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);
472 #ifdef CONFIG_IPV6_MROUTE
473 if (type == -1 || type == NETCONFA_MC_FORWARDING)
474 size += nla_total_size(4);
480 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
481 struct ipv6_devconf *devconf, u32 portid,
482 u32 seq, int event, unsigned int flags,
485 struct nlmsghdr *nlh;
486 struct netconfmsg *ncm;
488 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
493 ncm = nlmsg_data(nlh);
494 ncm->ncm_family = AF_INET6;
496 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
497 goto nla_put_failure;
499 /* type -1 is used for ALL */
500 if ((type == -1 || type == NETCONFA_FORWARDING) &&
501 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
502 goto nla_put_failure;
503 #ifdef CONFIG_IPV6_MROUTE
504 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
505 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
506 devconf->mc_forwarding) < 0)
507 goto nla_put_failure;
509 return nlmsg_end(skb, nlh);
512 nlmsg_cancel(skb, nlh);
516 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
517 struct ipv6_devconf *devconf)
522 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
526 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
527 RTM_NEWNETCONF, 0, type);
529 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
530 WARN_ON(err == -EMSGSIZE);
534 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
538 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
542 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
543 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
547 struct nlmsghdr *nlh,
550 struct net *net = sock_net(in_skb->sk);
551 struct nlattr *tb[NETCONFA_MAX+1];
552 struct netconfmsg *ncm;
554 struct ipv6_devconf *devconf;
555 struct inet6_dev *in6_dev;
556 struct net_device *dev;
560 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
561 devconf_ipv6_policy);
566 if (!tb[NETCONFA_IFINDEX])
569 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
571 case NETCONFA_IFINDEX_ALL:
572 devconf = net->ipv6.devconf_all;
574 case NETCONFA_IFINDEX_DEFAULT:
575 devconf = net->ipv6.devconf_dflt;
578 dev = __dev_get_by_index(net, ifindex);
581 in6_dev = __in6_dev_get(dev);
584 devconf = &in6_dev->cnf;
589 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
593 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
594 NETLINK_CB(in_skb).portid,
595 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
598 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
599 WARN_ON(err == -EMSGSIZE);
603 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
609 static void dev_forward_change(struct inet6_dev *idev)
611 struct net_device *dev;
612 struct inet6_ifaddr *ifa;
617 if (idev->cnf.forwarding)
618 dev_disable_lro(dev);
619 if (dev->flags & IFF_MULTICAST) {
620 if (idev->cnf.forwarding)
621 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
623 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
626 list_for_each_entry(ifa, &idev->addr_list, if_list) {
627 if (ifa->flags&IFA_F_TENTATIVE)
629 if (idev->cnf.forwarding)
630 addrconf_join_anycast(ifa);
632 addrconf_leave_anycast(ifa);
634 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
635 dev->ifindex, &idev->cnf);
639 static void addrconf_forward_change(struct net *net, __s32 newf)
641 struct net_device *dev;
642 struct inet6_dev *idev;
644 for_each_netdev(net, dev) {
645 idev = __in6_dev_get(dev);
647 int changed = (!idev->cnf.forwarding) ^ (!newf);
648 idev->cnf.forwarding = newf;
650 dev_forward_change(idev);
655 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
661 return restart_syscall();
663 net = (struct net *)table->extra2;
667 if (p == &net->ipv6.devconf_dflt->forwarding) {
668 if ((!newf) ^ (!old))
669 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
670 NETCONFA_IFINDEX_DEFAULT,
671 net->ipv6.devconf_dflt);
676 if (p == &net->ipv6.devconf_all->forwarding) {
677 net->ipv6.devconf_dflt->forwarding = newf;
678 addrconf_forward_change(net, newf);
679 if ((!newf) ^ (!old))
680 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
681 NETCONFA_IFINDEX_ALL,
682 net->ipv6.devconf_all);
683 } else if ((!newf) ^ (!old))
684 dev_forward_change((struct inet6_dev *)table->extra1);
688 rt6_purge_dflt_routers(net);
693 /* Nobody refers to this ifaddr, destroy it */
694 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
696 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
698 #ifdef NET_REFCNT_DEBUG
699 pr_debug("%s\n", __func__);
702 in6_dev_put(ifp->idev);
704 if (del_timer(&ifp->timer))
705 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
707 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
708 pr_warn("Freeing alive inet6 address %p\n", ifp);
717 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
720 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
723 * Each device address list is sorted in order of scope -
724 * global before linklocal.
726 list_for_each(p, &idev->addr_list) {
727 struct inet6_ifaddr *ifa
728 = list_entry(p, struct inet6_ifaddr, if_list);
729 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
733 list_add_tail(&ifp->if_list, p);
736 static u32 inet6_addr_hash(const struct in6_addr *addr)
738 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
741 /* On success it returns ifp with increased reference count */
743 static struct inet6_ifaddr *
744 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
745 int scope, u32 flags)
747 struct inet6_ifaddr *ifa = NULL;
751 int addr_type = ipv6_addr_type(addr);
753 if (addr_type == IPV6_ADDR_ANY ||
754 addr_type & IPV6_ADDR_MULTICAST ||
755 (!(idev->dev->flags & IFF_LOOPBACK) &&
756 addr_type & IPV6_ADDR_LOOPBACK))
757 return ERR_PTR(-EADDRNOTAVAIL);
761 err = -ENODEV; /*XXX*/
765 if (idev->cnf.disable_ipv6) {
770 spin_lock(&addrconf_hash_lock);
772 /* Ignore adding duplicate addresses on an interface */
773 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
774 ADBG(("ipv6_add_addr: already assigned\n"));
779 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
782 ADBG(("ipv6_add_addr: malloc failed\n"));
787 rt = addrconf_dst_alloc(idev, addr, false);
795 spin_lock_init(&ifa->lock);
796 spin_lock_init(&ifa->state_lock);
797 init_timer(&ifa->timer);
798 INIT_HLIST_NODE(&ifa->addr_lst);
799 ifa->timer.data = (unsigned long) ifa;
801 ifa->prefix_len = pfxlen;
802 ifa->flags = flags | IFA_F_TENTATIVE;
803 ifa->cstamp = ifa->tstamp = jiffies;
812 /* Add to big hash table */
813 hash = inet6_addr_hash(addr);
815 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
816 spin_unlock(&addrconf_hash_lock);
818 write_lock(&idev->lock);
819 /* Add to inet6_dev unicast addr list. */
820 ipv6_link_dev_addr(idev, ifa);
822 #ifdef CONFIG_IPV6_PRIVACY
823 if (ifa->flags&IFA_F_TEMPORARY) {
824 list_add(&ifa->tmp_list, &idev->tempaddr_list);
830 write_unlock(&idev->lock);
832 rcu_read_unlock_bh();
834 if (likely(err == 0))
835 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
843 spin_unlock(&addrconf_hash_lock);
847 /* This function wants to get referenced ifp and releases it before return */
849 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
851 struct inet6_ifaddr *ifa, *ifn;
852 struct inet6_dev *idev = ifp->idev;
854 int deleted = 0, onlink = 0;
855 unsigned long expires = jiffies;
857 spin_lock_bh(&ifp->state_lock);
859 ifp->state = INET6_IFADDR_STATE_DEAD;
860 spin_unlock_bh(&ifp->state_lock);
862 if (state == INET6_IFADDR_STATE_DEAD)
865 spin_lock_bh(&addrconf_hash_lock);
866 hlist_del_init_rcu(&ifp->addr_lst);
867 spin_unlock_bh(&addrconf_hash_lock);
869 write_lock_bh(&idev->lock);
870 #ifdef CONFIG_IPV6_PRIVACY
871 if (ifp->flags&IFA_F_TEMPORARY) {
872 list_del(&ifp->tmp_list);
874 in6_ifa_put(ifp->ifpub);
881 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
883 list_del_init(&ifp->if_list);
886 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
890 } else if (ifp->flags & IFA_F_PERMANENT) {
891 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
893 if (ifa->flags & IFA_F_PERMANENT) {
898 unsigned long lifetime;
903 spin_lock(&ifa->lock);
905 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
907 * Note: Because this address is
908 * not permanent, lifetime <
909 * LONG_MAX / HZ here.
911 if (time_before(expires,
912 ifa->tstamp + lifetime * HZ))
913 expires = ifa->tstamp + lifetime * HZ;
914 spin_unlock(&ifa->lock);
919 write_unlock_bh(&idev->lock);
921 addrconf_del_timer(ifp);
923 ipv6_ifa_notify(RTM_DELADDR, ifp);
925 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
928 * Purge or update corresponding prefix
930 * 1) we don't purge prefix here if address was not permanent.
931 * prefix is managed by its own lifetime.
932 * 2) if there're no addresses, delete prefix.
933 * 3) if there're still other permanent address(es),
934 * corresponding prefix is still permanent.
935 * 4) otherwise, update prefix lifetime to the
936 * longest valid lifetime among the corresponding
937 * addresses on the device.
938 * Note: subsequent RA will update lifetime.
942 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
943 struct in6_addr prefix;
945 struct net *net = dev_net(ifp->idev->dev);
946 struct flowi6 fl6 = {};
948 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
949 fl6.flowi6_oif = ifp->idev->dev->ifindex;
951 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
954 if (rt != net->ipv6.ip6_null_entry &&
955 addrconf_is_prefix_route(rt)) {
959 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
960 rt6_set_expires(rt, expires);
966 /* clean up prefsrc entries */
967 rt6_remove_prefsrc(ifp);
972 #ifdef CONFIG_IPV6_PRIVACY
973 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
975 struct inet6_dev *idev = ifp->idev;
976 struct in6_addr addr, *tmpaddr;
977 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
978 unsigned long regen_advance;
983 unsigned long now = jiffies;
985 write_lock(&idev->lock);
987 spin_lock_bh(&ift->lock);
988 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
989 spin_unlock_bh(&ift->lock);
996 if (idev->cnf.use_tempaddr <= 0) {
997 write_unlock(&idev->lock);
998 pr_info("%s: use_tempaddr is disabled\n", __func__);
1003 spin_lock_bh(&ifp->lock);
1004 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1005 idev->cnf.use_tempaddr = -1; /*XXX*/
1006 spin_unlock_bh(&ifp->lock);
1007 write_unlock(&idev->lock);
1008 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1015 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1016 __ipv6_try_regen_rndid(idev, tmpaddr);
1017 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1018 age = (now - ifp->tstamp) / HZ;
1019 tmp_valid_lft = min_t(__u32,
1021 idev->cnf.temp_valid_lft + age);
1022 tmp_prefered_lft = min_t(__u32,
1024 idev->cnf.temp_prefered_lft + age -
1025 idev->cnf.max_desync_factor);
1026 tmp_plen = ifp->prefix_len;
1027 max_addresses = idev->cnf.max_addresses;
1028 tmp_tstamp = ifp->tstamp;
1029 spin_unlock_bh(&ifp->lock);
1031 regen_advance = idev->cnf.regen_max_retry *
1032 idev->cnf.dad_transmits *
1033 idev->nd_parms->retrans_time / HZ;
1034 write_unlock(&idev->lock);
1036 /* A temporary address is created only if this calculated Preferred
1037 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1038 * an implementation must not create a temporary address with a zero
1039 * Preferred Lifetime.
1041 if (tmp_prefered_lft <= regen_advance) {
1048 addr_flags = IFA_F_TEMPORARY;
1049 /* set in addrconf_prefix_rcv() */
1050 if (ifp->flags & IFA_F_OPTIMISTIC)
1051 addr_flags |= IFA_F_OPTIMISTIC;
1053 ift = !max_addresses ||
1054 ipv6_count_addresses(idev) < max_addresses ?
