2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
60 #include <linux/sysctl.h>
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
68 #include <net/net_namespace.h>
72 #include <net/af_ieee802154.h>
73 #include <net/firewire.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86 #include <linux/random.h>
87 #include <linux/uaccess.h>
88 #include <asm/unaligned.h>
90 #include <linux/proc_fs.h>
91 #include <linux/seq_file.h>
92 #include <linux/export.h>
94 /* Set to 3 to get tracing... */
98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #define INFINITY_LIFE_TIME 0xFFFFFFFF
105 static inline u32 cstamp_delta(unsigned long cstamp)
107 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 static void addrconf_sysctl_register(struct inet6_dev *idev);
112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
125 static void ipv6_regen_rndid(unsigned long data);
127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
128 static int ipv6_count_addresses(struct inet6_dev *idev);
131 * Configured unicast address hash table
133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
134 static DEFINE_SPINLOCK(addrconf_hash_lock);
136 static void addrconf_verify(void);
137 static void addrconf_verify_rtnl(void);
138 static void addrconf_verify_work(struct work_struct *);
140 static struct workqueue_struct *addrconf_wq;
141 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
143 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
144 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_type_change(struct net_device *dev,
147 unsigned long event);
148 static int addrconf_ifdown(struct net_device *dev, int how);
150 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
152 const struct net_device *dev,
153 u32 flags, u32 noflags);
155 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
156 static void addrconf_dad_work(struct work_struct *w);
157 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
158 static void addrconf_dad_run(struct inet6_dev *idev);
159 static void addrconf_rs_timer(unsigned long data);
160 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
161 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
164 struct prefix_info *pinfo);
165 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
166 struct net_device *dev);
168 static struct ipv6_devconf ipv6_devconf __read_mostly = {
170 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
171 .mtu6 = IPV6_MIN_MTU,
173 .accept_redirects = 1,
175 .force_mld_version = 0,
176 .mldv1_unsolicited_report_interval = 10 * HZ,
177 .mldv2_unsolicited_report_interval = HZ,
179 .rtr_solicits = MAX_RTR_SOLICITATIONS,
180 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
181 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
183 .temp_valid_lft = TEMP_VALID_LIFETIME,
184 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
185 .regen_max_retry = REGEN_MAX_RETRY,
186 .max_desync_factor = MAX_DESYNC_FACTOR,
187 .max_addresses = IPV6_MAX_ADDRESSES,
188 .accept_ra_defrtr = 1,
189 .accept_ra_min_hop_limit= 1,
190 .accept_ra_pinfo = 1,
191 #ifdef CONFIG_IPV6_ROUTER_PREF
192 .accept_ra_rtr_pref = 1,
193 .rtr_probe_interval = 60 * HZ,
194 #ifdef CONFIG_IPV6_ROUTE_INFO
195 .accept_ra_rt_info_max_plen = 0,
199 .accept_source_route = 0, /* we do not accept RH0 by default. */
202 .suppress_frag_ndisc = 1,
205 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
207 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
208 .mtu6 = IPV6_MIN_MTU,
210 .accept_redirects = 1,
212 .force_mld_version = 0,
213 .mldv1_unsolicited_report_interval = 10 * HZ,
214 .mldv2_unsolicited_report_interval = HZ,
216 .rtr_solicits = MAX_RTR_SOLICITATIONS,
217 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
218 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
220 .temp_valid_lft = TEMP_VALID_LIFETIME,
221 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
222 .regen_max_retry = REGEN_MAX_RETRY,
223 .max_desync_factor = MAX_DESYNC_FACTOR,
224 .max_addresses = IPV6_MAX_ADDRESSES,
225 .accept_ra_defrtr = 1,
226 .accept_ra_min_hop_limit= 1,
227 .accept_ra_pinfo = 1,
228 #ifdef CONFIG_IPV6_ROUTER_PREF
229 .accept_ra_rtr_pref = 1,
230 .rtr_probe_interval = 60 * HZ,
231 #ifdef CONFIG_IPV6_ROUTE_INFO
232 .accept_ra_rt_info_max_plen = 0,
236 .accept_source_route = 0, /* we do not accept RH0 by default. */
239 .suppress_frag_ndisc = 1,
242 /* Check if a valid qdisc is available */
243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
245 return !qdisc_tx_is_noop(dev);
248 static void addrconf_del_rs_timer(struct inet6_dev *idev)
250 if (del_timer(&idev->rs_timer))
254 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
256 if (cancel_delayed_work(&ifp->dad_work))
260 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
263 if (!timer_pending(&idev->rs_timer))
265 mod_timer(&idev->rs_timer, jiffies + when);
268 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
271 if (!delayed_work_pending(&ifp->dad_work))
273 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
276 static int snmp6_alloc_dev(struct inet6_dev *idev)
280 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
281 if (!idev->stats.ipv6)
284 for_each_possible_cpu(i) {
285 struct ipstats_mib *addrconf_stats;
286 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
287 u64_stats_init(&addrconf_stats->syncp);
291 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
293 if (!idev->stats.icmpv6dev)
295 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
297 if (!idev->stats.icmpv6msgdev)
303 kfree(idev->stats.icmpv6dev);
305 free_percpu(idev->stats.ipv6);
310 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
312 struct inet6_dev *ndev;
316 if (dev->mtu < IPV6_MIN_MTU)
319 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
324 rwlock_init(&ndev->lock);
326 INIT_LIST_HEAD(&ndev->addr_list);
327 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
328 (unsigned long)ndev);
329 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
330 ndev->cnf.mtu6 = dev->mtu;
331 ndev->cnf.sysctl = NULL;
332 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
333 if (ndev->nd_parms == NULL) {
337 if (ndev->cnf.forwarding)
338 dev_disable_lro(dev);
339 /* We refer to the device */
342 if (snmp6_alloc_dev(ndev) < 0) {
344 "%s: cannot allocate memory for statistics; dev=%s.\n",
345 __func__, dev->name);
346 neigh_parms_release(&nd_tbl, ndev->nd_parms);
352 if (snmp6_register_dev(ndev) < 0) {
354 "%s: cannot create /proc/net/dev_snmp6/%s\n",
355 __func__, dev->name);
356 neigh_parms_release(&nd_tbl, ndev->nd_parms);
358 in6_dev_finish_destroy(ndev);
362 /* One reference from device. We must do this before
363 * we invoke __ipv6_regen_rndid().
367 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
368 ndev->cnf.accept_dad = -1;
370 #if IS_ENABLED(CONFIG_IPV6_SIT)
371 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
372 pr_info("%s: Disabled Multicast RS\n", dev->name);
373 ndev->cnf.rtr_solicits = 0;
377 INIT_LIST_HEAD(&ndev->tempaddr_list);
378 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
379 if ((dev->flags&IFF_LOOPBACK) ||
380 dev->type == ARPHRD_TUNNEL ||
381 dev->type == ARPHRD_TUNNEL6 ||
382 dev->type == ARPHRD_SIT ||
383 dev->type == ARPHRD_NONE) {
384 ndev->cnf.use_tempaddr = -1;
387 ipv6_regen_rndid((unsigned long) ndev);
390 ndev->token = in6addr_any;
392 if (netif_running(dev) && addrconf_qdisc_ok(dev))
393 ndev->if_flags |= IF_READY;
395 ipv6_mc_init_dev(ndev);
396 ndev->tstamp = jiffies;
397 addrconf_sysctl_register(ndev);
398 /* protected by rtnl_lock */
399 rcu_assign_pointer(dev->ip6_ptr, ndev);
401 /* Join interface-local all-node multicast group */
402 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
404 /* Join all-node multicast group */
405 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
407 /* Join all-router multicast group if forwarding is set */
408 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
409 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
414 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
416 struct inet6_dev *idev;
420 idev = __in6_dev_get(dev);
422 idev = ipv6_add_dev(dev);
427 if (dev->flags&IFF_UP)
432 static int inet6_netconf_msgsize_devconf(int type)
434 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
435 + nla_total_size(4); /* NETCONFA_IFINDEX */
437 /* type -1 is used for ALL */
438 if (type == -1 || type == NETCONFA_FORWARDING)
439 size += nla_total_size(4);
440 #ifdef CONFIG_IPV6_MROUTE
441 if (type == -1 || type == NETCONFA_MC_FORWARDING)
442 size += nla_total_size(4);
444 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
445 size += nla_total_size(4);
450 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
451 struct ipv6_devconf *devconf, u32 portid,
452 u32 seq, int event, unsigned int flags,
455 struct nlmsghdr *nlh;
456 struct netconfmsg *ncm;
458 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
463 ncm = nlmsg_data(nlh);
464 ncm->ncm_family = AF_INET6;
466 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
467 goto nla_put_failure;
469 /* type -1 is used for ALL */
470 if ((type == -1 || type == NETCONFA_FORWARDING) &&
471 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
472 goto nla_put_failure;
473 #ifdef CONFIG_IPV6_MROUTE
474 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
475 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
476 devconf->mc_forwarding) < 0)
477 goto nla_put_failure;
479 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
480 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
481 goto nla_put_failure;
483 return nlmsg_end(skb, nlh);
486 nlmsg_cancel(skb, nlh);
490 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
491 struct ipv6_devconf *devconf)
496 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
500 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
501 RTM_NEWNETCONF, 0, type);
503 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
504 WARN_ON(err == -EMSGSIZE);
508 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
511 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
514 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
515 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
516 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
517 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
520 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
521 struct nlmsghdr *nlh)
523 struct net *net = sock_net(in_skb->sk);
524 struct nlattr *tb[NETCONFA_MAX+1];
525 struct netconfmsg *ncm;
527 struct ipv6_devconf *devconf;
528 struct inet6_dev *in6_dev;
529 struct net_device *dev;
533 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
534 devconf_ipv6_policy);
539 if (!tb[NETCONFA_IFINDEX])
542 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
544 case NETCONFA_IFINDEX_ALL:
545 devconf = net->ipv6.devconf_all;
547 case NETCONFA_IFINDEX_DEFAULT:
548 devconf = net->ipv6.devconf_dflt;
551 dev = __dev_get_by_index(net, ifindex);
554 in6_dev = __in6_dev_get(dev);
557 devconf = &in6_dev->cnf;
562 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
566 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
567 NETLINK_CB(in_skb).portid,
568 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
571 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
572 WARN_ON(err == -EMSGSIZE);
576 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
581 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
582 struct netlink_callback *cb)
584 struct net *net = sock_net(skb->sk);
587 struct net_device *dev;
588 struct inet6_dev *idev;
589 struct hlist_head *head;
592 s_idx = idx = cb->args[1];
594 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
596 head = &net->dev_index_head[h];
598 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
600 hlist_for_each_entry_rcu(dev, head, index_hlist) {
603 idev = __in6_dev_get(dev);
607 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
609 NETLINK_CB(cb->skb).portid,
617 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
623 if (h == NETDEV_HASHENTRIES) {
624 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
625 net->ipv6.devconf_all,
626 NETLINK_CB(cb->skb).portid,
628 RTM_NEWNETCONF, NLM_F_MULTI,
634 if (h == NETDEV_HASHENTRIES + 1) {
635 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
636 net->ipv6.devconf_dflt,
637 NETLINK_CB(cb->skb).portid,
639 RTM_NEWNETCONF, NLM_F_MULTI,
653 static void dev_forward_change(struct inet6_dev *idev)
655 struct net_device *dev;
656 struct inet6_ifaddr *ifa;
661 if (idev->cnf.forwarding)
662 dev_disable_lro(dev);
663 if (dev->flags & IFF_MULTICAST) {
664 if (idev->cnf.forwarding) {
665 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
666 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
667 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
669 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
670 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
671 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
675 list_for_each_entry(ifa, &idev->addr_list, if_list) {
676 if (ifa->flags&IFA_F_TENTATIVE)
678 if (idev->cnf.forwarding)
679 addrconf_join_anycast(ifa);
681 addrconf_leave_anycast(ifa);
683 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
684 dev->ifindex, &idev->cnf);
688 static void addrconf_forward_change(struct net *net, __s32 newf)
690 struct net_device *dev;
691 struct inet6_dev *idev;
693 for_each_netdev(net, dev) {
694 idev = __in6_dev_get(dev);
696 int changed = (!idev->cnf.forwarding) ^ (!newf);
697 idev->cnf.forwarding = newf;
699 dev_forward_change(idev);
704 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
710 return restart_syscall();
712 net = (struct net *)table->extra2;
716 if (p == &net->ipv6.devconf_dflt->forwarding) {
717 if ((!newf) ^ (!old))
718 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
719 NETCONFA_IFINDEX_DEFAULT,
720 net->ipv6.devconf_dflt);
725 if (p == &net->ipv6.devconf_all->forwarding) {
726 int old_dflt = net->ipv6.devconf_dflt->forwarding;
728 net->ipv6.devconf_dflt->forwarding = newf;
729 if ((!newf) ^ (!old_dflt))
730 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
731 NETCONFA_IFINDEX_DEFAULT,
732 net->ipv6.devconf_dflt);
734 addrconf_forward_change(net, newf);
735 if ((!newf) ^ (!old))
736 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
737 NETCONFA_IFINDEX_ALL,
738 net->ipv6.devconf_all);
739 } else if ((!newf) ^ (!old))
740 dev_forward_change((struct inet6_dev *)table->extra1);
744 rt6_purge_dflt_routers(net);
749 /* Nobody refers to this ifaddr, destroy it */
750 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
752 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
754 #ifdef NET_REFCNT_DEBUG
755 pr_debug("%s\n", __func__);
758 in6_dev_put(ifp->idev);
760 if (cancel_delayed_work(&ifp->dad_work))
761 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
764 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
765 pr_warn("Freeing alive inet6 address %p\n", ifp);
774 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
777 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
780 * Each device address list is sorted in order of scope -
781 * global before linklocal.
