARN_ON(err == -EMSGSIZE);
5363 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5367 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5370 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5371 __s32 *array, int bytes)
5373 BUG_ON(bytes < (DEVCONF_MAX * 4));
5375 memset(array, 0, bytes);
5376 array[DEVCONF_FORWARDING] = cnf->forwarding;
5377 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5378 array[DEVCONF_MTU6] = cnf->mtu6;
5379 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5380 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5381 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5382 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5383 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5384 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5385 jiffies_to_msecs(cnf->rtr_solicit_interval);
5386 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5387 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5388 array[DEVCONF_RTR_SOLICIT_DELAY] =
5389 jiffies_to_msecs(cnf->rtr_solicit_delay);
5390 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5391 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5392 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5393 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5394 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5395 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5396 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5397 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5398 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5399 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5400 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5401 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5402 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5403 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5404 #ifdef CONFIG_IPV6_ROUTER_PREF
5405 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5406 array[DEVCONF_RTR_PROBE_INTERVAL] =
5407 jiffies_to_msecs(cnf->rtr_probe_interval);
5408 #ifdef CONFIG_IPV6_ROUTE_INFO
5409 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5410 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5413 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5414 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5415 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5416 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5417 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5419 #ifdef CONFIG_IPV6_MROUTE
5420 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5422 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5423 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5424 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5425 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5426 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5427 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5428 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5429 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5430 /* we omit DEVCONF_STABLE_SECRET for now */
5431 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5432 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5433 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5434 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5435 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5436 #ifdef CONFIG_IPV6_SEG6_HMAC
5437 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5439 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5440 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5441 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5442 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5445 static inline size_t inet6_ifla6_size(void)
5447 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5448 + nla_total_size(sizeof(struct ifla_cacheinfo))
5449 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5450 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5451 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5452 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5453 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5457 static inline size_t inet6_if_nlmsg_size(void)
5459 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5460 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5461 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5462 + nla_total_size(4) /* IFLA_MTU */
5463 + nla_total_size(4) /* IFLA_LINK */
5464 + nla_total_size(1) /* IFLA_OPERSTATE */
5465 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5468 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5472 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5475 /* Use put_unaligned() because stats may not be aligned for u64. */
5476 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5477 for (i = 1; i < ICMP6_MIB_MAX; i++)
5478 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5480 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5483 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5484 int bytes, size_t syncpoff)
5487 u64 buff[IPSTATS_MIB_MAX];
5488 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5492 memset(buff, 0, sizeof(buff));
5493 buff[0] = IPSTATS_MIB_MAX;
5495 for_each_possible_cpu(c) {
5496 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5497 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5500 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5501 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5504 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5508 case IFLA_INET6_STATS:
5509 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5510 offsetof(struct ipstats_mib, syncp));
5512 case IFLA_INET6_ICMP6STATS:
5513 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5518 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5519 u32 ext_filter_mask)
5522 struct ifla_cacheinfo ci;
5524 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5525 goto nla_put_failure;
5526 ci.max_reasm_len = IPV6_MAXPLEN;
5527 ci.tstamp = cstamp_delta(idev->tstamp);
5528 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5529 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5530 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5531 goto nla_put_failure;
5532 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5534 goto nla_put_failure;
5535 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5537 /* XXX - MC not implemented */
5539 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5542 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5544 goto nla_put_failure;
5545 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5547 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5549 goto nla_put_failure;
5550 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5552 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5554 goto nla_put_failure;
5556 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5557 goto nla_put_failure;
5559 read_lock_bh(&idev->lock);
5560 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5561 read_unlock_bh(&idev->lock);
5569 static size_t inet6_get_link_af_size(const struct net_device *dev,
5570 u32 ext_filter_mask)
5572 if (!__in6_dev_get(dev))
5575 return inet6_ifla6_size();
5578 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5579 u32 ext_filter_mask)
5581 struct inet6_dev *idev = __in6_dev_get(dev);
5586 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5592 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5594 struct inet6_ifaddr *ifp;
5595 struct net_device *dev = idev->dev;
5596 bool clear_token, update_rs = false;
5597 struct in6_addr ll_addr;
5603 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5605 if (!ipv6_accept_ra(idev))
5607 if (idev->cnf.rtr_solicits == 0)
5610 write_lock_bh(&idev->lock);
5612 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5613 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5615 write_unlock_bh(&idev->lock);
5617 clear_token = ipv6_addr_any(token);
5621 if (!idev->dead && (idev->if_flags & IF_READY) &&
5622 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5623 IFA_F_OPTIMISTIC)) {
5624 /* If we're not ready, then normal ifup will take care
5625 * of this. Otherwise, we need to request our rs here.
