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 && (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 old)
511 net = (struct net *)table->extra2;
512 if (p == &net->ipv6.devconf_dflt->forwarding)
515 if (!rtnl_trylock()) {
516 /* Restore the original values before restarting */
518 return restart_syscall();
521 if (p == &net->ipv6.devconf_all->forwarding) {
522 __s32 newf = net->ipv6.devconf_all->forwarding;
523 net->ipv6.devconf_dflt->forwarding = newf;
524 addrconf_forward_change(net, newf);
525 } else if ((!*p) ^ (!old))
526 dev_forward_change((struct inet6_dev *)table->extra1);
530 rt6_purge_dflt_routers(net);
535 /* Nobody refers to this ifaddr, destroy it */
536 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
538 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
540 #ifdef NET_REFCNT_DEBUG
541 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
544 in6_dev_put(ifp->idev);
546 if (del_timer(&ifp->timer))
547 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
549 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
550 pr_warning("Freeing alive inet6 address %p\n", ifp);
553 dst_release(&ifp->rt->dst);
559 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
562 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
565 * Each device address list is sorted in order of scope -
566 * global before linklocal.
568 list_for_each(p, &idev->addr_list) {
569 struct inet6_ifaddr *ifa
570 = list_entry(p, struct inet6_ifaddr, if_list);
571 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
575 list_add_tail(&ifp->if_list, p);
578 static u32 ipv6_addr_hash(const struct in6_addr *addr)
581 * We perform the hash function over the last 64 bits of the address
582 * This will include the IEEE address token on links that support it.
584 return jhash_2words((__force u32)addr->s6_addr32[2],
585 (__force u32)addr->s6_addr32[3], 0)
586 & (IN6_ADDR_HSIZE - 1);
589 /* On success it returns ifp with increased reference count */
591 static struct inet6_ifaddr *
592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
593 int scope, u32 flags)
595 struct inet6_ifaddr *ifa = NULL;
599 int addr_type = ipv6_addr_type(addr);
601 if (addr_type == IPV6_ADDR_ANY ||
602 addr_type & IPV6_ADDR_MULTICAST ||
603 (!(idev->dev->flags & IFF_LOOPBACK) &&
604 addr_type & IPV6_ADDR_LOOPBACK))
605 return ERR_PTR(-EADDRNOTAVAIL);
609 err = -ENODEV; /*XXX*/
613 if (idev->cnf.disable_ipv6) {
618 spin_lock(&addrconf_hash_lock);
620 /* Ignore adding duplicate addresses on an interface */
621 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
622 ADBG(("ipv6_add_addr: already assigned\n"));
627 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
630 ADBG(("ipv6_add_addr: malloc failed\n"));
635 rt = addrconf_dst_alloc(idev, addr, 0);
641 ipv6_addr_copy(&ifa->addr, addr);
643 spin_lock_init(&ifa->lock);
644 spin_lock_init(&ifa->state_lock);
645 init_timer(&ifa->timer);
646 INIT_HLIST_NODE(&ifa->addr_lst);
647 ifa->timer.data = (unsigned long) ifa;
649 ifa->prefix_len = pfxlen;
650 ifa->flags = flags | IFA_F_TENTATIVE;
651 ifa->cstamp = ifa->tstamp = jiffies;
656 * part one of RFC 4429, section 3.3
657 * We should not configure an address as
658 * optimistic if we do not yet know the link
659 * layer address of our nexhop router
662 if (dst_get_neighbour_raw(&rt->dst) == NULL)
663 ifa->flags &= ~IFA_F_OPTIMISTIC;
670 /* Add to big hash table */
671 hash = ipv6_addr_hash(addr);
673 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
674 spin_unlock(&addrconf_hash_lock);
676 write_lock(&idev->lock);
677 /* Add to inet6_dev unicast addr list. */
678 ipv6_link_dev_addr(idev, ifa);
680 #ifdef CONFIG_IPV6_PRIVACY
681 if (ifa->flags&IFA_F_TEMPORARY) {
682 list_add(&ifa->tmp_list, &idev->tempaddr_list);
688 write_unlock(&idev->lock);
690 rcu_read_unlock_bh();
692 if (likely(err == 0))
693 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
701 spin_unlock(&addrconf_hash_lock);
705 /* This function wants to get referenced ifp and releases it before return */
707 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
709 struct inet6_ifaddr *ifa, *ifn;
710 struct inet6_dev *idev = ifp->idev;
712 int deleted = 0, onlink = 0;
713 unsigned long expires = jiffies;
715 spin_lock_bh(&ifp->state_lock);
717 ifp->state = INET6_IFADDR_STATE_DEAD;
718 spin_unlock_bh(&ifp->state_lock);
720 if (state == INET6_IFADDR_STATE_DEAD)
723 spin_lock_bh(&addrconf_hash_lock);
724 hlist_del_init_rcu(&ifp->addr_lst);
725 spin_unlock_bh(&addrconf_hash_lock);
727 write_lock_bh(&idev->lock);
728 #ifdef CONFIG_IPV6_PRIVACY
729 if (ifp->flags&IFA_F_TEMPORARY) {
730 list_del(&ifp->tmp_list);
732 in6_ifa_put(ifp->ifpub);
739 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
741 list_del_init(&ifp->if_list);
744 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
748 } else if (ifp->flags & IFA_F_PERMANENT) {
749 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
751 if (ifa->flags & IFA_F_PERMANENT) {
756 unsigned long lifetime;
761 spin_lock(&ifa->lock);
763 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
765 * Note: Because this address is
766 * not permanent, lifetime <
767 * LONG_MAX / HZ here.
769 if (time_before(expires,
770 ifa->tstamp + lifetime * HZ))
771 expires = ifa->tstamp + lifetime * HZ;
772 spin_unlock(&ifa->lock);
777 write_unlock_bh(&idev->lock);
779 addrconf_del_timer(ifp);
781 ipv6_ifa_notify(RTM_DELADDR, ifp);
783 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
786 * Purge or update corresponding prefix
788 * 1) we don't purge prefix here if address was not permanent.
789 * prefix is managed by its own lifetime.
790 * 2) if there're no addresses, delete prefix.
791 * 3) if there're still other permanent address(es),
792 * corresponding prefix is still permanent.
793 * 4) otherwise, update prefix lifetime to the
794 * longest valid lifetime among the corresponding
795 * addresses on the device.
796 * Note: subsequent RA will update lifetime.
800 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
801 struct in6_addr prefix;
803 struct net *net = dev_net(ifp->idev->dev);
804 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
805 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
807 if (rt && addrconf_is_prefix_route(rt)) {
811 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
812 rt->rt6i_expires = expires;
813 rt->rt6i_flags |= RTF_EXPIRES;
816 dst_release(&rt->dst);
819 /* clean up prefsrc entries */
820 rt6_remove_prefsrc(ifp);
825 #ifdef CONFIG_IPV6_PRIVACY
826 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
828 struct inet6_dev *idev = ifp->idev;
829 struct in6_addr addr, *tmpaddr;
830 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
831 unsigned long regen_advance;
836 unsigned long now = jiffies;
838 write_lock(&idev->lock);
840 spin_lock_bh(&ift->lock);
841 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
842 spin_unlock_bh(&ift->lock);
849 if (idev->cnf.use_tempaddr <= 0) {
850 write_unlock(&idev->lock);
852 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
857 spin_lock_bh(&ifp->lock);
858 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
859 idev->cnf.use_tempaddr = -1; /*XXX*/
860 spin_unlock_bh(&ifp->lock);
861 write_unlock(&idev->lock);
863 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
869 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
870 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
871 spin_unlock_bh(&ifp->lock);
872 write_unlock(&idev->lock);
874 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
880 memcpy(&addr.s6_addr[8], idev->rndid, 8);
881 age = (now - ifp->tstamp) / HZ;
882 tmp_valid_lft = min_t(__u32,
884 idev->cnf.temp_valid_lft + age);
885 tmp_prefered_lft = min_t(__u32,
887 idev->cnf.temp_prefered_lft + age -
888 idev->cnf.max_desync_factor);
889 tmp_plen = ifp->prefix_len;
890 max_addresses = idev->cnf.max_addresses;
891 tmp_tstamp = ifp->tstamp;
892 spin_unlock_bh(&ifp->lock);
894 regen_advance = idev->cnf.regen_max_retry *
895 idev->cnf.dad_transmits *
896 idev->nd_parms->retrans_time / HZ;
897 write_unlock(&idev->lock);
899 /* A temporary address is created only if this calculated Preferred
900 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
901 * an implementation must not create a temporary address with a zero
902 * Preferred Lifetime.
904 if (tmp_prefered_lft <= regen_advance) {
911 addr_flags = IFA_F_TEMPORARY;
912 /* set in addrconf_prefix_rcv() */
913 if (ifp->flags & IFA_F_OPTIMISTIC)
914 addr_flags |= IFA_F_OPTIMISTIC;
916 ift = !max_addresses ||
917 ipv6_count_addresses(idev) < max_addresses ?
918 ipv6_add_addr(idev, &addr, tmp_plen,
919 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
921 if (!ift || IS_ERR(ift)) {
925 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
927 write_lock(&idev->lock);
931 spin_lock_bh(&ift->lock);
933 ift->valid_lft = tmp_valid_lft;
934 ift->prefered_lft = tmp_prefered_lft;
936 ift->tstamp = tmp_tstamp;
937 spin_unlock_bh(&ift->lock);
939 addrconf_dad_start(ift, 0);
948 * Choose an appropriate source address (RFC3484)
951 IPV6_SADDR_RULE_INIT = 0,
952 IPV6_SADDR_RULE_LOCAL,
953 IPV6_SADDR_RULE_SCOPE,
954 IPV6_SADDR_RULE_PREFERRED,
955 #ifdef CONFIG_IPV6_MIP6
959 IPV6_SADDR_RULE_LABEL,
960 #ifdef CONFIG_IPV6_PRIVACY
961 IPV6_SADDR_RULE_PRIVACY,
963 IPV6_SADDR_RULE_ORCHID,
964 IPV6_SADDR_RULE_PREFIX,
968 struct ipv6_saddr_score {
971 struct inet6_ifaddr *ifa;
972 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
977 struct ipv6_saddr_dst {
978 const struct in6_addr *addr;
985 static inline int ipv6_saddr_preferred(int type)
987 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
992 static int ipv6_get_saddr_eval(struct net *net,
993 struct ipv6_saddr_score *score,
994 struct ipv6_saddr_dst *dst,
999 if (i <= score->rule) {
1001 case IPV6_SADDR_RULE_SCOPE:
1002 ret = score->scopedist;
1004 case IPV6_SADDR_RULE_PREFIX:
1005 ret = score->matchlen;
1008 ret = !!test_bit(i, score->scorebits);
1014 case IPV6_SADDR_RULE_INIT:
1015 /* Rule 0: remember if hiscore is not ready yet */
1018 case IPV6_SADDR_RULE_LOCAL:
1019 /* Rule 1: Prefer same address */
1020 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1022 case IPV6_SADDR_RULE_SCOPE:
1023 /* Rule 2: Prefer appropriate scope
1028 * ---+--+-+---> scope
1030 * | d is scope of the destination.