783 list_for_each(p, &idev->addr_list) {
784 struct inet6_ifaddr *ifa
785 = list_entry(p, struct inet6_ifaddr, if_list);
786 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
790 list_add_tail(&ifp->if_list, p);
793 static u32 inet6_addr_hash(const struct in6_addr *addr)
795 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
798 /* On success it returns ifp with increased reference count */
800 static struct inet6_ifaddr *
801 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
802 const struct in6_addr *peer_addr, int pfxlen,
803 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
805 struct inet6_ifaddr *ifa = NULL;
809 int addr_type = ipv6_addr_type(addr);
811 if (addr_type == IPV6_ADDR_ANY ||
812 addr_type & IPV6_ADDR_MULTICAST ||
813 (!(idev->dev->flags & IFF_LOOPBACK) &&
814 addr_type & IPV6_ADDR_LOOPBACK))
815 return ERR_PTR(-EADDRNOTAVAIL);
819 err = -ENODEV; /*XXX*/
823 if (idev->cnf.disable_ipv6) {
828 spin_lock(&addrconf_hash_lock);
830 /* Ignore adding duplicate addresses on an interface */
831 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
832 ADBG("ipv6_add_addr: already assigned\n");
837 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
840 ADBG("ipv6_add_addr: malloc failed\n");
845 rt = addrconf_dst_alloc(idev, addr, false);
851 neigh_parms_data_state_setall(idev->nd_parms);
855 ifa->peer_addr = *peer_addr;
857 spin_lock_init(&ifa->lock);
858 spin_lock_init(&ifa->state_lock);
859 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
860 INIT_HLIST_NODE(&ifa->addr_lst);
862 ifa->prefix_len = pfxlen;
863 ifa->flags = flags | IFA_F_TENTATIVE;
864 ifa->valid_lft = valid_lft;
865 ifa->prefered_lft = prefered_lft;
866 ifa->cstamp = ifa->tstamp = jiffies;
867 ifa->tokenized = false;
876 /* Add to big hash table */
877 hash = inet6_addr_hash(addr);
879 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
880 spin_unlock(&addrconf_hash_lock);
882 write_lock(&idev->lock);
883 /* Add to inet6_dev unicast addr list. */
884 ipv6_link_dev_addr(idev, ifa);
886 if (ifa->flags&IFA_F_TEMPORARY) {
887 list_add(&ifa->tmp_list, &idev->tempaddr_list);
892 write_unlock(&idev->lock);
894 rcu_read_unlock_bh();
896 if (likely(err == 0))
897 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
905 spin_unlock(&addrconf_hash_lock);
909 enum cleanup_prefix_rt_t {
910 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
911 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
912 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
916 * Check, whether the prefix for ifp would still need a prefix route
917 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
920 * 1) we don't purge prefix if address was not permanent.
921 * prefix is managed by its own lifetime.
922 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
923 * 3) if there are no addresses, delete prefix.
924 * 4) if there are still other permanent address(es),
925 * corresponding prefix is still permanent.
926 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
927 * don't purge the prefix, assume user space is managing it.
928 * 6) otherwise, update prefix lifetime to the
929 * longest valid lifetime among the corresponding
930 * addresses on the device.
931 * Note: subsequent RA will update lifetime.
933 static enum cleanup_prefix_rt_t
934 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
936 struct inet6_ifaddr *ifa;
937 struct inet6_dev *idev = ifp->idev;
938 unsigned long lifetime;
939 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
943 list_for_each_entry(ifa, &idev->addr_list, if_list) {
946 if (ifa->prefix_len != ifp->prefix_len ||
947 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
950 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
951 return CLEANUP_PREFIX_RT_NOP;
953 action = CLEANUP_PREFIX_RT_EXPIRE;
955 spin_lock(&ifa->lock);
957 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
959 * Note: Because this address is
960 * not permanent, lifetime <
961 * LONG_MAX / HZ here.
963 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
964 *expires = ifa->tstamp + lifetime * HZ;
965 spin_unlock(&ifa->lock);
972 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
976 rt = addrconf_get_prefix_route(&ifp->addr,
979 0, RTF_GATEWAY | RTF_DEFAULT);
984 if (!(rt->rt6i_flags & RTF_EXPIRES))
985 rt6_set_expires(rt, expires);
992 /* This function wants to get referenced ifp and releases it before return */
994 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
997 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
998 unsigned long expires;
1002 spin_lock_bh(&ifp->state_lock);
1004 ifp->state = INET6_IFADDR_STATE_DEAD;
1005 spin_unlock_bh(&ifp->state_lock);
1007 if (state == INET6_IFADDR_STATE_DEAD)
1010 spin_lock_bh(&addrconf_hash_lock);
1011 hlist_del_init_rcu(&ifp->addr_lst);
1012 spin_unlock_bh(&addrconf_hash_lock);
1014 write_lock_bh(&ifp->idev->lock);
1016 if (ifp->flags&IFA_F_TEMPORARY) {
1017 list_del(&ifp->tmp_list);
1019 in6_ifa_put(ifp->ifpub);
1025 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1026 action = check_cleanup_prefix_route(ifp, &expires);
1028 list_del_init(&ifp->if_list);
1031 write_unlock_bh(&ifp->idev->lock);
1033 addrconf_del_dad_work(ifp);
1035 ipv6_ifa_notify(RTM_DELADDR, ifp);
1037 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1039 if (action != CLEANUP_PREFIX_RT_NOP) {
1040 cleanup_prefix_route(ifp, expires,
1041 action == CLEANUP_PREFIX_RT_DEL);
1044 /* clean up prefsrc entries */
1045 rt6_remove_prefsrc(ifp);
1050 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1052 struct inet6_dev *idev = ifp->idev;
1053 struct in6_addr addr, *tmpaddr;
1054 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1055 unsigned long regen_advance;
1059 unsigned long now = jiffies;
1061 write_lock_bh(&idev->lock);
1063 spin_lock_bh(&ift->lock);
1064 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1065 spin_unlock_bh(&ift->lock);
1072 if (idev->cnf.use_tempaddr <= 0) {
1073 write_unlock_bh(&idev->lock);
1074 pr_info("%s: use_tempaddr is disabled\n", __func__);
1079 spin_lock_bh(&ifp->lock);
1080 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1081 idev->cnf.use_tempaddr = -1; /*XXX*/
1082 spin_unlock_bh(&ifp->lock);
1083 write_unlock_bh(&idev->lock);
1084 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1091 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1092 __ipv6_try_regen_rndid(idev, tmpaddr);
1093 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1094 age = (now - ifp->tstamp) / HZ;
1095 tmp_valid_lft = min_t(__u32,
1097 idev->cnf.temp_valid_lft + age);
1098 tmp_prefered_lft = min_t(__u32,
1100 idev->cnf.temp_prefered_lft + age -
1101 idev->cnf.max_desync_factor);
1102 tmp_plen = ifp->prefix_len;
1103 tmp_tstamp = ifp->tstamp;
1104 spin_unlock_bh(&ifp->lock);
1106 regen_advance = idev->cnf.regen_max_retry *
1107 idev->cnf.dad_transmits *
1108 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1109 write_unlock_bh(&idev->lock);
1111 /* A temporary address is created only if this calculated Preferred
1112 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1113 * an implementation must not create a temporary address with a zero
1114 * Preferred Lifetime.
1115 * Use age calculation as in addrconf_verify to avoid unnecessary
1116 * temporary addresses being generated.
1118 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1119 if (tmp_prefered_lft <= regen_advance + age) {
1126 addr_flags = IFA_F_TEMPORARY;
1127 /* set in addrconf_prefix_rcv() */
1128 if (ifp->flags & IFA_F_OPTIMISTIC)
1129 addr_flags |= IFA_F_OPTIMISTIC;
1131 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1132 ipv6_addr_scope(&addr), addr_flags,
1133 tmp_valid_lft, tmp_prefered_lft);
1137 pr_info("%s: retry temporary address regeneration\n", __func__);
1139 write_lock_bh(&idev->lock);
1143 spin_lock_bh(&ift->lock);
1146 ift->tstamp = tmp_tstamp;
1147 spin_unlock_bh(&ift->lock);
1149 addrconf_dad_start(ift);
1157 * Choose an appropriate source address (RFC3484)
1160 IPV6_SADDR_RULE_INIT = 0,
1161 IPV6_SADDR_RULE_LOCAL,
1162 IPV6_SADDR_RULE_SCOPE,
1163 IPV6_SADDR_RULE_PREFERRED,
1164 #ifdef CONFIG_IPV6_MIP6
1165 IPV6_SADDR_RULE_HOA,
1167 IPV6_SADDR_RULE_OIF,
1168 IPV6_SADDR_RULE_LABEL,
1169 IPV6_SADDR_RULE_PRIVACY,
1170 IPV6_SADDR_RULE_ORCHID,
1171 IPV6_SADDR_RULE_PREFIX,
1175 struct ipv6_saddr_score {
1178 struct inet6_ifaddr *ifa;
1179 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1184 struct ipv6_saddr_dst {
1185 const struct in6_addr *addr;
1192 static inline int ipv6_saddr_preferred(int type)
1194 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1199 static int ipv6_get_saddr_eval(struct net *net,
1200 struct ipv6_saddr_score *score,
1201 struct ipv6_saddr_dst *dst,
1206 if (i <= score->rule) {
1208 case IPV6_SADDR_RULE_SCOPE:
1209 ret = score->scopedist;
1211 case IPV6_SADDR_RULE_PREFIX:
1212 ret = score->matchlen;
1215 ret = !!test_bit(i, score->scorebits);
1221 case IPV6_SADDR_RULE_INIT:
1222 /* Rule 0: remember if hiscore is not ready yet */
1225 case IPV6_SADDR_RULE_LOCAL:
1226 /* Rule 1: Prefer same address */
1227 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1229 case IPV6_SADDR_RULE_SCOPE:
1230 /* Rule 2: Prefer appropriate scope
1235 * ---+--+-+---> scope
1237 * | d is scope of the destination.
1239 * | \ <- smaller scope is better if
1240 * B-15 | \ if scope is enough for destination.
1241 * | ret = B - scope (-1 <= scope >= d <= 15).
1243 * |/ <- greater is better
1244 * -C / if scope is not enough for destination.
1245 * /| ret = scope - C (-1 <= d < scope <= 15).
1247 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1248 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1249 * Assume B = 0 and we get C > 29.
1251 ret = __ipv6_addr_src_scope(score->addr_type);
1252 if (ret >= dst->scope)
1255 ret -= 128; /* 30 is enough */
1256 score->scopedist = ret;
1258 case IPV6_SADDR_RULE_PREFERRED:
1259 /* Rule 3: Avoid deprecated and optimistic addresses */
1260 ret = ipv6_saddr_preferred(score->addr_type) ||
1261 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1263 #ifdef CONFIG_IPV6_MIP6
1264 case IPV6_SADDR_RULE_HOA:
1266 /* Rule 4: Prefer home address */
1267 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1268 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1272 case IPV6_SADDR_RULE_OIF:
1273 /* Rule 5: Prefer outgoing interface */
1274 ret = (!dst->ifindex ||
1275 dst->ifindex == score->ifa->idev->dev->ifindex);
1277 case IPV6_SADDR_RULE_LABEL:
1278 /* Rule 6: Prefer matching label */
1279 ret = ipv6_addr_label(net,
1280 &score->ifa->addr, score->addr_type,
1281 score->ifa->idev->dev->ifindex) == dst->label;
1283 case IPV6_SADDR_RULE_PRIVACY:
1285 /* Rule 7: Prefer public address
1286 * Note: prefer temporary address if use_tempaddr >= 2
1288 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1289 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1290 score->ifa->idev->cnf.use_tempaddr >= 2;
1291 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1294 case IPV6_SADDR_RULE_ORCHID:
1295 /* Rule 8-: Prefer ORCHID vs ORCHID or
1296 * non-ORCHID vs non-ORCHID
1298 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1299 ipv6_addr_orchid(dst->addr));
1301 case IPV6_SADDR_RULE_PREFIX:
1302 /* Rule 8: Use longest matching prefix */
1303 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1304 if (ret > score->ifa->prefix_len)
1305 ret = score->ifa->prefix_len;
1306 score->matchlen = ret;
1313 __set_bit(i, score->scorebits);
1319 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1320 const struct in6_addr *daddr, unsigned int prefs,
1321 struct in6_addr *saddr)
1323 struct ipv6_saddr_score scores[2],
1324 *score = &scores[0], *hiscore = &scores[1];
1325 struct ipv6_saddr_dst dst;
1326 struct net_device *dev;
1329 dst_type = __ipv6_addr_type(daddr);
1331 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1332 dst.scope = __ipv6_addr_src_scope(dst_type);
1333 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1337 hiscore->ifa = NULL;
1341 for_each_netdev_rcu(net, dev) {
1342 struct inet6_dev *idev;
1344 /* Candidate Source Address (section 4)
1345 * - multicast and link-local destination address,
1346 * the set of candidate source address MUST only
1347 * include addresses assigned to interfaces
1348 * belonging to the same link as the outgoing
1350 * (- For site-local destination addresses, the
1351 * set of candidate source addresses MUST only
1352 * include addresses assigned to interfaces
1353 * belonging to the same site as the outgoing
1356 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1357 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1358 dst.ifindex && dev->ifindex != dst.ifindex)
1361 idev = __in6_dev_get(dev);
1365 read_lock_bh(&idev->lock);
1366 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1370 * - Tentative Address (RFC2462 section 5.4)
1371 * - A tentative address is not considered
1372 * "assigned to an interface" in the traditional
1373 * sense, unless it is also flagged as optimistic.