5627 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5632 write_lock_bh(&idev->lock);
5635 idev->if_flags |= IF_RS_SENT;
5636 idev->rs_interval = rfc3315_s14_backoff_init(
5637 idev->cnf.rtr_solicit_interval);
5638 idev->rs_probes = 1;
5639 addrconf_mod_rs_timer(idev, idev->rs_interval);
5642 /* Well, that's kinda nasty ... */
5643 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5644 spin_lock(&ifp->lock);
5645 if (ifp->tokenized) {
5647 ifp->prefered_lft = 0;
5649 spin_unlock(&ifp->lock);
5652 write_unlock_bh(&idev->lock);
5653 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5654 addrconf_verify_rtnl();
5658 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5659 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5660 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5663 static int check_addr_gen_mode(int mode)
5665 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5666 mode != IN6_ADDR_GEN_MODE_NONE &&
5667 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5668 mode != IN6_ADDR_GEN_MODE_RANDOM)
5673 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5676 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5677 !idev->cnf.stable_secret.initialized &&
5678 !net->ipv6.devconf_dflt->stable_secret.initialized)
5683 static int inet6_validate_link_af(const struct net_device *dev,
5684 const struct nlattr *nla)
5686 struct nlattr *tb[IFLA_INET6_MAX + 1];
5687 struct inet6_dev *idev = NULL;
5691 idev = __in6_dev_get(dev);
5693 return -EAFNOSUPPORT;
5696 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5697 inet6_af_policy, NULL);
5701 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5704 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5705 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5707 if (check_addr_gen_mode(mode) < 0)
5709 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5716 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5718 struct inet6_dev *idev = __in6_dev_get(dev);
5719 struct nlattr *tb[IFLA_INET6_MAX + 1];
5722 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5725 if (tb[IFLA_INET6_TOKEN]) {
5726 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5731 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5732 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5734 idev->cnf.addr_gen_mode = mode;
5740 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5741 u32 portid, u32 seq, int event, unsigned int flags)
5743 struct net_device *dev = idev->dev;
5744 struct ifinfomsg *hdr;
5745 struct nlmsghdr *nlh;
5748 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5752 hdr = nlmsg_data(nlh);
5753 hdr->ifi_family = AF_INET6;
5755 hdr->ifi_type = dev->type;
5756 hdr->ifi_index = dev->ifindex;
5757 hdr->ifi_flags = dev_get_flags(dev);
5758 hdr->ifi_change = 0;
5760 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5762 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5763 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5764 (dev->ifindex != dev_get_iflink(dev) &&
5765 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5766 nla_put_u8(skb, IFLA_OPERSTATE,
5767 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5768 goto nla_put_failure;
5769 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5771 goto nla_put_failure;
5773 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5774 goto nla_put_failure;
5776 nla_nest_end(skb, protoinfo);
5777 nlmsg_end(skb, nlh);
5781 nlmsg_cancel(skb, nlh);
5785 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5786 struct netlink_ext_ack *extack)
5788 struct ifinfomsg *ifm;
5790 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5791 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5795 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5796 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5800 ifm = nlmsg_data(nlh);
5801 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5802 ifm->ifi_change || ifm->ifi_index) {
5803 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5810 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5812 struct net *net = sock_net(skb->sk);
5815 struct net_device *dev;
5816 struct inet6_dev *idev;
5817 struct hlist_head *head;
5819 /* only requests using strict checking can pass data to
5820 * influence the dump
5822 if (cb->strict_check) {
5823 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5830 s_idx = cb->args[1];
5833 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5835 head = &net->dev_index_head[h];
5836 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5839 idev = __in6_dev_get(dev);
5842 if (inet6_fill_ifinfo(skb, idev,
5843 NETLINK_CB(cb->skb).portid,
5845 RTM_NEWLINK, NLM_F_MULTI) < 0)
5859 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5861 struct sk_buff *skb;
5862 struct net *net = dev_net(idev->dev);
5865 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5869 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5871 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5872 WARN_ON(err == -EMSGSIZE);