1032 * | \ <- smaller scope is better if
1033 * B-15 | \ if scope is enough for destinaion.
1034 * | ret = B - scope (-1 <= scope >= d <= 15).
1036 * |/ <- greater is better
1037 * -C / if scope is not enough for destination.
1038 * /| ret = scope - C (-1 <= d < scope <= 15).
1040 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1041 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1042 * Assume B = 0 and we get C > 29.
1044 ret = __ipv6_addr_src_scope(score->addr_type);
1045 if (ret >= dst->scope)
1048 ret -= 128; /* 30 is enough */
1049 score->scopedist = ret;
1051 case IPV6_SADDR_RULE_PREFERRED:
1052 /* Rule 3: Avoid deprecated and optimistic addresses */
1053 ret = ipv6_saddr_preferred(score->addr_type) ||
1054 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1056 #ifdef CONFIG_IPV6_MIP6
1057 case IPV6_SADDR_RULE_HOA:
1059 /* Rule 4: Prefer home address */
1060 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1061 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1065 case IPV6_SADDR_RULE_OIF:
1066 /* Rule 5: Prefer outgoing interface */
1067 ret = (!dst->ifindex ||
1068 dst->ifindex == score->ifa->idev->dev->ifindex);
1070 case IPV6_SADDR_RULE_LABEL:
1071 /* Rule 6: Prefer matching label */
1072 ret = ipv6_addr_label(net,
1073 &score->ifa->addr, score->addr_type,
1074 score->ifa->idev->dev->ifindex) == dst->label;
1076 #ifdef CONFIG_IPV6_PRIVACY
1077 case IPV6_SADDR_RULE_PRIVACY:
1079 /* Rule 7: Prefer public address
1080 * Note: prefer temporary address if use_tempaddr >= 2
1082 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1083 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1084 score->ifa->idev->cnf.use_tempaddr >= 2;
1085 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1089 case IPV6_SADDR_RULE_ORCHID:
1090 /* Rule 8-: Prefer ORCHID vs ORCHID or
1091 * non-ORCHID vs non-ORCHID
1093 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1094 ipv6_addr_orchid(dst->addr));
1096 case IPV6_SADDR_RULE_PREFIX:
1097 /* Rule 8: Use longest matching prefix */
1098 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1106 __set_bit(i, score->scorebits);
1112 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1113 const struct in6_addr *daddr, unsigned int prefs,
1114 struct in6_addr *saddr)
1116 struct ipv6_saddr_score scores[2],
1117 *score = &scores[0], *hiscore = &scores[1];
1118 struct ipv6_saddr_dst dst;
1119 struct net_device *dev;
1122 dst_type = __ipv6_addr_type(daddr);
1124 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1125 dst.scope = __ipv6_addr_src_scope(dst_type);
1126 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1130 hiscore->ifa = NULL;
1134 for_each_netdev_rcu(net, dev) {
1135 struct inet6_dev *idev;
1137 /* Candidate Source Address (section 4)
1138 * - multicast and link-local destination address,
1139 * the set of candidate source address MUST only
1140 * include addresses assigned to interfaces
1141 * belonging to the same link as the outgoing
1143 * (- For site-local destination addresses, the
1144 * set of candidate source addresses MUST only
1145 * include addresses assigned to interfaces
1146 * belonging to the same site as the outgoing
1149 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1150 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1151 dst.ifindex && dev->ifindex != dst.ifindex)
1154 idev = __in6_dev_get(dev);
1158 read_lock_bh(&idev->lock);
1159 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1163 * - Tentative Address (RFC2462 section 5.4)
1164 * - A tentative address is not considered
1165 * "assigned to an interface" in the traditional
1166 * sense, unless it is also flagged as optimistic.
1167 * - Candidate Source Address (section 4)
1168 * - In any case, anycast addresses, multicast
1169 * addresses, and the unspecified address MUST
1170 * NOT be included in a candidate set.
1172 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1173 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1176 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1178 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1179 score->addr_type & IPV6_ADDR_MULTICAST)) {
1180 LIMIT_NETDEBUG(KERN_DEBUG
1181 "ADDRCONF: unspecified / multicast address "
1182 "assigned as unicast address on %s",
1188 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1190 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1191 int minihiscore, miniscore;
1193 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1194 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1196 if (minihiscore > miniscore) {
1197 if (i == IPV6_SADDR_RULE_SCOPE &&
1198 score->scopedist > 0) {
1201 * each remaining entry
1202 * has too small (not enough)
1203 * scope, because ifa entries
1204 * are sorted by their scope
1210 } else if (minihiscore < miniscore) {
1212 in6_ifa_put(hiscore->ifa);
1214 in6_ifa_hold(score->ifa);
1216 swap(hiscore, score);
1218 /* restore our iterator */
1219 score->ifa = hiscore->ifa;
1226 read_unlock_bh(&idev->lock);
1231 return -EADDRNOTAVAIL;
1233 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1234 in6_ifa_put(hiscore->ifa);
1237 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1239 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1240 unsigned char banned_flags)
1242 struct inet6_dev *idev;
1243 int err = -EADDRNOTAVAIL;
1246 idev = __in6_dev_get(dev);
1248 struct inet6_ifaddr *ifp;
1250 read_lock_bh(&idev->lock);
1251 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1252 if (ifp->scope == IFA_LINK &&
1253 !(ifp->flags & banned_flags)) {
1254 ipv6_addr_copy(addr, &ifp->addr);
1259 read_unlock_bh(&idev->lock);
1265 static int ipv6_count_addresses(struct inet6_dev *idev)
1268 struct inet6_ifaddr *ifp;
1270 read_lock_bh(&idev->lock);
1271 list_for_each_entry(ifp, &idev->addr_list, if_list)
1273 read_unlock_bh(&idev->lock);
1277 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1278 struct net_device *dev, int strict)
1280 struct inet6_ifaddr *ifp;
1281 struct hlist_node *node;
1282 unsigned int hash = ipv6_addr_hash(addr);
1285 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1286 if (!net_eq(dev_net(ifp->idev->dev), net))
1288 if (ipv6_addr_equal(&ifp->addr, addr) &&
1289 !(ifp->flags&IFA_F_TENTATIVE) &&
1290 (dev == NULL || ifp->idev->dev == dev ||
1291 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1292 rcu_read_unlock_bh();
1297 rcu_read_unlock_bh();
1300 EXPORT_SYMBOL(ipv6_chk_addr);
1302 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1303 struct net_device *dev)
1305 unsigned int hash = ipv6_addr_hash(addr);
1306 struct inet6_ifaddr *ifp;
1307 struct hlist_node *node;
1309 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1310 if (!net_eq(dev_net(ifp->idev->dev), net))
1312 if (ipv6_addr_equal(&ifp->addr, addr)) {
1313 if (dev == NULL || ifp->idev->dev == dev)
1320 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1322 struct inet6_dev *idev;
1323 struct inet6_ifaddr *ifa;
1328 idev = __in6_dev_get(dev);
1330 read_lock_bh(&idev->lock);
1331 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1332 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1337 read_unlock_bh(&idev->lock);
1343 EXPORT_SYMBOL(ipv6_chk_prefix);
1345 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1346 struct net_device *dev, int strict)
1348 struct inet6_ifaddr *ifp, *result = NULL;
1349 unsigned int hash = ipv6_addr_hash(addr);
1350 struct hlist_node *node;
1353 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1354 if (!net_eq(dev_net(ifp->idev->dev), net))
1356 if (ipv6_addr_equal(&ifp->addr, addr)) {
1357 if (dev == NULL || ifp->idev->dev == dev ||
1358 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1365 rcu_read_unlock_bh();
1370 /* Gets referenced address, destroys ifaddr */
1372 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1374 if (ifp->flags&IFA_F_PERMANENT) {
1375 spin_lock_bh(&ifp->lock);
1376 addrconf_del_timer(ifp);
1377 ifp->flags |= IFA_F_TENTATIVE;
1379 ifp->flags |= IFA_F_DADFAILED;
1380 spin_unlock_bh(&ifp->lock);
1382 ipv6_ifa_notify(0, ifp);
1384 #ifdef CONFIG_IPV6_PRIVACY
1385 } else if (ifp->flags&IFA_F_TEMPORARY) {
1386 struct inet6_ifaddr *ifpub;
1387 spin_lock_bh(&ifp->lock);
1390 in6_ifa_hold(ifpub);
1391 spin_unlock_bh(&ifp->lock);
1392 ipv6_create_tempaddr(ifpub, ifp);
1395 spin_unlock_bh(&ifp->lock);
1403 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1407 spin_lock(&ifp->state_lock);
1408 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1409 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1412 spin_unlock(&ifp->state_lock);
1417 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1419 struct inet6_dev *idev = ifp->idev;
1421 if (addrconf_dad_end(ifp)) {
1426 if (net_ratelimit())
1427 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1428 ifp->idev->dev->name, &ifp->addr);
1430 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1431 struct in6_addr addr;
1433 addr.s6_addr32[0] = htonl(0xfe800000);
1434 addr.s6_addr32[1] = 0;
1436 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1437 ipv6_addr_equal(&ifp->addr, &addr)) {
1438 /* DAD failed for link-local based on MAC address */
1439 idev->cnf.disable_ipv6 = 1;
1441 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1442 ifp->idev->dev->name);
1446 addrconf_dad_stop(ifp, 1);
1449 /* Join to solicited addr multicast group. */
1451 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1453 struct in6_addr maddr;
1455 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1458 addrconf_addr_solict_mult(addr, &maddr);
1459 ipv6_dev_mc_inc(dev, &maddr);
1462 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1464 struct in6_addr maddr;
1466 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1469 addrconf_addr_solict_mult(addr, &maddr);
1470 __ipv6_dev_mc_dec(idev, &maddr);
1473 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1475 struct in6_addr addr;
1476 if (ifp->prefix_len == 127) /* RFC 6164 */
1478 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1479 if (ipv6_addr_any(&addr))
1481 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1484 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1486 struct in6_addr addr;
1487 if (ifp->prefix_len == 127) /* RFC 6164 */
1489 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1490 if (ipv6_addr_any(&addr))
1492 __ipv6_dev_ac_dec(ifp->idev, &addr);
1495 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1497 if (dev->addr_len != ETH_ALEN)
1499 memcpy(eui, dev->dev_addr, 3);
1500 memcpy(eui + 5, dev->dev_addr + 3, 3);
1503 * The zSeries OSA network cards can be shared among various
1504 * OS instances, but the OSA cards have only one MAC address.
1505 * This leads to duplicate address conflicts in conjunction
1506 * with IPv6 if more than one instance uses the same card.
1508 * The driver for these cards can deliver a unique 16-bit
1509 * identifier for each instance sharing the same card. It is
1510 * placed instead of 0xFFFE in the interface identifier. The
1511 * "u" bit of the interface identifier is not inverted in this
1512 * case. Hence the resulting interface identifier has local
1513 * scope according to RFC2373.