1374 * - Candidate Source Address (section 4)
1375 * - In any case, anycast addresses, multicast
1376 * addresses, and the unspecified address MUST
1377 * NOT be included in a candidate set.
1379 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1380 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1383 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1385 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1386 score->addr_type & IPV6_ADDR_MULTICAST)) {
1387 LIMIT_NETDEBUG(KERN_DEBUG
1388 "ADDRCONF: unspecified / multicast address "
1389 "assigned as unicast address on %s",
1395 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1397 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1398 int minihiscore, miniscore;
1400 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1401 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1403 if (minihiscore > miniscore) {
1404 if (i == IPV6_SADDR_RULE_SCOPE &&
1405 score->scopedist > 0) {
1408 * each remaining entry
1409 * has too small (not enough)
1410 * scope, because ifa entries
1411 * are sorted by their scope
1417 } else if (minihiscore < miniscore) {
1419 in6_ifa_put(hiscore->ifa);
1421 in6_ifa_hold(score->ifa);
1423 swap(hiscore, score);
1425 /* restore our iterator */
1426 score->ifa = hiscore->ifa;
1433 read_unlock_bh(&idev->lock);
1438 return -EADDRNOTAVAIL;
1440 *saddr = hiscore->ifa->addr;
1441 in6_ifa_put(hiscore->ifa);
1444 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1446 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1449 struct inet6_ifaddr *ifp;
1450 int err = -EADDRNOTAVAIL;
1452 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1453 if (ifp->scope > IFA_LINK)
1455 if (ifp->scope == IFA_LINK &&
1456 !(ifp->flags & banned_flags)) {
1465 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1468 struct inet6_dev *idev;
1469 int err = -EADDRNOTAVAIL;
1472 idev = __in6_dev_get(dev);
1474 read_lock_bh(&idev->lock);
1475 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1476 read_unlock_bh(&idev->lock);
1482 static int ipv6_count_addresses(struct inet6_dev *idev)
1485 struct inet6_ifaddr *ifp;
1487 read_lock_bh(&idev->lock);
1488 list_for_each_entry(ifp, &idev->addr_list, if_list)
1490 read_unlock_bh(&idev->lock);
1494 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1495 const struct net_device *dev, int strict)
1497 struct inet6_ifaddr *ifp;
1498 unsigned int hash = inet6_addr_hash(addr);
1501 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1502 if (!net_eq(dev_net(ifp->idev->dev), net))
1504 if (ipv6_addr_equal(&ifp->addr, addr) &&
1505 !(ifp->flags&IFA_F_TENTATIVE) &&
1506 (dev == NULL || ifp->idev->dev == dev ||
1507 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1508 rcu_read_unlock_bh();
1513 rcu_read_unlock_bh();
1516 EXPORT_SYMBOL(ipv6_chk_addr);
1518 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1519 struct net_device *dev)
1521 unsigned int hash = inet6_addr_hash(addr);
1522 struct inet6_ifaddr *ifp;
1524 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1525 if (!net_eq(dev_net(ifp->idev->dev), net))
1527 if (ipv6_addr_equal(&ifp->addr, addr)) {
1528 if (dev == NULL || ifp->idev->dev == dev)
1535 /* Compares an address/prefix_len with addresses on device @dev.
1536 * If one is found it returns true.
1538 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1539 const unsigned int prefix_len, struct net_device *dev)
1541 struct inet6_dev *idev;
1542 struct inet6_ifaddr *ifa;
1546 idev = __in6_dev_get(dev);
1548 read_lock_bh(&idev->lock);
1549 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1550 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1554 read_unlock_bh(&idev->lock);
1560 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1562 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1564 struct inet6_dev *idev;
1565 struct inet6_ifaddr *ifa;
1570 idev = __in6_dev_get(dev);
1572 read_lock_bh(&idev->lock);
1573 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1574 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1579 read_unlock_bh(&idev->lock);
1584 EXPORT_SYMBOL(ipv6_chk_prefix);
1586 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1587 struct net_device *dev, int strict)
1589 struct inet6_ifaddr *ifp, *result = NULL;
1590 unsigned int hash = inet6_addr_hash(addr);
1593 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1594 if (!net_eq(dev_net(ifp->idev->dev), net))
1596 if (ipv6_addr_equal(&ifp->addr, addr)) {
1597 if (dev == NULL || ifp->idev->dev == dev ||
1598 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1605 rcu_read_unlock_bh();
1610 /* Gets referenced address, destroys ifaddr */
1612 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1615 ifp->flags |= IFA_F_DADFAILED;
1617 if (ifp->flags&IFA_F_TEMPORARY) {
1618 struct inet6_ifaddr *ifpub;
1619 spin_lock_bh(&ifp->lock);
1622 in6_ifa_hold(ifpub);
1623 spin_unlock_bh(&ifp->lock);
1624 ipv6_create_tempaddr(ifpub, ifp);
1627 spin_unlock_bh(&ifp->lock);
1630 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1631 spin_lock_bh(&ifp->lock);
1632 addrconf_del_dad_work(ifp);
1633 ifp->flags |= IFA_F_TENTATIVE;
1634 spin_unlock_bh(&ifp->lock);
1636 ipv6_ifa_notify(0, ifp);
1643 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1647 spin_lock_bh(&ifp->state_lock);
1648 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1649 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1652 spin_unlock_bh(&ifp->state_lock);
1657 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1659 struct inet6_dev *idev = ifp->idev;
1661 if (addrconf_dad_end(ifp)) {
1666 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1667 ifp->idev->dev->name, &ifp->addr);
1669 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1670 struct in6_addr addr;
1672 addr.s6_addr32[0] = htonl(0xfe800000);
1673 addr.s6_addr32[1] = 0;
1675 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1676 ipv6_addr_equal(&ifp->addr, &addr)) {
1677 /* DAD failed for link-local based on MAC address */
1678 idev->cnf.disable_ipv6 = 1;
1680 pr_info("%s: IPv6 being disabled!\n",
1681 ifp->idev->dev->name);
1685 spin_lock_bh(&ifp->state_lock);
1686 /* transition from _POSTDAD to _ERRDAD */
1687 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1688 spin_unlock_bh(&ifp->state_lock);
1690 addrconf_mod_dad_work(ifp, 0);
1694 /* Join to solicited addr multicast group.
1695 * caller must hold RTNL */
1696 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1698 struct in6_addr maddr;
1700 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1703 addrconf_addr_solict_mult(addr, &maddr);
1704 ipv6_dev_mc_inc(dev, &maddr);
1707 /* caller must hold RTNL */
1708 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1710 struct in6_addr maddr;
1712 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1715 addrconf_addr_solict_mult(addr, &maddr);
1716 __ipv6_dev_mc_dec(idev, &maddr);
1719 /* caller must hold RTNL */
1720 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1722 struct in6_addr addr;
1724 if (ifp->prefix_len >= 127) /* RFC 6164 */
1726 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1727 if (ipv6_addr_any(&addr))
1729 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1732 /* caller must hold RTNL */
1733 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1735 struct in6_addr addr;
1737 if (ifp->prefix_len >= 127) /* RFC 6164 */
1739 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1740 if (ipv6_addr_any(&addr))
1742 __ipv6_dev_ac_dec(ifp->idev, &addr);
1745 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1747 if (dev->addr_len != ETH_ALEN)
1749 memcpy(eui, dev->dev_addr, 3);
1750 memcpy(eui + 5, dev->dev_addr + 3, 3);
1753 * The zSeries OSA network cards can be shared among various
1754 * OS instances, but the OSA cards have only one MAC address.
1755 * This leads to duplicate address conflicts in conjunction
1756 * with IPv6 if more than one instance uses the same card.
1758 * The driver for these cards can deliver a unique 16-bit
1759 * identifier for each instance sharing the same card. It is
1760 * placed instead of 0xFFFE in the interface identifier. The
1761 * "u" bit of the interface identifier is not inverted in this
1762 * case. Hence the resulting interface identifier has local
1763 * scope according to RFC2373.
1766 eui[3] = (dev->dev_id >> 8) & 0xFF;
1767 eui[4] = dev->dev_id & 0xFF;
1776 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1778 if (dev->addr_len != IEEE802154_ADDR_LEN)
1780 memcpy(eui, dev->dev_addr, 8);
1785 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1787 union fwnet_hwaddr *ha;
1789 if (dev->addr_len != FWNET_ALEN)
1792 ha = (union fwnet_hwaddr *)dev->dev_addr;
1794 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1799 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1801 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1802 if (dev->addr_len != ARCNET_ALEN)
1805 eui[7] = *(u8 *)dev->dev_addr;
1809 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1811 if (dev->addr_len != INFINIBAND_ALEN)
1813 memcpy(eui, dev->dev_addr + 12, 8);
1818 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1822 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1823 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1824 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1825 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1826 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1827 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1831 memcpy(eui + 4, &addr, 4);
1835 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1837 if (dev->priv_flags & IFF_ISATAP)
1838 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1842 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1844 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1847 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1849 memcpy(eui, dev->perm_addr, 3);
1850 memcpy(eui + 5, dev->perm_addr + 3, 3);
1857 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1859 switch (dev->type) {
1862 return addrconf_ifid_eui48(eui, dev);
1864 return addrconf_ifid_arcnet(eui, dev);
1865 case ARPHRD_INFINIBAND:
1866 return addrconf_ifid_infiniband(eui, dev);
1868 return addrconf_ifid_sit(eui, dev);
1870 return addrconf_ifid_gre(eui, dev);
1871 case ARPHRD_6LOWPAN:
1872 case ARPHRD_IEEE802154:
1873 return addrconf_ifid_eui64(eui, dev);
1874 case ARPHRD_IEEE1394:
1875 return addrconf_ifid_ieee1394(eui, dev);
1876 case ARPHRD_TUNNEL6:
1877 return addrconf_ifid_ip6tnl(eui, dev);
1882 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1885 struct inet6_ifaddr *ifp;
1887 read_lock_bh(&idev->lock);
1888 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1889 if (ifp->scope > IFA_LINK)
1891 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1892 memcpy(eui, ifp->addr.s6_addr+8, 8);
1897 read_unlock_bh(&idev->lock);
1901 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1902 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1905 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1906 idev->rndid[0] &= ~0x02;
1909 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1910 * check if generated address is not inappropriate
1912 * - Reserved subnet anycast (RFC 2526)
1913 * 11111101 11....11 1xxxxxxx
1914 * - ISATAP (RFC4214) 6.1
1915 * 00-00-5E-FE-xx-xx-xx-xx
1917 * - XXX: already assigned to an address on the device
1919 if (idev->rndid[0] == 0xfd &&
1920 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1921 (idev->rndid[7]&0x80))
1923 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1924 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1926 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1931 static void ipv6_regen_rndid(unsigned long data)
1933 struct inet6_dev *idev = (struct inet6_dev *) data;
1934 unsigned long expires;
1937 write_lock_bh(&idev->lock);
1942 __ipv6_regen_rndid(idev);
1945 idev->cnf.temp_prefered_lft * HZ -
1946 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1947 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1948 idev->cnf.max_desync_factor * HZ;
1949 if (time_before(expires, jiffies)) {
1950 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1951 __func__, idev->dev->name);
1955 if (!mod_timer(&idev->regen_timer, expires))
1959 write_unlock_bh(&idev->lock);
1960 rcu_read_unlock_bh();
1964 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1966 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1967 __ipv6_regen_rndid(idev);
1975 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1976 unsigned long expires, u32 flags)
1978 struct fib6_config cfg = {
1979 .fc_table = RT6_TABLE_PREFIX,
1980 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1981 .fc_ifindex = dev->ifindex,
1982 .fc_expires = expires,
1984 .fc_flags = RTF_UP | flags,
1985 .fc_nlinfo.nl_net = dev_net(dev),
1986 .fc_protocol = RTPROT_KERNEL,
1991 /* Prevent useless cloning on PtP SIT.
1992 This thing is done here expecting that the whole
1993 class of non-broadcast devices need not cloning.
1995 #if IS_ENABLED(CONFIG_IPV6_SIT)
1996 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1997 cfg.fc_flags |= RTF_NONEXTHOP;
2000 ip6_route_add(&cfg);
2004 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2006 const struct net_device *dev,
2007 u32 flags, u32 noflags)
2009 struct fib6_node *fn;
2010 struct rt6_info *rt = NULL;
2011 struct fib6_table *table;
2013 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2017 read_lock_bh(&table->tb6_lock);
2018 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2021 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2022 if (rt->dst.dev->ifindex != dev->ifindex)
2024 if ((rt->rt6i_flags & flags) != flags)
2026 if ((rt->rt6i_flags & noflags) != 0)
2032 read_unlock_bh(&table->tb6_lock);
2037 /* Create "default" multicast route to the interface */
2039 static void addrconf_add_mroute(struct net_device *dev)
2041 struct fib6_config cfg = {
2042 .fc_table = RT6_TABLE_LOCAL,
2043 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2044 .fc_ifindex = dev->ifindex,
2047 .fc_nlinfo.nl_net = dev_net(dev),
2050 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2052 ip6_route_add(&cfg);
2055 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2057 struct inet6_dev *idev;
2061 idev = ipv6_find_idev(dev);
2063 return ERR_PTR(-ENOBUFS);
2065 if (idev->cnf.disable_ipv6)
2066 return ERR_PTR(-EACCES);
2068 /* Add default multicast route */
2069 if (!(dev->flags & IFF_LOOPBACK))
2070 addrconf_add_mroute(dev);
2075 static void manage_tempaddrs(struct inet6_dev *idev,
2076 struct inet6_ifaddr *ifp,
2077 __u32 valid_lft, __u32 prefered_lft,
2078 bool create, unsigned long now)
2081 struct inet6_ifaddr *ift;
2083 read_lock_bh(&idev->lock);
2084 /* update all temporary addresses in the list */
2085 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2086 int age, max_valid, max_prefered;
2088 if (ifp != ift->ifpub)
2091 /* RFC 4941 section 3.3:
2092 * If a received option will extend the lifetime of a public
2093 * address, the lifetimes of temporary addresses should
2094 * be extended, subject to the overall constraint that no
2095 * temporary addresses should ever remain "valid" or "preferred"
2096 * for a time longer than (TEMP_VALID_LIFETIME) or
2097 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2099 age = (now - ift->cstamp) / HZ;
2100 max_valid = idev->cnf.temp_valid_lft - age;
2104 max_prefered = idev->cnf.temp_prefered_lft -
2105 idev->cnf.max_desync_factor - age;
2106 if (max_prefered < 0)
2109 if (valid_lft > max_valid)
2110 valid_lft = max_valid;
2112 if (prefered_lft > max_prefered)
2113 prefered_lft = max_prefered;
2115 spin_lock(&ift->lock);
2117 ift->valid_lft = valid_lft;
2118 ift->prefered_lft = prefered_lft;
2120 if (prefered_lft > 0)
2121 ift->flags &= ~IFA_F_DEPRECATED;
2123 spin_unlock(&ift->lock);
2124 if (!(flags&IFA_F_TENTATIVE))
2125 ipv6_ifa_notify(0, ift);
2128 if ((create || list_empty(&idev->tempaddr_list)) &&
2129 idev->cnf.use_tempaddr > 0) {
2130 /* When a new public address is created as described
2131 * in [ADDRCONF], also create a new temporary address.
