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>
90 #include <linux/export.h>
92 /* Set to 3 to get tracing... */
96 #define ADBG(x) printk x
101 #define INFINITY_LIFE_TIME 0xFFFFFFFF
103 static inline u32 cstamp_delta(unsigned long cstamp)
105 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
108 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
109 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
110 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
113 static void addrconf_sysctl_register(struct inet6_dev *idev);
114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
125 #ifdef CONFIG_IPV6_PRIVACY
126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
128 static void ipv6_regen_rndid(unsigned long data);
131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
132 static int ipv6_count_addresses(struct inet6_dev *idev);
135 * Configured unicast address hash table
137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
138 static DEFINE_SPINLOCK(addrconf_hash_lock);
140 static void addrconf_verify(unsigned long);
142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
143 static DEFINE_SPINLOCK(addrconf_verify_lock);
145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
148 static void addrconf_type_change(struct net_device *dev,
149 unsigned long event);
150 static int addrconf_ifdown(struct net_device *dev, int how);
152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
153 static void addrconf_dad_timer(unsigned long data);
154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
155 static void addrconf_dad_run(struct inet6_dev *idev);
156 static void addrconf_rs_timer(unsigned long data);
157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
161 struct prefix_info *pinfo);
162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
163 struct net_device *dev);
165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
169 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
170 .mtu6 = IPV6_MIN_MTU,
172 .accept_redirects = 1,
174 .force_mld_version = 0,
176 .rtr_solicits = MAX_RTR_SOLICITATIONS,
177 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
178 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
179 #ifdef CONFIG_IPV6_PRIVACY
181 .temp_valid_lft = TEMP_VALID_LIFETIME,
182 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
183 .regen_max_retry = REGEN_MAX_RETRY,
184 .max_desync_factor = MAX_DESYNC_FACTOR,
186 .max_addresses = IPV6_MAX_ADDRESSES,
187 .accept_ra_defrtr = 1,
188 .accept_ra_pinfo = 1,
189 #ifdef CONFIG_IPV6_ROUTER_PREF
190 .accept_ra_rtr_pref = 1,
191 .rtr_probe_interval = 60 * HZ,
192 #ifdef CONFIG_IPV6_ROUTE_INFO
193 .accept_ra_rt_info_max_plen = 0,
197 .accept_source_route = 0, /* we do not accept RH0 by default. */
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
204 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
205 .mtu6 = IPV6_MIN_MTU,
207 .accept_redirects = 1,
210 .rtr_solicits = MAX_RTR_SOLICITATIONS,
211 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
212 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
213 #ifdef CONFIG_IPV6_PRIVACY
215 .temp_valid_lft = TEMP_VALID_LIFETIME,
216 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
217 .regen_max_retry = REGEN_MAX_RETRY,
218 .max_desync_factor = MAX_DESYNC_FACTOR,
220 .max_addresses = IPV6_MAX_ADDRESSES,
221 .accept_ra_defrtr = 1,
222 .accept_ra_pinfo = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 .accept_ra_rtr_pref = 1,
225 .rtr_probe_interval = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 .accept_ra_rt_info_max_plen = 0,
231 .accept_source_route = 0, /* we do not accept RH0 by default. */
236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
242 /* Check if a valid qdisc is available */
243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
245 return !qdisc_tx_is_noop(dev);
248 /* Check if a route is valid prefix route */
249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
251 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
256 if (del_timer(&ifp->timer))
260 enum addrconf_timer_t {
266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
267 enum addrconf_timer_t what,
270 if (!del_timer(&ifp->timer))
275 ifp->timer.function = addrconf_dad_timer;
278 ifp->timer.function = addrconf_rs_timer;
283 ifp->timer.expires = jiffies + when;
284 add_timer(&ifp->timer);
287 static int snmp6_alloc_dev(struct inet6_dev *idev)
289 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
290 sizeof(struct ipstats_mib),
291 __alignof__(struct ipstats_mib)) < 0)
293 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
295 if (!idev->stats.icmpv6dev)
297 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
299 if (!idev->stats.icmpv6msgdev)
305 kfree(idev->stats.icmpv6dev);
307 snmp_mib_free((void __percpu **)idev->stats.ipv6);
312 static void snmp6_free_dev(struct inet6_dev *idev)
314 kfree(idev->stats.icmpv6msgdev);
315 kfree(idev->stats.icmpv6dev);
316 snmp_mib_free((void __percpu **)idev->stats.ipv6);
319 /* Nobody refers to this device, we may destroy it. */
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
323 struct net_device *dev = idev->dev;
325 WARN_ON(!list_empty(&idev->addr_list));
326 WARN_ON(idev->mc_list != NULL);
328 #ifdef NET_REFCNT_DEBUG
329 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
333 pr_warning("Freeing alive inet6 device %p\n", idev);
336 snmp6_free_dev(idev);
337 kfree_rcu(idev, rcu);
340 EXPORT_SYMBOL(in6_dev_finish_destroy);
342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
344 struct inet6_dev *ndev;
348 if (dev->mtu < IPV6_MIN_MTU)
351 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
356 rwlock_init(&ndev->lock);
358 INIT_LIST_HEAD(&ndev->addr_list);
360 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
361 ndev->cnf.mtu6 = dev->mtu;
362 ndev->cnf.sysctl = NULL;
363 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 if (ndev->nd_parms == NULL) {
368 if (ndev->cnf.forwarding)
369 dev_disable_lro(dev);
370 /* We refer to the device */
373 if (snmp6_alloc_dev(ndev) < 0) {
375 "%s(): cannot allocate memory for statistics; dev=%s.\n",
376 __func__, dev->name));
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
383 if (snmp6_register_dev(ndev) < 0) {
385 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
386 __func__, dev->name));
387 neigh_parms_release(&nd_tbl, ndev->nd_parms);
389 in6_dev_finish_destroy(ndev);
393 /* One reference from device. We must do this before
394 * we invoke __ipv6_regen_rndid().
398 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
399 ndev->cnf.accept_dad = -1;
401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
402 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
404 "%s: Disabled Multicast RS\n",
406 ndev->cnf.rtr_solicits = 0;
410 #ifdef CONFIG_IPV6_PRIVACY
411 INIT_LIST_HEAD(&ndev->tempaddr_list);
412 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413 if ((dev->flags&IFF_LOOPBACK) ||
414 dev->type == ARPHRD_TUNNEL ||
415 dev->type == ARPHRD_TUNNEL6 ||
416 dev->type == ARPHRD_SIT ||
417 dev->type == ARPHRD_NONE) {
418 ndev->cnf.use_tempaddr = -1;
421 ipv6_regen_rndid((unsigned long) ndev);
425 if (netif_running(dev) && addrconf_qdisc_ok(dev))
426 ndev->if_flags |= IF_READY;
428 ipv6_mc_init_dev(ndev);
429 ndev->tstamp = jiffies;
430 addrconf_sysctl_register(ndev);
431 /* protected by rtnl_lock */
432 rcu_assign_pointer(dev->ip6_ptr, ndev);
434 /* Join all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
437 /* Join all-router multicast group if forwarding is set */
438 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
439 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
446 struct inet6_dev *idev;
450 idev = __in6_dev_get(dev);
452 idev = ipv6_add_dev(dev);
457 if (dev->flags&IFF_UP)
463 static void dev_forward_change(struct inet6_dev *idev)
465 struct net_device *dev;
466 struct inet6_ifaddr *ifa;
471 if (idev->cnf.forwarding)
472 dev_disable_lro(dev);
473 if (dev && (dev->flags & IFF_MULTICAST)) {
474 if (idev->cnf.forwarding)
475 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
477 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
480 list_for_each_entry(ifa, &idev->addr_list, if_list) {
481 if (ifa->flags&IFA_F_TENTATIVE)
483 if (idev->cnf.forwarding)
484 addrconf_join_anycast(ifa);
486 addrconf_leave_anycast(ifa);
491 static void addrconf_forward_change(struct net *net, __s32 newf)
493 struct net_device *dev;
494 struct inet6_dev *idev;
496 for_each_netdev(net, dev) {
497 idev = __in6_dev_get(dev);
499 int changed = (!idev->cnf.forwarding) ^ (!newf);
500 idev->cnf.forwarding = newf;
502 dev_forward_change(idev);
507 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
513 return restart_syscall();
515 net = (struct net *)table->extra2;
519 if (p == &net->ipv6.devconf_dflt->forwarding) {
524 if (p == &net->ipv6.devconf_all->forwarding) {
525 net->ipv6.devconf_dflt->forwarding = newf;
526 addrconf_forward_change(net, newf);
527 } else if ((!newf) ^ (!old))
528 dev_forward_change((struct inet6_dev *)table->extra1);
532 rt6_purge_dflt_routers(net);
537 /* Nobody refers to this ifaddr, destroy it */
538 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
540 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
542 #ifdef NET_REFCNT_DEBUG
543 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
546 in6_dev_put(ifp->idev);
548 if (del_timer(&ifp->timer))
549 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
551 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
552 pr_warning("Freeing alive inet6 address %p\n", ifp);
555 dst_release(&ifp->rt->dst);
561 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
564 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
567 * Each device address list is sorted in order of scope -
568 * global before linklocal.
570 list_for_each(p, &idev->addr_list) {
571 struct inet6_ifaddr *ifa
572 = list_entry(p, struct inet6_ifaddr, if_list);
573 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
577 list_add_tail(&ifp->if_list, p);
580 static u32 ipv6_addr_hash(const struct in6_addr *addr)
583 * We perform the hash function over the last 64 bits of the address
584 * This will include the IEEE address token on links that support it.
586 return jhash_2words((__force u32)addr->s6_addr32[2],
587 (__force u32)addr->s6_addr32[3], 0)
588 & (IN6_ADDR_HSIZE - 1);
591 /* On success it returns ifp with increased reference count */
593 static struct inet6_ifaddr *
594 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
595 int scope, u32 flags)
597 struct inet6_ifaddr *ifa = NULL;
601 int addr_type = ipv6_addr_type(addr);
603 if (addr_type == IPV6_ADDR_ANY ||
604 addr_type & IPV6_ADDR_MULTICAST ||
605 (!(idev->dev->flags & IFF_LOOPBACK) &&
606 addr_type & IPV6_ADDR_LOOPBACK))
607 return ERR_PTR(-EADDRNOTAVAIL);
611 err = -ENODEV; /*XXX*/
615 if (idev->cnf.disable_ipv6) {
620 spin_lock(&addrconf_hash_lock);
622 /* Ignore adding duplicate addresses on an interface */
623 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
624 ADBG(("ipv6_add_addr: already assigned\n"));
629 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
632 ADBG(("ipv6_add_addr: malloc failed\n"));
637 rt = addrconf_dst_alloc(idev, addr, false);
645 spin_lock_init(&ifa->lock);
646 spin_lock_init(&ifa->state_lock);
647 init_timer(&ifa->timer);
648 INIT_HLIST_NODE(&ifa->addr_lst);
649 ifa->timer.data = (unsigned long) ifa;
651 ifa->prefix_len = pfxlen;
652 ifa->flags = flags | IFA_F_TENTATIVE;
653 ifa->cstamp = ifa->tstamp = jiffies;
662 /* Add to big hash table */
663 hash = ipv6_addr_hash(addr);
665 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
666 spin_unlock(&addrconf_hash_lock);
668 write_lock(&idev->lock);
669 /* Add to inet6_dev unicast addr list. */
670 ipv6_link_dev_addr(idev, ifa);
672 #ifdef CONFIG_IPV6_PRIVACY
673 if (ifa->flags&IFA_F_TEMPORARY) {
674 list_add(&ifa->tmp_list, &idev->tempaddr_list);
680 write_unlock(&idev->lock);
682 rcu_read_unlock_bh();
684 if (likely(err == 0))
685 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
693 spin_unlock(&addrconf_hash_lock);
697 /* This function wants to get referenced ifp and releases it before return */
699 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
701 struct inet6_ifaddr *ifa, *ifn;
702 struct inet6_dev *idev = ifp->idev;
704 int deleted = 0, onlink = 0;
705 unsigned long expires = jiffies;
707 spin_lock_bh(&ifp->state_lock);
709 ifp->state = INET6_IFADDR_STATE_DEAD;
710 spin_unlock_bh(&ifp->state_lock);
712 if (state == INET6_IFADDR_STATE_DEAD)
715 spin_lock_bh(&addrconf_hash_lock);
716 hlist_del_init_rcu(&ifp->addr_lst);
717 spin_unlock_bh(&addrconf_hash_lock);
719 write_lock_bh(&idev->lock);
720 #ifdef CONFIG_IPV6_PRIVACY
721 if (ifp->flags&IFA_F_TEMPORARY) {
722 list_del(&ifp->tmp_list);
724 in6_ifa_put(ifp->ifpub);
731 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
733 list_del_init(&ifp->if_list);
736 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
740 } else if (ifp->flags & IFA_F_PERMANENT) {
741 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
743 if (ifa->flags & IFA_F_PERMANENT) {
748 unsigned long lifetime;
753 spin_lock(&ifa->lock);
755 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
757 * Note: Because this address is
758 * not permanent, lifetime <
759 * LONG_MAX / HZ here.