1516 eui[3] = (dev->dev_id >> 8) & 0xFF;
1517 eui[4] = dev->dev_id & 0xFF;
1526 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1528 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1529 if (dev->addr_len != ARCNET_ALEN)
1532 eui[7] = *(u8*)dev->dev_addr;
1536 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1538 if (dev->addr_len != INFINIBAND_ALEN)
1540 memcpy(eui, dev->dev_addr + 12, 8);
1545 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1549 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1550 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1551 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1552 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1553 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1554 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1558 memcpy(eui + 4, &addr, 4);
1562 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1564 if (dev->priv_flags & IFF_ISATAP)
1565 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1569 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1571 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1574 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1576 switch (dev->type) {
1579 case ARPHRD_IEEE802_TR:
1580 return addrconf_ifid_eui48(eui, dev);
1582 return addrconf_ifid_arcnet(eui, dev);
1583 case ARPHRD_INFINIBAND:
1584 return addrconf_ifid_infiniband(eui, dev);
1586 return addrconf_ifid_sit(eui, dev);
1588 return addrconf_ifid_gre(eui, dev);
1593 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1596 struct inet6_ifaddr *ifp;
1598 read_lock_bh(&idev->lock);
1599 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1600 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1601 memcpy(eui, ifp->addr.s6_addr+8, 8);
1606 read_unlock_bh(&idev->lock);
1610 #ifdef CONFIG_IPV6_PRIVACY
1611 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1612 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1615 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1616 idev->rndid[0] &= ~0x02;
1619 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1620 * check if generated address is not inappropriate
1622 * - Reserved subnet anycast (RFC 2526)
1623 * 11111101 11....11 1xxxxxxx
1624 * - ISATAP (RFC4214) 6.1
1625 * 00-00-5E-FE-xx-xx-xx-xx
1627 * - XXX: already assigned to an address on the device
1629 if (idev->rndid[0] == 0xfd &&
1630 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1631 (idev->rndid[7]&0x80))
1633 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1634 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1636 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1643 static void ipv6_regen_rndid(unsigned long data)
1645 struct inet6_dev *idev = (struct inet6_dev *) data;
1646 unsigned long expires;
1649 write_lock_bh(&idev->lock);
1654 if (__ipv6_regen_rndid(idev) < 0)
1658 idev->cnf.temp_prefered_lft * HZ -
1659 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1660 idev->cnf.max_desync_factor * HZ;
1661 if (time_before(expires, jiffies)) {
1663 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1668 if (!mod_timer(&idev->regen_timer, expires))
1672 write_unlock_bh(&idev->lock);
1673 rcu_read_unlock_bh();
1677 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1680 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1681 ret = __ipv6_regen_rndid(idev);
1691 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1692 unsigned long expires, u32 flags)
1694 struct fib6_config cfg = {
1695 .fc_table = RT6_TABLE_PREFIX,
1696 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1697 .fc_ifindex = dev->ifindex,
1698 .fc_expires = expires,
1700 .fc_flags = RTF_UP | flags,
1701 .fc_nlinfo.nl_net = dev_net(dev),
1702 .fc_protocol = RTPROT_KERNEL,
1705 ipv6_addr_copy(&cfg.fc_dst, pfx);
1707 /* Prevent useless cloning on PtP SIT.
1708 This thing is done here expecting that the whole
1709 class of non-broadcast devices need not cloning.
1711 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1712 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1713 cfg.fc_flags |= RTF_NONEXTHOP;
1716 ip6_route_add(&cfg);
1720 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1722 const struct net_device *dev,
1723 u32 flags, u32 noflags)
1725 struct fib6_node *fn;
1726 struct rt6_info *rt = NULL;
1727 struct fib6_table *table;
1729 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1733 write_lock_bh(&table->tb6_lock);
1734 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1737 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1738 if (rt->rt6i_dev->ifindex != dev->ifindex)
1740 if ((rt->rt6i_flags & flags) != flags)
1742 if ((rt->rt6i_flags & noflags) != 0)
1748 write_unlock_bh(&table->tb6_lock);
1753 /* Create "default" multicast route to the interface */
1755 static void addrconf_add_mroute(struct net_device *dev)
1757 struct fib6_config cfg = {
1758 .fc_table = RT6_TABLE_LOCAL,
1759 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1760 .fc_ifindex = dev->ifindex,
1763 .fc_nlinfo.nl_net = dev_net(dev),
1766 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1768 ip6_route_add(&cfg);
1771 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1772 static void sit_route_add(struct net_device *dev)
1774 struct fib6_config cfg = {
1775 .fc_table = RT6_TABLE_MAIN,
1776 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1777 .fc_ifindex = dev->ifindex,
1779 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1780 .fc_nlinfo.nl_net = dev_net(dev),
1783 /* prefix length - 96 bits "::d.d.d.d" */
1784 ip6_route_add(&cfg);
1788 static void addrconf_add_lroute(struct net_device *dev)
1790 struct in6_addr addr;
1792 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1793 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1796 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1798 struct inet6_dev *idev;
1802 idev = ipv6_find_idev(dev);
1804 return ERR_PTR(-ENOBUFS);
1806 if (idev->cnf.disable_ipv6)
1807 return ERR_PTR(-EACCES);
1809 /* Add default multicast route */
1810 if (!(dev->flags & IFF_LOOPBACK))
1811 addrconf_add_mroute(dev);
1813 /* Add link local route */
1814 addrconf_add_lroute(dev);
1818 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1820 struct prefix_info *pinfo;
1824 struct inet6_dev *in6_dev;
1825 struct net *net = dev_net(dev);
1827 pinfo = (struct prefix_info *) opt;
1829 if (len < sizeof(struct prefix_info)) {
1830 ADBG(("addrconf: prefix option too short\n"));
1835 * Validation checks ([ADDRCONF], page 19)
1838 addr_type = ipv6_addr_type(&pinfo->prefix);
1840 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1843 valid_lft = ntohl(pinfo->valid);
1844 prefered_lft = ntohl(pinfo->prefered);
1846 if (prefered_lft > valid_lft) {
1847 if (net_ratelimit())
1848 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1852 in6_dev = in6_dev_get(dev);
1854 if (in6_dev == NULL) {
1855 if (net_ratelimit())
1856 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1861 * Two things going on here:
1862 * 1) Add routes for on-link prefixes
1863 * 2) Configure prefixes with the auto flag set
1866 if (pinfo->onlink) {
1867 struct rt6_info *rt;
1868 unsigned long rt_expires;
1870 /* Avoid arithmetic overflow. Really, we could
1871 * save rt_expires in seconds, likely valid_lft,
1872 * but it would require division in fib gc, that it
1876 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1878 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1880 if (addrconf_finite_timeout(rt_expires))
1883 rt = addrconf_get_prefix_route(&pinfo->prefix,
1886 RTF_ADDRCONF | RTF_PREFIX_RT,
1887 RTF_GATEWAY | RTF_DEFAULT);
1890 /* Autoconf prefix route */
1891 if (valid_lft == 0) {
1894 } else if (addrconf_finite_timeout(rt_expires)) {
1896 rt->rt6i_expires = jiffies + rt_expires;
1897 rt->rt6i_flags |= RTF_EXPIRES;
1899 rt->rt6i_flags &= ~RTF_EXPIRES;
1900 rt->rt6i_expires = 0;
1902 } else if (valid_lft) {
1903 clock_t expires = 0;
1904 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1905 if (addrconf_finite_timeout(rt_expires)) {
1907 flags |= RTF_EXPIRES;
1908 expires = jiffies_to_clock_t(rt_expires);
1910 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1911 dev, expires, flags);
1914 dst_release(&rt->dst);
1917 /* Try to figure out our local address for this prefix */
1919 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1920 struct inet6_ifaddr * ifp;
1921 struct in6_addr addr;
1922 int create = 0, update_lft = 0;
1924 if (pinfo->prefix_len == 64) {
1925 memcpy(&addr, &pinfo->prefix, 8);
1926 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1927 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1928 in6_dev_put(in6_dev);
1933 if (net_ratelimit())
1934 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1936 in6_dev_put(in6_dev);
1941 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1943 if (ifp == NULL && valid_lft) {
1944 int max_addresses = in6_dev->cnf.max_addresses;
1947 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1948 if (in6_dev->cnf.optimistic_dad &&
1949 !net->ipv6.devconf_all->forwarding)
1950 addr_flags = IFA_F_OPTIMISTIC;
1953 /* Do not allow to create too much of autoconfigured
1954 * addresses; this would be too easy way to crash kernel.
1956 if (!max_addresses ||
1957 ipv6_count_addresses(in6_dev) < max_addresses)
1958 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1959 addr_type&IPV6_ADDR_SCOPE_MASK,
1962 if (!ifp || IS_ERR(ifp)) {
1963 in6_dev_put(in6_dev);
1967 update_lft = create = 1;
1968 ifp->cstamp = jiffies;
1969 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1975 #ifdef CONFIG_IPV6_PRIVACY
1976 struct inet6_ifaddr *ift;
1980 /* update lifetime (RFC2462 5.5.3 e) */
1981 spin_lock(&ifp->lock);
1983 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1984 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1987 if (!update_lft && stored_lft) {
1988 if (valid_lft > MIN_VALID_LIFETIME ||
1989 valid_lft > stored_lft)
1991 else if (stored_lft <= MIN_VALID_LIFETIME) {
1992 /* valid_lft <= stored_lft is always true */
1994 * RFC 4862 Section 5.5.3e:
1995 * "Note that the preferred lifetime of
1996 * the corresponding address is always
1997 * reset to the Preferred Lifetime in
1998 * the received Prefix Information
1999 * option, regardless of whether the
2000 * valid lifetime is also reset or
2003 * So if the preferred lifetime in
2004 * this advertisement is different
2005 * than what we have stored, but the
2006 * valid lifetime is invalid, just
2007 * reset prefered_lft.
2009 * We must set the valid lifetime
2010 * to the stored lifetime since we'll
2011 * be updating the timestamp below,
2012 * else we'll set it back to the
2015 if (prefered_lft != ifp->prefered_lft) {
2016 valid_lft = stored_lft;
2020 valid_lft = MIN_VALID_LIFETIME;
2021 if (valid_lft < prefered_lft)
2022 prefered_lft = valid_lft;
2028 ifp->valid_lft = valid_lft;
2029 ifp->prefered_lft = prefered_lft;
2032 ifp->flags &= ~IFA_F_DEPRECATED;
2033 spin_unlock(&ifp->lock);
2035 if (!(flags&IFA_F_TENTATIVE))
2036 ipv6_ifa_notify(0, ifp);
2038 spin_unlock(&ifp->lock);
2040 #ifdef CONFIG_IPV6_PRIVACY
2041 read_lock_bh(&in6_dev->lock);
2042 /* update all temporary addresses in the list */
2043 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2045 int age, max_valid, max_prefered;
2047 if (ifp != ift->ifpub)
2051 * RFC 4941 section 3.3:
2052 * If a received option will extend the lifetime
2053 * of a public address, the lifetimes of
2054 * temporary addresses should be extended,
2055 * subject to the overall constraint that no
2056 * temporary addresses should ever remain
2057 * "valid" or "preferred" for a time longer than
2058 * (TEMP_VALID_LIFETIME) or
2059 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2062 age = (now - ift->cstamp) / HZ;
2063 max_valid = in6_dev->cnf.temp_valid_lft - age;
2067 max_prefered = in6_dev->cnf.temp_prefered_lft -
2068 in6_dev->cnf.max_desync_factor -
2070 if (max_prefered < 0)
2073 if (valid_lft > max_valid)
2074 valid_lft = max_valid;
2076 if (prefered_lft > max_prefered)
2077 prefered_lft = max_prefered;
2079 spin_lock(&ift->lock);
2081 ift->valid_lft = valid_lft;
2082 ift->prefered_lft = prefered_lft;
2084 if (prefered_lft > 0)
2085 ift->flags &= ~IFA_F_DEPRECATED;
2087 spin_unlock(&ift->lock);
2088 if (!(flags&IFA_F_TENTATIVE))
2089 ipv6_ifa_notify(0, ift);
2092 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2094 * When a new public address is created as
2095 * described in [ADDRCONF], also create a new
2096 * temporary address. Also create a temporary
2097 * address if it's enabled but no temporary
2098 * address currently exists.