2132 * Also create a temporary address if it's enabled but
2133 * no temporary address currently exists.
2135 read_unlock_bh(&idev->lock);
2136 ipv6_create_tempaddr(ifp, NULL);
2138 read_unlock_bh(&idev->lock);
2142 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2144 struct prefix_info *pinfo;
2148 struct inet6_dev *in6_dev;
2149 struct net *net = dev_net(dev);
2151 pinfo = (struct prefix_info *) opt;
2153 if (len < sizeof(struct prefix_info)) {
2154 ADBG("addrconf: prefix option too short\n");
2159 * Validation checks ([ADDRCONF], page 19)
2162 addr_type = ipv6_addr_type(&pinfo->prefix);
2164 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2167 valid_lft = ntohl(pinfo->valid);
2168 prefered_lft = ntohl(pinfo->prefered);
2170 if (prefered_lft > valid_lft) {
2171 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2175 in6_dev = in6_dev_get(dev);
2177 if (in6_dev == NULL) {
2178 net_dbg_ratelimited("addrconf: device %s not configured\n",
2184 * Two things going on here:
2185 * 1) Add routes for on-link prefixes
2186 * 2) Configure prefixes with the auto flag set
2189 if (pinfo->onlink) {
2190 struct rt6_info *rt;
2191 unsigned long rt_expires;
2193 /* Avoid arithmetic overflow. Really, we could
2194 * save rt_expires in seconds, likely valid_lft,
2195 * but it would require division in fib gc, that it
2199 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2201 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2203 if (addrconf_finite_timeout(rt_expires))
2206 rt = addrconf_get_prefix_route(&pinfo->prefix,
2209 RTF_ADDRCONF | RTF_PREFIX_RT,
2210 RTF_GATEWAY | RTF_DEFAULT);
2213 /* Autoconf prefix route */
2214 if (valid_lft == 0) {
2217 } else if (addrconf_finite_timeout(rt_expires)) {
2219 rt6_set_expires(rt, jiffies + rt_expires);
2221 rt6_clean_expires(rt);
2223 } else if (valid_lft) {
2224 clock_t expires = 0;
2225 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2226 if (addrconf_finite_timeout(rt_expires)) {
2228 flags |= RTF_EXPIRES;
2229 expires = jiffies_to_clock_t(rt_expires);
2231 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2232 dev, expires, flags);
2237 /* Try to figure out our local address for this prefix */
2239 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2240 struct inet6_ifaddr *ifp;
2241 struct in6_addr addr;
2242 int create = 0, update_lft = 0;
2243 bool tokenized = false;
2245 if (pinfo->prefix_len == 64) {
2246 memcpy(&addr, &pinfo->prefix, 8);
2248 if (!ipv6_addr_any(&in6_dev->token)) {
2249 read_lock_bh(&in6_dev->lock);
2250 memcpy(addr.s6_addr + 8,
2251 in6_dev->token.s6_addr + 8, 8);
2252 read_unlock_bh(&in6_dev->lock);
2254 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2255 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2256 in6_dev_put(in6_dev);
2261 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2263 in6_dev_put(in6_dev);
2268 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2270 if (ifp == NULL && valid_lft) {
2271 int max_addresses = in6_dev->cnf.max_addresses;
2274 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2275 if (in6_dev->cnf.optimistic_dad &&
2276 !net->ipv6.devconf_all->forwarding && sllao)
2277 addr_flags = IFA_F_OPTIMISTIC;
2280 /* Do not allow to create too much of autoconfigured
2281 * addresses; this would be too easy way to crash kernel.
2283 if (!max_addresses ||
2284 ipv6_count_addresses(in6_dev) < max_addresses)
2285 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2287 addr_type&IPV6_ADDR_SCOPE_MASK,
2288 addr_flags, valid_lft,
2291 if (IS_ERR_OR_NULL(ifp)) {
2292 in6_dev_put(in6_dev);
2298 spin_lock_bh(&ifp->lock);
2299 ifp->flags |= IFA_F_MANAGETEMPADDR;
2300 ifp->cstamp = jiffies;
2301 ifp->tokenized = tokenized;
2302 spin_unlock_bh(&ifp->lock);
2303 addrconf_dad_start(ifp);
2311 /* update lifetime (RFC2462 5.5.3 e) */
2312 spin_lock(&ifp->lock);
2314 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2315 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2318 if (!update_lft && !create && stored_lft) {
2319 const u32 minimum_lft = min(
2320 stored_lft, (u32)MIN_VALID_LIFETIME);
2321 valid_lft = max(valid_lft, minimum_lft);
2323 /* RFC4862 Section 5.5.3e:
2324 * "Note that the preferred lifetime of the
2325 * corresponding address is always reset to
2326 * the Preferred Lifetime in the received
2327 * Prefix Information option, regardless of
2328 * whether the valid lifetime is also reset or
2331 * So we should always update prefered_lft here.
2337 ifp->valid_lft = valid_lft;
2338 ifp->prefered_lft = prefered_lft;
2341 ifp->flags &= ~IFA_F_DEPRECATED;
2342 spin_unlock(&ifp->lock);
2344 if (!(flags&IFA_F_TENTATIVE))
2345 ipv6_ifa_notify(0, ifp);
2347 spin_unlock(&ifp->lock);
2349 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2356 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2357 in6_dev_put(in6_dev);
2361 * Set destination address.
2362 * Special case for SIT interfaces where we create a new "virtual"
2365 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2367 struct in6_ifreq ireq;
2368 struct net_device *dev;
2374 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2377 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2383 #if IS_ENABLED(CONFIG_IPV6_SIT)
2384 if (dev->type == ARPHRD_SIT) {
2385 const struct net_device_ops *ops = dev->netdev_ops;
2387 struct ip_tunnel_parm p;
2389 err = -EADDRNOTAVAIL;
2390 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2393 memset(&p, 0, sizeof(p));
2394 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2398 p.iph.protocol = IPPROTO_IPV6;
2400 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2402 if (ops->ndo_do_ioctl) {
2403 mm_segment_t oldfs = get_fs();
2406 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2413 dev = __dev_get_by_name(net, p.name);
2416 err = dev_open(dev);
2427 * Manual configuration of address on an interface
2429 static int inet6_addr_add(struct net *net, int ifindex,
2430 const struct in6_addr *pfx,
2431 const struct in6_addr *peer_pfx,
2432 unsigned int plen, __u32 ifa_flags,
2433 __u32 prefered_lft, __u32 valid_lft)
2435 struct inet6_ifaddr *ifp;
2436 struct inet6_dev *idev;
2437 struct net_device *dev;
2441 unsigned long timeout;
2448 /* check the lifetime */
2449 if (!valid_lft || prefered_lft > valid_lft)
2452 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2455 dev = __dev_get_by_index(net, ifindex);
2459 idev = addrconf_add_dev(dev);
2461 return PTR_ERR(idev);
2463 scope = ipv6_addr_scope(pfx);
2465 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2466 if (addrconf_finite_timeout(timeout)) {
2467 expires = jiffies_to_clock_t(timeout * HZ);
2468 valid_lft = timeout;
2469 flags = RTF_EXPIRES;
2473 ifa_flags |= IFA_F_PERMANENT;
2476 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2477 if (addrconf_finite_timeout(timeout)) {
2479 ifa_flags |= IFA_F_DEPRECATED;
2480 prefered_lft = timeout;
2483 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2484 valid_lft, prefered_lft);
2487 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2488 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2493 * Note that section 3.1 of RFC 4429 indicates
2494 * that the Optimistic flag should not be set for
2495 * manually configured addresses
2497 addrconf_dad_start(ifp);
2498 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2499 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2502 addrconf_verify_rtnl();
2506 return PTR_ERR(ifp);
2509 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2510 const struct in6_addr *pfx, unsigned int plen)
2512 struct inet6_ifaddr *ifp;
2513 struct inet6_dev *idev;
2514 struct net_device *dev;
2519 dev = __dev_get_by_index(net, ifindex);
2523 if ((idev = __in6_dev_get(dev)) == NULL)
2526 read_lock_bh(&idev->lock);
2527 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2528 if (ifp->prefix_len == plen &&
2529 ipv6_addr_equal(pfx, &ifp->addr)) {
2531 read_unlock_bh(&idev->lock);
2533 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2534 (ifa_flags & IFA_F_MANAGETEMPADDR))
2535 manage_tempaddrs(idev, ifp, 0, 0, false,
2538 addrconf_verify_rtnl();
2542 read_unlock_bh(&idev->lock);
2543 return -EADDRNOTAVAIL;
2547 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2549 struct in6_ifreq ireq;
2552 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2555 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2559 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2560 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2561 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2566 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2568 struct in6_ifreq ireq;
2571 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2574 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2578 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2579 ireq.ifr6_prefixlen);
2584 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2585 int plen, int scope)
2587 struct inet6_ifaddr *ifp;
2589 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2590 scope, IFA_F_PERMANENT,
2591 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2593 spin_lock_bh(&ifp->lock);
2594 ifp->flags &= ~IFA_F_TENTATIVE;
2595 spin_unlock_bh(&ifp->lock);
2596 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2601 #if IS_ENABLED(CONFIG_IPV6_SIT)
2602 static void sit_add_v4_addrs(struct inet6_dev *idev)
2604 struct in6_addr addr;
2605 struct net_device *dev;
2606 struct net *net = dev_net(idev->dev);
2612 memset(&addr, 0, sizeof(struct in6_addr));
2613 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2615 if (idev->dev->flags&IFF_POINTOPOINT) {
2616 addr.s6_addr32[0] = htonl(0xfe800000);
2620 scope = IPV6_ADDR_COMPATv4;
2622 pflags |= RTF_NONEXTHOP;
2625 if (addr.s6_addr32[3]) {
2626 add_addr(idev, &addr, plen, scope);
2627 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2631 for_each_netdev(net, dev) {
2632 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2633 if (in_dev && (dev->flags & IFF_UP)) {
2634 struct in_ifaddr *ifa;
2638 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2640 addr.s6_addr32[3] = ifa->ifa_local;
2642 if (ifa->ifa_scope == RT_SCOPE_LINK)
2644 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2645 if (idev->dev->flags&IFF_POINTOPOINT)
2650 add_addr(idev, &addr, plen, flag);
2651 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2659 static void init_loopback(struct net_device *dev)
2661 struct inet6_dev *idev;
2662 struct net_device *sp_dev;
2663 struct inet6_ifaddr *sp_ifa;
2664 struct rt6_info *sp_rt;
2670 if ((idev = ipv6_find_idev(dev)) == NULL) {
2671 pr_debug("%s: add_dev failed\n", __func__);
2675 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2677 /* Add routes to other interface's IPv6 addresses */
2678 for_each_netdev(dev_net(dev), sp_dev) {
2679 if (!strcmp(sp_dev->name, dev->name))
2682 idev = __in6_dev_get(sp_dev);
2686 read_lock_bh(&idev->lock);
2687 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2689 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2693 /* This dst has been added to garbage list when
2694 * lo device down, release this obsolete dst and
2695 * reallocate a new router for ifa.