5876 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5880 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5883 static inline size_t inet6_prefix_nlmsg_size(void)
5885 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5886 + nla_total_size(sizeof(struct in6_addr))
5887 + nla_total_size(sizeof(struct prefix_cacheinfo));
5890 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5891 struct prefix_info *pinfo, u32 portid, u32 seq,
5892 int event, unsigned int flags)
5894 struct prefixmsg *pmsg;
5895 struct nlmsghdr *nlh;
5896 struct prefix_cacheinfo ci;
5898 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5902 pmsg = nlmsg_data(nlh);
5903 pmsg->prefix_family = AF_INET6;
5904 pmsg->prefix_pad1 = 0;
5905 pmsg->prefix_pad2 = 0;
5906 pmsg->prefix_ifindex = idev->dev->ifindex;
5907 pmsg->prefix_len = pinfo->prefix_len;
5908 pmsg->prefix_type = pinfo->type;
5909 pmsg->prefix_pad3 = 0;
5910 pmsg->prefix_flags = 0;
5912 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5913 if (pinfo->autoconf)
5914 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5916 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5917 goto nla_put_failure;
5918 ci.preferred_time = ntohl(pinfo->prefered);
5919 ci.valid_time = ntohl(pinfo->valid);
5920 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5921 goto nla_put_failure;
5922 nlmsg_end(skb, nlh);
5926 nlmsg_cancel(skb, nlh);
5930 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5931 struct prefix_info *pinfo)
5933 struct sk_buff *skb;
5934 struct net *net = dev_net(idev->dev);
5937 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5941 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5943 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5944 WARN_ON(err == -EMSGSIZE);
5948 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5952 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5955 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5957 struct net *net = dev_net(ifp->idev->dev);
5962 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5967 * If the address was optimistic
5968 * we inserted the route at the start of
5969 * our DAD process, so we don't need
5972 if (!rcu_access_pointer(ifp->rt->fib6_node))
5973 ip6_ins_rt(net, ifp->rt);
5974 if (ifp->idev->cnf.forwarding)
5975 addrconf_join_anycast(ifp);
5976 if (!ipv6_addr_any(&ifp->peer_addr))
5977 addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5978 ifp->idev->dev, 0, 0,
5982 if (ifp->idev->cnf.forwarding)
5983 addrconf_leave_anycast(ifp);
5984 addrconf_leave_solict(ifp->idev, &ifp->addr);
5985 if (!ipv6_addr_any(&ifp->peer_addr)) {
5986 struct fib6_info *rt;
5988 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5989 ifp->idev->dev, 0, 0,
5992 ip6_del_rt(net, rt);
5995 ip6_del_rt(net, ifp->rt);
5998 rt_genid_bump_ipv6(net);
6001 atomic_inc(&net->ipv6.dev_addr_genid);
6004 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6007 if (likely(ifp->idev->dead == 0))
6008 __ipv6_ifa_notify(event, ifp);
6009 rcu_read_unlock_bh();
6012 #ifdef CONFIG_SYSCTL
6015 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6016 void __user *buffer, size_t *lenp, loff_t *ppos)
6018 int *valp = ctl->data;
6021 struct ctl_table lctl;
6025 * ctl->data points to idev->cnf.forwarding, we should
6026 * not modify it until we get the rtnl lock.
6031 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6034 ret = addrconf_fixup_forwarding(ctl, valp, val);
6041 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6042 void __user *buffer, size_t *lenp, loff_t *ppos)
6044 struct inet6_dev *idev = ctl->extra1;
6045 int min_mtu = IPV6_MIN_MTU;
6046 struct ctl_table lctl;
6049 lctl.extra1 = &min_mtu;
6050 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6052 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6055 static void dev_disable_change(struct inet6_dev *idev)
6057 struct netdev_notifier_info info;
6059 if (!idev || !idev->dev)
6062 netdev_notifier_info_init(&info, idev->dev);
6063 if (idev->cnf.disable_ipv6)
6064 addrconf_notify(NULL, NETDEV_DOWN, &info);
6066 addrconf_notify(NULL, NETDEV_UP, &info);
6069 static void addrconf_disable_change(struct net *net, __s32 newf)
6071 struct net_device *dev;
6072 struct inet6_dev *idev;
6074 for_each_netdev(net, dev) {
6075 idev = __in6_dev_get(dev);
6077 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6078 idev->cnf.disable_ipv6 = newf;
6080 dev_disable_change(idev);
6085 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6090 if (!rtnl_trylock())
6091 return restart_syscall();
6093 net = (struct net *)table->extra2;
6097 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6102 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6103 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6104 addrconf_disable_change(net, newf);
6105 } else if ((!newf) ^ (!old))
6106 dev_disable_change((struct inet6_dev *)table->extra1);
6113 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6114 void __user *buffer, size_t *lenp, loff_t *ppos)
6116 int *valp = ctl->data;
6119 struct ctl_table lctl;
6123 * ctl->data points to idev->cnf.disable_ipv6, we should
6124 * not modify it until we get the rtnl lock.