2100 read_unlock_bh(&in6_dev->lock);
2101 ipv6_create_tempaddr(ifp, NULL);
2103 read_unlock_bh(&in6_dev->lock);
2110 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2111 in6_dev_put(in6_dev);
2115 * Set destination address.
2116 * Special case for SIT interfaces where we create a new "virtual"
2119 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2121 struct in6_ifreq ireq;
2122 struct net_device *dev;
2128 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2131 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2137 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2138 if (dev->type == ARPHRD_SIT) {
2139 const struct net_device_ops *ops = dev->netdev_ops;
2141 struct ip_tunnel_parm p;
2143 err = -EADDRNOTAVAIL;
2144 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2147 memset(&p, 0, sizeof(p));
2148 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2152 p.iph.protocol = IPPROTO_IPV6;
2154 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2156 if (ops->ndo_do_ioctl) {
2157 mm_segment_t oldfs = get_fs();
2160 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2167 dev = __dev_get_by_name(net, p.name);
2170 err = dev_open(dev);
2181 * Manual configuration of address on an interface
2183 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2184 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2187 struct inet6_ifaddr *ifp;
2188 struct inet6_dev *idev;
2189 struct net_device *dev;
2193 unsigned long timeout;
2200 /* check the lifetime */
2201 if (!valid_lft || prefered_lft > valid_lft)
2204 dev = __dev_get_by_index(net, ifindex);
2208 idev = addrconf_add_dev(dev);
2210 return PTR_ERR(idev);
2212 scope = ipv6_addr_scope(pfx);
2214 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2215 if (addrconf_finite_timeout(timeout)) {
2216 expires = jiffies_to_clock_t(timeout * HZ);
2217 valid_lft = timeout;
2218 flags = RTF_EXPIRES;
2222 ifa_flags |= IFA_F_PERMANENT;
2225 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2226 if (addrconf_finite_timeout(timeout)) {
2228 ifa_flags |= IFA_F_DEPRECATED;
2229 prefered_lft = timeout;
2232 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2235 spin_lock_bh(&ifp->lock);
2236 ifp->valid_lft = valid_lft;
2237 ifp->prefered_lft = prefered_lft;
2238 ifp->tstamp = jiffies;
2239 spin_unlock_bh(&ifp->lock);
2241 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2244 * Note that section 3.1 of RFC 4429 indicates
2245 * that the Optimistic flag should not be set for
2246 * manually configured addresses
2248 addrconf_dad_start(ifp, 0);
2254 return PTR_ERR(ifp);
2257 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2260 struct inet6_ifaddr *ifp;
2261 struct inet6_dev *idev;
2262 struct net_device *dev;
2267 dev = __dev_get_by_index(net, ifindex);
2271 if ((idev = __in6_dev_get(dev)) == NULL)
2274 read_lock_bh(&idev->lock);
2275 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2276 if (ifp->prefix_len == plen &&
2277 ipv6_addr_equal(pfx, &ifp->addr)) {
2279 read_unlock_bh(&idev->lock);
2283 /* If the last address is deleted administratively,
2284 disable IPv6 on this interface.
2286 if (list_empty(&idev->addr_list))
2287 addrconf_ifdown(idev->dev, 1);
2291 read_unlock_bh(&idev->lock);
2292 return -EADDRNOTAVAIL;
2296 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2298 struct in6_ifreq ireq;
2301 if (!capable(CAP_NET_ADMIN))
2304 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2308 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2309 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2310 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2315 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2317 struct in6_ifreq ireq;
2320 if (!capable(CAP_NET_ADMIN))
2323 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2327 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2328 ireq.ifr6_prefixlen);
2333 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2334 int plen, int scope)
2336 struct inet6_ifaddr *ifp;
2338 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2340 spin_lock_bh(&ifp->lock);
2341 ifp->flags &= ~IFA_F_TENTATIVE;
2342 spin_unlock_bh(&ifp->lock);
2343 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2348 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2349 static void sit_add_v4_addrs(struct inet6_dev *idev)
2351 struct in6_addr addr;
2352 struct net_device *dev;
2353 struct net *net = dev_net(idev->dev);
2358 memset(&addr, 0, sizeof(struct in6_addr));
2359 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2361 if (idev->dev->flags&IFF_POINTOPOINT) {
2362 addr.s6_addr32[0] = htonl(0xfe800000);
2365 scope = IPV6_ADDR_COMPATv4;
2368 if (addr.s6_addr32[3]) {
2369 add_addr(idev, &addr, 128, scope);
2373 for_each_netdev(net, dev) {
2374 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2375 if (in_dev && (dev->flags & IFF_UP)) {
2376 struct in_ifaddr * ifa;
2380 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2383 addr.s6_addr32[3] = ifa->ifa_local;
2385 if (ifa->ifa_scope == RT_SCOPE_LINK)
2387 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2388 if (idev->dev->flags&IFF_POINTOPOINT)
2392 if (idev->dev->flags&IFF_POINTOPOINT)
2397 add_addr(idev, &addr, plen, flag);
2404 static void init_loopback(struct net_device *dev)
2406 struct inet6_dev *idev;
2407 struct net_device *sp_dev;
2408 struct inet6_ifaddr *sp_ifa;
2409 struct rt6_info *sp_rt;
2415 if ((idev = ipv6_find_idev(dev)) == NULL) {
2416 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2420 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2422 /* Add routes to other interface's IPv6 addresses */
2423 for_each_netdev(dev_net(dev), sp_dev) {
2424 if (!strcmp(sp_dev->name, dev->name))
2427 idev = __in6_dev_get(sp_dev);
2431 read_lock_bh(&idev->lock);
2432 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2434 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2437 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2439 /* Failure cases are ignored */
2443 read_unlock_bh(&idev->lock);
2447 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2449 struct inet6_ifaddr * ifp;
2450 u32 addr_flags = IFA_F_PERMANENT;
2452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2453 if (idev->cnf.optimistic_dad &&
2454 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2455 addr_flags |= IFA_F_OPTIMISTIC;
2459 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2461 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2462 addrconf_dad_start(ifp, 0);
2467 static void addrconf_dev_config(struct net_device *dev)
2469 struct in6_addr addr;
2470 struct inet6_dev * idev;
2474 if ((dev->type != ARPHRD_ETHER) &&
2475 (dev->type != ARPHRD_FDDI) &&
2476 (dev->type != ARPHRD_IEEE802_TR) &&
2477 (dev->type != ARPHRD_ARCNET) &&
2478 (dev->type != ARPHRD_INFINIBAND)) {
2479 /* Alas, we support only Ethernet autoconfiguration. */
2483 idev = addrconf_add_dev(dev);
2487 memset(&addr, 0, sizeof(struct in6_addr));
2488 addr.s6_addr32[0] = htonl(0xFE800000);
2490 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2491 addrconf_add_linklocal(idev, &addr);
2494 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2495 static void addrconf_sit_config(struct net_device *dev)
2497 struct inet6_dev *idev;
2502 * Configure the tunnel with one of our IPv4
2503 * addresses... we should configure all of
2504 * our v4 addrs in the tunnel
2507 if ((idev = ipv6_find_idev(dev)) == NULL) {
2508 printk(KERN_DEBUG "init sit: add_dev failed\n");
2512 if (dev->priv_flags & IFF_ISATAP) {
2513 struct in6_addr addr;
2515 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2516 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2517 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2518 addrconf_add_linklocal(idev, &addr);
2522 sit_add_v4_addrs(idev);
2524 if (dev->flags&IFF_POINTOPOINT) {
2525 addrconf_add_mroute(dev);
2526 addrconf_add_lroute(dev);
2532 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2533 static void addrconf_gre_config(struct net_device *dev)
2535 struct inet6_dev *idev;
2536 struct in6_addr addr;
2538 pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2542 if ((idev = ipv6_find_idev(dev)) == NULL) {
2543 printk(KERN_DEBUG "init gre: add_dev failed\n");
2547 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2548 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2550 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2551 addrconf_add_linklocal(idev, &addr);
2556 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2558 struct in6_addr lladdr;
2560 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2561 addrconf_add_linklocal(idev, &lladdr);
2567 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2569 struct net_device *link_dev;
2570 struct net *net = dev_net(idev->dev);
2572 /* first try to inherit the link-local address from the link device */
2573 if (idev->dev->iflink &&
2574 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2575 if (!ipv6_inherit_linklocal(idev, link_dev))
2578 /* then try to inherit it from any device */
2579 for_each_netdev(net, link_dev) {
2580 if (!ipv6_inherit_linklocal(idev, link_dev))
2583 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2587 * Autoconfigure tunnel with a link-local address so routing protocols,
2588 * DHCPv6, MLD etc. can be run over the virtual link
2591 static void addrconf_ip6_tnl_config(struct net_device *dev)
2593 struct inet6_dev *idev;
2597 idev = addrconf_add_dev(dev);
2599 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2602 ip6_tnl_add_linklocal(idev);
2605 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2608 struct net_device *dev = (struct net_device *) data;
2609 struct inet6_dev *idev = __in6_dev_get(dev);
2610 int run_pending = 0;
2614 case NETDEV_REGISTER:
2615 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2616 idev = ipv6_add_dev(dev);
2618 return notifier_from_errno(-ENOMEM);
2624 if (dev->flags & IFF_SLAVE)
2627 if (event == NETDEV_UP) {
2628 if (!addrconf_qdisc_ok(dev)) {
2629 /* device is not ready yet. */
2631 "ADDRCONF(NETDEV_UP): %s: "
2632 "link is not ready\n",
2637 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2638 idev = ipv6_add_dev(dev);
2641 idev->if_flags |= IF_READY;
2645 if (!addrconf_qdisc_ok(dev)) {
2646 /* device is still not ready. */
2651 if (idev->if_flags & IF_READY)
2652 /* device is already configured. */
2654 idev->if_flags |= IF_READY;
2658 "ADDRCONF(NETDEV_CHANGE): %s: "
2659 "link becomes ready\n",
2665 switch (dev->type) {
2666 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2668 addrconf_sit_config(dev);
2671 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2673 addrconf_gre_config(dev);
2676 case ARPHRD_TUNNEL6:
2677 addrconf_ip6_tnl_config(dev);
2679 case ARPHRD_LOOPBACK:
2684 addrconf_dev_config(dev);
2690 addrconf_dad_run(idev);
2693 * If the MTU changed during the interface down,
2694 * when the interface up, the changed MTU must be
2695 * reflected in the idev as well as routers.
