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>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
97 /* Set to 3 to get tracing... */
101 #define ADBG(x) printk x
106 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108 static inline u32 cstamp_delta(unsigned long cstamp)
110 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
113 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
114 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
115 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
118 static void addrconf_sysctl_register(struct inet6_dev *idev);
119 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
121 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
125 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
130 #ifdef CONFIG_IPV6_PRIVACY
131 static void __ipv6_regen_rndid(struct inet6_dev *idev);
132 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
133 static void ipv6_regen_rndid(unsigned long data);
136 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
137 static int ipv6_count_addresses(struct inet6_dev *idev);
140 * Configured unicast address hash table
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
145 static void addrconf_verify(unsigned long);
147 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
148 static DEFINE_SPINLOCK(addrconf_verify_lock);
150 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
151 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
153 static void addrconf_type_change(struct net_device *dev,
154 unsigned long event);
155 static int addrconf_ifdown(struct net_device *dev, int how);
157 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
158 static void addrconf_dad_timer(unsigned long data);
159 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
160 static void addrconf_dad_run(struct inet6_dev *idev);
161 static void addrconf_rs_timer(unsigned long data);
162 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
165 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
166 struct prefix_info *pinfo);
167 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
168 struct net_device *dev);
170 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
174 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
175 .mtu6 = IPV6_MIN_MTU,
177 .accept_redirects = 1,
179 .force_mld_version = 0,
181 .rtr_solicits = MAX_RTR_SOLICITATIONS,
182 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
183 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
184 #ifdef CONFIG_IPV6_PRIVACY
186 .temp_valid_lft = TEMP_VALID_LIFETIME,
187 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
188 .regen_max_retry = REGEN_MAX_RETRY,
189 .max_desync_factor = MAX_DESYNC_FACTOR,
191 .max_addresses = IPV6_MAX_ADDRESSES,
192 .accept_ra_defrtr = 1,
193 .accept_ra_pinfo = 1,
194 #ifdef CONFIG_IPV6_ROUTER_PREF
195 .accept_ra_rtr_pref = 1,
196 .rtr_probe_interval = 60 * HZ,
197 #ifdef CONFIG_IPV6_ROUTE_INFO
198 .accept_ra_rt_info_max_plen = 0,
202 .accept_source_route = 0, /* we do not accept RH0 by default. */
207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
209 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
210 .mtu6 = IPV6_MIN_MTU,
212 .accept_redirects = 1,
215 .rtr_solicits = MAX_RTR_SOLICITATIONS,
216 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
217 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
218 #ifdef CONFIG_IPV6_PRIVACY
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,
225 .max_addresses = IPV6_MAX_ADDRESSES,
226 .accept_ra_defrtr = 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. */
241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
247 /* Check if a valid qdisc is available */
248 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
250 return !qdisc_tx_is_noop(dev);
253 /* Check if a route is valid prefix route */
254 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
256 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
259 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
261 if (del_timer(&ifp->timer))
265 enum addrconf_timer_t {
271 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
272 enum addrconf_timer_t what,
275 if (!del_timer(&ifp->timer))
280 ifp->timer.function = addrconf_dad_timer;
283 ifp->timer.function = addrconf_rs_timer;
288 ifp->timer.expires = jiffies + when;
289 add_timer(&ifp->timer);
292 static int snmp6_alloc_dev(struct inet6_dev *idev)
294 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
295 sizeof(struct ipstats_mib),
296 __alignof__(struct ipstats_mib)) < 0)
298 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
300 if (!idev->stats.icmpv6dev)
302 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
304 if (!idev->stats.icmpv6msgdev)
310 kfree(idev->stats.icmpv6dev);
312 snmp_mib_free((void __percpu **)idev->stats.ipv6);
317 static void snmp6_free_dev(struct inet6_dev *idev)
319 kfree(idev->stats.icmpv6msgdev);
320 kfree(idev->stats.icmpv6dev);
321 snmp_mib_free((void __percpu **)idev->stats.ipv6);
324 /* Nobody refers to this device, we may destroy it. */
326 void in6_dev_finish_destroy(struct inet6_dev *idev)
328 struct net_device *dev = idev->dev;
330 WARN_ON(!list_empty(&idev->addr_list));
331 WARN_ON(idev->mc_list != NULL);
333 #ifdef NET_REFCNT_DEBUG
334 pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
338 pr_warn("Freeing alive inet6 device %p\n", idev);
341 snmp6_free_dev(idev);
342 kfree_rcu(idev, rcu);
344 EXPORT_SYMBOL(in6_dev_finish_destroy);
346 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
348 struct inet6_dev *ndev;
352 if (dev->mtu < IPV6_MIN_MTU)
355 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
360 rwlock_init(&ndev->lock);
362 INIT_LIST_HEAD(&ndev->addr_list);
364 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
365 ndev->cnf.mtu6 = dev->mtu;
366 ndev->cnf.sysctl = NULL;
367 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
368 if (ndev->nd_parms == NULL) {
372 if (ndev->cnf.forwarding)
373 dev_disable_lro(dev);
374 /* We refer to the device */
377 if (snmp6_alloc_dev(ndev) < 0) {
379 "%s: cannot allocate memory for statistics; dev=%s.\n",
380 __func__, dev->name));
381 neigh_parms_release(&nd_tbl, ndev->nd_parms);
387 if (snmp6_register_dev(ndev) < 0) {
389 "%s: cannot create /proc/net/dev_snmp6/%s\n",
390 __func__, dev->name));
391 neigh_parms_release(&nd_tbl, ndev->nd_parms);
393 in6_dev_finish_destroy(ndev);
397 /* One reference from device. We must do this before
398 * we invoke __ipv6_regen_rndid().
402 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
403 ndev->cnf.accept_dad = -1;
405 #if IS_ENABLED(CONFIG_IPV6_SIT)
406 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
407 pr_info("%s: Disabled Multicast RS\n", dev->name);
408 ndev->cnf.rtr_solicits = 0;
412 #ifdef CONFIG_IPV6_PRIVACY
413 INIT_LIST_HEAD(&ndev->tempaddr_list);
414 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
415 if ((dev->flags&IFF_LOOPBACK) ||
416 dev->type == ARPHRD_TUNNEL ||
417 dev->type == ARPHRD_TUNNEL6 ||
418 dev->type == ARPHRD_SIT ||
419 dev->type == ARPHRD_NONE) {
420 ndev->cnf.use_tempaddr = -1;
423 ipv6_regen_rndid((unsigned long) ndev);
427 if (netif_running(dev) && addrconf_qdisc_ok(dev))
428 ndev->if_flags |= IF_READY;
430 ipv6_mc_init_dev(ndev);
431 ndev->tstamp = jiffies;
432 addrconf_sysctl_register(ndev);
433 /* protected by rtnl_lock */
434 rcu_assign_pointer(dev->ip6_ptr, ndev);
436 /* Join all-node multicast group */
437 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
439 /* Join all-router multicast group if forwarding is set */
440 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
441 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
446 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
448 struct inet6_dev *idev;
452 idev = __in6_dev_get(dev);
454 idev = ipv6_add_dev(dev);
459 if (dev->flags&IFF_UP)
464 static int inet6_netconf_msgsize_devconf(int type)
466 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
467 + nla_total_size(4); /* NETCONFA_IFINDEX */
469 /* type -1 is used for ALL */
470 if (type == -1 || type == NETCONFA_FORWARDING)
471 size += nla_total_size(4);
476 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
477 struct ipv6_devconf *devconf, u32 portid,
478 u32 seq, int event, unsigned int flags,
481 struct nlmsghdr *nlh;
482 struct netconfmsg *ncm;
484 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
489 ncm = nlmsg_data(nlh);
490 ncm->ncm_family = AF_INET6;
492 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
493 goto nla_put_failure;
495 /* type -1 is used for ALL */
496 if ((type == -1 || type == NETCONFA_FORWARDING) &&
497 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
498 goto nla_put_failure;
500 return nlmsg_end(skb, nlh);
503 nlmsg_cancel(skb, nlh);
507 static void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
508 struct ipv6_devconf *devconf)
513 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
517 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
518 RTM_NEWNETCONF, 0, type);
520 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
521 WARN_ON(err == -EMSGSIZE);
525 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
529 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
532 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
533 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
534 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
537 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
538 struct nlmsghdr *nlh,
541 struct net *net = sock_net(in_skb->sk);
542 struct nlattr *tb[NETCONFA_MAX+1];
543 struct netconfmsg *ncm;
545 struct ipv6_devconf *devconf;
546 struct inet6_dev *in6_dev;
547 struct net_device *dev;
551 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
552 devconf_ipv6_policy);
557 if (!tb[NETCONFA_IFINDEX])
560 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
562 case NETCONFA_IFINDEX_ALL:
563 devconf = net->ipv6.devconf_all;
565 case NETCONFA_IFINDEX_DEFAULT:
566 devconf = net->ipv6.devconf_dflt;
569 dev = __dev_get_by_index(net, ifindex);
572 in6_dev = __in6_dev_get(dev);
575 devconf = &in6_dev->cnf;
580 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
584 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
585 NETLINK_CB(in_skb).portid,
586 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
589 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
590 WARN_ON(err == -EMSGSIZE);
594 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
600 static void dev_forward_change(struct inet6_dev *idev)
602 struct net_device *dev;
603 struct inet6_ifaddr *ifa;
608 if (idev->cnf.forwarding)
609 dev_disable_lro(dev);
610 if (dev->flags & IFF_MULTICAST) {
611 if (idev->cnf.forwarding)
612 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
614 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
617 list_for_each_entry(ifa, &idev->addr_list, if_list) {
618 if (ifa->flags&IFA_F_TENTATIVE)
620 if (idev->cnf.forwarding)
621 addrconf_join_anycast(ifa);
623 addrconf_leave_anycast(ifa);
625 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
626 dev->ifindex, &idev->cnf);
630 static void addrconf_forward_change(struct net *net, __s32 newf)
632 struct net_device *dev;
633 struct inet6_dev *idev;
635 for_each_netdev(net, dev) {
636 idev = __in6_dev_get(dev);
638 int changed = (!idev->cnf.forwarding) ^ (!newf);
639 idev->cnf.forwarding = newf;
641 dev_forward_change(idev);
646 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
652 return restart_syscall();
654 net = (struct net *)table->extra2;
658 if (p == &net->ipv6.devconf_dflt->forwarding) {
659 if ((!newf) ^ (!old))
660 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
661 NETCONFA_IFINDEX_DEFAULT,
662 net->ipv6.devconf_dflt);
667 if (p == &net->ipv6.devconf_all->forwarding) {
668 net->ipv6.devconf_dflt->forwarding = newf;
669 addrconf_forward_change(net, newf);
670 if ((!newf) ^ (!old))
671 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
672 NETCONFA_IFINDEX_ALL,
673 net->ipv6.devconf_all);
674 } else if ((!newf) ^ (!old))
675 dev_forward_change((struct inet6_dev *)table->extra1);
679 rt6_purge_dflt_routers(net);
684 /* Nobody refers to this ifaddr, destroy it */
685 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
687 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
689 #ifdef NET_REFCNT_DEBUG
690 pr_debug("%s\n", __func__);
693 in6_dev_put(ifp->idev);
695 if (del_timer(&ifp->timer))
696 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
698 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
699 pr_warn("Freeing alive inet6 address %p\n", ifp);
708 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
711 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
714 * Each device address list is sorted in order of scope -
715 * global before linklocal.