1055 ipv6_add_addr(idev, &addr, tmp_plen,
1056 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1058 if (!ift || IS_ERR(ift)) {
1061 pr_info("%s: retry temporary address regeneration\n", __func__);
1063 write_lock(&idev->lock);
1067 spin_lock_bh(&ift->lock);
1069 ift->valid_lft = tmp_valid_lft;
1070 ift->prefered_lft = tmp_prefered_lft;
1072 ift->tstamp = tmp_tstamp;
1073 spin_unlock_bh(&ift->lock);
1075 addrconf_dad_start(ift);
1084 * Choose an appropriate source address (RFC3484)
1087 IPV6_SADDR_RULE_INIT = 0,
1088 IPV6_SADDR_RULE_LOCAL,
1089 IPV6_SADDR_RULE_SCOPE,
1090 IPV6_SADDR_RULE_PREFERRED,
1091 #ifdef CONFIG_IPV6_MIP6
1092 IPV6_SADDR_RULE_HOA,
1094 IPV6_SADDR_RULE_OIF,
1095 IPV6_SADDR_RULE_LABEL,
1096 #ifdef CONFIG_IPV6_PRIVACY
1097 IPV6_SADDR_RULE_PRIVACY,
1099 IPV6_SADDR_RULE_ORCHID,
1100 IPV6_SADDR_RULE_PREFIX,
1104 struct ipv6_saddr_score {
1107 struct inet6_ifaddr *ifa;
1108 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1113 struct ipv6_saddr_dst {
1114 const struct in6_addr *addr;
1121 static inline int ipv6_saddr_preferred(int type)
1123 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1128 static int ipv6_get_saddr_eval(struct net *net,
1129 struct ipv6_saddr_score *score,
1130 struct ipv6_saddr_dst *dst,
1135 if (i <= score->rule) {
1137 case IPV6_SADDR_RULE_SCOPE:
1138 ret = score->scopedist;
1140 case IPV6_SADDR_RULE_PREFIX:
1141 ret = score->matchlen;
1144 ret = !!test_bit(i, score->scorebits);
1150 case IPV6_SADDR_RULE_INIT:
1151 /* Rule 0: remember if hiscore is not ready yet */
1154 case IPV6_SADDR_RULE_LOCAL:
1155 /* Rule 1: Prefer same address */
1156 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1158 case IPV6_SADDR_RULE_SCOPE:
1159 /* Rule 2: Prefer appropriate scope
1164 * ---+--+-+---> scope
1166 * | d is scope of the destination.
1168 * | \ <- smaller scope is better if
1169 * B-15 | \ if scope is enough for destinaion.
1170 * | ret = B - scope (-1 <= scope >= d <= 15).
1172 * |/ <- greater is better
1173 * -C / if scope is not enough for destination.
1174 * /| ret = scope - C (-1 <= d < scope <= 15).
1176 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1177 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1178 * Assume B = 0 and we get C > 29.
1180 ret = __ipv6_addr_src_scope(score->addr_type);
1181 if (ret >= dst->scope)
1184 ret -= 128; /* 30 is enough */
1185 score->scopedist = ret;
1187 case IPV6_SADDR_RULE_PREFERRED:
1188 /* Rule 3: Avoid deprecated and optimistic addresses */
1189 ret = ipv6_saddr_preferred(score->addr_type) ||
1190 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1192 #ifdef CONFIG_IPV6_MIP6
1193 case IPV6_SADDR_RULE_HOA:
1195 /* Rule 4: Prefer home address */
1196 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1197 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1201 case IPV6_SADDR_RULE_OIF:
1202 /* Rule 5: Prefer outgoing interface */
1203 ret = (!dst->ifindex ||
1204 dst->ifindex == score->ifa->idev->dev->ifindex);
1206 case IPV6_SADDR_RULE_LABEL:
1207 /* Rule 6: Prefer matching label */
1208 ret = ipv6_addr_label(net,
1209 &score->ifa->addr, score->addr_type,
1210 score->ifa->idev->dev->ifindex) == dst->label;
1212 #ifdef CONFIG_IPV6_PRIVACY
1213 case IPV6_SADDR_RULE_PRIVACY:
1215 /* Rule 7: Prefer public address
1216 * Note: prefer temporary address if use_tempaddr >= 2
1218 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1219 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1220 score->ifa->idev->cnf.use_tempaddr >= 2;
1221 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1225 case IPV6_SADDR_RULE_ORCHID:
1226 /* Rule 8-: Prefer ORCHID vs ORCHID or
1227 * non-ORCHID vs non-ORCHID
1229 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1230 ipv6_addr_orchid(dst->addr));
1232 case IPV6_SADDR_RULE_PREFIX:
1233 /* Rule 8: Use longest matching prefix */
1234 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1235 if (ret > score->ifa->prefix_len)
1236 ret = score->ifa->prefix_len;
1237 score->matchlen = ret;
1244 __set_bit(i, score->scorebits);
1250 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1251 const struct in6_addr *daddr, unsigned int prefs,
1252 struct in6_addr *saddr)
1254 struct ipv6_saddr_score scores[2],
1255 *score = &scores[0], *hiscore = &scores[1];
1256 struct ipv6_saddr_dst dst;
1257 struct net_device *dev;
1260 dst_type = __ipv6_addr_type(daddr);
1262 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1263 dst.scope = __ipv6_addr_src_scope(dst_type);
1264 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1268 hiscore->ifa = NULL;
1272 for_each_netdev_rcu(net, dev) {
1273 struct inet6_dev *idev;
1275 /* Candidate Source Address (section 4)
1276 * - multicast and link-local destination address,
1277 * the set of candidate source address MUST only
1278 * include addresses assigned to interfaces
1279 * belonging to the same link as the outgoing
1281 * (- For site-local destination addresses, the
1282 * set of candidate source addresses MUST only
1283 * include addresses assigned to interfaces
1284 * belonging to the same site as the outgoing
1287 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1288 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1289 dst.ifindex && dev->ifindex != dst.ifindex)
1292 idev = __in6_dev_get(dev);
1296 read_lock_bh(&idev->lock);
1297 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1301 * - Tentative Address (RFC2462 section 5.4)
1302 * - A tentative address is not considered
1303 * "assigned to an interface" in the traditional
1304 * sense, unless it is also flagged as optimistic.
1305 * - Candidate Source Address (section 4)
1306 * - In any case, anycast addresses, multicast
1307 * addresses, and the unspecified address MUST
1308 * NOT be included in a candidate set.
1310 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1311 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1314 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1316 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1317 score->addr_type & IPV6_ADDR_MULTICAST)) {
1318 LIMIT_NETDEBUG(KERN_DEBUG
1319 "ADDRCONF: unspecified / multicast address "
1320 "assigned as unicast address on %s",
1326 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1328 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1329 int minihiscore, miniscore;
1331 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1332 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1334 if (minihiscore > miniscore) {
1335 if (i == IPV6_SADDR_RULE_SCOPE &&
1336 score->scopedist > 0) {
1339 * each remaining entry
1340 * has too small (not enough)
1341 * scope, because ifa entries
1342 * are sorted by their scope
1348 } else if (minihiscore < miniscore) {
1350 in6_ifa_put(hiscore->ifa);
1352 in6_ifa_hold(score->ifa);
1354 swap(hiscore, score);
1356 /* restore our iterator */
1357 score->ifa = hiscore->ifa;
1364 read_unlock_bh(&idev->lock);
1369 return -EADDRNOTAVAIL;
1371 *saddr = hiscore->ifa->addr;
1372 in6_ifa_put(hiscore->ifa);
1375 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1377 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1378 unsigned char banned_flags)
1380 struct inet6_dev *idev;
1381 int err = -EADDRNOTAVAIL;
1384 idev = __in6_dev_get(dev);
1386 struct inet6_ifaddr *ifp;
1388 read_lock_bh(&idev->lock);
1389 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1390 if (ifp->scope == IFA_LINK &&
1391 !(ifp->flags & banned_flags)) {
1397 read_unlock_bh(&idev->lock);
1403 static int ipv6_count_addresses(struct inet6_dev *idev)
1406 struct inet6_ifaddr *ifp;
1408 read_lock_bh(&idev->lock);
1409 list_for_each_entry(ifp, &idev->addr_list, if_list)
1411 read_unlock_bh(&idev->lock);
1415 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1416 struct net_device *dev, int strict)
1418 struct inet6_ifaddr *ifp;
1419 struct hlist_node *node;
1420 unsigned int hash = inet6_addr_hash(addr);
1423 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1424 if (!net_eq(dev_net(ifp->idev->dev), net))
1426 if (ipv6_addr_equal(&ifp->addr, addr) &&
1427 !(ifp->flags&IFA_F_TENTATIVE) &&
1428 (dev == NULL || ifp->idev->dev == dev ||
1429 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1430 rcu_read_unlock_bh();
1435 rcu_read_unlock_bh();
1438 EXPORT_SYMBOL(ipv6_chk_addr);
1440 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1441 struct net_device *dev)
1443 unsigned int hash = inet6_addr_hash(addr);
1444 struct inet6_ifaddr *ifp;
1445 struct hlist_node *node;
1447 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1448 if (!net_eq(dev_net(ifp->idev->dev), net))
1450 if (ipv6_addr_equal(&ifp->addr, addr)) {
1451 if (dev == NULL || ifp->idev->dev == dev)
1458 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1460 struct inet6_dev *idev;
1461 struct inet6_ifaddr *ifa;
1466 idev = __in6_dev_get(dev);
1468 read_lock_bh(&idev->lock);
1469 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1470 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1475 read_unlock_bh(&idev->lock);
1480 EXPORT_SYMBOL(ipv6_chk_prefix);
1482 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1483 struct net_device *dev, int strict)
1485 struct inet6_ifaddr *ifp, *result = NULL;
1486 unsigned int hash = inet6_addr_hash(addr);
1487 struct hlist_node *node;
1490 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1491 if (!net_eq(dev_net(ifp->idev->dev), net))
1493 if (ipv6_addr_equal(&ifp->addr, addr)) {
1494 if (dev == NULL || ifp->idev->dev == dev ||
1495 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1502 rcu_read_unlock_bh();
1507 /* Gets referenced address, destroys ifaddr */
1509 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1511 if (ifp->flags&IFA_F_PERMANENT) {
1512 spin_lock_bh(&ifp->lock);
1513 addrconf_del_timer(ifp);
1514 ifp->flags |= IFA_F_TENTATIVE;
1516 ifp->flags |= IFA_F_DADFAILED;
1517 spin_unlock_bh(&ifp->lock);
1519 ipv6_ifa_notify(0, ifp);
1521 #ifdef CONFIG_IPV6_PRIVACY
1522 } else if (ifp->flags&IFA_F_TEMPORARY) {
1523 struct inet6_ifaddr *ifpub;
1524 spin_lock_bh(&ifp->lock);
1527 in6_ifa_hold(ifpub);
1528 spin_unlock_bh(&ifp->lock);
1529 ipv6_create_tempaddr(ifpub, ifp);
1532 spin_unlock_bh(&ifp->lock);
1540 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1544 spin_lock(&ifp->state_lock);
1545 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1546 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1549 spin_unlock(&ifp->state_lock);
1554 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1556 struct inet6_dev *idev = ifp->idev;
1558 if (addrconf_dad_end(ifp)) {
1563 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1564 ifp->idev->dev->name, &ifp->addr);
1566 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1567 struct in6_addr addr;
1569 addr.s6_addr32[0] = htonl(0xfe800000);
1570 addr.s6_addr32[1] = 0;
1572 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1573 ipv6_addr_equal(&ifp->addr, &addr)) {
1574 /* DAD failed for link-local based on MAC address */
1575 idev->cnf.disable_ipv6 = 1;
1577 pr_info("%s: IPv6 being disabled!\n",
1578 ifp->idev->dev->name);
1582 addrconf_dad_stop(ifp, 1);
1585 /* Join to solicited addr multicast group. */
1587 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1589 struct in6_addr maddr;
1591 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1594 addrconf_addr_solict_mult(addr, &maddr);
1595 ipv6_dev_mc_inc(dev, &maddr);
1598 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1600 struct in6_addr maddr;
1602 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1605 addrconf_addr_solict_mult(addr, &maddr);
1606 __ipv6_dev_mc_dec(idev, &maddr);
1609 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1611 struct in6_addr addr;
1612 if (ifp->prefix_len == 127) /* RFC 6164 */
1614 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1615 if (ipv6_addr_any(&addr))
1617 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1620 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1622 struct in6_addr addr;
1623 if (ifp->prefix_len == 127) /* RFC 6164 */
1625 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1626 if (ipv6_addr_any(&addr))
1628 __ipv6_dev_ac_dec(ifp->idev, &addr);
1631 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1633 if (dev->addr_len != ETH_ALEN)
1635 memcpy(eui, dev->dev_addr, 3);
1636 memcpy(eui + 5, dev->dev_addr + 3, 3);
1639 * The zSeries OSA network cards can be shared among various
1640 * OS instances, but the OSA cards have only one MAC address.