6129 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6132 ret = addrconf_disable_ipv6(ctl, valp, val);
6139 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6140 void __user *buffer, size_t *lenp, loff_t *ppos)
6142 int *valp = ctl->data;
6147 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6150 if (write && old != new) {
6151 struct net *net = ctl->extra2;
6153 if (!rtnl_trylock())
6154 return restart_syscall();
6156 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6157 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6158 NETCONFA_PROXY_NEIGH,
6159 NETCONFA_IFINDEX_DEFAULT,
6160 net->ipv6.devconf_dflt);
6161 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6162 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6163 NETCONFA_PROXY_NEIGH,
6164 NETCONFA_IFINDEX_ALL,
6165 net->ipv6.devconf_all);
6167 struct inet6_dev *idev = ctl->extra1;
6169 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6170 NETCONFA_PROXY_NEIGH,
6180 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6181 void __user *buffer, size_t *lenp,
6186 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6187 struct net *net = (struct net *)ctl->extra2;
6188 struct ctl_table tmp = {
6190 .maxlen = sizeof(new_val),
6194 if (!rtnl_trylock())
6195 return restart_syscall();
6197 new_val = *((u32 *)ctl->data);
6199 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6204 if (check_addr_gen_mode(new_val) < 0) {
6210 if (check_stable_privacy(idev, net, new_val) < 0) {
6215 if (idev->cnf.addr_gen_mode != new_val) {
6216 idev->cnf.addr_gen_mode = new_val;
6217 addrconf_dev_config(idev->dev);
6219 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6220 struct net_device *dev;
6222 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6223 for_each_netdev(net, dev) {
6224 idev = __in6_dev_get(dev);
6226 idev->cnf.addr_gen_mode != new_val) {
6227 idev->cnf.addr_gen_mode = new_val;
6228 addrconf_dev_config(idev->dev);
6233 *((u32 *)ctl->data) = new_val;
6242 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6243 void __user *buffer, size_t *lenp,
6247 struct in6_addr addr;
6248 char str[IPV6_MAX_STRLEN];
6249 struct ctl_table lctl = *ctl;
6250 struct net *net = ctl->extra2;
6251 struct ipv6_stable_secret *secret = ctl->data;
6253 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6256 lctl.maxlen = IPV6_MAX_STRLEN;
6259 if (!rtnl_trylock())
6260 return restart_syscall();
6262 if (!write && !secret->initialized) {
6267 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6268 if (err >= sizeof(str)) {
6273 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6277 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6282 secret->initialized = true;
6283 secret->secret = addr;
6285 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6286 struct net_device *dev;
6288 for_each_netdev(net, dev) {
6289 struct inet6_dev *idev = __in6_dev_get(dev);
6292 idev->cnf.addr_gen_mode =
6293 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6297 struct inet6_dev *idev = ctl->extra1;
6299 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6309 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6311 void __user *buffer,
6315 int *valp = ctl->data;
6318 struct ctl_table lctl;
6321 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6322 * we should not modify it until we get the rtnl lock.