717 list_for_each(p, &idev->addr_list) {
718 struct inet6_ifaddr *ifa
719 = list_entry(p, struct inet6_ifaddr, if_list);
720 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
724 list_add_tail(&ifp->if_list, p);
727 static u32 inet6_addr_hash(const struct in6_addr *addr)
729 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
732 /* On success it returns ifp with increased reference count */
734 static struct inet6_ifaddr *
735 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
736 int scope, u32 flags)
738 struct inet6_ifaddr *ifa = NULL;
742 int addr_type = ipv6_addr_type(addr);
744 if (addr_type == IPV6_ADDR_ANY ||
745 addr_type & IPV6_ADDR_MULTICAST ||
746 (!(idev->dev->flags & IFF_LOOPBACK) &&
747 addr_type & IPV6_ADDR_LOOPBACK))
748 return ERR_PTR(-EADDRNOTAVAIL);
752 err = -ENODEV; /*XXX*/
756 if (idev->cnf.disable_ipv6) {
761 spin_lock(&addrconf_hash_lock);
763 /* Ignore adding duplicate addresses on an interface */
764 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
765 ADBG(("ipv6_add_addr: already assigned\n"));
770 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
773 ADBG(("ipv6_add_addr: malloc failed\n"));
778 rt = addrconf_dst_alloc(idev, addr, false);
786 spin_lock_init(&ifa->lock);
787 spin_lock_init(&ifa->state_lock);
788 init_timer(&ifa->timer);
789 INIT_HLIST_NODE(&ifa->addr_lst);
790 ifa->timer.data = (unsigned long) ifa;
792 ifa->prefix_len = pfxlen;
793 ifa->flags = flags | IFA_F_TENTATIVE;
794 ifa->cstamp = ifa->tstamp = jiffies;
803 /* Add to big hash table */
804 hash = inet6_addr_hash(addr);
806 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
807 spin_unlock(&addrconf_hash_lock);
809 write_lock(&idev->lock);
810 /* Add to inet6_dev unicast addr list. */
811 ipv6_link_dev_addr(idev, ifa);
813 #ifdef CONFIG_IPV6_PRIVACY
814 if (ifa->flags&IFA_F_TEMPORARY) {
815 list_add(&ifa->tmp_list, &idev->tempaddr_list);
821 write_unlock(&idev->lock);
823 rcu_read_unlock_bh();
825 if (likely(err == 0))
826 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
834 spin_unlock(&addrconf_hash_lock);
838 /* This function wants to get referenced ifp and releases it before return */
840 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
842 struct inet6_ifaddr *ifa, *ifn;
843 struct inet6_dev *idev = ifp->idev;
845 int deleted = 0, onlink = 0;
846 unsigned long expires = jiffies;
848 spin_lock_bh(&ifp->state_lock);
850 ifp->state = INET6_IFADDR_STATE_DEAD;
851 spin_unlock_bh(&ifp->state_lock);
853 if (state == INET6_IFADDR_STATE_DEAD)
856 spin_lock_bh(&addrconf_hash_lock);
857 hlist_del_init_rcu(&ifp->addr_lst);
858 spin_unlock_bh(&addrconf_hash_lock);
860 write_lock_bh(&idev->lock);
861 #ifdef CONFIG_IPV6_PRIVACY
862 if (ifp->flags&IFA_F_TEMPORARY) {
863 list_del(&ifp->tmp_list);
865 in6_ifa_put(ifp->ifpub);
872 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
874 list_del_init(&ifp->if_list);
877 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
881 } else if (ifp->flags & IFA_F_PERMANENT) {
882 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
884 if (ifa->flags & IFA_F_PERMANENT) {
889 unsigned long lifetime;
894 spin_lock(&ifa->lock);
896 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
898 * Note: Because this address is
899 * not permanent, lifetime <
900 * LONG_MAX / HZ here.
902 if (time_before(expires,
903 ifa->tstamp + lifetime * HZ))
904 expires = ifa->tstamp + lifetime * HZ;
905 spin_unlock(&ifa->lock);
910 write_unlock_bh(&idev->lock);
912 addrconf_del_timer(ifp);
914 ipv6_ifa_notify(RTM_DELADDR, ifp);
916 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
919 * Purge or update corresponding prefix
921 * 1) we don't purge prefix here if address was not permanent.
922 * prefix is managed by its own lifetime.
923 * 2) if there're no addresses, delete prefix.
924 * 3) if there're still other permanent address(es),
925 * corresponding prefix is still permanent.
926 * 4) otherwise, update prefix lifetime to the
927 * longest valid lifetime among the corresponding
928 * addresses on the device.
929 * Note: subsequent RA will update lifetime.
933 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
934 struct in6_addr prefix;
936 struct net *net = dev_net(ifp->idev->dev);
937 struct flowi6 fl6 = {};
939 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
940 fl6.flowi6_oif = ifp->idev->dev->ifindex;
942 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
945 if (rt != net->ipv6.ip6_null_entry &&
946 addrconf_is_prefix_route(rt)) {
950 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
951 rt6_set_expires(rt, expires);
957 /* clean up prefsrc entries */
958 rt6_remove_prefsrc(ifp);
963 #ifdef CONFIG_IPV6_PRIVACY
964 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
966 struct inet6_dev *idev = ifp->idev;
967 struct in6_addr addr, *tmpaddr;
968 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
969 unsigned long regen_advance;
974 unsigned long now = jiffies;
976 write_lock(&idev->lock);
978 spin_lock_bh(&ift->lock);
979 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
980 spin_unlock_bh(&ift->lock);
987 if (idev->cnf.use_tempaddr <= 0) {
988 write_unlock(&idev->lock);
989 pr_info("%s: use_tempaddr is disabled\n", __func__);
994 spin_lock_bh(&ifp->lock);
995 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
996 idev->cnf.use_tempaddr = -1; /*XXX*/
997 spin_unlock_bh(&ifp->lock);
998 write_unlock(&idev->lock);
999 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1006 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1007 __ipv6_try_regen_rndid(idev, tmpaddr);
1008 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1009 age = (now - ifp->tstamp) / HZ;
1010 tmp_valid_lft = min_t(__u32,
1012 idev->cnf.temp_valid_lft + age);
1013 tmp_prefered_lft = min_t(__u32,
1015 idev->cnf.temp_prefered_lft + age -
1016 idev->cnf.max_desync_factor);
1017 tmp_plen = ifp->prefix_len;
1018 max_addresses = idev->cnf.max_addresses;
1019 tmp_tstamp = ifp->tstamp;
1020 spin_unlock_bh(&ifp->lock);
1022 regen_advance = idev->cnf.regen_max_retry *
1023 idev->cnf.dad_transmits *
1024 idev->nd_parms->retrans_time / HZ;
1025 write_unlock(&idev->lock);
1027 /* A temporary address is created only if this calculated Preferred
1028 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1029 * an implementation must not create a temporary address with a zero
1030 * Preferred Lifetime.
1032 if (tmp_prefered_lft <= regen_advance) {
1039 addr_flags = IFA_F_TEMPORARY;
1040 /* set in addrconf_prefix_rcv() */
1041 if (ifp->flags & IFA_F_OPTIMISTIC)
1042 addr_flags |= IFA_F_OPTIMISTIC;
1044 ift = !max_addresses ||
1045 ipv6_count_addresses(idev) < max_addresses ?
1046 ipv6_add_addr(idev, &addr, tmp_plen,
1047 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1049 if (!ift || IS_ERR(ift)) {
1052 pr_info("%s: retry temporary address regeneration\n", __func__);
1054 write_lock(&idev->lock);
1058 spin_lock_bh(&ift->lock);
1060 ift->valid_lft = tmp_valid_lft;
1061 ift->prefered_lft = tmp_prefered_lft;
1063 ift->tstamp = tmp_tstamp;
1064 spin_unlock_bh(&ift->lock);
1066 addrconf_dad_start(ift);
1075 * Choose an appropriate source address (RFC3484)
1078 IPV6_SADDR_RULE_INIT = 0,
1079 IPV6_SADDR_RULE_LOCAL,
1080 IPV6_SADDR_RULE_SCOPE,
1081 IPV6_SADDR_RULE_PREFERRED,
1082 #ifdef CONFIG_IPV6_MIP6
1083 IPV6_SADDR_RULE_HOA,
1085 IPV6_SADDR_RULE_OIF,
1086 IPV6_SADDR_RULE_LABEL,
1087 #ifdef CONFIG_IPV6_PRIVACY
1088 IPV6_SADDR_RULE_PRIVACY,
1090 IPV6_SADDR_RULE_ORCHID,
1091 IPV6_SADDR_RULE_PREFIX,
1095 struct ipv6_saddr_score {
1098 struct inet6_ifaddr *ifa;
1099 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1104 struct ipv6_saddr_dst {
1105 const struct in6_addr *addr;
1112 static inline int ipv6_saddr_preferred(int type)
1114 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1119 static int ipv6_get_saddr_eval(struct net *net,
1120 struct ipv6_saddr_score *score,
1121 struct ipv6_saddr_dst *dst,
1126 if (i <= score->rule) {
1128 case IPV6_SADDR_RULE_SCOPE:
1129 ret = score->scopedist;
1131 case IPV6_SADDR_RULE_PREFIX:
1132 ret = score->matchlen;
1135 ret = !!test_bit(i, score->scorebits);
1141 case IPV6_SADDR_RULE_INIT:
1142 /* Rule 0: remember if hiscore is not ready yet */
1145 case IPV6_SADDR_RULE_LOCAL:
1146 /* Rule 1: Prefer same address */
1147 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1149 case IPV6_SADDR_RULE_SCOPE:
1150 /* Rule 2: Prefer appropriate scope
1155 * ---+--+-+---> scope
1157 * | d is scope of the destination.
1159 * | \ <- smaller scope is better if
1160 * B-15 | \ if scope is enough for destinaion.
1161 * | ret = B - scope (-1 <= scope >= d <= 15).
1163 * |/ <- greater is better
1164 * -C / if scope is not enough for destination.
1165 * /| ret = scope - C (-1 <= d < scope <= 15).
1167 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1168 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1169 * Assume B = 0 and we get C > 29.
1171 ret = __ipv6_addr_src_scope(score->addr_type);
1172 if (ret >= dst->scope)
1175 ret -= 128; /* 30 is enough */
1176 score->scopedist = ret;
1178 case IPV6_SADDR_RULE_PREFERRED:
1179 /* Rule 3: Avoid deprecated and optimistic addresses */
1180 ret = ipv6_saddr_preferred(score->addr_type) ||
1181 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1183 #ifdef CONFIG_IPV6_MIP6
1184 case IPV6_SADDR_RULE_HOA:
1186 /* Rule 4: Prefer home address */
1187 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1188 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1192 case IPV6_SADDR_RULE_OIF:
1193 /* Rule 5: Prefer outgoing interface */
1194 ret = (!dst->ifindex ||
1195 dst->ifindex == score->ifa->idev->dev->ifindex);
1197 case IPV6_SADDR_RULE_LABEL:
1198 /* Rule 6: Prefer matching label */
1199 ret = ipv6_addr_label(net,
1200 &score->ifa->addr, score->addr_type,
1201 score->ifa->idev->dev->ifindex) == dst->label;
1203 #ifdef CONFIG_IPV6_PRIVACY
1204 case IPV6_SADDR_RULE_PRIVACY:
1206 /* Rule 7: Prefer public address
1207 * Note: prefer temporary address if use_tempaddr >= 2
1209 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1210 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1211 score->ifa->idev->cnf.use_tempaddr >= 2;
1212 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1216 case IPV6_SADDR_RULE_ORCHID:
1217 /* Rule 8-: Prefer ORCHID vs ORCHID or
1218 * non-ORCHID vs non-ORCHID
1220 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1221 ipv6_addr_orchid(dst->addr));
1223 case IPV6_SADDR_RULE_PREFIX:
1224 /* Rule 8: Use longest matching prefix */
1225 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1226 if (ret > score->ifa->prefix_len)
1227 ret = score->ifa->prefix_len;
1228 score->matchlen = ret;
1235 __set_bit(i, score->scorebits);
1241 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1242 const struct in6_addr *daddr, unsigned int prefs,
1243 struct in6_addr *saddr)
1245 struct ipv6_saddr_score scores[2],
1246 *score = &scores[0], *hiscore = &scores[1];
1247 struct ipv6_saddr_dst dst;
1248 struct net_device *dev;
1251 dst_type = __ipv6_addr_type(daddr);
1253 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1254 dst.scope = __ipv6_addr_src_scope(dst_type);
1255 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1259 hiscore->ifa = NULL;
1263 for_each_netdev_rcu(net, dev) {
1264 struct inet6_dev *idev;
1266 /* Candidate Source Address (section 4)
1267 * - multicast and link-local destination address,
1268 * the set of candidate source address MUST only
1269 * include addresses assigned to interfaces
1270 * belonging to the same link as the outgoing
1272 * (- For site-local destination addresses, the
1273 * set of candidate source addresses MUST only
1274 * include addresses assigned to interfaces
1275 * belonging to the same site as the outgoing
1278 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1279 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1280 dst.ifindex && dev->ifindex != dst.ifindex)
1283 idev = __in6_dev_get(dev);
1287 read_lock_bh(&idev->lock);
1288 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1292 * - Tentative Address (RFC2462 section 5.4)
1293 * - A tentative address is not considered
1294 * "assigned to an interface" in the traditional
1295 * sense, unless it is also flagged as optimistic.
