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 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
58 #include <linux/sysctl.h>
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
65 #include <net/net_namespace.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
91 /* Set to 3 to get tracing... */
95 #define ADBG(x) printk x
100 #define INFINITY_LIFE_TIME 0xFFFFFFFF
102 static inline u32 cstamp_delta(unsigned long cstamp)
104 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
107 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
108 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
109 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
112 static void addrconf_sysctl_register(struct inet6_dev *idev);
113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
124 #ifdef CONFIG_IPV6_PRIVACY
125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
127 static void ipv6_regen_rndid(unsigned long data);
130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
131 static int ipv6_count_addresses(struct inet6_dev *idev);
134 * Configured unicast address hash table
136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
137 static DEFINE_SPINLOCK(addrconf_hash_lock);
139 static void addrconf_verify(unsigned long);
141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
142 static DEFINE_SPINLOCK(addrconf_verify_lock);
144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
147 static void addrconf_type_change(struct net_device *dev,
148 unsigned long event);
149 static int addrconf_ifdown(struct net_device *dev, int how);
151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
152 static void addrconf_dad_timer(unsigned long data);
153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
154 static void addrconf_dad_run(struct inet6_dev *idev);
155 static void addrconf_rs_timer(unsigned long data);
156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
160 struct prefix_info *pinfo);
161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
162 struct net_device *dev);
164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
168 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
169 .mtu6 = IPV6_MIN_MTU,
171 .accept_redirects = 1,
173 .force_mld_version = 0,
175 .rtr_solicits = MAX_RTR_SOLICITATIONS,
176 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
177 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
178 #ifdef CONFIG_IPV6_PRIVACY
180 .temp_valid_lft = TEMP_VALID_LIFETIME,
181 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
182 .regen_max_retry = REGEN_MAX_RETRY,
183 .max_desync_factor = MAX_DESYNC_FACTOR,
185 .max_addresses = IPV6_MAX_ADDRESSES,
186 .accept_ra_defrtr = 1,
187 .accept_ra_pinfo = 1,
188 #ifdef CONFIG_IPV6_ROUTER_PREF
189 .accept_ra_rtr_pref = 1,
190 .rtr_probe_interval = 60 * HZ,
191 #ifdef CONFIG_IPV6_ROUTE_INFO
192 .accept_ra_rt_info_max_plen = 0,
196 .accept_source_route = 0, /* we do not accept RH0 by default. */
201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
204 .mtu6 = IPV6_MIN_MTU,
206 .accept_redirects = 1,
209 .rtr_solicits = MAX_RTR_SOLICITATIONS,
210 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
211 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
212 #ifdef CONFIG_IPV6_PRIVACY
214 .temp_valid_lft = TEMP_VALID_LIFETIME,
215 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
216 .regen_max_retry = REGEN_MAX_RETRY,
217 .max_desync_factor = MAX_DESYNC_FACTOR,
219 .max_addresses = IPV6_MAX_ADDRESSES,
220 .accept_ra_defrtr = 1,
221 .accept_ra_pinfo = 1,
222 #ifdef CONFIG_IPV6_ROUTER_PREF
223 .accept_ra_rtr_pref = 1,
224 .rtr_probe_interval = 60 * HZ,
225 #ifdef CONFIG_IPV6_ROUTE_INFO
226 .accept_ra_rt_info_max_plen = 0,
230 .accept_source_route = 0, /* we do not accept RH0 by default. */
235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
241 /* Check if a valid qdisc is available */
242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
244 return !qdisc_tx_is_noop(dev);
247 /* Check if a route is valid prefix route */
248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
250 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
255 if (del_timer(&ifp->timer))
259 enum addrconf_timer_t {
265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
266 enum addrconf_timer_t what,
269 if (!del_timer(&ifp->timer))
274 ifp->timer.function = addrconf_dad_timer;
277 ifp->timer.function = addrconf_rs_timer;
282 ifp->timer.expires = jiffies + when;
283 add_timer(&ifp->timer);
286 static int snmp6_alloc_dev(struct inet6_dev *idev)
288 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
289 sizeof(struct ipstats_mib),
290 __alignof__(struct ipstats_mib)) < 0)
292 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
293 sizeof(struct icmpv6_mib),
294 __alignof__(struct icmpv6_mib)) < 0)
296 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
297 sizeof(struct icmpv6msg_mib),
298 __alignof__(struct icmpv6msg_mib)) < 0)
304 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
306 snmp_mib_free((void __percpu **)idev->stats.ipv6);
311 static void snmp6_free_dev(struct inet6_dev *idev)
313 snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
314 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
315 snmp_mib_free((void __percpu **)idev->stats.ipv6);
318 /* Nobody refers to this device, we may destroy it. */
320 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
322 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
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 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
338 pr_warning("Freeing alive inet6 device %p\n", idev);
341 snmp6_free_dev(idev);
342 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
345 EXPORT_SYMBOL(in6_dev_finish_destroy);
347 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
349 struct inet6_dev *ndev;
353 if (dev->mtu < IPV6_MIN_MTU)
356 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
361 rwlock_init(&ndev->lock);
363 INIT_LIST_HEAD(&ndev->addr_list);
365 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
366 ndev->cnf.mtu6 = dev->mtu;
367 ndev->cnf.sysctl = NULL;
368 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
369 if (ndev->nd_parms == NULL) {
373 if (ndev->cnf.forwarding)
374 dev_disable_lro(dev);
375 /* We refer to the device */
378 if (snmp6_alloc_dev(ndev) < 0) {
380 "%s(): cannot allocate memory for statistics; dev=%s.\n",
381 __func__, dev->name));
382 neigh_parms_release(&nd_tbl, ndev->nd_parms);
384 in6_dev_finish_destroy(ndev);
388 if (snmp6_register_dev(ndev) < 0) {
390 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
391 __func__, dev->name));
392 neigh_parms_release(&nd_tbl, ndev->nd_parms);
394 in6_dev_finish_destroy(ndev);
398 /* One reference from device. We must do this before
399 * we invoke __ipv6_regen_rndid().
403 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
404 ndev->cnf.accept_dad = -1;
406 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
407 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
409 "%s: Disabled Multicast RS\n",
411 ndev->cnf.rtr_solicits = 0;
415 #ifdef CONFIG_IPV6_PRIVACY
416 INIT_LIST_HEAD(&ndev->tempaddr_list);
417 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
418 if ((dev->flags&IFF_LOOPBACK) ||
419 dev->type == ARPHRD_TUNNEL ||
420 dev->type == ARPHRD_TUNNEL6 ||
421 dev->type == ARPHRD_SIT ||
422 dev->type == ARPHRD_NONE) {
424 "%s: Disabled Privacy Extensions\n",
426 ndev->cnf.use_tempaddr = -1;
429 ipv6_regen_rndid((unsigned long) ndev);
433 if (netif_running(dev) && addrconf_qdisc_ok(dev))
434 ndev->if_flags |= IF_READY;
436 ipv6_mc_init_dev(ndev);
437 ndev->tstamp = jiffies;
438 addrconf_sysctl_register(ndev);
439 /* protected by rtnl_lock */
440 rcu_assign_pointer(dev->ip6_ptr, ndev);
442 /* Join all-node multicast group */
443 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
448 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
450 struct inet6_dev *idev;
454 idev = __in6_dev_get(dev);
456 idev = ipv6_add_dev(dev);
461 if (dev->flags&IFF_UP)
467 static void dev_forward_change(struct inet6_dev *idev)
469 struct net_device *dev;
470 struct inet6_ifaddr *ifa;
475 if (idev->cnf.forwarding)
476 dev_disable_lro(dev);
477 if (dev && (dev->flags & IFF_MULTICAST)) {
478 if (idev->cnf.forwarding)
479 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
481 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
484 list_for_each_entry(ifa, &idev->addr_list, if_list) {
485 if (ifa->flags&IFA_F_TENTATIVE)
487 if (idev->cnf.forwarding)
488 addrconf_join_anycast(ifa);
490 addrconf_leave_anycast(ifa);
495 static void addrconf_forward_change(struct net *net, __s32 newf)
497 struct net_device *dev;
498 struct inet6_dev *idev;
501 for_each_netdev_rcu(net, dev) {
502 idev = __in6_dev_get(dev);
504 int changed = (!idev->cnf.forwarding) ^ (!newf);
505 idev->cnf.forwarding = newf;
507 dev_forward_change(idev);
513 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
517 net = (struct net *)table->extra2;
518 if (p == &net->ipv6.devconf_dflt->forwarding)
521 if (!rtnl_trylock()) {
522 /* Restore the original values before restarting */
524 return restart_syscall();
527 if (p == &net->ipv6.devconf_all->forwarding) {
528 __s32 newf = net->ipv6.devconf_all->forwarding;
529 net->ipv6.devconf_dflt->forwarding = newf;
530 addrconf_forward_change(net, newf);
531 } else if ((!*p) ^ (!old))
532 dev_forward_change((struct inet6_dev *)table->extra1);
536 rt6_purge_dflt_routers(net);
541 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
543 struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
547 /* Nobody refers to this ifaddr, destroy it */
548 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
550 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
552 #ifdef NET_REFCNT_DEBUG
553 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
556 in6_dev_put(ifp->idev);
558 if (del_timer(&ifp->timer))
559 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
561 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
562 pr_warning("Freeing alive inet6 address %p\n", ifp);
565 dst_release(&ifp->rt->dst);
567 call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
571 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
574 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
577 * Each device address list is sorted in order of scope -
578 * global before linklocal.