2697 if (idev->cnf.mtu6 != dev->mtu &&
2698 dev->mtu >= IPV6_MIN_MTU) {
2699 rt6_mtu_change(dev, dev->mtu);
2700 idev->cnf.mtu6 = dev->mtu;
2702 idev->tstamp = jiffies;
2703 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2706 * If the changed mtu during down is lower than
2707 * IPV6_MIN_MTU stop IPv6 on this interface.
2709 if (dev->mtu < IPV6_MIN_MTU)
2710 addrconf_ifdown(dev, 1);
2714 case NETDEV_CHANGEMTU:
2715 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2716 rt6_mtu_change(dev, dev->mtu);
2717 idev->cnf.mtu6 = dev->mtu;
2721 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2722 idev = ipv6_add_dev(dev);
2728 * MTU falled under IPV6_MIN_MTU.
2729 * Stop IPv6 on this interface.
2733 case NETDEV_UNREGISTER:
2735 * Remove all addresses from this interface.
2737 addrconf_ifdown(dev, event != NETDEV_DOWN);
2740 case NETDEV_CHANGENAME:
2742 snmp6_unregister_dev(idev);
2743 addrconf_sysctl_unregister(idev);
2744 addrconf_sysctl_register(idev);
2745 err = snmp6_register_dev(idev);
2747 return notifier_from_errno(err);
2751 case NETDEV_PRE_TYPE_CHANGE:
2752 case NETDEV_POST_TYPE_CHANGE:
2753 addrconf_type_change(dev, event);
2761 * addrconf module should be notified of a device going up
2763 static struct notifier_block ipv6_dev_notf = {
2764 .notifier_call = addrconf_notify,
2767 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2769 struct inet6_dev *idev;
2772 idev = __in6_dev_get(dev);
2774 if (event == NETDEV_POST_TYPE_CHANGE)
2775 ipv6_mc_remap(idev);
2776 else if (event == NETDEV_PRE_TYPE_CHANGE)
2777 ipv6_mc_unmap(idev);
2780 static int addrconf_ifdown(struct net_device *dev, int how)
2782 struct net *net = dev_net(dev);
2783 struct inet6_dev *idev;
2784 struct inet6_ifaddr *ifa;
2789 rt6_ifdown(net, dev);
2790 neigh_ifdown(&nd_tbl, dev);
2792 idev = __in6_dev_get(dev);
2797 * Step 1: remove reference to ipv6 device from parent device.
2803 /* protected by rtnl_lock */
2804 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2806 /* Step 1.5: remove snmp6 entry */
2807 snmp6_unregister_dev(idev);
2811 /* Step 2: clear hash table */
2812 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2813 struct hlist_head *h = &inet6_addr_lst[i];
2814 struct hlist_node *n;
2816 spin_lock_bh(&addrconf_hash_lock);
2818 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2819 if (ifa->idev == idev) {
2820 hlist_del_init_rcu(&ifa->addr_lst);
2821 addrconf_del_timer(ifa);
2825 spin_unlock_bh(&addrconf_hash_lock);
2828 write_lock_bh(&idev->lock);
2830 /* Step 2: clear flags for stateless addrconf */
2832 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2834 #ifdef CONFIG_IPV6_PRIVACY
2835 if (how && del_timer(&idev->regen_timer))
2838 /* Step 3: clear tempaddr list */
2839 while (!list_empty(&idev->tempaddr_list)) {
2840 ifa = list_first_entry(&idev->tempaddr_list,
2841 struct inet6_ifaddr, tmp_list);
2842 list_del(&ifa->tmp_list);
2843 write_unlock_bh(&idev->lock);
2844 spin_lock_bh(&ifa->lock);
2847 in6_ifa_put(ifa->ifpub);
2850 spin_unlock_bh(&ifa->lock);
2852 write_lock_bh(&idev->lock);
2856 while (!list_empty(&idev->addr_list)) {
2857 ifa = list_first_entry(&idev->addr_list,
2858 struct inet6_ifaddr, if_list);
2859 addrconf_del_timer(ifa);
2861 list_del(&ifa->if_list);
2863 write_unlock_bh(&idev->lock);
2865 spin_lock_bh(&ifa->state_lock);
2867 ifa->state = INET6_IFADDR_STATE_DEAD;
2868 spin_unlock_bh(&ifa->state_lock);
2870 if (state != INET6_IFADDR_STATE_DEAD) {
2871 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2872 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2876 write_lock_bh(&idev->lock);
2879 write_unlock_bh(&idev->lock);
2881 /* Step 5: Discard multicast list */
2883 ipv6_mc_destroy_dev(idev);
2887 idev->tstamp = jiffies;
2889 /* Last: Shot the device (if unregistered) */
2891 addrconf_sysctl_unregister(idev);
2892 neigh_parms_release(&nd_tbl, idev->nd_parms);
2893 neigh_ifdown(&nd_tbl, dev);
2899 static void addrconf_rs_timer(unsigned long data)
2901 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2902 struct inet6_dev *idev = ifp->idev;
2904 read_lock(&idev->lock);
2905 if (idev->dead || !(idev->if_flags & IF_READY))
2908 if (idev->cnf.forwarding)
2911 /* Announcement received after solicitation was sent */
2912 if (idev->if_flags & IF_RA_RCVD)
2915 spin_lock(&ifp->lock);
2916 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2917 /* The wait after the last probe can be shorter */
2918 addrconf_mod_timer(ifp, AC_RS,
2919 (ifp->probes == idev->cnf.rtr_solicits) ?
2920 idev->cnf.rtr_solicit_delay :
2921 idev->cnf.rtr_solicit_interval);
2922 spin_unlock(&ifp->lock);
2924 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2926 spin_unlock(&ifp->lock);
2928 * Note: we do not support deprecated "all on-link"
2929 * assumption any longer.
2931 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2936 read_unlock(&idev->lock);
2941 * Duplicate Address Detection
2943 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2945 unsigned long rand_num;
2946 struct inet6_dev *idev = ifp->idev;
2948 if (ifp->flags & IFA_F_OPTIMISTIC)
2951 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2953 ifp->probes = idev->cnf.dad_transmits;
2954 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2957 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2959 struct inet6_dev *idev = ifp->idev;
2960 struct net_device *dev = idev->dev;
2962 addrconf_join_solict(dev, &ifp->addr);
2964 net_srandom(ifp->addr.s6_addr32[3]);
2966 read_lock_bh(&idev->lock);
2967 spin_lock(&ifp->lock);
2968 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2971 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2972 idev->cnf.accept_dad < 1 ||
2973 !(ifp->flags&IFA_F_TENTATIVE) ||
2974 ifp->flags & IFA_F_NODAD) {
2975 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2976 spin_unlock(&ifp->lock);
2977 read_unlock_bh(&idev->lock);
2979 addrconf_dad_completed(ifp);
2983 if (!(idev->if_flags & IF_READY)) {
2984 spin_unlock(&ifp->lock);
2985 read_unlock_bh(&idev->lock);
2987 * If the device is not ready:
2988 * - keep it tentative if it is a permanent address.
2989 * - otherwise, kill it.
2992 addrconf_dad_stop(ifp, 0);
2997 * Optimistic nodes can start receiving
3000 if (ifp->flags & IFA_F_OPTIMISTIC)
3001 ip6_ins_rt(ifp->rt);
3003 addrconf_dad_kick(ifp);
3005 spin_unlock(&ifp->lock);
3006 read_unlock_bh(&idev->lock);
3009 static void addrconf_dad_timer(unsigned long data)
3011 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3012 struct inet6_dev *idev = ifp->idev;
3013 struct in6_addr mcaddr;
3015 if (!ifp->probes && addrconf_dad_end(ifp))
3018 read_lock(&idev->lock);
3019 if (idev->dead || !(idev->if_flags & IF_READY)) {
3020 read_unlock(&idev->lock);
3024 spin_lock(&ifp->lock);
3025 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3026 spin_unlock(&ifp->lock);
3027 read_unlock(&idev->lock);
3031 if (ifp->probes == 0) {
3033 * DAD was successful
3036 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3037 spin_unlock(&ifp->lock);
3038 read_unlock(&idev->lock);
3040 addrconf_dad_completed(ifp);
3046 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3047 spin_unlock(&ifp->lock);
3048 read_unlock(&idev->lock);
3050 /* send a neighbour solicitation for our addr */
3051 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3052 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3057 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3059 struct net_device *dev = ifp->idev->dev;
3062 * Configure the address for reception. Now it is valid.
3065 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3067 /* If added prefix is link local and we are prepared to process
3068 router advertisements, start sending router solicitations.