2697 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
2698 ip6_rt_put(sp_ifa->rt);
2705 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2707 /* Failure cases are ignored */
2708 if (!IS_ERR(sp_rt)) {
2713 read_unlock_bh(&idev->lock);
2717 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2719 struct inet6_ifaddr *ifp;
2720 u32 addr_flags = IFA_F_PERMANENT;
2722 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2723 if (idev->cnf.optimistic_dad &&
2724 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2725 addr_flags |= IFA_F_OPTIMISTIC;
2729 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2730 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2732 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2733 addrconf_dad_start(ifp);
2738 static void addrconf_dev_config(struct net_device *dev)
2740 struct in6_addr addr;
2741 struct inet6_dev *idev;
2745 if ((dev->type != ARPHRD_ETHER) &&
2746 (dev->type != ARPHRD_FDDI) &&
2747 (dev->type != ARPHRD_ARCNET) &&
2748 (dev->type != ARPHRD_INFINIBAND) &&
2749 (dev->type != ARPHRD_IEEE802154) &&
2750 (dev->type != ARPHRD_IEEE1394) &&
2751 (dev->type != ARPHRD_TUNNEL6) &&
2752 (dev->type != ARPHRD_6LOWPAN)) {
2753 /* Alas, we support only Ethernet autoconfiguration. */
2757 idev = addrconf_add_dev(dev);
2761 memset(&addr, 0, sizeof(struct in6_addr));
2762 addr.s6_addr32[0] = htonl(0xFE800000);
2764 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2765 addrconf_add_linklocal(idev, &addr);
2768 #if IS_ENABLED(CONFIG_IPV6_SIT)
2769 static void addrconf_sit_config(struct net_device *dev)
2771 struct inet6_dev *idev;
2776 * Configure the tunnel with one of our IPv4
2777 * addresses... we should configure all of
2778 * our v4 addrs in the tunnel
2781 if ((idev = ipv6_find_idev(dev)) == NULL) {
2782 pr_debug("%s: add_dev failed\n", __func__);
2786 if (dev->priv_flags & IFF_ISATAP) {
2787 struct in6_addr addr;
2789 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2790 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2791 addrconf_add_linklocal(idev, &addr);
2795 sit_add_v4_addrs(idev);
2797 if (dev->flags&IFF_POINTOPOINT)
2798 addrconf_add_mroute(dev);
2802 #if IS_ENABLED(CONFIG_NET_IPGRE)
2803 static void addrconf_gre_config(struct net_device *dev)
2805 struct inet6_dev *idev;
2806 struct in6_addr addr;
2810 if ((idev = ipv6_find_idev(dev)) == NULL) {
2811 pr_debug("%s: add_dev failed\n", __func__);
2815 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2816 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2817 addrconf_add_linklocal(idev, &addr);
2819 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2823 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2826 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2827 struct inet6_dev *idev = __in6_dev_get(dev);
2828 struct net *net = dev_net(dev);
2829 int run_pending = 0;
2833 case NETDEV_REGISTER:
2834 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2835 idev = ipv6_add_dev(dev);
2837 return notifier_from_errno(-ENOMEM);
2843 if (dev->flags & IFF_SLAVE)
2846 if (event == NETDEV_UP) {
2847 if (!addrconf_qdisc_ok(dev)) {
2848 /* device is not ready yet. */
2849 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2854 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2855 idev = ipv6_add_dev(dev);
2858 idev->if_flags |= IF_READY;
2862 if (!addrconf_qdisc_ok(dev)) {
2863 /* device is still not ready. */
2868 if (idev->if_flags & IF_READY)
2869 /* device is already configured. */
2871 idev->if_flags |= IF_READY;
2874 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2880 switch (dev->type) {
2881 #if IS_ENABLED(CONFIG_IPV6_SIT)
2883 addrconf_sit_config(dev);
2886 #if IS_ENABLED(CONFIG_NET_IPGRE)
2888 addrconf_gre_config(dev);
2891 case ARPHRD_LOOPBACK:
2896 addrconf_dev_config(dev);
2902 addrconf_dad_run(idev);
2905 * If the MTU changed during the interface down,
2906 * when the interface up, the changed MTU must be
2907 * reflected in the idev as well as routers.
2909 if (idev->cnf.mtu6 != dev->mtu &&
2910 dev->mtu >= IPV6_MIN_MTU) {
2911 rt6_mtu_change(dev, dev->mtu);
2912 idev->cnf.mtu6 = dev->mtu;
2914 idev->tstamp = jiffies;
2915 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2918 * If the changed mtu during down is lower than
2919 * IPV6_MIN_MTU stop IPv6 on this interface.
2921 if (dev->mtu < IPV6_MIN_MTU)
2922 addrconf_ifdown(dev, dev != net->loopback_dev);
2926 case NETDEV_CHANGEMTU:
2927 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2928 rt6_mtu_change(dev, dev->mtu);
2929 idev->cnf.mtu6 = dev->mtu;
2933 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2934 idev = ipv6_add_dev(dev);
2940 * if MTU under IPV6_MIN_MTU.
2941 * Stop IPv6 on this interface.
2943 addrconf_ifdown(dev, dev != net->loopback_dev);
2947 case NETDEV_UNREGISTER:
2949 * Remove all addresses from this interface.
2951 addrconf_ifdown(dev, event != NETDEV_DOWN);
2954 case NETDEV_CHANGENAME:
2956 snmp6_unregister_dev(idev);
2957 addrconf_sysctl_unregister(idev);
2958 addrconf_sysctl_register(idev);
2959 err = snmp6_register_dev(idev);
2961 return notifier_from_errno(err);
2965 case NETDEV_PRE_TYPE_CHANGE:
2966 case NETDEV_POST_TYPE_CHANGE:
2967 addrconf_type_change(dev, event);
2975 * addrconf module should be notified of a device going up
2977 static struct notifier_block ipv6_dev_notf = {
2978 .notifier_call = addrconf_notify,
2979 .priority = ADDRCONF_NOTIFY_PRIORITY,
2982 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2984 struct inet6_dev *idev;
2987 idev = __in6_dev_get(dev);
2989 if (event == NETDEV_POST_TYPE_CHANGE)
2990 ipv6_mc_remap(idev);
2991 else if (event == NETDEV_PRE_TYPE_CHANGE)
2992 ipv6_mc_unmap(idev);
2995 static int addrconf_ifdown(struct net_device *dev, int how)
2997 struct net *net = dev_net(dev);
2998 struct inet6_dev *idev;
2999 struct inet6_ifaddr *ifa;
3004 rt6_ifdown(net, dev);
3005 neigh_ifdown(&nd_tbl, dev);
3007 idev = __in6_dev_get(dev);
3012 * Step 1: remove reference to ipv6 device from parent device.
3018 /* protected by rtnl_lock */
3019 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3021 /* Step 1.5: remove snmp6 entry */
3022 snmp6_unregister_dev(idev);
3026 /* Step 2: clear hash table */
3027 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3028 struct hlist_head *h = &inet6_addr_lst[i];
3030 spin_lock_bh(&addrconf_hash_lock);
3032 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3033 if (ifa->idev == idev) {
3034 hlist_del_init_rcu(&ifa->addr_lst);
3035 addrconf_del_dad_work(ifa);
3039 spin_unlock_bh(&addrconf_hash_lock);
3042 write_lock_bh(&idev->lock);
3044 addrconf_del_rs_timer(idev);
3046 /* Step 2: clear flags for stateless addrconf */
3048 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3050 if (how && del_timer(&idev->regen_timer))
3053 /* Step 3: clear tempaddr list */
3054 while (!list_empty(&idev->tempaddr_list)) {
3055 ifa = list_first_entry(&idev->tempaddr_list,
3056 struct inet6_ifaddr, tmp_list);
3057 list_del(&ifa->tmp_list);
3058 write_unlock_bh(&idev->lock);
3059 spin_lock_bh(&ifa->lock);
3062 in6_ifa_put(ifa->ifpub);
3065 spin_unlock_bh(&ifa->lock);
3067 write_lock_bh(&idev->lock);
3070 while (!list_empty(&idev->addr_list)) {
3071 ifa = list_first_entry(&idev->addr_list,
3072 struct inet6_ifaddr, if_list);
3073 addrconf_del_dad_work(ifa);
3075 list_del(&ifa->if_list);
3077 write_unlock_bh(&idev->lock);
3079 spin_lock_bh(&ifa->state_lock);
3081 ifa->state = INET6_IFADDR_STATE_DEAD;
3082 spin_unlock_bh(&ifa->state_lock);
3084 if (state != INET6_IFADDR_STATE_DEAD) {
3085 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3086 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3090 write_lock_bh(&idev->lock);
3093 write_unlock_bh(&idev->lock);
3095 /* Step 5: Discard multicast list */
3097 ipv6_mc_destroy_dev(idev);
3101 idev->tstamp = jiffies;
3103 /* Last: Shot the device (if unregistered) */
3105 addrconf_sysctl_unregister(idev);
3106 neigh_parms_release(&nd_tbl, idev->nd_parms);
3107 neigh_ifdown(&nd_tbl, dev);
3113 static void addrconf_rs_timer(unsigned long data)
3115 struct inet6_dev *idev = (struct inet6_dev *)data;
3116 struct net_device *dev = idev->dev;
3117 struct in6_addr lladdr;
3119 write_lock(&idev->lock);
3120 if (idev->dead || !(idev->if_flags & IF_READY))
3123 if (!ipv6_accept_ra(idev))
3126 /* Announcement received after solicitation was sent */
3127 if (idev->if_flags & IF_RA_RCVD)
3130 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3131 write_unlock(&idev->lock);
3132 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3133 ndisc_send_rs(dev, &lladdr,
3134 &in6addr_linklocal_allrouters);
3138 write_lock(&idev->lock);
3139 /* The wait after the last probe can be shorter */
3140 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3141 idev->cnf.rtr_solicits) ?
3142 idev->cnf.rtr_solicit_delay :
3143 idev->cnf.rtr_solicit_interval);
3146 * Note: we do not support deprecated "all on-link"
3147 * assumption any longer.
3149 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3153 write_unlock(&idev->lock);
3159 * Duplicate Address Detection
3161 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3163 unsigned long rand_num;
3164 struct inet6_dev *idev = ifp->idev;
3166 if (ifp->flags & IFA_F_OPTIMISTIC)
3169 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3171 ifp->dad_probes = idev->cnf.dad_transmits;
3172 addrconf_mod_dad_work(ifp, rand_num);
3175 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3177 struct inet6_dev *idev = ifp->idev;
3178 struct net_device *dev = idev->dev;
3180 addrconf_join_solict(dev, &ifp->addr);
3182 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3184 read_lock_bh(&idev->lock);
3185 spin_lock(&ifp->lock);
3186 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3189 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3190 idev->cnf.accept_dad < 1 ||
3191 !(ifp->flags&IFA_F_TENTATIVE) ||
3192 ifp->flags & IFA_F_NODAD) {
3193 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3194 spin_unlock(&ifp->lock);
3195 read_unlock_bh(&idev->lock);
3197 addrconf_dad_completed(ifp);
3201 if (!(idev->if_flags & IF_READY)) {
3202 spin_unlock(&ifp->lock);
3203 read_unlock_bh(&idev->lock);
3205 * If the device is not ready:
3206 * - keep it tentative if it is a permanent address.
3207 * - otherwise, kill it.
3210 addrconf_dad_stop(ifp, 0);
3215 * Optimistic nodes can start receiving
3218 if (ifp->flags & IFA_F_OPTIMISTIC)
3219 ip6_ins_rt(ifp->rt);
3221 addrconf_dad_kick(ifp);
3223 spin_unlock(&ifp->lock);
3224 read_unlock_bh(&idev->lock);
3227 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3229 bool begin_dad = false;
3231 spin_lock_bh(&ifp->state_lock);
3232 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3233 ifp->state = INET6_IFADDR_STATE_PREDAD;
3236 spin_unlock_bh(&ifp->state_lock);
3239 addrconf_mod_dad_work(ifp, 0);
3242 static void addrconf_dad_work(struct work_struct *w)
3244 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3245 struct inet6_ifaddr,
3247 struct inet6_dev *idev = ifp->idev;
3248 struct in6_addr mcaddr;
3254 } action = DAD_PROCESS;
3258 spin_lock_bh(&ifp->state_lock);
3259 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3261 ifp->state = INET6_IFADDR_STATE_DAD;
3262 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3264 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3266 spin_unlock_bh(&ifp->state_lock);
3268 if (action == DAD_BEGIN) {
3269 addrconf_dad_begin(ifp);
3271 } else if (action == DAD_ABORT) {
3273 addrconf_dad_stop(ifp, 1);
3277 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3280 write_lock_bh(&idev->lock);
3281 if (idev->dead || !(idev->if_flags & IF_READY)) {
3282 write_unlock_bh(&idev->lock);
3286 spin_lock(&ifp->lock);
3287 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3288 spin_unlock(&ifp->lock);
3289 write_unlock_bh(&idev->lock);
3293 if (ifp->dad_probes == 0) {
3295 * DAD was successful
3298 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3299 spin_unlock(&ifp->lock);
3300 write_unlock_bh(&idev->lock);
3302 addrconf_dad_completed(ifp);
3308 addrconf_mod_dad_work(ifp,
3309 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3310 spin_unlock(&ifp->lock);
3311 write_unlock_bh(&idev->lock);
3313 /* send a neighbour solicitation for our addr */
3314 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3315 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3321 /* ifp->idev must be at least read locked */
3322 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3324 struct inet6_ifaddr *ifpiter;
3325 struct inet6_dev *idev = ifp->idev;
3327 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3328 if (ifpiter->scope > IFA_LINK)
3330 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3331 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3332 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3339 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3341 struct net_device *dev = ifp->idev->dev;
3342 struct in6_addr lladdr;
3343 bool send_rs, send_mld;
3345 addrconf_del_dad_work(ifp);
3348 * Configure the address for reception. Now it is valid.
3351 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3353 /* If added prefix is link local and we are prepared to process
3354 router advertisements, start sending router solicitations.
3357 read_lock_bh(&ifp->idev->lock);
3358 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3359 send_rs = send_mld &&
3360 ipv6_accept_ra(ifp->idev) &&
3361 ifp->idev->cnf.rtr_solicits > 0 &&
3362 (dev->flags&IFF_LOOPBACK) == 0;
3363 read_unlock_bh(&ifp->idev->lock);
3365 /* While dad is in progress mld report's source address is in6_addrany.
3366 * Resend with proper ll now.