580 list_for_each(p, &idev->addr_list) {
581 struct inet6_ifaddr *ifa
582 = list_entry(p, struct inet6_ifaddr, if_list);
583 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
587 list_add_tail(&ifp->if_list, p);
590 static u32 ipv6_addr_hash(const struct in6_addr *addr)
593 * We perform the hash function over the last 64 bits of the address
594 * This will include the IEEE address token on links that support it.
596 return jhash_2words((__force u32)addr->s6_addr32[2],
597 (__force u32)addr->s6_addr32[3], 0)
598 & (IN6_ADDR_HSIZE - 1);
601 /* On success it returns ifp with increased reference count */
603 static struct inet6_ifaddr *
604 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
605 int scope, u32 flags)
607 struct inet6_ifaddr *ifa = NULL;
611 int addr_type = ipv6_addr_type(addr);
613 if (addr_type == IPV6_ADDR_ANY ||
614 addr_type & IPV6_ADDR_MULTICAST ||
615 (!(idev->dev->flags & IFF_LOOPBACK) &&
616 addr_type & IPV6_ADDR_LOOPBACK))
617 return ERR_PTR(-EADDRNOTAVAIL);
621 err = -ENODEV; /*XXX*/
625 if (idev->cnf.disable_ipv6) {
630 spin_lock(&addrconf_hash_lock);
632 /* Ignore adding duplicate addresses on an interface */
633 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
634 ADBG(("ipv6_add_addr: already assigned\n"));
639 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
642 ADBG(("ipv6_add_addr: malloc failed\n"));
647 rt = addrconf_dst_alloc(idev, addr, 0);
653 ipv6_addr_copy(&ifa->addr, addr);
655 spin_lock_init(&ifa->lock);
656 spin_lock_init(&ifa->state_lock);
657 init_timer(&ifa->timer);
658 INIT_HLIST_NODE(&ifa->addr_lst);
659 ifa->timer.data = (unsigned long) ifa;
661 ifa->prefix_len = pfxlen;
662 ifa->flags = flags | IFA_F_TENTATIVE;
663 ifa->cstamp = ifa->tstamp = jiffies;
668 * part one of RFC 4429, section 3.3
669 * We should not configure an address as
670 * optimistic if we do not yet know the link
671 * layer address of our nexhop router
674 if (rt->rt6i_nexthop == NULL)
675 ifa->flags &= ~IFA_F_OPTIMISTIC;
682 /* Add to big hash table */
683 hash = ipv6_addr_hash(addr);
685 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
686 spin_unlock(&addrconf_hash_lock);
688 write_lock(&idev->lock);
689 /* Add to inet6_dev unicast addr list. */
690 ipv6_link_dev_addr(idev, ifa);
692 #ifdef CONFIG_IPV6_PRIVACY
693 if (ifa->flags&IFA_F_TEMPORARY) {
694 list_add(&ifa->tmp_list, &idev->tempaddr_list);
700 write_unlock(&idev->lock);
702 rcu_read_unlock_bh();
704 if (likely(err == 0))
705 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
713 spin_unlock(&addrconf_hash_lock);
717 /* This function wants to get referenced ifp and releases it before return */
719 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
721 struct inet6_ifaddr *ifa, *ifn;
722 struct inet6_dev *idev = ifp->idev;
725 int deleted = 0, onlink = 0;
726 unsigned long expires = jiffies;
728 hash = ipv6_addr_hash(&ifp->addr);
730 spin_lock_bh(&ifp->state_lock);
732 ifp->state = INET6_IFADDR_STATE_DEAD;
733 spin_unlock_bh(&ifp->state_lock);
735 if (state == INET6_IFADDR_STATE_DEAD)
738 spin_lock_bh(&addrconf_hash_lock);
739 hlist_del_init_rcu(&ifp->addr_lst);
740 spin_unlock_bh(&addrconf_hash_lock);
742 write_lock_bh(&idev->lock);
743 #ifdef CONFIG_IPV6_PRIVACY
744 if (ifp->flags&IFA_F_TEMPORARY) {
745 list_del(&ifp->tmp_list);
747 in6_ifa_put(ifp->ifpub);
754 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
756 list_del_init(&ifp->if_list);
759 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
763 } else if (ifp->flags & IFA_F_PERMANENT) {
764 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
766 if (ifa->flags & IFA_F_PERMANENT) {
771 unsigned long lifetime;
776 spin_lock(&ifa->lock);
778 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
780 * Note: Because this address is
781 * not permanent, lifetime <
782 * LONG_MAX / HZ here.
784 if (time_before(expires,
785 ifa->tstamp + lifetime * HZ))
786 expires = ifa->tstamp + lifetime * HZ;
787 spin_unlock(&ifa->lock);
792 write_unlock_bh(&idev->lock);
794 addrconf_del_timer(ifp);
796 ipv6_ifa_notify(RTM_DELADDR, ifp);
798 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
801 * Purge or update corresponding prefix
803 * 1) we don't purge prefix here if address was not permanent.
804 * prefix is managed by its own lifetime.
805 * 2) if there're no addresses, delete prefix.
806 * 3) if there're still other permanent address(es),
807 * corresponding prefix is still permanent.
808 * 4) otherwise, update prefix lifetime to the
809 * longest valid lifetime among the corresponding
810 * addresses on the device.
811 * Note: subsequent RA will update lifetime.
815 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
816 struct in6_addr prefix;
818 struct net *net = dev_net(ifp->idev->dev);
819 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
820 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
822 if (rt && addrconf_is_prefix_route(rt)) {
826 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
827 rt->rt6i_expires = expires;
828 rt->rt6i_flags |= RTF_EXPIRES;
831 dst_release(&rt->dst);
838 #ifdef CONFIG_IPV6_PRIVACY
839 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
841 struct inet6_dev *idev = ifp->idev;
842 struct in6_addr addr, *tmpaddr;
843 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
844 unsigned long regen_advance;
850 write_lock(&idev->lock);
852 spin_lock_bh(&ift->lock);
853 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
854 spin_unlock_bh(&ift->lock);
861 if (idev->cnf.use_tempaddr <= 0) {
862 write_unlock(&idev->lock);
864 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
869 spin_lock_bh(&ifp->lock);
870 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
871 idev->cnf.use_tempaddr = -1; /*XXX*/
872 spin_unlock_bh(&ifp->lock);
873 write_unlock(&idev->lock);
875 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
881 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
882 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
883 spin_unlock_bh(&ifp->lock);
884 write_unlock(&idev->lock);
886 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
892 memcpy(&addr.s6_addr[8], idev->rndid, 8);
893 age = (jiffies - ifp->tstamp) / HZ;
894 tmp_valid_lft = min_t(__u32,
896 idev->cnf.temp_valid_lft + age);
897 tmp_prefered_lft = min_t(__u32,
899 idev->cnf.temp_prefered_lft + age -
900 idev->cnf.max_desync_factor);
901 tmp_plen = ifp->prefix_len;
902 max_addresses = idev->cnf.max_addresses;
903 tmp_cstamp = ifp->cstamp;
904 tmp_tstamp = ifp->tstamp;
905 spin_unlock_bh(&ifp->lock);
907 regen_advance = idev->cnf.regen_max_retry *
908 idev->cnf.dad_transmits *
909 idev->nd_parms->retrans_time / HZ;
910 write_unlock(&idev->lock);
912 /* A temporary address is created only if this calculated Preferred
913 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
914 * an implementation must not create a temporary address with a zero
915 * Preferred Lifetime.