1296 * - Candidate Source Address (section 4)
1297 * - In any case, anycast addresses, multicast
1298 * addresses, and the unspecified address MUST
1299 * NOT be included in a candidate set.
1301 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1302 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1305 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1307 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1308 score->addr_type & IPV6_ADDR_MULTICAST)) {
1309 LIMIT_NETDEBUG(KERN_DEBUG
1310 "ADDRCONF: unspecified / multicast address "
1311 "assigned as unicast address on %s",
1317 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1319 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1320 int minihiscore, miniscore;
1322 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1323 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1325 if (minihiscore > miniscore) {
1326 if (i == IPV6_SADDR_RULE_SCOPE &&
1327 score->scopedist > 0) {
1330 * each remaining entry
1331 * has too small (not enough)
1332 * scope, because ifa entries
1333 * are sorted by their scope
1339 } else if (minihiscore < miniscore) {
1341 in6_ifa_put(hiscore->ifa);
1343 in6_ifa_hold(score->ifa);
1345 swap(hiscore, score);
1347 /* restore our iterator */
1348 score->ifa = hiscore->ifa;
1355 read_unlock_bh(&idev->lock);
1360 return -EADDRNOTAVAIL;
1362 *saddr = hiscore->ifa->addr;
1363 in6_ifa_put(hiscore->ifa);
1366 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1368 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1369 unsigned char banned_flags)
1371 struct inet6_dev *idev;
1372 int err = -EADDRNOTAVAIL;
1375 idev = __in6_dev_get(dev);
1377 struct inet6_ifaddr *ifp;
1379 read_lock_bh(&idev->lock);
1380 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1381 if (ifp->scope == IFA_LINK &&
1382 !(ifp->flags & banned_flags)) {
1388 read_unlock_bh(&idev->lock);
1394 static int ipv6_count_addresses(struct inet6_dev *idev)
1397 struct inet6_ifaddr *ifp;
1399 read_lock_bh(&idev->lock);
1400 list_for_each_entry(ifp, &idev->addr_list, if_list)
1402 read_unlock_bh(&idev->lock);
1406 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1407 struct net_device *dev, int strict)
1409 struct inet6_ifaddr *ifp;
1410 struct hlist_node *node;
1411 unsigned int hash = inet6_addr_hash(addr);
1414 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1415 if (!net_eq(dev_net(ifp->idev->dev), net))
1417 if (ipv6_addr_equal(&ifp->addr, addr) &&
1418 !(ifp->flags&IFA_F_TENTATIVE) &&
1419 (dev == NULL || ifp->idev->dev == dev ||
1420 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1421 rcu_read_unlock_bh();
1426 rcu_read_unlock_bh();
1429 EXPORT_SYMBOL(ipv6_chk_addr);
1431 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1432 struct net_device *dev)
1434 unsigned int hash = inet6_addr_hash(addr);
1435 struct inet6_ifaddr *ifp;
1436 struct hlist_node *node;
1438 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1439 if (!net_eq(dev_net(ifp->idev->dev), net))
1441 if (ipv6_addr_equal(&ifp->addr, addr)) {
1442 if (dev == NULL || ifp->idev->dev == dev)
1449 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1451 struct inet6_dev *idev;
1452 struct inet6_ifaddr *ifa;
1457 idev = __in6_dev_get(dev);
1459 read_lock_bh(&idev->lock);
1460 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1461 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1466 read_unlock_bh(&idev->lock);
1471 EXPORT_SYMBOL(ipv6_chk_prefix);
1473 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1474 struct net_device *dev, int strict)
1476 struct inet6_ifaddr *ifp, *result = NULL;
1477 unsigned int hash = inet6_addr_hash(addr);
1478 struct hlist_node *node;
1481 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1482 if (!net_eq(dev_net(ifp->idev->dev), net))
1484 if (ipv6_addr_equal(&ifp->addr, addr)) {
1485 if (dev == NULL || ifp->idev->dev == dev ||
1486 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1493 rcu_read_unlock_bh();
1498 /* Gets referenced address, destroys ifaddr */
1500 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1502 if (ifp->flags&IFA_F_PERMANENT) {
1503 spin_lock_bh(&ifp->lock);
1504 addrconf_del_timer(ifp);
1505 ifp->flags |= IFA_F_TENTATIVE;
1507 ifp->flags |= IFA_F_DADFAILED;
1508 spin_unlock_bh(&ifp->lock);
1510 ipv6_ifa_notify(0, ifp);
1512 #ifdef CONFIG_IPV6_PRIVACY
1513 } else if (ifp->flags&IFA_F_TEMPORARY) {
1514 struct inet6_ifaddr *ifpub;
1515 spin_lock_bh(&ifp->lock);
1518 in6_ifa_hold(ifpub);
1519 spin_unlock_bh(&ifp->lock);
1520 ipv6_create_tempaddr(ifpub, ifp);
1523 spin_unlock_bh(&ifp->lock);
1531 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1535 spin_lock(&ifp->state_lock);
1536 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1537 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1540 spin_unlock(&ifp->state_lock);
1545 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1547 struct inet6_dev *idev = ifp->idev;
1549 if (addrconf_dad_end(ifp)) {
1554 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1555 ifp->idev->dev->name, &ifp->addr);
1557 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1558 struct in6_addr addr;
1560 addr.s6_addr32[0] = htonl(0xfe800000);
1561 addr.s6_addr32[1] = 0;
1563 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1564 ipv6_addr_equal(&ifp->addr, &addr)) {
1565 /* DAD failed for link-local based on MAC address */
1566 idev->cnf.disable_ipv6 = 1;
1568 pr_info("%s: IPv6 being disabled!\n",
1569 ifp->idev->dev->name);
1573 addrconf_dad_stop(ifp, 1);
1576 /* Join to solicited addr multicast group. */
1578 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1580 struct in6_addr maddr;
1582 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1585 addrconf_addr_solict_mult(addr, &maddr);
1586 ipv6_dev_mc_inc(dev, &maddr);
1589 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1591 struct in6_addr maddr;
1593 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1596 addrconf_addr_solict_mult(addr, &maddr);
1597 __ipv6_dev_mc_dec(idev, &maddr);
1600 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1602 struct in6_addr addr;
1603 if (ifp->prefix_len == 127) /* RFC 6164 */
1605 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1606 if (ipv6_addr_any(&addr))
1608 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1611 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1613 struct in6_addr addr;
1614 if (ifp->prefix_len == 127) /* RFC 6164 */
1616 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1617 if (ipv6_addr_any(&addr))
1619 __ipv6_dev_ac_dec(ifp->idev, &addr);
1622 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1624 if (dev->addr_len != ETH_ALEN)
1626 memcpy(eui, dev->dev_addr, 3);
1627 memcpy(eui + 5, dev->dev_addr + 3, 3);
1630 * The zSeries OSA network cards can be shared among various
1631 * OS instances, but the OSA cards have only one MAC address.
1632 * This leads to duplicate address conflicts in conjunction
1633 * with IPv6 if more than one instance uses the same card.
1635 * The driver for these cards can deliver a unique 16-bit
1636 * identifier for each instance sharing the same card. It is
1637 * placed instead of 0xFFFE in the interface identifier. The
1638 * "u" bit of the interface identifier is not inverted in this
1639 * case. Hence the resulting interface identifier has local
1640 * scope according to RFC2373.
1643 eui[3] = (dev->dev_id >> 8) & 0xFF;
1644 eui[4] = dev->dev_id & 0xFF;
1653 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1655 if (dev->addr_len != IEEE802154_ADDR_LEN)
1657 memcpy(eui, dev->dev_addr, 8);
1661 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1663 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1664 if (dev->addr_len != ARCNET_ALEN)
1667 eui[7] = *(u8 *)dev->dev_addr;
1671 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1673 if (dev->addr_len != INFINIBAND_ALEN)
1675 memcpy(eui, dev->dev_addr + 12, 8);
1680 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1684 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1685 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1686 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1687 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1688 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1689 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1693 memcpy(eui + 4, &addr, 4);
1697 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1699 if (dev->priv_flags & IFF_ISATAP)
1700 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1704 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1706 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1709 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1711 switch (dev->type) {
1714 return addrconf_ifid_eui48(eui, dev);
1716 return addrconf_ifid_arcnet(eui, dev);
1717 case ARPHRD_INFINIBAND:
1718 return addrconf_ifid_infiniband(eui, dev);
1720 return addrconf_ifid_sit(eui, dev);
1722 return addrconf_ifid_gre(eui, dev);
1723 case ARPHRD_IEEE802154:
1724 return addrconf_ifid_eui64(eui, dev);
1729 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1732 struct inet6_ifaddr *ifp;
1734 read_lock_bh(&idev->lock);
1735 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1736 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1737 memcpy(eui, ifp->addr.s6_addr+8, 8);
1742 read_unlock_bh(&idev->lock);
1746 #ifdef CONFIG_IPV6_PRIVACY
1747 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1748 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1751 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1752 idev->rndid[0] &= ~0x02;
1755 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1756 * check if generated address is not inappropriate
1758 * - Reserved subnet anycast (RFC 2526)
1759 * 11111101 11....11 1xxxxxxx
1760 * - ISATAP (RFC4214) 6.1
1761 * 00-00-5E-FE-xx-xx-xx-xx
1763 * - XXX: already assigned to an address on the device
1765 if (idev->rndid[0] == 0xfd &&
1766 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1767 (idev->rndid[7]&0x80))
1769 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1770 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1772 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1777 static void ipv6_regen_rndid(unsigned long data)
1779 struct inet6_dev *idev = (struct inet6_dev *) data;
1780 unsigned long expires;
1783 write_lock_bh(&idev->lock);
1788 __ipv6_regen_rndid(idev);
1791 idev->cnf.temp_prefered_lft * HZ -
1792 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1793 idev->cnf.max_desync_factor * HZ;
1794 if (time_before(expires, jiffies)) {
1795 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1796 __func__, idev->dev->name);
1800 if (!mod_timer(&idev->regen_timer, expires))
1804 write_unlock_bh(&idev->lock);
1805 rcu_read_unlock_bh();
1809 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1811 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1812 __ipv6_regen_rndid(idev);
1821 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1822 unsigned long expires, u32 flags)
1824 struct fib6_config cfg = {
1825 .fc_table = RT6_TABLE_PREFIX,
1826 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1827 .fc_ifindex = dev->ifindex,
1828 .fc_expires = expires,
1830 .fc_flags = RTF_UP | flags,
1831 .fc_nlinfo.nl_net = dev_net(dev),
1832 .fc_protocol = RTPROT_KERNEL,
1837 /* Prevent useless cloning on PtP SIT.
1838 This thing is done here expecting that the whole
1839 class of non-broadcast devices need not cloning.