6327 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6330 ret = addrconf_fixup_linkdown(ctl, valp, val);
6337 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6341 rt->dst.flags |= DST_NOPOLICY;
6343 rt->dst.flags &= ~DST_NOPOLICY;
6348 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6350 struct inet6_ifaddr *ifa;
6352 read_lock_bh(&idev->lock);
6353 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6354 spin_lock(&ifa->lock);
6356 /* host routes only use builtin fib6_nh */
6357 struct fib6_nh *nh = ifa->rt->fib6_nh;
6361 ifa->rt->dst_nopolicy = val ? true : false;
6362 if (nh->rt6i_pcpu) {
6363 for_each_possible_cpu(cpu) {
6364 struct rt6_info **rtp;
6366 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6367 addrconf_set_nopolicy(*rtp, val);
6372 spin_unlock(&ifa->lock);
6374 read_unlock_bh(&idev->lock);
6378 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6380 struct inet6_dev *idev;
6383 if (!rtnl_trylock())
6384 return restart_syscall();
6388 net = (struct net *)ctl->extra2;
6389 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6394 if (valp == &net->ipv6.devconf_all->disable_policy) {
6395 struct net_device *dev;
6397 for_each_netdev(net, dev) {
6398 idev = __in6_dev_get(dev);
6400 addrconf_disable_policy_idev(idev, val);
6403 idev = (struct inet6_dev *)ctl->extra1;
6404 addrconf_disable_policy_idev(idev, val);
6412 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6413 void __user *buffer, size_t *lenp,
6416 int *valp = ctl->data;
6419 struct ctl_table lctl;
6424 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6426 if (write && (*valp != val))
6427 ret = addrconf_disable_policy(ctl, valp, val);
6435 static int minus_one = -1;
6436 static const int two_five_five = 255;
6438 static const struct ctl_table addrconf_sysctl[] = {
6440 .procname = "forwarding",
6441 .data = &ipv6_devconf.forwarding,
6442 .maxlen = sizeof(int),
6444 .proc_handler = addrconf_sysctl_forward,
6447 .procname = "hop_limit",
6448 .data = &ipv6_devconf.hop_limit,
6449 .maxlen = sizeof(int),
6451 .proc_handler = proc_dointvec_minmax,
6452 .extra1 = (void *)SYSCTL_ONE,
6453 .extra2 = (void *)&two_five_five,
6457 .data = &ipv6_devconf.mtu6,
6458 .maxlen = sizeof(int),
6460 .proc_handler = addrconf_sysctl_mtu,
6463 .procname = "accept_ra",
6464 .data = &ipv6_devconf.accept_ra,
6465 .maxlen = sizeof(int),
6467 .proc_handler = proc_dointvec,
6470 .procname = "accept_redirects",
6471 .data = &ipv6_devconf.accept_redirects,
6472 .maxlen = sizeof(int),
6474 .proc_handler = proc_dointvec,
6477 .procname = "autoconf",
6478 .data = &ipv6_devconf.autoconf,
6479 .maxlen = sizeof(int),
6481 .proc_handler = proc_dointvec,
6484 .procname = "dad_transmits",
6485 .data = &ipv6_devconf.dad_transmits,
6486 .maxlen = sizeof(int),
6488 .proc_handler = proc_dointvec,
6491 .procname = "router_solicitations",
6492 .data = &ipv6_devconf.rtr_solicits,
6493 .maxlen = sizeof(int),
6495 .proc_handler = proc_dointvec_minmax,
6496 .extra1 = &minus_one,
6499 .procname = "router_solicitation_interval",
6500 .data = &ipv6_devconf.rtr_solicit_interval,
6501 .maxlen = sizeof(int),
6503 .proc_handler = proc_dointvec_jiffies,
6506 .procname = "router_solicitation_max_interval",
6507 .data = &ipv6_devconf.rtr_solicit_max_interval,
6508 .maxlen = sizeof(int),
6510 .proc_handler = proc_dointvec_jiffies,
6513 .procname = "router_solicitation_delay",
6514 .data = &ipv6_devconf.rtr_solicit_delay,
6515 .maxlen = sizeof(int),
6517 .proc_handler = proc_dointvec_jiffies,
6520 .procname = "force_mld_version",
6521 .data = &ipv6_devconf.force_mld_version,
6522 .maxlen = sizeof(int),
6524 .proc_handler = proc_dointvec,
6527 .procname = "mldv1_unsolicited_report_interval",
6529 &ipv6_devconf.mldv1_unsolicited_report_interval,
6530 .maxlen = sizeof(int),
6532 .proc_handler = proc_dointvec_ms_jiffies,
6535 .procname = "mldv2_unsolicited_report_interval",
6537 &ipv6_devconf.mldv2_unsolicited_report_interval,
6538 .maxlen = sizeof(int),
6540 .proc_handler = proc_dointvec_ms_jiffies,
6543 .procname = "use_tempaddr",
6544 .data = &ipv6_devconf.use_tempaddr,
6545 .maxlen = sizeof(int),
6547 .proc_handler = proc_dointvec,