3071 if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3072 ifp->idev->cnf.accept_ra == 2) &&
3073 ifp->idev->cnf.rtr_solicits > 0 &&
3074 (dev->flags&IFF_LOOPBACK) == 0 &&
3075 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3077 * If a host as already performed a random delay
3078 * [...] as part of DAD [...] there is no need
3079 * to delay again before sending the first RS
3081 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3083 spin_lock_bh(&ifp->lock);
3085 ifp->idev->if_flags |= IF_RS_SENT;
3086 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3087 spin_unlock_bh(&ifp->lock);
3091 static void addrconf_dad_run(struct inet6_dev *idev)
3093 struct inet6_ifaddr *ifp;
3095 read_lock_bh(&idev->lock);
3096 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3097 spin_lock(&ifp->lock);
3098 if (ifp->flags & IFA_F_TENTATIVE &&
3099 ifp->state == INET6_IFADDR_STATE_DAD)
3100 addrconf_dad_kick(ifp);
3101 spin_unlock(&ifp->lock);
3103 read_unlock_bh(&idev->lock);
3106 #ifdef CONFIG_PROC_FS
3107 struct if6_iter_state {
3108 struct seq_net_private p;
3112 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3114 struct inet6_ifaddr *ifa = NULL;
3115 struct if6_iter_state *state = seq->private;
3116 struct net *net = seq_file_net(seq);
3118 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3119 struct hlist_node *n;
3120 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3122 if (net_eq(dev_net(ifa->idev->dev), net))
3128 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3129 struct inet6_ifaddr *ifa)
3131 struct if6_iter_state *state = seq->private;
3132 struct net *net = seq_file_net(seq);
3133 struct hlist_node *n = &ifa->addr_lst;
3135 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3136 if (net_eq(dev_net(ifa->idev->dev), net))
3139 while (++state->bucket < IN6_ADDR_HSIZE) {
3140 hlist_for_each_entry_rcu_bh(ifa, n,
3141 &inet6_addr_lst[state->bucket], addr_lst) {
3142 if (net_eq(dev_net(ifa->idev->dev), net))
3150 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3152 struct inet6_ifaddr *ifa = if6_get_first(seq);
3155 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3157 return pos ? NULL : ifa;
3160 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3164 return if6_get_idx(seq, *pos);
3167 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3169 struct inet6_ifaddr *ifa;
3171 ifa = if6_get_next(seq, v);
3176 static void if6_seq_stop(struct seq_file *seq, void *v)
3179 rcu_read_unlock_bh();
3182 static int if6_seq_show(struct seq_file *seq, void *v)
3184 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3185 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3187 ifp->idev->dev->ifindex,
3191 ifp->idev->dev->name);
3195 static const struct seq_operations if6_seq_ops = {
3196 .start = if6_seq_start,
3197 .next = if6_seq_next,
3198 .show = if6_seq_show,
3199 .stop = if6_seq_stop,
3202 static int if6_seq_open(struct inode *inode, struct file *file)
3204 return seq_open_net(inode, file, &if6_seq_ops,
3205 sizeof(struct if6_iter_state));
3208 static const struct file_operations if6_fops = {
3209 .owner = THIS_MODULE,
3210 .open = if6_seq_open,
3212 .llseek = seq_lseek,
3213 .release = seq_release_net,
3216 static int __net_init if6_proc_net_init(struct net *net)
3218 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3223 static void __net_exit if6_proc_net_exit(struct net *net)
3225 proc_net_remove(net, "if_inet6");
3228 static struct pernet_operations if6_proc_net_ops = {
3229 .init = if6_proc_net_init,
3230 .exit = if6_proc_net_exit,
3233 int __init if6_proc_init(void)
3235 return register_pernet_subsys(&if6_proc_net_ops);
3238 void if6_proc_exit(void)
3240 unregister_pernet_subsys(&if6_proc_net_ops);
3242 #endif /* CONFIG_PROC_FS */
3244 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3245 /* Check if address is a home address configured on any interface. */
3246 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3249 struct inet6_ifaddr *ifp = NULL;
3250 struct hlist_node *n;
3251 unsigned int hash = ipv6_addr_hash(addr);
3254 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3255 if (!net_eq(dev_net(ifp->idev->dev), net))
3257 if (ipv6_addr_equal(&ifp->addr, addr) &&
3258 (ifp->flags & IFA_F_HOMEADDRESS)) {
3263 rcu_read_unlock_bh();
3269 * Periodic address status verification
3272 static void addrconf_verify(unsigned long foo)
3274 unsigned long now, next, next_sec, next_sched;
3275 struct inet6_ifaddr *ifp;
3276 struct hlist_node *node;
3280 spin_lock(&addrconf_verify_lock);
3282 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3284 del_timer(&addr_chk_timer);
3286 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3288 hlist_for_each_entry_rcu_bh(ifp, node,
3289 &inet6_addr_lst[i], addr_lst) {
3292 if (ifp->flags & IFA_F_PERMANENT)
3295 spin_lock(&ifp->lock);
3296 /* We try to batch several events at once. */
3297 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3299 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3300 age >= ifp->valid_lft) {
3301 spin_unlock(&ifp->lock);
3305 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3306 spin_unlock(&ifp->lock);
3308 } else if (age >= ifp->prefered_lft) {
3309 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3312 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3314 ifp->flags |= IFA_F_DEPRECATED;
3317 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3318 next = ifp->tstamp + ifp->valid_lft * HZ;
3320 spin_unlock(&ifp->lock);
3325 ipv6_ifa_notify(0, ifp);
3329 #ifdef CONFIG_IPV6_PRIVACY
3330 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3331 !(ifp->flags&IFA_F_TENTATIVE)) {
3332 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3333 ifp->idev->cnf.dad_transmits *
3334 ifp->idev->nd_parms->retrans_time / HZ;
3336 if (age >= ifp->prefered_lft - regen_advance) {
3337 struct inet6_ifaddr *ifpub = ifp->ifpub;
3338 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3339 next = ifp->tstamp + ifp->prefered_lft * HZ;
3340 if (!ifp->regen_count && ifpub) {
3343 in6_ifa_hold(ifpub);
3344 spin_unlock(&ifp->lock);
3346 spin_lock(&ifpub->lock);
3347 ifpub->regen_count = 0;
3348 spin_unlock(&ifpub->lock);
3349 ipv6_create_tempaddr(ifpub, ifp);
3354 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3355 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3356 spin_unlock(&ifp->lock);
3359 /* ifp->prefered_lft <= ifp->valid_lft */
3360 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3361 next = ifp->tstamp + ifp->prefered_lft * HZ;
3362 spin_unlock(&ifp->lock);
3367 next_sec = round_jiffies_up(next);
3370 /* If rounded timeout is accurate enough, accept it. */
3371 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3372 next_sched = next_sec;
3374 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3375 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3376 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3378 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3379 now, next, next_sec, next_sched));
3381 addr_chk_timer.expires = next_sched;
3382 add_timer(&addr_chk_timer);
3383 spin_unlock(&addrconf_verify_lock);
3384 rcu_read_unlock_bh();
3387 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3389 struct in6_addr *pfx = NULL;
3392 pfx = nla_data(addr);
3395 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3398 pfx = nla_data(local);
3404 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3405 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3406 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3407 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3411 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3413 struct net *net = sock_net(skb->sk);
3414 struct ifaddrmsg *ifm;
3415 struct nlattr *tb[IFA_MAX+1];
3416 struct in6_addr *pfx;
3419 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3423 ifm = nlmsg_data(nlh);
3424 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3428 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3431 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3432 u32 prefered_lft, u32 valid_lft)
3436 unsigned long timeout;
3438 if (!valid_lft || (prefered_lft > valid_lft))
3441 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3442 if (addrconf_finite_timeout(timeout)) {
3443 expires = jiffies_to_clock_t(timeout * HZ);
3444 valid_lft = timeout;
3445 flags = RTF_EXPIRES;
3449 ifa_flags |= IFA_F_PERMANENT;
3452 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3453 if (addrconf_finite_timeout(timeout)) {
3455 ifa_flags |= IFA_F_DEPRECATED;
3456 prefered_lft = timeout;
3459 spin_lock_bh(&ifp->lock);
3460 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3461 ifp->tstamp = jiffies;
3462 ifp->valid_lft = valid_lft;
3463 ifp->prefered_lft = prefered_lft;
3465 spin_unlock_bh(&ifp->lock);
3466 if (!(ifp->flags&IFA_F_TENTATIVE))
3467 ipv6_ifa_notify(0, ifp);
3469 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3477 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3479 struct net *net = sock_net(skb->sk);
3480 struct ifaddrmsg *ifm;
3481 struct nlattr *tb[IFA_MAX+1];
3482 struct in6_addr *pfx;
3483 struct inet6_ifaddr *ifa;
3484 struct net_device *dev;
3485 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3489 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3493 ifm = nlmsg_data(nlh);
3494 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3498 if (tb[IFA_CACHEINFO]) {
3499 struct ifa_cacheinfo *ci;
3501 ci = nla_data(tb[IFA_CACHEINFO]);
3502 valid_lft = ci->ifa_valid;
3503 preferred_lft = ci->ifa_prefered;
3505 preferred_lft = INFINITY_LIFE_TIME;
3506 valid_lft = INFINITY_LIFE_TIME;
3509 dev = __dev_get_by_index(net, ifm->ifa_index);
3513 /* We ignore other flags so far. */
3514 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3516 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3519 * It would be best to check for !NLM_F_CREATE here but
3520 * userspace alreay relies on not having to provide this.
3522 return inet6_addr_add(net, ifm->ifa_index, pfx,
3523 ifm->ifa_prefixlen, ifa_flags,
3524 preferred_lft, valid_lft);
3527 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3528 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3531 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3538 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3539 u8 scope, int ifindex)
3541 struct ifaddrmsg *ifm;
3543 ifm = nlmsg_data(nlh);
3544 ifm->ifa_family = AF_INET6;
3545 ifm->ifa_prefixlen = prefixlen;
3546 ifm->ifa_flags = flags;
3547 ifm->ifa_scope = scope;
3548 ifm->ifa_index = ifindex;
3551 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3552 unsigned long tstamp, u32 preferred, u32 valid)
3554 struct ifa_cacheinfo ci;
3556 ci.cstamp = cstamp_delta(cstamp);
3557 ci.tstamp = cstamp_delta(tstamp);
3558 ci.ifa_prefered = preferred;
3559 ci.ifa_valid = valid;
3561 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3564 static inline int rt_scope(int ifa_scope)
3566 if (ifa_scope & IFA_HOST)
3567 return RT_SCOPE_HOST;
3568 else if (ifa_scope & IFA_LINK)
3569 return RT_SCOPE_LINK;
3570 else if (ifa_scope & IFA_SITE)
3571 return RT_SCOPE_SITE;
3573 return RT_SCOPE_UNIVERSE;
3576 static inline int inet6_ifaddr_msgsize(void)
3578 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3579 + nla_total_size(16) /* IFA_ADDRESS */
3580 + nla_total_size(sizeof(struct ifa_cacheinfo));
3583 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3584 u32 pid, u32 seq, int event, unsigned int flags)
3586 struct nlmsghdr *nlh;
3587 u32 preferred, valid;
3589 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3593 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3594 ifa->idev->dev->ifindex);
3596 if (!(ifa->flags&IFA_F_PERMANENT)) {
3597 preferred = ifa->prefered_lft;
3598 valid = ifa->valid_lft;
3599 if (preferred != INFINITY_LIFE_TIME) {
3600 long tval = (jiffies - ifa->tstamp)/HZ;
3601 if (preferred > tval)
3605 if (valid != INFINITY_LIFE_TIME) {
3613 preferred = INFINITY_LIFE_TIME;
3614 valid = INFINITY_LIFE_TIME;
3617 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3618 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3619 nlmsg_cancel(skb, nlh);
3623 return nlmsg_end(skb, nlh);
3626 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3627 u32 pid, u32 seq, int event, u16 flags)
3629 struct nlmsghdr *nlh;
3630 u8 scope = RT_SCOPE_UNIVERSE;
3631 int ifindex = ifmca->idev->dev->ifindex;
3633 if (ipv6_addr_scope(&ifmca->mca_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_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3642 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3643 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3644 nlmsg_cancel(skb, nlh);
3648 return nlmsg_end(skb, nlh);
3651 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3652 u32 pid, u32 seq, int event, unsigned int flags)
3654 struct nlmsghdr *nlh;