1841 #if IS_ENABLED(CONFIG_IPV6_SIT)
1842 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1843 cfg.fc_flags |= RTF_NONEXTHOP;
1846 ip6_route_add(&cfg);
1850 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1852 const struct net_device *dev,
1853 u32 flags, u32 noflags)
1855 struct fib6_node *fn;
1856 struct rt6_info *rt = NULL;
1857 struct fib6_table *table;
1859 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1863 read_lock_bh(&table->tb6_lock);
1864 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1867 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1868 if (rt->dst.dev->ifindex != dev->ifindex)
1870 if ((rt->rt6i_flags & flags) != flags)
1872 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1878 read_unlock_bh(&table->tb6_lock);
1883 /* Create "default" multicast route to the interface */
1885 static void addrconf_add_mroute(struct net_device *dev)
1887 struct fib6_config cfg = {
1888 .fc_table = RT6_TABLE_LOCAL,
1889 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1890 .fc_ifindex = dev->ifindex,
1893 .fc_nlinfo.nl_net = dev_net(dev),
1896 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1898 ip6_route_add(&cfg);
1901 #if IS_ENABLED(CONFIG_IPV6_SIT)
1902 static void sit_route_add(struct net_device *dev)
1904 struct fib6_config cfg = {
1905 .fc_table = RT6_TABLE_MAIN,
1906 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1907 .fc_ifindex = dev->ifindex,
1909 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1910 .fc_nlinfo.nl_net = dev_net(dev),
1913 /* prefix length - 96 bits "::d.d.d.d" */
1914 ip6_route_add(&cfg);
1918 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1920 struct inet6_dev *idev;
1924 idev = ipv6_find_idev(dev);
1926 return ERR_PTR(-ENOBUFS);
1928 if (idev->cnf.disable_ipv6)
1929 return ERR_PTR(-EACCES);
1931 /* Add default multicast route */
1932 if (!(dev->flags & IFF_LOOPBACK))
1933 addrconf_add_mroute(dev);
1938 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1940 struct prefix_info *pinfo;
1944 struct inet6_dev *in6_dev;
1945 struct net *net = dev_net(dev);
1947 pinfo = (struct prefix_info *) opt;
1949 if (len < sizeof(struct prefix_info)) {
1950 ADBG(("addrconf: prefix option too short\n"));
1955 * Validation checks ([ADDRCONF], page 19)
1958 addr_type = ipv6_addr_type(&pinfo->prefix);
1960 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1963 valid_lft = ntohl(pinfo->valid);
1964 prefered_lft = ntohl(pinfo->prefered);
1966 if (prefered_lft > valid_lft) {
1967 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1971 in6_dev = in6_dev_get(dev);
1973 if (in6_dev == NULL) {
1974 net_dbg_ratelimited("addrconf: device %s not configured\n",
1980 * Two things going on here:
1981 * 1) Add routes for on-link prefixes
1982 * 2) Configure prefixes with the auto flag set
1985 if (pinfo->onlink) {
1986 struct rt6_info *rt;
1987 unsigned long rt_expires;
1989 /* Avoid arithmetic overflow. Really, we could
1990 * save rt_expires in seconds, likely valid_lft,
1991 * but it would require division in fib gc, that it
1995 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1997 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1999 if (addrconf_finite_timeout(rt_expires))
2002 rt = addrconf_get_prefix_route(&pinfo->prefix,
2005 RTF_ADDRCONF | RTF_PREFIX_RT,
2006 RTF_GATEWAY | RTF_DEFAULT);
2009 /* Autoconf prefix route */
2010 if (valid_lft == 0) {
2013 } else if (addrconf_finite_timeout(rt_expires)) {
2015 rt6_set_expires(rt, jiffies + rt_expires);
2017 rt6_clean_expires(rt);
2019 } else if (valid_lft) {
2020 clock_t expires = 0;
2021 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2022 if (addrconf_finite_timeout(rt_expires)) {
2024 flags |= RTF_EXPIRES;
2025 expires = jiffies_to_clock_t(rt_expires);
2027 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2028 dev, expires, flags);
2033 /* Try to figure out our local address for this prefix */
2035 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2036 struct inet6_ifaddr *ifp;
2037 struct in6_addr addr;
2038 int create = 0, update_lft = 0;
2040 if (pinfo->prefix_len == 64) {
2041 memcpy(&addr, &pinfo->prefix, 8);
2042 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2043 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2044 in6_dev_put(in6_dev);
2049 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2051 in6_dev_put(in6_dev);
2056 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2058 if (ifp == NULL && valid_lft) {
2059 int max_addresses = in6_dev->cnf.max_addresses;
2062 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2063 if (in6_dev->cnf.optimistic_dad &&
2064 !net->ipv6.devconf_all->forwarding && sllao)
2065 addr_flags = IFA_F_OPTIMISTIC;
2068 /* Do not allow to create too much of autoconfigured
2069 * addresses; this would be too easy way to crash kernel.
2071 if (!max_addresses ||
2072 ipv6_count_addresses(in6_dev) < max_addresses)
2073 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2074 addr_type&IPV6_ADDR_SCOPE_MASK,
2077 if (!ifp || IS_ERR(ifp)) {
2078 in6_dev_put(in6_dev);
2082 update_lft = create = 1;
2083 ifp->cstamp = jiffies;
2084 addrconf_dad_start(ifp);
2090 #ifdef CONFIG_IPV6_PRIVACY
2091 struct inet6_ifaddr *ift;
2095 /* update lifetime (RFC2462 5.5.3 e) */
2096 spin_lock(&ifp->lock);
2098 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2099 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2102 if (!update_lft && stored_lft) {
2103 if (valid_lft > MIN_VALID_LIFETIME ||
2104 valid_lft > stored_lft)
2106 else if (stored_lft <= MIN_VALID_LIFETIME) {
2107 /* valid_lft <= stored_lft is always true */
2109 * RFC 4862 Section 5.5.3e:
2110 * "Note that the preferred lifetime of
2111 * the corresponding address is always
2112 * reset to the Preferred Lifetime in
2113 * the received Prefix Information
2114 * option, regardless of whether the
2115 * valid lifetime is also reset or
2118 * So if the preferred lifetime in
2119 * this advertisement is different
2120 * than what we have stored, but the
2121 * valid lifetime is invalid, just
2122 * reset prefered_lft.
2124 * We must set the valid lifetime
2125 * to the stored lifetime since we'll
2126 * be updating the timestamp below,
2127 * else we'll set it back to the
2130 if (prefered_lft != ifp->prefered_lft) {
2131 valid_lft = stored_lft;
2135 valid_lft = MIN_VALID_LIFETIME;
2136 if (valid_lft < prefered_lft)
2137 prefered_lft = valid_lft;
2143 ifp->valid_lft = valid_lft;
2144 ifp->prefered_lft = prefered_lft;
2147 ifp->flags &= ~IFA_F_DEPRECATED;
2148 spin_unlock(&ifp->lock);
2150 if (!(flags&IFA_F_TENTATIVE))
2151 ipv6_ifa_notify(0, ifp);
2153 spin_unlock(&ifp->lock);
2155 #ifdef CONFIG_IPV6_PRIVACY
2156 read_lock_bh(&in6_dev->lock);
2157 /* update all temporary addresses in the list */
2158 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2160 int age, max_valid, max_prefered;
2162 if (ifp != ift->ifpub)
2166 * RFC 4941 section 3.3:
2167 * If a received option will extend the lifetime
2168 * of a public address, the lifetimes of
2169 * temporary addresses should be extended,
2170 * subject to the overall constraint that no
2171 * temporary addresses should ever remain
2172 * "valid" or "preferred" for a time longer than
2173 * (TEMP_VALID_LIFETIME) or
2174 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2177 age = (now - ift->cstamp) / HZ;
2178 max_valid = in6_dev->cnf.temp_valid_lft - age;
2182 max_prefered = in6_dev->cnf.temp_prefered_lft -
2183 in6_dev->cnf.max_desync_factor -
2185 if (max_prefered < 0)
2188 if (valid_lft > max_valid)
2189 valid_lft = max_valid;
2191 if (prefered_lft > max_prefered)
2192 prefered_lft = max_prefered;
2194 spin_lock(&ift->lock);
2196 ift->valid_lft = valid_lft;
2197 ift->prefered_lft = prefered_lft;
2199 if (prefered_lft > 0)
2200 ift->flags &= ~IFA_F_DEPRECATED;
2202 spin_unlock(&ift->lock);
2203 if (!(flags&IFA_F_TENTATIVE))
2204 ipv6_ifa_notify(0, ift);
2207 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2209 * When a new public address is created as
2210 * described in [ADDRCONF], also create a new
2211 * temporary address. Also create a temporary
2212 * address if it's enabled but no temporary
2213 * address currently exists.
2215 read_unlock_bh(&in6_dev->lock);
2216 ipv6_create_tempaddr(ifp, NULL);
2218 read_unlock_bh(&in6_dev->lock);
2225 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2226 in6_dev_put(in6_dev);
2230 * Set destination address.
2231 * Special case for SIT interfaces where we create a new "virtual"
2234 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2236 struct in6_ifreq ireq;
2237 struct net_device *dev;
2243 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2246 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2252 #if IS_ENABLED(CONFIG_IPV6_SIT)
2253 if (dev->type == ARPHRD_SIT) {
2254 const struct net_device_ops *ops = dev->netdev_ops;
2256 struct ip_tunnel_parm p;
2258 err = -EADDRNOTAVAIL;
2259 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2262 memset(&p, 0, sizeof(p));
2263 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2267 p.iph.protocol = IPPROTO_IPV6;
2269 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2271 if (ops->ndo_do_ioctl) {
2272 mm_segment_t oldfs = get_fs();
2275 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2282 dev = __dev_get_by_name(net, p.name);
2285 err = dev_open(dev);
2296 * Manual configuration of address on an interface
2298 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2299 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2302 struct inet6_ifaddr *ifp;
2303 struct inet6_dev *idev;
2304 struct net_device *dev;
2308 unsigned long timeout;
2315 /* check the lifetime */
2316 if (!valid_lft || prefered_lft > valid_lft)
2319 dev = __dev_get_by_index(net, ifindex);
2323 idev = addrconf_add_dev(dev);
2325 return PTR_ERR(idev);
2327 scope = ipv6_addr_scope(pfx);
2329 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2330 if (addrconf_finite_timeout(timeout)) {
2331 expires = jiffies_to_clock_t(timeout * HZ);
2332 valid_lft = timeout;
2333 flags = RTF_EXPIRES;
2337 ifa_flags |= IFA_F_PERMANENT;
2340 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2341 if (addrconf_finite_timeout(timeout)) {
2343 ifa_flags |= IFA_F_DEPRECATED;
2344 prefered_lft = timeout;
2347 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2350 spin_lock_bh(&ifp->lock);
2351 ifp->valid_lft = valid_lft;
2352 ifp->prefered_lft = prefered_lft;
2353 ifp->tstamp = jiffies;
2354 spin_unlock_bh(&ifp->lock);
2356 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2359 * Note that section 3.1 of RFC 4429 indicates
2360 * that the Optimistic flag should not be set for
2361 * manually configured addresses
2363 addrconf_dad_start(ifp);
2369 return PTR_ERR(ifp);
2372 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2375 struct inet6_ifaddr *ifp;
2376 struct inet6_dev *idev;
2377 struct net_device *dev;
2382 dev = __dev_get_by_index(net, ifindex);
2386 if ((idev = __in6_dev_get(dev)) == NULL)
2389 read_lock_bh(&idev->lock);
2390 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2391 if (ifp->prefix_len == plen &&
2392 ipv6_addr_equal(pfx, &ifp->addr)) {
2394 read_unlock_bh(&idev->lock);
2398 /* If the last address is deleted administratively,
2399 disable IPv6 on this interface.