917 if (tmp_prefered_lft <= regen_advance) {
924 addr_flags = IFA_F_TEMPORARY;
925 /* set in addrconf_prefix_rcv() */
926 if (ifp->flags & IFA_F_OPTIMISTIC)
927 addr_flags |= IFA_F_OPTIMISTIC;
929 ift = !max_addresses ||
930 ipv6_count_addresses(idev) < max_addresses ?
931 ipv6_add_addr(idev, &addr, tmp_plen,
932 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
934 if (!ift || IS_ERR(ift)) {
938 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
940 write_lock(&idev->lock);
944 spin_lock_bh(&ift->lock);
946 ift->valid_lft = tmp_valid_lft;
947 ift->prefered_lft = tmp_prefered_lft;
948 ift->cstamp = tmp_cstamp;
949 ift->tstamp = tmp_tstamp;
950 spin_unlock_bh(&ift->lock);
952 addrconf_dad_start(ift, 0);
961 * Choose an appropriate source address (RFC3484)
964 IPV6_SADDR_RULE_INIT = 0,
965 IPV6_SADDR_RULE_LOCAL,
966 IPV6_SADDR_RULE_SCOPE,
967 IPV6_SADDR_RULE_PREFERRED,
968 #ifdef CONFIG_IPV6_MIP6
972 IPV6_SADDR_RULE_LABEL,
973 #ifdef CONFIG_IPV6_PRIVACY
974 IPV6_SADDR_RULE_PRIVACY,
976 IPV6_SADDR_RULE_ORCHID,
977 IPV6_SADDR_RULE_PREFIX,
981 struct ipv6_saddr_score {
984 struct inet6_ifaddr *ifa;
985 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
990 struct ipv6_saddr_dst {
991 const struct in6_addr *addr;
998 static inline int ipv6_saddr_preferred(int type)
1000 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1005 static int ipv6_get_saddr_eval(struct net *net,
1006 struct ipv6_saddr_score *score,
1007 struct ipv6_saddr_dst *dst,
1012 if (i <= score->rule) {
1014 case IPV6_SADDR_RULE_SCOPE:
1015 ret = score->scopedist;
1017 case IPV6_SADDR_RULE_PREFIX:
1018 ret = score->matchlen;
1021 ret = !!test_bit(i, score->scorebits);
1027 case IPV6_SADDR_RULE_INIT:
1028 /* Rule 0: remember if hiscore is not ready yet */
1031 case IPV6_SADDR_RULE_LOCAL:
1032 /* Rule 1: Prefer same address */
1033 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1035 case IPV6_SADDR_RULE_SCOPE:
1036 /* Rule 2: Prefer appropriate scope
1041 * ---+--+-+---> scope
1043 * | d is scope of the destination.
1045 * | \ <- smaller scope is better if
1046 * B-15 | \ if scope is enough for destinaion.
1047 * | ret = B - scope (-1 <= scope >= d <= 15).
1049 * |/ <- greater is better
1050 * -C / if scope is not enough for destination.
1051 * /| ret = scope - C (-1 <= d < scope <= 15).
1053 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1054 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1055 * Assume B = 0 and we get C > 29.
1057 ret = __ipv6_addr_src_scope(score->addr_type);
1058 if (ret >= dst->scope)
1061 ret -= 128; /* 30 is enough */
1062 score->scopedist = ret;
1064 case IPV6_SADDR_RULE_PREFERRED:
1065 /* Rule 3: Avoid deprecated and optimistic addresses */
1066 ret = ipv6_saddr_preferred(score->addr_type) ||
1067 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1069 #ifdef CONFIG_IPV6_MIP6
1070 case IPV6_SADDR_RULE_HOA:
1072 /* Rule 4: Prefer home address */
1073 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1074 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1078 case IPV6_SADDR_RULE_OIF:
1079 /* Rule 5: Prefer outgoing interface */
1080 ret = (!dst->ifindex ||
1081 dst->ifindex == score->ifa->idev->dev->ifindex);
1083 case IPV6_SADDR_RULE_LABEL:
1084 /* Rule 6: Prefer matching label */
1085 ret = ipv6_addr_label(net,
1086 &score->ifa->addr, score->addr_type,
1087 score->ifa->idev->dev->ifindex) == dst->label;
1089 #ifdef CONFIG_IPV6_PRIVACY
1090 case IPV6_SADDR_RULE_PRIVACY:
1092 /* Rule 7: Prefer public address
1093 * Note: prefer temprary address if use_tempaddr >= 2
1095 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1096 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1097 score->ifa->idev->cnf.use_tempaddr >= 2;
1098 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1102 case IPV6_SADDR_RULE_ORCHID:
1103 /* Rule 8-: Prefer ORCHID vs ORCHID or
1104 * non-ORCHID vs non-ORCHID
1106 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1107 ipv6_addr_orchid(dst->addr));
1109 case IPV6_SADDR_RULE_PREFIX:
1110 /* Rule 8: Use longest matching prefix */
1111 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1119 __set_bit(i, score->scorebits);
1125 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1126 const struct in6_addr *daddr, unsigned int prefs,
1127 struct in6_addr *saddr)
1129 struct ipv6_saddr_score scores[2],
1130 *score = &scores[0], *hiscore = &scores[1];
1131 struct ipv6_saddr_dst dst;
1132 struct net_device *dev;
1135 dst_type = __ipv6_addr_type(daddr);
1137 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1138 dst.scope = __ipv6_addr_src_scope(dst_type);
1139 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1143 hiscore->ifa = NULL;
1147 for_each_netdev_rcu(net, dev) {
1148 struct inet6_dev *idev;
1150 /* Candidate Source Address (section 4)
1151 * - multicast and link-local destination address,
1152 * the set of candidate source address MUST only
1153 * include addresses assigned to interfaces
1154 * belonging to the same link as the outgoing
1156 * (- For site-local destination addresses, the
1157 * set of candidate source addresses MUST only
1158 * include addresses assigned to interfaces
1159 * belonging to the same site as the outgoing
1162 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1163 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1164 dst.ifindex && dev->ifindex != dst.ifindex)
1167 idev = __in6_dev_get(dev);
1171 read_lock_bh(&idev->lock);
1172 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1176 * - Tentative Address (RFC2462 section 5.4)
1177 * - A tentative address is not considered
1178 * "assigned to an interface" in the traditional
1179 * sense, unless it is also flagged as optimistic.
1180 * - Candidate Source Address (section 4)
1181 * - In any case, anycast addresses, multicast
1182 * addresses, and the unspecified address MUST
1183 * NOT be included in a candidate set.