3655 u8 scope = RT_SCOPE_UNIVERSE;
3656 int ifindex = ifaca->aca_idev->dev->ifindex;
3658 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3659 scope = RT_SCOPE_SITE;
3661 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3665 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3666 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3667 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3668 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3669 nlmsg_cancel(skb, nlh);
3673 return nlmsg_end(skb, nlh);
3682 /* called with rcu_read_lock() */
3683 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3684 struct netlink_callback *cb, enum addr_type_t type,
3685 int s_ip_idx, int *p_ip_idx)
3687 struct ifmcaddr6 *ifmca;
3688 struct ifacaddr6 *ifaca;
3690 int ip_idx = *p_ip_idx;
3692 read_lock_bh(&idev->lock);
3694 case UNICAST_ADDR: {
3695 struct inet6_ifaddr *ifa;
3697 /* unicast address incl. temp addr */
3698 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3699 if (++ip_idx < s_ip_idx)
3701 err = inet6_fill_ifaddr(skb, ifa,
3702 NETLINK_CB(cb->skb).pid,
3711 case MULTICAST_ADDR:
3712 /* multicast address */
3713 for (ifmca = idev->mc_list; ifmca;
3714 ifmca = ifmca->next, ip_idx++) {
3715 if (ip_idx < s_ip_idx)
3717 err = inet6_fill_ifmcaddr(skb, ifmca,
3718 NETLINK_CB(cb->skb).pid,
3727 /* anycast address */
3728 for (ifaca = idev->ac_list; ifaca;
3729 ifaca = ifaca->aca_next, ip_idx++) {
3730 if (ip_idx < s_ip_idx)
3732 err = inet6_fill_ifacaddr(skb, ifaca,
3733 NETLINK_CB(cb->skb).pid,
3744 read_unlock_bh(&idev->lock);
3749 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3750 enum addr_type_t type)
3752 struct net *net = sock_net(skb->sk);
3755 int s_idx, s_ip_idx;
3756 struct net_device *dev;
3757 struct inet6_dev *idev;
3758 struct hlist_head *head;
3759 struct hlist_node *node;
3762 s_idx = idx = cb->args[1];
3763 s_ip_idx = ip_idx = cb->args[2];
3766 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3768 head = &net->dev_index_head[h];
3769 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3772 if (h > s_h || idx > s_idx)
3775 idev = __in6_dev_get(dev);
3779 if (in6_dump_addrs(idev, skb, cb, type,
3780 s_ip_idx, &ip_idx) <= 0)
3790 cb->args[2] = ip_idx;
3795 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3797 enum addr_type_t type = UNICAST_ADDR;
3799 return inet6_dump_addr(skb, cb, type);
3802 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3804 enum addr_type_t type = MULTICAST_ADDR;
3806 return inet6_dump_addr(skb, cb, type);
3810 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3812 enum addr_type_t type = ANYCAST_ADDR;
3814 return inet6_dump_addr(skb, cb, type);
3817 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3820 struct net *net = sock_net(in_skb->sk);
3821 struct ifaddrmsg *ifm;
3822 struct nlattr *tb[IFA_MAX+1];
3823 struct in6_addr *addr = NULL;
3824 struct net_device *dev = NULL;
3825 struct inet6_ifaddr *ifa;
3826 struct sk_buff *skb;
3829 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3833 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3839 ifm = nlmsg_data(nlh);
3841 dev = __dev_get_by_index(net, ifm->ifa_index);
3843 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3845 err = -EADDRNOTAVAIL;
3849 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3855 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3856 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3858 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3859 WARN_ON(err == -EMSGSIZE);
3863 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3870 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3872 struct sk_buff *skb;
3873 struct net *net = dev_net(ifa->idev->dev);
3876 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3880 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3882 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3883 WARN_ON(err == -EMSGSIZE);
3887 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3891 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3894 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3895 __s32 *array, int bytes)
3897 BUG_ON(bytes < (DEVCONF_MAX * 4));
3899 memset(array, 0, bytes);
3900 array[DEVCONF_FORWARDING] = cnf->forwarding;
3901 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3902 array[DEVCONF_MTU6] = cnf->mtu6;
3903 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3904 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3905 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3906 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3907 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3908 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3909 jiffies_to_msecs(cnf->rtr_solicit_interval);
3910 array[DEVCONF_RTR_SOLICIT_DELAY] =
3911 jiffies_to_msecs(cnf->rtr_solicit_delay);
3912 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3913 #ifdef CONFIG_IPV6_PRIVACY
3914 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3915 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3916 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3917 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3918 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3920 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3921 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3922 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3923 #ifdef CONFIG_IPV6_ROUTER_PREF
3924 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3925 array[DEVCONF_RTR_PROBE_INTERVAL] =
3926 jiffies_to_msecs(cnf->rtr_probe_interval);
3927 #ifdef CONFIG_IPV6_ROUTE_INFO
3928 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3931 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3932 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3933 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3934 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3936 #ifdef CONFIG_IPV6_MROUTE
3937 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3939 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3940 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3941 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3944 static inline size_t inet6_ifla6_size(void)
3946 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3947 + nla_total_size(sizeof(struct ifla_cacheinfo))
3948 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3949 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3950 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3953 static inline size_t inet6_if_nlmsg_size(void)
3955 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3956 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3957 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3958 + nla_total_size(4) /* IFLA_MTU */
3959 + nla_total_size(4) /* IFLA_LINK */
3960 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3963 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3964 int items, int bytes)
3967 int pad = bytes - sizeof(u64) * items;
3970 /* Use put_unaligned() because stats may not be aligned for u64. */
3971 put_unaligned(items, &stats[0]);
3972 for (i = 1; i < items; i++)
3973 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3975 memset(&stats[items], 0, pad);
3978 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3979 int items, int bytes, size_t syncpoff)
3982 int pad = bytes - sizeof(u64) * items;
3985 /* Use put_unaligned() because stats may not be aligned for u64. */
3986 put_unaligned(items, &stats[0]);
3987 for (i = 1; i < items; i++)
3988 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3990 memset(&stats[items], 0, pad);
3993 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3997 case IFLA_INET6_STATS:
3998 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3999 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4001 case IFLA_INET6_ICMP6STATS:
4002 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4007 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4010 struct ifla_cacheinfo ci;
4012 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
4014 ci.max_reasm_len = IPV6_MAXPLEN;
4015 ci.tstamp = cstamp_delta(idev->tstamp);
4016 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4017 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4018 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
4020 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4022 goto nla_put_failure;
4023 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4025 /* XXX - MC not implemented */
4027 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4029 goto nla_put_failure;
4030 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4032 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4034 goto nla_put_failure;
4035 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4043 static size_t inet6_get_link_af_size(const struct net_device *dev)
4045 if (!__in6_dev_get(dev))
4048 return inet6_ifla6_size();
4051 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4053 struct inet6_dev *idev = __in6_dev_get(dev);
4058 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4064 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4065 u32 pid, u32 seq, int event, unsigned int flags)
4067 struct net_device *dev = idev->dev;
4068 struct ifinfomsg *hdr;
4069 struct nlmsghdr *nlh;
4072 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4076 hdr = nlmsg_data(nlh);
4077 hdr->ifi_family = AF_INET6;
4079 hdr->ifi_type = dev->type;
4080 hdr->ifi_index = dev->ifindex;
4081 hdr->ifi_flags = dev_get_flags(dev);
4082 hdr->ifi_change = 0;
4084 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4087 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4089 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4090 if (dev->ifindex != dev->iflink)
4091 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4093 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4094 if (protoinfo == NULL)
4095 goto nla_put_failure;
4097 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4098 goto nla_put_failure;
4100 nla_nest_end(skb, protoinfo);
4101 return nlmsg_end(skb, nlh);
4104 nlmsg_cancel(skb, nlh);
4108 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4110 struct net *net = sock_net(skb->sk);
4113 struct net_device *dev;
4114 struct inet6_dev *idev;
4115 struct hlist_head *head;
4116 struct hlist_node *node;
4119 s_idx = cb->args[1];
4122 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4124 head = &net->dev_index_head[h];
4125 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4128 idev = __in6_dev_get(dev);
4131 if (inet6_fill_ifinfo(skb, idev,
4132 NETLINK_CB(cb->skb).pid,
4134 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4148 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4150 struct sk_buff *skb;
4151 struct net *net = dev_net(idev->dev);
4154 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4158 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4160 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4161 WARN_ON(err == -EMSGSIZE);
4165 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4169 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4172 static inline size_t inet6_prefix_nlmsg_size(void)
4174 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4175 + nla_total_size(sizeof(struct in6_addr))
4176 + nla_total_size(sizeof(struct prefix_cacheinfo));
4179 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4180 struct prefix_info *pinfo, u32 pid, u32 seq,
4181 int event, unsigned int flags)
4183 struct prefixmsg *pmsg;
4184 struct nlmsghdr *nlh;
4185 struct prefix_cacheinfo ci;
4187 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4191 pmsg = nlmsg_data(nlh);
4192 pmsg->prefix_family = AF_INET6;
4193 pmsg->prefix_pad1 = 0;
4194 pmsg->prefix_pad2 = 0;
4195 pmsg->prefix_ifindex = idev->dev->ifindex;
4196 pmsg->prefix_len = pinfo->prefix_len;
4197 pmsg->prefix_type = pinfo->type;
4198 pmsg->prefix_pad3 = 0;
4199 pmsg->prefix_flags = 0;
4201 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4202 if (pinfo->autoconf)
4203 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4205 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4207 ci.preferred_time = ntohl(pinfo->prefered);
4208 ci.valid_time = ntohl(pinfo->valid);
4209 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4211 return nlmsg_end(skb, nlh);
4214 nlmsg_cancel(skb, nlh);
4218 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4219 struct prefix_info *pinfo)
4221 struct sk_buff *skb;
4222 struct net *net = dev_net(idev->dev);
4225 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4229 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4231 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4232 WARN_ON(err == -EMSGSIZE);