761 if (time_before(expires,
762 ifa->tstamp + lifetime * HZ))
763 expires = ifa->tstamp + lifetime * HZ;
764 spin_unlock(&ifa->lock);
769 write_unlock_bh(&idev->lock);
771 addrconf_del_timer(ifp);
773 ipv6_ifa_notify(RTM_DELADDR, ifp);
775 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
778 * Purge or update corresponding prefix
780 * 1) we don't purge prefix here if address was not permanent.
781 * prefix is managed by its own lifetime.
782 * 2) if there're no addresses, delete prefix.
783 * 3) if there're still other permanent address(es),
784 * corresponding prefix is still permanent.
785 * 4) otherwise, update prefix lifetime to the
786 * longest valid lifetime among the corresponding
787 * addresses on the device.
788 * Note: subsequent RA will update lifetime.
792 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
793 struct in6_addr prefix;
795 struct net *net = dev_net(ifp->idev->dev);
796 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
797 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
799 if (rt && addrconf_is_prefix_route(rt)) {
803 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
804 rt6_set_expires(rt, expires);
807 dst_release(&rt->dst);
810 /* clean up prefsrc entries */
811 rt6_remove_prefsrc(ifp);
816 #ifdef CONFIG_IPV6_PRIVACY
817 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
819 struct inet6_dev *idev = ifp->idev;
820 struct in6_addr addr, *tmpaddr;
821 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
822 unsigned long regen_advance;
827 unsigned long now = jiffies;
829 write_lock(&idev->lock);
831 spin_lock_bh(&ift->lock);
832 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
833 spin_unlock_bh(&ift->lock);
840 if (idev->cnf.use_tempaddr <= 0) {
841 write_unlock(&idev->lock);
843 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
848 spin_lock_bh(&ifp->lock);
849 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
850 idev->cnf.use_tempaddr = -1; /*XXX*/
851 spin_unlock_bh(&ifp->lock);
852 write_unlock(&idev->lock);
854 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
860 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
861 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
862 spin_unlock_bh(&ifp->lock);
863 write_unlock(&idev->lock);
865 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
871 memcpy(&addr.s6_addr[8], idev->rndid, 8);
872 age = (now - ifp->tstamp) / HZ;
873 tmp_valid_lft = min_t(__u32,
875 idev->cnf.temp_valid_lft + age);
876 tmp_prefered_lft = min_t(__u32,
878 idev->cnf.temp_prefered_lft + age -
879 idev->cnf.max_desync_factor);
880 tmp_plen = ifp->prefix_len;
881 max_addresses = idev->cnf.max_addresses;
882 tmp_tstamp = ifp->tstamp;
883 spin_unlock_bh(&ifp->lock);
885 regen_advance = idev->cnf.regen_max_retry *
886 idev->cnf.dad_transmits *
887 idev->nd_parms->retrans_time / HZ;
888 write_unlock(&idev->lock);
890 /* A temporary address is created only if this calculated Preferred
891 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
892 * an implementation must not create a temporary address with a zero
893 * Preferred Lifetime.
895 if (tmp_prefered_lft <= regen_advance) {
902 addr_flags = IFA_F_TEMPORARY;
903 /* set in addrconf_prefix_rcv() */
904 if (ifp->flags & IFA_F_OPTIMISTIC)
905 addr_flags |= IFA_F_OPTIMISTIC;
907 ift = !max_addresses ||
908 ipv6_count_addresses(idev) < max_addresses ?
909 ipv6_add_addr(idev, &addr, tmp_plen,
910 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
912 if (!ift || IS_ERR(ift)) {
916 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
918 write_lock(&idev->lock);
922 spin_lock_bh(&ift->lock);
924 ift->valid_lft = tmp_valid_lft;
925 ift->prefered_lft = tmp_prefered_lft;
927 ift->tstamp = tmp_tstamp;
928 spin_unlock_bh(&ift->lock);
930 addrconf_dad_start(ift, 0);
939 * Choose an appropriate source address (RFC3484)
942 IPV6_SADDR_RULE_INIT = 0,
943 IPV6_SADDR_RULE_LOCAL,
944 IPV6_SADDR_RULE_SCOPE,
945 IPV6_SADDR_RULE_PREFERRED,
946 #ifdef CONFIG_IPV6_MIP6
950 IPV6_SADDR_RULE_LABEL,
951 #ifdef CONFIG_IPV6_PRIVACY
952 IPV6_SADDR_RULE_PRIVACY,
954 IPV6_SADDR_RULE_ORCHID,
955 IPV6_SADDR_RULE_PREFIX,
959 struct ipv6_saddr_score {
962 struct inet6_ifaddr *ifa;
963 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
968 struct ipv6_saddr_dst {
969 const struct in6_addr *addr;
976 static inline int ipv6_saddr_preferred(int type)
978 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
983 static int ipv6_get_saddr_eval(struct net *net,
984 struct ipv6_saddr_score *score,
985 struct ipv6_saddr_dst *dst,
990 if (i <= score->rule) {
992 case IPV6_SADDR_RULE_SCOPE:
993 ret = score->scopedist;
995 case IPV6_SADDR_RULE_PREFIX:
996 ret = score->matchlen;
999 ret = !!test_bit(i, score->scorebits);
1005 case IPV6_SADDR_RULE_INIT:
1006 /* Rule 0: remember if hiscore is not ready yet */
1009 case IPV6_SADDR_RULE_LOCAL:
1010 /* Rule 1: Prefer same address */
1011 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1013 case IPV6_SADDR_RULE_SCOPE:
1014 /* Rule 2: Prefer appropriate scope
1019 * ---+--+-+---> scope
1021 * | d is scope of the destination.
1023 * | \ <- smaller scope is better if
1024 * B-15 | \ if scope is enough for destinaion.
1025 * | ret = B - scope (-1 <= scope >= d <= 15).
1027 * |/ <- greater is better
1028 * -C / if scope is not enough for destination.
1029 * /| ret = scope - C (-1 <= d < scope <= 15).
1031 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1032 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1033 * Assume B = 0 and we get C > 29.
1035 ret = __ipv6_addr_src_scope(score->addr_type);
1036 if (ret >= dst->scope)
1039 ret -= 128; /* 30 is enough */
1040 score->scopedist = ret;
1042 case IPV6_SADDR_RULE_PREFERRED:
1043 /* Rule 3: Avoid deprecated and optimistic addresses */
1044 ret = ipv6_saddr_preferred(score->addr_type) ||
1045 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1047 #ifdef CONFIG_IPV6_MIP6
1048 case IPV6_SADDR_RULE_HOA:
1050 /* Rule 4: Prefer home address */
1051 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1052 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1056 case IPV6_SADDR_RULE_OIF:
1057 /* Rule 5: Prefer outgoing interface */
1058 ret = (!dst->ifindex ||
1059 dst->ifindex == score->ifa->idev->dev->ifindex);
1061 case IPV6_SADDR_RULE_LABEL:
1062 /* Rule 6: Prefer matching label */
1063 ret = ipv6_addr_label(net,
1064 &score->ifa->addr, score->addr_type,
1065 score->ifa->idev->dev->ifindex) == dst->label;
1067 #ifdef CONFIG_IPV6_PRIVACY
1068 case IPV6_SADDR_RULE_PRIVACY:
1070 /* Rule 7: Prefer public address
1071 * Note: prefer temporary address if use_tempaddr >= 2
1073 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1074 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1075 score->ifa->idev->cnf.use_tempaddr >= 2;
1076 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1080 case IPV6_SADDR_RULE_ORCHID:
1081 /* Rule 8-: Prefer ORCHID vs ORCHID or
1082 * non-ORCHID vs non-ORCHID
1084 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1085 ipv6_addr_orchid(dst->addr));
1087 case IPV6_SADDR_RULE_PREFIX:
1088 /* Rule 8: Use longest matching prefix */
1089 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1097 __set_bit(i, score->scorebits);
1103 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1104 const struct in6_addr *daddr, unsigned int prefs,
1105 struct in6_addr *saddr)
1107 struct ipv6_saddr_score scores[2],
1108 *score = &scores[0], *hiscore = &scores[1];
1109 struct ipv6_saddr_dst dst;
1110 struct net_device *dev;
1113 dst_type = __ipv6_addr_type(daddr);
1115 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1116 dst.scope = __ipv6_addr_src_scope(dst_type);
1117 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1121 hiscore->ifa = NULL;
1125 for_each_netdev_rcu(net, dev) {
1126 struct inet6_dev *idev;
1128 /* Candidate Source Address (section 4)
1129 * - multicast and link-local destination address,
1130 * the set of candidate source address MUST only
1131 * include addresses assigned to interfaces
1132 * belonging to the same link as the outgoing
1134 * (- For site-local destination addresses, the
1135 * set of candidate source addresses MUST only
1136 * include addresses assigned to interfaces
1137 * belonging to the same site as the outgoing
1140 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1141 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1142 dst.ifindex && dev->ifindex != dst.ifindex)
1145 idev = __in6_dev_get(dev);
1149 read_lock_bh(&idev->lock);
1150 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1154 * - Tentative Address (RFC2462 section 5.4)
1155 * - A tentative address is not considered
1156 * "assigned to an interface" in the traditional
1157 * sense, unless it is also flagged as optimistic.
1158 * - Candidate Source Address (section 4)
1159 * - In any case, anycast addresses, multicast
1160 * addresses, and the unspecified address MUST
1161 * NOT be included in a candidate set.