1185 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1186 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1189 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1191 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1192 score->addr_type & IPV6_ADDR_MULTICAST)) {
1193 LIMIT_NETDEBUG(KERN_DEBUG
1194 "ADDRCONF: unspecified / multicast address "
1195 "assigned as unicast address on %s",
1201 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1203 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1204 int minihiscore, miniscore;
1206 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1207 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1209 if (minihiscore > miniscore) {
1210 if (i == IPV6_SADDR_RULE_SCOPE &&
1211 score->scopedist > 0) {
1214 * each remaining entry
1215 * has too small (not enough)
1216 * scope, because ifa entries
1217 * are sorted by their scope
1223 } else if (minihiscore < miniscore) {
1225 in6_ifa_put(hiscore->ifa);
1227 in6_ifa_hold(score->ifa);
1229 swap(hiscore, score);
1231 /* restore our iterator */
1232 score->ifa = hiscore->ifa;
1239 read_unlock_bh(&idev->lock);
1244 return -EADDRNOTAVAIL;
1246 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1247 in6_ifa_put(hiscore->ifa);
1250 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1252 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1253 unsigned char banned_flags)
1255 struct inet6_dev *idev;
1256 int err = -EADDRNOTAVAIL;
1259 idev = __in6_dev_get(dev);
1261 struct inet6_ifaddr *ifp;
1263 read_lock_bh(&idev->lock);
1264 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1265 if (ifp->scope == IFA_LINK &&
1266 !(ifp->flags & banned_flags)) {
1267 ipv6_addr_copy(addr, &ifp->addr);
1272 read_unlock_bh(&idev->lock);
1278 static int ipv6_count_addresses(struct inet6_dev *idev)
1281 struct inet6_ifaddr *ifp;
1283 read_lock_bh(&idev->lock);
1284 list_for_each_entry(ifp, &idev->addr_list, if_list)
1286 read_unlock_bh(&idev->lock);
1290 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1291 struct net_device *dev, int strict)
1293 struct inet6_ifaddr *ifp;
1294 struct hlist_node *node;
1295 unsigned int hash = ipv6_addr_hash(addr);
1298 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1299 if (!net_eq(dev_net(ifp->idev->dev), net))
1301 if (ipv6_addr_equal(&ifp->addr, addr) &&
1302 !(ifp->flags&IFA_F_TENTATIVE) &&
1303 (dev == NULL || ifp->idev->dev == dev ||
1304 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1305 rcu_read_unlock_bh();
1310 rcu_read_unlock_bh();
1313 EXPORT_SYMBOL(ipv6_chk_addr);
1315 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1316 struct net_device *dev)
1318 unsigned int hash = ipv6_addr_hash(addr);
1319 struct inet6_ifaddr *ifp;
1320 struct hlist_node *node;
1322 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1323 if (!net_eq(dev_net(ifp->idev->dev), net))
1325 if (ipv6_addr_equal(&ifp->addr, addr)) {
1326 if (dev == NULL || ifp->idev->dev == dev)
1333 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1335 struct inet6_dev *idev;
1336 struct inet6_ifaddr *ifa;
1341 idev = __in6_dev_get(dev);
1343 read_lock_bh(&idev->lock);
1344 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1345 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1350 read_unlock_bh(&idev->lock);
1356 EXPORT_SYMBOL(ipv6_chk_prefix);
1358 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1359 struct net_device *dev, int strict)
1361 struct inet6_ifaddr *ifp, *result = NULL;
1362 unsigned int hash = ipv6_addr_hash(addr);
1363 struct hlist_node *node;
1366 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1367 if (!net_eq(dev_net(ifp->idev->dev), net))
1369 if (ipv6_addr_equal(&ifp->addr, addr)) {
1370 if (dev == NULL || ifp->idev->dev == dev ||
1371 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1378 rcu_read_unlock_bh();
1383 /* Gets referenced address, destroys ifaddr */
1385 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1387 if (ifp->flags&IFA_F_PERMANENT) {
1388 spin_lock_bh(&ifp->lock);
1389 addrconf_del_timer(ifp);
1390 ifp->flags |= IFA_F_TENTATIVE;
1392 ifp->flags |= IFA_F_DADFAILED;
1393 spin_unlock_bh(&ifp->lock);
1395 ipv6_ifa_notify(0, ifp);
1397 #ifdef CONFIG_IPV6_PRIVACY
1398 } else if (ifp->flags&IFA_F_TEMPORARY) {
1399 struct inet6_ifaddr *ifpub;
1400 spin_lock_bh(&ifp->lock);
1403 in6_ifa_hold(ifpub);
1404 spin_unlock_bh(&ifp->lock);
1405 ipv6_create_tempaddr(ifpub, ifp);
1408 spin_unlock_bh(&ifp->lock);
1416 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1420 spin_lock(&ifp->state_lock);
1421 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1422 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1425 spin_unlock(&ifp->state_lock);
1430 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1432 struct inet6_dev *idev = ifp->idev;
1434 if (addrconf_dad_end(ifp)) {
1439 if (net_ratelimit())
1440 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1441 ifp->idev->dev->name, &ifp->addr);
1443 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1444 struct in6_addr addr;
1446 addr.s6_addr32[0] = htonl(0xfe800000);
1447 addr.s6_addr32[1] = 0;
1449 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1450 ipv6_addr_equal(&ifp->addr, &addr)) {
1451 /* DAD failed for link-local based on MAC address */
1452 idev->cnf.disable_ipv6 = 1;
1454 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1455 ifp->idev->dev->name);
1459 addrconf_dad_stop(ifp, 1);
1462 /* Join to solicited addr multicast group. */
1464 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1466 struct in6_addr maddr;
1468 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1471 addrconf_addr_solict_mult(addr, &maddr);
1472 ipv6_dev_mc_inc(dev, &maddr);
1475 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1477 struct in6_addr maddr;
1479 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1482 addrconf_addr_solict_mult(addr, &maddr);
1483 __ipv6_dev_mc_dec(idev, &maddr);
1486 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1488 struct in6_addr addr;
1489 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1490 if (ipv6_addr_any(&addr))
1492 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1495 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1497 struct in6_addr addr;
1498 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1499 if (ipv6_addr_any(&addr))
1501 __ipv6_dev_ac_dec(ifp->idev, &addr);
1504 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1506 if (dev->addr_len != ETH_ALEN)
1508 memcpy(eui, dev->dev_addr, 3);
1509 memcpy(eui + 5, dev->dev_addr + 3, 3);
1512 * The zSeries OSA network cards can be shared among various
1513 * OS instances, but the OSA cards have only one MAC address.
1514 * This leads to duplicate address conflicts in conjunction
1515 * with IPv6 if more than one instance uses the same card.
1517 * The driver for these cards can deliver a unique 16-bit
1518 * identifier for each instance sharing the same card. It is
1519 * placed instead of 0xFFFE in the interface identifier. The
1520 * "u" bit of the interface identifier is not inverted in this
1521 * case. Hence the resulting interface identifier has local
1522 * scope according to RFC2373.
1525 eui[3] = (dev->dev_id >> 8) & 0xFF;
1526 eui[4] = dev->dev_id & 0xFF;
1535 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1537 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1538 if (dev->addr_len != ARCNET_ALEN)
1541 eui[7] = *(u8*)dev->dev_addr;
1545 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1547 if (dev->addr_len != INFINIBAND_ALEN)
1549 memcpy(eui, dev->dev_addr + 12, 8);
1554 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1558 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1559 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1560 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1561 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1562 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1563 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1567 memcpy(eui + 4, &addr, 4);
1571 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1573 if (dev->priv_flags & IFF_ISATAP)
1574 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1578 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1580 switch (dev->type) {
1583 case ARPHRD_IEEE802_TR:
1584 return addrconf_ifid_eui48(eui, dev);
1586 return addrconf_ifid_arcnet(eui, dev);
1587 case ARPHRD_INFINIBAND:
1588 return addrconf_ifid_infiniband(eui, dev);
1590 return addrconf_ifid_sit(eui, dev);
1595 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1598 struct inet6_ifaddr *ifp;
1600 read_lock_bh(&idev->lock);
1601 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1602 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1603 memcpy(eui, ifp->addr.s6_addr+8, 8);
1608 read_unlock_bh(&idev->lock);
1612 #ifdef CONFIG_IPV6_PRIVACY
1613 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1614 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1617 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1618 idev->rndid[0] &= ~0x02;
1621 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1622 * check if generated address is not inappropriate
1624 * - Reserved subnet anycast (RFC 2526)
1625 * 11111101 11....11 1xxxxxxx
1626 * - ISATAP (RFC4214) 6.1
1627 * 00-00-5E-FE-xx-xx-xx-xx
1629 * - XXX: already assigned to an address on the device
1631 if (idev->rndid[0] == 0xfd &&
1632 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1633 (idev->rndid[7]&0x80))
1635 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1636 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1638 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1645 static void ipv6_regen_rndid(unsigned long data)
1647 struct inet6_dev *idev = (struct inet6_dev *) data;
1648 unsigned long expires;
1651 write_lock_bh(&idev->lock);
1656 if (__ipv6_regen_rndid(idev) < 0)
1660 idev->cnf.temp_prefered_lft * HZ -
1661 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1662 idev->cnf.max_desync_factor * HZ;
1663 if (time_before(expires, jiffies)) {
1665 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1670 if (!mod_timer(&idev->regen_timer, expires))
1674 write_unlock_bh(&idev->lock);
1675 rcu_read_unlock_bh();
1679 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1682 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1683 ret = __ipv6_regen_rndid(idev);
1693 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1694 unsigned long expires, u32 flags)
1696 struct fib6_config cfg = {
1697 .fc_table = RT6_TABLE_PREFIX,
1698 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1699 .fc_ifindex = dev->ifindex,
1700 .fc_expires = expires,
1702 .fc_flags = RTF_UP | flags,
1703 .fc_nlinfo.nl_net = dev_net(dev),
1704 .fc_protocol = RTPROT_KERNEL,
1707 ipv6_addr_copy(&cfg.fc_dst, pfx);
1709 /* Prevent useless cloning on PtP SIT.