1641 * This leads to duplicate address conflicts in conjunction
1642 * with IPv6 if more than one instance uses the same card.
1644 * The driver for these cards can deliver a unique 16-bit
1645 * identifier for each instance sharing the same card. It is
1646 * placed instead of 0xFFFE in the interface identifier. The
1647 * "u" bit of the interface identifier is not inverted in this
1648 * case. Hence the resulting interface identifier has local
1649 * scope according to RFC2373.
1652 eui[3] = (dev->dev_id >> 8) & 0xFF;
1653 eui[4] = dev->dev_id & 0xFF;
1662 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1664 if (dev->addr_len != IEEE802154_ADDR_LEN)
1666 memcpy(eui, dev->dev_addr, 8);
1670 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1672 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1673 if (dev->addr_len != ARCNET_ALEN)
1676 eui[7] = *(u8 *)dev->dev_addr;
1680 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1682 if (dev->addr_len != INFINIBAND_ALEN)
1684 memcpy(eui, dev->dev_addr + 12, 8);
1689 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1693 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1694 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1695 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1696 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1697 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1698 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1702 memcpy(eui + 4, &addr, 4);
1706 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1708 if (dev->priv_flags & IFF_ISATAP)
1709 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1713 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1715 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1718 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1720 switch (dev->type) {
1723 return addrconf_ifid_eui48(eui, dev);
1725 return addrconf_ifid_arcnet(eui, dev);
1726 case ARPHRD_INFINIBAND:
1727 return addrconf_ifid_infiniband(eui, dev);
1729 return addrconf_ifid_sit(eui, dev);
1731 return addrconf_ifid_gre(eui, dev);
1732 case ARPHRD_IEEE802154:
1733 return addrconf_ifid_eui64(eui, dev);
1738 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1741 struct inet6_ifaddr *ifp;
1743 read_lock_bh(&idev->lock);
1744 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1745 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1746 memcpy(eui, ifp->addr.s6_addr+8, 8);
1751 read_unlock_bh(&idev->lock);
1755 #ifdef CONFIG_IPV6_PRIVACY
1756 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1757 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1760 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1761 idev->rndid[0] &= ~0x02;
1764 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1765 * check if generated address is not inappropriate
1767 * - Reserved subnet anycast (RFC 2526)
1768 * 11111101 11....11 1xxxxxxx
1769 * - ISATAP (RFC4214) 6.1
1770 * 00-00-5E-FE-xx-xx-xx-xx
1772 * - XXX: already assigned to an address on the device
1774 if (idev->rndid[0] == 0xfd &&
1775 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1776 (idev->rndid[7]&0x80))
1778 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1779 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1781 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1786 static void ipv6_regen_rndid(unsigned long data)
1788 struct inet6_dev *idev = (struct inet6_dev *) data;
1789 unsigned long expires;
1792 write_lock_bh(&idev->lock);
1797 __ipv6_regen_rndid(idev);
1800 idev->cnf.temp_prefered_lft * HZ -
1801 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1802 idev->cnf.max_desync_factor * HZ;
1803 if (time_before(expires, jiffies)) {
1804 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1805 __func__, idev->dev->name);
1809 if (!mod_timer(&idev->regen_timer, expires))
1813 write_unlock_bh(&idev->lock);
1814 rcu_read_unlock_bh();
1818 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1820 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1821 __ipv6_regen_rndid(idev);
1830 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1831 unsigned long expires, u32 flags)
1833 struct fib6_config cfg = {
1834 .fc_table = RT6_TABLE_PREFIX,
1835 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1836 .fc_ifindex = dev->ifindex,
1837 .fc_expires = expires,
1839 .fc_flags = RTF_UP | flags,
1840 .fc_nlinfo.nl_net = dev_net(dev),
1841 .fc_protocol = RTPROT_KERNEL,
1846 /* Prevent useless cloning on PtP SIT.
1847 This thing is done here expecting that the whole
1848 class of non-broadcast devices need not cloning.
1850 #if IS_ENABLED(CONFIG_IPV6_SIT)
1851 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1852 cfg.fc_flags |= RTF_NONEXTHOP;
1855 ip6_route_add(&cfg);
1859 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1861 const struct net_device *dev,
1862 u32 flags, u32 noflags)
1864 struct fib6_node *fn;
1865 struct rt6_info *rt = NULL;
1866 struct fib6_table *table;
1868 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1872 read_lock_bh(&table->tb6_lock);
1873 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1876 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1877 if (rt->dst.dev->ifindex != dev->ifindex)
1879 if ((rt->rt6i_flags & flags) != flags)
1881 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1887 read_unlock_bh(&table->tb6_lock);
1892 /* Create "default" multicast route to the interface */
1894 static void addrconf_add_mroute(struct net_device *dev)
1896 struct fib6_config cfg = {
1897 .fc_table = RT6_TABLE_LOCAL,
1898 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1899 .fc_ifindex = dev->ifindex,
1902 .fc_nlinfo.nl_net = dev_net(dev),
1905 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1907 ip6_route_add(&cfg);
1910 #if IS_ENABLED(CONFIG_IPV6_SIT)
1911 static void sit_route_add(struct net_device *dev)
1913 struct fib6_config cfg = {
1914 .fc_table = RT6_TABLE_MAIN,
1915 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1916 .fc_ifindex = dev->ifindex,
1918 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1919 .fc_nlinfo.nl_net = dev_net(dev),
1922 /* prefix length - 96 bits "::d.d.d.d" */
1923 ip6_route_add(&cfg);
1927 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1929 struct inet6_dev *idev;
1933 idev = ipv6_find_idev(dev);
1935 return ERR_PTR(-ENOBUFS);
1937 if (idev->cnf.disable_ipv6)
1938 return ERR_PTR(-EACCES);
1940 /* Add default multicast route */
1941 if (!(dev->flags & IFF_LOOPBACK))
1942 addrconf_add_mroute(dev);
1947 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1949 struct prefix_info *pinfo;
1953 struct inet6_dev *in6_dev;
1954 struct net *net = dev_net(dev);
1956 pinfo = (struct prefix_info *) opt;
1958 if (len < sizeof(struct prefix_info)) {
1959 ADBG(("addrconf: prefix option too short\n"));
1964 * Validation checks ([ADDRCONF], page 19)
1967 addr_type = ipv6_addr_type(&pinfo->prefix);
1969 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1972 valid_lft = ntohl(pinfo->valid);
1973 prefered_lft = ntohl(pinfo->prefered);
1975 if (prefered_lft > valid_lft) {
1976 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1980 in6_dev = in6_dev_get(dev);
1982 if (in6_dev == NULL) {
1983 net_dbg_ratelimited("addrconf: device %s not configured\n",
1989 * Two things going on here:
1990 * 1) Add routes for on-link prefixes
1991 * 2) Configure prefixes with the auto flag set
1994 if (pinfo->onlink) {
1995 struct rt6_info *rt;
1996 unsigned long rt_expires;
1998 /* Avoid arithmetic overflow. Really, we could
1999 * save rt_expires in seconds, likely valid_lft,
2000 * but it would require division in fib gc, that it
2004 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2006 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2008 if (addrconf_finite_timeout(rt_expires))
2011 rt = addrconf_get_prefix_route(&pinfo->prefix,
2014 RTF_ADDRCONF | RTF_PREFIX_RT,
2015 RTF_GATEWAY | RTF_DEFAULT);
2018 /* Autoconf prefix route */
2019 if (valid_lft == 0) {
2022 } else if (addrconf_finite_timeout(rt_expires)) {
2024 rt6_set_expires(rt, jiffies + rt_expires);
2026 rt6_clean_expires(rt);
2028 } else if (valid_lft) {
2029 clock_t expires = 0;
2030 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2031 if (addrconf_finite_timeout(rt_expires)) {
2033 flags |= RTF_EXPIRES;
2034 expires = jiffies_to_clock_t(rt_expires);
2036 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2037 dev, expires, flags);
2042 /* Try to figure out our local address for this prefix */
2044 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2045 struct inet6_ifaddr *ifp;
2046 struct in6_addr addr;
2047 int create = 0, update_lft = 0;
2049 if (pinfo->prefix_len == 64) {
2050 memcpy(&addr, &pinfo->prefix, 8);
2051 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2052 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2053 in6_dev_put(in6_dev);
2058 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2060 in6_dev_put(in6_dev);
2065 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2067 if (ifp == NULL && valid_lft) {
2068 int max_addresses = in6_dev->cnf.max_addresses;
2071 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2072 if (in6_dev->cnf.optimistic_dad &&
2073 !net->ipv6.devconf_all->forwarding && sllao)
2074 addr_flags = IFA_F_OPTIMISTIC;
2077 /* Do not allow to create too much of autoconfigured
2078 * addresses; this would be too easy way to crash kernel.
2080 if (!max_addresses ||
2081 ipv6_count_addresses(in6_dev) < max_addresses)
2082 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2083 addr_type&IPV6_ADDR_SCOPE_MASK,
2086 if (!ifp || IS_ERR(ifp)) {
2087 in6_dev_put(in6_dev);
2091 update_lft = create = 1;
2092 ifp->cstamp = jiffies;
2093 addrconf_dad_start(ifp);
2099 #ifdef CONFIG_IPV6_PRIVACY
2100 struct inet6_ifaddr *ift;
2104 /* update lifetime (RFC2462 5.5.3 e) */
2105 spin_lock(&ifp->lock);
2107 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2108 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2111 if (!update_lft && stored_lft) {
2112 if (valid_lft > MIN_VALID_LIFETIME ||
2113 valid_lft > stored_lft)
2115 else if (stored_lft <= MIN_VALID_LIFETIME) {
2116 /* valid_lft <= stored_lft is always true */
2118 * RFC 4862 Section 5.5.3e:
2119 * "Note that the preferred lifetime of
2120 * the corresponding address is always
2121 * reset to the Preferred Lifetime in
2122 * the received Prefix Information
2123 * option, regardless of whether the
2124 * valid lifetime is also reset or
2127 * So if the preferred lifetime in
2128 * this advertisement is different
2129 * than what we have stored, but the
2130 * valid lifetime is invalid, just
2131 * reset prefered_lft.