6550 .procname = "temp_valid_lft",
6551 .data = &ipv6_devconf.temp_valid_lft,
6552 .maxlen = sizeof(int),
6554 .proc_handler = proc_dointvec,
6557 .procname = "temp_prefered_lft",
6558 .data = &ipv6_devconf.temp_prefered_lft,
6559 .maxlen = sizeof(int),
6561 .proc_handler = proc_dointvec,
6564 .procname = "regen_max_retry",
6565 .data = &ipv6_devconf.regen_max_retry,
6566 .maxlen = sizeof(int),
6568 .proc_handler = proc_dointvec,
6571 .procname = "max_desync_factor",
6572 .data = &ipv6_devconf.max_desync_factor,
6573 .maxlen = sizeof(int),
6575 .proc_handler = proc_dointvec,
6578 .procname = "max_addresses",
6579 .data = &ipv6_devconf.max_addresses,
6580 .maxlen = sizeof(int),
6582 .proc_handler = proc_dointvec,
6585 .procname = "accept_ra_defrtr",
6586 .data = &ipv6_devconf.accept_ra_defrtr,
6587 .maxlen = sizeof(int),
6589 .proc_handler = proc_dointvec,
6592 .procname = "accept_ra_min_hop_limit",
6593 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6594 .maxlen = sizeof(int),
6596 .proc_handler = proc_dointvec,
6599 .procname = "accept_ra_pinfo",
6600 .data = &ipv6_devconf.accept_ra_pinfo,
6601 .maxlen = sizeof(int),
6603 .proc_handler = proc_dointvec,
6605 #ifdef CONFIG_IPV6_ROUTER_PREF
6607 .procname = "accept_ra_rtr_pref",
6608 .data = &ipv6_devconf.accept_ra_rtr_pref,
6609 .maxlen = sizeof(int),
6611 .proc_handler = proc_dointvec,
6614 .procname = "router_probe_interval",
6615 .data = &ipv6_devconf.rtr_probe_interval,
6616 .maxlen = sizeof(int),
6618 .proc_handler = proc_dointvec_jiffies,
6620 #ifdef CONFIG_IPV6_ROUTE_INFO
6622 .procname = "accept_ra_rt_info_min_plen",
6623 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6624 .maxlen = sizeof(int),
6626 .proc_handler = proc_dointvec,
6629 .procname = "accept_ra_rt_info_max_plen",
6630 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6631 .maxlen = sizeof(int),
6633 .proc_handler = proc_dointvec,
6638 .procname = "proxy_ndp",
6639 .data = &ipv6_devconf.proxy_ndp,
6640 .maxlen = sizeof(int),
6642 .proc_handler = addrconf_sysctl_proxy_ndp,
6645 .procname = "accept_source_route",
6646 .data = &ipv6_devconf.accept_source_route,
6647 .maxlen = sizeof(int),
6649 .proc_handler = proc_dointvec,
6651 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6653 .procname = "optimistic_dad",
6654 .data = &ipv6_devconf.optimistic_dad,
6655 .maxlen = sizeof(int),
6657 .proc_handler = proc_dointvec,
6660 .procname = "use_optimistic",
6661 .data = &ipv6_devconf.use_optimistic,
6662 .maxlen = sizeof(int),
6664 .proc_handler = proc_dointvec,
6667 #ifdef CONFIG_IPV6_MROUTE
6669 .procname = "mc_forwarding",
6670 .data = &ipv6_devconf.mc_forwarding,
6671 .maxlen = sizeof(int),
6673 .proc_handler = proc_dointvec,
6677 .procname = "disable_ipv6",
6678 .data = &ipv6_devconf.disable_ipv6,
6679 .maxlen = sizeof(int),
6681 .proc_handler = addrconf_sysctl_disable,
6684 .procname = "accept_dad",
6685 .data = &ipv6_devconf.accept_dad,
6686 .maxlen = sizeof(int),
6688 .proc_handler = proc_dointvec,
6691 .procname = "force_tllao",
6692 .data = &ipv6_devconf.force_tllao,
6693 .maxlen = sizeof(int),
6695 .proc_handler = proc_dointvec
6698 .procname = "ndisc_notify",
6699 .data = &ipv6_devconf.ndisc_notify,
6700 .maxlen = sizeof(int),
6702 .proc_handler = proc_dointvec
6705 .procname = "suppress_frag_ndisc",
6706 .data = &ipv6_devconf.suppress_frag_ndisc,
6707 .maxlen = sizeof(int),
6709 .proc_handler = proc_dointvec
6712 .procname = "accept_ra_from_local",
6713 .data = &ipv6_devconf.accept_ra_from_local,
6714 .maxlen = sizeof(int),
6716 .proc_handler = proc_dointvec,
6719 .procname = "accept_ra_mtu",
6720 .data = &ipv6_devconf.accept_ra_mtu,
6721 .maxlen = sizeof(int),
6723 .proc_handler = proc_dointvec,
6726 .procname = "stable_secret",
6727 .data = &ipv6_devconf.stable_secret,
6728 .maxlen = IPV6_MAX_STRLEN,
6730 .proc_handler = addrconf_sysctl_stable_secret,
6733 .procname = "use_oif_addrs_only",
6734 .data = &ipv6_devconf.use_oif_addrs_only,
6735 .maxlen = sizeof(int),
6737 .proc_handler = proc_dointvec,
6740 .procname = "ignore_routes_with_linkdown",
6741 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6742 .maxlen = sizeof(int),