1710 This thing is done here expecting that the whole
1711 class of non-broadcast devices need not cloning.
1713 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1714 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1715 cfg.fc_flags |= RTF_NONEXTHOP;
1718 ip6_route_add(&cfg);
1721 /* Create "default" multicast route to the interface */
1723 static void addrconf_add_mroute(struct net_device *dev)
1725 struct fib6_config cfg = {
1726 .fc_table = RT6_TABLE_LOCAL,
1727 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1728 .fc_ifindex = dev->ifindex,
1731 .fc_nlinfo.nl_net = dev_net(dev),
1734 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1736 ip6_route_add(&cfg);
1739 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1740 static void sit_route_add(struct net_device *dev)
1742 struct fib6_config cfg = {
1743 .fc_table = RT6_TABLE_MAIN,
1744 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1745 .fc_ifindex = dev->ifindex,
1747 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1748 .fc_nlinfo.nl_net = dev_net(dev),
1751 /* prefix length - 96 bits "::d.d.d.d" */
1752 ip6_route_add(&cfg);
1756 static void addrconf_add_lroute(struct net_device *dev)
1758 struct in6_addr addr;
1760 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1761 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1764 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1766 struct inet6_dev *idev;
1770 idev = ipv6_find_idev(dev);
1772 return ERR_PTR(-ENOBUFS);
1774 if (idev->cnf.disable_ipv6)
1775 return ERR_PTR(-EACCES);
1777 /* Add default multicast route */
1778 addrconf_add_mroute(dev);
1780 /* Add link local route */
1781 addrconf_add_lroute(dev);
1785 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1787 struct prefix_info *pinfo;
1791 struct inet6_dev *in6_dev;
1792 struct net *net = dev_net(dev);
1794 pinfo = (struct prefix_info *) opt;
1796 if (len < sizeof(struct prefix_info)) {
1797 ADBG(("addrconf: prefix option too short\n"));
1802 * Validation checks ([ADDRCONF], page 19)
1805 addr_type = ipv6_addr_type(&pinfo->prefix);
1807 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1810 valid_lft = ntohl(pinfo->valid);
1811 prefered_lft = ntohl(pinfo->prefered);
1813 if (prefered_lft > valid_lft) {
1814 if (net_ratelimit())
1815 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1819 in6_dev = in6_dev_get(dev);
1821 if (in6_dev == NULL) {
1822 if (net_ratelimit())
1823 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1828 * Two things going on here:
1829 * 1) Add routes for on-link prefixes
1830 * 2) Configure prefixes with the auto flag set
1833 if (pinfo->onlink) {
1834 struct rt6_info *rt;
1835 unsigned long rt_expires;
1837 /* Avoid arithmetic overflow. Really, we could
1838 * save rt_expires in seconds, likely valid_lft,
1839 * but it would require division in fib gc, that it
1843 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1845 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1847 if (addrconf_finite_timeout(rt_expires))
1850 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1853 if (rt && addrconf_is_prefix_route(rt)) {
1854 /* Autoconf prefix route */
1855 if (valid_lft == 0) {
1858 } else if (addrconf_finite_timeout(rt_expires)) {
1860 rt->rt6i_expires = jiffies + rt_expires;
1861 rt->rt6i_flags |= RTF_EXPIRES;
1863 rt->rt6i_flags &= ~RTF_EXPIRES;
1864 rt->rt6i_expires = 0;
1866 } else if (valid_lft) {
1867 clock_t expires = 0;
1868 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1869 if (addrconf_finite_timeout(rt_expires)) {
1871 flags |= RTF_EXPIRES;
1872 expires = jiffies_to_clock_t(rt_expires);
1874 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1875 dev, expires, flags);
1878 dst_release(&rt->dst);
1881 /* Try to figure out our local address for this prefix */
1883 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1884 struct inet6_ifaddr * ifp;
1885 struct in6_addr addr;
1886 int create = 0, update_lft = 0;
1888 if (pinfo->prefix_len == 64) {
1889 memcpy(&addr, &pinfo->prefix, 8);
1890 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1891 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1892 in6_dev_put(in6_dev);
1897 if (net_ratelimit())
1898 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1900 in6_dev_put(in6_dev);
1905 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1907 if (ifp == NULL && valid_lft) {
1908 int max_addresses = in6_dev->cnf.max_addresses;
1911 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1912 if (in6_dev->cnf.optimistic_dad &&
1913 !net->ipv6.devconf_all->forwarding)
1914 addr_flags = IFA_F_OPTIMISTIC;
1917 /* Do not allow to create too much of autoconfigured
1918 * addresses; this would be too easy way to crash kernel.
1920 if (!max_addresses ||
1921 ipv6_count_addresses(in6_dev) < max_addresses)
1922 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1923 addr_type&IPV6_ADDR_SCOPE_MASK,
1926 if (!ifp || IS_ERR(ifp)) {
1927 in6_dev_put(in6_dev);
1931 update_lft = create = 1;
1932 ifp->cstamp = jiffies;
1933 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1939 #ifdef CONFIG_IPV6_PRIVACY
1940 struct inet6_ifaddr *ift;
1944 /* update lifetime (RFC2462 5.5.3 e) */
1945 spin_lock(&ifp->lock);
1947 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1948 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1951 if (!update_lft && stored_lft) {
1952 if (valid_lft > MIN_VALID_LIFETIME ||
1953 valid_lft > stored_lft)
1955 else if (stored_lft <= MIN_VALID_LIFETIME) {
1956 /* valid_lft <= stored_lft is always true */
1958 * RFC 4862 Section 5.5.3e:
1959 * "Note that the preferred lifetime of
1960 * the corresponding address is always
1961 * reset to the Preferred Lifetime in
1962 * the received Prefix Information
1963 * option, regardless of whether the
1964 * valid lifetime is also reset or
1967 * So if the preferred lifetime in
1968 * this advertisement is different
1969 * than what we have stored, but the
1970 * valid lifetime is invalid, just
1971 * reset prefered_lft.
1973 * We must set the valid lifetime
1974 * to the stored lifetime since we'll
1975 * be updating the timestamp below,
1976 * else we'll set it back to the
1979 if (prefered_lft != ifp->prefered_lft) {
1980 valid_lft = stored_lft;
1984 valid_lft = MIN_VALID_LIFETIME;
1985 if (valid_lft < prefered_lft)
1986 prefered_lft = valid_lft;
1992 ifp->valid_lft = valid_lft;
1993 ifp->prefered_lft = prefered_lft;
1996 ifp->flags &= ~IFA_F_DEPRECATED;
1997 spin_unlock(&ifp->lock);
1999 if (!(flags&IFA_F_TENTATIVE))
2000 ipv6_ifa_notify(0, ifp);
2002 spin_unlock(&ifp->lock);
2004 #ifdef CONFIG_IPV6_PRIVACY
2005 read_lock_bh(&in6_dev->lock);
2006 /* update all temporary addresses in the list */
2007 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2009 * When adjusting the lifetimes of an existing
2010 * temporary address, only lower the lifetimes.