2133 * We must set the valid lifetime
2134 * to the stored lifetime since we'll
2135 * be updating the timestamp below,
2136 * else we'll set it back to the
2139 if (prefered_lft != ifp->prefered_lft) {
2140 valid_lft = stored_lft;
2144 valid_lft = MIN_VALID_LIFETIME;
2145 if (valid_lft < prefered_lft)
2146 prefered_lft = valid_lft;
2152 ifp->valid_lft = valid_lft;
2153 ifp->prefered_lft = prefered_lft;
2156 ifp->flags &= ~IFA_F_DEPRECATED;
2157 spin_unlock(&ifp->lock);
2159 if (!(flags&IFA_F_TENTATIVE))
2160 ipv6_ifa_notify(0, ifp);
2162 spin_unlock(&ifp->lock);
2164 #ifdef CONFIG_IPV6_PRIVACY
2165 read_lock_bh(&in6_dev->lock);
2166 /* update all temporary addresses in the list */
2167 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2169 int age, max_valid, max_prefered;
2171 if (ifp != ift->ifpub)
2175 * RFC 4941 section 3.3:
2176 * If a received option will extend the lifetime
2177 * of a public address, the lifetimes of
2178 * temporary addresses should be extended,
2179 * subject to the overall constraint that no
2180 * temporary addresses should ever remain
2181 * "valid" or "preferred" for a time longer than
2182 * (TEMP_VALID_LIFETIME) or
2183 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2186 age = (now - ift->cstamp) / HZ;
2187 max_valid = in6_dev->cnf.temp_valid_lft - age;
2191 max_prefered = in6_dev->cnf.temp_prefered_lft -
2192 in6_dev->cnf.max_desync_factor -
2194 if (max_prefered < 0)
2197 if (valid_lft > max_valid)
2198 valid_lft = max_valid;
2200 if (prefered_lft > max_prefered)
2201 prefered_lft = max_prefered;
2203 spin_lock(&ift->lock);
2205 ift->valid_lft = valid_lft;
2206 ift->prefered_lft = prefered_lft;
2208 if (prefered_lft > 0)
2209 ift->flags &= ~IFA_F_DEPRECATED;
2211 spin_unlock(&ift->lock);
2212 if (!(flags&IFA_F_TENTATIVE))
2213 ipv6_ifa_notify(0, ift);
2216 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2218 * When a new public address is created as
2219 * described in [ADDRCONF], also create a new
2220 * temporary address. Also create a temporary
2221 * address if it's enabled but no temporary
2222 * address currently exists.
2224 read_unlock_bh(&in6_dev->lock);
2225 ipv6_create_tempaddr(ifp, NULL);
2227 read_unlock_bh(&in6_dev->lock);
2234 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2235 in6_dev_put(in6_dev);
2239 * Set destination address.
2240 * Special case for SIT interfaces where we create a new "virtual"
2243 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2245 struct in6_ifreq ireq;
2246 struct net_device *dev;
2252 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2255 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2261 #if IS_ENABLED(CONFIG_IPV6_SIT)
2262 if (dev->type == ARPHRD_SIT) {
2263 const struct net_device_ops *ops = dev->netdev_ops;
2265 struct ip_tunnel_parm p;
2267 err = -EADDRNOTAVAIL;
2268 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2271 memset(&p, 0, sizeof(p));
2272 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2276 p.iph.protocol = IPPROTO_IPV6;
2278 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2280 if (ops->ndo_do_ioctl) {
2281 mm_segment_t oldfs = get_fs();
2284 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2291 dev = __dev_get_by_name(net, p.name);
2294 err = dev_open(dev);
2305 * Manual configuration of address on an interface
2307 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2308 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2311 struct inet6_ifaddr *ifp;
2312 struct inet6_dev *idev;
2313 struct net_device *dev;
2317 unsigned long timeout;
2324 /* check the lifetime */
2325 if (!valid_lft || prefered_lft > valid_lft)
2328 dev = __dev_get_by_index(net, ifindex);
2332 idev = addrconf_add_dev(dev);
2334 return PTR_ERR(idev);
2336 scope = ipv6_addr_scope(pfx);
2338 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2339 if (addrconf_finite_timeout(timeout)) {
2340 expires = jiffies_to_clock_t(timeout * HZ);
2341 valid_lft = timeout;
2342 flags = RTF_EXPIRES;
2346 ifa_flags |= IFA_F_PERMANENT;
2349 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2350 if (addrconf_finite_timeout(timeout)) {
2352 ifa_flags |= IFA_F_DEPRECATED;
2353 prefered_lft = timeout;
2356 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2359 spin_lock_bh(&ifp->lock);
2360 ifp->valid_lft = valid_lft;
2361 ifp->prefered_lft = prefered_lft;
2362 ifp->tstamp = jiffies;
2363 spin_unlock_bh(&ifp->lock);
2365 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2368 * Note that section 3.1 of RFC 4429 indicates
2369 * that the Optimistic flag should not be set for
2370 * manually configured addresses
2372 addrconf_dad_start(ifp);
2378 return PTR_ERR(ifp);
2381 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2384 struct inet6_ifaddr *ifp;
2385 struct inet6_dev *idev;
2386 struct net_device *dev;
2391 dev = __dev_get_by_index(net, ifindex);
2395 if ((idev = __in6_dev_get(dev)) == NULL)
2398 read_lock_bh(&idev->lock);
2399 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2400 if (ifp->prefix_len == plen &&
2401 ipv6_addr_equal(pfx, &ifp->addr)) {
2403 read_unlock_bh(&idev->lock);
2407 /* If the last address is deleted administratively,
2408 disable IPv6 on this interface.
2410 if (list_empty(&idev->addr_list))
2411 addrconf_ifdown(idev->dev, 1);
2415 read_unlock_bh(&idev->lock);
2416 return -EADDRNOTAVAIL;
2420 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2422 struct in6_ifreq ireq;
2425 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2428 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2432 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2433 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2434 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2439 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2441 struct in6_ifreq ireq;
2444 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2447 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2451 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2452 ireq.ifr6_prefixlen);
2457 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2458 int plen, int scope)
2460 struct inet6_ifaddr *ifp;
2462 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2464 spin_lock_bh(&ifp->lock);
2465 ifp->flags &= ~IFA_F_TENTATIVE;
2466 spin_unlock_bh(&ifp->lock);
2467 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2472 #if IS_ENABLED(CONFIG_IPV6_SIT)
2473 static void sit_add_v4_addrs(struct inet6_dev *idev)
2475 struct in6_addr addr;
2476 struct net_device *dev;
2477 struct net *net = dev_net(idev->dev);
2482 memset(&addr, 0, sizeof(struct in6_addr));
2483 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2485 if (idev->dev->flags&IFF_POINTOPOINT) {
2486 addr.s6_addr32[0] = htonl(0xfe800000);
2489 scope = IPV6_ADDR_COMPATv4;
2492 if (addr.s6_addr32[3]) {
2493 add_addr(idev, &addr, 128, scope);
2497 for_each_netdev(net, dev) {
2498 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2499 if (in_dev && (dev->flags & IFF_UP)) {
2500 struct in_ifaddr *ifa;
2504 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2507 addr.s6_addr32[3] = ifa->ifa_local;
2509 if (ifa->ifa_scope == RT_SCOPE_LINK)
2511 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2512 if (idev->dev->flags&IFF_POINTOPOINT)
2516 if (idev->dev->flags&IFF_POINTOPOINT)
2521 add_addr(idev, &addr, plen, flag);
2528 static void init_loopback(struct net_device *dev)
2530 struct inet6_dev *idev;
2536 if ((idev = ipv6_find_idev(dev)) == NULL) {
2537 pr_debug("%s: add_dev failed\n", __func__);
2541 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2544 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2546 struct inet6_ifaddr *ifp;
2547 u32 addr_flags = IFA_F_PERMANENT;
2549 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2550 if (idev->cnf.optimistic_dad &&
2551 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2552 addr_flags |= IFA_F_OPTIMISTIC;
2556 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2558 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2559 addrconf_dad_start(ifp);
2564 static void addrconf_dev_config(struct net_device *dev)
2566 struct in6_addr addr;
2567 struct inet6_dev *idev;
2571 if ((dev->type != ARPHRD_ETHER) &&
2572 (dev->type != ARPHRD_FDDI) &&
2573 (dev->type != ARPHRD_ARCNET) &&
2574 (dev->type != ARPHRD_INFINIBAND) &&
2575 (dev->type != ARPHRD_IEEE802154)) {
2576 /* Alas, we support only Ethernet autoconfiguration. */
2580 idev = addrconf_add_dev(dev);
2584 memset(&addr, 0, sizeof(struct in6_addr));
2585 addr.s6_addr32[0] = htonl(0xFE800000);
2587 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2588 addrconf_add_linklocal(idev, &addr);
2591 #if IS_ENABLED(CONFIG_IPV6_SIT)
2592 static void addrconf_sit_config(struct net_device *dev)
2594 struct inet6_dev *idev;
2599 * Configure the tunnel with one of our IPv4
2600 * addresses... we should configure all of
2601 * our v4 addrs in the tunnel
2604 if ((idev = ipv6_find_idev(dev)) == NULL) {
2605 pr_debug("%s: add_dev failed\n", __func__);
2609 if (dev->priv_flags & IFF_ISATAP) {
2610 struct in6_addr addr;
2612 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2613 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2614 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2615 addrconf_add_linklocal(idev, &addr);
2619 sit_add_v4_addrs(idev);
2621 if (dev->flags&IFF_POINTOPOINT)
2622 addrconf_add_mroute(dev);
2628 #if IS_ENABLED(CONFIG_NET_IPGRE)
2629 static void addrconf_gre_config(struct net_device *dev)
2631 struct inet6_dev *idev;
2632 struct in6_addr addr;
2634 pr_info("%s(%s)\n", __func__, dev->name);
2638 if ((idev = ipv6_find_idev(dev)) == NULL) {
2639 pr_debug("%s: add_dev failed\n", __func__);
2643 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2644 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2646 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2647 addrconf_add_linklocal(idev, &addr);
2652 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2654 struct in6_addr lladdr;
2656 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2657 addrconf_add_linklocal(idev, &lladdr);
2663 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2665 struct net_device *link_dev;
2666 struct net *net = dev_net(idev->dev);
2668 /* first try to inherit the link-local address from the link device */
2669 if (idev->dev->iflink &&
2670 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2671 if (!ipv6_inherit_linklocal(idev, link_dev))
2674 /* then try to inherit it from any device */
2675 for_each_netdev(net, link_dev) {
2676 if (!ipv6_inherit_linklocal(idev, link_dev))
2679 pr_debug("init ip6-ip6: add_linklocal failed\n");
2683 * Autoconfigure tunnel with a link-local address so routing protocols,
2684 * DHCPv6, MLD etc. can be run over the virtual link
2687 static void addrconf_ip6_tnl_config(struct net_device *dev)
2689 struct inet6_dev *idev;
2693 idev = addrconf_add_dev(dev);
2695 pr_debug("init ip6-ip6: add_dev failed\n");
2698 ip6_tnl_add_linklocal(idev);
2701 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2704 struct net_device *dev = (struct net_device *) data;
2705 struct inet6_dev *idev = __in6_dev_get(dev);
2706 int run_pending = 0;
2710 case NETDEV_REGISTER:
2711 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2712 idev = ipv6_add_dev(dev);
2714 return notifier_from_errno(-ENOMEM);
2720 if (dev->flags & IFF_SLAVE)
2723 if (event == NETDEV_UP) {
2724 if (!addrconf_qdisc_ok(dev)) {
2725 /* device is not ready yet. */
2726 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2731 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2732 idev = ipv6_add_dev(dev);
2735 idev->if_flags |= IF_READY;
2739 if (!addrconf_qdisc_ok(dev)) {
2740 /* device is still not ready. */
2745 if (idev->if_flags & IF_READY)
2746 /* device is already configured. */
2748 idev->if_flags |= IF_READY;
2751 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2757 switch (dev->type) {
2758 #if IS_ENABLED(CONFIG_IPV6_SIT)
2760 addrconf_sit_config(dev);
2763 #if IS_ENABLED(CONFIG_NET_IPGRE)
2765 addrconf_gre_config(dev);
2768 case ARPHRD_TUNNEL6:
2769 addrconf_ip6_tnl_config(dev);
2771 case ARPHRD_LOOPBACK:
2776 addrconf_dev_config(dev);
2782 addrconf_dad_run(idev);
2785 * If the MTU changed during the interface down,
2786 * when the interface up, the changed MTU must be
2787 * reflected in the idev as well as routers.