3369 ipv6_mc_dad_complete(ifp->idev);
3373 * If a host as already performed a random delay
3374 * [...] as part of DAD [...] there is no need
3375 * to delay again before sending the first RS
3377 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3379 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3381 write_lock_bh(&ifp->idev->lock);
3382 spin_lock(&ifp->lock);
3383 ifp->idev->rs_probes = 1;
3384 ifp->idev->if_flags |= IF_RS_SENT;
3385 addrconf_mod_rs_timer(ifp->idev,
3386 ifp->idev->cnf.rtr_solicit_interval);
3387 spin_unlock(&ifp->lock);
3388 write_unlock_bh(&ifp->idev->lock);
3392 static void addrconf_dad_run(struct inet6_dev *idev)
3394 struct inet6_ifaddr *ifp;
3396 read_lock_bh(&idev->lock);
3397 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3398 spin_lock(&ifp->lock);
3399 if (ifp->flags & IFA_F_TENTATIVE &&
3400 ifp->state == INET6_IFADDR_STATE_DAD)
3401 addrconf_dad_kick(ifp);
3402 spin_unlock(&ifp->lock);
3404 read_unlock_bh(&idev->lock);
3407 #ifdef CONFIG_PROC_FS
3408 struct if6_iter_state {
3409 struct seq_net_private p;
3414 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3416 struct inet6_ifaddr *ifa = NULL;
3417 struct if6_iter_state *state = seq->private;
3418 struct net *net = seq_file_net(seq);
3421 /* initial bucket if pos is 0 */
3427 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3428 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3430 if (!net_eq(dev_net(ifa->idev->dev), net))
3432 /* sync with offset */
3433 if (p < state->offset) {
3441 /* prepare for next bucket */
3448 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3449 struct inet6_ifaddr *ifa)
3451 struct if6_iter_state *state = seq->private;
3452 struct net *net = seq_file_net(seq);
3454 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3455 if (!net_eq(dev_net(ifa->idev->dev), net))
3461 while (++state->bucket < IN6_ADDR_HSIZE) {
3463 hlist_for_each_entry_rcu_bh(ifa,
3464 &inet6_addr_lst[state->bucket], addr_lst) {
3465 if (!net_eq(dev_net(ifa->idev->dev), net))
3475 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3479 return if6_get_first(seq, *pos);
3482 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3484 struct inet6_ifaddr *ifa;
3486 ifa = if6_get_next(seq, v);
3491 static void if6_seq_stop(struct seq_file *seq, void *v)
3494 rcu_read_unlock_bh();
3497 static int if6_seq_show(struct seq_file *seq, void *v)
3499 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3500 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3502 ifp->idev->dev->ifindex,
3506 ifp->idev->dev->name);
3510 static const struct seq_operations if6_seq_ops = {
3511 .start = if6_seq_start,
3512 .next = if6_seq_next,
3513 .show = if6_seq_show,
3514 .stop = if6_seq_stop,
3517 static int if6_seq_open(struct inode *inode, struct file *file)
3519 return seq_open_net(inode, file, &if6_seq_ops,
3520 sizeof(struct if6_iter_state));
3523 static const struct file_operations if6_fops = {
3524 .owner = THIS_MODULE,
3525 .open = if6_seq_open,
3527 .llseek = seq_lseek,
3528 .release = seq_release_net,
3531 static int __net_init if6_proc_net_init(struct net *net)
3533 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3538 static void __net_exit if6_proc_net_exit(struct net *net)
3540 remove_proc_entry("if_inet6", net->proc_net);
3543 static struct pernet_operations if6_proc_net_ops = {
3544 .init = if6_proc_net_init,
3545 .exit = if6_proc_net_exit,
3548 int __init if6_proc_init(void)
3550 return register_pernet_subsys(&if6_proc_net_ops);
3553 void if6_proc_exit(void)
3555 unregister_pernet_subsys(&if6_proc_net_ops);
3557 #endif /* CONFIG_PROC_FS */
3559 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3560 /* Check if address is a home address configured on any interface. */
3561 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3564 struct inet6_ifaddr *ifp = NULL;
3565 unsigned int hash = inet6_addr_hash(addr);
3568 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3569 if (!net_eq(dev_net(ifp->idev->dev), net))
3571 if (ipv6_addr_equal(&ifp->addr, addr) &&
3572 (ifp->flags & IFA_F_HOMEADDRESS)) {
3577 rcu_read_unlock_bh();
3583 * Periodic address status verification
3586 static void addrconf_verify_rtnl(void)
3588 unsigned long now, next, next_sec, next_sched;
3589 struct inet6_ifaddr *ifp;
3596 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3598 cancel_delayed_work(&addr_chk_work);
3600 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3602 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3605 /* When setting preferred_lft to a value not zero or
3606 * infinity, while valid_lft is infinity
3607 * IFA_F_PERMANENT has a non-infinity life time.
3609 if ((ifp->flags & IFA_F_PERMANENT) &&
3610 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3613 spin_lock(&ifp->lock);
3614 /* We try to batch several events at once. */
3615 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3617 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3618 age >= ifp->valid_lft) {
3619 spin_unlock(&ifp->lock);
3623 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3624 spin_unlock(&ifp->lock);
3626 } else if (age >= ifp->prefered_lft) {
3627 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3630 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3632 ifp->flags |= IFA_F_DEPRECATED;
3635 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3636 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3637 next = ifp->tstamp + ifp->valid_lft * HZ;
3639 spin_unlock(&ifp->lock);
3644 ipv6_ifa_notify(0, ifp);
3648 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3649 !(ifp->flags&IFA_F_TENTATIVE)) {
3650 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3651 ifp->idev->cnf.dad_transmits *
3652 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3654 if (age >= ifp->prefered_lft - regen_advance) {
3655 struct inet6_ifaddr *ifpub = ifp->ifpub;
3656 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3657 next = ifp->tstamp + ifp->prefered_lft * HZ;
3658 if (!ifp->regen_count && ifpub) {
3661 in6_ifa_hold(ifpub);
3662 spin_unlock(&ifp->lock);
3664 spin_lock(&ifpub->lock);
3665 ifpub->regen_count = 0;
3666 spin_unlock(&ifpub->lock);
3667 ipv6_create_tempaddr(ifpub, ifp);
3672 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3673 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3674 spin_unlock(&ifp->lock);
3676 /* ifp->prefered_lft <= ifp->valid_lft */
3677 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3678 next = ifp->tstamp + ifp->prefered_lft * HZ;
3679 spin_unlock(&ifp->lock);
3684 next_sec = round_jiffies_up(next);
3687 /* If rounded timeout is accurate enough, accept it. */
3688 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3689 next_sched = next_sec;
3691 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3692 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3693 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3695 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3696 now, next, next_sec, next_sched);
3697 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3698 rcu_read_unlock_bh();
3701 static void addrconf_verify_work(struct work_struct *w)
3704 addrconf_verify_rtnl();
3708 static void addrconf_verify(void)
3710 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3713 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3714 struct in6_addr **peer_pfx)
3716 struct in6_addr *pfx = NULL;
3721 pfx = nla_data(addr);
3724 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3726 pfx = nla_data(local);
3732 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3733 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3734 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3735 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3736 [IFA_FLAGS] = { .len = sizeof(u32) },
3740 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3742 struct net *net = sock_net(skb->sk);
3743 struct ifaddrmsg *ifm;
3744 struct nlattr *tb[IFA_MAX+1];
3745 struct in6_addr *pfx, *peer_pfx;
3749 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3753 ifm = nlmsg_data(nlh);
3754 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3758 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3760 /* We ignore other flags so far. */
3761 ifa_flags &= IFA_F_MANAGETEMPADDR;
3763 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3764 ifm->ifa_prefixlen);
3767 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3768 u32 prefered_lft, u32 valid_lft)
3772 unsigned long timeout;
3773 bool was_managetempaddr;
3774 bool had_prefixroute;
3778 if (!valid_lft || (prefered_lft > valid_lft))
3781 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3782 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3785 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3786 if (addrconf_finite_timeout(timeout)) {
3787 expires = jiffies_to_clock_t(timeout * HZ);
3788 valid_lft = timeout;
3789 flags = RTF_EXPIRES;
3793 ifa_flags |= IFA_F_PERMANENT;
3796 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3797 if (addrconf_finite_timeout(timeout)) {
3799 ifa_flags |= IFA_F_DEPRECATED;
3800 prefered_lft = timeout;
3803 spin_lock_bh(&ifp->lock);
3804 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3805 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3806 !(ifp->flags & IFA_F_NOPREFIXROUTE);
3807 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3808 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3809 IFA_F_NOPREFIXROUTE);
3810 ifp->flags |= ifa_flags;
3811 ifp->tstamp = jiffies;
3812 ifp->valid_lft = valid_lft;
3813 ifp->prefered_lft = prefered_lft;
3815 spin_unlock_bh(&ifp->lock);
3816 if (!(ifp->flags&IFA_F_TENTATIVE))
3817 ipv6_ifa_notify(0, ifp);
3819 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3820 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3822 } else if (had_prefixroute) {
3823 enum cleanup_prefix_rt_t action;
3824 unsigned long rt_expires;
3826 write_lock_bh(&ifp->idev->lock);
3827 action = check_cleanup_prefix_route(ifp, &rt_expires);
3828 write_unlock_bh(&ifp->idev->lock);
3830 if (action != CLEANUP_PREFIX_RT_NOP) {
3831 cleanup_prefix_route(ifp, rt_expires,
3832 action == CLEANUP_PREFIX_RT_DEL);
3836 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3837 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3838 valid_lft = prefered_lft = 0;
3839 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3840 !was_managetempaddr, jiffies);
3843 addrconf_verify_rtnl();
3849 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3851 struct net *net = sock_net(skb->sk);
3852 struct ifaddrmsg *ifm;
3853 struct nlattr *tb[IFA_MAX+1];
3854 struct in6_addr *pfx, *peer_pfx;
3855 struct inet6_ifaddr *ifa;
3856 struct net_device *dev;
3857 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3861 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3865 ifm = nlmsg_data(nlh);
3866 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3870 if (tb[IFA_CACHEINFO]) {
3871 struct ifa_cacheinfo *ci;
3873 ci = nla_data(tb[IFA_CACHEINFO]);
3874 valid_lft = ci->ifa_valid;
3875 preferred_lft = ci->ifa_prefered;
3877 preferred_lft = INFINITY_LIFE_TIME;
3878 valid_lft = INFINITY_LIFE_TIME;
3881 dev = __dev_get_by_index(net, ifm->ifa_index);
3885 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3887 /* We ignore other flags so far. */
3888 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3889 IFA_F_NOPREFIXROUTE;
3891 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3894 * It would be best to check for !NLM_F_CREATE here but
3895 * userspace already relies on not having to provide this.