2401 if (list_empty(&idev->addr_list))
2402 addrconf_ifdown(idev->dev, 1);
2406 read_unlock_bh(&idev->lock);
2407 return -EADDRNOTAVAIL;
2411 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2413 struct in6_ifreq ireq;
2416 if (!capable(CAP_NET_ADMIN))
2419 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2423 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2424 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2425 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2430 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2432 struct in6_ifreq ireq;
2435 if (!capable(CAP_NET_ADMIN))
2438 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2442 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2443 ireq.ifr6_prefixlen);
2448 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2449 int plen, int scope)
2451 struct inet6_ifaddr *ifp;
2453 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2455 spin_lock_bh(&ifp->lock);
2456 ifp->flags &= ~IFA_F_TENTATIVE;
2457 spin_unlock_bh(&ifp->lock);
2458 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2463 #if IS_ENABLED(CONFIG_IPV6_SIT)
2464 static void sit_add_v4_addrs(struct inet6_dev *idev)
2466 struct in6_addr addr;
2467 struct net_device *dev;
2468 struct net *net = dev_net(idev->dev);
2473 memset(&addr, 0, sizeof(struct in6_addr));
2474 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2476 if (idev->dev->flags&IFF_POINTOPOINT) {
2477 addr.s6_addr32[0] = htonl(0xfe800000);
2480 scope = IPV6_ADDR_COMPATv4;
2483 if (addr.s6_addr32[3]) {
2484 add_addr(idev, &addr, 128, scope);
2488 for_each_netdev(net, dev) {
2489 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2490 if (in_dev && (dev->flags & IFF_UP)) {
2491 struct in_ifaddr *ifa;
2495 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2498 addr.s6_addr32[3] = ifa->ifa_local;
2500 if (ifa->ifa_scope == RT_SCOPE_LINK)
2502 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2503 if (idev->dev->flags&IFF_POINTOPOINT)
2507 if (idev->dev->flags&IFF_POINTOPOINT)
2512 add_addr(idev, &addr, plen, flag);
2519 static void init_loopback(struct net_device *dev)
2521 struct inet6_dev *idev;
2527 if ((idev = ipv6_find_idev(dev)) == NULL) {
2528 pr_debug("%s: add_dev failed\n", __func__);
2532 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2535 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2537 struct inet6_ifaddr *ifp;
2538 u32 addr_flags = IFA_F_PERMANENT;
2540 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2541 if (idev->cnf.optimistic_dad &&
2542 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2543 addr_flags |= IFA_F_OPTIMISTIC;
2547 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2549 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2550 addrconf_dad_start(ifp);
2555 static void addrconf_dev_config(struct net_device *dev)
2557 struct in6_addr addr;
2558 struct inet6_dev *idev;
2562 if ((dev->type != ARPHRD_ETHER) &&
2563 (dev->type != ARPHRD_FDDI) &&
2564 (dev->type != ARPHRD_ARCNET) &&
2565 (dev->type != ARPHRD_INFINIBAND) &&
2566 (dev->type != ARPHRD_IEEE802154)) {
2567 /* Alas, we support only Ethernet autoconfiguration. */
2571 idev = addrconf_add_dev(dev);
2575 memset(&addr, 0, sizeof(struct in6_addr));
2576 addr.s6_addr32[0] = htonl(0xFE800000);
2578 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2579 addrconf_add_linklocal(idev, &addr);
2582 #if IS_ENABLED(CONFIG_IPV6_SIT)
2583 static void addrconf_sit_config(struct net_device *dev)
2585 struct inet6_dev *idev;
2590 * Configure the tunnel with one of our IPv4
2591 * addresses... we should configure all of
2592 * our v4 addrs in the tunnel
2595 if ((idev = ipv6_find_idev(dev)) == NULL) {
2596 pr_debug("%s: add_dev failed\n", __func__);
2600 if (dev->priv_flags & IFF_ISATAP) {
2601 struct in6_addr addr;
2603 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2604 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2605 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2606 addrconf_add_linklocal(idev, &addr);
2610 sit_add_v4_addrs(idev);
2612 if (dev->flags&IFF_POINTOPOINT)
2613 addrconf_add_mroute(dev);
2619 #if IS_ENABLED(CONFIG_NET_IPGRE)
2620 static void addrconf_gre_config(struct net_device *dev)
2622 struct inet6_dev *idev;
2623 struct in6_addr addr;
2625 pr_info("%s(%s)\n", __func__, dev->name);
2629 if ((idev = ipv6_find_idev(dev)) == NULL) {
2630 pr_debug("%s: add_dev failed\n", __func__);
2634 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2635 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2637 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2638 addrconf_add_linklocal(idev, &addr);
2643 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2645 struct in6_addr lladdr;
2647 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2648 addrconf_add_linklocal(idev, &lladdr);
2654 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2656 struct net_device *link_dev;
2657 struct net *net = dev_net(idev->dev);
2659 /* first try to inherit the link-local address from the link device */
2660 if (idev->dev->iflink &&
2661 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2662 if (!ipv6_inherit_linklocal(idev, link_dev))
2665 /* then try to inherit it from any device */
2666 for_each_netdev(net, link_dev) {
2667 if (!ipv6_inherit_linklocal(idev, link_dev))
2670 pr_debug("init ip6-ip6: add_linklocal failed\n");
2674 * Autoconfigure tunnel with a link-local address so routing protocols,
2675 * DHCPv6, MLD etc. can be run over the virtual link
2678 static void addrconf_ip6_tnl_config(struct net_device *dev)
2680 struct inet6_dev *idev;
2684 idev = addrconf_add_dev(dev);
2686 pr_debug("init ip6-ip6: add_dev failed\n");
2689 ip6_tnl_add_linklocal(idev);
2692 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2695 struct net_device *dev = (struct net_device *) data;
2696 struct inet6_dev *idev = __in6_dev_get(dev);
2697 int run_pending = 0;
2701 case NETDEV_REGISTER:
2702 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2703 idev = ipv6_add_dev(dev);
2705 return notifier_from_errno(-ENOMEM);
2711 if (dev->flags & IFF_SLAVE)
2714 if (event == NETDEV_UP) {
2715 if (!addrconf_qdisc_ok(dev)) {
2716 /* device is not ready yet. */
2717 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2722 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2723 idev = ipv6_add_dev(dev);
2726 idev->if_flags |= IF_READY;
2730 if (!addrconf_qdisc_ok(dev)) {
2731 /* device is still not ready. */
2736 if (idev->if_flags & IF_READY)
2737 /* device is already configured. */
2739 idev->if_flags |= IF_READY;
2742 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2748 switch (dev->type) {
2749 #if IS_ENABLED(CONFIG_IPV6_SIT)
2751 addrconf_sit_config(dev);
2754 #if IS_ENABLED(CONFIG_NET_IPGRE)
2756 addrconf_gre_config(dev);
2759 case ARPHRD_TUNNEL6:
2760 addrconf_ip6_tnl_config(dev);
2762 case ARPHRD_LOOPBACK:
2767 addrconf_dev_config(dev);
2773 addrconf_dad_run(idev);
2776 * If the MTU changed during the interface down,
2777 * when the interface up, the changed MTU must be
2778 * reflected in the idev as well as routers.
2780 if (idev->cnf.mtu6 != dev->mtu &&
2781 dev->mtu >= IPV6_MIN_MTU) {
2782 rt6_mtu_change(dev, dev->mtu);
2783 idev->cnf.mtu6 = dev->mtu;
2785 idev->tstamp = jiffies;
2786 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2789 * If the changed mtu during down is lower than
2790 * IPV6_MIN_MTU stop IPv6 on this interface.
2792 if (dev->mtu < IPV6_MIN_MTU)
2793 addrconf_ifdown(dev, 1);
2797 case NETDEV_CHANGEMTU:
2798 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2799 rt6_mtu_change(dev, dev->mtu);
2800 idev->cnf.mtu6 = dev->mtu;
2804 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2805 idev = ipv6_add_dev(dev);
2811 * MTU falled under IPV6_MIN_MTU.
2812 * Stop IPv6 on this interface.
2816 case NETDEV_UNREGISTER:
2818 * Remove all addresses from this interface.
2820 addrconf_ifdown(dev, event != NETDEV_DOWN);
2823 case NETDEV_CHANGENAME:
2825 snmp6_unregister_dev(idev);
2826 addrconf_sysctl_unregister(idev);
2827 addrconf_sysctl_register(idev);
2828 err = snmp6_register_dev(idev);
2830 return notifier_from_errno(err);
2834 case NETDEV_PRE_TYPE_CHANGE:
2835 case NETDEV_POST_TYPE_CHANGE:
2836 addrconf_type_change(dev, event);
2844 * addrconf module should be notified of a device going up
2846 static struct notifier_block ipv6_dev_notf = {
2847 .notifier_call = addrconf_notify,
2850 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2852 struct inet6_dev *idev;
2855 idev = __in6_dev_get(dev);
2857 if (event == NETDEV_POST_TYPE_CHANGE)
2858 ipv6_mc_remap(idev);
2859 else if (event == NETDEV_PRE_TYPE_CHANGE)
2860 ipv6_mc_unmap(idev);
2863 static int addrconf_ifdown(struct net_device *dev, int how)
2865 struct net *net = dev_net(dev);
2866 struct inet6_dev *idev;
2867 struct inet6_ifaddr *ifa;
2872 rt6_ifdown(net, dev);
2873 neigh_ifdown(&nd_tbl, dev);
2875 idev = __in6_dev_get(dev);
2880 * Step 1: remove reference to ipv6 device from parent device.
2886 /* protected by rtnl_lock */
2887 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2889 /* Step 1.5: remove snmp6 entry */
2890 snmp6_unregister_dev(idev);
2894 /* Step 2: clear hash table */
2895 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2896 struct hlist_head *h = &inet6_addr_lst[i];
2897 struct hlist_node *n;
2899 spin_lock_bh(&addrconf_hash_lock);
2901 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2902 if (ifa->idev == idev) {
2903 hlist_del_init_rcu(&ifa->addr_lst);
2904 addrconf_del_timer(ifa);
2908 spin_unlock_bh(&addrconf_hash_lock);
2911 write_lock_bh(&idev->lock);
2913 /* Step 2: clear flags for stateless addrconf */
2915 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2917 #ifdef CONFIG_IPV6_PRIVACY
2918 if (how && del_timer(&idev->regen_timer))
2921 /* Step 3: clear tempaddr list */
2922 while (!list_empty(&idev->tempaddr_list)) {
2923 ifa = list_first_entry(&idev->tempaddr_list,
2924 struct inet6_ifaddr, tmp_list);
2925 list_del(&ifa->tmp_list);
2926 write_unlock_bh(&idev->lock);
2927 spin_lock_bh(&ifa->lock);
2930 in6_ifa_put(ifa->ifpub);
2933 spin_unlock_bh(&ifa->lock);
2935 write_lock_bh(&idev->lock);
2939 while (!list_empty(&idev->addr_list)) {
2940 ifa = list_first_entry(&idev->addr_list,
2941 struct inet6_ifaddr, if_list);
2942 addrconf_del_timer(ifa);
2944 list_del(&ifa->if_list);
2946 write_unlock_bh(&idev->lock);
2948 spin_lock_bh(&ifa->state_lock);
2950 ifa->state = INET6_IFADDR_STATE_DEAD;
2951 spin_unlock_bh(&ifa->state_lock);
2953 if (state != INET6_IFADDR_STATE_DEAD) {
2954 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2955 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2959 write_lock_bh(&idev->lock);
2962 write_unlock_bh(&idev->lock);
2964 /* Step 5: Discard multicast list */
2966 ipv6_mc_destroy_dev(idev);
2970 idev->tstamp = jiffies;
2972 /* Last: Shot the device (if unregistered) */
2974 addrconf_sysctl_unregister(idev);
2975 neigh_parms_release(&nd_tbl, idev->nd_parms);
2976 neigh_ifdown(&nd_tbl, dev);
2982 static void addrconf_rs_timer(unsigned long data)
2984 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2985 struct inet6_dev *idev = ifp->idev;
2987 read_lock(&idev->lock);
2988 if (idev->dead || !(idev->if_flags & IF_READY))
2991 if (idev->cnf.forwarding)
2994 /* Announcement received after solicitation was sent */
2995 if (idev->if_flags & IF_RA_RCVD)
2998 spin_lock(&ifp->lock);
2999 if (ifp->probes++ < idev->cnf.rtr_solicits) {
3000 /* The wait after the last probe can be shorter */
3001 addrconf_mod_timer(ifp, AC_RS,
3002 (ifp->probes == idev->cnf.rtr_solicits) ?
3003 idev->cnf.rtr_solicit_delay :
3004 idev->cnf.rtr_solicit_interval);
3005 spin_unlock(&ifp->lock);
3007 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3009 spin_unlock(&ifp->lock);
3011 * Note: we do not support deprecated "all on-link"
3012 * assumption any longer.
3014 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3018 read_unlock(&idev->lock);
3023 * Duplicate Address Detection
3025 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3027 unsigned long rand_num;
3028 struct inet6_dev *idev = ifp->idev;
3030 if (ifp->flags & IFA_F_OPTIMISTIC)
3033 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3035 ifp->probes = idev->cnf.dad_transmits;
3036 addrconf_mod_timer(ifp, AC_DAD, rand_num);
3039 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3041 struct inet6_dev *idev = ifp->idev;
3042 struct net_device *dev = idev->dev;
3044 addrconf_join_solict(dev, &ifp->addr);
3046 net_srandom(ifp->addr.s6_addr32[3]);
3048 read_lock_bh(&idev->lock);
3049 spin_lock(&ifp->lock);
3050 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3053 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3054 idev->cnf.accept_dad < 1 ||
3055 !(ifp->flags&IFA_F_TENTATIVE) ||
3056 ifp->flags & IFA_F_NODAD) {
3057 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3058 spin_unlock(&ifp->lock);
3059 read_unlock_bh(&idev->lock);
3061 addrconf_dad_completed(ifp);
3065 if (!(idev->if_flags & IF_READY)) {
3066 spin_unlock(&ifp->lock);
3067 read_unlock_bh(&idev->lock);
3069 * If the device is not ready:
3070 * - keep it tentative if it is a permanent address.
3071 * - otherwise, kill it.