2789 if (idev->cnf.mtu6 != dev->mtu &&
2790 dev->mtu >= IPV6_MIN_MTU) {
2791 rt6_mtu_change(dev, dev->mtu);
2792 idev->cnf.mtu6 = dev->mtu;
2794 idev->tstamp = jiffies;
2795 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2798 * If the changed mtu during down is lower than
2799 * IPV6_MIN_MTU stop IPv6 on this interface.
2801 if (dev->mtu < IPV6_MIN_MTU)
2802 addrconf_ifdown(dev, 1);
2806 case NETDEV_CHANGEMTU:
2807 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2808 rt6_mtu_change(dev, dev->mtu);
2809 idev->cnf.mtu6 = dev->mtu;
2813 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2814 idev = ipv6_add_dev(dev);
2820 * MTU falled under IPV6_MIN_MTU.
2821 * Stop IPv6 on this interface.
2825 case NETDEV_UNREGISTER:
2827 * Remove all addresses from this interface.
2829 addrconf_ifdown(dev, event != NETDEV_DOWN);
2832 case NETDEV_CHANGENAME:
2834 snmp6_unregister_dev(idev);
2835 addrconf_sysctl_unregister(idev);
2836 addrconf_sysctl_register(idev);
2837 err = snmp6_register_dev(idev);
2839 return notifier_from_errno(err);
2843 case NETDEV_PRE_TYPE_CHANGE:
2844 case NETDEV_POST_TYPE_CHANGE:
2845 addrconf_type_change(dev, event);
2853 * addrconf module should be notified of a device going up
2855 static struct notifier_block ipv6_dev_notf = {
2856 .notifier_call = addrconf_notify,
2859 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2861 struct inet6_dev *idev;
2864 idev = __in6_dev_get(dev);
2866 if (event == NETDEV_POST_TYPE_CHANGE)
2867 ipv6_mc_remap(idev);
2868 else if (event == NETDEV_PRE_TYPE_CHANGE)
2869 ipv6_mc_unmap(idev);
2872 static int addrconf_ifdown(struct net_device *dev, int how)
2874 struct net *net = dev_net(dev);
2875 struct inet6_dev *idev;
2876 struct inet6_ifaddr *ifa;
2881 rt6_ifdown(net, dev);
2882 neigh_ifdown(&nd_tbl, dev);
2884 idev = __in6_dev_get(dev);
2889 * Step 1: remove reference to ipv6 device from parent device.
2895 /* protected by rtnl_lock */
2896 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2898 /* Step 1.5: remove snmp6 entry */
2899 snmp6_unregister_dev(idev);
2903 /* Step 2: clear hash table */
2904 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2905 struct hlist_head *h = &inet6_addr_lst[i];
2906 struct hlist_node *n;
2908 spin_lock_bh(&addrconf_hash_lock);
2910 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2911 if (ifa->idev == idev) {
2912 hlist_del_init_rcu(&ifa->addr_lst);
2913 addrconf_del_timer(ifa);
2917 spin_unlock_bh(&addrconf_hash_lock);
2920 write_lock_bh(&idev->lock);
2922 /* Step 2: clear flags for stateless addrconf */
2924 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2926 #ifdef CONFIG_IPV6_PRIVACY
2927 if (how && del_timer(&idev->regen_timer))
2930 /* Step 3: clear tempaddr list */
2931 while (!list_empty(&idev->tempaddr_list)) {
2932 ifa = list_first_entry(&idev->tempaddr_list,
2933 struct inet6_ifaddr, tmp_list);
2934 list_del(&ifa->tmp_list);
2935 write_unlock_bh(&idev->lock);
2936 spin_lock_bh(&ifa->lock);
2939 in6_ifa_put(ifa->ifpub);
2942 spin_unlock_bh(&ifa->lock);
2944 write_lock_bh(&idev->lock);
2948 while (!list_empty(&idev->addr_list)) {
2949 ifa = list_first_entry(&idev->addr_list,
2950 struct inet6_ifaddr, if_list);
2951 addrconf_del_timer(ifa);
2953 list_del(&ifa->if_list);
2955 write_unlock_bh(&idev->lock);
2957 spin_lock_bh(&ifa->state_lock);
2959 ifa->state = INET6_IFADDR_STATE_DEAD;
2960 spin_unlock_bh(&ifa->state_lock);
2962 if (state != INET6_IFADDR_STATE_DEAD) {
2963 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2964 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2968 write_lock_bh(&idev->lock);
2971 write_unlock_bh(&idev->lock);
2973 /* Step 5: Discard multicast list */
2975 ipv6_mc_destroy_dev(idev);
2979 idev->tstamp = jiffies;
2981 /* Last: Shot the device (if unregistered) */
2983 addrconf_sysctl_unregister(idev);
2984 neigh_parms_release(&nd_tbl, idev->nd_parms);
2985 neigh_ifdown(&nd_tbl, dev);
2991 static void addrconf_rs_timer(unsigned long data)
2993 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2994 struct inet6_dev *idev = ifp->idev;
2996 read_lock(&idev->lock);
2997 if (idev->dead || !(idev->if_flags & IF_READY))
3000 if (!ipv6_accept_ra(idev))
3003 /* Announcement received after solicitation was sent */
3004 if (idev->if_flags & IF_RA_RCVD)
3007 spin_lock(&ifp->lock);
3008 if (ifp->probes++ < idev->cnf.rtr_solicits) {
3009 /* The wait after the last probe can be shorter */
3010 addrconf_mod_timer(ifp, AC_RS,
3011 (ifp->probes == idev->cnf.rtr_solicits) ?
3012 idev->cnf.rtr_solicit_delay :
3013 idev->cnf.rtr_solicit_interval);
3014 spin_unlock(&ifp->lock);
3016 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3018 spin_unlock(&ifp->lock);
3020 * Note: we do not support deprecated "all on-link"
3021 * assumption any longer.
3023 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3027 read_unlock(&idev->lock);
3032 * Duplicate Address Detection
3034 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3036 unsigned long rand_num;
3037 struct inet6_dev *idev = ifp->idev;
3039 if (ifp->flags & IFA_F_OPTIMISTIC)
3042 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3044 ifp->probes = idev->cnf.dad_transmits;
3045 addrconf_mod_timer(ifp, AC_DAD, rand_num);
3048 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3050 struct inet6_dev *idev = ifp->idev;
3051 struct net_device *dev = idev->dev;
3053 addrconf_join_solict(dev, &ifp->addr);
3055 net_srandom(ifp->addr.s6_addr32[3]);
3057 read_lock_bh(&idev->lock);
3058 spin_lock(&ifp->lock);
3059 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3062 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3063 idev->cnf.accept_dad < 1 ||
3064 !(ifp->flags&IFA_F_TENTATIVE) ||
3065 ifp->flags & IFA_F_NODAD) {
3066 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3067 spin_unlock(&ifp->lock);
3068 read_unlock_bh(&idev->lock);
3070 addrconf_dad_completed(ifp);
3074 if (!(idev->if_flags & IF_READY)) {
3075 spin_unlock(&ifp->lock);
3076 read_unlock_bh(&idev->lock);
3078 * If the device is not ready:
3079 * - keep it tentative if it is a permanent address.
3080 * - otherwise, kill it.
3083 addrconf_dad_stop(ifp, 0);
3088 * Optimistic nodes can start receiving
3091 if (ifp->flags & IFA_F_OPTIMISTIC)
3092 ip6_ins_rt(ifp->rt);
3094 addrconf_dad_kick(ifp);
3096 spin_unlock(&ifp->lock);
3097 read_unlock_bh(&idev->lock);
3100 static void addrconf_dad_timer(unsigned long data)
3102 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3103 struct inet6_dev *idev = ifp->idev;
3104 struct in6_addr mcaddr;
3106 if (!ifp->probes && addrconf_dad_end(ifp))
3109 read_lock(&idev->lock);
3110 if (idev->dead || !(idev->if_flags & IF_READY)) {
3111 read_unlock(&idev->lock);
3115 spin_lock(&ifp->lock);
3116 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3117 spin_unlock(&ifp->lock);
3118 read_unlock(&idev->lock);
3122 if (ifp->probes == 0) {
3124 * DAD was successful
3127 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3128 spin_unlock(&ifp->lock);
3129 read_unlock(&idev->lock);
3131 addrconf_dad_completed(ifp);
3137 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3138 spin_unlock(&ifp->lock);
3139 read_unlock(&idev->lock);
3141 /* send a neighbour solicitation for our addr */
3142 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3143 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3148 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3150 struct net_device *dev = ifp->idev->dev;
3153 * Configure the address for reception. Now it is valid.
3156 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3158 /* If added prefix is link local and we are prepared to process
3159 router advertisements, start sending router solicitations.