2011 * Implementations must not increase the
2012 * lifetimes of an existing temporary address
2013 * when processing a Prefix Information Option.
2015 if (ifp != ift->ifpub)
2018 spin_lock(&ift->lock);
2020 if (ift->valid_lft > valid_lft &&
2021 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2022 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2023 if (ift->prefered_lft > prefered_lft &&
2024 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2025 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2026 spin_unlock(&ift->lock);
2027 if (!(flags&IFA_F_TENTATIVE))
2028 ipv6_ifa_notify(0, ift);
2031 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2033 * When a new public address is created as described in [ADDRCONF],
2034 * also create a new temporary address. Also create a temporary
2035 * address if it's enabled but no temporary address currently exists.
2037 read_unlock_bh(&in6_dev->lock);
2038 ipv6_create_tempaddr(ifp, NULL);
2040 read_unlock_bh(&in6_dev->lock);
2047 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2048 in6_dev_put(in6_dev);
2052 * Set destination address.
2053 * Special case for SIT interfaces where we create a new "virtual"
2056 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2058 struct in6_ifreq ireq;
2059 struct net_device *dev;
2065 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2068 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2074 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2075 if (dev->type == ARPHRD_SIT) {
2076 const struct net_device_ops *ops = dev->netdev_ops;
2078 struct ip_tunnel_parm p;
2080 err = -EADDRNOTAVAIL;
2081 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2084 memset(&p, 0, sizeof(p));
2085 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2089 p.iph.protocol = IPPROTO_IPV6;
2091 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2093 if (ops->ndo_do_ioctl) {
2094 mm_segment_t oldfs = get_fs();
2097 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2104 dev = __dev_get_by_name(net, p.name);
2107 err = dev_open(dev);
2118 * Manual configuration of address on an interface
2120 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2121 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2124 struct inet6_ifaddr *ifp;
2125 struct inet6_dev *idev;
2126 struct net_device *dev;
2130 unsigned long timeout;
2137 /* check the lifetime */
2138 if (!valid_lft || prefered_lft > valid_lft)
2141 dev = __dev_get_by_index(net, ifindex);
2145 idev = addrconf_add_dev(dev);
2147 return PTR_ERR(idev);
2149 scope = ipv6_addr_scope(pfx);
2151 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2152 if (addrconf_finite_timeout(timeout)) {
2153 expires = jiffies_to_clock_t(timeout * HZ);
2154 valid_lft = timeout;
2155 flags = RTF_EXPIRES;
2159 ifa_flags |= IFA_F_PERMANENT;
2162 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2163 if (addrconf_finite_timeout(timeout)) {
2165 ifa_flags |= IFA_F_DEPRECATED;
2166 prefered_lft = timeout;
2169 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2172 spin_lock_bh(&ifp->lock);
2173 ifp->valid_lft = valid_lft;
2174 ifp->prefered_lft = prefered_lft;
2175 ifp->tstamp = jiffies;
2176 spin_unlock_bh(&ifp->lock);
2178 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2181 * Note that section 3.1 of RFC 4429 indicates
2182 * that the Optimistic flag should not be set for
2183 * manually configured addresses
2185 addrconf_dad_start(ifp, 0);
2191 return PTR_ERR(ifp);
2194 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2197 struct inet6_ifaddr *ifp;
2198 struct inet6_dev *idev;
2199 struct net_device *dev;
2204 dev = __dev_get_by_index(net, ifindex);
2208 if ((idev = __in6_dev_get(dev)) == NULL)
2211 read_lock_bh(&idev->lock);
2212 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2213 if (ifp->prefix_len == plen &&
2214 ipv6_addr_equal(pfx, &ifp->addr)) {
2216 read_unlock_bh(&idev->lock);
2220 /* If the last address is deleted administratively,
2221 disable IPv6 on this interface.
2223 if (list_empty(&idev->addr_list))
2224 addrconf_ifdown(idev->dev, 1);
2228 read_unlock_bh(&idev->lock);
2229 return -EADDRNOTAVAIL;
2233 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2235 struct in6_ifreq ireq;
2238 if (!capable(CAP_NET_ADMIN))
2241 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2245 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2246 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2247 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2252 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2254 struct in6_ifreq ireq;
2257 if (!capable(CAP_NET_ADMIN))
2260 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2264 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2265 ireq.ifr6_prefixlen);
2270 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2271 int plen, int scope)
2273 struct inet6_ifaddr *ifp;
2275 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2277 spin_lock_bh(&ifp->lock);
2278 ifp->flags &= ~IFA_F_TENTATIVE;
2279 spin_unlock_bh(&ifp->lock);
2280 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2285 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2286 static void sit_add_v4_addrs(struct inet6_dev *idev)
2288 struct in6_addr addr;
2289 struct net_device *dev;
2290 struct net *net = dev_net(idev->dev);
2295 memset(&addr, 0, sizeof(struct in6_addr));
2296 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2298 if (idev->dev->flags&IFF_POINTOPOINT) {
2299 addr.s6_addr32[0] = htonl(0xfe800000);
2302 scope = IPV6_ADDR_COMPATv4;
2305 if (addr.s6_addr32[3]) {
2306 add_addr(idev, &addr, 128, scope);
2310 for_each_netdev(net, dev) {
2311 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2312 if (in_dev && (dev->flags & IFF_UP)) {
2313 struct in_ifaddr * ifa;
2317 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2320 addr.s6_addr32[3] = ifa->ifa_local;
2322 if (ifa->ifa_scope == RT_SCOPE_LINK)
2324 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2325 if (idev->dev->flags&IFF_POINTOPOINT)
2329 if (idev->dev->flags&IFF_POINTOPOINT)
2334 add_addr(idev, &addr, plen, flag);
2341 static void init_loopback(struct net_device *dev)
2343 struct inet6_dev *idev;
2349 if ((idev = ipv6_find_idev(dev)) == NULL) {
2350 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2354 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2357 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2359 struct inet6_ifaddr * ifp;
2360 u32 addr_flags = IFA_F_PERMANENT;
2362 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2363 if (idev->cnf.optimistic_dad &&
2364 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2365 addr_flags |= IFA_F_OPTIMISTIC;
2369 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2371 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2372 addrconf_dad_start(ifp, 0);
2377 static void addrconf_dev_config(struct net_device *dev)
2379 struct in6_addr addr;
2380 struct inet6_dev * idev;
2384 if ((dev->type != ARPHRD_ETHER) &&
2385 (dev->type != ARPHRD_FDDI) &&
2386 (dev->type != ARPHRD_IEEE802_TR) &&
2387 (dev->type != ARPHRD_ARCNET) &&
2388 (dev->type != ARPHRD_INFINIBAND)) {
2389 /* Alas, we support only Ethernet autoconfiguration. */
2393 idev = addrconf_add_dev(dev);
2397 memset(&addr, 0, sizeof(struct in6_addr));
2398 addr.s6_addr32[0] = htonl(0xFE800000);
2400 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2401 addrconf_add_linklocal(idev, &addr);
2404 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2405 static void addrconf_sit_config(struct net_device *dev)
2407 struct inet6_dev *idev;
2412 * Configure the tunnel with one of our IPv4
2413 * addresses... we should configure all of
2414 * our v4 addrs in the tunnel
2417 if ((idev = ipv6_find_idev(dev)) == NULL) {
2418 printk(KERN_DEBUG "init sit: add_dev failed\n");