1163 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1164 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1167 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1169 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1170 score->addr_type & IPV6_ADDR_MULTICAST)) {
1171 LIMIT_NETDEBUG(KERN_DEBUG
1172 "ADDRCONF: unspecified / multicast address "
1173 "assigned as unicast address on %s",
1179 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1181 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1182 int minihiscore, miniscore;
1184 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1185 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1187 if (minihiscore > miniscore) {
1188 if (i == IPV6_SADDR_RULE_SCOPE &&
1189 score->scopedist > 0) {
1192 * each remaining entry
1193 * has too small (not enough)
1194 * scope, because ifa entries
1195 * are sorted by their scope
1201 } else if (minihiscore < miniscore) {
1203 in6_ifa_put(hiscore->ifa);
1205 in6_ifa_hold(score->ifa);
1207 swap(hiscore, score);
1209 /* restore our iterator */
1210 score->ifa = hiscore->ifa;
1217 read_unlock_bh(&idev->lock);
1222 return -EADDRNOTAVAIL;
1224 *saddr = hiscore->ifa->addr;
1225 in6_ifa_put(hiscore->ifa);
1228 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1230 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1231 unsigned char banned_flags)
1233 struct inet6_dev *idev;
1234 int err = -EADDRNOTAVAIL;
1237 idev = __in6_dev_get(dev);
1239 struct inet6_ifaddr *ifp;
1241 read_lock_bh(&idev->lock);
1242 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1243 if (ifp->scope == IFA_LINK &&
1244 !(ifp->flags & banned_flags)) {
1250 read_unlock_bh(&idev->lock);
1256 static int ipv6_count_addresses(struct inet6_dev *idev)
1259 struct inet6_ifaddr *ifp;
1261 read_lock_bh(&idev->lock);
1262 list_for_each_entry(ifp, &idev->addr_list, if_list)
1264 read_unlock_bh(&idev->lock);
1268 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1269 struct net_device *dev, int strict)
1271 struct inet6_ifaddr *ifp;
1272 struct hlist_node *node;
1273 unsigned int hash = ipv6_addr_hash(addr);
1276 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1277 if (!net_eq(dev_net(ifp->idev->dev), net))
1279 if (ipv6_addr_equal(&ifp->addr, addr) &&
1280 !(ifp->flags&IFA_F_TENTATIVE) &&
1281 (dev == NULL || ifp->idev->dev == dev ||
1282 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1283 rcu_read_unlock_bh();
1288 rcu_read_unlock_bh();
1291 EXPORT_SYMBOL(ipv6_chk_addr);
1293 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1294 struct net_device *dev)
1296 unsigned int hash = ipv6_addr_hash(addr);
1297 struct inet6_ifaddr *ifp;
1298 struct hlist_node *node;
1300 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1301 if (!net_eq(dev_net(ifp->idev->dev), net))
1303 if (ipv6_addr_equal(&ifp->addr, addr)) {
1304 if (dev == NULL || ifp->idev->dev == dev)
1311 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1313 struct inet6_dev *idev;
1314 struct inet6_ifaddr *ifa;
1319 idev = __in6_dev_get(dev);
1321 read_lock_bh(&idev->lock);
1322 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1323 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1328 read_unlock_bh(&idev->lock);
1334 EXPORT_SYMBOL(ipv6_chk_prefix);
1336 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1337 struct net_device *dev, int strict)
1339 struct inet6_ifaddr *ifp, *result = NULL;
1340 unsigned int hash = ipv6_addr_hash(addr);
1341 struct hlist_node *node;
1344 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1345 if (!net_eq(dev_net(ifp->idev->dev), net))
1347 if (ipv6_addr_equal(&ifp->addr, addr)) {
1348 if (dev == NULL || ifp->idev->dev == dev ||
1349 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1356 rcu_read_unlock_bh();
1361 /* Gets referenced address, destroys ifaddr */
1363 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1365 if (ifp->flags&IFA_F_PERMANENT) {
1366 spin_lock_bh(&ifp->lock);
1367 addrconf_del_timer(ifp);
1368 ifp->flags |= IFA_F_TENTATIVE;
1370 ifp->flags |= IFA_F_DADFAILED;
1371 spin_unlock_bh(&ifp->lock);
1373 ipv6_ifa_notify(0, ifp);
1375 #ifdef CONFIG_IPV6_PRIVACY
1376 } else if (ifp->flags&IFA_F_TEMPORARY) {
1377 struct inet6_ifaddr *ifpub;
1378 spin_lock_bh(&ifp->lock);
1381 in6_ifa_hold(ifpub);
1382 spin_unlock_bh(&ifp->lock);
1383 ipv6_create_tempaddr(ifpub, ifp);
1386 spin_unlock_bh(&ifp->lock);
1394 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1398 spin_lock(&ifp->state_lock);
1399 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1400 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1403 spin_unlock(&ifp->state_lock);
1408 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1410 struct inet6_dev *idev = ifp->idev;
1412 if (addrconf_dad_end(ifp)) {
1417 if (net_ratelimit())
1418 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1419 ifp->idev->dev->name, &ifp->addr);
1421 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1422 struct in6_addr addr;
1424 addr.s6_addr32[0] = htonl(0xfe800000);
1425 addr.s6_addr32[1] = 0;
1427 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1428 ipv6_addr_equal(&ifp->addr, &addr)) {
1429 /* DAD failed for link-local based on MAC address */
1430 idev->cnf.disable_ipv6 = 1;
1432 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1433 ifp->idev->dev->name);
1437 addrconf_dad_stop(ifp, 1);
1440 /* Join to solicited addr multicast group. */
1442 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1444 struct in6_addr maddr;
1446 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1449 addrconf_addr_solict_mult(addr, &maddr);
1450 ipv6_dev_mc_inc(dev, &maddr);
1453 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1455 struct in6_addr maddr;
1457 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1460 addrconf_addr_solict_mult(addr, &maddr);
1461 __ipv6_dev_mc_dec(idev, &maddr);
1464 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1466 struct in6_addr addr;
1467 if (ifp->prefix_len == 127) /* RFC 6164 */
1469 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1470 if (ipv6_addr_any(&addr))
1472 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1475 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1477 struct in6_addr addr;
1478 if (ifp->prefix_len == 127) /* RFC 6164 */
1480 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1481 if (ipv6_addr_any(&addr))
1483 __ipv6_dev_ac_dec(ifp->idev, &addr);
1486 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1488 if (dev->addr_len != ETH_ALEN)
1490 memcpy(eui, dev->dev_addr, 3);
1491 memcpy(eui + 5, dev->dev_addr + 3, 3);
1494 * The zSeries OSA network cards can be shared among various
1495 * OS instances, but the OSA cards have only one MAC address.
1496 * This leads to duplicate address conflicts in conjunction
1497 * with IPv6 if more than one instance uses the same card.
1499 * The driver for these cards can deliver a unique 16-bit
1500 * identifier for each instance sharing the same card. It is
1501 * placed instead of 0xFFFE in the interface identifier. The
1502 * "u" bit of the interface identifier is not inverted in this
1503 * case. Hence the resulting interface identifier has local
1504 * scope according to RFC2373.
1507 eui[3] = (dev->dev_id >> 8) & 0xFF;
1508 eui[4] = dev->dev_id & 0xFF;
1517 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1519 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1520 if (dev->addr_len != ARCNET_ALEN)
1523 eui[7] = *(u8*)dev->dev_addr;
1527 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1529 if (dev->addr_len != INFINIBAND_ALEN)
1531 memcpy(eui, dev->dev_addr + 12, 8);
1536 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1540 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1541 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1542 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1543 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1544 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1545 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1549 memcpy(eui + 4, &addr, 4);
1553 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1555 if (dev->priv_flags & IFF_ISATAP)
1556 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1560 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1562 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1565 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1567 switch (dev->type) {
1570 case ARPHRD_IEEE802_TR:
1571 return addrconf_ifid_eui48(eui, dev);
1573 return addrconf_ifid_arcnet(eui, dev);
1574 case ARPHRD_INFINIBAND:
1575 return addrconf_ifid_infiniband(eui, dev);
1577 return addrconf_ifid_sit(eui, dev);
1579 return addrconf_ifid_gre(eui, dev);
1584 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1587 struct inet6_ifaddr *ifp;
1589 read_lock_bh(&idev->lock);
1590 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1591 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1592 memcpy(eui, ifp->addr.s6_addr+8, 8);
1597 read_unlock_bh(&idev->lock);
1601 #ifdef CONFIG_IPV6_PRIVACY
1602 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1603 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1606 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1607 idev->rndid[0] &= ~0x02;
1610 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1611 * check if generated address is not inappropriate
1613 * - Reserved subnet anycast (RFC 2526)
1614 * 11111101 11....11 1xxxxxxx
1615 * - ISATAP (RFC4214) 6.1
1616 * 00-00-5E-FE-xx-xx-xx-xx
1618 * - XXX: already assigned to an address on the device
1620 if (idev->rndid[0] == 0xfd &&
1621 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1622 (idev->rndid[7]&0x80))
1624 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1625 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1627 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1634 static void ipv6_regen_rndid(unsigned long data)
1636 struct inet6_dev *idev = (struct inet6_dev *) data;
1637 unsigned long expires;
1640 write_lock_bh(&idev->lock);
1645 if (__ipv6_regen_rndid(idev) < 0)
1649 idev->cnf.temp_prefered_lft * HZ -
1650 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1651 idev->cnf.max_desync_factor * HZ;
1652 if (time_before(expires, jiffies)) {
1654 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1659 if (!mod_timer(&idev->regen_timer, expires))
1663 write_unlock_bh(&idev->lock);
1664 rcu_read_unlock_bh();
1668 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1671 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1672 ret = __ipv6_regen_rndid(idev);
1682 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1683 unsigned long expires, u32 flags)
1685 struct fib6_config cfg = {
1686 .fc_table = RT6_TABLE_PREFIX,
1687 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1688 .fc_ifindex = dev->ifindex,
1689 .fc_expires = expires,
1691 .fc_flags = RTF_UP | flags,
1692 .fc_nlinfo.nl_net = dev_net(dev),
1693 .fc_protocol = RTPROT_KERNEL,
1698 /* Prevent useless cloning on PtP SIT.
1699 This thing is done here expecting that the whole
1700 class of non-broadcast devices need not cloning.
1702 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1703 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1704 cfg.fc_flags |= RTF_NONEXTHOP;
1707 ip6_route_add(&cfg);
1711 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1713 const struct net_device *dev,
1714 u32 flags, u32 noflags)
1716 struct fib6_node *fn;
1717 struct rt6_info *rt = NULL;
1718 struct fib6_table *table;
1720 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1724 write_lock_bh(&table->tb6_lock);
1725 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1728 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1729 if (rt->dst.dev->ifindex != dev->ifindex)
1731 if ((rt->rt6i_flags & flags) != flags)
1733 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1739 write_unlock_bh(&table->tb6_lock);
1744 /* Create "default" multicast route to the interface */
1746 static void addrconf_add_mroute(struct net_device *dev)
1748 struct fib6_config cfg = {
1749 .fc_table = RT6_TABLE_LOCAL,
1750 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1751 .fc_ifindex = dev->ifindex,
1754 .fc_nlinfo.nl_net = dev_net(dev),
1757 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1759 ip6_route_add(&cfg);
1762 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1763 static void sit_route_add(struct net_device *dev)
1765 struct fib6_config cfg = {
1766 .fc_table = RT6_TABLE_MAIN,
1767 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1768 .fc_ifindex = dev->ifindex,
1770 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1771 .fc_nlinfo.nl_net = dev_net(dev),
1774 /* prefix length - 96 bits "::d.d.d.d" */
1775 ip6_route_add(&cfg);
1779 static void addrconf_add_lroute(struct net_device *dev)
1781 struct in6_addr addr;
1783 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1784 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1787 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1789 struct inet6_dev *idev;
1793 idev = ipv6_find_idev(dev);
1795 return ERR_PTR(-ENOBUFS);
1797 if (idev->cnf.disable_ipv6)
1798 return ERR_PTR(-EACCES);
1800 /* Add default multicast route */
1801 if (!(dev->flags & IFF_LOOPBACK))
1802 addrconf_add_mroute(dev);
1804 /* Add link local route */
1805 addrconf_add_lroute(dev);
1809 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1811 struct prefix_info *pinfo;
1815 struct inet6_dev *in6_dev;
1816 struct net *net = dev_net(dev);
1818 pinfo = (struct prefix_info *) opt;
1820 if (len < sizeof(struct prefix_info)) {
1821 ADBG(("addrconf: prefix option too short\n"));
1826 * Validation checks ([ADDRCONF], page 19)
1829 addr_type = ipv6_addr_type(&pinfo->prefix);
1831 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1834 valid_lft = ntohl(pinfo->valid);
1835 prefered_lft = ntohl(pinfo->prefered);
1837 if (prefered_lft > valid_lft) {
1838 if (net_ratelimit())
1839 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1843 in6_dev = in6_dev_get(dev);
1845 if (in6_dev == NULL) {
1846 if (net_ratelimit())
1847 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1852 * Two things going on here:
1853 * 1) Add routes for on-link prefixes
1854 * 2) Configure prefixes with the auto flag set
1857 if (pinfo->onlink) {
1858 struct rt6_info *rt;
1859 unsigned long rt_expires;
1861 /* Avoid arithmetic overflow. Really, we could
1862 * save rt_expires in seconds, likely valid_lft,
1863 * but it would require division in fib gc, that it
1867 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1869 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1871 if (addrconf_finite_timeout(rt_expires))
1874 rt = addrconf_get_prefix_route(&pinfo->prefix,
1877 RTF_ADDRCONF | RTF_PREFIX_RT,
1878 RTF_GATEWAY | RTF_DEFAULT);
1881 /* Autoconf prefix route */
1882 if (valid_lft == 0) {
1885 } else if (addrconf_finite_timeout(rt_expires)) {
1887 rt6_set_expires(rt, jiffies + rt_expires);
1889 rt6_clean_expires(rt);
1891 } else if (valid_lft) {
1892 clock_t expires = 0;
1893 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1894 if (addrconf_finite_timeout(rt_expires)) {
1896 flags |= RTF_EXPIRES;
1897 expires = jiffies_to_clock_t(rt_expires);
1899 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1900 dev, expires, flags);
1903 dst_release(&rt->dst);
1906 /* Try to figure out our local address for this prefix */
1908 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1909 struct inet6_ifaddr * ifp;
1910 struct in6_addr addr;
1911 int create = 0, update_lft = 0;
1913 if (pinfo->prefix_len == 64) {
1914 memcpy(&addr, &pinfo->prefix, 8);
1915 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1916 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1917 in6_dev_put(in6_dev);
1922 if (net_ratelimit())
1923 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1925 in6_dev_put(in6_dev);
1930 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1932 if (ifp == NULL && valid_lft) {
1933 int max_addresses = in6_dev->cnf.max_addresses;
1936 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1937 if (in6_dev->cnf.optimistic_dad &&
1938 !net->ipv6.devconf_all->forwarding && sllao)
1939 addr_flags = IFA_F_OPTIMISTIC;
1942 /* Do not allow to create too much of autoconfigured
1943 * addresses; this would be too easy way to crash kernel.