3162 if (ipv6_accept_ra(ifp->idev) &&
3163 ifp->idev->cnf.rtr_solicits > 0 &&
3164 (dev->flags&IFF_LOOPBACK) == 0 &&
3165 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3167 * If a host as already performed a random delay
3168 * [...] as part of DAD [...] there is no need
3169 * to delay again before sending the first RS
3171 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3173 spin_lock_bh(&ifp->lock);
3175 ifp->idev->if_flags |= IF_RS_SENT;
3176 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3177 spin_unlock_bh(&ifp->lock);
3181 static void addrconf_dad_run(struct inet6_dev *idev)
3183 struct inet6_ifaddr *ifp;
3185 read_lock_bh(&idev->lock);
3186 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3187 spin_lock(&ifp->lock);
3188 if (ifp->flags & IFA_F_TENTATIVE &&
3189 ifp->state == INET6_IFADDR_STATE_DAD)
3190 addrconf_dad_kick(ifp);
3191 spin_unlock(&ifp->lock);
3193 read_unlock_bh(&idev->lock);
3196 #ifdef CONFIG_PROC_FS
3197 struct if6_iter_state {
3198 struct seq_net_private p;
3203 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3205 struct inet6_ifaddr *ifa = NULL;
3206 struct if6_iter_state *state = seq->private;
3207 struct net *net = seq_file_net(seq);
3210 /* initial bucket if pos is 0 */
3216 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3217 struct hlist_node *n;
3218 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3220 if (!net_eq(dev_net(ifa->idev->dev), net))
3222 /* sync with offset */
3223 if (p < state->offset) {
3231 /* prepare for next bucket */
3238 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3239 struct inet6_ifaddr *ifa)
3241 struct if6_iter_state *state = seq->private;
3242 struct net *net = seq_file_net(seq);
3243 struct hlist_node *n = &ifa->addr_lst;
3245 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3246 if (!net_eq(dev_net(ifa->idev->dev), net))
3252 while (++state->bucket < IN6_ADDR_HSIZE) {
3254 hlist_for_each_entry_rcu_bh(ifa, n,
3255 &inet6_addr_lst[state->bucket], addr_lst) {
3256 if (!net_eq(dev_net(ifa->idev->dev), net))
3266 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3270 return if6_get_first(seq, *pos);
3273 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3275 struct inet6_ifaddr *ifa;
3277 ifa = if6_get_next(seq, v);
3282 static void if6_seq_stop(struct seq_file *seq, void *v)
3285 rcu_read_unlock_bh();
3288 static int if6_seq_show(struct seq_file *seq, void *v)
3290 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3291 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3293 ifp->idev->dev->ifindex,
3297 ifp->idev->dev->name);
3301 static const struct seq_operations if6_seq_ops = {
3302 .start = if6_seq_start,
3303 .next = if6_seq_next,
3304 .show = if6_seq_show,
3305 .stop = if6_seq_stop,
3308 static int if6_seq_open(struct inode *inode, struct file *file)
3310 return seq_open_net(inode, file, &if6_seq_ops,
3311 sizeof(struct if6_iter_state));
3314 static const struct file_operations if6_fops = {
3315 .owner = THIS_MODULE,
3316 .open = if6_seq_open,
3318 .llseek = seq_lseek,
3319 .release = seq_release_net,
3322 static int __net_init if6_proc_net_init(struct net *net)
3324 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3329 static void __net_exit if6_proc_net_exit(struct net *net)
3331 proc_net_remove(net, "if_inet6");
3334 static struct pernet_operations if6_proc_net_ops = {
3335 .init = if6_proc_net_init,
3336 .exit = if6_proc_net_exit,
3339 int __init if6_proc_init(void)
3341 return register_pernet_subsys(&if6_proc_net_ops);
3344 void if6_proc_exit(void)
3346 unregister_pernet_subsys(&if6_proc_net_ops);
3348 #endif /* CONFIG_PROC_FS */
3350 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3351 /* Check if address is a home address configured on any interface. */
3352 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3355 struct inet6_ifaddr *ifp = NULL;
3356 struct hlist_node *n;
3357 unsigned int hash = inet6_addr_hash(addr);
3360 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3361 if (!net_eq(dev_net(ifp->idev->dev), net))
3363 if (ipv6_addr_equal(&ifp->addr, addr) &&
3364 (ifp->flags & IFA_F_HOMEADDRESS)) {
3369 rcu_read_unlock_bh();
3375 * Periodic address status verification
3378 static void addrconf_verify(unsigned long foo)
3380 unsigned long now, next, next_sec, next_sched;
3381 struct inet6_ifaddr *ifp;
3382 struct hlist_node *node;
3386 spin_lock(&addrconf_verify_lock);
3388 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3390 del_timer(&addr_chk_timer);
3392 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3394 hlist_for_each_entry_rcu_bh(ifp, node,
3395 &inet6_addr_lst[i], addr_lst) {
3398 if (ifp->flags & IFA_F_PERMANENT)
3401 spin_lock(&ifp->lock);
3402 /* We try to batch several events at once. */
3403 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3405 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3406 age >= ifp->valid_lft) {
3407 spin_unlock(&ifp->lock);
3411 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3412 spin_unlock(&ifp->lock);
3414 } else if (age >= ifp->prefered_lft) {
3415 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3418 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3420 ifp->flags |= IFA_F_DEPRECATED;
3423 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3424 next = ifp->tstamp + ifp->valid_lft * HZ;
3426 spin_unlock(&ifp->lock);
3431 ipv6_ifa_notify(0, ifp);
3435 #ifdef CONFIG_IPV6_PRIVACY
3436 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3437 !(ifp->flags&IFA_F_TENTATIVE)) {
3438 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3439 ifp->idev->cnf.dad_transmits *
3440 ifp->idev->nd_parms->retrans_time / HZ;
3442 if (age >= ifp->prefered_lft - regen_advance) {
3443 struct inet6_ifaddr *ifpub = ifp->ifpub;
3444 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3445 next = ifp->tstamp + ifp->prefered_lft * HZ;
3446 if (!ifp->regen_count && ifpub) {
3449 in6_ifa_hold(ifpub);
3450 spin_unlock(&ifp->lock);
3452 spin_lock(&ifpub->lock);
3453 ifpub->regen_count = 0;
3454 spin_unlock(&ifpub->lock);
3455 ipv6_create_tempaddr(ifpub, ifp);
3460 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3461 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3462 spin_unlock(&ifp->lock);
3465 /* ifp->prefered_lft <= ifp->valid_lft */
3466 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3467 next = ifp->tstamp + ifp->prefered_lft * HZ;
3468 spin_unlock(&ifp->lock);
3473 next_sec = round_jiffies_up(next);
3476 /* If rounded timeout is accurate enough, accept it. */
3477 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3478 next_sched = next_sec;
3480 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3481 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3482 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3484 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3485 now, next, next_sec, next_sched));
3487 addr_chk_timer.expires = next_sched;
3488 add_timer(&addr_chk_timer);
3489 spin_unlock(&addrconf_verify_lock);
3490 rcu_read_unlock_bh();
3493 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3495 struct in6_addr *pfx = NULL;
3498 pfx = nla_data(addr);
3501 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3504 pfx = nla_data(local);
3510 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3511 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3512 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3513 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3517 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3519 struct net *net = sock_net(skb->sk);
3520 struct ifaddrmsg *ifm;
3521 struct nlattr *tb[IFA_MAX+1];
3522 struct in6_addr *pfx;
3525 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3529 ifm = nlmsg_data(nlh);
3530 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3534 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3537 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3538 u32 prefered_lft, u32 valid_lft)
3542 unsigned long timeout;
3544 if (!valid_lft || (prefered_lft > valid_lft))
3547 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3548 if (addrconf_finite_timeout(timeout)) {
3549 expires = jiffies_to_clock_t(timeout * HZ);
3550 valid_lft = timeout;
3551 flags = RTF_EXPIRES;
3555 ifa_flags |= IFA_F_PERMANENT;
3558 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3559 if (addrconf_finite_timeout(timeout)) {
3561 ifa_flags |= IFA_F_DEPRECATED;
3562 prefered_lft = timeout;
3565 spin_lock_bh(&ifp->lock);
3566 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3567 ifp->tstamp = jiffies;
3568 ifp->valid_lft = valid_lft;
3569 ifp->prefered_lft = prefered_lft;
3571 spin_unlock_bh(&ifp->lock);
3572 if (!(ifp->flags&IFA_F_TENTATIVE))
3573 ipv6_ifa_notify(0, ifp);
3575 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3583 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3585 struct net *net = sock_net(skb->sk);
3586 struct ifaddrmsg *ifm;
3587 struct nlattr *tb[IFA_MAX+1];
3588 struct in6_addr *pfx;
3589 struct inet6_ifaddr *ifa;
3590 struct net_device *dev;
3591 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3595 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3599 ifm = nlmsg_data(nlh);
3600 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3604 if (tb[IFA_CACHEINFO]) {
3605 struct ifa_cacheinfo *ci;
3607 ci = nla_data(tb[IFA_CACHEINFO]);
3608 valid_lft = ci->ifa_valid;
3609 preferred_lft = ci->ifa_prefered;
3611 preferred_lft = INFINITY_LIFE_TIME;
3612 valid_lft = INFINITY_LIFE_TIME;
3615 dev = __dev_get_by_index(net, ifm->ifa_index);
3619 /* We ignore other flags so far. */
3620 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3622 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3625 * It would be best to check for !NLM_F_CREATE here but
3626 * userspace alreay relies on not having to provide this.
3628 return inet6_addr_add(net, ifm->ifa_index, pfx,
3629 ifm->ifa_prefixlen, ifa_flags,
3630 preferred_lft, valid_lft);
3633 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3634 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3637 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3644 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3645 u8 scope, int ifindex)
3647 struct ifaddrmsg *ifm;
3649 ifm = nlmsg_data(nlh);
3650 ifm->ifa_family = AF_INET6;
3651 ifm->ifa_prefixlen = prefixlen;
3652 ifm->ifa_flags = flags;
3653 ifm->ifa_scope = scope;
3654 ifm->ifa_index = ifindex;
3657 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3658 unsigned long tstamp, u32 preferred, u32 valid)
3660 struct ifa_cacheinfo ci;
3662 ci.cstamp = cstamp_delta(cstamp);
3663 ci.tstamp = cstamp_delta(tstamp);
3664 ci.ifa_prefered = preferred;
3665 ci.ifa_valid = valid;
3667 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3670 static inline int rt_scope(int ifa_scope)
3672 if (ifa_scope & IFA_HOST)
3673 return RT_SCOPE_HOST;
3674 else if (ifa_scope & IFA_LINK)
3675 return RT_SCOPE_LINK;
3676 else if (ifa_scope & IFA_SITE)
3677 return RT_SCOPE_SITE;
3679 return RT_SCOPE_UNIVERSE;
3682 static inline int inet6_ifaddr_msgsize(void)
3684 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3685 + nla_total_size(16) /* IFA_ADDRESS */
3686 + nla_total_size(sizeof(struct ifa_cacheinfo));
3689 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3690 u32 portid, u32 seq, int event, unsigned int flags)
3692 struct nlmsghdr *nlh;
3693 u32 preferred, valid;
3695 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3699 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3700 ifa->idev->dev->ifindex);
3702 if (!(ifa->flags&IFA_F_PERMANENT)) {
3703 preferred = ifa->prefered_lft;
3704 valid = ifa->valid_lft;
3705 if (preferred != INFINITY_LIFE_TIME) {
3706 long tval = (jiffies - ifa->tstamp)/HZ;
3707 if (preferred > tval)
3711 if (valid != INFINITY_LIFE_TIME) {
3719 preferred = INFINITY_LIFE_TIME;
3720 valid = INFINITY_LIFE_TIME;
3723 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3724 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3725 nlmsg_cancel(skb, nlh);
3729 return nlmsg_end(skb, nlh);
3732 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3733 u32 portid, u32 seq, int event, u16 flags)
3735 struct nlmsghdr *nlh;
3736 u8 scope = RT_SCOPE_UNIVERSE;
3737 int ifindex = ifmca->idev->dev->ifindex;
3739 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3740 scope = RT_SCOPE_SITE;
3742 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3746 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3747 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3748 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3749 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3750 nlmsg_cancel(skb, nlh);
3754 return nlmsg_end(skb, nlh);
3757 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3758 u32 portid, u32 seq, int event, unsigned int flags)
3760 struct nlmsghdr *nlh;
3761 u8 scope = RT_SCOPE_UNIVERSE;
3762 int ifindex = ifaca->aca_idev->dev->ifindex;
3764 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3765 scope = RT_SCOPE_SITE;
3767 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3771 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3772 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3773 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3774 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3775 nlmsg_cancel(skb, nlh);
3779 return nlmsg_end(skb, nlh);
3788 /* called with rcu_read_lock() */
3789 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3790 struct netlink_callback *cb, enum addr_type_t type,
3791 int s_ip_idx, int *p_ip_idx)
3793 struct ifmcaddr6 *ifmca;
3794 struct ifacaddr6 *ifaca;
3796 int ip_idx = *p_ip_idx;
3798 read_lock_bh(&idev->lock);
3800 case UNICAST_ADDR: {
3801 struct inet6_ifaddr *ifa;
3803 /* unicast address incl. temp addr */
3804 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3805 if (++ip_idx < s_ip_idx)
3807 err = inet6_fill_ifaddr(skb, ifa,
3808 NETLINK_CB(cb->skb).portid,
3817 case MULTICAST_ADDR:
3818 /* multicast address */
3819 for (ifmca = idev->mc_list; ifmca;
3820 ifmca = ifmca->next, ip_idx++) {
3821 if (ip_idx < s_ip_idx)
3823 err = inet6_fill_ifmcaddr(skb, ifmca,