6744 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6747 .procname = "drop_unicast_in_l2_multicast",
6748 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6749 .maxlen = sizeof(int),
6751 .proc_handler = proc_dointvec,
6754 .procname = "drop_unsolicited_na",
6755 .data = &ipv6_devconf.drop_unsolicited_na,
6756 .maxlen = sizeof(int),
6758 .proc_handler = proc_dointvec,
6761 .procname = "keep_addr_on_down",
6762 .data = &ipv6_devconf.keep_addr_on_down,
6763 .maxlen = sizeof(int),
6765 .proc_handler = proc_dointvec,
6769 .procname = "seg6_enabled",
6770 .data = &ipv6_devconf.seg6_enabled,
6771 .maxlen = sizeof(int),
6773 .proc_handler = proc_dointvec,
6775 #ifdef CONFIG_IPV6_SEG6_HMAC
6777 .procname = "seg6_require_hmac",
6778 .data = &ipv6_devconf.seg6_require_hmac,
6779 .maxlen = sizeof(int),
6781 .proc_handler = proc_dointvec,
6785 .procname = "enhanced_dad",
6786 .data = &ipv6_devconf.enhanced_dad,
6787 .maxlen = sizeof(int),
6789 .proc_handler = proc_dointvec,
6792 .procname = "addr_gen_mode",
6793 .data = &ipv6_devconf.addr_gen_mode,
6794 .maxlen = sizeof(int),
6796 .proc_handler = addrconf_sysctl_addr_gen_mode,
6799 .procname = "disable_policy",
6800 .data = &ipv6_devconf.disable_policy,
6801 .maxlen = sizeof(int),
6803 .proc_handler = addrconf_sysctl_disable_policy,
6806 .procname = "ndisc_tclass",
6807 .data = &ipv6_devconf.ndisc_tclass,
6808 .maxlen = sizeof(int),
6810 .proc_handler = proc_dointvec_minmax,
6811 .extra1 = (void *)SYSCTL_ZERO,
6812 .extra2 = (void *)&two_five_five,
6819 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6820 struct inet6_dev *idev, struct ipv6_devconf *p)
6823 struct ctl_table *table;
6824 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6826 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6830 for (i = 0; table[i].data; i++) {
6831 table[i].data += (char *)p - (char *)&ipv6_devconf;
6832 /* If one of these is already set, then it is not safe to
6833 * overwrite either of them: this makes proc_dointvec_minmax
6836 if (!table[i].extra1 && !table[i].extra2) {
6837 table[i].extra1 = idev; /* embedded; no ref */
6838 table[i].extra2 = net;
6842 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6844 p->sysctl_header = register_net_sysctl(net, path, table);
6845 if (!p->sysctl_header)
6848 if (!strcmp(dev_name, "all"))
6849 ifindex = NETCONFA_IFINDEX_ALL;
6850 else if (!strcmp(dev_name, "default"))
6851 ifindex = NETCONFA_IFINDEX_DEFAULT;
6853 ifindex = idev->dev->ifindex;
6854 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6864 static void __addrconf_sysctl_unregister(struct net *net,
6865 struct ipv6_devconf *p, int ifindex)
6867 struct ctl_table *table;
6869 if (!p->sysctl_header)
6872 table = p->sysctl_header->ctl_table_arg;
6873 unregister_net_sysctl_table(p->sysctl_header);
6874 p->sysctl_header = NULL;
6877 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6880 static int addrconf_sysctl_register(struct inet6_dev *idev)
6884 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6887 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6888 &ndisc_ifinfo_sysctl_change);
6891 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6894 neigh_sysctl_unregister(idev->nd_parms);
6899 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6901 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6902 idev->dev->ifindex);
6903 neigh_sysctl_unregister(idev->nd_parms);
6909 static int __net_init addrconf_init_net(struct net *net)
6912 struct ipv6_devconf *all, *dflt;
6914 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6918 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6920 goto err_alloc_dflt;
6922 if (IS_ENABLED(CONFIG_SYSCTL) &&
6923 sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6924 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6925 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6928 /* these will be inherited by all namespaces */
6929 dflt->autoconf = ipv6_defaults.autoconf;
6930 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6932 dflt->stable_secret.initialized = false;
6933 all->stable_secret.initialized = false;
6935 net->ipv6.devconf_all = all;
6936 net->ipv6.devconf_dflt = dflt;