3897 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3898 ifm->ifa_prefixlen, ifa_flags,
3899 preferred_lft, valid_lft);
3902 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3903 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3906 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3913 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3914 u8 scope, int ifindex)
3916 struct ifaddrmsg *ifm;
3918 ifm = nlmsg_data(nlh);
3919 ifm->ifa_family = AF_INET6;
3920 ifm->ifa_prefixlen = prefixlen;
3921 ifm->ifa_flags = flags;
3922 ifm->ifa_scope = scope;
3923 ifm->ifa_index = ifindex;
3926 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3927 unsigned long tstamp, u32 preferred, u32 valid)
3929 struct ifa_cacheinfo ci;
3931 ci.cstamp = cstamp_delta(cstamp);
3932 ci.tstamp = cstamp_delta(tstamp);
3933 ci.ifa_prefered = preferred;
3934 ci.ifa_valid = valid;
3936 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3939 static inline int rt_scope(int ifa_scope)
3941 if (ifa_scope & IFA_HOST)
3942 return RT_SCOPE_HOST;
3943 else if (ifa_scope & IFA_LINK)
3944 return RT_SCOPE_LINK;
3945 else if (ifa_scope & IFA_SITE)
3946 return RT_SCOPE_SITE;
3948 return RT_SCOPE_UNIVERSE;
3951 static inline int inet6_ifaddr_msgsize(void)
3953 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3954 + nla_total_size(16) /* IFA_LOCAL */
3955 + nla_total_size(16) /* IFA_ADDRESS */
3956 + nla_total_size(sizeof(struct ifa_cacheinfo))
3957 + nla_total_size(4) /* IFA_FLAGS */;
3960 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3961 u32 portid, u32 seq, int event, unsigned int flags)
3963 struct nlmsghdr *nlh;
3964 u32 preferred, valid;
3966 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3970 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3971 ifa->idev->dev->ifindex);
3973 if (!((ifa->flags&IFA_F_PERMANENT) &&
3974 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3975 preferred = ifa->prefered_lft;
3976 valid = ifa->valid_lft;
3977 if (preferred != INFINITY_LIFE_TIME) {
3978 long tval = (jiffies - ifa->tstamp)/HZ;
3979 if (preferred > tval)
3983 if (valid != INFINITY_LIFE_TIME) {
3991 preferred = INFINITY_LIFE_TIME;
3992 valid = INFINITY_LIFE_TIME;
3995 if (!ipv6_addr_any(&ifa->peer_addr)) {
3996 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3997 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
4000 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4003 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4006 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4009 return nlmsg_end(skb, nlh);
4012 nlmsg_cancel(skb, nlh);
4016 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4017 u32 portid, u32 seq, int event, u16 flags)
4019 struct nlmsghdr *nlh;
4020 u8 scope = RT_SCOPE_UNIVERSE;
4021 int ifindex = ifmca->idev->dev->ifindex;
4023 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4024 scope = RT_SCOPE_SITE;
4026 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4030 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4031 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4032 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4033 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4034 nlmsg_cancel(skb, nlh);
4038 return nlmsg_end(skb, nlh);
4041 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4042 u32 portid, u32 seq, int event, unsigned int flags)
4044 struct nlmsghdr *nlh;
4045 u8 scope = RT_SCOPE_UNIVERSE;
4046 int ifindex = ifaca->aca_idev->dev->ifindex;
4048 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4049 scope = RT_SCOPE_SITE;
4051 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4055 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4056 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4057 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4058 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4059 nlmsg_cancel(skb, nlh);
4063 return nlmsg_end(skb, nlh);
4072 /* called with rcu_read_lock() */
4073 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4074 struct netlink_callback *cb, enum addr_type_t type,
4075 int s_ip_idx, int *p_ip_idx)
4077 struct ifmcaddr6 *ifmca;
4078 struct ifacaddr6 *ifaca;
4080 int ip_idx = *p_ip_idx;
4082 read_lock_bh(&idev->lock);
4084 case UNICAST_ADDR: {
4085 struct inet6_ifaddr *ifa;
4087 /* unicast address incl. temp addr */
4088 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4089 if (ip_idx < s_ip_idx)
4091 err = inet6_fill_ifaddr(skb, ifa,
4092 NETLINK_CB(cb->skb).portid,
4098 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4104 case MULTICAST_ADDR:
4105 /* multicast address */
4106 for (ifmca = idev->mc_list; ifmca;
4107 ifmca = ifmca->next, ip_idx++) {
4108 if (ip_idx < s_ip_idx)
4110 err = inet6_fill_ifmcaddr(skb, ifmca,
4111 NETLINK_CB(cb->skb).portid,
4120 /* anycast address */
4121 for (ifaca = idev->ac_list; ifaca;
4122 ifaca = ifaca->aca_next, ip_idx++) {
4123 if (ip_idx < s_ip_idx)
4125 err = inet6_fill_ifacaddr(skb, ifaca,
4126 NETLINK_CB(cb->skb).portid,
4137 read_unlock_bh(&idev->lock);
4142 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4143 enum addr_type_t type)
4145 struct net *net = sock_net(skb->sk);
4148 int s_idx, s_ip_idx;
4149 struct net_device *dev;
4150 struct inet6_dev *idev;
4151 struct hlist_head *head;
4154 s_idx = idx = cb->args[1];
4155 s_ip_idx = ip_idx = cb->args[2];
4158 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4159 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4161 head = &net->dev_index_head[h];
4162 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4165 if (h > s_h || idx > s_idx)
4168 idev = __in6_dev_get(dev);
4172 if (in6_dump_addrs(idev, skb, cb, type,
4173 s_ip_idx, &ip_idx) <= 0)
4183 cb->args[2] = ip_idx;
4188 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4190 enum addr_type_t type = UNICAST_ADDR;
4192 return inet6_dump_addr(skb, cb, type);
4195 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4197 enum addr_type_t type = MULTICAST_ADDR;
4199 return inet6_dump_addr(skb, cb, type);
4203 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4205 enum addr_type_t type = ANYCAST_ADDR;
4207 return inet6_dump_addr(skb, cb, type);
4210 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4212 struct net *net = sock_net(in_skb->sk);
4213 struct ifaddrmsg *ifm;
4214 struct nlattr *tb[IFA_MAX+1];
4215 struct in6_addr *addr = NULL, *peer;
4216 struct net_device *dev = NULL;
4217 struct inet6_ifaddr *ifa;
4218 struct sk_buff *skb;
4221 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4225 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4231 ifm = nlmsg_data(nlh);
4233 dev = __dev_get_by_index(net, ifm->ifa_index);
4235 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4237 err = -EADDRNOTAVAIL;
4241 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4247 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4248 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4250 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4251 WARN_ON(err == -EMSGSIZE);
4255 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4262 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4264 struct sk_buff *skb;
4265 struct net *net = dev_net(ifa->idev->dev);
4268 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4272 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4274 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4275 WARN_ON(err == -EMSGSIZE);
4279 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4283 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4286 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4287 __s32 *array, int bytes)
4289 BUG_ON(bytes < (DEVCONF_MAX * 4));
4291 memset(array, 0, bytes);
4292 array[DEVCONF_FORWARDING] = cnf->forwarding;
4293 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4294 array[DEVCONF_MTU6] = cnf->mtu6;
4295 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4296 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4297 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4298 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4299 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4300 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4301 jiffies_to_msecs(cnf->rtr_solicit_interval);
4302 array[DEVCONF_RTR_SOLICIT_DELAY] =
4303 jiffies_to_msecs(cnf->rtr_solicit_delay);
4304 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4305 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4306 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4307 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4308 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4309 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4310 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4311 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4312 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4313 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4314 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4315 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4316 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4317 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4318 #ifdef CONFIG_IPV6_ROUTER_PREF
4319 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4320 array[DEVCONF_RTR_PROBE_INTERVAL] =
4321 jiffies_to_msecs(cnf->rtr_probe_interval);
4322 #ifdef CONFIG_IPV6_ROUTE_INFO
4323 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4326 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4327 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4328 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4329 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4331 #ifdef CONFIG_IPV6_MROUTE
4332 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4334 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4335 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4336 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4337 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4338 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4341 static inline size_t inet6_ifla6_size(void)
4343 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4344 + nla_total_size(sizeof(struct ifla_cacheinfo))
4345 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4346 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4347 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4348 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4351 static inline size_t inet6_if_nlmsg_size(void)
4353 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4354 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4355 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4356 + nla_total_size(4) /* IFLA_MTU */
4357 + nla_total_size(4) /* IFLA_LINK */
4358 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4361 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4362 int items, int bytes)
4365 int pad = bytes - sizeof(u64) * items;
4368 /* Use put_unaligned() because stats may not be aligned for u64. */
4369 put_unaligned(items, &stats[0]);
4370 for (i = 1; i < items; i++)
4371 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4373 memset(&stats[items], 0, pad);
4376 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4377 int items, int bytes, size_t syncpoff)
4380 int pad = bytes - sizeof(u64) * items;
4383 /* Use put_unaligned() because stats may not be aligned for u64. */
4384 put_unaligned(items, &stats[0]);
4385 for (i = 1; i < items; i++)
4386 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4388 memset(&stats[items], 0, pad);
4391 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4395 case IFLA_INET6_STATS:
4396 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4397 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4399 case IFLA_INET6_ICMP6STATS:
4400 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4405 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4408 struct ifla_cacheinfo ci;
4410 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4411 goto nla_put_failure;
4412 ci.max_reasm_len = IPV6_MAXPLEN;
4413 ci.tstamp = cstamp_delta(idev->tstamp);
4414 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4415 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4416 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4417 goto nla_put_failure;
4418 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4420 goto nla_put_failure;
4421 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4423 /* XXX - MC not implemented */
4425 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4427 goto nla_put_failure;
4428 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4430 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4432 goto nla_put_failure;
4433 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4435 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4437 goto nla_put_failure;
4438 read_lock_bh(&idev->lock);
4439 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4440 read_unlock_bh(&idev->lock);
4448 static size_t inet6_get_link_af_size(const struct net_device *dev)
4450 if (!__in6_dev_get(dev))
4453 return inet6_ifla6_size();
4456 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4458 struct inet6_dev *idev = __in6_dev_get(dev);
4463 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4469 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4471 struct inet6_ifaddr *ifp;
4472 struct net_device *dev = idev->dev;
4473 bool clear_token, update_rs = false;
4474 struct in6_addr ll_addr;
4480 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4482 if (!ipv6_accept_ra(idev))
4484 if (idev->cnf.rtr_solicits <= 0)
4487 write_lock_bh(&idev->lock);
4489 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4490 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4492 write_unlock_bh(&idev->lock);
4494 clear_token = ipv6_addr_any(token);
4498 if (!idev->dead && (idev->if_flags & IF_READY) &&
4499 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4500 IFA_F_OPTIMISTIC)) {
4501 /* If we're not ready, then normal ifup will take care
4502 * of this. Otherwise, we need to request our rs here.
4504 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4509 write_lock_bh(&idev->lock);
4512 idev->if_flags |= IF_RS_SENT;
4513 idev->rs_probes = 1;
4514 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4517 /* Well, that's kinda nasty ... */
4518 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4519 spin_lock(&ifp->lock);
4520 if (ifp->tokenized) {
4522 ifp->prefered_lft = 0;
4524 spin_unlock(&ifp->lock);
4527 write_unlock_bh(&idev->lock);
4528 addrconf_verify_rtnl();
4532 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4535 struct inet6_dev *idev = __in6_dev_get(dev);
4536 struct nlattr *tb[IFLA_INET6_MAX + 1];
4539 return -EAFNOSUPPORT;
4541 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4544 if (tb[IFLA_INET6_TOKEN])
4545 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4550 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4551 u32 portid, u32 seq, int event, unsigned int flags)
4553 struct net_device *dev = idev->dev;
4554 struct ifinfomsg *hdr;
4555 struct nlmsghdr *nlh;
4558 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4562 hdr = nlmsg_data(nlh);
4563 hdr->ifi_family = AF_INET6;
4565 hdr->ifi_type = dev->type;
4566 hdr->ifi_index = dev->ifindex;
4567 hdr->ifi_flags = dev_get_flags(dev);
4568 hdr->ifi_change = 0;
4570 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4572 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4573 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4574 (dev->ifindex != dev->iflink &&
4575 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4576 goto nla_put_failure;
4577 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4578 if (protoinfo == NULL)
4579 goto nla_put_failure;
4581 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4582 goto nla_put_failure;
4584 nla_nest_end(skb, protoinfo);
4585 return nlmsg_end(skb, nlh);
4588 nlmsg_cancel(skb, nlh);
4592 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4594 struct net *net = sock_net(skb->sk);
4597 struct net_device *dev;
4598 struct inet6_dev *idev;
4599 struct hlist_head *head;
4602 s_idx = cb->args[1];
4605 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4607 head = &net->dev_index_head[h];
4608 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4611 idev = __in6_dev_get(dev);
4614 if (inet6_fill_ifinfo(skb, idev,
4615 NETLINK_CB(cb->skb).portid,
4617 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4631 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4633 struct sk_buff *skb;
4634 struct net *net = dev_net(idev->dev);
4637 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4641 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4643 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4644 WARN_ON(err == -EMSGSIZE);
4648 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4652 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4655 static inline size_t inet6_prefix_nlmsg_size(void)
4657 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4658 + nla_total_size(sizeof(struct in6_addr))
4659 + nla_total_size(sizeof(struct prefix_cacheinfo));
4662 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4663 struct prefix_info *pinfo, u32 portid, u32 seq,
4664 int event, unsigned int flags)
4666 struct prefixmsg *pmsg;
4667 struct nlmsghdr *nlh;
4668 struct prefix_cacheinfo ci;
4670 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4674 pmsg = nlmsg_data(nlh);
4675 pmsg->prefix_family = AF_INET6;
4676 pmsg->prefix_pad1 = 0;
4677 pmsg->prefix_pad2 = 0;
4678 pmsg->prefix_ifindex = idev->dev->ifindex;
4679 pmsg->prefix_len = pinfo->prefix_len;
4680 pmsg->prefix_type = pinfo->type;
4681 pmsg->prefix_pad3 = 0;
4682 pmsg->prefix_flags = 0;
4684 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4685 if (pinfo->autoconf)
4686 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4688 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4689 goto nla_put_failure;
4690 ci.preferred_time = ntohl(pinfo->prefered);
4691 ci.valid_time = ntohl(pinfo->valid);
4692 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4693 goto nla_put_failure;
4694 return nlmsg_end(skb, nlh);
4697 nlmsg_cancel(skb, nlh);
4701 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4702 struct prefix_info *pinfo)
4704 struct sk_buff *skb;
4705 struct net *net = dev_net(idev->dev);
4708 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4712 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4714 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4715 WARN_ON(err == -EMSGSIZE);
4719 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4723 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4726 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4728 struct net *net = dev_net(ifp->idev->dev);
4733 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4738 * If the address was optimistic
4739 * we inserted the route at the start of
4740 * our DAD process, so we don't need
4743 if (!rcu_access_pointer(ifp->rt->rt6i_node))
4744 ip6_ins_rt(ifp->rt);
4745 if (ifp->idev->cnf.forwarding)
4746 addrconf_join_anycast(ifp);
4747 if (!ipv6_addr_any(&ifp->peer_addr))
4748 addrconf_prefix_route(&ifp->peer_addr, 128,
4749 ifp->idev->dev, 0, 0);
4752 if (ifp->idev->cnf.forwarding)
4753 addrconf_leave_anycast(ifp);
4754 addrconf_leave_solict(ifp->idev, &ifp->addr);
4755 if (!ipv6_addr_any(&ifp->peer_addr)) {
4756 struct rt6_info *rt;
4757 struct net_device *dev = ifp->idev->dev;
4759 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4767 dst_hold(&ifp->rt->dst);
4769 if (ip6_del_rt(ifp->rt))
4770 dst_free(&ifp->rt->dst);
4772 rt_genid_bump_ipv6(net);
4775 atomic_inc(&net->ipv6.dev_addr_genid);
4778 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4781 if (likely(ifp->idev->dead == 0))
4782 __ipv6_ifa_notify(event, ifp);
4783 rcu_read_unlock_bh();
4786 #ifdef CONFIG_SYSCTL
4789 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4790 void __user *buffer, size_t *lenp, loff_t *ppos)
4792 int *valp = ctl->data;
4795 struct ctl_table lctl;
4799 * ctl->data points to idev->cnf.forwarding, we should
4800 * not modify it until we get the rtnl lock.