3824 NETLINK_CB(cb->skb).portid,
3833 /* anycast address */
3834 for (ifaca = idev->ac_list; ifaca;
3835 ifaca = ifaca->aca_next, ip_idx++) {
3836 if (ip_idx < s_ip_idx)
3838 err = inet6_fill_ifacaddr(skb, ifaca,
3839 NETLINK_CB(cb->skb).portid,
3850 read_unlock_bh(&idev->lock);
3855 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3856 enum addr_type_t type)
3858 struct net *net = sock_net(skb->sk);
3861 int s_idx, s_ip_idx;
3862 struct net_device *dev;
3863 struct inet6_dev *idev;
3864 struct hlist_head *head;
3865 struct hlist_node *node;
3868 s_idx = idx = cb->args[1];
3869 s_ip_idx = ip_idx = cb->args[2];
3872 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3874 head = &net->dev_index_head[h];
3875 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3878 if (h > s_h || idx > s_idx)
3881 idev = __in6_dev_get(dev);
3885 if (in6_dump_addrs(idev, skb, cb, type,
3886 s_ip_idx, &ip_idx) <= 0)
3896 cb->args[2] = ip_idx;
3901 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3903 enum addr_type_t type = UNICAST_ADDR;
3905 return inet6_dump_addr(skb, cb, type);
3908 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3910 enum addr_type_t type = MULTICAST_ADDR;
3912 return inet6_dump_addr(skb, cb, type);
3916 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3918 enum addr_type_t type = ANYCAST_ADDR;
3920 return inet6_dump_addr(skb, cb, type);
3923 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3926 struct net *net = sock_net(in_skb->sk);
3927 struct ifaddrmsg *ifm;
3928 struct nlattr *tb[IFA_MAX+1];
3929 struct in6_addr *addr = NULL;
3930 struct net_device *dev = NULL;
3931 struct inet6_ifaddr *ifa;
3932 struct sk_buff *skb;
3935 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3939 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3945 ifm = nlmsg_data(nlh);
3947 dev = __dev_get_by_index(net, ifm->ifa_index);
3949 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3951 err = -EADDRNOTAVAIL;
3955 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3961 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3962 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3964 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3965 WARN_ON(err == -EMSGSIZE);
3969 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3976 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3978 struct sk_buff *skb;
3979 struct net *net = dev_net(ifa->idev->dev);
3982 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3986 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3988 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3989 WARN_ON(err == -EMSGSIZE);
3993 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3997 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4000 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4001 __s32 *array, int bytes)
4003 BUG_ON(bytes < (DEVCONF_MAX * 4));
4005 memset(array, 0, bytes);
4006 array[DEVCONF_FORWARDING] = cnf->forwarding;
4007 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4008 array[DEVCONF_MTU6] = cnf->mtu6;
4009 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4010 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4011 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4012 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4013 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4014 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4015 jiffies_to_msecs(cnf->rtr_solicit_interval);
4016 array[DEVCONF_RTR_SOLICIT_DELAY] =
4017 jiffies_to_msecs(cnf->rtr_solicit_delay);
4018 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4019 #ifdef CONFIG_IPV6_PRIVACY
4020 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4021 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4022 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4023 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4024 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4026 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4027 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4028 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4029 #ifdef CONFIG_IPV6_ROUTER_PREF
4030 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4031 array[DEVCONF_RTR_PROBE_INTERVAL] =
4032 jiffies_to_msecs(cnf->rtr_probe_interval);
4033 #ifdef CONFIG_IPV6_ROUTE_INFO
4034 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4037 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4038 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4039 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4040 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4042 #ifdef CONFIG_IPV6_MROUTE
4043 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4045 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4046 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4047 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4048 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4051 static inline size_t inet6_ifla6_size(void)
4053 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4054 + nla_total_size(sizeof(struct ifla_cacheinfo))
4055 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4056 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4057 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4060 static inline size_t inet6_if_nlmsg_size(void)
4062 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4063 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4064 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4065 + nla_total_size(4) /* IFLA_MTU */
4066 + nla_total_size(4) /* IFLA_LINK */
4067 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4070 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4071 int items, int bytes)
4074 int pad = bytes - sizeof(u64) * items;
4077 /* Use put_unaligned() because stats may not be aligned for u64. */
4078 put_unaligned(items, &stats[0]);
4079 for (i = 1; i < items; i++)
4080 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4082 memset(&stats[items], 0, pad);
4085 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4086 int items, int bytes, size_t syncpoff)
4089 int pad = bytes - sizeof(u64) * items;
4092 /* Use put_unaligned() because stats may not be aligned for u64. */
4093 put_unaligned(items, &stats[0]);
4094 for (i = 1; i < items; i++)
4095 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4097 memset(&stats[items], 0, pad);
4100 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4104 case IFLA_INET6_STATS:
4105 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4106 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4108 case IFLA_INET6_ICMP6STATS:
4109 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4114 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4117 struct ifla_cacheinfo ci;
4119 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4120 goto nla_put_failure;
4121 ci.max_reasm_len = IPV6_MAXPLEN;
4122 ci.tstamp = cstamp_delta(idev->tstamp);
4123 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4124 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4125 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4126 goto nla_put_failure;
4127 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4129 goto nla_put_failure;
4130 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4132 /* XXX - MC not implemented */
4134 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4136 goto nla_put_failure;
4137 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4139 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4141 goto nla_put_failure;
4142 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4150 static size_t inet6_get_link_af_size(const struct net_device *dev)
4152 if (!__in6_dev_get(dev))
4155 return inet6_ifla6_size();
4158 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4160 struct inet6_dev *idev = __in6_dev_get(dev);
4165 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4171 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4172 u32 portid, u32 seq, int event, unsigned int flags)
4174 struct net_device *dev = idev->dev;
4175 struct ifinfomsg *hdr;
4176 struct nlmsghdr *nlh;
4179 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4183 hdr = nlmsg_data(nlh);
4184 hdr->ifi_family = AF_INET6;
4186 hdr->ifi_type = dev->type;
4187 hdr->ifi_index = dev->ifindex;
4188 hdr->ifi_flags = dev_get_flags(dev);
4189 hdr->ifi_change = 0;
4191 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4193 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4194 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4195 (dev->ifindex != dev->iflink &&
4196 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4197 goto nla_put_failure;
4198 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4199 if (protoinfo == NULL)
4200 goto nla_put_failure;
4202 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4203 goto nla_put_failure;
4205 nla_nest_end(skb, protoinfo);
4206 return nlmsg_end(skb, nlh);
4209 nlmsg_cancel(skb, nlh);
4213 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4215 struct net *net = sock_net(skb->sk);
4218 struct net_device *dev;
4219 struct inet6_dev *idev;
4220 struct hlist_head *head;
4221 struct hlist_node *node;
4224 s_idx = cb->args[1];
4227 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4229 head = &net->dev_index_head[h];
4230 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4233 idev = __in6_dev_get(dev);
4236 if (inet6_fill_ifinfo(skb, idev,
4237 NETLINK_CB(cb->skb).portid,
4239 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4253 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4255 struct sk_buff *skb;
4256 struct net *net = dev_net(idev->dev);
4259 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4263 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4265 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4266 WARN_ON(err == -EMSGSIZE);
4270 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4274 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4277 static inline size_t inet6_prefix_nlmsg_size(void)
4279 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4280 + nla_total_size(sizeof(struct in6_addr))
4281 + nla_total_size(sizeof(struct prefix_cacheinfo));
4284 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4285 struct prefix_info *pinfo, u32 portid, u32 seq,
4286 int event, unsigned int flags)
4288 struct prefixmsg *pmsg;
4289 struct nlmsghdr *nlh;
4290 struct prefix_cacheinfo ci;
4292 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4296 pmsg = nlmsg_data(nlh);
4297 pmsg->prefix_family = AF_INET6;
4298 pmsg->prefix_pad1 = 0;
4299 pmsg->prefix_pad2 = 0;
4300 pmsg->prefix_ifindex = idev->dev->ifindex;
4301 pmsg->prefix_len = pinfo->prefix_len;
4302 pmsg->prefix_type = pinfo->type;
4303 pmsg->prefix_pad3 = 0;
4304 pmsg->prefix_flags = 0;
4306 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4307 if (pinfo->autoconf)
4308 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4310 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4311 goto nla_put_failure;
4312 ci.preferred_time = ntohl(pinfo->prefered);
4313 ci.valid_time = ntohl(pinfo->valid);
4314 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4315 goto nla_put_failure;
4316 return nlmsg_end(skb, nlh);
4319 nlmsg_cancel(skb, nlh);
4323 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4324 struct prefix_info *pinfo)
4326 struct sk_buff *skb;
4327 struct net *net = dev_net(idev->dev);
4330 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4334 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4336 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4337 WARN_ON(err == -EMSGSIZE);
4341 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4345 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4348 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4350 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4355 * If the address was optimistic
4356 * we inserted the route at the start of
4357 * our DAD process, so we don't need
4360 if (!(ifp->rt->rt6i_node))
4361 ip6_ins_rt(ifp->rt);
4362 if (ifp->idev->cnf.forwarding)
4363 addrconf_join_anycast(ifp);
4366 if (ifp->idev->cnf.forwarding)
4367 addrconf_leave_anycast(ifp);
4368 addrconf_leave_solict(ifp->idev, &ifp->addr);
4369 dst_hold(&ifp->rt->dst);
4371 if (ip6_del_rt(ifp->rt))
4372 dst_free(&ifp->rt->dst);
4377 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4380 if (likely(ifp->idev->dead == 0))
4381 __ipv6_ifa_notify(event, ifp);
4382 rcu_read_unlock_bh();
4385 #ifdef CONFIG_SYSCTL
4388 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4389 void __user *buffer, size_t *lenp, loff_t *ppos)
4391 int *valp = ctl->data;
4398 * ctl->data points to idev->cnf.forwarding, we should
4399 * not modify it until we get the rtnl lock.
4404 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4407 ret = addrconf_fixup_forwarding(ctl, valp, val);
4413 static void dev_disable_change(struct inet6_dev *idev)
4415 if (!idev || !idev->dev)
4418 if (idev->cnf.disable_ipv6)
4419 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4421 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4424 static void addrconf_disable_change(struct net *net, __s32 newf)
4426 struct net_device *dev;
4427 struct inet6_dev *idev;
4430 for_each_netdev_rcu(net, dev) {
4431 idev = __in6_dev_get(dev);
4433 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4434 idev->cnf.disable_ipv6 = newf;
4436 dev_disable_change(idev);
4442 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4447 if (!rtnl_trylock())
4448 return restart_syscall();
4450 net = (struct net *)table->extra2;
4454 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4459 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4460 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4461 addrconf_disable_change(net, newf);
4462 } else if ((!newf) ^ (!old))
4463 dev_disable_change((struct inet6_dev *)table->extra1);
4470 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4471 void __user *buffer, size_t *lenp, loff_t *ppos)
4473 int *valp = ctl->data;
4480 * ctl->data points to idev->cnf.disable_ipv6, we should
4481 * not modify it until we get the rtnl lock.