6938 #ifdef CONFIG_SYSCTL
6939 err = __addrconf_sysctl_register(net, "all", NULL, all);
6943 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6949 #ifdef CONFIG_SYSCTL
6951 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6961 static void __net_exit addrconf_exit_net(struct net *net)
6963 #ifdef CONFIG_SYSCTL
6964 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6965 NETCONFA_IFINDEX_DEFAULT);
6966 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6967 NETCONFA_IFINDEX_ALL);
6969 kfree(net->ipv6.devconf_dflt);
6970 kfree(net->ipv6.devconf_all);
6973 static struct pernet_operations addrconf_ops = {
6974 .init = addrconf_init_net,
6975 .exit = addrconf_exit_net,
6978 static struct rtnl_af_ops inet6_ops __read_mostly = {
6980 .fill_link_af = inet6_fill_link_af,
6981 .get_link_af_size = inet6_get_link_af_size,
6982 .validate_link_af = inet6_validate_link_af,
6983 .set_link_af = inet6_set_link_af,
6987 * Init / cleanup code
6990 int __init addrconf_init(void)
6992 struct inet6_dev *idev;
6995 err = ipv6_addr_label_init();
6997 pr_crit("%s: cannot initialize default policy table: %d\n",
7002 err = register_pernet_subsys(&addrconf_ops);
7006 addrconf_wq = create_workqueue("ipv6_addrconf");
7012 /* The addrconf netdev notifier requires that loopback_dev
7013 * has it's ipv6 private information allocated and setup
7014 * before it can bring up and give link-local addresses
7015 * to other devices which are up.
7017 * Unfortunately, loopback_dev is not necessarily the first
7018 * entry in the global dev_base list of net devices. In fact,
7019 * it is likely to be the very last entry on that list.
7020 * So this causes the notifier registry below to try and
7021 * give link-local addresses to all devices besides loopback_dev
7022 * first, then loopback_dev, which cases all the non-loopback_dev
7023 * devices to fail to get a link-local address.
7025 * So, as a temporary fix, allocate the ipv6 structure for
7026 * loopback_dev first by hand.
7027 * Longer term, all of the dependencies ipv6 has upon the loopback
7028 * device and it being up should be removed.
7031 idev = ipv6_add_dev(init_net.loopback_dev);
7034 err = PTR_ERR(idev);
7038 ip6_route_init_special_entries();
7040 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7041 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7043 register_netdevice_notifier(&ipv6_dev_notf);
7047 rtnl_af_register(&inet6_ops);
7049 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7050 NULL, inet6_dump_ifinfo, 0);
7054 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7055 inet6_rtm_newaddr, NULL, 0);
7058 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7059 inet6_rtm_deladdr, NULL, 0);
7062 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7063 inet6_rtm_getaddr, inet6_dump_ifaddr,
7064 RTNL_FLAG_DOIT_UNLOCKED);
7067 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7068 NULL, inet6_dump_ifmcaddr, 0);
7071 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7072 NULL, inet6_dump_ifacaddr, 0);
7075 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7076 inet6_netconf_get_devconf,
7077 inet6_netconf_dump_devconf,
7078 RTNL_FLAG_DOIT_UNLOCKED);
7081 err = ipv6_addr_label_rtnl_register();
7087 rtnl_unregister_all(PF_INET6);
7088 rtnl_af_unregister(&inet6_ops);
7089 unregister_netdevice_notifier(&ipv6_dev_notf);
7091 destroy_workqueue(addrconf_wq);
7093 unregister_pernet_subsys(&addrconf_ops);
7095 ipv6_addr_label_cleanup();
7100 void addrconf_cleanup(void)
7102 struct net_device *dev;
7105 unregister_netdevice_notifier(&ipv6_dev_notf);
7106 unregister_pernet_subsys(&addrconf_ops);
7107 ipv6_addr_label_cleanup();
7109 rtnl_af_unregister(&inet6_ops);
7113 /* clean dev list */
7114 for_each_netdev(&init_net, dev) {
7115 if (__in6_dev_get(dev) == NULL)
7117 addrconf_ifdown(dev, 1);
7119 addrconf_ifdown(init_net.loopback_dev, 2);
7124 spin_lock_bh(&addrconf_hash_lock);
7125 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7126 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7127 spin_unlock_bh(&addrconf_hash_lock);
7128 cancel_delayed_work(&addr_chk_work);
7131 destroy_workqueue(addrconf_wq);