1945 if (!max_addresses ||
1946 ipv6_count_addresses(in6_dev) < max_addresses)
1947 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1948 addr_type&IPV6_ADDR_SCOPE_MASK,
1951 if (!ifp || IS_ERR(ifp)) {
1952 in6_dev_put(in6_dev);
1956 update_lft = create = 1;
1957 ifp->cstamp = jiffies;
1958 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1964 #ifdef CONFIG_IPV6_PRIVACY
1965 struct inet6_ifaddr *ift;
1969 /* update lifetime (RFC2462 5.5.3 e) */
1970 spin_lock(&ifp->lock);
1972 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1973 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1976 if (!update_lft && stored_lft) {
1977 if (valid_lft > MIN_VALID_LIFETIME ||
1978 valid_lft > stored_lft)
1980 else if (stored_lft <= MIN_VALID_LIFETIME) {
1981 /* valid_lft <= stored_lft is always true */
1983 * RFC 4862 Section 5.5.3e:
1984 * "Note that the preferred lifetime of
1985 * the corresponding address is always
1986 * reset to the Preferred Lifetime in
1987 * the received Prefix Information
1988 * option, regardless of whether the
1989 * valid lifetime is also reset or
1992 * So if the preferred lifetime in
1993 * this advertisement is different
1994 * than what we have stored, but the
1995 * valid lifetime is invalid, just
1996 * reset prefered_lft.
1998 * We must set the valid lifetime
1999 * to the stored lifetime since we'll
2000 * be updating the timestamp below,
2001 * else we'll set it back to the
2004 if (prefered_lft != ifp->prefered_lft) {
2005 valid_lft = stored_lft;
2009 valid_lft = MIN_VALID_LIFETIME;
2010 if (valid_lft < prefered_lft)
2011 prefered_lft = valid_lft;
2017 ifp->valid_lft = valid_lft;
2018 ifp->prefered_lft = prefered_lft;
2021 ifp->flags &= ~IFA_F_DEPRECATED;
2022 spin_unlock(&ifp->lock);
2024 if (!(flags&IFA_F_TENTATIVE))
2025 ipv6_ifa_notify(0, ifp);
2027 spin_unlock(&ifp->lock);
2029 #ifdef CONFIG_IPV6_PRIVACY
2030 read_lock_bh(&in6_dev->lock);
2031 /* update all temporary addresses in the list */
2032 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2034 int age, max_valid, max_prefered;
2036 if (ifp != ift->ifpub)
2040 * RFC 4941 section 3.3:
2041 * If a received option will extend the lifetime
2042 * of a public address, the lifetimes of
2043 * temporary addresses should be extended,
2044 * subject to the overall constraint that no
2045 * temporary addresses should ever remain
2046 * "valid" or "preferred" for a time longer than
2047 * (TEMP_VALID_LIFETIME) or
2048 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2051 age = (now - ift->cstamp) / HZ;
2052 max_valid = in6_dev->cnf.temp_valid_lft - age;
2056 max_prefered = in6_dev->cnf.temp_prefered_lft -
2057 in6_dev->cnf.max_desync_factor -
2059 if (max_prefered < 0)
2062 if (valid_lft > max_valid)
2063 valid_lft = max_valid;
2065 if (prefered_lft > max_prefered)
2066 prefered_lft = max_prefered;
2068 spin_lock(&ift->lock);
2070 ift->valid_lft = valid_lft;
2071 ift->prefered_lft = prefered_lft;
2073 if (prefered_lft > 0)
2074 ift->flags &= ~IFA_F_DEPRECATED;
2076 spin_unlock(&ift->lock);
2077 if (!(flags&IFA_F_TENTATIVE))
2078 ipv6_ifa_notify(0, ift);
2081 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2083 * When a new public address is created as
2084 * described in [ADDRCONF], also create a new
2085 * temporary address. Also create a temporary
2086 * address if it's enabled but no temporary
2087 * address currently exists.
2089 read_unlock_bh(&in6_dev->lock);
2090 ipv6_create_tempaddr(ifp, NULL);
2092 read_unlock_bh(&in6_dev->lock);
2099 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2100 in6_dev_put(in6_dev);
2104 * Set destination address.
2105 * Special case for SIT interfaces where we create a new "virtual"
2108 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2110 struct in6_ifreq ireq;
2111 struct net_device *dev;
2117 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2120 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2126 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2127 if (dev->type == ARPHRD_SIT) {
2128 const struct net_device_ops *ops = dev->netdev_ops;
2130 struct ip_tunnel_parm p;
2132 err = -EADDRNOTAVAIL;
2133 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2136 memset(&p, 0, sizeof(p));
2137 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2141 p.iph.protocol = IPPROTO_IPV6;
2143 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2145 if (ops->ndo_do_ioctl) {
2146 mm_segment_t oldfs = get_fs();
2149 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2156 dev = __dev_get_by_name(net, p.name);
2159 err = dev_open(dev);
2170 * Manual configuration of address on an interface
2172 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2173 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2176 struct inet6_ifaddr *ifp;
2177 struct inet6_dev *idev;
2178 struct net_device *dev;
2182 unsigned long timeout;
2189 /* check the lifetime */
2190 if (!valid_lft || prefered_lft > valid_lft)
2193 dev = __dev_get_by_index(net, ifindex);
2197 idev = addrconf_add_dev(dev);
2199 return PTR_ERR(idev);
2201 scope = ipv6_addr_scope(pfx);
2203 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2204 if (addrconf_finite_timeout(timeout)) {
2205 expires = jiffies_to_clock_t(timeout * HZ);
2206 valid_lft = timeout;
2207 flags = RTF_EXPIRES;
2211 ifa_flags |= IFA_F_PERMANENT;
2214 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2215 if (addrconf_finite_timeout(timeout)) {
2217 ifa_flags |= IFA_F_DEPRECATED;
2218 prefered_lft = timeout;
2221 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2224 spin_lock_bh(&ifp->lock);
2225 ifp->valid_lft = valid_lft;
2226 ifp->prefered_lft = prefered_lft;
2227 ifp->tstamp = jiffies;
2228 spin_unlock_bh(&ifp->lock);
2230 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2233 * Note that section 3.1 of RFC 4429 indicates
2234 * that the Optimistic flag should not be set for
2235 * manually configured addresses
2237 addrconf_dad_start(ifp, 0);
2243 return PTR_ERR(ifp);
2246 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2249 struct inet6_ifaddr *ifp;
2250 struct inet6_dev *idev;
2251 struct net_device *dev;
2256 dev = __dev_get_by_index(net, ifindex);
2260 if ((idev = __in6_dev_get(dev)) == NULL)
2263 read_lock_bh(&idev->lock);
2264 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2265 if (ifp->prefix_len == plen &&
2266 ipv6_addr_equal(pfx, &ifp->addr)) {
2268 read_unlock_bh(&idev->lock);
2272 /* If the last address is deleted administratively,
2273 disable IPv6 on this interface.
2275 if (list_empty(&idev->addr_list))
2276 addrconf_ifdown(idev->dev, 1);
2280 read_unlock_bh(&idev->lock);
2281 return -EADDRNOTAVAIL;
2285 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2287 struct in6_ifreq ireq;
2290 if (!capable(CAP_NET_ADMIN))
2293 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2297 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2298 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2299 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2304 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2306 struct in6_ifreq ireq;
2309 if (!capable(CAP_NET_ADMIN))
2312 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2316 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2317 ireq.ifr6_prefixlen);
2322 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2323 int plen, int scope)
2325 struct inet6_ifaddr *ifp;
2327 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2329 spin_lock_bh(&ifp->lock);
2330 ifp->flags &= ~IFA_F_TENTATIVE;
2331 spin_unlock_bh(&ifp->lock);
2332 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2337 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2338 static void sit_add_v4_addrs(struct inet6_dev *idev)
2340 struct in6_addr addr;
2341 struct net_device *dev;
2342 struct net *net = dev_net(idev->dev);
2347 memset(&addr, 0, sizeof(struct in6_addr));
2348 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2350 if (idev->dev->flags&IFF_POINTOPOINT) {
2351 addr.s6_addr32[0] = htonl(0xfe800000);
2354 scope = IPV6_ADDR_COMPATv4;
2357 if (addr.s6_addr32[3]) {
2358 add_addr(idev, &addr, 128, scope);
2362 for_each_netdev(net, dev) {
2363 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2364 if (in_dev && (dev->flags & IFF_UP)) {
2365 struct in_ifaddr * ifa;
2369 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2372 addr.s6_addr32[3] = ifa->ifa_local;
2374 if (ifa->ifa_scope == RT_SCOPE_LINK)
2376 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2377 if (idev->dev->flags&IFF_POINTOPOINT)
2381 if (idev->dev->flags&IFF_POINTOPOINT)
2386 add_addr(idev, &addr, plen, flag);
2393 static void init_loopback(struct net_device *dev)
2395 struct inet6_dev *idev;
2401 if ((idev = ipv6_find_idev(dev)) == NULL) {
2402 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2406 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2409 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2411 struct inet6_ifaddr * ifp;
2412 u32 addr_flags = IFA_F_PERMANENT;
2414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2415 if (idev->cnf.optimistic_dad &&
2416 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2417 addr_flags |= IFA_F_OPTIMISTIC;
2421 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2423 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2424 addrconf_dad_start(ifp, 0);
2429 static void addrconf_dev_config(struct net_device *dev)
2431 struct in6_addr addr;
2432 struct inet6_dev * idev;
2436 if ((dev->type != ARPHRD_ETHER) &&
2437 (dev->type != ARPHRD_FDDI) &&
2438 (dev->type != ARPHRD_IEEE802_TR) &&
2439 (dev->type != ARPHRD_ARCNET) &&
2440 (dev->type != ARPHRD_INFINIBAND)) {
2441 /* Alas, we support only Ethernet autoconfiguration. */
2445 idev = addrconf_add_dev(dev);
2449 memset(&addr, 0, sizeof(struct in6_addr));
2450 addr.s6_addr32[0] = htonl(0xFE800000);
2452 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2453 addrconf_add_linklocal(idev, &addr);
2456 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2457 static void addrconf_sit_config(struct net_device *dev)
2459 struct inet6_dev *idev;
2464 * Configure the tunnel with one of our IPv4
2465 * addresses... we should configure all of
2466 * our v4 addrs in the tunnel
2469 if ((idev = ipv6_find_idev(dev)) == NULL) {
2470 printk(KERN_DEBUG "init sit: add_dev failed\n");
2474 if (dev->priv_flags & IFF_ISATAP) {
2475 struct in6_addr addr;
2477 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2478 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2479 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2480 addrconf_add_linklocal(idev, &addr);
2484 sit_add_v4_addrs(idev);
2486 if (dev->flags&IFF_POINTOPOINT) {
2487 addrconf_add_mroute(dev);
2488 addrconf_add_lroute(dev);
2494 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2495 static void addrconf_gre_config(struct net_device *dev)
2497 struct inet6_dev *idev;
2498 struct in6_addr addr;
2500 pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2504 if ((idev = ipv6_find_idev(dev)) == NULL) {
2505 printk(KERN_DEBUG "init gre: add_dev failed\n");
2509 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2510 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2512 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2513 addrconf_add_linklocal(idev, &addr);
2518 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2520 struct in6_addr lladdr;
2522 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2523 addrconf_add_linklocal(idev, &lladdr);
2529 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2531 struct net_device *link_dev;
2532 struct net *net = dev_net(idev->dev);
2534 /* first try to inherit the link-local address from the link device */
2535 if (idev->dev->iflink &&
2536 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2537 if (!ipv6_inherit_linklocal(idev, link_dev))
2540 /* then try to inherit it from any device */
2541 for_each_netdev(net, link_dev) {
2542 if (!ipv6_inherit_linklocal(idev, link_dev))
2545 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2549 * Autoconfigure tunnel with a link-local address so routing protocols,
2550 * DHCPv6, MLD etc. can be run over the virtual link
2553 static void addrconf_ip6_tnl_config(struct net_device *dev)
2555 struct inet6_dev *idev;
2559 idev = addrconf_add_dev(dev);
2561 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2564 ip6_tnl_add_linklocal(idev);
2567 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2570 struct net_device *dev = (struct net_device *) data;
2571 struct inet6_dev *idev = __in6_dev_get(dev);
2572 int run_pending = 0;
2576 case NETDEV_REGISTER:
2577 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2578 idev = ipv6_add_dev(dev);
2580 return notifier_from_errno(-ENOMEM);
2586 if (dev->flags & IFF_SLAVE)
2589 if (event == NETDEV_UP) {
2590 if (!addrconf_qdisc_ok(dev)) {
2591 /* device is not ready yet. */
2593 "ADDRCONF(NETDEV_UP): %s: "
2594 "link is not ready\n",
2599 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2600 idev = ipv6_add_dev(dev);
2603 idev->if_flags |= IF_READY;
2607 if (!addrconf_qdisc_ok(dev)) {
2608 /* device is still not ready. */
2613 if (idev->if_flags & IF_READY)
2614 /* device is already configured. */
2616 idev->if_flags |= IF_READY;
2620 "ADDRCONF(NETDEV_CHANGE): %s: "
2621 "link becomes ready\n",
2627 switch (dev->type) {
2628 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2630 addrconf_sit_config(dev);
2633 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2635 addrconf_gre_config(dev);
2638 case ARPHRD_TUNNEL6:
2639 addrconf_ip6_tnl_config(dev);
2641 case ARPHRD_LOOPBACK:
2646 addrconf_dev_config(dev);
2652 addrconf_dad_run(idev);
2655 * If the MTU changed during the interface down,
2656 * when the interface up, the changed MTU must be
2657 * reflected in the idev as well as routers.