4805 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4808 ret = addrconf_fixup_forwarding(ctl, valp, val);
4815 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
4816 void __user *buffer, size_t *lenp, loff_t *ppos)
4818 struct inet6_dev *idev = ctl->extra1;
4819 int min_mtu = IPV6_MIN_MTU;
4820 struct ctl_table lctl;
4823 lctl.extra1 = &min_mtu;
4824 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
4826 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
4829 static void dev_disable_change(struct inet6_dev *idev)
4831 struct netdev_notifier_info info;
4833 if (!idev || !idev->dev)
4836 netdev_notifier_info_init(&info, idev->dev);
4837 if (idev->cnf.disable_ipv6)
4838 addrconf_notify(NULL, NETDEV_DOWN, &info);
4840 addrconf_notify(NULL, NETDEV_UP, &info);
4843 static void addrconf_disable_change(struct net *net, __s32 newf)
4845 struct net_device *dev;
4846 struct inet6_dev *idev;
4848 for_each_netdev(net, dev) {
4849 idev = __in6_dev_get(dev);
4851 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4852 idev->cnf.disable_ipv6 = newf;
4854 dev_disable_change(idev);
4859 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4864 if (!rtnl_trylock())
4865 return restart_syscall();
4867 net = (struct net *)table->extra2;
4871 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4876 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4877 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4878 addrconf_disable_change(net, newf);
4879 } else if ((!newf) ^ (!old))
4880 dev_disable_change((struct inet6_dev *)table->extra1);
4887 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4888 void __user *buffer, size_t *lenp, loff_t *ppos)
4890 int *valp = ctl->data;
4893 struct ctl_table lctl;
4897 * ctl->data points to idev->cnf.disable_ipv6, we should
4898 * not modify it until we get the rtnl lock.
4903 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4906 ret = addrconf_disable_ipv6(ctl, valp, val);
4913 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4914 void __user *buffer, size_t *lenp, loff_t *ppos)
4916 int *valp = ctl->data;
4921 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4924 if (write && old != new) {
4925 struct net *net = ctl->extra2;
4927 if (!rtnl_trylock())
4928 return restart_syscall();
4930 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4931 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4932 NETCONFA_IFINDEX_DEFAULT,
4933 net->ipv6.devconf_dflt);
4934 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4935 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4936 NETCONFA_IFINDEX_ALL,
4937 net->ipv6.devconf_all);
4939 struct inet6_dev *idev = ctl->extra1;
4941 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4952 static struct addrconf_sysctl_table
4954 struct ctl_table_header *sysctl_header;
4955 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4956 } addrconf_sysctl __read_mostly = {
4957 .sysctl_header = NULL,
4960 .procname = "forwarding",
4961 .data = &ipv6_devconf.forwarding,
4962 .maxlen = sizeof(int),
4964 .proc_handler = addrconf_sysctl_forward,
4967 .procname = "hop_limit",
4968 .data = &ipv6_devconf.hop_limit,
4969 .maxlen = sizeof(int),
4971 .proc_handler = proc_dointvec,
4975 .data = &ipv6_devconf.mtu6,
4976 .maxlen = sizeof(int),
4978 .proc_handler = addrconf_sysctl_mtu,
4981 .procname = "accept_ra",
4982 .data = &ipv6_devconf.accept_ra,
4983 .maxlen = sizeof(int),
4985 .proc_handler = proc_dointvec,
4988 .procname = "accept_redirects",
4989 .data = &ipv6_devconf.accept_redirects,
4990 .maxlen = sizeof(int),
4992 .proc_handler = proc_dointvec,
4995 .procname = "autoconf",
4996 .data = &ipv6_devconf.autoconf,
4997 .maxlen = sizeof(int),
4999 .proc_handler = proc_dointvec,
5002 .procname = "dad_transmits",
5003 .data = &ipv6_devconf.dad_transmits,
5004 .maxlen = sizeof(int),
5006 .proc_handler = proc_dointvec,
5009 .procname = "router_solicitations",
5010 .data = &ipv6_devconf.rtr_solicits,
5011 .maxlen = sizeof(int),
5013 .proc_handler = proc_dointvec,
5016 .procname = "router_solicitation_interval",
5017 .data = &ipv6_devconf.rtr_solicit_interval,
5018 .maxlen = sizeof(int),
5020 .proc_handler = proc_dointvec_jiffies,
5023 .procname = "router_solicitation_delay",
5024 .data = &ipv6_devconf.rtr_solicit_delay,
5025 .maxlen = sizeof(int),
5027 .proc_handler = proc_dointvec_jiffies,
5030 .procname = "force_mld_version",
5031 .data = &ipv6_devconf.force_mld_version,
5032 .maxlen = sizeof(int),
5034 .proc_handler = proc_dointvec,
5037 .procname = "mldv1_unsolicited_report_interval",
5039 &ipv6_devconf.mldv1_unsolicited_report_interval,
5040 .maxlen = sizeof(int),
5042 .proc_handler = proc_dointvec_ms_jiffies,
5045 .procname = "mldv2_unsolicited_report_interval",
5047 &ipv6_devconf.mldv2_unsolicited_report_interval,
5048 .maxlen = sizeof(int),
5050 .proc_handler = proc_dointvec_ms_jiffies,
5053 .procname = "use_tempaddr",
5054 .data = &ipv6_devconf.use_tempaddr,
5055 .maxlen = sizeof(int),
5057 .proc_handler = proc_dointvec,
5060 .procname = "temp_valid_lft",
5061 .data = &ipv6_devconf.temp_valid_lft,
5062 .maxlen = sizeof(int),
5064 .proc_handler = proc_dointvec,
5067 .procname = "temp_prefered_lft",
5068 .data = &ipv6_devconf.temp_prefered_lft,
5069 .maxlen = sizeof(int),
5071 .proc_handler = proc_dointvec,
5074 .procname = "regen_max_retry",
5075 .data = &ipv6_devconf.regen_max_retry,
5076 .maxlen = sizeof(int),
5078 .proc_handler = proc_dointvec,
5081 .procname = "max_desync_factor",
5082 .data = &ipv6_devconf.max_desync_factor,
5083 .maxlen = sizeof(int),
5085 .proc_handler = proc_dointvec,
5088 .procname = "max_addresses",
5089 .data = &ipv6_devconf.max_addresses,
5090 .maxlen = sizeof(int),
5092 .proc_handler = proc_dointvec,
5095 .procname = "accept_ra_defrtr",
5096 .data = &ipv6_devconf.accept_ra_defrtr,
5097 .maxlen = sizeof(int),
5099 .proc_handler = proc_dointvec,
5102 .procname = "accept_ra_min_hop_limit",
5103 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5104 .maxlen = sizeof(int),
5106 .proc_handler = proc_dointvec,
5109 .procname = "accept_ra_pinfo",
5110 .data = &ipv6_devconf.accept_ra_pinfo,
5111 .maxlen = sizeof(int),
5113 .proc_handler = proc_dointvec,
5115 #ifdef CONFIG_IPV6_ROUTER_PREF
5117 .procname = "accept_ra_rtr_pref",
5118 .data = &ipv6_devconf.accept_ra_rtr_pref,
5119 .maxlen = sizeof(int),
5121 .proc_handler = proc_dointvec,
5124 .procname = "router_probe_interval",
5125 .data = &ipv6_devconf.rtr_probe_interval,
5126 .maxlen = sizeof(int),
5128 .proc_handler = proc_dointvec_jiffies,
5130 #ifdef CONFIG_IPV6_ROUTE_INFO
5132 .procname = "accept_ra_rt_info_max_plen",
5133 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5134 .maxlen = sizeof(int),
5136 .proc_handler = proc_dointvec,
5141 .procname = "proxy_ndp",
5142 .data = &ipv6_devconf.proxy_ndp,
5143 .maxlen = sizeof(int),
5145 .proc_handler = addrconf_sysctl_proxy_ndp,
5148 .procname = "accept_source_route",
5149 .data = &ipv6_devconf.accept_source_route,
5150 .maxlen = sizeof(int),
5152 .proc_handler = proc_dointvec,
5154 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5156 .procname = "optimistic_dad",
5157 .data = &ipv6_devconf.optimistic_dad,
5158 .maxlen = sizeof(int),
5160 .proc_handler = proc_dointvec,
5164 #ifdef CONFIG_IPV6_MROUTE
5166 .procname = "mc_forwarding",
5167 .data = &ipv6_devconf.mc_forwarding,
5168 .maxlen = sizeof(int),
5170 .proc_handler = proc_dointvec,
5174 .procname = "disable_ipv6",
5175 .data = &ipv6_devconf.disable_ipv6,
5176 .maxlen = sizeof(int),
5178 .proc_handler = addrconf_sysctl_disable,
5181 .procname = "accept_dad",
5182 .data = &ipv6_devconf.accept_dad,
5183 .maxlen = sizeof(int),
5185 .proc_handler = proc_dointvec,
5188 .procname = "force_tllao",
5189 .data = &ipv6_devconf.force_tllao,
5190 .maxlen = sizeof(int),
5192 .proc_handler = proc_dointvec
5195 .procname = "ndisc_notify",
5196 .data = &ipv6_devconf.ndisc_notify,
5197 .maxlen = sizeof(int),
5199 .proc_handler = proc_dointvec
5202 .procname = "suppress_frag_ndisc",
5203 .data = &ipv6_devconf.suppress_frag_ndisc,
5204 .maxlen = sizeof(int),
5206 .proc_handler = proc_dointvec
5214 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5215 struct inet6_dev *idev, struct ipv6_devconf *p)
5218 struct addrconf_sysctl_table *t;
5219 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5221 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5225 for (i = 0; t->addrconf_vars[i].data; i++) {
5226 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5227 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5228 t->addrconf_vars[i].extra2 = net;
5231 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5233 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5234 if (t->sysctl_header == NULL)
5246 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5248 struct addrconf_sysctl_table *t;
5250 if (p->sysctl == NULL)
5255 unregister_net_sysctl_table(t->sysctl_header);
5259 static void addrconf_sysctl_register(struct inet6_dev *idev)
5261 neigh_sysctl_register(idev->dev, idev->nd_parms,
5262 &ndisc_ifinfo_sysctl_change);
5263 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5267 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5269 __addrconf_sysctl_unregister(&idev->cnf);
5270 neigh_sysctl_unregister(idev->nd_parms);
5276 static int __net_init addrconf_init_net(struct net *net)
5279 struct ipv6_devconf *all, *dflt;
5281 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5285 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5287 goto err_alloc_dflt;
5289 /* these will be inherited by all namespaces */
5290 dflt->autoconf = ipv6_defaults.autoconf;
5291 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5293 net->ipv6.devconf_all = all;
5294 net->ipv6.devconf_dflt = dflt;
5296 #ifdef CONFIG_SYSCTL
5297 err = __addrconf_sysctl_register(net, "all", NULL, all);
5301 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5307 #ifdef CONFIG_SYSCTL
5309 __addrconf_sysctl_unregister(all);
5319 static void __net_exit addrconf_exit_net(struct net *net)
5321 #ifdef CONFIG_SYSCTL
5322 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5323 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5325 if (!net_eq(net, &init_net)) {
5326 kfree(net->ipv6.devconf_dflt);
5327 kfree(net->ipv6.devconf_all);
5331 static struct pernet_operations addrconf_ops = {
5332 .init = addrconf_init_net,
5333 .exit = addrconf_exit_net,
5336 static struct rtnl_af_ops inet6_ops = {
5338 .fill_link_af = inet6_fill_link_af,
5339 .get_link_af_size = inet6_get_link_af_size,
5340 .set_link_af = inet6_set_link_af,
5344 * Init / cleanup code
5347 int __init addrconf_init(void)
5351 err = ipv6_addr_label_init();
5353 pr_crit("%s: cannot initialize default policy table: %d\n",
5358 err = register_pernet_subsys(&addrconf_ops);
5362 addrconf_wq = create_workqueue("ipv6_addrconf");
5368 /* The addrconf netdev notifier requires that loopback_dev
5369 * has it's ipv6 private information allocated and setup
5370 * before it can bring up and give link-local addresses
5371 * to other devices which are up.
5373 * Unfortunately, loopback_dev is not necessarily the first
5374 * entry in the global dev_base list of net devices. In fact,
5375 * it is likely to be the very last entry on that list.
5376 * So this causes the notifier registry below to try and
5377 * give link-local addresses to all devices besides loopback_dev
5378 * first, then loopback_dev, which cases all the non-loopback_dev
5379 * devices to fail to get a link-local address.
5381 * So, as a temporary fix, allocate the ipv6 structure for
5382 * loopback_dev first by hand.
5383 * Longer term, all of the dependencies ipv6 has upon the loopback
5384 * device and it being up should be removed.
5387 if (!ipv6_add_dev(init_net.loopback_dev))
5393 ip6_route_init_special_entries();
5395 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5396 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5398 register_netdevice_notifier(&ipv6_dev_notf);
5402 rtnl_af_register(&inet6_ops);
5404 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5409 /* Only the first call to __rtnl_register can fail */
5410 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5411 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5412 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5413 inet6_dump_ifaddr, NULL);
5414 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5415 inet6_dump_ifmcaddr, NULL);
5416 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5417 inet6_dump_ifacaddr, NULL);
5418 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5419 inet6_netconf_dump_devconf, NULL);
5421 ipv6_addr_label_rtnl_register();
5425 rtnl_af_unregister(&inet6_ops);
5426 unregister_netdevice_notifier(&ipv6_dev_notf);
5428 destroy_workqueue(addrconf_wq);
5430 unregister_pernet_subsys(&addrconf_ops);
5432 ipv6_addr_label_cleanup();
5437 void addrconf_cleanup(void)
5439 struct net_device *dev;
5442 unregister_netdevice_notifier(&ipv6_dev_notf);
5443 unregister_pernet_subsys(&addrconf_ops);
5444 ipv6_addr_label_cleanup();
5448 __rtnl_af_unregister(&inet6_ops);
5450 /* clean dev list */
5451 for_each_netdev(&init_net, dev) {
5452 if (__in6_dev_get(dev) == NULL)
5454 addrconf_ifdown(dev, 1);
5456 addrconf_ifdown(init_net.loopback_dev, 2);
5461 spin_lock_bh(&addrconf_hash_lock);
5462 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5463 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5464 spin_unlock_bh(&addrconf_hash_lock);
5465 cancel_delayed_work(&addr_chk_work);
5468 destroy_workqueue(addrconf_wq);