3074 addrconf_dad_stop(ifp, 0);
3079 * Optimistic nodes can start receiving
3082 if (ifp->flags & IFA_F_OPTIMISTIC)
3083 ip6_ins_rt(ifp->rt);
3085 addrconf_dad_kick(ifp);
3087 spin_unlock(&ifp->lock);
3088 read_unlock_bh(&idev->lock);
3091 static void addrconf_dad_timer(unsigned long data)
3093 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3094 struct inet6_dev *idev = ifp->idev;
3095 struct in6_addr mcaddr;
3097 if (!ifp->probes && addrconf_dad_end(ifp))
3100 read_lock(&idev->lock);
3101 if (idev->dead || !(idev->if_flags & IF_READY)) {
3102 read_unlock(&idev->lock);
3106 spin_lock(&ifp->lock);
3107 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3108 spin_unlock(&ifp->lock);
3109 read_unlock(&idev->lock);
3113 if (ifp->probes == 0) {
3115 * DAD was successful
3118 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3119 spin_unlock(&ifp->lock);
3120 read_unlock(&idev->lock);
3122 addrconf_dad_completed(ifp);
3128 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3129 spin_unlock(&ifp->lock);
3130 read_unlock(&idev->lock);
3132 /* send a neighbour solicitation for our addr */
3133 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3134 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3139 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3141 struct net_device *dev = ifp->idev->dev;
3144 * Configure the address for reception. Now it is valid.
3147 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3149 /* If added prefix is link local and we are prepared to process
3150 router advertisements, start sending router solicitations.
3153 if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3154 ifp->idev->cnf.accept_ra == 2) &&
3155 ifp->idev->cnf.rtr_solicits > 0 &&
3156 (dev->flags&IFF_LOOPBACK) == 0 &&
3157 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3159 * If a host as already performed a random delay
3160 * [...] as part of DAD [...] there is no need
3161 * to delay again before sending the first RS
3163 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3165 spin_lock_bh(&ifp->lock);
3167 ifp->idev->if_flags |= IF_RS_SENT;
3168 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3169 spin_unlock_bh(&ifp->lock);
3173 static void addrconf_dad_run(struct inet6_dev *idev)
3175 struct inet6_ifaddr *ifp;
3177 read_lock_bh(&idev->lock);
3178 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3179 spin_lock(&ifp->lock);
3180 if (ifp->flags & IFA_F_TENTATIVE &&
3181 ifp->state == INET6_IFADDR_STATE_DAD)
3182 addrconf_dad_kick(ifp);
3183 spin_unlock(&ifp->lock);
3185 read_unlock_bh(&idev->lock);
3188 #ifdef CONFIG_PROC_FS
3189 struct if6_iter_state {
3190 struct seq_net_private p;
3195 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3197 struct inet6_ifaddr *ifa = NULL;
3198 struct if6_iter_state *state = seq->private;
3199 struct net *net = seq_file_net(seq);
3202 /* initial bucket if pos is 0 */
3208 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3209 struct hlist_node *n;
3210 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3212 if (!net_eq(dev_net(ifa->idev->dev), net))
3214 /* sync with offset */
3215 if (p < state->offset) {
3223 /* prepare for next bucket */
3230 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3231 struct inet6_ifaddr *ifa)
3233 struct if6_iter_state *state = seq->private;
3234 struct net *net = seq_file_net(seq);
3235 struct hlist_node *n = &ifa->addr_lst;
3237 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3238 if (!net_eq(dev_net(ifa->idev->dev), net))
3244 while (++state->bucket < IN6_ADDR_HSIZE) {
3246 hlist_for_each_entry_rcu_bh(ifa, n,
3247 &inet6_addr_lst[state->bucket], addr_lst) {
3248 if (!net_eq(dev_net(ifa->idev->dev), net))
3258 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3262 return if6_get_first(seq, *pos);
3265 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3267 struct inet6_ifaddr *ifa;
3269 ifa = if6_get_next(seq, v);
3274 static void if6_seq_stop(struct seq_file *seq, void *v)
3277 rcu_read_unlock_bh();
3280 static int if6_seq_show(struct seq_file *seq, void *v)
3282 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3283 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3285 ifp->idev->dev->ifindex,
3289 ifp->idev->dev->name);
3293 static const struct seq_operations if6_seq_ops = {
3294 .start = if6_seq_start,
3295 .next = if6_seq_next,
3296 .show = if6_seq_show,
3297 .stop = if6_seq_stop,
3300 static int if6_seq_open(struct inode *inode, struct file *file)
3302 return seq_open_net(inode, file, &if6_seq_ops,
3303 sizeof(struct if6_iter_state));
3306 static const struct file_operations if6_fops = {
3307 .owner = THIS_MODULE,
3308 .open = if6_seq_open,
3310 .llseek = seq_lseek,
3311 .release = seq_release_net,
3314 static int __net_init if6_proc_net_init(struct net *net)
3316 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3321 static void __net_exit if6_proc_net_exit(struct net *net)
3323 proc_net_remove(net, "if_inet6");
3326 static struct pernet_operations if6_proc_net_ops = {
3327 .init = if6_proc_net_init,
3328 .exit = if6_proc_net_exit,
3331 int __init if6_proc_init(void)
3333 return register_pernet_subsys(&if6_proc_net_ops);
3336 void if6_proc_exit(void)
3338 unregister_pernet_subsys(&if6_proc_net_ops);
3340 #endif /* CONFIG_PROC_FS */
3342 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3343 /* Check if address is a home address configured on any interface. */
3344 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3347 struct inet6_ifaddr *ifp = NULL;
3348 struct hlist_node *n;
3349 unsigned int hash = inet6_addr_hash(addr);
3352 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3353 if (!net_eq(dev_net(ifp->idev->dev), net))
3355 if (ipv6_addr_equal(&ifp->addr, addr) &&
3356 (ifp->flags & IFA_F_HOMEADDRESS)) {
3361 rcu_read_unlock_bh();
3367 * Periodic address status verification
3370 static void addrconf_verify(unsigned long foo)
3372 unsigned long now, next, next_sec, next_sched;
3373 struct inet6_ifaddr *ifp;
3374 struct hlist_node *node;
3378 spin_lock(&addrconf_verify_lock);
3380 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3382 del_timer(&addr_chk_timer);
3384 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3386 hlist_for_each_entry_rcu_bh(ifp, node,
3387 &inet6_addr_lst[i], addr_lst) {
3390 if (ifp->flags & IFA_F_PERMANENT)
3393 spin_lock(&ifp->lock);
3394 /* We try to batch several events at once. */
3395 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3397 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3398 age >= ifp->valid_lft) {
3399 spin_unlock(&ifp->lock);
3403 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3404 spin_unlock(&ifp->lock);
3406 } else if (age >= ifp->prefered_lft) {
3407 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3410 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3412 ifp->flags |= IFA_F_DEPRECATED;
3415 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3416 next = ifp->tstamp + ifp->valid_lft * HZ;
3418 spin_unlock(&ifp->lock);
3423 ipv6_ifa_notify(0, ifp);
3427 #ifdef CONFIG_IPV6_PRIVACY
3428 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3429 !(ifp->flags&IFA_F_TENTATIVE)) {
3430 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3431 ifp->idev->cnf.dad_transmits *
3432 ifp->idev->nd_parms->retrans_time / HZ;
3434 if (age >= ifp->prefered_lft - regen_advance) {
3435 struct inet6_ifaddr *ifpub = ifp->ifpub;
3436 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3437 next = ifp->tstamp + ifp->prefered_lft * HZ;
3438 if (!ifp->regen_count && ifpub) {
3441 in6_ifa_hold(ifpub);
3442 spin_unlock(&ifp->lock);
3444 spin_lock(&ifpub->lock);
3445 ifpub->regen_count = 0;
3446 spin_unlock(&ifpub->lock);
3447 ipv6_create_tempaddr(ifpub, ifp);
3452 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3453 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3454 spin_unlock(&ifp->lock);
3457 /* ifp->prefered_lft <= ifp->valid_lft */
3458 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3459 next = ifp->tstamp + ifp->prefered_lft * HZ;
3460 spin_unlock(&ifp->lock);
3465 next_sec = round_jiffies_up(next);
3468 /* If rounded timeout is accurate enough, accept it. */
3469 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3470 next_sched = next_sec;
3472 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3473 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3474 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3476 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3477 now, next, next_sec, next_sched));
3479 addr_chk_timer.expires = next_sched;
3480 add_timer(&addr_chk_timer);
3481 spin_unlock(&addrconf_verify_lock);
3482 rcu_read_unlock_bh();
3485 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3487 struct in6_addr *pfx = NULL;
3490 pfx = nla_data(addr);
3493 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3496 pfx = nla_data(local);
3502 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3503 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3504 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3505 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3509 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3511 struct net *net = sock_net(skb->sk);
3512 struct ifaddrmsg *ifm;
3513 struct nlattr *tb[IFA_MAX+1];
3514 struct in6_addr *pfx;
3517 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3521 ifm = nlmsg_data(nlh);
3522 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3526 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3529 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3530 u32 prefered_lft, u32 valid_lft)
3534 unsigned long timeout;
3536 if (!valid_lft || (prefered_lft > valid_lft))
3539 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3540 if (addrconf_finite_timeout(timeout)) {
3541 expires = jiffies_to_clock_t(timeout * HZ);
3542 valid_lft = timeout;
3543 flags = RTF_EXPIRES;
3547 ifa_flags |= IFA_F_PERMANENT;
3550 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3551 if (addrconf_finite_timeout(timeout)) {
3553 ifa_flags |= IFA_F_DEPRECATED;
3554 prefered_lft = timeout;
3557 spin_lock_bh(&ifp->lock);
3558 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3559 ifp->tstamp = jiffies;
3560 ifp->valid_lft = valid_lft;
3561 ifp->prefered_lft = prefered_lft;
3563 spin_unlock_bh(&ifp->lock);
3564 if (!(ifp->flags&IFA_F_TENTATIVE))
3565 ipv6_ifa_notify(0, ifp);
3567 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3575 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3577 struct net *net = sock_net(skb->sk);
3578 struct ifaddrmsg *ifm;
3579 struct nlattr *tb[IFA_MAX+1];
3580 struct in6_addr *pfx;
3581 struct inet6_ifaddr *ifa;
3582 struct net_device *dev;
3583 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3587 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3591 ifm = nlmsg_data(nlh);
3592 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3596 if (tb[IFA_CACHEINFO]) {
3597 struct ifa_cacheinfo *ci;
3599 ci = nla_data(tb[IFA_CACHEINFO]);
3600 valid_lft = ci->ifa_valid;
3601 preferred_lft = ci->ifa_prefered;
3603 preferred_lft = INFINITY_LIFE_TIME;
3604 valid_lft = INFINITY_LIFE_TIME;
3607 dev = __dev_get_by_index(net, ifm->ifa_index);
3611 /* We ignore other flags so far. */
3612 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3614 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3617 * It would be best to check for !NLM_F_CREATE here but
3618 * userspace alreay relies on not having to provide this.