2659 if (idev->cnf.mtu6 != dev->mtu &&
2660 dev->mtu >= IPV6_MIN_MTU) {
2661 rt6_mtu_change(dev, dev->mtu);
2662 idev->cnf.mtu6 = dev->mtu;
2664 idev->tstamp = jiffies;
2665 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2668 * If the changed mtu during down is lower than
2669 * IPV6_MIN_MTU stop IPv6 on this interface.
2671 if (dev->mtu < IPV6_MIN_MTU)
2672 addrconf_ifdown(dev, 1);
2676 case NETDEV_CHANGEMTU:
2677 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2678 rt6_mtu_change(dev, dev->mtu);
2679 idev->cnf.mtu6 = dev->mtu;
2683 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2684 idev = ipv6_add_dev(dev);
2690 * MTU falled under IPV6_MIN_MTU.
2691 * Stop IPv6 on this interface.
2695 case NETDEV_UNREGISTER:
2697 * Remove all addresses from this interface.
2699 addrconf_ifdown(dev, event != NETDEV_DOWN);
2702 case NETDEV_CHANGENAME:
2704 snmp6_unregister_dev(idev);
2705 addrconf_sysctl_unregister(idev);
2706 addrconf_sysctl_register(idev);
2707 err = snmp6_register_dev(idev);
2709 return notifier_from_errno(err);
2713 case NETDEV_PRE_TYPE_CHANGE:
2714 case NETDEV_POST_TYPE_CHANGE:
2715 addrconf_type_change(dev, event);
2723 * addrconf module should be notified of a device going up
2725 static struct notifier_block ipv6_dev_notf = {
2726 .notifier_call = addrconf_notify,
2729 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2731 struct inet6_dev *idev;
2734 idev = __in6_dev_get(dev);
2736 if (event == NETDEV_POST_TYPE_CHANGE)
2737 ipv6_mc_remap(idev);
2738 else if (event == NETDEV_PRE_TYPE_CHANGE)
2739 ipv6_mc_unmap(idev);
2742 static int addrconf_ifdown(struct net_device *dev, int how)
2744 struct net *net = dev_net(dev);
2745 struct inet6_dev *idev;
2746 struct inet6_ifaddr *ifa;
2751 rt6_ifdown(net, dev);
2752 neigh_ifdown(&nd_tbl, dev);
2754 idev = __in6_dev_get(dev);
2759 * Step 1: remove reference to ipv6 device from parent device.
2765 /* protected by rtnl_lock */
2766 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2768 /* Step 1.5: remove snmp6 entry */
2769 snmp6_unregister_dev(idev);
2773 /* Step 2: clear hash table */
2774 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2775 struct hlist_head *h = &inet6_addr_lst[i];
2776 struct hlist_node *n;
2778 spin_lock_bh(&addrconf_hash_lock);
2780 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2781 if (ifa->idev == idev) {
2782 hlist_del_init_rcu(&ifa->addr_lst);
2783 addrconf_del_timer(ifa);
2787 spin_unlock_bh(&addrconf_hash_lock);
2790 write_lock_bh(&idev->lock);
2792 /* Step 2: clear flags for stateless addrconf */
2794 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2796 #ifdef CONFIG_IPV6_PRIVACY
2797 if (how && del_timer(&idev->regen_timer))
2800 /* Step 3: clear tempaddr list */
2801 while (!list_empty(&idev->tempaddr_list)) {
2802 ifa = list_first_entry(&idev->tempaddr_list,
2803 struct inet6_ifaddr, tmp_list);
2804 list_del(&ifa->tmp_list);
2805 write_unlock_bh(&idev->lock);
2806 spin_lock_bh(&ifa->lock);
2809 in6_ifa_put(ifa->ifpub);
2812 spin_unlock_bh(&ifa->lock);
2814 write_lock_bh(&idev->lock);
2818 while (!list_empty(&idev->addr_list)) {
2819 ifa = list_first_entry(&idev->addr_list,
2820 struct inet6_ifaddr, if_list);
2821 addrconf_del_timer(ifa);
2823 list_del(&ifa->if_list);
2825 write_unlock_bh(&idev->lock);
2827 spin_lock_bh(&ifa->state_lock);
2829 ifa->state = INET6_IFADDR_STATE_DEAD;
2830 spin_unlock_bh(&ifa->state_lock);
2832 if (state != INET6_IFADDR_STATE_DEAD) {
2833 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2834 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2838 write_lock_bh(&idev->lock);
2841 write_unlock_bh(&idev->lock);
2843 /* Step 5: Discard multicast list */
2845 ipv6_mc_destroy_dev(idev);
2849 idev->tstamp = jiffies;
2851 /* Last: Shot the device (if unregistered) */
2853 addrconf_sysctl_unregister(idev);
2854 neigh_parms_release(&nd_tbl, idev->nd_parms);
2855 neigh_ifdown(&nd_tbl, dev);
2861 static void addrconf_rs_timer(unsigned long data)
2863 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2864 struct inet6_dev *idev = ifp->idev;
2866 read_lock(&idev->lock);
2867 if (idev->dead || !(idev->if_flags & IF_READY))
2870 if (idev->cnf.forwarding)
2873 /* Announcement received after solicitation was sent */
2874 if (idev->if_flags & IF_RA_RCVD)
2877 spin_lock(&ifp->lock);
2878 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2879 /* The wait after the last probe can be shorter */
2880 addrconf_mod_timer(ifp, AC_RS,
2881 (ifp->probes == idev->cnf.rtr_solicits) ?
2882 idev->cnf.rtr_solicit_delay :
2883 idev->cnf.rtr_solicit_interval);
2884 spin_unlock(&ifp->lock);
2886 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2888 spin_unlock(&ifp->lock);
2890 * Note: we do not support deprecated "all on-link"
2891 * assumption any longer.
2893 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2898 read_unlock(&idev->lock);
2903 * Duplicate Address Detection
2905 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2907 unsigned long rand_num;
2908 struct inet6_dev *idev = ifp->idev;
2910 if (ifp->flags & IFA_F_OPTIMISTIC)
2913 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2915 ifp->probes = idev->cnf.dad_transmits;
2916 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2919 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2921 struct inet6_dev *idev = ifp->idev;
2922 struct net_device *dev = idev->dev;
2924 addrconf_join_solict(dev, &ifp->addr);
2926 net_srandom(ifp->addr.s6_addr32[3]);
2928 read_lock_bh(&idev->lock);
2929 spin_lock(&ifp->lock);
2930 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2933 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2934 idev->cnf.accept_dad < 1 ||
2935 !(ifp->flags&IFA_F_TENTATIVE) ||
2936 ifp->flags & IFA_F_NODAD) {
2937 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2938 spin_unlock(&ifp->lock);
2939 read_unlock_bh(&idev->lock);
2941 addrconf_dad_completed(ifp);
2945 if (!(idev->if_flags & IF_READY)) {
2946 spin_unlock(&ifp->lock);
2947 read_unlock_bh(&idev->lock);
2949 * If the device is not ready:
2950 * - keep it tentative if it is a permanent address.
2951 * - otherwise, kill it.
2954 addrconf_dad_stop(ifp, 0);
2959 * Optimistic nodes can start receiving
2962 if (ifp->flags & IFA_F_OPTIMISTIC)
2963 ip6_ins_rt(ifp->rt);
2965 addrconf_dad_kick(ifp);
2967 spin_unlock(&ifp->lock);
2968 read_unlock_bh(&idev->lock);
2971 static void addrconf_dad_timer(unsigned long data)
2973 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2974 struct inet6_dev *idev = ifp->idev;
2975 struct in6_addr mcaddr;
2977 if (!ifp->probes && addrconf_dad_end(ifp))
2980 read_lock(&idev->lock);
2981 if (idev->dead || !(idev->if_flags & IF_READY)) {
2982 read_unlock(&idev->lock);
2986 spin_lock(&ifp->lock);
2987 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2988 spin_unlock(&ifp->lock);
2989 read_unlock(&idev->lock);
2993 if (ifp->probes == 0) {
2995 * DAD was successful
2998 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2999 spin_unlock(&ifp->lock);
3000 read_unlock(&idev->lock);
3002 addrconf_dad_completed(ifp);
3008 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3009 spin_unlock(&ifp->lock);
3010 read_unlock(&idev->lock);
3012 /* send a neighbour solicitation for our addr */
3013 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3014 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3019 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3021 struct net_device *dev = ifp->idev->dev;
3024 * Configure the address for reception. Now it is valid.
3027 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3029 /* If added prefix is link local and we are prepared to process
3030 router advertisements, start sending router solicitations.