4486 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4489 ret = addrconf_disable_ipv6(ctl, valp, val);
4495 static struct addrconf_sysctl_table
4497 struct ctl_table_header *sysctl_header;
4498 ctl_table addrconf_vars[DEVCONF_MAX+1];
4499 } addrconf_sysctl __read_mostly = {
4500 .sysctl_header = NULL,
4503 .procname = "forwarding",
4504 .data = &ipv6_devconf.forwarding,
4505 .maxlen = sizeof(int),
4507 .proc_handler = addrconf_sysctl_forward,
4510 .procname = "hop_limit",
4511 .data = &ipv6_devconf.hop_limit,
4512 .maxlen = sizeof(int),
4514 .proc_handler = proc_dointvec,
4518 .data = &ipv6_devconf.mtu6,
4519 .maxlen = sizeof(int),
4521 .proc_handler = proc_dointvec,
4524 .procname = "accept_ra",
4525 .data = &ipv6_devconf.accept_ra,
4526 .maxlen = sizeof(int),
4528 .proc_handler = proc_dointvec,
4531 .procname = "accept_redirects",
4532 .data = &ipv6_devconf.accept_redirects,
4533 .maxlen = sizeof(int),
4535 .proc_handler = proc_dointvec,
4538 .procname = "autoconf",
4539 .data = &ipv6_devconf.autoconf,
4540 .maxlen = sizeof(int),
4542 .proc_handler = proc_dointvec,
4545 .procname = "dad_transmits",
4546 .data = &ipv6_devconf.dad_transmits,
4547 .maxlen = sizeof(int),
4549 .proc_handler = proc_dointvec,
4552 .procname = "router_solicitations",
4553 .data = &ipv6_devconf.rtr_solicits,
4554 .maxlen = sizeof(int),
4556 .proc_handler = proc_dointvec,
4559 .procname = "router_solicitation_interval",
4560 .data = &ipv6_devconf.rtr_solicit_interval,
4561 .maxlen = sizeof(int),
4563 .proc_handler = proc_dointvec_jiffies,
4566 .procname = "router_solicitation_delay",
4567 .data = &ipv6_devconf.rtr_solicit_delay,
4568 .maxlen = sizeof(int),
4570 .proc_handler = proc_dointvec_jiffies,
4573 .procname = "force_mld_version",
4574 .data = &ipv6_devconf.force_mld_version,
4575 .maxlen = sizeof(int),
4577 .proc_handler = proc_dointvec,
4579 #ifdef CONFIG_IPV6_PRIVACY
4581 .procname = "use_tempaddr",
4582 .data = &ipv6_devconf.use_tempaddr,
4583 .maxlen = sizeof(int),
4585 .proc_handler = proc_dointvec,
4588 .procname = "temp_valid_lft",
4589 .data = &ipv6_devconf.temp_valid_lft,
4590 .maxlen = sizeof(int),
4592 .proc_handler = proc_dointvec,
4595 .procname = "temp_prefered_lft",
4596 .data = &ipv6_devconf.temp_prefered_lft,
4597 .maxlen = sizeof(int),
4599 .proc_handler = proc_dointvec,
4602 .procname = "regen_max_retry",
4603 .data = &ipv6_devconf.regen_max_retry,
4604 .maxlen = sizeof(int),
4606 .proc_handler = proc_dointvec,
4609 .procname = "max_desync_factor",
4610 .data = &ipv6_devconf.max_desync_factor,
4611 .maxlen = sizeof(int),
4613 .proc_handler = proc_dointvec,
4617 .procname = "max_addresses",
4618 .data = &ipv6_devconf.max_addresses,
4619 .maxlen = sizeof(int),
4621 .proc_handler = proc_dointvec,
4624 .procname = "accept_ra_defrtr",
4625 .data = &ipv6_devconf.accept_ra_defrtr,
4626 .maxlen = sizeof(int),
4628 .proc_handler = proc_dointvec,
4631 .procname = "accept_ra_pinfo",
4632 .data = &ipv6_devconf.accept_ra_pinfo,
4633 .maxlen = sizeof(int),
4635 .proc_handler = proc_dointvec,
4637 #ifdef CONFIG_IPV6_ROUTER_PREF
4639 .procname = "accept_ra_rtr_pref",
4640 .data = &ipv6_devconf.accept_ra_rtr_pref,
4641 .maxlen = sizeof(int),
4643 .proc_handler = proc_dointvec,
4646 .procname = "router_probe_interval",
4647 .data = &ipv6_devconf.rtr_probe_interval,
4648 .maxlen = sizeof(int),
4650 .proc_handler = proc_dointvec_jiffies,
4652 #ifdef CONFIG_IPV6_ROUTE_INFO
4654 .procname = "accept_ra_rt_info_max_plen",
4655 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4656 .maxlen = sizeof(int),
4658 .proc_handler = proc_dointvec,
4663 .procname = "proxy_ndp",
4664 .data = &ipv6_devconf.proxy_ndp,
4665 .maxlen = sizeof(int),
4667 .proc_handler = proc_dointvec,
4670 .procname = "accept_source_route",
4671 .data = &ipv6_devconf.accept_source_route,
4672 .maxlen = sizeof(int),
4674 .proc_handler = proc_dointvec,
4676 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4678 .procname = "optimistic_dad",
4679 .data = &ipv6_devconf.optimistic_dad,
4680 .maxlen = sizeof(int),
4682 .proc_handler = proc_dointvec,
4686 #ifdef CONFIG_IPV6_MROUTE
4688 .procname = "mc_forwarding",
4689 .data = &ipv6_devconf.mc_forwarding,
4690 .maxlen = sizeof(int),
4692 .proc_handler = proc_dointvec,
4696 .procname = "disable_ipv6",
4697 .data = &ipv6_devconf.disable_ipv6,
4698 .maxlen = sizeof(int),
4700 .proc_handler = addrconf_sysctl_disable,
4703 .procname = "accept_dad",
4704 .data = &ipv6_devconf.accept_dad,
4705 .maxlen = sizeof(int),
4707 .proc_handler = proc_dointvec,
4710 .procname = "force_tllao",
4711 .data = &ipv6_devconf.force_tllao,
4712 .maxlen = sizeof(int),
4714 .proc_handler = proc_dointvec
4717 .procname = "ndisc_notify",
4718 .data = &ipv6_devconf.ndisc_notify,
4719 .maxlen = sizeof(int),
4721 .proc_handler = proc_dointvec
4729 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4730 struct inet6_dev *idev, struct ipv6_devconf *p)
4733 struct addrconf_sysctl_table *t;
4734 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4736 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4740 for (i = 0; t->addrconf_vars[i].data; i++) {
4741 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4742 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4743 t->addrconf_vars[i].extra2 = net;
4746 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4748 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4749 if (t->sysctl_header == NULL)
4761 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4763 struct addrconf_sysctl_table *t;
4765 if (p->sysctl == NULL)
4770 unregister_net_sysctl_table(t->sysctl_header);
4774 static void addrconf_sysctl_register(struct inet6_dev *idev)
4776 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4777 &ndisc_ifinfo_sysctl_change);
4778 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4782 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4784 __addrconf_sysctl_unregister(&idev->cnf);
4785 neigh_sysctl_unregister(idev->nd_parms);
4791 static int __net_init addrconf_init_net(struct net *net)
4794 struct ipv6_devconf *all, *dflt;
4797 all = &ipv6_devconf;
4798 dflt = &ipv6_devconf_dflt;
4800 if (!net_eq(net, &init_net)) {
4801 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4805 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4807 goto err_alloc_dflt;
4809 /* these will be inherited by all namespaces */
4810 dflt->autoconf = ipv6_defaults.autoconf;
4811 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4814 net->ipv6.devconf_all = all;
4815 net->ipv6.devconf_dflt = dflt;
4817 #ifdef CONFIG_SYSCTL
4818 err = __addrconf_sysctl_register(net, "all", NULL, all);
4822 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4828 #ifdef CONFIG_SYSCTL
4830 __addrconf_sysctl_unregister(all);
4840 static void __net_exit addrconf_exit_net(struct net *net)
4842 #ifdef CONFIG_SYSCTL
4843 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4844 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4846 if (!net_eq(net, &init_net)) {
4847 kfree(net->ipv6.devconf_dflt);
4848 kfree(net->ipv6.devconf_all);
4852 static struct pernet_operations addrconf_ops = {
4853 .init = addrconf_init_net,
4854 .exit = addrconf_exit_net,
4861 int register_inet6addr_notifier(struct notifier_block *nb)
4863 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4865 EXPORT_SYMBOL(register_inet6addr_notifier);
4867 int unregister_inet6addr_notifier(struct notifier_block *nb)
4869 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4871 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4873 static struct rtnl_af_ops inet6_ops = {
4875 .fill_link_af = inet6_fill_link_af,
4876 .get_link_af_size = inet6_get_link_af_size,
4880 * Init / cleanup code
4883 int __init addrconf_init(void)
4887 err = ipv6_addr_label_init();
4889 pr_crit("%s: cannot initialize default policy table: %d\n",
4894 err = register_pernet_subsys(&addrconf_ops);
4898 /* The addrconf netdev notifier requires that loopback_dev
4899 * has it's ipv6 private information allocated and setup
4900 * before it can bring up and give link-local addresses
4901 * to other devices which are up.
4903 * Unfortunately, loopback_dev is not necessarily the first
4904 * entry in the global dev_base list of net devices. In fact,
4905 * it is likely to be the very last entry on that list.
4906 * So this causes the notifier registry below to try and
4907 * give link-local addresses to all devices besides loopback_dev
4908 * first, then loopback_dev, which cases all the non-loopback_dev
4909 * devices to fail to get a link-local address.
4911 * So, as a temporary fix, allocate the ipv6 structure for
4912 * loopback_dev first by hand.
4913 * Longer term, all of the dependencies ipv6 has upon the loopback
4914 * device and it being up should be removed.
4917 if (!ipv6_add_dev(init_net.loopback_dev))
4923 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4924 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4926 register_netdevice_notifier(&ipv6_dev_notf);
4930 err = rtnl_af_register(&inet6_ops);
4934 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4939 /* Only the first call to __rtnl_register can fail */
4940 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4941 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4942 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4943 inet6_dump_ifaddr, NULL);
4944 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4945 inet6_dump_ifmcaddr, NULL);
4946 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4947 inet6_dump_ifacaddr, NULL);
4948 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4951 ipv6_addr_label_rtnl_register();
4955 rtnl_af_unregister(&inet6_ops);
4957 unregister_netdevice_notifier(&ipv6_dev_notf);
4959 unregister_pernet_subsys(&addrconf_ops);
4961 ipv6_addr_label_cleanup();
4966 void addrconf_cleanup(void)
4968 struct net_device *dev;
4971 unregister_netdevice_notifier(&ipv6_dev_notf);
4972 unregister_pernet_subsys(&addrconf_ops);
4973 ipv6_addr_label_cleanup();
4977 __rtnl_af_unregister(&inet6_ops);
4979 /* clean dev list */
4980 for_each_netdev(&init_net, dev) {
4981 if (__in6_dev_get(dev) == NULL)
4983 addrconf_ifdown(dev, 1);
4985 addrconf_ifdown(init_net.loopback_dev, 2);
4990 spin_lock_bh(&addrconf_hash_lock);
4991 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4992 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4993 spin_unlock_bh(&addrconf_hash_lock);
4995 del_timer(&addr_chk_timer);