2422 if (dev->priv_flags & IFF_ISATAP) {
2423 struct in6_addr addr;
2425 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2426 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2427 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2428 addrconf_add_linklocal(idev, &addr);
2432 sit_add_v4_addrs(idev);
2434 if (dev->flags&IFF_POINTOPOINT) {
2435 addrconf_add_mroute(dev);
2436 addrconf_add_lroute(dev);
2443 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2445 struct in6_addr lladdr;
2447 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2448 addrconf_add_linklocal(idev, &lladdr);
2454 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2456 struct net_device *link_dev;
2457 struct net *net = dev_net(idev->dev);
2459 /* first try to inherit the link-local address from the link device */
2460 if (idev->dev->iflink &&
2461 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2462 if (!ipv6_inherit_linklocal(idev, link_dev))
2465 /* then try to inherit it from any device */
2466 for_each_netdev(net, link_dev) {
2467 if (!ipv6_inherit_linklocal(idev, link_dev))
2470 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2474 * Autoconfigure tunnel with a link-local address so routing protocols,
2475 * DHCPv6, MLD etc. can be run over the virtual link
2478 static void addrconf_ip6_tnl_config(struct net_device *dev)
2480 struct inet6_dev *idev;
2484 idev = addrconf_add_dev(dev);
2486 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2489 ip6_tnl_add_linklocal(idev);
2492 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2495 struct net_device *dev = (struct net_device *) data;
2496 struct inet6_dev *idev = __in6_dev_get(dev);
2497 int run_pending = 0;
2501 case NETDEV_REGISTER:
2502 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2503 idev = ipv6_add_dev(dev);
2505 return notifier_from_errno(-ENOMEM);
2511 if (dev->flags & IFF_SLAVE)
2514 if (event == NETDEV_UP) {
2515 if (!addrconf_qdisc_ok(dev)) {
2516 /* device is not ready yet. */
2518 "ADDRCONF(NETDEV_UP): %s: "
2519 "link is not ready\n",
2524 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2525 idev = ipv6_add_dev(dev);
2528 idev->if_flags |= IF_READY;
2532 if (!addrconf_qdisc_ok(dev)) {
2533 /* device is still not ready. */
2538 if (idev->if_flags & IF_READY)
2539 /* device is already configured. */
2541 idev->if_flags |= IF_READY;
2545 "ADDRCONF(NETDEV_CHANGE): %s: "
2546 "link becomes ready\n",
2552 switch (dev->type) {
2553 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2555 addrconf_sit_config(dev);
2558 case ARPHRD_TUNNEL6:
2559 addrconf_ip6_tnl_config(dev);
2561 case ARPHRD_LOOPBACK:
2566 addrconf_dev_config(dev);
2572 addrconf_dad_run(idev);
2575 * If the MTU changed during the interface down,
2576 * when the interface up, the changed MTU must be
2577 * reflected in the idev as well as routers.
2579 if (idev->cnf.mtu6 != dev->mtu &&
2580 dev->mtu >= IPV6_MIN_MTU) {
2581 rt6_mtu_change(dev, dev->mtu);
2582 idev->cnf.mtu6 = dev->mtu;
2584 idev->tstamp = jiffies;
2585 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2588 * If the changed mtu during down is lower than
2589 * IPV6_MIN_MTU stop IPv6 on this interface.
2591 if (dev->mtu < IPV6_MIN_MTU)
2592 addrconf_ifdown(dev, 1);
2596 case NETDEV_CHANGEMTU:
2597 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2598 rt6_mtu_change(dev, dev->mtu);
2599 idev->cnf.mtu6 = dev->mtu;
2603 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2604 idev = ipv6_add_dev(dev);
2610 * MTU falled under IPV6_MIN_MTU.
2611 * Stop IPv6 on this interface.
2615 case NETDEV_UNREGISTER:
2617 * Remove all addresses from this interface.
2619 addrconf_ifdown(dev, event != NETDEV_DOWN);
2622 case NETDEV_CHANGENAME:
2624 snmp6_unregister_dev(idev);
2625 addrconf_sysctl_unregister(idev);
2626 addrconf_sysctl_register(idev);
2627 err = snmp6_register_dev(idev);
2629 return notifier_from_errno(err);
2633 case NETDEV_PRE_TYPE_CHANGE:
2634 case NETDEV_POST_TYPE_CHANGE:
2635 addrconf_type_change(dev, event);
2643 * addrconf module should be notified of a device going up
2645 static struct notifier_block ipv6_dev_notf = {
2646 .notifier_call = addrconf_notify,
2649 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2651 struct inet6_dev *idev;
2654 idev = __in6_dev_get(dev);
2656 if (event == NETDEV_POST_TYPE_CHANGE)
2657 ipv6_mc_remap(idev);
2658 else if (event == NETDEV_PRE_TYPE_CHANGE)
2659 ipv6_mc_unmap(idev);
2662 static int addrconf_ifdown(struct net_device *dev, int how)
2664 struct net *net = dev_net(dev);
2665 struct inet6_dev *idev;
2666 struct inet6_ifaddr *ifa;
2667 LIST_HEAD(keep_list);
2672 rt6_ifdown(net, dev);
2673 neigh_ifdown(&nd_tbl, dev);
2675 idev = __in6_dev_get(dev);
2680 * Step 1: remove reference to ipv6 device from parent device.
2686 /* protected by rtnl_lock */
2687 rcu_assign_pointer(dev->ip6_ptr, NULL);
2689 /* Step 1.5: remove snmp6 entry */
2690 snmp6_unregister_dev(idev);
2694 write_lock_bh(&idev->lock);
2696 /* Step 2: clear flags for stateless addrconf */
2698 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2700 #ifdef CONFIG_IPV6_PRIVACY
2701 if (how && del_timer(&idev->regen_timer))
2704 /* Step 3: clear tempaddr list */
2705 while (!list_empty(&idev->tempaddr_list)) {
2706 ifa = list_first_entry(&idev->tempaddr_list,
2707 struct inet6_ifaddr, tmp_list);
2708 list_del(&ifa->tmp_list);
2709 write_unlock_bh(&idev->lock);
2710 spin_lock_bh(&ifa->lock);
2713 in6_ifa_put(ifa->ifpub);
2716 spin_unlock_bh(&ifa->lock);
2718 write_lock_bh(&idev->lock);
2722 while (!list_empty(&idev->addr_list)) {
2723 ifa = list_first_entry(&idev->addr_list,
2724 struct inet6_ifaddr, if_list);
2725 addrconf_del_timer(ifa);
2727 /* If just doing link down, and address is permanent
2728 and not link-local, then retain it. */
2730 (ifa->flags&IFA_F_PERMANENT) &&
2731 !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2732 list_move_tail(&ifa->if_list, &keep_list);
2734 /* If not doing DAD on this address, just keep it. */
2735 if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2736 idev->cnf.accept_dad <= 0 ||
2737 (ifa->flags & IFA_F_NODAD))
2740 /* If it was tentative already, no need to notify */
2741 if (ifa->flags & IFA_F_TENTATIVE)
2744 /* Flag it for later restoration when link comes up */
2745 ifa->flags |= IFA_F_TENTATIVE;
2746 ifa->state = INET6_IFADDR_STATE_DAD;
2748 list_del(&ifa->if_list);
2750 /* clear hash table */
2751 spin_lock_bh(&addrconf_hash_lock);
2752 hlist_del_init_rcu(&ifa->addr_lst);
2753 spin_unlock_bh(&addrconf_hash_lock);
2755 write_unlock_bh(&idev->lock);
2756 spin_lock_bh(&ifa->state_lock);
2758 ifa->state = INET6_IFADDR_STATE_DEAD;
2759 spin_unlock_bh(&ifa->state_lock);
2761 if (state != INET6_IFADDR_STATE_DEAD) {
2762 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2763 atomic_notifier_call_chain(&inet6addr_chain,
2768 write_lock_bh(&idev->lock);
2772 list_splice(&keep_list, &idev->addr_list);
2774 write_unlock_bh(&idev->lock);
2776 /* Step 5: Discard multicast list */
2778 ipv6_mc_destroy_dev(idev);
2782 idev->tstamp = jiffies;
2784 /* Last: Shot the device (if unregistered) */
2786 addrconf_sysctl_unregister(idev);
2787 neigh_parms_release(&nd_tbl, idev->nd_parms);
2788 neigh_ifdown(&nd_tbl, dev);
2794 static void addrconf_rs_timer(unsigned long data)
2796 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2797 struct inet6_dev *idev = ifp->idev;
2799 read_lock(&idev->lock);
2800 if (idev->dead || !(idev->if_flags & IF_READY))
2803 if (idev->cnf.forwarding)
2806 /* Announcement received after solicitation was sent */
2807 if (idev->if_flags & IF_RA_RCVD)
2810 spin_lock(&ifp->lock);
2811 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2812 /* The wait after the last probe can be shorter */
2813 addrconf_mod_timer(ifp, AC_RS,
2814 (ifp->probes == idev->cnf.rtr_solicits) ?