3033 if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3034 ifp->idev->cnf.accept_ra == 2) &&
3035 ifp->idev->cnf.rtr_solicits > 0 &&
3036 (dev->flags&IFF_LOOPBACK) == 0 &&
3037 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3039 * If a host as already performed a random delay
3040 * [...] as part of DAD [...] there is no need
3041 * to delay again before sending the first RS
3043 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3045 spin_lock_bh(&ifp->lock);
3047 ifp->idev->if_flags |= IF_RS_SENT;
3048 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3049 spin_unlock_bh(&ifp->lock);
3053 static void addrconf_dad_run(struct inet6_dev *idev)
3055 struct inet6_ifaddr *ifp;
3057 read_lock_bh(&idev->lock);
3058 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3059 spin_lock(&ifp->lock);
3060 if (ifp->flags & IFA_F_TENTATIVE &&
3061 ifp->state == INET6_IFADDR_STATE_DAD)
3062 addrconf_dad_kick(ifp);
3063 spin_unlock(&ifp->lock);
3065 read_unlock_bh(&idev->lock);
3068 #ifdef CONFIG_PROC_FS
3069 struct if6_iter_state {
3070 struct seq_net_private p;
3075 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3077 struct inet6_ifaddr *ifa = NULL;
3078 struct if6_iter_state *state = seq->private;
3079 struct net *net = seq_file_net(seq);
3082 /* initial bucket if pos is 0 */
3088 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3089 struct hlist_node *n;
3090 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3092 /* sync with offset */
3093 if (p < state->offset) {
3098 if (net_eq(dev_net(ifa->idev->dev), net))
3102 /* prepare for next bucket */
3109 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3110 struct inet6_ifaddr *ifa)
3112 struct if6_iter_state *state = seq->private;
3113 struct net *net = seq_file_net(seq);
3114 struct hlist_node *n = &ifa->addr_lst;
3116 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3118 if (net_eq(dev_net(ifa->idev->dev), net))
3122 while (++state->bucket < IN6_ADDR_HSIZE) {
3124 hlist_for_each_entry_rcu_bh(ifa, n,
3125 &inet6_addr_lst[state->bucket], addr_lst) {
3127 if (net_eq(dev_net(ifa->idev->dev), net))
3135 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3139 return if6_get_first(seq, *pos);
3142 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3144 struct inet6_ifaddr *ifa;
3146 ifa = if6_get_next(seq, v);
3151 static void if6_seq_stop(struct seq_file *seq, void *v)
3154 rcu_read_unlock_bh();
3157 static int if6_seq_show(struct seq_file *seq, void *v)
3159 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3160 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3162 ifp->idev->dev->ifindex,
3166 ifp->idev->dev->name);
3170 static const struct seq_operations if6_seq_ops = {
3171 .start = if6_seq_start,
3172 .next = if6_seq_next,
3173 .show = if6_seq_show,
3174 .stop = if6_seq_stop,
3177 static int if6_seq_open(struct inode *inode, struct file *file)
3179 return seq_open_net(inode, file, &if6_seq_ops,
3180 sizeof(struct if6_iter_state));
3183 static const struct file_operations if6_fops = {
3184 .owner = THIS_MODULE,
3185 .open = if6_seq_open,
3187 .llseek = seq_lseek,
3188 .release = seq_release_net,
3191 static int __net_init if6_proc_net_init(struct net *net)
3193 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3198 static void __net_exit if6_proc_net_exit(struct net *net)
3200 proc_net_remove(net, "if_inet6");
3203 static struct pernet_operations if6_proc_net_ops = {
3204 .init = if6_proc_net_init,
3205 .exit = if6_proc_net_exit,
3208 int __init if6_proc_init(void)
3210 return register_pernet_subsys(&if6_proc_net_ops);
3213 void if6_proc_exit(void)
3215 unregister_pernet_subsys(&if6_proc_net_ops);
3217 #endif /* CONFIG_PROC_FS */
3219 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3220 /* Check if address is a home address configured on any interface. */
3221 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3224 struct inet6_ifaddr *ifp = NULL;
3225 struct hlist_node *n;
3226 unsigned int hash = ipv6_addr_hash(addr);
3229 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3230 if (!net_eq(dev_net(ifp->idev->dev), net))
3232 if (ipv6_addr_equal(&ifp->addr, addr) &&
3233 (ifp->flags & IFA_F_HOMEADDRESS)) {
3238 rcu_read_unlock_bh();
3244 * Periodic address status verification
3247 static void addrconf_verify(unsigned long foo)
3249 unsigned long now, next, next_sec, next_sched;
3250 struct inet6_ifaddr *ifp;
3251 struct hlist_node *node;
3255 spin_lock(&addrconf_verify_lock);
3257 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3259 del_timer(&addr_chk_timer);
3261 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3263 hlist_for_each_entry_rcu_bh(ifp, node,
3264 &inet6_addr_lst[i], addr_lst) {
3267 if (ifp->flags & IFA_F_PERMANENT)
3270 spin_lock(&ifp->lock);
3271 /* We try to batch several events at once. */
3272 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3274 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3275 age >= ifp->valid_lft) {
3276 spin_unlock(&ifp->lock);
3280 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3281 spin_unlock(&ifp->lock);
3283 } else if (age >= ifp->prefered_lft) {
3284 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3287 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3289 ifp->flags |= IFA_F_DEPRECATED;
3292 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3293 next = ifp->tstamp + ifp->valid_lft * HZ;
3295 spin_unlock(&ifp->lock);
3300 ipv6_ifa_notify(0, ifp);
3304 #ifdef CONFIG_IPV6_PRIVACY
3305 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3306 !(ifp->flags&IFA_F_TENTATIVE)) {
3307 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3308 ifp->idev->cnf.dad_transmits *
3309 ifp->idev->nd_parms->retrans_time / HZ;
3311 if (age >= ifp->prefered_lft - regen_advance) {
3312 struct inet6_ifaddr *ifpub = ifp->ifpub;
3313 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3314 next = ifp->tstamp + ifp->prefered_lft * HZ;
3315 if (!ifp->regen_count && ifpub) {
3318 in6_ifa_hold(ifpub);
3319 spin_unlock(&ifp->lock);
3321 spin_lock(&ifpub->lock);
3322 ifpub->regen_count = 0;
3323 spin_unlock(&ifpub->lock);
3324 ipv6_create_tempaddr(ifpub, ifp);
3329 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3330 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3331 spin_unlock(&ifp->lock);
3334 /* ifp->prefered_lft <= ifp->valid_lft */
3335 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3336 next = ifp->tstamp + ifp->prefered_lft * HZ;
3337 spin_unlock(&ifp->lock);
3342 next_sec = round_jiffies_up(next);
3345 /* If rounded timeout is accurate enough, accept it. */
3346 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3347 next_sched = next_sec;
3349 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3350 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3351 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3353 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3354 now, next, next_sec, next_sched));
3356 addr_chk_timer.expires = next_sched;
3357 add_timer(&addr_chk_timer);
3358 spin_unlock(&addrconf_verify_lock);
3359 rcu_read_unlock_bh();
3362 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3364 struct in6_addr *pfx = NULL;
3367 pfx = nla_data(addr);
3370 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3373 pfx = nla_data(local);
3379 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3380 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3381 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3382 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3386 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3388 struct net *net = sock_net(skb->sk);
3389 struct ifaddrmsg *ifm;
3390 struct nlattr *tb[IFA_MAX+1];
3391 struct in6_addr *pfx;
3394 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3398 ifm = nlmsg_data(nlh);
3399 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3403 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3406 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3407 u32 prefered_lft, u32 valid_lft)
3411 unsigned long timeout;
3413 if (!valid_lft || (prefered_lft > valid_lft))
3416 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3417 if (addrconf_finite_timeout(timeout)) {
3418 expires = jiffies_to_clock_t(timeout * HZ);
3419 valid_lft = timeout;
3420 flags = RTF_EXPIRES;
3424 ifa_flags |= IFA_F_PERMANENT;
3427 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3428 if (addrconf_finite_timeout(timeout)) {
3430 ifa_flags |= IFA_F_DEPRECATED;
3431 prefered_lft = timeout;
3434 spin_lock_bh(&ifp->lock);
3435 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3436 ifp->tstamp = jiffies;
3437 ifp->valid_lft = valid_lft;
3438 ifp->prefered_lft = prefered_lft;
3440 spin_unlock_bh(&ifp->lock);
3441 if (!(ifp->flags&IFA_F_TENTATIVE))
3442 ipv6_ifa_notify(0, ifp);
3444 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3452 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3454 struct net *net = sock_net(skb->sk);
3455 struct ifaddrmsg *ifm;
3456 struct nlattr *tb[IFA_MAX+1];
3457 struct in6_addr *pfx;
3458 struct inet6_ifaddr *ifa;
3459 struct net_device *dev;
3460 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3464 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3468 ifm = nlmsg_data(nlh);
3469 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3473 if (tb[IFA_CACHEINFO]) {
3474 struct ifa_cacheinfo *ci;
3476 ci = nla_data(tb[IFA_CACHEINFO]);
3477 valid_lft = ci->ifa_valid;
3478 preferred_lft = ci->ifa_prefered;
3480 preferred_lft = INFINITY_LIFE_TIME;
3481 valid_lft = INFINITY_LIFE_TIME;
3484 dev = __dev_get_by_index(net, ifm->ifa_index);
3488 /* We ignore other flags so far. */
3489 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3491 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3494 * It would be best to check for !NLM_F_CREATE here but
3495 * userspace alreay relies on not having to provide this.