4236 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4240 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4243 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4245 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4250 * If the address was optimistic
4251 * we inserted the route at the start of
4252 * our DAD process, so we don't need
4255 if (!(ifp->rt->rt6i_node))
4256 ip6_ins_rt(ifp->rt);
4257 if (ifp->idev->cnf.forwarding)
4258 addrconf_join_anycast(ifp);
4261 if (ifp->idev->cnf.forwarding)
4262 addrconf_leave_anycast(ifp);
4263 addrconf_leave_solict(ifp->idev, &ifp->addr);
4264 dst_hold(&ifp->rt->dst);
4266 if (ip6_del_rt(ifp->rt))
4267 dst_free(&ifp->rt->dst);
4272 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4275 if (likely(ifp->idev->dead == 0))
4276 __ipv6_ifa_notify(event, ifp);
4277 rcu_read_unlock_bh();
4280 #ifdef CONFIG_SYSCTL
4283 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4284 void __user *buffer, size_t *lenp, loff_t *ppos)
4286 int *valp = ctl->data;
4291 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4294 ret = addrconf_fixup_forwarding(ctl, valp, val);
4300 static void dev_disable_change(struct inet6_dev *idev)
4302 if (!idev || !idev->dev)
4305 if (idev->cnf.disable_ipv6)
4306 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4308 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4311 static void addrconf_disable_change(struct net *net, __s32 newf)
4313 struct net_device *dev;
4314 struct inet6_dev *idev;
4317 for_each_netdev_rcu(net, dev) {
4318 idev = __in6_dev_get(dev);
4320 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4321 idev->cnf.disable_ipv6 = newf;
4323 dev_disable_change(idev);
4329 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4333 net = (struct net *)table->extra2;
4335 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4338 if (!rtnl_trylock()) {
4339 /* Restore the original values before restarting */
4341 return restart_syscall();
4344 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4345 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4346 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4347 addrconf_disable_change(net, newf);
4348 } else if ((!*p) ^ (!old))
4349 dev_disable_change((struct inet6_dev *)table->extra1);
4356 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4357 void __user *buffer, size_t *lenp, loff_t *ppos)
4359 int *valp = ctl->data;
4364 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4367 ret = addrconf_disable_ipv6(ctl, valp, val);
4373 static struct addrconf_sysctl_table
4375 struct ctl_table_header *sysctl_header;
4376 ctl_table addrconf_vars[DEVCONF_MAX+1];
4378 } addrconf_sysctl __read_mostly = {
4379 .sysctl_header = NULL,
4382 .procname = "forwarding",
4383 .data = &ipv6_devconf.forwarding,
4384 .maxlen = sizeof(int),
4386 .proc_handler = addrconf_sysctl_forward,
4389 .procname = "hop_limit",
4390 .data = &ipv6_devconf.hop_limit,
4391 .maxlen = sizeof(int),
4393 .proc_handler = proc_dointvec,
4397 .data = &ipv6_devconf.mtu6,
4398 .maxlen = sizeof(int),
4400 .proc_handler = proc_dointvec,
4403 .procname = "accept_ra",
4404 .data = &ipv6_devconf.accept_ra,
4405 .maxlen = sizeof(int),
4407 .proc_handler = proc_dointvec,
4410 .procname = "accept_redirects",
4411 .data = &ipv6_devconf.accept_redirects,
4412 .maxlen = sizeof(int),
4414 .proc_handler = proc_dointvec,
4417 .procname = "autoconf",
4418 .data = &ipv6_devconf.autoconf,
4419 .maxlen = sizeof(int),
4421 .proc_handler = proc_dointvec,
4424 .procname = "dad_transmits",
4425 .data = &ipv6_devconf.dad_transmits,
4426 .maxlen = sizeof(int),
4428 .proc_handler = proc_dointvec,
4431 .procname = "router_solicitations",
4432 .data = &ipv6_devconf.rtr_solicits,
4433 .maxlen = sizeof(int),
4435 .proc_handler = proc_dointvec,
4438 .procname = "router_solicitation_interval",
4439 .data = &ipv6_devconf.rtr_solicit_interval,
4440 .maxlen = sizeof(int),
4442 .proc_handler = proc_dointvec_jiffies,
4445 .procname = "router_solicitation_delay",
4446 .data = &ipv6_devconf.rtr_solicit_delay,
4447 .maxlen = sizeof(int),
4449 .proc_handler = proc_dointvec_jiffies,
4452 .procname = "force_mld_version",
4453 .data = &ipv6_devconf.force_mld_version,
4454 .maxlen = sizeof(int),
4456 .proc_handler = proc_dointvec,
4458 #ifdef CONFIG_IPV6_PRIVACY
4460 .procname = "use_tempaddr",
4461 .data = &ipv6_devconf.use_tempaddr,
4462 .maxlen = sizeof(int),
4464 .proc_handler = proc_dointvec,
4467 .procname = "temp_valid_lft",
4468 .data = &ipv6_devconf.temp_valid_lft,
4469 .maxlen = sizeof(int),
4471 .proc_handler = proc_dointvec,
4474 .procname = "temp_prefered_lft",
4475 .data = &ipv6_devconf.temp_prefered_lft,
4476 .maxlen = sizeof(int),
4478 .proc_handler = proc_dointvec,
4481 .procname = "regen_max_retry",
4482 .data = &ipv6_devconf.regen_max_retry,
4483 .maxlen = sizeof(int),
4485 .proc_handler = proc_dointvec,
4488 .procname = "max_desync_factor",
4489 .data = &ipv6_devconf.max_desync_factor,
4490 .maxlen = sizeof(int),
4492 .proc_handler = proc_dointvec,
4496 .procname = "max_addresses",
4497 .data = &ipv6_devconf.max_addresses,
4498 .maxlen = sizeof(int),
4500 .proc_handler = proc_dointvec,
4503 .procname = "accept_ra_defrtr",
4504 .data = &ipv6_devconf.accept_ra_defrtr,
4505 .maxlen = sizeof(int),
4507 .proc_handler = proc_dointvec,
4510 .procname = "accept_ra_pinfo",
4511 .data = &ipv6_devconf.accept_ra_pinfo,
4512 .maxlen = sizeof(int),
4514 .proc_handler = proc_dointvec,
4516 #ifdef CONFIG_IPV6_ROUTER_PREF
4518 .procname = "accept_ra_rtr_pref",
4519 .data = &ipv6_devconf.accept_ra_rtr_pref,
4520 .maxlen = sizeof(int),
4522 .proc_handler = proc_dointvec,
4525 .procname = "router_probe_interval",
4526 .data = &ipv6_devconf.rtr_probe_interval,
4527 .maxlen = sizeof(int),
4529 .proc_handler = proc_dointvec_jiffies,
4531 #ifdef CONFIG_IPV6_ROUTE_INFO
4533 .procname = "accept_ra_rt_info_max_plen",
4534 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4535 .maxlen = sizeof(int),
4537 .proc_handler = proc_dointvec,
4542 .procname = "proxy_ndp",
4543 .data = &ipv6_devconf.proxy_ndp,
4544 .maxlen = sizeof(int),
4546 .proc_handler = proc_dointvec,
4549 .procname = "accept_source_route",
4550 .data = &ipv6_devconf.accept_source_route,
4551 .maxlen = sizeof(int),
4553 .proc_handler = proc_dointvec,
4555 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4557 .procname = "optimistic_dad",
4558 .data = &ipv6_devconf.optimistic_dad,
4559 .maxlen = sizeof(int),
4561 .proc_handler = proc_dointvec,
4565 #ifdef CONFIG_IPV6_MROUTE
4567 .procname = "mc_forwarding",
4568 .data = &ipv6_devconf.mc_forwarding,
4569 .maxlen = sizeof(int),
4571 .proc_handler = proc_dointvec,
4575 .procname = "disable_ipv6",
4576 .data = &ipv6_devconf.disable_ipv6,
4577 .maxlen = sizeof(int),
4579 .proc_handler = addrconf_sysctl_disable,
4582 .procname = "accept_dad",
4583 .data = &ipv6_devconf.accept_dad,
4584 .maxlen = sizeof(int),
4586 .proc_handler = proc_dointvec,
4589 .procname = "force_tllao",
4590 .data = &ipv6_devconf.force_tllao,
4591 .maxlen = sizeof(int),
4593 .proc_handler = proc_dointvec
4601 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4602 struct inet6_dev *idev, struct ipv6_devconf *p)
4605 struct addrconf_sysctl_table *t;
4607 #define ADDRCONF_CTL_PATH_DEV 3
4609 struct ctl_path addrconf_ctl_path[] = {
4610 { .procname = "net", },
4611 { .procname = "ipv6", },
4612 { .procname = "conf", },
4613 { /* to be set */ },
4618 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4622 for (i = 0; t->addrconf_vars[i].data; i++) {
4623 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4624 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4625 t->addrconf_vars[i].extra2 = net;
4629 * Make a copy of dev_name, because '.procname' is regarded as const
4630 * by sysctl and we wouldn't want anyone to change it under our feet
4631 * (see SIOCSIFNAME).
4633 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4637 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4639 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4641 if (t->sysctl_header == NULL)
4655 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4657 struct addrconf_sysctl_table *t;
4659 if (p->sysctl == NULL)
4664 unregister_net_sysctl_table(t->sysctl_header);
4669 static void addrconf_sysctl_register(struct inet6_dev *idev)
4671 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4672 &ndisc_ifinfo_sysctl_change);
4673 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4677 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4679 __addrconf_sysctl_unregister(&idev->cnf);
4680 neigh_sysctl_unregister(idev->nd_parms);
4686 static int __net_init addrconf_init_net(struct net *net)
4689 struct ipv6_devconf *all, *dflt;
4691 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
4695 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4697 goto err_alloc_dflt;
4699 /* these will be inherited by all namespaces */
4700 dflt->autoconf = ipv6_defaults.autoconf;
4701 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4703 net->ipv6.devconf_all = all;
4704 net->ipv6.devconf_dflt = dflt;
4706 #ifdef CONFIG_SYSCTL
4707 err = __addrconf_sysctl_register(net, "all", NULL, all);
4711 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4717 #ifdef CONFIG_SYSCTL
4719 __addrconf_sysctl_unregister(all);
4729 static void __net_exit addrconf_exit_net(struct net *net)
4731 #ifdef CONFIG_SYSCTL
4732 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4733 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4735 if (!net_eq(net, &init_net)) {
4736 kfree(net->ipv6.devconf_dflt);
4737 kfree(net->ipv6.devconf_all);
4741 static struct pernet_operations addrconf_ops = {
4742 .init = addrconf_init_net,
4743 .exit = addrconf_exit_net,
4750 int register_inet6addr_notifier(struct notifier_block *nb)
4752 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4754 EXPORT_SYMBOL(register_inet6addr_notifier);
4756 int unregister_inet6addr_notifier(struct notifier_block *nb)
4758 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4760 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4762 static struct rtnl_af_ops inet6_ops = {
4764 .fill_link_af = inet6_fill_link_af,
4765 .get_link_af_size = inet6_get_link_af_size,
4769 * Init / cleanup code
4772 int __init addrconf_init(void)
4776 err = ipv6_addr_label_init();
4778 printk(KERN_CRIT "IPv6 Addrconf:"
4779 " cannot initialize default policy table: %d.\n", err);
4783 err = register_pernet_subsys(&addrconf_ops);
4787 /* The addrconf netdev notifier requires that loopback_dev
4788 * has it's ipv6 private information allocated and setup
4789 * before it can bring up and give link-local addresses
4790 * to other devices which are up.
4792 * Unfortunately, loopback_dev is not necessarily the first
4793 * entry in the global dev_base list of net devices. In fact,
4794 * it is likely to be the very last entry on that list.
4795 * So this causes the notifier registry below to try and
4796 * give link-local addresses to all devices besides loopback_dev
4797 * first, then loopback_dev, which cases all the non-loopback_dev
4798 * devices to fail to get a link-local address.
4800 * So, as a temporary fix, allocate the ipv6 structure for
4801 * loopback_dev first by hand.
4802 * Longer term, all of the dependencies ipv6 has upon the loopback
4803 * device and it being up should be removed.
4806 if (!ipv6_add_dev(init_net.loopback_dev))
4812 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4813 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4815 register_netdevice_notifier(&ipv6_dev_notf);
4819 err = rtnl_af_register(&inet6_ops);
4823 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4828 /* Only the first call to __rtnl_register can fail */
4829 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4830 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4831 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4832 inet6_dump_ifaddr, NULL);
4833 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4834 inet6_dump_ifmcaddr, NULL);
4835 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4836 inet6_dump_ifacaddr, NULL);
4838 ipv6_addr_label_rtnl_register();
4842 rtnl_af_unregister(&inet6_ops);
4844 unregister_netdevice_notifier(&ipv6_dev_notf);
4846 unregister_pernet_subsys(&addrconf_ops);
4848 ipv6_addr_label_cleanup();
4853 void addrconf_cleanup(void)
4855 struct net_device *dev;
4858 unregister_netdevice_notifier(&ipv6_dev_notf);
4859 unregister_pernet_subsys(&addrconf_ops);
4860 ipv6_addr_label_cleanup();
4864 __rtnl_af_unregister(&inet6_ops);
4866 /* clean dev list */
4867 for_each_netdev(&init_net, dev) {
4868 if (__in6_dev_get(dev) == NULL)
4870 addrconf_ifdown(dev, 1);
4872 addrconf_ifdown(init_net.loopback_dev, 2);
4877 spin_lock_bh(&addrconf_hash_lock);
4878 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4879 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4880 spin_unlock_bh(&addrconf_hash_lock);
4882 del_timer(&addr_chk_timer);