2815 idev->cnf.rtr_solicit_delay :
2816 idev->cnf.rtr_solicit_interval);
2817 spin_unlock(&ifp->lock);
2819 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2821 spin_unlock(&ifp->lock);
2823 * Note: we do not support deprecated "all on-link"
2824 * assumption any longer.
2826 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2831 read_unlock(&idev->lock);
2836 * Duplicate Address Detection
2838 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2840 unsigned long rand_num;
2841 struct inet6_dev *idev = ifp->idev;
2843 if (ifp->flags & IFA_F_OPTIMISTIC)
2846 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2848 ifp->probes = idev->cnf.dad_transmits;
2849 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2852 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2854 struct inet6_dev *idev = ifp->idev;
2855 struct net_device *dev = idev->dev;
2857 addrconf_join_solict(dev, &ifp->addr);
2859 net_srandom(ifp->addr.s6_addr32[3]);
2861 read_lock_bh(&idev->lock);
2862 spin_lock(&ifp->lock);
2863 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2866 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2867 idev->cnf.accept_dad < 1 ||
2868 !(ifp->flags&IFA_F_TENTATIVE) ||
2869 ifp->flags & IFA_F_NODAD) {
2870 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2871 spin_unlock(&ifp->lock);
2872 read_unlock_bh(&idev->lock);
2874 addrconf_dad_completed(ifp);
2878 if (!(idev->if_flags & IF_READY)) {
2879 spin_unlock(&ifp->lock);
2880 read_unlock_bh(&idev->lock);
2882 * If the device is not ready:
2883 * - keep it tentative if it is a permanent address.
2884 * - otherwise, kill it.
2887 addrconf_dad_stop(ifp, 0);
2892 * Optimistic nodes can start receiving
2895 if (ifp->flags & IFA_F_OPTIMISTIC)
2896 ip6_ins_rt(ifp->rt);
2898 addrconf_dad_kick(ifp);
2900 spin_unlock(&ifp->lock);
2901 read_unlock_bh(&idev->lock);
2904 static void addrconf_dad_timer(unsigned long data)
2906 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2907 struct inet6_dev *idev = ifp->idev;
2908 struct in6_addr mcaddr;
2910 if (!ifp->probes && addrconf_dad_end(ifp))
2913 read_lock(&idev->lock);
2914 if (idev->dead || !(idev->if_flags & IF_READY)) {
2915 read_unlock(&idev->lock);
2919 spin_lock(&ifp->lock);
2920 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2921 spin_unlock(&ifp->lock);
2922 read_unlock(&idev->lock);
2926 if (ifp->probes == 0) {
2928 * DAD was successful
2931 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2932 spin_unlock(&ifp->lock);
2933 read_unlock(&idev->lock);
2935 addrconf_dad_completed(ifp);
2941 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2942 spin_unlock(&ifp->lock);
2943 read_unlock(&idev->lock);
2945 /* send a neighbour solicitation for our addr */
2946 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2947 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2952 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2954 struct net_device *dev = ifp->idev->dev;
2957 * Configure the address for reception. Now it is valid.
2960 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2962 /* If added prefix is link local and forwarding is off,
2963 start sending router solicitations.
2966 if ((ifp->idev->cnf.forwarding == 0 ||
2967 ifp->idev->cnf.forwarding == 2) &&
2968 ifp->idev->cnf.rtr_solicits > 0 &&
2969 (dev->flags&IFF_LOOPBACK) == 0 &&
2970 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2972 * If a host as already performed a random delay
2973 * [...] as part of DAD [...] there is no need
2974 * to delay again before sending the first RS
2976 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2978 spin_lock_bh(&ifp->lock);
2980 ifp->idev->if_flags |= IF_RS_SENT;
2981 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2982 spin_unlock_bh(&ifp->lock);
2986 static void addrconf_dad_run(struct inet6_dev *idev)
2988 struct inet6_ifaddr *ifp;
2990 read_lock_bh(&idev->lock);
2991 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2992 spin_lock(&ifp->lock);
2993 if (ifp->flags & IFA_F_TENTATIVE &&
2994 ifp->state == INET6_IFADDR_STATE_DAD)
2995 addrconf_dad_kick(ifp);
2996 spin_unlock(&ifp->lock);
2998 read_unlock_bh(&idev->lock);
3001 #ifdef CONFIG_PROC_FS
3002 struct if6_iter_state {
3003 struct seq_net_private p;
3007 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3009 struct inet6_ifaddr *ifa = NULL;
3010 struct if6_iter_state *state = seq->private;
3011 struct net *net = seq_file_net(seq);
3013 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3014 struct hlist_node *n;
3015 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3017 if (net_eq(dev_net(ifa->idev->dev), net))
3023 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3024 struct inet6_ifaddr *ifa)
3026 struct if6_iter_state *state = seq->private;
3027 struct net *net = seq_file_net(seq);
3028 struct hlist_node *n = &ifa->addr_lst;
3030 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3031 if (net_eq(dev_net(ifa->idev->dev), net))
3034 while (++state->bucket < IN6_ADDR_HSIZE) {
3035 hlist_for_each_entry_rcu_bh(ifa, n,
3036 &inet6_addr_lst[state->bucket], addr_lst) {
3037 if (net_eq(dev_net(ifa->idev->dev), net))
3045 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3047 struct inet6_ifaddr *ifa = if6_get_first(seq);
3050 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3052 return pos ? NULL : ifa;
3055 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3059 return if6_get_idx(seq, *pos);
3062 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3064 struct inet6_ifaddr *ifa;
3066 ifa = if6_get_next(seq, v);
3071 static void if6_seq_stop(struct seq_file *seq, void *v)
3074 rcu_read_unlock_bh();
3077 static int if6_seq_show(struct seq_file *seq, void *v)
3079 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3080 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3082 ifp->idev->dev->ifindex,
3086 ifp->idev->dev->name);
3090 static const struct seq_operations if6_seq_ops = {
3091 .start = if6_seq_start,
3092 .next = if6_seq_next,
3093 .show = if6_seq_show,
3094 .stop = if6_seq_stop,
3097 static int if6_seq_open(struct inode *inode, struct file *file)
3099 return seq_open_net(inode, file, &if6_seq_ops,
3100 sizeof(struct if6_iter_state));
3103 static const struct file_operations if6_fops = {
3104 .owner = THIS_MODULE,
3105 .open = if6_seq_open,
3107 .llseek = seq_lseek,
3108 .release = seq_release_net,
3111 static int __net_init if6_proc_net_init(struct net *net)
3113 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3118 static void __net_exit if6_proc_net_exit(struct net *net)
3120 proc_net_remove(net, "if_inet6");
3123 static struct pernet_operations if6_proc_net_ops = {
3124 .init = if6_proc_net_init,
3125 .exit = if6_proc_net_exit,
3128 int __init if6_proc_init(void)
3130 return register_pernet_subsys(&if6_proc_net_ops);
3133 void if6_proc_exit(void)
3135 unregister_pernet_subsys(&if6_proc_net_ops);
3137 #endif /* CONFIG_PROC_FS */
3139 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3140 /* Check if address is a home address configured on any interface. */
3141 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3144 struct inet6_ifaddr *ifp = NULL;
3145 struct hlist_node *n;
3146 unsigned int hash = ipv6_addr_hash(addr);
3149 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3150 if (!net_eq(dev_net(ifp->idev->dev), net))