3497 return inet6_addr_add(net, ifm->ifa_index, pfx,
3498 ifm->ifa_prefixlen, ifa_flags,
3499 preferred_lft, valid_lft);
3502 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3503 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3506 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3513 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3514 u8 scope, int ifindex)
3516 struct ifaddrmsg *ifm;
3518 ifm = nlmsg_data(nlh);
3519 ifm->ifa_family = AF_INET6;
3520 ifm->ifa_prefixlen = prefixlen;
3521 ifm->ifa_flags = flags;
3522 ifm->ifa_scope = scope;
3523 ifm->ifa_index = ifindex;
3526 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3527 unsigned long tstamp, u32 preferred, u32 valid)
3529 struct ifa_cacheinfo ci;
3531 ci.cstamp = cstamp_delta(cstamp);
3532 ci.tstamp = cstamp_delta(tstamp);
3533 ci.ifa_prefered = preferred;
3534 ci.ifa_valid = valid;
3536 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3539 static inline int rt_scope(int ifa_scope)
3541 if (ifa_scope & IFA_HOST)
3542 return RT_SCOPE_HOST;
3543 else if (ifa_scope & IFA_LINK)
3544 return RT_SCOPE_LINK;
3545 else if (ifa_scope & IFA_SITE)
3546 return RT_SCOPE_SITE;
3548 return RT_SCOPE_UNIVERSE;
3551 static inline int inet6_ifaddr_msgsize(void)
3553 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3554 + nla_total_size(16) /* IFA_ADDRESS */
3555 + nla_total_size(sizeof(struct ifa_cacheinfo));
3558 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3559 u32 pid, u32 seq, int event, unsigned int flags)
3561 struct nlmsghdr *nlh;
3562 u32 preferred, valid;
3564 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3568 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3569 ifa->idev->dev->ifindex);
3571 if (!(ifa->flags&IFA_F_PERMANENT)) {
3572 preferred = ifa->prefered_lft;
3573 valid = ifa->valid_lft;
3574 if (preferred != INFINITY_LIFE_TIME) {
3575 long tval = (jiffies - ifa->tstamp)/HZ;
3576 if (preferred > tval)
3580 if (valid != INFINITY_LIFE_TIME) {
3588 preferred = INFINITY_LIFE_TIME;
3589 valid = INFINITY_LIFE_TIME;
3592 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3593 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3594 nlmsg_cancel(skb, nlh);
3598 return nlmsg_end(skb, nlh);
3601 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3602 u32 pid, u32 seq, int event, u16 flags)
3604 struct nlmsghdr *nlh;
3605 u8 scope = RT_SCOPE_UNIVERSE;
3606 int ifindex = ifmca->idev->dev->ifindex;
3608 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3609 scope = RT_SCOPE_SITE;
3611 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3615 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3616 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3617 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3618 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3619 nlmsg_cancel(skb, nlh);
3623 return nlmsg_end(skb, nlh);
3626 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3627 u32 pid, u32 seq, int event, unsigned int flags)
3629 struct nlmsghdr *nlh;
3630 u8 scope = RT_SCOPE_UNIVERSE;
3631 int ifindex = ifaca->aca_idev->dev->ifindex;
3633 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3634 scope = RT_SCOPE_SITE;
3636 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3640 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3641 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3642 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3643 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3644 nlmsg_cancel(skb, nlh);
3648 return nlmsg_end(skb, nlh);
3657 /* called with rcu_read_lock() */
3658 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3659 struct netlink_callback *cb, enum addr_type_t type,
3660 int s_ip_idx, int *p_ip_idx)
3662 struct ifmcaddr6 *ifmca;
3663 struct ifacaddr6 *ifaca;
3665 int ip_idx = *p_ip_idx;
3667 read_lock_bh(&idev->lock);
3669 case UNICAST_ADDR: {
3670 struct inet6_ifaddr *ifa;
3672 /* unicast address incl. temp addr */
3673 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3674 if (++ip_idx < s_ip_idx)
3676 err = inet6_fill_ifaddr(skb, ifa,
3677 NETLINK_CB(cb->skb).pid,
3686 case MULTICAST_ADDR:
3687 /* multicast address */
3688 for (ifmca = idev->mc_list; ifmca;
3689 ifmca = ifmca->next, ip_idx++) {
3690 if (ip_idx < s_ip_idx)
3692 err = inet6_fill_ifmcaddr(skb, ifmca,
3693 NETLINK_CB(cb->skb).pid,
3702 /* anycast address */
3703 for (ifaca = idev->ac_list; ifaca;
3704 ifaca = ifaca->aca_next, ip_idx++) {
3705 if (ip_idx < s_ip_idx)
3707 err = inet6_fill_ifacaddr(skb, ifaca,
3708 NETLINK_CB(cb->skb).pid,
3719 read_unlock_bh(&idev->lock);
3724 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3725 enum addr_type_t type)
3727 struct net *net = sock_net(skb->sk);
3730 int s_idx, s_ip_idx;
3731 struct net_device *dev;
3732 struct inet6_dev *idev;
3733 struct hlist_head *head;
3734 struct hlist_node *node;
3737 s_idx = idx = cb->args[1];
3738 s_ip_idx = ip_idx = cb->args[2];
3741 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3743 head = &net->dev_index_head[h];
3744 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3747 if (h > s_h || idx > s_idx)
3750 idev = __in6_dev_get(dev);
3754 if (in6_dump_addrs(idev, skb, cb, type,
3755 s_ip_idx, &ip_idx) <= 0)
3765 cb->args[2] = ip_idx;
3770 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3772 enum addr_type_t type = UNICAST_ADDR;
3774 return inet6_dump_addr(skb, cb, type);
3777 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3779 enum addr_type_t type = MULTICAST_ADDR;
3781 return inet6_dump_addr(skb, cb, type);
3785 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3787 enum addr_type_t type = ANYCAST_ADDR;
3789 return inet6_dump_addr(skb, cb, type);
3792 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3795 struct net *net = sock_net(in_skb->sk);
3796 struct ifaddrmsg *ifm;
3797 struct nlattr *tb[IFA_MAX+1];
3798 struct in6_addr *addr = NULL;
3799 struct net_device *dev = NULL;
3800 struct inet6_ifaddr *ifa;
3801 struct sk_buff *skb;
3804 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3808 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3814 ifm = nlmsg_data(nlh);
3816 dev = __dev_get_by_index(net, ifm->ifa_index);
3818 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3820 err = -EADDRNOTAVAIL;
3824 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3830 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3831 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3833 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3834 WARN_ON(err == -EMSGSIZE);
3838 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3845 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3847 struct sk_buff *skb;
3848 struct net *net = dev_net(ifa->idev->dev);
3851 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3855 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3857 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3858 WARN_ON(err == -EMSGSIZE);
3862 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3866 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3869 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3870 __s32 *array, int bytes)
3872 BUG_ON(bytes < (DEVCONF_MAX * 4));
3874 memset(array, 0, bytes);
3875 array[DEVCONF_FORWARDING] = cnf->forwarding;
3876 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3877 array[DEVCONF_MTU6] = cnf->mtu6;
3878 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3879 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3880 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3881 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3882 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3883 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3884 jiffies_to_msecs(cnf->rtr_solicit_interval);
3885 array[DEVCONF_RTR_SOLICIT_DELAY] =
3886 jiffies_to_msecs(cnf->rtr_solicit_delay);
3887 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3888 #ifdef CONFIG_IPV6_PRIVACY
3889 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3890 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3891 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3892 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3893 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3895 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3896 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3897 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3898 #ifdef CONFIG_IPV6_ROUTER_PREF
3899 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3900 array[DEVCONF_RTR_PROBE_INTERVAL] =
3901 jiffies_to_msecs(cnf->rtr_probe_interval);
3902 #ifdef CONFIG_IPV6_ROUTE_INFO
3903 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3906 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3907 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3908 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3909 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3911 #ifdef CONFIG_IPV6_MROUTE
3912 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3914 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3915 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3916 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3919 static inline size_t inet6_ifla6_size(void)
3921 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3922 + nla_total_size(sizeof(struct ifla_cacheinfo))
3923 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3924 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3925 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3928 static inline size_t inet6_if_nlmsg_size(void)
3930 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3931 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3932 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3933 + nla_total_size(4) /* IFLA_MTU */
3934 + nla_total_size(4) /* IFLA_LINK */
3935 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3938 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3939 int items, int bytes)
3942 int pad = bytes - sizeof(u64) * items;
3945 /* Use put_unaligned() because stats may not be aligned for u64. */
3946 put_unaligned(items, &stats[0]);
3947 for (i = 1; i < items; i++)
3948 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3950 memset(&stats[items], 0, pad);
3953 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3954 int items, int bytes, size_t syncpoff)
3957 int pad = bytes - sizeof(u64) * items;
3960 /* Use put_unaligned() because stats may not be aligned for u64. */
3961 put_unaligned(items, &stats[0]);
3962 for (i = 1; i < items; i++)
3963 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3965 memset(&stats[items], 0, pad);
3968 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3972 case IFLA_INET6_STATS:
3973 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3974 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3976 case IFLA_INET6_ICMP6STATS:
3977 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3982 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3985 struct ifla_cacheinfo ci;
3987 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3989 ci.max_reasm_len = IPV6_MAXPLEN;
3990 ci.tstamp = cstamp_delta(idev->tstamp);
3991 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3992 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3993 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3995 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3997 goto nla_put_failure;
3998 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4000 /* XXX - MC not implemented */
4002 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4004 goto nla_put_failure;
4005 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4007 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4009 goto nla_put_failure;
4010 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4018 static size_t inet6_get_link_af_size(const struct net_device *dev)
4020 if (!__in6_dev_get(dev))
4023 return inet6_ifla6_size();
4026 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4028 struct inet6_dev *idev = __in6_dev_get(dev);
4033 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4039 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4040 u32 pid, u32 seq, int event, unsigned int flags)
4042 struct net_device *dev = idev->dev;
4043 struct ifinfomsg *hdr;
4044 struct nlmsghdr *nlh;
4047 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4051 hdr = nlmsg_data(nlh);
4052 hdr->ifi_family = AF_INET6;
4054 hdr->ifi_type = dev->type;
4055 hdr->ifi_index = dev->ifindex;
4056 hdr->ifi_flags = dev_get_flags(dev);
4057 hdr->ifi_change = 0;
4059 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4062 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4064 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4065 if (dev->ifindex != dev->iflink)
4066 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4068 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4069 if (protoinfo == NULL)
4070 goto nla_put_failure;
4072 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4073 goto nla_put_failure;
4075 nla_nest_end(skb, protoinfo);
4076 return nlmsg_end(skb, nlh);
4079 nlmsg_cancel(skb, nlh);
4083 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4085 struct net *net = sock_net(skb->sk);
4088 struct net_device *dev;
4089 struct inet6_dev *idev;
4090 struct hlist_head *head;
4091 struct hlist_node *node;
4094 s_idx = cb->args[1];
4097 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4099 head = &net->dev_index_head[h];
4100 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4103 idev = __in6_dev_get(dev);
4106 if (inet6_fill_ifinfo(skb, idev,
4107 NETLINK_CB(cb->skb).pid,
4109 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4123 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4125 struct sk_buff *skb;
4126 struct net *net = dev_net(idev->dev);
4129 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4133 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4135 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4136 WARN_ON(err == -EMSGSIZE);
4140 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4144 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4147 static inline size_t inet6_prefix_nlmsg_size(void)
4149 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4150 + nla_total_size(sizeof(struct in6_addr))
4151 + nla_total_size(sizeof(struct prefix_cacheinfo));
4154 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4155 struct prefix_info *pinfo, u32 pid, u32 seq,
4156 int event, unsigned int flags)
4158 struct prefixmsg *pmsg;
4159 struct nlmsghdr *nlh;
4160 struct prefix_cacheinfo ci;
4162 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4166 pmsg = nlmsg_data(nlh);
4167 pmsg->prefix_family = AF_INET6;
4168 pmsg->prefix_pad1 = 0;
4169 pmsg->prefix_pad2 = 0;
4170 pmsg->prefix_ifindex = idev->dev->ifindex;
4171 pmsg->prefix_len = pinfo->prefix_len;
4172 pmsg->prefix_type = pinfo->type;
4173 pmsg->prefix_pad3 = 0;
4174 pmsg->prefix_flags = 0;
4176 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4177 if (pinfo->autoconf)
4178 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4180 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4182 ci.preferred_time = ntohl(pinfo->prefered);
4183 ci.valid_time = ntohl(pinfo->valid);
4184 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4186 return nlmsg_end(skb, nlh);
4189 nlmsg_cancel(skb, nlh);
4193 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4194 struct prefix_info *pinfo)
4196 struct sk_buff *skb;
4197 struct net *net = dev_net(idev->dev);
4200 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4204 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4206 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4207 WARN_ON(err == -EMSGSIZE);
4211 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4215 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4218 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4220 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4225 * If the address was optimistic
4226 * we inserted the route at the start of
4227 * our DAD process, so we don't need
4230 if (!(ifp->rt->rt6i_node))
4231 ip6_ins_rt(ifp->rt);
4232 if (ifp->idev->cnf.forwarding)
4233 addrconf_join_anycast(ifp);
4236 if (ifp->idev->cnf.forwarding)
4237 addrconf_leave_anycast(ifp);
4238 addrconf_leave_solict(ifp->idev, &ifp->addr);
4239 dst_hold(&ifp->rt->dst);
4241 if (ip6_del_rt(ifp->rt))
4242 dst_free(&ifp->rt->dst);
4247 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4250 if (likely(ifp->idev->dead == 0))
4251 __ipv6_ifa_notify(event, ifp);
4252 rcu_read_unlock_bh();
4255 #ifdef CONFIG_SYSCTL
4258 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4259 void __user *buffer, size_t *lenp, loff_t *ppos)
4261 int *valp = ctl->data;
4268 * ctl->data points to idev->cnf.forwarding, we should
4269 * not modify it until we get the rtnl lock.
4274 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4277 ret = addrconf_fixup_forwarding(ctl, valp, val);
4283 static void dev_disable_change(struct inet6_dev *idev)
4285 if (!idev || !idev->dev)
4288 if (idev->cnf.disable_ipv6)
4289 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4291 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4294 static void addrconf_disable_change(struct net *net, __s32 newf)
4296 struct net_device *dev;
4297 struct inet6_dev *idev;
4300 for_each_netdev_rcu(net, dev) {
4301 idev = __in6_dev_get(dev);
4303 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4304 idev->cnf.disable_ipv6 = newf;
4306 dev_disable_change(idev);
4312 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4317 if (!rtnl_trylock())
4318 return restart_syscall();
4320 net = (struct net *)table->extra2;
4324 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4329 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4330 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4331 addrconf_disable_change(net, newf);
4332 } else if ((!newf) ^ (!old))
4333 dev_disable_change((struct inet6_dev *)table->extra1);
4340 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4341 void __user *buffer, size_t *lenp, loff_t *ppos)
4343 int *valp = ctl->data;
4350 * ctl->data points to idev->cnf.disable_ipv6, we should
4351 * not modify it until we get the rtnl lock.