3620 return inet6_addr_add(net, ifm->ifa_index, pfx,
3621 ifm->ifa_prefixlen, ifa_flags,
3622 preferred_lft, valid_lft);
3625 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3626 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3629 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3636 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3637 u8 scope, int ifindex)
3639 struct ifaddrmsg *ifm;
3641 ifm = nlmsg_data(nlh);
3642 ifm->ifa_family = AF_INET6;
3643 ifm->ifa_prefixlen = prefixlen;
3644 ifm->ifa_flags = flags;
3645 ifm->ifa_scope = scope;
3646 ifm->ifa_index = ifindex;
3649 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3650 unsigned long tstamp, u32 preferred, u32 valid)
3652 struct ifa_cacheinfo ci;
3654 ci.cstamp = cstamp_delta(cstamp);
3655 ci.tstamp = cstamp_delta(tstamp);
3656 ci.ifa_prefered = preferred;
3657 ci.ifa_valid = valid;
3659 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3662 static inline int rt_scope(int ifa_scope)
3664 if (ifa_scope & IFA_HOST)
3665 return RT_SCOPE_HOST;
3666 else if (ifa_scope & IFA_LINK)
3667 return RT_SCOPE_LINK;
3668 else if (ifa_scope & IFA_SITE)
3669 return RT_SCOPE_SITE;
3671 return RT_SCOPE_UNIVERSE;
3674 static inline int inet6_ifaddr_msgsize(void)
3676 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3677 + nla_total_size(16) /* IFA_ADDRESS */
3678 + nla_total_size(sizeof(struct ifa_cacheinfo));
3681 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3682 u32 portid, u32 seq, int event, unsigned int flags)
3684 struct nlmsghdr *nlh;
3685 u32 preferred, valid;
3687 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3691 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3692 ifa->idev->dev->ifindex);
3694 if (!(ifa->flags&IFA_F_PERMANENT)) {
3695 preferred = ifa->prefered_lft;
3696 valid = ifa->valid_lft;
3697 if (preferred != INFINITY_LIFE_TIME) {
3698 long tval = (jiffies - ifa->tstamp)/HZ;
3699 if (preferred > tval)
3703 if (valid != INFINITY_LIFE_TIME) {
3711 preferred = INFINITY_LIFE_TIME;
3712 valid = INFINITY_LIFE_TIME;
3715 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3716 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3717 nlmsg_cancel(skb, nlh);
3721 return nlmsg_end(skb, nlh);
3724 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3725 u32 portid, u32 seq, int event, u16 flags)
3727 struct nlmsghdr *nlh;
3728 u8 scope = RT_SCOPE_UNIVERSE;
3729 int ifindex = ifmca->idev->dev->ifindex;
3731 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3732 scope = RT_SCOPE_SITE;
3734 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3738 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3739 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3740 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3741 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3742 nlmsg_cancel(skb, nlh);
3746 return nlmsg_end(skb, nlh);
3749 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3750 u32 portid, u32 seq, int event, unsigned int flags)
3752 struct nlmsghdr *nlh;
3753 u8 scope = RT_SCOPE_UNIVERSE;
3754 int ifindex = ifaca->aca_idev->dev->ifindex;
3756 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3757 scope = RT_SCOPE_SITE;
3759 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3763 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3764 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3765 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3766 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3767 nlmsg_cancel(skb, nlh);
3771 return nlmsg_end(skb, nlh);
3780 /* called with rcu_read_lock() */
3781 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3782 struct netlink_callback *cb, enum addr_type_t type,
3783 int s_ip_idx, int *p_ip_idx)
3785 struct ifmcaddr6 *ifmca;
3786 struct ifacaddr6 *ifaca;
3788 int ip_idx = *p_ip_idx;
3790 read_lock_bh(&idev->lock);
3792 case UNICAST_ADDR: {
3793 struct inet6_ifaddr *ifa;
3795 /* unicast address incl. temp addr */
3796 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3797 if (++ip_idx < s_ip_idx)
3799 err = inet6_fill_ifaddr(skb, ifa,
3800 NETLINK_CB(cb->skb).portid,
3809 case MULTICAST_ADDR:
3810 /* multicast address */
3811 for (ifmca = idev->mc_list; ifmca;
3812 ifmca = ifmca->next, ip_idx++) {
3813 if (ip_idx < s_ip_idx)
3815 err = inet6_fill_ifmcaddr(skb, ifmca,
3816 NETLINK_CB(cb->skb).portid,
3825 /* anycast address */
3826 for (ifaca = idev->ac_list; ifaca;
3827 ifaca = ifaca->aca_next, ip_idx++) {
3828 if (ip_idx < s_ip_idx)
3830 err = inet6_fill_ifacaddr(skb, ifaca,
3831 NETLINK_CB(cb->skb).portid,
3842 read_unlock_bh(&idev->lock);
3847 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3848 enum addr_type_t type)
3850 struct net *net = sock_net(skb->sk);
3853 int s_idx, s_ip_idx;
3854 struct net_device *dev;
3855 struct inet6_dev *idev;
3856 struct hlist_head *head;
3857 struct hlist_node *node;
3860 s_idx = idx = cb->args[1];
3861 s_ip_idx = ip_idx = cb->args[2];
3864 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3866 head = &net->dev_index_head[h];
3867 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3870 if (h > s_h || idx > s_idx)
3873 idev = __in6_dev_get(dev);
3877 if (in6_dump_addrs(idev, skb, cb, type,
3878 s_ip_idx, &ip_idx) <= 0)
3888 cb->args[2] = ip_idx;
3893 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3895 enum addr_type_t type = UNICAST_ADDR;
3897 return inet6_dump_addr(skb, cb, type);
3900 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3902 enum addr_type_t type = MULTICAST_ADDR;
3904 return inet6_dump_addr(skb, cb, type);
3908 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3910 enum addr_type_t type = ANYCAST_ADDR;
3912 return inet6_dump_addr(skb, cb, type);
3915 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3918 struct net *net = sock_net(in_skb->sk);
3919 struct ifaddrmsg *ifm;
3920 struct nlattr *tb[IFA_MAX+1];
3921 struct in6_addr *addr = NULL;
3922 struct net_device *dev = NULL;
3923 struct inet6_ifaddr *ifa;
3924 struct sk_buff *skb;
3927 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3931 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3937 ifm = nlmsg_data(nlh);
3939 dev = __dev_get_by_index(net, ifm->ifa_index);
3941 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3943 err = -EADDRNOTAVAIL;
3947 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3953 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3954 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3956 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3957 WARN_ON(err == -EMSGSIZE);
3961 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3968 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3970 struct sk_buff *skb;
3971 struct net *net = dev_net(ifa->idev->dev);
3974 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3978 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3980 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3981 WARN_ON(err == -EMSGSIZE);
3985 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3989 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3992 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3993 __s32 *array, int bytes)
3995 BUG_ON(bytes < (DEVCONF_MAX * 4));
3997 memset(array, 0, bytes);
3998 array[DEVCONF_FORWARDING] = cnf->forwarding;
3999 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4000 array[DEVCONF_MTU6] = cnf->mtu6;
4001 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4002 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4003 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4004 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4005 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4006 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4007 jiffies_to_msecs(cnf->rtr_solicit_interval);
4008 array[DEVCONF_RTR_SOLICIT_DELAY] =
4009 jiffies_to_msecs(cnf->rtr_solicit_delay);
4010 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4011 #ifdef CONFIG_IPV6_PRIVACY
4012 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4013 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4014 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4015 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4016 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4018 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4019 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4020 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4021 #ifdef CONFIG_IPV6_ROUTER_PREF
4022 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4023 array[DEVCONF_RTR_PROBE_INTERVAL] =
4024 jiffies_to_msecs(cnf->rtr_probe_interval);
4025 #ifdef CONFIG_IPV6_ROUTE_INFO
4026 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4029 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4030 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4031 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4032 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4034 #ifdef CONFIG_IPV6_MROUTE
4035 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4037 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4038 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4039 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4040 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4043 static inline size_t inet6_ifla6_size(void)
4045 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4046 + nla_total_size(sizeof(struct ifla_cacheinfo))
4047 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4048 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4049 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4052 static inline size_t inet6_if_nlmsg_size(void)
4054 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4055 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4056 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4057 + nla_total_size(4) /* IFLA_MTU */
4058 + nla_total_size(4) /* IFLA_LINK */
4059 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4062 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4063 int items, int bytes)
4066 int pad = bytes - sizeof(u64) * items;
4069 /* Use put_unaligned() because stats may not be aligned for u64. */
4070 put_unaligned(items, &stats[0]);
4071 for (i = 1; i < items; i++)
4072 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4074 memset(&stats[items], 0, pad);
4077 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4078 int items, int bytes, size_t syncpoff)
4081 int pad = bytes - sizeof(u64) * items;
4084 /* Use put_unaligned() because stats may not be aligned for u64. */
4085 put_unaligned(items, &stats[0]);
4086 for (i = 1; i < items; i++)
4087 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4089 memset(&stats[items], 0, pad);
4092 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4096 case IFLA_INET6_STATS:
4097 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4098 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4100 case IFLA_INET6_ICMP6STATS:
4101 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4106 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4109 struct ifla_cacheinfo ci;
4111 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4112 goto nla_put_failure;
4113 ci.max_reasm_len = IPV6_MAXPLEN;
4114 ci.tstamp = cstamp_delta(idev->tstamp);
4115 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4116 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4117 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4118 goto nla_put_failure;
4119 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4121 goto nla_put_failure;
4122 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4124 /* XXX - MC not implemented */
4126 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4128 goto nla_put_failure;
4129 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4131 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4133 goto nla_put_failure;
4134 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4142 static size_t inet6_get_link_af_size(const struct net_device *dev)
4144 if (!__in6_dev_get(dev))
4147 return inet6_ifla6_size();
4150 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4152 struct inet6_dev *idev = __in6_dev_get(dev);
4157 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4163 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4164 u32 portid, u32 seq, int event, unsigned int flags)
4166 struct net_device *dev = idev->dev;
4167 struct ifinfomsg *hdr;
4168 struct nlmsghdr *nlh;
4171 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4175 hdr = nlmsg_data(nlh);
4176 hdr->ifi_family = AF_INET6;
4178 hdr->ifi_type = dev->type;
4179 hdr->ifi_index = dev->ifindex;
4180 hdr->ifi_flags = dev_get_flags(dev);
4181 hdr->ifi_change = 0;
4183 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4185 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4186 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4187 (dev->ifindex != dev->iflink &&
4188 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4189 goto nla_put_failure;
4190 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4191 if (protoinfo == NULL)
4192 goto nla_put_failure;
4194 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4195 goto nla_put_failure;
4197 nla_nest_end(skb, protoinfo);
4198 return nlmsg_end(skb, nlh);
4201 nlmsg_cancel(skb, nlh);
4205 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4207 struct net *net = sock_net(skb->sk);
4210 struct net_device *dev;
4211 struct inet6_dev *idev;
4212 struct hlist_head *head;
4213 struct hlist_node *node;
4216 s_idx = cb->args[1];
4219 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4221 head = &net->dev_index_head[h];
4222 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4225 idev = __in6_dev_get(dev);
4228 if (inet6_fill_ifinfo(skb, idev,
4229 NETLINK_CB(cb->skb).portid,
4231 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4245 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4247 struct sk_buff *skb;
4248 struct net *net = dev_net(idev->dev);
4251 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4255 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4257 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4258 WARN_ON(err == -EMSGSIZE);
4262 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4266 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4269 static inline size_t inet6_prefix_nlmsg_size(void)
4271 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4272 + nla_total_size(sizeof(struct in6_addr))
4273 + nla_total_size(sizeof(struct prefix_cacheinfo));
4276 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4277 struct prefix_info *pinfo, u32 portid, u32 seq,
4278 int event, unsigned int flags)
4280 struct prefixmsg *pmsg;
4281 struct nlmsghdr *nlh;
4282 struct prefix_cacheinfo ci;
4284 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4288 pmsg = nlmsg_data(nlh);
4289 pmsg->prefix_family = AF_INET6;
4290 pmsg->prefix_pad1 = 0;
4291 pmsg->prefix_pad2 = 0;
4292 pmsg->prefix_ifindex = idev->dev->ifindex;
4293 pmsg->prefix_len = pinfo->prefix_len;
4294 pmsg->prefix_type = pinfo->type;
4295 pmsg->prefix_pad3 = 0;
4296 pmsg->prefix_flags = 0;
4298 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4299 if (pinfo->autoconf)
4300 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4302 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4303 goto nla_put_failure;
4304 ci.preferred_time = ntohl(pinfo->prefered);
4305 ci.valid_time = ntohl(pinfo->valid);
4306 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4307 goto nla_put_failure;
4308 return nlmsg_end(skb, nlh);
4311 nlmsg_cancel(skb, nlh);
4315 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4316 struct prefix_info *pinfo)
4318 struct sk_buff *skb;
4319 struct net *net = dev_net(idev->dev);
4322 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4326 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4328 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4329 WARN_ON(err == -EMSGSIZE);
4333 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4337 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4340 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4342 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4347 * If the address was optimistic
4348 * we inserted the route at the start of
4349 * our DAD process, so we don't need
4352 if (!(ifp->rt->rt6i_node))
4353 ip6_ins_rt(ifp->rt);
4354 if (ifp->idev->cnf.forwarding)
4355 addrconf_join_anycast(ifp);
4358 if (ifp->idev->cnf.forwarding)
4359 addrconf_leave_anycast(ifp);
4360 addrconf_leave_solict(ifp->idev, &ifp->addr);
4361 dst_hold(&ifp->rt->dst);
4363 if (ip6_del_rt(ifp->rt))
4364 dst_free(&ifp->rt->dst);
4369 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4372 if (likely(ifp->idev->dead == 0))
4373 __ipv6_ifa_notify(event, ifp);
4374 rcu_read_unlock_bh();
4377 #ifdef CONFIG_SYSCTL
4380 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4381 void __user *buffer, size_t *lenp, loff_t *ppos)
4383 int *valp = ctl->data;
4390 * ctl->data points to idev->cnf.forwarding, we should
4391 * not modify it until we get the rtnl lock.