4356 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4359 ret = addrconf_disable_ipv6(ctl, valp, val);
4365 static struct addrconf_sysctl_table
4367 struct ctl_table_header *sysctl_header;
4368 ctl_table addrconf_vars[DEVCONF_MAX+1];
4370 } addrconf_sysctl __read_mostly = {
4371 .sysctl_header = NULL,
4374 .procname = "forwarding",
4375 .data = &ipv6_devconf.forwarding,
4376 .maxlen = sizeof(int),
4378 .proc_handler = addrconf_sysctl_forward,
4381 .procname = "hop_limit",
4382 .data = &ipv6_devconf.hop_limit,
4383 .maxlen = sizeof(int),
4385 .proc_handler = proc_dointvec,
4389 .data = &ipv6_devconf.mtu6,
4390 .maxlen = sizeof(int),
4392 .proc_handler = proc_dointvec,
4395 .procname = "accept_ra",
4396 .data = &ipv6_devconf.accept_ra,
4397 .maxlen = sizeof(int),
4399 .proc_handler = proc_dointvec,
4402 .procname = "accept_redirects",
4403 .data = &ipv6_devconf.accept_redirects,
4404 .maxlen = sizeof(int),
4406 .proc_handler = proc_dointvec,
4409 .procname = "autoconf",
4410 .data = &ipv6_devconf.autoconf,
4411 .maxlen = sizeof(int),
4413 .proc_handler = proc_dointvec,
4416 .procname = "dad_transmits",
4417 .data = &ipv6_devconf.dad_transmits,
4418 .maxlen = sizeof(int),
4420 .proc_handler = proc_dointvec,
4423 .procname = "router_solicitations",
4424 .data = &ipv6_devconf.rtr_solicits,
4425 .maxlen = sizeof(int),
4427 .proc_handler = proc_dointvec,
4430 .procname = "router_solicitation_interval",
4431 .data = &ipv6_devconf.rtr_solicit_interval,
4432 .maxlen = sizeof(int),
4434 .proc_handler = proc_dointvec_jiffies,
4437 .procname = "router_solicitation_delay",
4438 .data = &ipv6_devconf.rtr_solicit_delay,
4439 .maxlen = sizeof(int),
4441 .proc_handler = proc_dointvec_jiffies,
4444 .procname = "force_mld_version",
4445 .data = &ipv6_devconf.force_mld_version,
4446 .maxlen = sizeof(int),
4448 .proc_handler = proc_dointvec,
4450 #ifdef CONFIG_IPV6_PRIVACY
4452 .procname = "use_tempaddr",
4453 .data = &ipv6_devconf.use_tempaddr,
4454 .maxlen = sizeof(int),
4456 .proc_handler = proc_dointvec,
4459 .procname = "temp_valid_lft",
4460 .data = &ipv6_devconf.temp_valid_lft,
4461 .maxlen = sizeof(int),
4463 .proc_handler = proc_dointvec,
4466 .procname = "temp_prefered_lft",
4467 .data = &ipv6_devconf.temp_prefered_lft,
4468 .maxlen = sizeof(int),
4470 .proc_handler = proc_dointvec,
4473 .procname = "regen_max_retry",
4474 .data = &ipv6_devconf.regen_max_retry,
4475 .maxlen = sizeof(int),
4477 .proc_handler = proc_dointvec,
4480 .procname = "max_desync_factor",
4481 .data = &ipv6_devconf.max_desync_factor,
4482 .maxlen = sizeof(int),
4484 .proc_handler = proc_dointvec,
4488 .procname = "max_addresses",
4489 .data = &ipv6_devconf.max_addresses,
4490 .maxlen = sizeof(int),
4492 .proc_handler = proc_dointvec,
4495 .procname = "accept_ra_defrtr",
4496 .data = &ipv6_devconf.accept_ra_defrtr,
4497 .maxlen = sizeof(int),
4499 .proc_handler = proc_dointvec,
4502 .procname = "accept_ra_pinfo",
4503 .data = &ipv6_devconf.accept_ra_pinfo,
4504 .maxlen = sizeof(int),
4506 .proc_handler = proc_dointvec,
4508 #ifdef CONFIG_IPV6_ROUTER_PREF
4510 .procname = "accept_ra_rtr_pref",
4511 .data = &ipv6_devconf.accept_ra_rtr_pref,
4512 .maxlen = sizeof(int),
4514 .proc_handler = proc_dointvec,
4517 .procname = "router_probe_interval",
4518 .data = &ipv6_devconf.rtr_probe_interval,
4519 .maxlen = sizeof(int),
4521 .proc_handler = proc_dointvec_jiffies,
4523 #ifdef CONFIG_IPV6_ROUTE_INFO
4525 .procname = "accept_ra_rt_info_max_plen",
4526 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4527 .maxlen = sizeof(int),
4529 .proc_handler = proc_dointvec,
4534 .procname = "proxy_ndp",
4535 .data = &ipv6_devconf.proxy_ndp,
4536 .maxlen = sizeof(int),
4538 .proc_handler = proc_dointvec,
4541 .procname = "accept_source_route",
4542 .data = &ipv6_devconf.accept_source_route,
4543 .maxlen = sizeof(int),
4545 .proc_handler = proc_dointvec,
4547 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4549 .procname = "optimistic_dad",
4550 .data = &ipv6_devconf.optimistic_dad,
4551 .maxlen = sizeof(int),
4553 .proc_handler = proc_dointvec,
4557 #ifdef CONFIG_IPV6_MROUTE
4559 .procname = "mc_forwarding",
4560 .data = &ipv6_devconf.mc_forwarding,
4561 .maxlen = sizeof(int),
4563 .proc_handler = proc_dointvec,
4567 .procname = "disable_ipv6",
4568 .data = &ipv6_devconf.disable_ipv6,
4569 .maxlen = sizeof(int),
4571 .proc_handler = addrconf_sysctl_disable,
4574 .procname = "accept_dad",
4575 .data = &ipv6_devconf.accept_dad,
4576 .maxlen = sizeof(int),
4578 .proc_handler = proc_dointvec,
4581 .procname = "force_tllao",
4582 .data = &ipv6_devconf.force_tllao,
4583 .maxlen = sizeof(int),
4585 .proc_handler = proc_dointvec
4593 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4594 struct inet6_dev *idev, struct ipv6_devconf *p)
4597 struct addrconf_sysctl_table *t;
4599 #define ADDRCONF_CTL_PATH_DEV 3
4601 struct ctl_path addrconf_ctl_path[] = {
4602 { .procname = "net", },
4603 { .procname = "ipv6", },
4604 { .procname = "conf", },
4605 { /* to be set */ },
4610 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4614 for (i = 0; t->addrconf_vars[i].data; i++) {
4615 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4616 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4617 t->addrconf_vars[i].extra2 = net;
4621 * Make a copy of dev_name, because '.procname' is regarded as const
4622 * by sysctl and we wouldn't want anyone to change it under our feet
4623 * (see SIOCSIFNAME).
4625 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4629 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4631 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4633 if (t->sysctl_header == NULL)
4647 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4649 struct addrconf_sysctl_table *t;
4651 if (p->sysctl == NULL)
4656 unregister_net_sysctl_table(t->sysctl_header);
4661 static void addrconf_sysctl_register(struct inet6_dev *idev)
4663 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4664 &ndisc_ifinfo_sysctl_change);
4665 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4669 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4671 __addrconf_sysctl_unregister(&idev->cnf);
4672 neigh_sysctl_unregister(idev->nd_parms);
4678 static int __net_init addrconf_init_net(struct net *net)
4681 struct ipv6_devconf *all, *dflt;
4684 all = &ipv6_devconf;
4685 dflt = &ipv6_devconf_dflt;
4687 if (!net_eq(net, &init_net)) {
4688 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4692 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4694 goto err_alloc_dflt;
4696 /* these will be inherited by all namespaces */
4697 dflt->autoconf = ipv6_defaults.autoconf;
4698 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4701 net->ipv6.devconf_all = all;
4702 net->ipv6.devconf_dflt = dflt;
4704 #ifdef CONFIG_SYSCTL
4705 err = __addrconf_sysctl_register(net, "all", NULL, all);
4709 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4715 #ifdef CONFIG_SYSCTL
4717 __addrconf_sysctl_unregister(all);
4727 static void __net_exit addrconf_exit_net(struct net *net)
4729 #ifdef CONFIG_SYSCTL
4730 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4731 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4733 if (!net_eq(net, &init_net)) {
4734 kfree(net->ipv6.devconf_dflt);
4735 kfree(net->ipv6.devconf_all);
4739 static struct pernet_operations addrconf_ops = {
4740 .init = addrconf_init_net,
4741 .exit = addrconf_exit_net,
4748 int register_inet6addr_notifier(struct notifier_block *nb)
4750 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4752 EXPORT_SYMBOL(register_inet6addr_notifier);
4754 int unregister_inet6addr_notifier(struct notifier_block *nb)
4756 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4758 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4760 static struct rtnl_af_ops inet6_ops = {
4762 .fill_link_af = inet6_fill_link_af,
4763 .get_link_af_size = inet6_get_link_af_size,
4767 * Init / cleanup code
4770 int __init addrconf_init(void)
4774 err = ipv6_addr_label_init();
4776 printk(KERN_CRIT "IPv6 Addrconf:"
4777 " cannot initialize default policy table: %d.\n", err);
4781 err = register_pernet_subsys(&addrconf_ops);
4785 /* The addrconf netdev notifier requires that loopback_dev
4786 * has it's ipv6 private information allocated and setup
4787 * before it can bring up and give link-local addresses
4788 * to other devices which are up.
4790 * Unfortunately, loopback_dev is not necessarily the first
4791 * entry in the global dev_base list of net devices. In fact,
4792 * it is likely to be the very last entry on that list.
4793 * So this causes the notifier registry below to try and
4794 * give link-local addresses to all devices besides loopback_dev
4795 * first, then loopback_dev, which cases all the non-loopback_dev
4796 * devices to fail to get a link-local address.
4798 * So, as a temporary fix, allocate the ipv6 structure for
4799 * loopback_dev first by hand.
4800 * Longer term, all of the dependencies ipv6 has upon the loopback
4801 * device and it being up should be removed.
4804 if (!ipv6_add_dev(init_net.loopback_dev))
4810 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4811 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4813 register_netdevice_notifier(&ipv6_dev_notf);
4817 err = rtnl_af_register(&inet6_ops);
4821 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4826 /* Only the first call to __rtnl_register can fail */
4827 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4828 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4829 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4830 inet6_dump_ifaddr, NULL);
4831 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4832 inet6_dump_ifmcaddr, NULL);
4833 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4834 inet6_dump_ifacaddr, NULL);
4836 ipv6_addr_label_rtnl_register();
4840 rtnl_af_unregister(&inet6_ops);
4842 unregister_netdevice_notifier(&ipv6_dev_notf);
4844 unregister_pernet_subsys(&addrconf_ops);
4846 ipv6_addr_label_cleanup();
4851 void addrconf_cleanup(void)
4853 struct net_device *dev;
4856 unregister_netdevice_notifier(&ipv6_dev_notf);
4857 unregister_pernet_subsys(&addrconf_ops);
4858 ipv6_addr_label_cleanup();
4862 __rtnl_af_unregister(&inet6_ops);
4864 /* clean dev list */
4865 for_each_netdev(&init_net, dev) {
4866 if (__in6_dev_get(dev) == NULL)
4868 addrconf_ifdown(dev, 1);
4870 addrconf_ifdown(init_net.loopback_dev, 2);
4875 spin_lock_bh(&addrconf_hash_lock);
4876 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4877 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4878 spin_unlock_bh(&addrconf_hash_lock);
4880 del_timer(&addr_chk_timer);