4396 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4399 ret = addrconf_fixup_forwarding(ctl, valp, val);
4405 static void dev_disable_change(struct inet6_dev *idev)
4407 if (!idev || !idev->dev)
4410 if (idev->cnf.disable_ipv6)
4411 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4413 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4416 static void addrconf_disable_change(struct net *net, __s32 newf)
4418 struct net_device *dev;
4419 struct inet6_dev *idev;
4422 for_each_netdev_rcu(net, dev) {
4423 idev = __in6_dev_get(dev);
4425 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4426 idev->cnf.disable_ipv6 = newf;
4428 dev_disable_change(idev);
4434 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4439 if (!rtnl_trylock())
4440 return restart_syscall();
4442 net = (struct net *)table->extra2;
4446 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4451 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4452 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4453 addrconf_disable_change(net, newf);
4454 } else if ((!newf) ^ (!old))
4455 dev_disable_change((struct inet6_dev *)table->extra1);
4462 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4463 void __user *buffer, size_t *lenp, loff_t *ppos)
4465 int *valp = ctl->data;
4472 * ctl->data points to idev->cnf.disable_ipv6, we should
4473 * not modify it until we get the rtnl lock.
4478 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4481 ret = addrconf_disable_ipv6(ctl, valp, val);
4487 static struct addrconf_sysctl_table
4489 struct ctl_table_header *sysctl_header;
4490 ctl_table addrconf_vars[DEVCONF_MAX+1];
4491 } addrconf_sysctl __read_mostly = {
4492 .sysctl_header = NULL,
4495 .procname = "forwarding",
4496 .data = &ipv6_devconf.forwarding,
4497 .maxlen = sizeof(int),
4499 .proc_handler = addrconf_sysctl_forward,
4502 .procname = "hop_limit",
4503 .data = &ipv6_devconf.hop_limit,
4504 .maxlen = sizeof(int),
4506 .proc_handler = proc_dointvec,
4510 .data = &ipv6_devconf.mtu6,
4511 .maxlen = sizeof(int),
4513 .proc_handler = proc_dointvec,
4516 .procname = "accept_ra",
4517 .data = &ipv6_devconf.accept_ra,
4518 .maxlen = sizeof(int),
4520 .proc_handler = proc_dointvec,
4523 .procname = "accept_redirects",
4524 .data = &ipv6_devconf.accept_redirects,
4525 .maxlen = sizeof(int),
4527 .proc_handler = proc_dointvec,
4530 .procname = "autoconf",
4531 .data = &ipv6_devconf.autoconf,
4532 .maxlen = sizeof(int),
4534 .proc_handler = proc_dointvec,
4537 .procname = "dad_transmits",
4538 .data = &ipv6_devconf.dad_transmits,
4539 .maxlen = sizeof(int),
4541 .proc_handler = proc_dointvec,
4544 .procname = "router_solicitations",
4545 .data = &ipv6_devconf.rtr_solicits,
4546 .maxlen = sizeof(int),
4548 .proc_handler = proc_dointvec,
4551 .procname = "router_solicitation_interval",
4552 .data = &ipv6_devconf.rtr_solicit_interval,
4553 .maxlen = sizeof(int),
4555 .proc_handler = proc_dointvec_jiffies,
4558 .procname = "router_solicitation_delay",
4559 .data = &ipv6_devconf.rtr_solicit_delay,
4560 .maxlen = sizeof(int),
4562 .proc_handler = proc_dointvec_jiffies,
4565 .procname = "force_mld_version",
4566 .data = &ipv6_devconf.force_mld_version,
4567 .maxlen = sizeof(int),
4569 .proc_handler = proc_dointvec,
4571 #ifdef CONFIG_IPV6_PRIVACY
4573 .procname = "use_tempaddr",
4574 .data = &ipv6_devconf.use_tempaddr,
4575 .maxlen = sizeof(int),
4577 .proc_handler = proc_dointvec,
4580 .procname = "temp_valid_lft",
4581 .data = &ipv6_devconf.temp_valid_lft,
4582 .maxlen = sizeof(int),
4584 .proc_handler = proc_dointvec,
4587 .procname = "temp_prefered_lft",
4588 .data = &ipv6_devconf.temp_prefered_lft,
4589 .maxlen = sizeof(int),
4591 .proc_handler = proc_dointvec,
4594 .procname = "regen_max_retry",
4595 .data = &ipv6_devconf.regen_max_retry,
4596 .maxlen = sizeof(int),
4598 .proc_handler = proc_dointvec,
4601 .procname = "max_desync_factor",
4602 .data = &ipv6_devconf.max_desync_factor,
4603 .maxlen = sizeof(int),
4605 .proc_handler = proc_dointvec,
4609 .procname = "max_addresses",
4610 .data = &ipv6_devconf.max_addresses,
4611 .maxlen = sizeof(int),
4613 .proc_handler = proc_dointvec,
4616 .procname = "accept_ra_defrtr",
4617 .data = &ipv6_devconf.accept_ra_defrtr,
4618 .maxlen = sizeof(int),
4620 .proc_handler = proc_dointvec,
4623 .procname = "accept_ra_pinfo",
4624 .data = &ipv6_devconf.accept_ra_pinfo,
4625 .maxlen = sizeof(int),
4627 .proc_handler = proc_dointvec,
4629 #ifdef CONFIG_IPV6_ROUTER_PREF
4631 .procname = "accept_ra_rtr_pref",
4632 .data = &ipv6_devconf.accept_ra_rtr_pref,
4633 .maxlen = sizeof(int),
4635 .proc_handler = proc_dointvec,
4638 .procname = "router_probe_interval",
4639 .data = &ipv6_devconf.rtr_probe_interval,
4640 .maxlen = sizeof(int),
4642 .proc_handler = proc_dointvec_jiffies,
4644 #ifdef CONFIG_IPV6_ROUTE_INFO
4646 .procname = "accept_ra_rt_info_max_plen",
4647 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4648 .maxlen = sizeof(int),
4650 .proc_handler = proc_dointvec,
4655 .procname = "proxy_ndp",
4656 .data = &ipv6_devconf.proxy_ndp,
4657 .maxlen = sizeof(int),
4659 .proc_handler = proc_dointvec,
4662 .procname = "accept_source_route",
4663 .data = &ipv6_devconf.accept_source_route,
4664 .maxlen = sizeof(int),
4666 .proc_handler = proc_dointvec,
4668 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4670 .procname = "optimistic_dad",
4671 .data = &ipv6_devconf.optimistic_dad,
4672 .maxlen = sizeof(int),
4674 .proc_handler = proc_dointvec,
4678 #ifdef CONFIG_IPV6_MROUTE
4680 .procname = "mc_forwarding",
4681 .data = &ipv6_devconf.mc_forwarding,
4682 .maxlen = sizeof(int),
4684 .proc_handler = proc_dointvec,
4688 .procname = "disable_ipv6",
4689 .data = &ipv6_devconf.disable_ipv6,
4690 .maxlen = sizeof(int),
4692 .proc_handler = addrconf_sysctl_disable,
4695 .procname = "accept_dad",
4696 .data = &ipv6_devconf.accept_dad,
4697 .maxlen = sizeof(int),
4699 .proc_handler = proc_dointvec,
4702 .procname = "force_tllao",
4703 .data = &ipv6_devconf.force_tllao,
4704 .maxlen = sizeof(int),
4706 .proc_handler = proc_dointvec
4709 .procname = "ndisc_notify",
4710 .data = &ipv6_devconf.ndisc_notify,
4711 .maxlen = sizeof(int),
4713 .proc_handler = proc_dointvec
4721 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4722 struct inet6_dev *idev, struct ipv6_devconf *p)
4725 struct addrconf_sysctl_table *t;
4726 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4728 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4732 for (i = 0; t->addrconf_vars[i].data; i++) {
4733 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4734 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4735 t->addrconf_vars[i].extra2 = net;
4738 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4740 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4741 if (t->sysctl_header == NULL)
4753 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4755 struct addrconf_sysctl_table *t;
4757 if (p->sysctl == NULL)
4762 unregister_net_sysctl_table(t->sysctl_header);
4766 static void addrconf_sysctl_register(struct inet6_dev *idev)
4768 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4769 &ndisc_ifinfo_sysctl_change);
4770 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4774 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4776 __addrconf_sysctl_unregister(&idev->cnf);
4777 neigh_sysctl_unregister(idev->nd_parms);
4783 static int __net_init addrconf_init_net(struct net *net)
4786 struct ipv6_devconf *all, *dflt;
4789 all = &ipv6_devconf;
4790 dflt = &ipv6_devconf_dflt;
4792 if (!net_eq(net, &init_net)) {
4793 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4797 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4799 goto err_alloc_dflt;
4801 /* these will be inherited by all namespaces */
4802 dflt->autoconf = ipv6_defaults.autoconf;
4803 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4806 net->ipv6.devconf_all = all;
4807 net->ipv6.devconf_dflt = dflt;
4809 #ifdef CONFIG_SYSCTL
4810 err = __addrconf_sysctl_register(net, "all", NULL, all);
4814 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4820 #ifdef CONFIG_SYSCTL
4822 __addrconf_sysctl_unregister(all);
4832 static void __net_exit addrconf_exit_net(struct net *net)
4834 #ifdef CONFIG_SYSCTL
4835 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4836 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4838 if (!net_eq(net, &init_net)) {
4839 kfree(net->ipv6.devconf_dflt);
4840 kfree(net->ipv6.devconf_all);
4844 static struct pernet_operations addrconf_ops = {
4845 .init = addrconf_init_net,
4846 .exit = addrconf_exit_net,
4853 int register_inet6addr_notifier(struct notifier_block *nb)
4855 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4857 EXPORT_SYMBOL(register_inet6addr_notifier);
4859 int unregister_inet6addr_notifier(struct notifier_block *nb)
4861 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4863 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4865 static struct rtnl_af_ops inet6_ops = {
4867 .fill_link_af = inet6_fill_link_af,
4868 .get_link_af_size = inet6_get_link_af_size,
4872 * Init / cleanup code
4875 int __init addrconf_init(void)
4879 err = ipv6_addr_label_init();
4881 pr_crit("%s: cannot initialize default policy table: %d\n",
4886 err = register_pernet_subsys(&addrconf_ops);
4890 /* The addrconf netdev notifier requires that loopback_dev
4891 * has it's ipv6 private information allocated and setup
4892 * before it can bring up and give link-local addresses
4893 * to other devices which are up.
4895 * Unfortunately, loopback_dev is not necessarily the first
4896 * entry in the global dev_base list of net devices. In fact,
4897 * it is likely to be the very last entry on that list.
4898 * So this causes the notifier registry below to try and
4899 * give link-local addresses to all devices besides loopback_dev
4900 * first, then loopback_dev, which cases all the non-loopback_dev
4901 * devices to fail to get a link-local address.
4903 * So, as a temporary fix, allocate the ipv6 structure for
4904 * loopback_dev first by hand.
4905 * Longer term, all of the dependencies ipv6 has upon the loopback
4906 * device and it being up should be removed.
4909 if (!ipv6_add_dev(init_net.loopback_dev))
4915 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4916 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4918 register_netdevice_notifier(&ipv6_dev_notf);
4922 err = rtnl_af_register(&inet6_ops);
4926 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4931 /* Only the first call to __rtnl_register can fail */
4932 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4933 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4934 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4935 inet6_dump_ifaddr, NULL);
4936 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4937 inet6_dump_ifmcaddr, NULL);
4938 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4939 inet6_dump_ifacaddr, NULL);
4940 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4943 ipv6_addr_label_rtnl_register();
4947 rtnl_af_unregister(&inet6_ops);
4949 unregister_netdevice_notifier(&ipv6_dev_notf);
4951 unregister_pernet_subsys(&addrconf_ops);
4953 ipv6_addr_label_cleanup();
4958 void addrconf_cleanup(void)
4960 struct net_device *dev;
4963 unregister_netdevice_notifier(&ipv6_dev_notf);
4964 unregister_pernet_subsys(&addrconf_ops);
4965 ipv6_addr_label_cleanup();
4969 __rtnl_af_unregister(&inet6_ops);
4971 /* clean dev list */
4972 for_each_netdev(&init_net, dev) {
4973 if (__in6_dev_get(dev) == NULL)
4975 addrconf_ifdown(dev, 1);
4977 addrconf_ifdown(init_net.loopback_dev, 2);
4982 spin_lock_bh(&addrconf_hash_lock);
4983 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4984 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4985 spin_unlock_bh(&addrconf_hash_lock);
4987 del_timer(&addr_chk_timer);